HANDBOOK OF BIOLOGICAL CONFOCAL MICROSCOPY THIRD EDITION
HANDBOOK OF BIOLOGICAL CONFOCAL MICROSCOPY THIRD EDITION
Editor James B. Pawley Department of Zoology University of Wisconsin Madison, Wisconsin
~ Springer
James B. Pawley Department of Zoology University of Wisconsin Madison, WI 53706 USA
Library of Congress Control Number: 2005926334 ISBN 10: 0-387-25921-X ISBN 13: 987-0-387-25921-5 Printed on acid-free paper. © 2006, 1995, 1989 Springer Science+Business Media, LLC All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed is forbidden. The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights. 9
8
7
6
springer. com
5
4
3
To my wonderful wife, Christine, who is hoping that we still get along once she begins to see me more often, and to the friends and partners of all the 123 authors, similarly oppressed.
Preface to the Third Edition
Once the second edition was safely off to the printer, the llO authors breathed a sigh of relief and relaxed, secure in the belief that they would "never have to do that again." That lasted for 10 years. When we finally awoke, it seemed that a lot had happened. In particular, people were trying to use the Handbook as a textbook even though it lacked the practical chapters needed. There had been tremendous progress in lasers and fiber-optics and in our understanding of the mechanisms underlying photobleaching and phototoxicity. It was time for a new book. I contacted "the usual suspects" and almost all agreed as long as the deadline was still a year away. That was in 2002. Three years later, most of the old chapters have been substantially or totally rewritten. Although 12 of the chapters are on topics that have either been rendered obsolete by improvements in instrumentation or changes in research interest have been dropped, some have been replaced by chapters on similar topics. To make the Handbook of more use as a textbook, we have added an extended appendix about practical multi photon imaging and another describing the operation of CCO cameras in some detail. There is a new series of practical chapters on confocal microscopy and the selection of dyes, as well as on ion imaging, and on methods for studying brain slices, embryos, biofilms and plants (two). There is also a new chapter describing in some detail how such components as interference filters, acousto-optical devices, and galvanometers are made and what parameters limit their performance. The single chapter on 3D image analysis now has the company of two more on automated 3D image analysis and a third on high-content screening and a fourth on database management. Chapters have been added describing techniques that have only recently come to the fore, such as patterned-illumination fluorescence microscopy, fluorescence resonance energy transfer (FRET) and the generation and detection of second- and third-harmonic signals. In addition, new imaging techniques such as stimulated emission depletion (STED), coherent anti-Stokes Raman (CARS) imaging and selected plane illumination (SPIM) now have their own chapters and there are also chapters that connect the world of 3D light microscopy to the
larger world of micro-CT and micro-MRI and the smaller world revealed by the scanning and transmission electron microscopes. To round out the story we even have a chapter on what PowerPoint does to the results, and the annotated bibliography has been updated and extended. As with the previous editions, the editor enjoyed a tremendous amount of good will and cooperation from the l24 authors involved. Both I, and the light microscopy community in general, owe them all a great debt of gratitude. On a more personal note, I would like to thank Kathy Lyons and her associates at Springer for their unstinting support on one of the biggest books they have done in microscopy and the assistance of her co-workers at Chernow Editorial Services, Barbara Chernow and Kathy Cleghorn. Helen Noeldner was again willing to work long hours to keep all the manuscripts straight in spite of my best effort to confuse them. Thanks are also due to Bill Feeny, the Zoology Department artist, for the innumerable figures that he rescued, reconstructed and otherwise returned to life. If the hidden agenda of the first edition was photon efficiency, and of the second, spherical aberration, the message of the third edition is definitely that all raw, 3D data sets should be deconvolved (or at least 3D-Gaussian filtered) before being viewed or measured. Not only is this required to meet the Nyquist reconstruction criterion, it also greatly reduces the apparent effects of Poisson Noise by effectively averaging the signal over the 50-100 voxels needed to make a Nyquist-sampled, 3D image of a single point object. This last factor allows one to obtain acceptable images using much less excitation, thereby reducing the chance that studies of living cells will be compromised by artifacts caused by phototoxicity. As evermore studies in 3D light microscopy are carried out on living cells, nothing is more important. Now we need dyes that produce less toxicity because they do not cross to the triplet state and photodetectors that operate with lower noise and higher quantum efficiency! That will take another book. James B. Pawley January 2006
vii
Preface to the Second Edition
Confocal microscopy is a good idea that was invented, forgotten and then reinvented about once every decade in the years between 1957 and 1985. However, when White and Amos demonstrated an instrument that was sufficiently user-friendly to become the ideal tool for the 3D localization of specific, fluorescent labels in biological specimens, the field finally took off. Soon after the publication of their 1985 article in the Journal o.f Cell Biology, requests to fund the purchase of similar equipment increased at such a rate that, in the fall of 1988, the U.S. National Science Foundation (NSF) realized that it needed some hard information about the capabilities of this new technique. They funded a two-day symposium on the subject as part of the August 1989 annual meeting of the Electron Microscope Society of America and also financed the publication of 18 papers by the participants as The Handbook (it Biological Confocal Microscopy for free distribution at the meeting. This first edition of the Handbook differed from most of the many other compiled volumes on the subject in that, rather than each author concentrating on his or her own work, an outline for the entire book was written first, and then authors were solicited to cover particular aspects of the instrumentation or its use. Although the necessity of having a volume ready for distribution by August 1989 imposed stringent deadlines on the authors and required the typography and printing to be done locally, every effort was made to try to edit the chapters so that they fit together to form a cohesive whole. The success of the project was due almost entirely to the enthusiasm the authors had for sharing their knowledge of this fascinating subject with a wider audience. Manuscripts originally expected to be 10 pages in length ended up being more than twice this length, and several were more that 50 pages long. The resulting volume included chapters that described and compared each of the component parts of the microscope itself (laser and conventional light sources, intermediate optics, alternative scanning systems, objectives, pinholes, detectors, and antecedent and related optical techniques), chapters that discussed the digital aspects of data acquisition (pixelation, digitization, and display and measurement of 3D data sets) and chapters that reviewed the properties of fluorescent dyes, the techniques of 3D specimen preparation, and the fundamental limitations and practical complexities of quantitative confocal fluorescence imaging. An annotated bibliography of the field was also included. If this first book had any underlying theme, it was probably the importance of photon efficiency. This came about because, as the chapters came together, it became clear that technical limitations of the early instruments, in combination with suboptimal operating techniques, often had an effect such that the signal actually recorded was only about 1% of the expected signal. The Handbook included several concrete suggestions for increasing this fraction, and it is a pleasure to report that instruments incorporating many of these improvements now demonstrate an efficiency figure that is closer to 10-20%. Because of the widespread acceptance of the NSF-sponsored volume by users of the confocal microscope, a revised edition (the "red book") was published by Plenum in 1990. Although this hard-
cover version included over 40 new figures, updated tabular information and over 1,400 typographical improvements, it was otherwise generally very similar to the initial offering. However, the past five years has seen a virtual explosion in the field of biological confocal microscopy. As it became more and more evident that the original Handbook could no longer claim to cover the entire field, I contacted the original set of authors about producing an updated edition. Remembering the frantic urgency that had typified the production of the first edition, I did this with some trepidation; but I need not have worried. The response was uniformly enthusiastic, and several authors were not only willing to completely revise their original chapters but also volunteered to write additional chapters describing several new areas. The response from the 17 new authors was similarly enthusiastic. The final product includes 37 chapters (15 updated from the first edition, 21 new ones, and an annotated bibliography) and is almost three times as long as the original. Chapters covering confocal operation in the UV, in the transmission mode, and when scanning at video rates using a variety of either pointscanning or line-scanning techniques have been added. The use of pulsed laser sources for both two-photon excitation and fluorescence-lifetime imaging is covered in depth, and there is an entire chapter on the functional principles of modern fiberoptic components and the manifold ways that these can be applied to confocal microscopy. In addition, chapters on the joys and perils of observing living specimens in the confocal microscope and on the detection of gold-conjugated labels now complement a revised version of the earlier chapter describing the preparation of dead specimens. No less than 3 of the new chapters address the comparative advantages of the confocal and widefield/deconvolution methods of obtaining 3D data sets from biological specimens with the minimum possible damage. Although each of these chapters proceeds from a very different perspective (algebraic optics, actual measurements, and minimum-entrope image processing), I believe that together they give a balanced view of this complex and important subject and make it clear that the confocal microscope could still be improved if the present photodetector were replaced with one having a higher quantum efficiency. The longest chapter in the book describes the inner workings of the 17 currently available systems applicable to the analysis and display of 3D digital image data, and there is now also a chapter describing the features of all of the current hardware systems for the storage, display, and hardcopy output of 3D and 4D image data sets. The subtext of this second edition is probably an increased recognition of the extent to which the resolution and signal strength of confocal images can be degraded by spherical aberration introduced whenever there is a refractive-index mismatch, such as that occurring when an oil-immersion objective is used with an aqueous specimen. Not only is an entirely new chapter devoted to the SUbject, but many other authors emphasize the same point in their chapters. Again, the manufacturers have responded with the introduction of a number of superb new water-immersion objectives to simplify confocal observations of living specimens; these are also described. ix
x
Preface to Second Edition
On the subject of optics there are also two chapters on realtime 3D imaging. In one, the approach is to combine a high speed slit scanner with rapid motion of the focus plane, while the other demonstrates the truth of the almost paradoxical premise that it can be useful to actually increase the chromatic aberration of an objective if it is to be used to examine surface height in the backscattered light mode with "white" light. Of more interest to those wishing to improve axial resolution in the fluorescent mode is the chapter describing new, high-resolution techniques that combine either two or even three confocal objectives with two-photon excitation to improve resolution to a level heretofore believed to be impossible. Finally, there is a tutorial chapter intended for the novice user, as well as two appendixes. The first appendix describes the relationship between real-space and optical coordinates, while the second provides a compilation of the optical path layouts of the major commercial confocal instruments. The topics in this book cover a very wide range of disciplines. While this is good in that it shows the integrating nature of the field, it can lead to problems with notation when optical physicists, experts on information theory, microscope designers, and just plain biologists have to try to agree on a common system of notation. In the first edition, we did not even try to overcome this problem. Although this led to some confusion, I must confess that my efforts to remedy the problem in the present volume have not been totally successful. Index of refraction has been rendered as 11, so that n can be reserved for the number of quantum events; where t has been used for thickness, we have tried to use italics, so that t could be used for time as a variable and T for temperature, while specific times (lifetimes, pulse times) are shown as T or 't; wherever x, y and z are used as directions, we have italicized them, while we have tried to keep r as actual dimensions in the x-y plane (rp = pinhole radius, r, = slit width, rd = detector diameter, etc.); and numerical aperture appears almost everywhere as NA but becomes ANA in some equations. Perhaps most debatable was my decision to try to save space by replacing the word "wavelength" with Ie in the body of the text. On reflection, this change probably did not repay, in space, the interruption of the reader that it produces, but, unfortunately, by the time this became evident, it was too late to change it. In spite of our best efforts, problems arose because, while authors wanted to fit in with the book as whole, they also, understandably, wished to remain consistent with their previous publications. I would like to thank them all for their cooperation on this complex issue, and I hope that our efforts at consistency have not introduced any errors into the text. This brings us to the Index. There was not enough time to prepare an index for the NSF version. One was put together for the "red book," but it was somewhat less extensive than one might have wished for a handbook. This time, when faced with the need to do it all again, and also having all of the text in electronic form, I was mindful of the two opposing indexing concepts currently pervasive in the popular culture. What one might call the minimalist view of indexing comes from the Douglas Adams book The Hitchhiker's Guide to the Galaxy, where the original entry for Earth is "Harmless;" this is only slightly improved later by being updated to "Mostly Harmless." The opposing view was crystallized by
Barry Commoner as: "Everything is connected to everything else" - a concept amply demonstrated within the field of confocal microscopy. Trying to steer a middle course between these two extremes, I have concocted a new Index that is over twelve times the size of the previous one (now with nearly 7,000 topics and about twice that many page listings), while the book itself has almost tripled. This Index contains entries for almost every diagram, plot, image, and table in the book. It also lists under "Summaries" the pages of the summary sections that conclude most chapters and contain their "take-home lessons." The listing "chapter" refers to an entire chapter starting on the page noted and dealing predominantly with the listed topic. Although subjects in the text are extensively cross-indexed, literally "connecting everything to everything else" would have required another book. I settled for making sure that each text topic appeared at least once under all of the Index topics that seemed appropriate, but I did not attempt to list all the pages in each chapter on which a term was mentioned. As a result, the reader would probably be well-advised to look for additional information on the pages adjacent to (usually following) those pages listed in the Index. I beg indulgence for all of the "inevitable omissions." Confocal microscopy is not the only technology to have developed over the last five years. Constant improvements in the international digital communication network have brought e-mail and electronic file transfer into the normal working Ii ves of most of the authors, and this made the editing of the present edition much more of a two-way process. Chapters could be modified to fit better with their neighbors, returned, checked, and resubmitted all in a matter of days, even when the authors concerned were in Australia, Taiwan, and Europe. Although this process added a welcome level of flexibility not present for the earlier book, it also imposed an additional strain on the authors, who often were just congratulating themselves on finally getting their chapter "out the door" only to have them reappear with a lot of suggested changes and requests for expansion to cover additional areas. Again, the authors responded to this challenge in the most positive manner possible, and this seems the most appropriate place to record my sincere thanks to them for the cooperative spirit that they invariably displayed. Thanks are also due to Helen Noeldner, who provided the order and secretarial assistance without which we could not have succeeded; to Mary Born, my editor at Plenum, whose kind voice prevented me from jumping out of my twelfth-floor window on several occasions; to those manufacturers who provided support for publishing some of the color figures and to their representatives for providing the diagrams and other information included in Appendix 2; to NSF, which provided me with grant DIR-90-17534, to my wife; Christine, who toiled many late nights on the Index; and to my family (and doubtless the families of the authors), who gave me their precious time to help get this project finished. All of these contributed everything that they could in an effort to make this the most comprehensive, accurate, and useful volume on the subject possible. We all hope that you will think we have succeeded. James B. Pawley January 1995
Contents
Preface to the Third Edition ................... . Preface to the Second Edition ................. . Contributors ............................... .
Vll
IX
xxv
CHAPTER 1: FOUNDATIONS OF CONFOCAL SCANNED IMAGING IN LIGHT MICROSCOPY
. .
36 36
.
37
. .
39 41
. . .
43 43 43
. . . . .
44
. . . . . . . . . . . . .
51 51 52
CHAPTER 3: SPECIAL OPTICAL ELEMENTS
Shinya Inoue Light Microscopy ........................... . Lateral Resolution .......................... . Axial Resolution ........................... . Depth of Field ............................ . Confocal Imaging ........................... . Impact of Video ............................ . Nipkow Disk ............................. . Electron-Beam-Scanning Television ............. . Impact of Modern Video ..................... . Lasers and Microscopy ...................... . Holography .............................. . Laser Illumination .......................... . Laser-Illuminated Confocal Microscopes ......... . Confocal Laser-Scanning Microscope ........... . Two- and Multi-Photon Microscopy ............ . Is Laser-Scanning Confocal Microsopy a Cure-All? ................................ . Speed of Image or Data Acquisition ............. . Yokogawa Disk-Scanning Confocal System ....... . Depth of Field in Phase-Dependent Imaging ....... . Other Optical and Mechanical Factors Affecting Confocal Microscopy ...................... . Lens Aberration ........................... . Unintentional Beam Deviation ................. . Contrast Transfer and Resolution in Confocal Versus Non-Confocal Microscopy ............ . Summary ................................. .
I I
3
4 4 5 5 6
7 7 7 8 9 9 10
II II 12 13
13 13 15 16 16
Jens Rietdorf and Ernst H.K. Stelzer Introduction ............................... Regulating the Intensity ...................... Wavelength Selective Filtering Devices .......... Selecting the Wavelength of the Illumination and the Detected Light ........................ Separating the Light Paths .................... Conventional Filters ........................ Interference Filters ......................... Dichroic and Polarizing Beam-Splitters .......... Filters and Dispersive Elements for Multi-Channel Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Mechanical Scanners ........................ Galvanometer Scanners ...................... General Specifications ....................... Acousto-Optical Components ................. Acousto-Optical Deflectors ................... Acousto-Optical Modulators .................. Acousto-Optical Tunable Filters ................ Acousto-Optical Beam-Splitters ................ Electro-Optical Modulators ................... Piezoelectric Scanners ....................... Polarizing Elements ......................... Removing Excess Light .......................
44 45 45 50
54 54
56 56 56 56 57 57 57 58
CHAPTER 4: POINTS, PIXELS, AND GRAY LEVELS: DIGITIZING IMAGE DATA
James B. Pawley
CHAPTER 2: FUNDAMENTAL LIMITS IN CONFOCAL MICROSCOPY
James B. Pawley Introduction ............................... What Limits? ............................. Counting Statistics: The Importance of n ......... Source Brightness .......................... Specimen Response: Dye Saturation ............ A Typical Problem ......................... Practical Photon Efficiency ................... Losses in the Optical System .................. Detection and Measurement Losses ............. Where Have All the Photons Gone? .............
Resolution: How Much Is Enough? ............. Can Resolution Be Too High? ................. Limitations Imposed by Spatial and Temporal Quantization ............................ Practical Considerations Relating Resolution to Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion ................................
. . . . . . . . . .
20 20 20 21 21 24 24 25 28 33
Contrast Transfer Function, Points, and Pixels Pixels, Images, and the Contrast Transfer Function .. Digitization and Pixels ....................... Digitization of Images ....................... How Big Should a Pixel Be? Sampling and Quantum Noise .......................... The Nyquist Criterion ....................... Estimating the Expected Resolution of an Image ... The Story So Far .......................... Reality Check? ............................. Is Over-Sampling Ever Wise? ................. Under-Sampling? ......................... Digitizing Trade-Offs ........................
. . . . . . . . . . .
59 59
62 62 63 64
65 68 68 68 68 68 xi
xii
Contents
Nyquist Reconstruction: "Deconvolution lite" .... . Some Special Cases ........................ . Gray Levels, "Noise," and Photodetector Performance ............................. . Optical Density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Zone System: Quantified Photography ........ . Linearity: Do We Need It? ................... . Gray Levels in Images Recorded Using Charge-Coupled Devices: The Intensity Spread Function ................................ . What Counts as Noise? ...................... . Measuring the Intensity Spread Function ........ . Calibrating a Charge-Coupled Device to Measure the ISF . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . "Fixed-Pattern" Noise ....................... . Gain-Register Charge-Coupled Devices ......... . Multiplicative Noise ........................ . Trade-Offs ................................ .
68
70
71 71 71 72
74 74
75 75 76 76 77
79
CHAPTER 5: LASER SOURCES FOR CONFOCAL MICROSCOPY
Enrico Gratton and Martin J. vandeVen Introduction ............................... . Laser Power Requirements ................... . The Basic Laser ............................. . Principle of Operation ....................... . Pumping Power Requirements ................. . Laser Modes: Longitudinal (Axial) and Transverse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Coherent Properties of Laser Light ............. . Phase Randomization: Scrambling the Coherence Properties of Laser Light ................... . Measures to Reduce the Coherence Length of Laser Light . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Heat Removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Other Installation Requirements ................ . Attenuation of Laser Beams .................. . Stabilization of Intensity, Wavelength, and Beam Position in Lasers ......................... . Sources of Noise in Lasers ................... . Spatial Beam Characteristics .................. . Laser Requirements for Biological Confocal Laser Scanning Microscopy-Related Techniques ..... . Optical Tweezers . . . . . . . . . . . . . . . . . . . . . . . . . . . Total Internal Reflection Microscopy ............ . Confocal Raman Confocal Laser Scanning Microscopy for Chemical Imaging ..................... . Non-Linear Confocal Microscopy .............. . Nanosurgery and Microdissection .............. . Types of Lasers ............................. . Continuous Wave Lasers ..................... . Gas Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Dye Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Solid-State Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . Thin Disk Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . Pulsed Lasers .............................. . Classification of Pulsed Laser Systems ........... . Nitrogen Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . Excimer Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Metal Vapor Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . Dye Lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
80 80 81 82 82 82 83 83 84 84 84 85 85 85 85
89 89 89 89
90 90 90 90 90 90 103 103
109 110 III
112 112 112 112
Modulated Diode Lasers ..................... . Diode Pumped Solid State Laser in Pulsed Mode ... . Ultrafast Diode Pumped Solid State Lasers ....... . Titanium-Sapphire and Related Ultrafast Lasers .... . White Light Continuum Lasers ................ . Ultrafast Fiber Lasers . . . . . . . . . . . . . . . . . . . . . . . . Wavelength Expansion Through Non-linear Techniques .............................. . Second and Higher Harmonic Generation: SHG, THG, FHG Label-Free Microscopy ........... . Sum or Difference Mixing .................... . Optical Parametric Oscillators and Optical Parametric Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Pulse Length Measurement ................... . Maintenance .............................. . Maintenance of Active Laser Media ............. . Maintenance of Pumping Media ............... . Maintenance of the Optical Resonator ........... . Maintenance of Other System Components ....... . Troubleshooting ............................ . Safety Precautions .......................... . Beam Stops . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Curtains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Laser Goggles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Screens . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Exposure Effects, Warning Signs, and Interlocks ... . Infrared Paper . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion ................................ .
112 112 112 112 113
113 114 114 114 114
115 115 115 116
116 116
117 117 118 118 118 118 118 118 118
CHAPTER 6: NON-LASER LIGHT SOURCES FOR THREE-DIMENSIONAL MICROSCOPY
Andreas Nolte, James B. Pawley, and Lutz Haring Introduction ............................... . General Remarks on Choice of Excitation Light Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Scrambling and Filtering the Light .............. . Types of Sources and Their Features ............ . Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Wavelength . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Stability in Time and Wavelength .............. . Radiance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Measuring What Comes Through the Illumination System ....................... . The Bare Minimum ......................... . Types of Confocal Microscopes That Can Use Non-Laser light Sources ................... . Tandem Scanning: Basic Description ............ . Single-Sided Disk Scanning: Basic Description .... . Exposure Time and Source Brightness .......... . Future Trends .............................. .
126 126 131 132
132 135 136 137 138 139 139
141 141 141
141 143
CHAPTER 7: OBJECTIVE LENSES FOR CONFOCAL MICROSCOPY
H. Ernst Keller Introduction ............................... Aberrations of Refractive Systems .............. Defocusing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Monochromatic Aberrations ................... Chromatic Aberrations ......................
. . . . .
145 146 146
147
152
Contents Finite Versus Infinity Optics ................... Working Distance .......................... Optical Materials ........................... Anti-Reflection Coatings ..................... Transmission of Microscope Objectives .......... Conclusion ................................
. . . . . .
156 157 158 158 158 160
CHAPTER 8: THE CONTRAST FORMATION IN OPTICAL MICROSCOPY
Ping-Chin Cheng Introduction ............................... . Sources of Contrast ......................... . Absorption Contrast ........................ . Scattering and Reflection Contrast .............. . Phase Contrast ............................ . Fluorescence Contrast ....................... . Contrast Related to Excitation Wavelength Change ................................ . Negative Contrast .......................... . Special Concerns in Ultraviolet and Near-Infrared Range Confocal Microscopy ................ . Total Internal Reflection Contrast ............... . Harmonic Generation Contrast ................. . Geometric Contrast ......................... . z-Contrast in Confocal Microscopy ............. . Total Internal Refraction Fluorescence Contrast .... . Fluorescence Resonant Energy Transfer .......... . Fluorescence Recovery After Photobleaching (FRAP and FLIP) ........................ . Structural Contrast .......................... . Harmonic Generation Contrast ................. . Birefringence Contrast ...................... . Derived Contrast (Synthetic Contrast) .......... . Ratiometric ............................... . Deconvolution ............................ . Movement Contrast (Subtraction of Previous Image) .................................... . Spectral Unmixing and Color Reassignment ....... . Effects of the Specimen: Spherical Aberration and Optical Heterogeneity ..................... . Mounting Medium Selection .. , ............... . Artificial Contrast ........................... . Contrast Resulting from Instrument Vibration and Ambient Lighting ........................ . Contrast Resulting from Interference of Cover Glass Surfaces ........................... . Background Level and Ghost Images from the Transmission Illuminator ................... . Contrast Resulting from Differences in Photobleaching Dynamics .................. . Effect of Spectral Leakage and Signal Imbalance Between Different Channels ................. . New Contrasts: Fluorescence Lifetime and Coherent Antistokes Raman Spectroscopy ............. . Summary ................................. .
162 163 163 167 171 172
173 173 174 177
179 180 180 180 184 187 188 188 188 188 189 189 190 190 192 198 201 201 201 201 202 203
204 204
CHAPTER 9: THE INTERMEDIATE OPTICAL SYSTEM OF LASER-SCANNING CONFOCAL MICROSCOPES
Ernst H.K. Stelzer Introduction ............................... . Design Principles ........................... .
207 207
Overview ................................ Telecentricity ............................. The Scanning System ....................... The Back-Focal Planes ...................... Practical Requirements ...................... Diffraction Limit ........................... Geometric Distortion ........................ Evaluation of the Illumination and Detection Systems ................................. Influence of Optical Elements ................. Errors ................................... Evaluation of Optical Arrangements ............. Evaluation of Scanner Arrangements ............ Scanners ................................. Attachment to Microscopes ................... Merit Functions ........................... Multi-Fluorescence ......................... Special Setups ............................. Setups for Fluorescence Recovery After Photobleaching Experiments ................ Setups for Fluorescence Resonance Energy Transfer Experiments ............................ Setups for the Integration of Optical Tweezers ..... Setups for the Integration of Laser Cutters ........ Setups for the Observation of Living Specimens .... Miniaturization and Computer Control ......... Thermal Stability ........................... Vibration Isolation .......................... Conclusions and Future Prospects .............
xiii
. . . . . . .
207 207 208 210 210 210 211
. . . . . . . . . .
211 211 211 212 213 215 217 217
.
218
. . . . . . . .
218 218 218 219 219 219 219 219
. . . .
221 221 221 222
. . . .
222 223 223 223
. . . . . . . . . .
223 224 224 224 225 225 225 228 228 228
.
229
. . . .
231 231 232 232
.
233
217
218
CHAPTER 10: DISK-SCANNING CONFOCAL MICROSCOPY
Derek Toomre and James B. Pawley Introduction ............................... Background ............................... Living Cell Imaging: Probing the Future ......... A Need for Speed and Less Photobleaching ....... Advantages and Limitations of Confocal Laser-Scanning Microscopes ................ Other Imaging and Deconvolution .............. Confocal Disk-Scanning Microscopy ............ Nipkow Disk - An Innovation ................ A Renaissance - Advantages of Disk-Scanning Confocal Imaging ........................ Disadvantages ............................. Critical Parameters in Pinhole and Slit Disks ..... Fill Factor and Spacing Interval F .............. Lateral Resolution .......................... Pinhole/Slit Size ........................... Axial Resolution ........................... Types of Disk-Scanning Confocals ............. General Considerations ...................... Disk Scanners for Backscattered Light Imaging .... CARV, DSU, and Other Disk-Scanning Confocal Microscopes ............................ The Yokogawa Microlens - An Illuminating Approach .............................. New Fast Slit Scanner - Zeiss LSM51 0 LIVE .... New Detectors - A Critical Component ........ Image Intensifiers .......................... On-Chip Electron MUltiplying Charge-Coupled Device ................................
xiv
Contents
Electron Multiplication Charge-Coupled Devices and Disk Scanners ........................... Applications and Examples of Confocal Disk-Scanning Microscopes ................. Comparison with Epi-Fluorescence Imaging ....... Fast 3D/4D Imaging ........................ Blazingly Fast Confocal Imaging ............... Future Developments? ....................... Summary ..................... " ..........
.
234
. . . . . .
235 235 235 235 236 237
239 240 240 241 242 243 243 245 245 245 247 248 248 248 249 250 250
CHAPTER 12: PHOTON DETECTORS FOR CONFOCAL MICROSCOPY
Jonathan Art Introduction ............................... The Quantal Nature of Light .................. Interaction of Photons with Materials ........... Thermal Effects ........................... Direct Etfects ............................. Photoconductivity .......................... Photovoltaic .............................. Photoemissive ............................. Comparison of Detectors ..................... Noise Internal to Detectors ................... Noise in Internal Detectors ................... Noise in Photoemissive Devices ............... Statistics of Photon Flux and Detectors .......... Representing the Pixel Value .................. Conversion Techniques ...................... Assessment of Devices ....................... Point Detection Assessment and Optimization ..... Field Detection Assessment and Optimization ..... Detectors Present and Future .................
. . . . . . . . . . . . . . . . . . .
251 251 252 252 252 252 252 254 255 256 256 256 257 258 259 260 260 261 262
.
268
. . .
270 276 276
Introduction Definitions ............................... What Is the Microscopist Trying to Achieve? ...... Criteria for Choosing a Visualization System ...... Why Do We Want to Visualize Multi-Dimensional Laser-Scanning Microscopy Data? ............ Data and Dimensional Reduction ............... Objective or Subjective Visualization? ........... Prefiltering ............................... Identifying Unknown Structures ............... Highlighting Previously Elucidated Structures ..... Visualization for Multi-Dimensional Measurements ........................... What Confocal Laser Scanning Microscopy Images Can the Visualization System Handle? ........ Image Data: How Are Image Values Represented in the Program? .......................... What Dimensions Can the Images and Views Have? ............................ Standard File Formats for Calibration and Interpretation ........................... How Will the System Generate the Reconstructed Views? .................................. Assessing the Four Basic Steps in the Generation of Reconstructed Views .................... Loading the Image Subregion ................. Choosing a View: The 5D Image Display Space .... Mapping the Image Space into the Display Space ... How Do 3D Visualizations Retain the z-Information? ........................... Mapping the Data Values into the Display ........ How Can Intensities Be Used to Retain z-Information? ........................... Hidden-Object Removal ..................... Adding Realism to the View .................. How Can I Make Measurements Using the Reconstructed Views? ...................... Conclusion ................................
. . .
280 280 280 281
. . . . . .
281 281 281 281 281 284
.
284
.
286
.
286
.
286
.
288
.
290
. . . .
290 290 291 294
. .
296 300
. . .
304 304 306
. .
312 312
CHAPTER 15: AUTOMATED THREEDIMENSIONAL IMAGE ANALYSIS METHODS FOR CONFOCAL MICROSCOPY
Badrinath Roysam, Gang Lin, Muhammad-Amri Abdul-Karim, Omar Al-Kofahi, Khalid Al-Kofahi, William Shain, Donald H. Szarowsk, and James N. Turner
CHAPTER 13: STRUCTURED ILLUMINATION METHODS
Rainer Heintzmann Introduction ............................... .
265 266
N.S. White
Rimas Ju§kaitis . . . . . . . . . . . . . . . . .
. .
CHAPTER 14: VISUALIZATION SYSTEMS FOR MULTI-DIMENSIONAL MICROSCOPY IMAGES
CHAPTER 11: MEASURING THE REAL POINT SPREAD FUNCTION OF HIGH NUMERICAL APERTURE MICROSCOPE OBJECTIVE LENSES
Introduction ............................... Measuring Point Spread Function .............. Fiber-Optic Interferometer .................... Point Spread Function Measurements ............ Chromatic Aberrations ...................... Apparatus ................................ Axial Shift ............................... Pupil Function ............................. Phase-Shifting Interferometry ................. Zernike Polynomial Fit ...................... Restoration of a 3D Point Spread Function ........ Empty Aperture ........................... Miscellanea ............................... Temperature Variations ...................... Polarization Effects ......................... Apodization .............................. Conclusion ................................
Experimental Considerations .................. Pattern Generation ......................... Computing Optical Sections from Structured-Illumination Data ................ Resolution Improvement by Structured Illumination ............................. Nonlinear Structured Illumination .............. Summary .................................
265
Introduction ............................... . Types of Automated Image Analysis Studies ...... .
316 318
Contents Common Types of Biological Image Objects ..... . Specimen Preparation and Image Preprocessing Methods ................................ . Data Collection Guidelines for Image Analysis Purposes ............................... . Image Preprocessing Methods ................. . General Segmentation Methods Applicable to Confocal Data ........................... . Bottom-Up Segmentation Methods ............. . Top-Down Segmentation Methods .............. . Hybrid Segmentation Methods Combining Bottom-Up and Top-Down Processing .................. . Example Illustrating Blob Segmentation ......... . Model-Based Object Merging ................. . Example Illustrating Segmentation of Tube-Like Objects ................................. . Skeletonization Methods ..................... . Vectorization Methods ....................... . Example Combining Tube and Blob Segmentation ............................ . Registration and Montage Synthesis Methods .... . Methods for Quantitative Morphometry ........ . Methods for Validating the Segmentation and Making Corrections ....................... . Analysis of Morphometric Data ................ . Discussion, Conclusion, and Future Directions ... .
319 319 319 320 321 321 322 322 322 323 324 324 324 328 328 331 333 334 335
CHAPTER 16: FLUOROPHORES FOR CONFOCAL MICROSCOPY: PHOTOPHYSICS AND PHOTOCHEMISTRY
Roger Y. Tsien, Lauren Ernst, and Alan Waggoner Introduction ............................... Photo physical Problems Related to High Intensity Excitation ............................... Singlet State Saturation ...................... Triplet State Saturation ...................... Contaminating Background Signals ............. What Is the Optimal Intensity? ................ Photodestruction of Fluorophores and Biological Specimens .............................. Dependency on Intensity or Its Time Integral? ..... Strategies for Signal Optimization in the Face of Photobleaching ........................... Light Collection Efficiency ................... Spatial Resolution .......................... Protective Agents .......................... Fluorophore Concentration ................... Choice of Fluorophore ...................... Fluorescent Labels for Antibodies, Other Proteins, and DNA Probes .......................... Fluorescent Organic Dyes .................... Phycobiliproteins .......................... DNA Probes .............................. Luminescent Nanocrystals .................... Fluorescent Lanthanide Chelates ............... Fluorescent Indicators for Dynamic Intracellular Parameters .............................. Membrane Potentials ........................ Ion Concentrations ......................... pH Indicators ............................. Ca'+ Indicators ............................
.
338
. . . . .
338 338 339 339 340
. .
340 340
. . . . . .
341 341 341 341 342 342
. . . . . .
342 342 343 343 343 345
. . . . .
346 346 346 346 346
xv
. . .
347 347 347
. . . . .
348 348 348 348 348
. . .
353 353 353
. . . . .
356 357 357 358 358
. . . . . . . .
360 361 361 362 362 362 363 364
Introduction ............................... . Characteristics of Fixatives ................... . Glutaraldehyde ............................ . Formaldehyde ............................. . Fixation Staining and Mounting Methods ........ . Glutaraldehyde Fixation ..................... . pH Shift/Formaldehyde Fixation ............... . Immunofluorescence Staining ................. . Mounting the Specimen ...................... . Critical Evaluation of Light Microscopy Fixation and Mounting Methods ........................ . Use of the Cell Height to Evaluate the Fixation Method ......................... . Use of Cell Height to Evaluate Mounting Media ................................. . Well-Defined Structures Can Be Used to Evaluate Fixation Methods ........................ . Comparison of In Vivo Labeled Cell Organelles with lmmunolabeled Cell Organelles .............. . General Notes ............................. . Labeling Samples with Two or More Probes ...... .
368 368 369 369 370 370 370 371
Oxygen Sensor ............................ cAMP Indicators ........................... Fatty Acid Indicator ........................ Genetically Expressed Intracellular Fluorescent Indicators ............................... Green Fluorescent Protein .................... Ligand-Binding Modules ..................... Ion Indicators ............................. Future Developments .......................
CHAPTER 17: PRACTICAL CONSIDERATIONS IN THE SELECTION AND APPLICATION OF FLUORESCENT PROBES
lain D. johnson Introduction ............................... Selection Criteria for Dyes and Probes .......... Organic Dyes ............................. Fluorescent Proteins: Green Fluorescent Protein and Phycobiliproteins ......................... Quantum Dots ............................ Multi-Photon Excitation ..................... Introducing the Probe to the Specimen ......... Loading Methods .......................... Target Abundance and Autofluorescence Considerations .......................... Interactions of Probes and Specimens .......... Localization and Metabolism .................. Perturbation and Cytotoxicity ................. Under the Microscope ....................... Photobleaching ............................ Phototoxicity ............................. Summary .................................
CHAPTER 18: GUIDING PRINCIPLES OF SPECIMEN PRESERVATION FOR CONFOCAL FLUORESCENCE MICROSCOPY
Rohert Bacallao, Sadaf Sohrah, and Carrie Phillips
371
371 372 373
373 374 374 375
xvi
Contents
Triple Labeling ............................ . Preparation of Tissue Specimens ............... . Labeling Thick Sections ..................... . Refractive Index Mismatch ................... . Screening Antibodies on Glutaraldehyde-Fixed Specimens ............................. . Microwave Fixation ........ ................ . Conclusion .................................
375 376 376 377 377 377 378
CHAPTER 19: CONFOCAL MICROSCOPY OF LIVING CELLS
Michael E. Dailey, Erik Manders, David R. Soil, and Mark Terasaki Introduction ............................... . Overview of Living-Cell Confocal Imaging Techniques .............................. . Time-Lapse Fluorescence Imaging .............. . Multi-Channel Time-Lapse Fluorescence Imaging .. . Spectral Imaging and Linear Unmixing .......... . Fluorescence Recovery After Photobleaching Fluorescence Loss in Photobleaching ............ . Fluorescence Resonance Energy Transfer ......... . Fluorescence Lifetime Imaging ................ . Fluorescence Correlation Spectroscopy .......... . Fluorescence Speckle Microscopy .............. . Photo-UncaginglPhotoactivation ............... . Optical TweezerslLaser Trapping ............... . Physiological Fluorescence Imaging ............. . Combining Fluorescence and Other Imaging Modalities .............................. . General Considerations for Confocal Microscopy of Living Cells ............................ . Maintenance of Living Cells and Tissue Preparations ............................ . Fluorescent Probes ......................... . Minimizing Photodynamic Damage ............. . The Online Confocal Community .............. . A Convenient Test Specimen .................. . Specific Example I: Visualizing Chromatin Dynamics Using Very Low Light Levels ........ . Phototoxicity ............................. . Reduction of Phototoxicity ................... . Improving Image Quality in Low-Dose M'ICroSCOpy ............................. . Low-Dose Imaging Conclusion ................ . Specific Example II: Multi-Dimensional Imaging of Microglial Cell Behaviors in Live Rodent Brain Slices ............................... Preparation of Central Nervous System Tissue Slices ..... ...... ...... ..... ........... . Fluorescent Staining ........................ . Maintaining Tissue Health on the Microscope Stage ................................. . Imaging Methods .......................... . Imaging Deep Within Tissue .................. . Keeping Cells in Focus ...................... . Handling the Data .......................... . Results ................................... Conclusion ........ .... ................. .. . Future Directions ....... .... ................ .
381 382 382 382 382 382 382 382 382 383 383 383 383 383 383 386 387 387 389 390 390 390 390 391 391 391
392 393 393 393 394 395 395 395 396 396 398
CHAPTER 20: ABBERATIONS IN CONFOCAL AND MULTI-PHOTON FLUORESCENCE MICROSCOPY INDUCED BY REFRACTIVE INDEX MISMATCH
Alexander Egner and Stefan W Hell Introduction ....... . .................. , ..... The Situation ............................... Theory ................................... . Results of Theoretical Calculations ............. . Experiments ............................... . Other Considerations ....................... . Dry Objectives ............................ . Refractive Index, Wavelength, and Temperature .... . Spherical Aberration Correction ................ . Conclusion ................................. Consequences ............................. . Practical Strategies to Reduce Refractive Index Mismatch ...............................
404 404 404 407 409 4lO 410 411 411 412 412 412
CHAPTER 21: INTERACTION OF LIGHT WITH BOTANICAL SPECIMENS
Ping-Chin Cheng Introduction ............................... . Light Attenuation in Plant Tissue ............... . Linear Absorption .......................... . Nonlinear Absorption ....................... . Scattering ................................ . Refractive Index Heterogeneity ................ . Birefringent Structures in Plant Cells ........... . Fluorescence Properties of Plants .............. . Changes in Emission Spectra Depending on One- Versus Two-Photon Excitation ........... . Microspectroscopy ......................... . Light-Specimen Interaction (Fluorescence Emission) .............................. . Harmonic Generation Properties .............. . The Effect of Fixation on the Optical Properties of Plants ................................ . Living Plant Cells ........................... . Callus, Suspension Culture Cells and Protoplasts ............................. . Meristem ................................. Stem and Root ............................. Microspores and Pollen Grains ................ . Cuticles, Hairs, and Waxes ................... . Storage Structures .......................... . Mineral Deposits ........................... . Primary and Secondary Cell Walls .............. . Fungi ................................... . Conclusion .................................
414 414 414 416 417 418 420 421 421 421 425 428 428 429 429 430 430 431 434 435 436 438 438 439
CHAPTER 22: SIGNAL-TO-NOISE RATIO IN CONFOCAL MICROSCOPES
Colin J.R. Sheppard, Xiaosong Can, Min Cu, and Maitreyee Roy Introduction ............................... , Sources of Noise ............................ Shot Noise and Quantum Efficiency ............. . Background Noise ......................... .
442 442 442 443
Contents .
444
. . . . . .
445 445 446 447 447 448
. . . . . .
448 448 450 450 451 451
Introduction ............................... . The Point Spread Function: Imaging as a Convolution ............................. . Limits to Linearity and Shift Invariance .......... . Deconvolution ............................. . Practical Differences ........................ . Temporal Resolution ........................ . Combination of Charged-Coupled Device and Confocal Imaging ............................... . Integration of Fluorescence Intensity ............ . Resolution, Sensitivity, and Noise .............. . Fluorescence Excitation ...................... . Fluorescent Light Detection ................... . Gain Register Charge-Coupled Devices .......... . Out-of-Focus Light ......................... . Model Specimens .......................... . The Best Solution: Deconvolving Confocal Data ... . Practical Comparisons ....................... . Conclusion ................................ . Summary ................................. .
453
Signal Level in Confocal Microscopes ........... Signal-to-Noise Ratio for Confocal Microscopes ............................. Q" N I, and Stain Level ..................... N2 and Detectability ........................ Multi-Photon Fluorescence Microscopy .......... Designs of Confocal Microscopes .............. Sampling .................................. Comparative Performance of Fluorescence Microscopes ............................. Bleaching-Limited Performance ................ Saturation-Limited Performance ................ Effects of Scanning Speed .................... 3D Imaging .............................. Summary ................................. CHAPTER 23: COMPARISON OF WIDEFIELDIDECONVOLUTION AND CONFOCAL MICROSCOPY FOR THREEDIMENSIONAL IMAGING Peter 1. Shaw
453 457 457 458 458 458 459 459 459 459 460 461 461 461 463 466 467
CHAPTER 24: BLIND DECONVOLUTION Timothy 1. Holmes, David Biggs, and Asad Abu-Tarif
Introduction ............................... Purposes of Deconvolution ................... Advantages and Limitations .................. Principles ................................. Data Collection Model ...................... Maximum likelihood Estimation ............... Algorithms ............................... Different Approaches ........................ 3D ..................................... 20 Image Filtering ......................... Data Corrections ........................... light Source and Optics Alignment ............ Newest Developments ....................... Subpixel ................................. Polarized Light ............................ Live Imaging .............................
. . . . . . . . . . . . . . . .
468 468 468 472 472 472 472 475 475 476 477 477 478 478 479 480
More Examples ............................. Blind Deconvolution and Spherical Aberration . . . . . . Widefield Fluorescence Simulation .............. Spinning-Disk Confocal ...................... Two Photon ............................... Speed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Future Directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . Summary of Main Points . . . . . . . . . . . . . . . . . . . . . .
xvii 480 480 481 481 481 482 483 483
CHAPTER 25: IMAGE ENHANCEMENT BY DECONVOLUTION Mark B. Cannell, Angus McMorland, and Christian Soeller
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Background ................................ Image Formation ............................ Forwards: Convolution and the Imaging System ... Properties of the Point Spread Function . . . . . . . . . . Quantifying the Point Spread Function . . . . . . . . . . . The Missing Cone Problem .................... Noise ..................................... Deconvolution Algorithms. . . . . . . . . . . . . . . . . . . . . Nearest-Neighbor Deconvolution . . . . . . . . . . . . . . . . Wiener Filtering ............................ Nonlinear Constrained Iterative Deconvolution Algorithms .............................. Comparison of Methods ......................
488 488 489 490 492 492 494 495 495 495 496 496 497
CHAPTER 26: FIBER-OPTICS IN SCANNING OPTICAL MICROSCOPY Peter Delaney and Martin Harris
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Key Fiber Technologies Relevant to Scanning Microscopy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Glass Made from Gas and Its Transmission Properties ........... . . . . . . . . . . . . . . . . . . . . Step Index and Gradient Index Optical Fibers. . . . . . . Modes in Optical Fibers ...................... Evanescent Wave and Polarization Effects in Optical Fibers . . . . . . . . . . . . . . . . . . . . . . . . . . . . Polarization-Maintaining Fibers . . . . . . . . . . . . . . . . . Fused Biconical Taper Couplers: Fiber-Optic Beam-Splitters ........................... Microstructure Fibers ........................ Fiber Image Transfer Bundles .................. Key Functions of Fibers in Optical Microscopes ... Optical Fiber for Delivering Light . . . . . . . . . . . . . . . Optical Fiber as a Detection Aperture. . . . . . . . . . . . . Same Fiber for Both Source and Confocal Detection ............................... Fiber Delivery for Nonlinear Microscopy with Femtosecond Lasers ....................... Large Core Fibers as Source or Detection Apertures ............................... Benchtop Scanning Microscopes Exploiting Fiber Components. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Miniaturized Scanning Confocal Microscope Imaging Heads ............................ Miniature Confocal Imaging Heads Based on Coherent Imaging Bundles . . . . . . . . . . . . . . . . . . .
501 501 501 501 502 503 503 503 504 504 505 505 506 506 507 507 507 508 508
xviii
Contents
Resolution and Optical Efficiency of Bundles ...... Bundle Imagers for In Vivo Studies in Animals ..... Scan Heads Based on Single Fibers with Miniature Scanning Mechanisms ..................... Vibrating the Fiber Tip ...................... Vibrating the Lens and Fiber .................. Scanning with Micromirrors .................. Scanning Fiber Confocal Microscopes for In Vivo Imaging in Animals ....................... Implementations for Clinical Endomicroscopy .... Summary .................................
. .
509 509
. . . .
510 510 510 511
. . .
512 513 513
CHAPTER 27: FLUORESCENCE LIFETIME IMAGING IN SCANNING MICROSCOPY
. . . . . . . . . . .
543 543 544 544 545 545 545 545 545 545 545
CHAPTER 29: MULTI FOCAL MULTI-PHOTON MICROSCOPY
H. C. Gerritsen, A. Draaijer, D.l. van den Heuvel, and A. V. Agronskaia
Introduction ............................... Fluorescence, Lifetime, and Quantum Efficiency .............................. Fluorescence Lifetime Spectroscopy ............ Fluorescence Lifetime Imaging Applications ...... Fluorescence Resonance Energy Transfer ......... Fluorescence lifetime Imaging Methods ........ Introduction .............................. Lifetime Sensing in the Frequency Domain ....... Fluorescence Lifetime Sensing in the Time Domain ............................... Comparison of Confocal Fluorescence Lifetime Imaging Methods ........................ Applications ............................... Multi-Labeling and Segmentation .............. Ion-Concentration Determination ............... Probes for Fluorescence Lifetime Microscopy ..... Summary .................................
Chromophores (Fluorophores and Caged Compounds) ............................. Two-Photon Absorption Cross-Sections .......... Caged Compounds ......................... Cell Viability During Imaging ................. Applications ............................... Calcium Imaging .......................... Uncaging and Photobleaching ................. Autofluorescence .......................... Developmental Biology ...................... In Vivo (Intact Animal) Imaging ................ Outlook ..................................
iorg Bewersdorj; Alexander Egner, and Stefan W Hell
.
516
. . . . . . .
516 516 516 517 518 518 518
.
520
. . . . . .
523 527
527 528 530 532
CHAPTER 28: MULTI-PHOTON MOLECULAR EXCITATION IN LASER-SCANNING MICROSCOPY
Introduction ............................... . Background .............................. . Determination of the Optimum Degree of Parallelization ........................... . Experimental Realization ..................... . A Multi-Focal Multi-Photon Microscopy Setup Using a Nipkow-Type Microlens Array ............. . Resolution ............................... . Time Multiplexing as a Solution to Interfocal Crosstalk .............................. . Alternative Realizations ..................... . Advanced Variants of Multi-Focal Multi-Photon Microscopy .............................. . Space MUltiplexing ......................... . Fluorescence Lifetime Imaging ................ . Second Harmonic Generation Multi-Focal Multi-Photon Microscopy ............................. . Multi-Focal Multi-Photon Microscopy-4Pi Microscopy ............................. . Imaging Applications ........................ . limitations ................................ . Current Developments ...................... . Summary ................................. .
550 550 550 551 551 552 553 554 555 555 555 556 556 556 556 558 559
Winfried Denk, David W Piston, and Watt W Webb
Introduction ............................... Physical Principles of Multi-Photon Excitation and Their Implications for Image Formation .... Physics of Multi-Photon Excitation ............. Optical Pulse Length ........................ Excitation Localization ...................... De~ction ................................ Wavelengths .............................. Resolution ............................... Photodamage: Heating and Bleaching ........... Instrumentation ............................ Lasers and the Choice of Excitation Wavelengths ... Detection ................................ Optical Aberrations ......................... Pulse Spreading Due to Group Delay Dispersion ... Control of Laser Power ...................... Resonance and Non-Mechanical Scanning ........
.
535
. . . . . . . . . . . . . . .
535 535 537 538 538 538 539 539 540 540 541 542 543 543 543
CHAPTER 30: 4Pi MICROSCOPY iorg Bewersdorf, Alexander Egner, and Stefan W Hell
Introduction ............................... Theoretical Background ...................... The Point Spread Function ................... The z-Response and the Axial Resolution ......... The Optical Transfer Function ................. Multi-Focal Multi-Photon Microscopy-4Pi Microscopy .............................. Space lnvariance of the Point Spread Function ..... Live Mammalian Cell 4Pi Imaging ............. Type C 4Pi Microscopy with the Leica TCS 4PI ... Resolution ............................... Type C 4Pi Imaging in Living Cells ............. Summary and Outlook ......................
. . . . . . . . . . . .
561 562 562 563 563 563
564 564 565 567 568 568
Contents
Coherent Anti-Stokes Raman Scattering Correlation Spectroscopy ............................ . Coherent Anti-Stokes Raman Scattering Microscopy Imaging of Biological Samples ............... . Conclusions and Perspectives ................. .
CHAPTER 31: NANOSCALE RESOLUTION WITH FOCUSED LIGHT: STIMULATED EMISSION DEPLETION AND OTHER REVERSIBLE SATURABLE OPTICAL FLUORESCENCE TRANSITIONS MICROSCOPY CONCEPTS
xix
602 603 604
Stefan W. Hell, Katrin I. Willig, Marcus Dyba, Stefan lakobs, Lars Kastrup, and Volker Westphal The Resolution Issue ........................ . Breaking the Diffraction Barrier: The Concept of Reversible Saturable Optical Fluorescence Transitions .............................. . Different Approaches of Reversible Saturable Optical Fluorescence Transitions Microscopy ......... . Stimulated Emission Depletion Microscopy ...... . Challenges and Outlook ..................... .
571
571 573 574
577
CHAPTER 32: MASS STORAGE, DISPLAY, AND HARD COpy
Guy Cox Introduction ............................... Mass Storage .............................. Data Compression .......................... Removable Storage Media .................... Random-Access Devices ..................... Solid State Devices ......................... Display ................................... Monitors ................................. Liquid Crystal Displays ...................... Data Projectors ............................ Hard Copy ................................ Photographic Systems ....................... Digital Printers ............................ Conclusion ................................ Summary ................................. Bulk Storage .............................. Display .................................. Hard Copy ...............................
. . . . . . . . . . . . . . . . . .
580 580 580 585 586 588 588 588 589 590 590 590 591 593 593 593 594 594
CHAPTER 34: RELATED METHODS FOR THREE-DIMENSIONAL IMAGING
l. Michael Tyszka, Seth W. Ruffins, lamey P. Weichert, Michael l. Paulus, and Scott E. Fraser Introduction .............................. . Surface Imaging Microscopy and Episcopic Fluorescence Image Capture ................ . Optical Coherence Tomography ............... . Optical Projection Tomography ............... . Light Sheet Microscopy ...................... . Optical Setup ............................. . Micro-Computerized Tomography Imaging ...... . Operating Principle ......................... . Contrast and Dose .......................... . Computed Tomography Scanning Systems ........ . Magnetic Resonance Microscopy .............. . Basic Principles of Nuclear Magnetic Resonance ............................. . Magnetic Resonance Image Formation ........... . Magnetic Resonance Microscopy Hardware ....... . Strengths and Limitations of Magnetic Resonance Microscopy ............................. . Image Contrast in Magnetic Resonance Microscopy .. Magnetic Resonance Microscopy Applications ..... . Future Development of Magnetic Resonance Microscopy ............................. . Conclusion ................................ .
607 607 609 610
613 613 614 614 614 615 618 618 619 622 622 622
623 624 624
CHAPTER 35: TUTORIAL ON PRACTICAL CONFOCAL MICROSCOPY AND USE OF THE CONFOCAL TEST SPECIMEN
Victoria Centonze and lames B. Pawley CHAPTER 33: COHERENT ANTI-STOKES RAMAN SCATTERING MICROSCOPY
X. Sunney Xie, li-Xin Cheng, and Eric Potma Introduction ............................... . Unique Features of Coherent Anti-Stokes Raman Scattering Under the Tight-Focusing Condition .. Forward and Backward Detected Coherent Anti-Stokes Raman Scattering ............... . Optimal Laser Sources for Coherent Anti-Stokes Raman Scattering Microscopy ............... . Suppression of the Non-Resonant Background ... . Use of Picosecond Instead of Femtosecond Pulses .,. Epi-Detection ............................. . Polarization-Sensitive Detection ................ . Time-Resolved Coherent Anti-Stokes Raman Scattering Dctection .............................. . Phase Control of Excitation Pulses .............. . Multiplex Coherent Anti-Stokes Raman Scattering Microspectroscopy ........................ .
595 596 597 599 600 600 600 600 600 600 602
Introduction ............................... Getting Started ............................ Bleaching - The Only Thing That Really Matters ................................ Getting a Good Confocal Image ............... Simultaneous Detection of Backscattered Light and Fluorescence ......................... New Controls .............................. Photon Efficiency .......................... Pinhole Size .............................. Stray Light ............................... Is the Back-Focal Plane Filled? ................ Pinhole Summary .......................... Statistical Considerations in Confocal Microscopy ............................. The Importance of Pixel Size .................. Measuring Pixel Size ....................... Over-Sampling and Under-Sampling ............ Nyquist Reconstruction and Deconvolution ....... Pixel Size Summary ........................
. .
627 627
. .
627 629
. . . . . . .
631 631 631 631 632 633 633
. . . . . .
633 634 635 635 635 636
xx
Contents
Using a Test Specimen ....................... Why Use a Test Specimen? ................... Description of the Test Specimen ............... Using the Test Specimen ..................... The Diatom: A Natural 3D Test Specimen ........ Reasons for Poor Performance ................ Sampling Problems ......................... Optical Problems .......................... Imaging Depth ............................ Singlet-State Saturation ...................... Which 3D Method Is Best? ................... Optimal 3D Light Microscopy Summary ......... Things to Remember About Deconvolution ....... Decision Time ............................ Multi-Photon Versus Single-Photon Excitation .... Widefield Versus Beam Scanning ............... Summary .................................
. . . . . . . . . . . . . . . . .
636 636 636 637 638 640 640 640 643 643 644 646 646 646 646 647 647
Alan R. Hibbs, Glen MacDonald, and Karl Garsha
. . . . . . . . . .
650 650 650 650 651 652 660 663 663 664
. . . . . . .
664 664 667 667 668 668 670
.
683
. . . . .
684 684 685 685 685
. . .
686 686 687
. . . . . .
690 691 691 693 696 697
. . . .
698 698 698 699
. . . . .
703 704 704 70S 70S
. . . .
706 708 710 710
.
714
. .
717 719
Introduction ............................... .
722
CHAPTER 39: PHOTOBLEACHING
Introduction ............................... Photobleaching ............................ Photobleaching Mechanisms .................. Reducing Photobleaching .................... Photobleaching at the Single-Molecule Level ..... Photobleaching of Single Molecules ............ Photobleaching and Photocycling of Single Fluorescent Proteins ...................... Bleaching and Autofluorescence ............... Other Fluorescent Proteins .................... Conclusion ................................
CHAPTER 40: NONLINEAR (HARMONIC GENERATION) OPTICAL MICROSCOPY Ping-Chin Cheng and C.K. Sun
672 673 674 674 675 675 678
CHAPTER 41: IMAGING BRAIN SLICES
Jan Huisken, Jim Swoger, Steffen Lindek, and Ernst H.K. Stelzer
672
CHAPTER 38: CElL DAMAGE DURING MULTI-PHOTON MICROSCOPY Karsten Konig
Introduction ............................... . Photochemical Damage in Multi-Photon Microscopes ............................. .
682 682
Introduction ............................... Harmonic Generation ....................... Second Harmonic Generation ................. Third Harmonic Generation ................... Multi-Photon Absorption and Fluorescence ....... light Sources and Detectors for Second Harmonic Generation and Third Harmonic Generation Imaging ................................. Nonlinear Optical Microscopy Setup ........... Optically Active Biological Structures ........... Optically Active Structures in Plants ............ Optically Active Structures in Animal Tissues ................................ Polarization Dependence of Second Harmonic Generation ............................. Summary .................................
CHAPTER 37: SElECTIVE PLANE ILLUMINATION MICROSCOPY
Introduction ............................... . Combining light Sheet Illumination and Orthogonal Detection ..................... . Selective Plane Illumination Microscopy Setup ... . Lateral Resolution .......................... . light Sheet Thickness and Axial Resolution ...... . Applications ............................... . Processing Selective Plane Illumination Microscopy Images/Multi-View Reconstruction ........... . Summary ................................. .
. .
Alberto Diaspro, Giuseppe Chirico, Cesare Usai, Paola Ramoino, and Jurek Dobrucki
CHAPTER 36: PRACTICAL CONFOCAL MICROSCOPY
The Art of Imaging by Confocal Microscopy ..... Balancing Multiple Parameters ................ Monitoring Instrument Performance ............ Illumination Source ......................... Scan Raster and Focus Positioning .............. Optical Performance and Objective Lenses ........ Signal Detection ........................... Optimizing Multi-Labeling Applications ......... Control Samples Establish the Limits ............ Separation of Fluorescence into Spectral Regions ... Sequential Channel Collection to Minimize Bleed-Through .......................... Spectral Unmixing ......................... Colocalization ............................. Image Collection for Colocalization ............. Quantifying Colocalization ................... Spatial Deconvolution in Colocalization Studies .... Discussion ................................
Absorbers and Targets in Biological Specimens .... Laser Exposure Parameters ................... Evidence for Near Infrared-Induced Reactive Oxygen Species Formation ........................ Evidence for Near Infrared-Induced DNA Strand Breaks ................................ Photodynamic-Induced Effects ................. Photothermal Damage ....................... Damage by Optical Breakdown ............... Modifications of Ultrastructure ................ Influence of Ultrashort Near Infrared Pulses on Reproductive Behavior ..................... Nanosurgery ............................... Conclusion ................................
680 682
Ayumu Tashiro, Gloster Aaron, Dmitriy Aronov, Rosa Cossart, Daniella Dumitriu, Vivian Fenstermaker, Jesse Goldberg, Farid Hamzei-Sichani, Yuji Ikegaya, Sila Konur, Jason MacLean, Boaz Nemet, Volodymyr Nikolenko, Carlos Portera-Cailliau, and Rafael Yuste
Contents
Making Brain Slices ......................... Acute Slices .............................. Cultured Slices ............................ Labeling Cells .............................. Biolistic Transfection ....................... Genetic Manipulation with Dominant-Negative and Constitutively Active Mutants ............... Diolistics and Calistics ...................... Dye Injection with Whole-Cell Patch Clamp ...... Slice Loading and "Painting" with Acetoxymethyl Ester Indicators .......................... Grccn Fluorescent Protein Transgenic Mice ....... Imaging Slices ............................. Two-Photon Imaging of Slices ................. Slice Chamber Protocol ...................... Choice of Objectives ........................ Beam Collimation and Pulse Broadening ......... Image Production, Resolution, and z-Sectioning .... Choice of Indicators for Two-Photon Imaging of Calcium ............................... Photodamage ............................. Second Harmonic Imaging ................... Silicon-Intensified Target Camera Imaging ........ Morphological Processing and Analysis ......... Biocytin Protocol .......................... Anatomy with a Two-Photon/Neurolucida System ................................ Correlated Electron Microscopy ................ Morphological Classification of Neurons Using Cluster Analysis ......................... Image Processing ........................... Compensation for the Drift and the Vibration of thc Slices .............................. Alignment Based on the Overlap Between Images ................................ Alignment Based on the Center of Mass ......... Online Cell Detection of Neurons .............. Image De-Noising Using Wavelets .............. Summary .................................
. . . . .
722 722 724 724 724
. . .
725 726 726
. . . . . . . .
726 727 727 727 727 727 728 729
. . . . . .
729 729 729
. .
731
730 730 730 731
. .
731
.
732
. . . . .
732
732
732
733 734 734
CHAPTER 42: FLUORESCENT ION MEASUREMENT
Mark B. Cannell and Stephen H. Cody Introduction ............................... The Limiting Case ......................... Choice of Indicator ......................... Introducing the Indicators into Cells ........... Care of Fluorescent Probes ................... Interpretation of Measurements ............... Kinetics ................................... Calibration ................................ Conclusion ................................
. . . . . . . . .
736 736
737 738 739 740 741 742 745
CHAPTER 43: CONFOCAL AND MULTI-PHOTON IMAGING OF LIVING EMBRYOS
Jefl Hardin Introduction ............................... . Into the Depths: Embryos Are Thick, Refractile, and Susceptible to Photodamage ............ . Imaging Embryos Often Requires "40" Imaging ... .
746 746 746
The Quest for Better Resolution: Aberration and the Challenge of Imaging Thick Embryos ....... . Embryos Are Highly Scattering and Refractile Specimens ............................. . Imaging Embryos Involves Inherent Trade-Offs .... . Common Themes in Living Embryo Imaging Have System-Specific Solutions .................. . Dealing with Depth: Strategies for Imaging Thick Specimens .......................... . Avoiding the Thickness Dilemma: Going Small .... . Grazing the Surface: Superficial Optical Sections Are Often Sufficient ...................... . Up from the Deep: Explants Can Reduce the Thickness of Specimens Dramatically ......... . Multi-Photon Microscopy Can Penetrate More Deeply into Specimens .................... . Selective Plane Illumination Can Provide Optical Sectioning in Very Thick Specimens .......... . Deconvolution and Other Post-Acquisition Processing ............................. . Striving for Speed: Strategies for Reducing Specimen Exposure ....................... . Simple Solutions: Reducing Image Dimensions, Increasing Slice Spacing, and Scan Speed ....... . Disk-Scanning Confocal Microscopy Allows High-Speed Acquisition .................... . Additional Hardware Improvements Can Increase Acquisition Speed ........................ . Localizing Label: Strategies for Increasing Effective Contrast in Thick Specimens ......... . Addition of Labeled Proteins to Embryos ......... . Expressing Green Fluorescent Protein and mRFP Constructs in Embryos Allows Dynamic Analysis of Embryos at Multiple Wavelengths ............ . Using Selective Labeling to Reduce the Number of Labeled Structures ...................... . Bulk Vital Labeling Can Enhance Contrast ........ . Seeing in Space: Strategies for 40 Visualization ... . Depicting Embryos in Time and Space: 2D + Time Versus 3D + Time ........................ . Other Uses for Confocal and Multi-Photon Microscopy in Imaging and Manipulating Embryos ................................ . Multi-Photon-Based Ablation ................. . Fluorescence Resonance Energy Transfer ......... . Conclusions: A Bright Future for 3D Imaging of Living Embryos ........................... .
xxi
747 747 747 748 748 748 748 748 749 751 751 753 753 754 754 755 756
756 757
760 761 762
764 764 764 766
CHAPTER 44: IMAGING PLANT CELLS
Nuno Moreno, Susan Bougourd, Jim Haseloff, and Jose A. Feij6 Introduction ............................... The Ever Present Problem of Autofluorescence ... Single-Photon Confocal Microscopy ............ Staining Plant Tissues ....................... Clearing Intact Plant Material ................. 3D Reconstruction ......................... 3D Segmentation .......................... Two-Photon Excitation: Are Two Better Than One? .............................. Improved Signal-to-Noise Ratio and Dynamic Range .................................
. . . . . . .
774 774 775 776
.
778
.
778
769
770 772
xxii
Contents
Imaging Thick/Opaque Specimens .............. Fading, Vital Imaging, and Cell Viability ......... Two-Photon Imaging of Plant Cells and Organelles .............................. Two-Photon Excitation Imaging of Green Fluorescent Protein ................................ Dynamic Imaging ........................... Deconvolution ............................. Conclusion ................................
. .
779 779
.
782
. . . .
782 783 784 785
. .
803
. .
804 806
.
809
. . . . .
810 81l
803
CHAPTER 46: AUTOMATED CONFOCAL IMAGING AND HIGH-CONTENT SCREENING FOR CYTOMICS
Maria A. DeBernardi, Stephen M. Hewitt, and Andres Kriete
CHAPTER 45: PRACTICAL FLUORESCENCE RESONANCE ENERGY TRANSFER OR MOLECULAR NANOBIOSCOPY OF LIVING CELLS
Irina Majoul, Yiwei Jia, and Rainer Duden Introduction ............................... . How to Make a Good Science ................. . Beauty, Functionality, Cell Cycle, and Living-Cell Imaging ...................... . Fluorescence Resonance Energy Transfer Theory .. . Fluorescent Proteins and Fluorescence Resonance Energy Transfer ........................... . Qualitative Analysis ........................ . Preparation ............................... . Nanobioscopy of Protein-Protein Interactions with Fluorescence Resonance Energy Transfer ... Methods of Fluorescence Resonance Energy Transfer Measurement ........................... . Sensitized Emission of Acceptor ............... . Donor Fluorescence ........................ . Acceptor Bleach ........................... . Fluorescent Proteins as Fluorescence Resonance Energy Transfer Pairs ...................... . Cyan Fluorescent Protein and Yellow Fluorescent Protein - The Commonly Used Fluorescence Resonance Energy Transfer Pair .............. . Cyan Fluorescent Protein or Green Fluorescent Protein Forms a Fluorescence Resonance Energy Transfer Pair with mRFPl ......................... . Fluorescence Resonance Energy Transfer-Based Sensors ................................ . Fluorescence Resonance Energy Transfer and Other Complementary Methods .................. . Fluorescence Resonance Energy Transfer and Fluorescence Lifetime Imaging Microscope ..... . Fluorescence Recovery After Photobleaching and Fluorescence Loss in Photobleaching .......... . Fluorescence Resonance Energy Transfer and Fluorescence Correlation Spectroscopy ......... . Fluorescence Resonance Energy Transfer and Total Internal Reflection Fluorescence .............. . Quantum Dots and Fluorescence Resonance Energy Transfer ............................... . Cloning and Expression of Fluorescent Constructs for Fluorescence Resonance Energy Transfer ... . Cloning of Fluorescent Chimeras ............... . Functional Activity of Expressed Constructs ....... . Expression and Over-Expression ............... . Methods for Introducing Chromophores into Living Cells .............................. . Electroporation ............................ .
Transfection Reagents ....................... Microinjection ............................ Future Perspectives: 3D Microscopy, Biological Complexity, and In Vivo Molecular Imaging .... In Vivo Molecular Imaging ....................
788 788 790 790 794 795 795 795 795 795 796 797 798
798
798 798 799 799 801 801 801 801
801 801 802 802 803
803
Introduction ............................... Platforms Used for Automated Confocal Imaging ................................. Types of Assays ............................. 3D Cell Microarray Assays .................... Data Management and Image Informatics ....... Conclusion ................................
815 816 817
CHAPTER 47: AUTOMATED INTERPRETATION OF SUBCELLULAR LOCATION PATTERNS FROM THREE DIMENSIONAL CONFOCAL MICROSCOPY
Ting Zhao and Robert F. Murphy Introduction ............................... Protein Subcellular Location .................. Overview of 2D Dataset Analysis .............. High-Resolution 3D Datasets ................. 3DHeLa ................................. 3D3T3 .................................. Image Acquisition Considerations When Using Automated Analysis ...................... Image Processing and Analysis ................ Segmentation of Multi-Cell Images and Preprocessing ........................... 3D Subcellular Location Features .............. Automated Classification of Location Patterns .... Classification of 3DHeLa Dataset .............. Downsampled Images with Different Gray Scales ... Clustering of Location Patterns: Location Proteomics .............................. Exclusion of Outliers ....................... Determination of Optimal Clustering ............ Statistical Comparison of Location Patterns ...... Image Database Systems ..................... Future Directions ...........................
. . . . . .
818
818 818 820 820 820
. .
821 822
. . . . .
822 822 824 824 824
. . . . . .
825 825 825 826 827 827
. . . .
829 830 831 832
. . . .
832 835 836 838
CHAPTER 48: DISPLAY AND PRESENTATION SOFTWARE
Felix Margadant Introduction ............................... Testing ................................... "Static" Image Performance .................. Brightness ............................... Resolution: Changing the Display Size of Your Images ................................ Compression .............................. Motion Pictures ............................ Coding Limitations .........................
Contents Up-Sampling or Frame Rate Matching ........... Motion Picture Artifacts ..................... The MPEG Formats ........................ MPEG Display Formats ..................... Very High Resolutions ...................... Movie Compression and Entropy ............... Performance Benchmark ..................... Storing Your Presentation for Remote Use ........ Taking Your Presentation on the Road: Digital Rights Management and Overlaying ................
. . . . . . . .
838 839 840 840 841 841 841 842
.
844
CHAPTER 49: WHEN LIGHT MICROSCOPE RESOLUTION IS NOT ENOUGH: CORRELATIONAL LIGHT MICROSCOPY AND ELECTRON MICROSCOPY Paul Sims, Ralph Albrecht, James B. Pawley, Victoria Centonze, Thomas Deerinck, and Jeff Hardin
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Early Correlative Microscopy .................. Early 4D Microscopy ........................ Correlative Light Microscope/Electron Microscope Today ................................... Light Microscope and Electron Microscope Have Different Requirements ..................... Finding the Same Cell Structure in Two Different Types of Microscope: Light Microscope/Scanning Electron Microscope ....................... Finding the Same Cell Structure in Two Different Types of Microscope: Light Microscope/Transmission Electron Microscope ....................... Cryo-Immobilization Followed by Post-Embedding Confocal Laser Scanning Microscopy on Thin Sections ................................ Tiled Montage Transmission Electron Microscope Images Aid Correlation ..................... Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
846 846 846 846 846
850
852
856 858 860
867 867 867
CHAPTER 51: CONFOCAL MICROSCOPY OF BIOFILMS - SPATIOTEMPORAL APPROACHES R.J. Palmer, Jr., Janus A.J. Haagensen, Thomas R. Neu, and Claus Sternberg
Introduction ............................... . Sample Presentation ........................ . Flowcells and Other Perfusion Chambers ......... . Water-Immersible Lenses .................... . Upright Versus Inverted Microscopes ............ . Setup of a Flow Chamber System Setup - A Practical Example ............................... . Making Bacteria Fluorescent .................. . Fluorescent Proteins ........................ . Stains .................................... . Nucleic Acid Stains ......................... . Live/Dead Stain ........................... . Fluorescence In Situ Hybridization .............. . General Procedure for Embedding of Flowcell-Grown Biofilms for Fluorescence In Situ Hybridization .. . Antibodies ............................... . Preparation of Labeled Primary Antibodies ........ . Imaging Bacteria Without Fluorescence .......... . Imaging Extracellular Polymeric Substances in Biofilms ............................... . Application of Two-Photon Laser-Scanning Microscopy for Biofilm Analysis .............. . Limitations of Confocal Laser Scanning Microscopy and Two-Photon Laser-Scanning Microscopy in Biofilm Analysis ......................... . Temporal Experiments ....................... . Time-Lapse Confocal Imaging ................. . Summary and Future Directions ............... .
870 870 870 872 872 872 873 873 874 874 875 875 876 877 878 879 879 882
884 885 885 887
CHAPTER 52: BIBLIOGRAPHY OF CONFOCAL MICROSCOPY
CHAPTER 50: DATABASES FOR TWO- AND THREE-DIMENSIONAL MICROSCOPICAL IMAGES IN BIOLOGY
Robert H. Webb
Steffen Lindek, Nicholas J. Salmon, and Ernst H.K. Stelzer
Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Data and Metadata Management in Microscopes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Recent Developments ........................ Image Information Management. . . . . . . . . . . . . . . . The Aims of Modern Microscope System Design . . . Instrument Database Model ................... System Requirements ........................ Image Database Model ....................... Selected Projects ............................ BioImagc ......... . . . . . . . . . . . . . . . . . . . . . . . . Biomedical Image Library . . . . . . . . . . . . . . . . . . . . . Scientific Image DataBase. . . . . . . . . . . . . . . . . . . . . Other Projects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Criteria and Requirements for Microscopy Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . User Interface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Query by Content . . . . . . . . . . . . . . . . . . . . . . . . . . .
Metadata Structure ......................... . Digital Rights Management ................... . Future Prospects ........................... .
xxiii
861 861 861 862 862 864 864 864 865 865 866 866 866
A. B. C. D. E.
F. G. H. I.
J. K.
l. M. N. O. P.
866 866 866
Q. R. S.
Book and Review Articles ................. Historical Interest ....................... Theory (Mostly) ......................... Technical .............................. General ............................... Adaptive Optics ......................... Differential ............................. Display ................................ Fiber-Optic Confocal Microscopes .......... Index Mismatch ......................... Multiplex .............................. Nonlinear .............................. Polarization ............................ Profilometry ............................ Point Spread Function .................... Pupil Engineering ........................ Thickness .............................. Turbidity ............................... Variants on the Main Theme ...............
. . . . . . . . . . . . . . . . . . .
889 889 890 891 891 892 892 892 893 893 894 894 894 895 895 896 896 896 897
xxiv
Contents
APPENDIX 1: PRACTICAL TIPS FOR TWO-PHOTON MICROSCOPY
APPENDIX 3: MORE THAN YOU EVER REALLY WANTED TO KNOW ABOUT CHARGE-COUPLED DEVICES
Mark B. Cannell, Angus McMarland, and Christian Saeller
James B. Pawley
Introduction ............................... Laser Safety ............................... Laser Alignment ............................ Testing Alignment and System Performance ...... Laser Settings and Operation ................. Monitoring Laser Performance ................ Power Levels and Trouble-Shooting ............ Choice of Pulse Length ...................... Controlling Laser Power ..................... Am I Seeing Two-photon Excited Fluorescence or. .. . ...................... Stray Light and Non-Descanned Detection ...... Laser Power Adjustment for Imaging at Depth .... Simultaneous Imaging of Multiple Labels ........ Minimize Exposure During Orientation and Parameter Setting ......................... Ultraviolet-Excited Fluorochromes .............
. . . . . . . . .
900 900 900 900 90 I 901 903 903 903
. . . .
904 904 904 904
. .
905 905
APPENDIX 2: LIGHT PATHS OF THE CURRENT COMMERCIAL CONFOCAL LIGHT MICROSCOPES USED IN BIOLOGY
James B. Pawley Introduction ............................... BD-CARV II ............................... LaVision-BioTec TriM-Scope .................. Leica TCS SP2 AOBS-MPRS ................... Nikon C1si ................................ Olympus Fluoview 1000-DSU ................. Visitech VT Infinity-VT-eye .................... Yokogawa CSU 22 .......................... Zeiss LSM-5-LlVE Fast Slit Scanner-LSM 510 META-FCS ..............................
. . . . . . . .
906 907 907 910 911 912 914 915
.
916
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Part I: How Charge-Coupled Devices Work. . . . . . . Charge Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . Readout Methods ........................... What Could Go Wrong? ...................... Quantum Efficiency ......................... Edge Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Charge Loss ............................... Leakage or "Dark Charge" .................... Blooming . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Incomplete Charge Transfer . . . . . . . . . . . . . . . . . . . . Charge Amplifiers ........................... What Is a Charge Amplifier? ....... . . . . . . . . . . . . FET Amplifier Performance . . . . . . . . . . . . . . . . . . . . Noise Sources in the Charge-Coupled Device . . . . . Fixed Pattern Noise. . . . . . . . . . . . . . . . . . . . . . . . . . Noise from the Charge Amplifier. . . . . . . . . . . . . . . . Where Is Zero? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A New Idea: The Gain Register Amplifier!! ..... . . . Of Course, There Is One Snag! ................. Part II: Evaluating a Charge-Coupled Device. . . . . . A. Important Charge-Coupled Device Specs for Live-Cell Stuff! ........................ B. Things That Are (Almost!) Irrelevant When Choosing a Charge-Coupled Device for Live-Cell Microscopy ........................... C. A Test You Can Do Yourself!!! . . . . . . . . . . . . . . . D. Intensified Charge-Coupled Devices . . . . . . . . . . .
919 919 919 920 920 920 921 921 921 921 923 923 923 924 924 924 925 925 925 926 927
Index.....................................
933
927
929 930 930
Contributors
Gloster Aaron Department of Biological Sciences, Columbia University, New York, NY 10027, USA Muhammad-Amri Abdul-Karim Center for Subsurface Sensing and Imaging Systems, Rensselaer Polytechnic Institute, Troy, NY 12180, USA Asad Abu-Tarif AutoQuant Imaging, Inc., Watervliet, NY 12189, USA A.V. Agronskaia Department of Molecular Biophysics, Utrecht University, The Netherlands Ralph Albrecht Department of Animal Sciences, University of Wisconsin-Madison, Madison, WI, 53706, USA Khalid AI-Kofahi Center for Subsurface Sensing and Imaging Systems, Rensselaer Polytechnic Institute, Troy, NY 12180, USA Omar AI-Kofahi Center for Subsurface Sensing and Imaging Systems, Rensselaer Polytechnic Institute, Troy, NY 12180, USA Dmitriy Aronov Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Susan Bougourd Department of Biology, University of York, York YOlO 5YW, UK Mark B. Cannell School of Medicine and Health Sciences, University of Auckland, New Zealand Victoria Centonze Department of Cellular and Structural Biology, Optical Imaging Facility, University of Texas Health Science Center at San Antonio, San Antonio, TX 78229, USA Ji-Xin Cheng Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA and Department of Biomedical Engineering, Purdue University, West Lafayette, IN 47907, USA Ping-Chin Cheng Department of Electrical Engineering, Advanced Microscopy and Imaging Laboratory, State University of New York, Buffalo, NY 14260, USA and Department of Biological Sciences, National University of Singapore, Singapore Giuseppe Chirico Department of Physics, University of Milano Bicocca, Milan, Italy and INFM, The National Institute for the Physics of Matter, Italy
Jonathan Art Department of Anatomy and Cell Biology, University of Illinois College of Medicine, Chicago, IL 60612, USA
Stephen H. Cody Central Resource for Advanced Microscopy, Ludwig Institute for Cancer Research, PO Royal Melbourne Hospital, Parkville Victoria, 3050, Australia
Robert Bacallao Division of Nephrology, Indiana University School of Medicine, Indianapolis, IN 40202, USA
Rosa Cossart Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Jorg Bewersdorf Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, 37070 Gottingen, Germany
Guy Cox Electron Microscope Unit, University of Sydney, NSW 2006, Australia
David Biggs AutoQuant Imaging, Inc., Watervliet, NY 12189, USA
Michael E. Dailey Department of Biological Sciences, University of Iowa, Iowa City, IA 52242, USA
Maria A. DeBernardi Department of Biology, The Johns Hopkins University, Baltimore, MD 20850, USA Thomas Deerinck National Center for Microscopy and Imaging Research, University of California-San Diego, La Jolla, CA 92093, USA Peter Delaney Optiscan Pty. Ltd., Mount Waverley MDC, Victoria, 3149, Australia Winfried Denk Max-Planck Institute for Medical Research, Heidelberg, Germany Alberto Diaspro Department of Physics, IFOM, LAMBSMicroScoBio Research Center, University of Genoa, Genoa, Italy and INFM, The National Institute for the Physics of Matter, Italy Jurek Dobrucki Department of Biophysics, Laboratory of Confocal Microscopy and Image Analysis, Jagiellonian University, Krak6w, Poland A. Draaijer Department of Analytical Sciences, TNO-Voeding, Utrechtseweg 48, 3704 HE Utrecht, The Netherlands Rainer Duden Royal Holloway University of London, School of Biological Sciences, Egham TW20 OWEX, UK Daniella Dumitriu Department of Biological Sciences, Columbia University, New York, NY 10027, USA Marcus Dyba Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, D-37018 Goettigen, Germany xxv
xxvi
Contributors
Alexander Egner Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, 37070 Gottingen, Germany lauren Ernst Department of Biological Sciences and Molecular Biosensors and Imaging Center, Carnegie Mellon University, Pittsburgh, PA 15213, USA Jose A. Feij6 Department Biologia Vegetal, Universidade de Lisboa, Faculdade de Ciencias, Campo Grande, C2, PT-1749-016 Lisboa, Portugal Vivian Fenstermaker Department of Biological Sciences, Columbia University, New York, NY 10027, USA Scott E. Fraser Division of Biology, Beckman Institute (139-74), California Institute of Technology, Pasadena, CA 91125, USA Xiaosong Gan Centre for Microphotonics, Swinburne University, Australia Karl Garsha Imaging Technology Gray, Beckman Institute for Advanced Science Techology, Universtiy of Illinois at Urbana-Champaign, IL 61801, USA H.C. Gerritsen Department of Molecular Biophysics, Utrecht University, 3508 TA Utrecht, The Netherlands Jesse Goldberg Department of Biological Sciences, Columbia University, New York, NY 10027, USA Enrico Gratton Department of Physics, University of Illinois at Urbana-Champaign, Laboratory for Fluorescence Dynamics, Urbana, IL 61801, USA
Farid Hamzei-Sichani Department of Biological Sciences, Columbia University, New York, NY 10027, USA Jeff Hardin Department of Zoology, University of Wisconsin-Madison, Madison, WI 53706, USA Martin Harris Optiscan Pty. Ltd., Mount Waverley MDC, Victoria, 3149, Australia Jim Haseloff Department of Plant Sciences, University of Cambridge, Cambridge CB2 3EA, UK Rainer Heintzmann Randall Division, King's College London, London SEl lUL, UK Stefan W. Hell Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, D-37018 Goettigen, Germany Stephen M. Hewitt Tissue Array Research Program, Laboratory of Pathology, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD 20886, USA Alan R. Hibbs Biocon, 2 Vista Court, Melbourne, VIC 3135, Australia Timothy J. Holmes AutoQuant Imaging, Inc., Watervliet, NY 12189, USA and Rensselaer Polytechnic Institute, Troy, NY 12180, USA lutz Horing Carl Zeiss AG, 73447 Oberkochen, Germany Jan Huisken Light Microscopy Group, European Molecular Biology Laboratory (EMBL), 69012 Heidelberg, Germany Yuji Ikegaya Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Yiwei Jia SEG, Olympus America Inc., Melville, NY 11747, USA lain D. Johnson Molecular Probes, Inc., Eugene, OR 97402, USA Rimas Juskaitis Department of Engineering Science, University of Oxford, Oxford OXl 3PJ, UK lars Kastrup Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, D-37108 Gottigen, Germany H. Ernst Keller Carl Zeiss, Inc., Thornwood, NY 10594, USA Karsten Konig Fraunhofer Institute of Biomedical Technology, D-661119 Saarbrticken, Germany Sila Konur Department of Biological Sciences, Columbia University, New York, NY 10027, USA Andres Kriete Coriell Institute for Medical Research and Drexel University, Philadelphia, PA 19104, USA Gang lin Center for Subsurface Sensing and Imaging Systems, Rensselaer Polytechnic Institute, Troy, NY 12180, USA Steffen Lindek Light Microscopy Group, European Molecular Biology Laboratory (EMBL), 69012 Heidelberg, Germany Glen MacDonald V.M. Bloedel Hearing Research Center, University of Washington, Seattle, WA 98195, USA
Min Gu Centre for Microphotonics, Swinburne University, Australia
Shinya Inoue Marine Biological Laboratory, Woods Hole, MA 02543, USA
Jason Maclean Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Janus A.J. Haagensen Center for Biomedical Microbiology, BioCentrum, Technical University of Denmark, 2800, Lyngby, Denmark
Stefan Jakobs Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, D-37108 Gottigen, Germany
Irina Majoul Royal Holloway University of London, School of Biological Sciences, Egham TW20 OWEX, UK
Contributors
xxvii
Erik Manders
David W. Piston
Christian Soeller
Centre for Advanced Microscopy, Section of Molecular Cytology, Swammerdam Institute for Life Sciences, Faculty of Science, University of Amsterdam, 1090 GB Amsterdam, The Netherlands
Vanderbilt University, Nashville, TN 37232, USA
School of Medicine and Health Sciences, University of Auckland, New Zealand
Felix Margadant
Carlos Portera-Cailliau
Sadaf Sohrab
Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Division of Nephrology, Indianapolis, IN 40202, USA
Eric Potma
CarverlEmil Witschi Professor in the Biological Sciences and Director, W.M. Keck Dynamic Image Analysis Facility, Department of Biological Sciences, University of Iowa, Iowa City, IA 52242, USA
Zurich, Switzerland
Angus McMorland School of Medicine and Health Sciences, University of Auckland, New Zealand
Nuno Moreno
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA
Paola Ramoino
Centro de Biologia do Desenvolvimento, Instituto Gulbenkian de Ciencia, PT-2780156 Oeiras, Portugal
Department for the Study of the Territory and Its Resources, University of Genoa, Italy
Robert F. Murphy Departments of Biomedical Engineering and Biological Sciences and Center for Bioimage Informatics, Carnegie Mellon University, Pittsburgh, PA 15213, USA
Jens Rietdorf
Boaz Nemet Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Thomas R. Neu Department of River Ecology, UFZ Center for Environmental Research Leipzig-Halle, 39114, Magdeburg, Germany
Volodymyr Nikolenko Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Andreas Nolte Carl Zeiss MicroImaging GmbH, 37081, Goettingen, Germany
R.J. Palmer, Jr. Oral Infection and Immunity Branch, National Institute of Dental and Craniofacial Research, National Institutes of Health, Bethesda, MD 20886, USA
Light Microscopy Group, European Molecular Biology Laboratory, 69012 Heidelberg, Germany
Maitreyee Roy Department of Physical Optics, School of Physics, University of Sydney, NSW 2006, Australia
Badrinath Roysam Center for Subsurface Sensing and Imaging Systems, Rensselaer Polytechnic Institute, Troy, NY 12180, USA
Seth W. Ruffins Division of Biology, Beckman Institute (139-74), California Institute of Technology, Pasadena, CA 91125, USA
Nicholas J. Salmon SLS Software Technologies GmbH, Heidelberg, Germany
William Shain New York State Department of Health, The Wadsworth Center, Albany, NY 12201, USA
Peter J. Shaw Department of Cell and Developmental Biology, John Innes Centre, Colney, Norwich NR4 7UH, UK
James B. Pawley Department of Zoology, University of Wisconsin-Madison, Madison, WI 53706, USA
Carrie Phillips Division of Nephrology, Indianapolis, IN 40202, USA
Ernst H.K. Stelzer Light Microscopy Group, European Molecular Biology Laboratory 69012 Heidelberg, Germany Claus Sternberg Center for Biomedical Microbiology, BioCentrum, Technical University of Denmark, 2800, Lyngby, Denmark C.K. Sun Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan, Republic of China
Jim Swoger Light Microscopy Group, European Molecular Biology Laboratory 69012 Heidelberg, Germany
Donald H. Szarowsk New York State Department of Health, The Wadsworth Center, Albany, NY 12201, USA Ayumu Tashiro Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Mark Terasaki Department of Physiology, University of Connecticut Health Center, Farmington, CT 06032, USA
Derek Toomre
Colin J.R. Sheppard
Department of Cell Biology, Yale University School of Medicine, New Haven, CT 06520, USA
Division of Bioengineering, National University of Singapore, Singapore 117576, and Department of Diagnostic Radiology, National University of Singapore, 119074, Singapore
Roger Y. Tsien Department of Pharmacology 0647, School of Medicine, University of California, San Diego, CA 92093, USA
Michael J. Paulus ORNL and CTI-Concorde Microsystems, LLC, Knoxville, TN 37932, USA
David R. 5011
Paul Sims Department of Zoology, University of Wisconsin-Madison, Madison, WI 53706, USA
James N. Turner New York State Department of Health, The Wadsworth Center, Albany, NY 12201, USA
xxviii
Contributors
J. Michael Tyszka
Alan Waggoner
N.S. White
Division of Biology, Beckman Institute (139-74), California Institute of Technology, Pasadena, CA 91125, USA
Department of Biological Sciences and Molecular Biosensors and Imaging Center, Carnegie Mellon University, Pittsburgh, PA 15213, USA
Sir William Dunn School of Pathology, University of Oxford, Oxford OXI 3RE,
Cesare Usai
Robert H. Webb
Institute of Biophysics, National Research Council, Genoa, Italy
Schepens Eye Research Institute and Wellman Center for Photomedicine, Boston, MA 02114, USA
O.J. van den Heuvel Department of Molecular Biophysics, Utrecht University, 3508 TA Utrecht, The Netherlands
UK Katrin I. Willig Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, D-37018 Gottigen, Germany X. Sunney Xie
Watt W. Webb School of Applied and Engineering Physics, Cornell University, Ithaca, NY 14853, USA
Department of Chemistry and Chemical Biology, Harvard University, Cambridge, MA 02138, USA
Rafael Yuste Martin J. vandeVen
Jamey P. Weichert
Department of Cell Physiology, Microftuorimetry Section, Biomedical Institute, Limburg University Center and Trans National University Limburg, University Campus Building D, and Institute for Materials Research IMOIIMOMEC, Wetenschapspark 1, Diepenbeek, Belgium
Radiology Department, University of Wisconsin-Madison, BX 3252 Clinical Science Center, Madison, WI 53792, USA
Department of Biological Sciences, Columbia University, New York, NY 10027, USA
Ting Zhao Volker Westphal Department of NanoBiophotonics, Max Planck Institute for Biophysical Chemistry, D-37018 Gottigen, Germany
Departments of Biomedical Engineering and Biological Sciences and Center for Bioimage Informatics, Carnegie Mellon University, Pittsburgh, PA 15213, USA
1
Foundations of Confocal Scanned Imaging in Light Microscopy Shinya Inoue
Seldom has the introdu ction of a new instrument generated as instant an excitement amon g biologists as the laser-scanning confocal microscope. With the new microscope, one can slice incredibly clean, thin optical sections out of thick fluorescent specimen s; view specimens in planes tilted to, and even running parallel to, the line of sight; penetrate deep into light-scattering tissue s; gain impressive three-dimensional (3D) views at very high resolut ion; obtain differential interference or phase-contrast images in exact register with confocal fluorescence images ; and improve the precis ion of microphotometry. While the instrument that engendered such excitement became commercially available first in 1987, the optical and electronic theory and the technolo gy that led to this sudden emergence had been brewing for several decades. The developm ent of this microscope stems from several roots, including light microscopy, confoca l imaging, video and sca nning microscopy, and coher ent or laser-illuminated optics (see historic overview in Table l .l ). In this chapter, I will first discuss some basic principles relating to lateral and axia l resolution as well as depth of field in light microscopy, highlight some history that lays a foundation to the development of laser-scanning confocal microscopy, and end with some general remarks regarding the new microscopes, includ ing a disk-scannin g confoca l system.
LIGHT MICROSCOPY Lateral Resolution 1 The foundations of modem light microscopy were estab lished a century ago by Ernst Abbe ( 1873, 1884). He demonstrated how the diffraction of light by the specimen, and by the objective lens, determined image resolution; defined the conditions needed to design a lens whose resolution was diffraction limited (rather than limited by chromatic and spherica l aberrations); and establ ished the role of the objective and condenser numerical apertur es (NA) on image resoluti on (Eq. I). Thu s,
d . = min
1.22A. 0 NA Obj + NA cond
(1)
where dm;n is the minimum spacing in a periodic grating that can ju st be resolved. d min is expressed as lateral distance in the specimen space; A.n is the wavelength of light in vacuum; and NA obj and NA cond are the numerical apertures of the objective and condenser lenses, respecti vely. The NA is the produ ct of the sine of the halfangle (o) of the cone of light either acceptable by the objec tive lens or emerging from the condenser lens and the refractive indexes (ll) of the imbibing medium between the specimen and the objective or condenser lens, respectively. Equation I demonstrates that, in addition to the wavelength and the NA of the objec tive lens, the condenser NA also affects image resolution in the microscope. For object s that are illuminated fully coher ently (a condition that pertain s when NA co nd approaches 0, namely when the condenser iris is closed down to a pinhole), the minimum resolvable lateral spacing increases (i.e., the resolution decreases) by a factor of 2 compared to the case when the condenser iris is opened so that NA cond = NAo bj ' As the conden ser iris is opened and NAco nd become s larger, the illumin ation become s progressively less coherent and resolution increases. [Note, however, that laser beams tend to illuminat e object s coherently even when the condenser iris is not closed down (see Chapter 5, this volume).] Equation I describes the relation between NA and resoluti on for line-grating objects. A complementary method of defining the limit of resolution uses point objects instead of line grating s. The image, in focus, of an infinitely small luminous object point is itself not infinitely small, but is a circular Airy diffra ction image with a central bright disk and progressively weaker concentric dark and bright rings. The radiu s rA iry of the first dark ring around the central disk of the Airy diffraction image depend s on A. and the NA of the objecti ve: r A ir y =
A. 0.6] N~ .
(2)
ob j
I
For extensive discussions o n modem microscope lens des ign and aberra tions and a more rigorous treatment of the optical principle s and applicat ions of light microscopy than is appropriate for this revi sed chapter, refer to co mplementary chapters in Handbook of Opt ics (e.g., Inoue and Oldenbourg, 1994) and in Video Microscopy , 2nd edition (Inoue and Spring. 1997).
Shinya Inou e
>
where r Airy is expre ssed as distance in the specimen plane . When there exist two points of light separated by a small distance d in the specimen plane, their diffraction images lie side by side in the image plane. The images of two equall y bright spots are said to be resolved if d is larger or equal to the radius of the Airy disk. This is the Rayleigh criterion, and it relies on the assumpt ion that the two point sources radiate incoherently. If the two point sources emit light coherently, their amplitud e rather than
Marine Biologi cal Laborator y, Woods Hole, Ma ssachu setts 025 43
Handbook of Biological Confo cal Microscopy , Third Edition , edited by James B. Pawley, Springer Scienc e+Bu siness Media, LLC , New York, 2006.
2
Chapter 1 • S. Inoue
TABLE 1.1. Historic Overviewa Confocal Microscopy
Microscopy
Video (Microscopy) Nipkow (1884)
Abbe (1873, 1884)"" Berek (1927)d
Zworykin (1934) Zernicke (1935)"" Gabor (1948)" Hopkins (1951)"
Flory (1951) Young and Roberts (1951) Flying spot'
Linfoot and Wolf (1953) 3D diffraction by annuL apert,a,d Tolardo di Francia (1955) Limited field h Nomarski (1955)" Linfoot and Wolf (1956) 3D diffraction pattern",d Ingelstam (1956) Resolution and info, theory"
Montgomery et ai, (1956) Flying spot UVe
Minsky Patent (1957) Insighta,b,,·,d Stage scanning Kubota and Inoue (1959)"" Smith and Osterberg (1961)" Freed and Engle (1962) Flying spot' Harris (1964 )a,h Ellis (1966) Holomicrography" Petnin et al. (1968) Tandem scanningd,g Davidovits and Egger (1971) Laser illumination Lens scanning d Hellwarth and Christensen (1974) Second harmonic generationc,d
Hoffman and Gross (1975) Modulation contrast C
Sheppard and Choudhury (1977) Theorya,b,d Sheppard et at. (1978) Stage scanning,hl Cremer and Cremer (1978) Auto-focus stage scanningd "4-rc-point illumination,,",b.d Brakenhoff et al. (1979) Specimen scana,h,d" Koester (1980) Scanning mirror"
Ellis (1978) Single sideband edge enhancement microscopyc.d
Castleman (1979) Digital image processing"'"'' Quate (1980) Acoustic microscopy"'c Inoue (1981)," Allen et ai, (1981a,b)"' Fuchs et at. (19821 Agard and Sedat (1983 )"",d
Cox and Sheppard (1983) Digital recording d,' Aslund et al. (1983) 2-mirror laser scanningd Hamilton et al. (1984) Differential phaseC.d Wilson and Sheppard (1984) Extended depth of field"d,cJ Boyde (1985a) Nipkow typed" Carlsson et at. (1985) Laser scan, Stacks of confocal images d," Wijnaendts van Resandt et al. (1985) xz-viewd,e
Ellis (1985) Light scrambler"
Sher and Barry (1985Y Fay et at. (1985)"",d
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
3
TABLE 1.1. (Continued) Microscopy
Confocal Microscopy Suzuki-Horikawa (1986) Video-rate laser scan' Acousto-optical modulator No exit pinhole Xiao and Kino (1987) Nipkow typed Amos et al. (l987),·d., McCarthy and Walker (1988) Nipkow typed Denk et al. (1990) Two photon,,·,··d,( Hell and Wichmann (1994) PSF reduction by stimulated emission depletion"·b.d
Cox and Sheppard (1986)
Video (Microscopy) Inoue (\ 986) Overview, How to a .,.,!
Castleman (1987)a
Ellis (1988) Scanned aperture phase contrast,,·,··d
Oldenbourg and Mei (1995) LC-pol system,··d Ichihara et al. (1996) High-throughput spinning diskd,fg Conchello et al. (1997) Aperture scanninga.b.,.d! Gustaffson et al. (2000) Structured illumination".!,·"d Volkmer et al. (2001) Coherent anti-stokes raman' Inoue et al. (2002) Fluorescence pol'
Inoue et al. (2001 a,b) Centrifuge pol scope'·K
"Diffraction theory. h Superresolution. , Contrast modes. d Optical sectioning/depth of field. 'Stereo. 13D in objective space. 'High speed.
their intensity distribution in the image must be considered, and resolution generally decreases. The impact of the quality and NA of the condenser on the lateral resolution of point objects was considered by Hopkins and Barham (1950). Their results are similar to, but not strictly identical with, the case of line-grating objects (see Born and Wolf, 1980). It is important to realize that these resolution criteria apply only to objective lenses used under conditions in which the image is free from significant aberrations (see Chapters 6, 7, 8, 9, 11, and 22, this volume; Inoue and Oldenbourg, 1994; Chapters 2 and 3 in Inoue and Spring, 1997). This implies several things: • A well-corrected clean objective lens is used within the wavelengths of light and diameter of field for which the lens was designed (commonly in conjunction with specific oculars and/or tube lenses). • The refractive index, dispersion, and thickness of the coverslip and immersion media are those specified for the particular objective lens. • The correct tube length and ancillary optics are used and the optics are axially aligned. • The full complement of image-forming rays and light waves leaving all points of the objective-lens aperture converge to the image plane without obstruction. • The condenser aperture IS homogeneously and fully illuminated. • The condenser is properly focused to produce Kohler illumination.
These considerations for resolution assume that the specimen is viewed in conventional widefield (WF) microscopy. When the (instantaneous) field of view becomes extremely small, as in confocal microscopy, the resolution can in fact be greater than when the field of view is not so limited. We shall return to this point later. 2
Axial Resolution We now turn to the axial (z-axis) resolution, measured along the optical axis of the microscope, that is, perpendicular to the plane of focus in which the lateral resolution was considered. To define axial resolution, it is customary to use the 3D diffraction image of a point source that is formed near the focal plane. In the case of lateral resolution, that is, the resolution in the plane of focus, the Rayleigh criterion makes use of the infocus diffraction images (the central cross-section of the 3D diffraction pattern) of two point sources and the minimum distance that they can approach each other laterally, yet still be distinguished as two. Similarly, axial resolution can be defined by the minimum distance
2
Note also that one's ability to determine the location of an object is not determined by the resolution limit of the system. In fact, the location of an object (diffraction pattern) can be determined under a microscope with precisions that are many times. or even orders of magnitude, greater than the resolution limit (e.g., Denk and Webb, 1987; also see Inoue, 1989).
4
Chapter 1 • S. Inoue
that the diffraction images of two points can approach each other along the axis of the microscope, yet still be seen as two. To define this minimum distance, we use again the diffraction image of an infinitely small point object and ask for the location of the first minimum along the axis of the microscope. The precise distribution of energy in the image-forming light above and below focus, especially for high NA objective lenses, cannot be deduced by geometric ray tracing but must be derived from wave optics. The wave optical studies of Linfoot and Wolf (1956) show that the image of a point source produced by a diffraction-limited optical system (e.g., a well-designed and properly used light microscope) is not only periodic around the point of focus in the focal plane, but is also periodic above and below the focal plane along the axis of the microscope. [Such 3D diffraction images (including those produced in the presence of lens aberrations) are presented photographically by Cagnet et al. (1962; also see Chapter 7, this volume, Fig. 7.4). The intensity distribution calculated by Linfoot and Wolf for an aberration-free system is reproduced in Born and Wolf (1980) and also in Inoue and Spring (1997, Fig. 2-30). The 3D pattern of a point source formed by a lens possessing an annular aperture was calculated by Linfoot and Wolf (1953).] The distance from the center of the 3D diffraction pattern to the first axial minimum (in object space dimensions) is given by
z min
=
2"- 11 0
(NA
.)2
(3)
ob]
where 11 is the refractive index of the object medium. Zmin corresponds to the distance by which we have to raise the microscope objective in order to focus the first intensity minimum observed along the axis of the 3D diffraction pattern instead of the central maximum. 3 As with the lateral resolution limit, we can use Zmin as a measure of the limit of axial resolution of the microscope optics. Note, however, that Zmin shrinks inversely proportionally with the square of the NA obj , in contrast to the lateral resolution limit which shrinks with the first power of the NA obj ' Thus, the ratio of axialto-lateral resolution (ZminlrAiry = 3.2811INAobj) is substantially larger than "- and is inversely proportional to the NA of the objective lens.
Depth of Field The depth of field of a microscope is the depth of the image (measured along the microscope axis translated into distances in the specimen space) that appears to be sharply in focus at one setting of the fine-focus adjustment. In brightfield microscopy, this depth should be approximately equal to the axial resolution, at least in theory. The actual depth of field has been determined experimentally, and the contribution of various factors that affect the measurement have been explored by Berek (1927). According to Berek, the depth of field is affected by (1) the geometric and diffraction-limited spreading, above and below the plane of focus, of the light beam that arose from a single point in the specimen; (2) the accommodation of the observer's eye; and (3) the final magnification of the image. The second factor becomes irrelevant when the image is not viewed directly through the ocular but is instead focused onto a thin detector (as in video
3
As discussed later, the distance 2min can be reduced significantly below the classical limit given by Eq. 1.3, for example, by reducing the effective pointspread function by special use of two-photon confocal imaging.
microscopy or confocal microscopy with a minute exit pinhole). The third factor should also disappear once the total magnification is raised sufficiently, so that the unit diffraction image becomes significantly larger than the resolution element of the detector (e.g., Hansen in Inoue, 1986; Castleman, 1987, 1993; Schotten, 1993; Inoue and Spring, 1997, Section 12.2). When the detector can be considered to be infinitely thin and made up of resolution elements spaced sufficiently (at least 2-fold) finer than the Airy disk radius, then one need only to consider the diffraction-limited depth of field. In that case, the depth of field is taken to be
1 0. - -(z 4 min+ -z min- )
(4)
that is, one quarter of the distance between the first axial minima above (Zmin') and below (Zmin-) the central maximum in the 3D Airy pattern converted to distances in specimen space (see Eq. 3; Zmin' and Zmin- correspond to Zl and -Zl in Chapter 7, Fig. 7.4, this volume). In conventional fluorescence and darkfield microscopy, the light arising from each image point produces significant intensity within a solid cone that reaches a considerable distance above and below focus (as seen in the point-spread functions for these modes of microscopy; e.g., Streibl, 1985; also Chapters 11 and 23, this volume). Therefore, fluorescent (or light-scattering) objects that are out of focus produce unwanted light that is collected by the objective and reduces the contrast of the signal from the region in focus. For these reasons, the depth of field may be difficult to measure or even to define precisely in conventional fluorescence and darkfield microscopy. Put another way, one could say that when objects that are not infinitely thin are observed in conventional fluorescence or darkfield microscopy, the apparent depth of field is very much greater than the axial resolution. The unwanted light that expands the apparent depth of field is exactly what confocal imaging eliminates. Thus, we can view only those fluorescent and light-scattering objects that lie within the depth that is given by the axial resolution of the microscope and attain the desired shallow depth of field. As mentioned earlier, the lateral resolution of a microscope is also a function of the size of the field observed at anyone instant. Tolardo di Francia (1955) suggested, and Ingelstam (1956) argued on the basis of information theory, that one gains lateral resolution by a factor of -J2 as the field of view becomes vanishingly small. These theoretical considerations set the stage for the development of confocal imaging.
CONFOCAL IMAGING As a young postdoctoral fellow at Harvard University in 1957, Marvin Minsky applied for a patent for a microscope that used a stage-scanning confocal optical system. Not only was the conception farsighted, but his insight into the potential application and significance of confocal microscopy was nothing short of remarkable. [See the delightful article by Minsky (1988) that shows even greater insight into the significance of confocal imaging than do the following extracts culled from his patent application.] In Minsky's embodiment of the confocal microscope, the conventional microscope condenser is replaced by a lens identical to the objective lens. The field of illumination is limited by a pinhole, positioned on the microscope axis. A reduced image of this pinhole
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
is projected onto the specimen by the "condenser." The field of view is also restricted by a second (or exit) pinhole in the image plane placed confocally to the illuminated spot in the specimen and to the first pinhole (Fig. 1.1). Instead of trans-illuminating the specimen with a separate "condenser" and objective lens, the confocal microscope could also be used in the epi-illuminating mode, making a single objective lens serve as both the condenser and the objective lens (Fig. 1.2). Using either transmitted or epi-illumination, the specimen is scanned with a point of light by moving the specimen over short distances in a raster pattern. (The specimen stage was supported on two orthogonally vibrating tuning forks driven by electromagnets at 60 and 6000 Hz.) The variation in the amount of light, modulated by the specimen and passing the second pinhole, is captured by a photoelectric cell. The photoelectric current is amplified and modulates the beam intensity of a long-persistence cathode-ray tube (CRT) scanned in synchrony with the tuning forks. As a result, the image of the specimen is displayed on the CRT. The ratio of scanning distances between the electron beam and the specimen provides image magnification, which is variable and can be very large. With this stage-scanning confocal microscope, Minsky says, light scattered from parts other than the illuminated point on the specimen is rejected from the optical system (by the exit pinhole) to an extent never before realized . As pointed out in the patent application, there are several advantages to such an optical system: Reduced blurring of the image from light scattering Increased effective resolution Improved signal-to-noise ratio Permits unusually clear examination of thick, light-scattering objects • xy-scan possible over wide areas of the specimen • Inclusion of a z-scan is possible • Electronic adjustment of magnification
• • • •
,
==er
5
\
p
FIGURE 1.2. Optical path in epi-illuminated confocal microscope. The entrance pinhole. A, point D in the specimen, S, and exit pinhole. B, are confocal points as in Figure 1.1. A partial, or dichromatic, mirror, M \0 transmits the illuminating beam a-b-c and reflects the beam d-e which passed D and was reflected by the mirror, M 2 , on which the specimen is lying. Only the reflected beam that passes point D focuses onto the detector pinhole and reaches the photocell, P. A single lens, 0, replaces the condenser and objective lenses in Figure 1.1. (After Minsky. 1957.)
• Especially well suited for making quantitative studies of the optical properties of the specimen • An infinite number of aperture planes in the microscope are potentially available for modulating the aperture with darkfield stops, annuli, phase plates, etc . • Complex contrast effects can be provided with comparatively simple equipment • Permits use of less complex objective lenses, including those for long working distance, ultraviolet (UV), or infrared imaging, as they need to be corrected only for a single axial point. The high-resolution acoustic microscope developed by Quate and co-workers (Quate, 1980) and the laser disk, video, and audio recorder/players are object-scanning-type confocal microscopes. The designers of these instruments take advantage of the fact that only a single axial point is focused or scanned (see, e.g., Inoue and Spring, 1997, Sect. 11.10).
IMPACT OF VIDEO FIGURE 1.1 . Optical path in simple confocal microscope. The condenser lens, C, forms an image of the first pinhole, A, onto a confocal spot, D. in the specim en, S. The objective lens, 0 , forms an image of D onto the second (exit) pinholc, B, which is confocal with D and A. Another point, such as E in the specimen. would not be in focus with A, so that the illumination would be less. In addition. most of the light, g- h, scattered from E would not pass the second pinhole, B. The light reaching the phototube. P, from E is thus greatly attenuated co mpared to that from the confocal point. D. In addition, the exit pinhole can be made small enough to exclude the diffraction rings in the image of D, so that the resolving power of the microscope is improved. As the specimen is scanned . the phototube provides a signal of the light passing through sequential spec imen points D" D2 • D3 , etc. (not shown). D" D2 , D" etc., can lie in the focal plane as in conventional microscopy or perpendicular to it, or at any angle defined by the scanning pattern, so that optica l sections can be made in or at angles tilted from the conventional image plane. Because. in the stagescanning system, the small scanning spot. D, lies exactly on the axis of the microscope, the lenses C and 0 can be considerably less sophisticated than conventional microscope lenses. which must form images from points some distance away from the lens axis. (After Minsky, 1957.)
Nipkow Disk Just about the same time that Abbe in Jena laid the foundation for modem light microscopy, a young student in Berlin, Paul Nipkow (1884), figured out how to convert a two-dimensional (2D) optical image into an electrical signal that could be transmitted as a onedimensional (lD), or serial, time-dependent signal, over a single cable (as in a Morse code). Prior to Nipkow, most attempts at the electrical transmission of optical images involved the use of multiple detectors and as many cables. Nipkow dissected the image by scanning over it in a raster pattern, using a spinning opaque wheel perforated by a series of rectangular holes. The successive holes, placed a constant angle apart around the center of the disk but on constantly decreasing radii (i.e .. arranged as an Archimedes spiral), generated the rasterscanning pattern (Fig. 1.3). The brightness of each image element, thus scanned by the raster, was picked up by a photocell. The
6
Chapter 1 • S. Inoue
o
D
l:1
A
B
FIGURE 1.3. Nipkow disk. The perforations in the opaque disk, A, which is rotating at a constant velocity, scan the image in a raster pattern as shown in B. (After Zworykin and Morton, 1954.)
output of the photocell reflected the brightness of the sequentially scanned image elements and drove a neon bulb that, viewed through another (part of the) Nipkow disk, reproduced the desired picture. A similar type of scanner disk, but with multiple, centrosymmetric sets of spirally placed holes, was used by Mojmir Petnin and co-workers at Prague and New Haven to develop their epiilluminated tandem-scanning confocal microscope (TSM) (Egger and Petran, 1967; Petran et al., 1968). In Petnin's microscope, holes on a portion of the spinning disk placed in front of the lightsource collector lens are imaged onto the specimen by the objective lens. Each point of light reflected or scattered by the specimen is focused by the same objective lens back onto the centrosymmetric portion of the Nipkow disk. The pinholes at this region exclude the light originating from points in the specimen not illuminated by the first set of pinholes, giving rise to confocal operation (Chapter 10, this volume). As with Nipkow's initial attempt at television, this TSM tends to suffer from the low fraction (1 %-2%) of light that is transmitted through the source pinholes. Also, very high mechanical precision is required for fabricating the symmetrical Nipkow disk and for spinning it exactly on axis. In addition, some of the advantages pointed out by Minsky for the stage-scanning type confocal optics are lost because the objective lenses are no longer focusing a single axial point of light. However, for biological applications, the tandem-scanning system provides the decided advantage that the specimen remains stationary. As a result, the speed of the raster scan is not limited by the mass of the specimen support as it is in stage scanning, and the scanning system is unlikely to introduce any geometrical distortion. Thus, with a TSM, one can observe objects that reflect or scatter light moderately strongly, in real time, either by using a television or photographic camera or by observing the image directly through the eyepiece. In addition to Petnin and co-workers, Alan Boyde (l985a) in London took advantage of the good axial discrimination and lightpenetrating capability of the tandem scanning confocal microscope and pioneered its use for viewing biological objects. In particular, he used it for imaging below the surface of hard tissue such as bone and teeth to visualize the cells and lacunae found there (see
Lewin, 1985). Boyde also provides striking stereoscopic images obtained with the tandem-scanning confocal microscope (Boyde, 1985b, 1987). Gordon Kino and co-workers at Stanford University have designed a confocal microscope using a Nipkow disk in a manner that differs somewhat from the Petnin type (Xiao and Kino, 1987; also see Chapter 10, this volume). In the Kino type, the rays illuminating the specimen and those scattered by the specimen traverse the same set of pinholes on the spinning Nipkow disk rather than those that are centro symmetrical. By using a special lowreflection Nipkow disk, tilted somewhat to the optic axis of the microscope, and by employing crossed polarizers and a quarterwave birefringent plate to further reduce the spurious reflections from the disk, they are able to use only one side of it, thus alleviating some of the alignment difficulties of the Petran type. More recently, spinning-disk confocal units have been vastly improved by adding microlenses to the pinholes. As described later (see "Yokogawa Disk-Scanning Confocal System" below), the microlens-equipped disk-scanning systems effectively provide video-rate and faster confocal full-frame imaging and in real color. While in part depending on older technology, the confocalscanning unit (CSU) systems turn out to have certain advantages not achievable with point-scanning confocal systems.
Electron-Beam-Scanning Television While Nipkow's invention laid the conceptual groundwork for television, raster scanning based on a mechanical device was simply too inefficient for practical television. Thus, it was not until five decades after Nipkow, following the advent of vacuum tube and electronic technology, that Zworykin (1934) and his colleagues at RCA were able to devise a practical television system. These workers developed the image iconoscope, an image-storagetype electron-beam-scanning image pickup tube. The image iconoscope, coupled with a CRT for picture display, permitted very rapid, "inertialess" switching and scanning of the image and picture elements. With these major breakthroughs, television not only became practical for broadcasting but emerged as a tool that could be applied to microscopy (see Inoue and Spring. 1997, Sects. 1.1 and 1.2). An early application of video (the picture portion of television) was the flying spot UV microscope of Young and Roberts (1951). With this microscope, the specimen remains stationary and single object points are scanned serially in a raster pattern by a moving spot of UV light emitted by the face of a special high-intensity UVCRT. The optical elements (condenser) of the microscope demagnify this moving spot onto the specimen, which modulates its brightness. The modulated UV light is then picked up by a phototube and amplified electronically before being displayed on a visible-light CRT scanned in synchrony with the UV-CRT. Young and Roberts point out that by illuminating only a single specimen point at a time with a flying-spot microscope, flare is reduced and the image becomes a closer rendition of the specimen's optical properties than that obtained with a non-scanning microscope. They also point out that for these same reasons - and because a photoelectric detector can provide a sequential, linear output of the brightness of each specimen point - quantitative analysis becomes possible with a flying-spot microscope. In addition, they note that the electronic photo detector raises the sensitivity of image capture by perhaps two orders of magnitude compared to photography. It should be noted that the flare which would otherwise arise from the unilluminated parts of the specimen is significantly
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
reduced with a flying-spot microscope (see Sheppard and Choudhury, 1977), even though the exit pinhole used in a confocal microscope is not present. Thus, for example, Wilke et al. (1983) and Suzuki and Hirokawa (1986) developed laser-scanning flyingspot microscopes (coupled with digital image processors) to raise image contrast (at video rate) in fluorescence, differentialinterference-contrast (DIC), and brightfield microscopy. Naturally, the exit pinhole in a confocal system is very much more effective at excluding unwanted light arising from different layers or portions of the specimen not currently illuminated by the source "pinhole," but it does so at the cost of reduced image brightness, lower scanning speed, and increased instrumental complexity and price. While the flying-spot, or beam-scanning, microscope was developed and applied in UV microscopy for about a decade after its introduction, its further development as an imaging device was eclipsed for some time by the need and the opportunity to develop automated microscopy for rapid cell sorting and diagnosis. Here, the aim was not the imaging of cell structures as such but rather the rapid and efficient classification of cells based on their biochemical characteristics, taking advantage of the emerging power of high-speed digital computers. The size, shape, absorbance, light scattering, or light emission of cells (labeled with specific fluorescent markers) was used either to classify the cells by scanning the slide under a microscope or to sort the cells at very high rates as the cells traversed a monitoring laser beam in a flow cell or a Coulter-type cell separator.
Impact of Modern Video Meanwhile, starting in the late 1970s, the introduction of new solid state devices, especially large-scale integrated circuits and related technology, led to dramatic improvements in the performance and availability, and reduction in price, of industrial-grade video cameras, video tape recorders, and display devices. Concurrently, ever more compact and powerful digital computers and imageprocessing systems appeared in rapid succession. These advances led to the birth of modern video microscopy, which in turn brought about a revitalized interest in the power and use of the light microscope (for reviews see Allen el al., 1981 a,b; Inoue, 1981 , 1989; Allen, 1985 ; Inoue and Spring, 1997). In brief, dynamic structures in living cell s could now be visualized with a clarity, speed, and resolution never achieved before in DIe, fluorescence, polarized-light, darkfield, and other modes of microscopy ; the growth and shortening of individual molecular filaments of tubulin and f-actin, and their gliding motion and interaction with motor molecules, could be followed in real time directly on the monitor screen; and the changing concentration and distribution of ions and specific protein molecules tagged with fluorescent reporter molecules could be followed, moment by moment, in physiologically active cells (Chapters 19, 29, and 42, this volume) . In addition to its immediate impact on cellular and molecular biology, video microscopy and digital image processing also stimulated the exploration of other new approaches in light microscopy along several fronts. These include the development of ratio imaging and new reporter dyes for quantitative measurement of local intracellular pH , calcium ion concentration, etc. (Tanasugarn et al. , 1984; Bright et al., 1989; Tsien , 1989; Chapters 16, 19, and 29, this volume); the computational extraction of pure optical sections from whole-mount specimens in fluorescence microscopy (based on deconvolution of multi-layered images utilizing knowledge of the microscope's point-spread function; Agard and Sedat,
7
1983 ; Agard et al., 1989; see also Chapters 23, 24, and 25, this volume); 3D imaging including stereoscopy (Brakenhoff et al., 1986, 1989; Inoue and Inoue, 1986; Aslund et aI. , 1987; Stevens etal., 1994; Inoue and Spring, 1997, Sect. 12.7.7); and, finally, the further development of laser-scanning microscopy and confocal microscopy.
LASERS AND MICROSCOPY
Holography In 1960, Maiman announced the development of the first operating laser. However, "his initial paper, which would have made his findings known in a more traditional fashion, was rejected for publication by the editors of Physical Review Letters - this to their everlasting chagrin." (For historic accounts including this quotation and a comprehensive discussion of the principles and application of lasers and holography, see Sects. 14.2 and 14.3 in Hecht, 1987 ; see also Chapter 5, this volume.) Shortly thereafter, two types of applications of lasers were sought in microscopy. One took advantage of the high degree of monochromaticity and the attendant long coherence length. Coherence length is the distance over which the laser waves could be shifted in path and still remain coherent enough to display clear interference phenomena (note that, in fact, this reflects a very high degree of temporal coherence). These characteristics made the laser an ideal source for holography (Leith and Upatnieks, 1963, 1964). To explore the use of holography with the microscope, Ellis (1966) introduced a conventional light microscope into one of two beams split from a laser. When this beam was combined with the other beam passing outside of the microscope, the two beams could be made to interfere in a plane above the ocular. The closely spaced interference fringes were recorded on very fine-grained photographic film to produce the hologram. What Ellis found was that the coherence length of the laser beam was so long that the hologram constructed as described above could be viewed not only to reconstruct an image of the specimen being magnified by the microscope, but also to reconstruct images of the inside of the microscope. Indeed, in the hologram one could see the whole optical train and interior of the microscope, staIting with the substage condenser assembly, the specimen, the objective lens and its back aperture, the interior of the body tube up to the ocular, and even the light shield placed above it! This made it possible for Ellis to view the hologram through appropriately positioned stops, phase plates, etc., and to generate contrast from the specimen in imaging modes such as darkfield or oblique illumination, phase contrast, etc ., after the hologram itself had been recorded. In other words, the state of the specimen at a given point of time could be reconstructed and viewed after the fact in contrast modes different from the one present when the hologram was recorded. In principle, holomicrography presents many intriguing possibilities including 3D imaging. But the very virtue of the long coherence length of the laser beam means that the hologram also registers all the defects and dirt in the microscope. Without laser illumination, the optical noise produced by these defects would be far out of focus . With a laser illuminating the whole field of view of the microscope, the interference fringes from these oUl-of-focws defects intrude into the holographic image of the specimen where they are prominently superimposed. Because of this problem, holomicrography has so far not been widely used. [However, see
8
Chapter 1 • S. Inoue
Sharnoff et al. (1986), who have figured out how to obtain holomicrograms that display only the changes taking place in the specimen (contracting muscle striations) over an interval of time and thus eliminate the fixed-patterned optical noise.]
Laser Illumination Another practical application of lasers in microscopy is its use as an intense, monochromatic light source. Lasers can produce light beams with a very high degree of monochromaticity and polarization, implying a high degree of coherence. Some lasers also generate beams with very high intensity. Thus, an appropriate laser could serve as a valuable light source in those modes of microscopy where monochromaticity, high intensity, and a high degree of coherence and polarization are important. To use the laser as an effective light source for microscopy, three conditions must be satisfied: • Both the microscope's field of view and the condenser aperture must appropriately be filled. • The coherence length of the laser beam (i.e., the temporal coherence) must be reduced to eliminate interference from outof-focus defects. • The coherence at the image plane must be reduced to eliminate laser "speckle" and to maximize image resolution. In fact, these three conditions are not totally independent, but they do specify the conditions that must be met. One of the following five approaches can be used to fulfill these conditions (see also Chapter 6, this volume).
Spinning-Disk Scrambler The laser beam, expanded to fill the desired field, is passed through a spinning ground-glass diffuser placed in front of the beam expander lens (Hard et al., 1977). The ground glass diffuses the light so that the condenser aperture is automatically filled. However, if the ground glass were not moving, small regions of its irregular surface would act as coherent scatterers and the image field would still be filled with laser speckle. Spinning or vibrating the ground glass reduces the temporal coherence of each of the coherent scattering points to a period shorter than the integration time of the image sensor. Thus, when averaged over the period of the motion, the field also becomes uniformly illuminated. This approach, while simple to understand, can result in considerable light loss at the diffuser. Also, inhomogeneity of the diffuser's texture can give rise to concentric rings of varying brightnesses which traverse the field.
Oscillating-Fiber Scrambler The laser beam is focused onto the entrance end of a singlestranded multi-mode optical fiber whose output end lies at the focal point of a beam-expanding lens. This lens projects an enlarged image of the fiber tip to fill the condenser aperture. The fiber, which is fixed at both ends, is vibrated at some point along its length. The field and aperture are then uniformly filled with incoherent light with little loss of intensity (Ellis, 1979). If the fiber were not vibrated, the simple fact that the light beam is transmitted through the fiber could make the laser beam highly multi-modal. That would reduce the lateral coherence of the beam at the aperture plane, but the image would still be filled with speckle. Vibration that reduces the temporal coherence of the beam below the integration time of the image sensor integrates out the speckle without loss of light (see also Chapter 6, this volume).
Multi-Length Fiber Scrambler None of the mechanical scramblers mentioned above can be used where speckles have to be removed within extremely brief time periods. For example, in a centrifuge polarizing microscope, the laser output must be made spatially incoherent within the few nanoseconds required to freeze the image of the specimen flying through the field of the objective lens at speeds up to 100 m/s (Inoue et al., 200tb). Our solution for reducing the coherence of the laser pulse was to introduce a fiber bundle made up of up to 100 fibers of multiple lengths between two multi-mode single-fiber scramblers. The first multi-mode fiber introduced some phase randomizing effects, while the multi-length fibers provided a fiber bundle output whose phase varied depending on the length of the fiber. However, the intensities of the fiber output varied depending on their location in the bundle. The final multi-mode fiber made the non-uniform brightness of the bundle output homogeneous, so that the microscope condenser received uniform illumination. Thus, without using any mechanically moving parts, the phase of the monochromatic laser beam is randomized, and speckles are eliminated from the field image, while the whole condenser aperture is filled uniformly.
Field Scanning The field is scanned by a minute focused spot (the diffraction image) of a single-mode laser beam that has been expanded to fill the condenser aperture (as in a laser-scanning confocal microscope). Thus, the specimen is scanned point by point, and the signal light reflected, transmitted, or emitted by the specimen is collected and focused by the objective lens. This imaging mode avoids the generation of speckle from laser-illuminated specimens because speckle arises from the interference between the coherent light waves scattered from different parts of a specimen. (This optical setup is less effective at removing speckle when a smooth reflecting surface is presented slightly away from the plane of focus.) This fourth approach leads to field-scanning microscopy. A focused spot of laser light can be made to scan the field as in a flying-spot microscope, or the specimen can be moved and scanned through a fixed focus point. Alternatively, an exit pinhole and beam scanners can be added to generate a laser-scanning confocal microscope.
Aperture Scanning The minute diffraction image of a single-mode laser is focused by a beam expander onto an off-axis point on the condenser aperture. The small spot is scanned (made to precess) over the condenser aperture in such a way that the field is uniformly illuminated. At any instant of time, the specimen is illuminated by a tilted collimated beam of light emerging from the condenser and originating from the illuminated aperture point. Selected regions of the aperture are filled in rapid succession by scanning the spot, so that the whole field is illuminated by collimated, coherent beams at successively changing azimuth angles. The rapid scanning of the source reduces the temporal coherence of illumination at the object plane to less than the response time of the image detector. Nevertheless, the lateral coherence is maintained for each instantaneous beam that illuminates the specimen (Ellis, 1988). Ellis has argued the theoretical advantage provided by this fifth approach and has demonstrated its practical attractiveness. With aperture scanning, one gains new degrees of freedom for optical image processing because the aperture function (which controls the image transfer function of the microscope) can be regulated dynamically for each point of the aperture. The image resolution
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
and the shallow depth of field that can be achieved with aperturescanning phase-contrast microscopy is most impressive (see, e.g., Inoue and Spring, 1997, Fig. 2-47).
Laser-Illuminated Confocal Microscopes During the early 1970s, Egger and co-workers at Yale University developed a laser-illuminated confocal microscope in which the objective lens was oscillated in order to scan the beam over the specimen. Davidovits and Egger obtained a U.S. patent on this microscope (1972; see review by Egger, 1989). A few years later, Sheppard and Choudhury (1977) provided a thorough theoretical analysis on various modes of confocal and laser-scanning microscopy. The following year, Sheppard et al. (1978) and Wilson et al. (1980) described an epi-illuminating confocal microscope of the stage-scanning type, equipped with a laser source and a photomultiplier tube (PMT) as the detector, using a novel specimen holder. The specimen holder, supported on four taut steel wires running parallel to the optical axis, allowed precise z-axis positioning as well as fairly rapid voice-coil-actuated scanning of the specimen in the xy-plane. Using this instrument, Sheppard et al. demonstrated the value of the confocal system particularly for examining integrated circuit chips. With stagescanning confocal imaging, optical sections and profile images could be displayed on a slow-scan monitor over areas very much larger than can be contained within the field of view of any given objective lens by conventional microscopy. These authors capitalized on the fact that the confocal signal falls off extremely sharply with depth, and the image is therefore completely dark for regions of the specimen that are not near the confocal focus plane. For example, with a tilted integrated circuit chip, only the portion of the surface within the shallow depth of field (at any selected z-value) could be displayed, as a strip-shaped region elongated parallel to the chip's axis of tilt. Other areas of the image were dark and devoid of structure. Conversely, by combining all the xy-scan images made during a slow z-scan, they could produce a final "extended focus" image of the whole tilted surface, which demonstrated maximum spatial resolution on all features throughout the focus range (Wilson and Sheppard, 1984; Wilson, 1985) (Chapter 22, this volume). This could be done even when the specimen surface was not a single tilted plane but was wavy or consisted of complex surfaces. In their monograph Scanning Optical Microscopy, Wilson and Sheppard (1984) show shallow optical sections of insect antennae shining on a dark background. They also show stereo-pair images of the same object consisting of two "extended focus" images made by focusing along two focal axes that were tilted by several degrees relative to the optical axis. Extended-focus images demonstrate that the confocal system can either decrease or increase the effective depth of field without loss of resolution. As described in the final section of this article, the lateral resolution that is practically attainable can be improved by using confocal optics. In addition, the removal of the extraneous light contributed by out-of-focus objects dramatically improves the contrast and gives rise to a brilliantly sharp image. Sheppard et al. also managed to display different regions on the surface of an integrated circuit chip with varying intensity or pseudocolor corresponding to the height of the region. This is possible because the amount of light reflected by an (untilted) step on the surface of the chip and passing the second pinhole varies with the distance of the reflecting surface from the focal plane. The authors also showed that, by processing the photoelectric signal electronically, the edges of the steps alone could be outlined or the
9
gradient of the steps could be displayed in a DIC-like image (Hamilton and Wilson, 1984). [For the basics of digital image processing, see Castleman (1979), Baxes (1984), Gonzales and Wintz (1987), Chapter 12 in Inoue and Spring (1997), and Chapter 14, this volume.] The integrated circuit chip could also be displayed with contrast reflecting the status of the local circuit elements, for example, reflecting its temperature or the amount of photo-induced current flowing through the circuit, superimposed on the confocal image of the chip made with reflected light (Wilson and Sheppard, 1984). In addition to the Oxford group, the brothers Cremer and Cremer (1978) of Heidelberg designed a specimen-scanning laserilluminated confocal microscope. This epi-fluorescence system was equipped with (1) a circular exit pinhole, in front of the first PMT, whose diameter was equal to the principal maximum of the diffraction pattern; and (2) an annular aperture, in front of a second PMT, whose opening corresponded to the first subsidiary maximum of the diffraction pattern. The output of the two PMTs was used to provide autofocus as well as displays of surface contour and fluorescent intensity distribution. In the 1978 article, the Cremers also discussed the possibility of laser spot illumination using a "41t-point hologram" that could, at least in principle, provide long working distance relative to the small spot size that could be produced.
CONFOCAL LASER-SCANNING MICROSCOPE In addition to those already mentioned, the pioneering work of the Oxford electrical engineering group was followed in several European laboratories by Brakenhoff et al. (1979, 1985), Wijnaendts van Resandt et al. (1985), and Carlsson et al. (1985). These investigators respectively developed the stage-scanning confocal microscope further, verified the theory of confocal imaging, and expanded its application into cell biology. I shall defer further discussions on these important contributions to authors of other chapters in this volume. In the meantime, video microscopy and digital image processing were also advancing at a rapid rate. These circumstances culminated in the development of the confocal laser-scanning microscope (CLSM, Figs. lA, l.5; Aslund et al., 1983, 1987) and publication of its biological application by Carlsson et al. (1985), Amos et al. (1987), and White et al. (1987). The publications were followed shortly by introduction of laserscanning confocal microscopes to the market by Sarastro, BioRad, Olympus, Zeiss, and Leitz. It was White, Amos, and Fordham of the Cambridge group that first enraptured the world's biological community with their exquisite and convincing illustrations of the power of the CLSM. Here at last was a microscope that could generate clear, thin optical sectioned images, totally free of out-offocus fluorescence, from whole embryos or cells and at NAs as high as lA. Not only could one obtain such remarkable opticalsectioned fluorescence images in a matter of seconds, but x-z sections (providing views at right angles to the normal direction of observation) could also be captured and rapidly displayed on the monitor. A series of optical sections (stored in the memory of the built-in or add-on digital image processor) could be converted into 3D images or displayed as stereo pairs. The confocal fluorescent optical sections could also be displayed side by side with nonconfocal brightfield or phase-contrast images, acquired concurrently using the transmitted portion of the scanning laser beam. These images could also be displayed superimposed on top of each other, for example, with each image coded in different pseudocolor, but unlike similar image pairs produced by conventional
10
Chapter 1 • S. Inoue FIGURE 1.4. Schematic of laser-scanning confocal microscope. (From Aslund et al., 1987.)
Work stotion Operator Eyepiece TV-Monitor Focus motor
Tope
microscopes, the two images were in exact register and showed no parallax as each was generated by the same scanning spot. Most of the laser-scanning systems discussed in this section employed epi-iUumination using some form of mechanical scanning devices. They could not readily be applied to confocal imaging of transmitted light, for example, for high-extinction polarization or DIC microscopy. Nevertheless, Goldstein et al. (1990) developed a system using an Image Dissector Tube which, in principle, should be able to provide confocal imaging in the trans-illumination mode. Such an approach may eventually lead to workable transmission laser-scanning confocal microscopes with multiple contrast modes.
TWO- AND MULTI-PHOTON MICROSCOPY As noted, conventional point-scanning confocal microscopes dramatically reduce the contribution of fluorescence from out-offocus regions of the specimen. Nevertheless, regions of the specimen above and below the focal plane are exposed to the
Fi ...........
Del.
®
Scanner x
~::::::::r~
Objective
@) FIGURE 1.5. Depth discrimination in a laser-scanning confocal fluorescent microscope. Compare with Figure 1.2. (Courtesy of Dr. H. Kapitza, Carl Zeiss, Oberkochen.)
full cones of intense excitation light, converging and diverging from the illuminated spot. Thus, with conventional confocal microscopy, biological specimens tend to suffer from photoninduced damage and rapid bleaching of fluorescence, while the fraction of the short wavelength excitation beam that reaches the focal plane is reduced by absorption in the intervening material. Many of these shortcomings are circumvented by two- and multiphoton microscopy. By focusing a pulse of very intense laser beam with twice the wavelength (half the frequency) of the standard short wavelength excitation beam, and within a period shorter than the fluorescence decay time of the fluorophore, the coherently interfering photons can excite molecules at half the wavelength of the long wavelength laser, and do so selectively in the focused spot. In other words, the output of an intense near-infrared (IR) laser induces fluorescence in a blue or UV excitable fluorophore at, and only at, the focused spot where the coherent electromagnetic field strength is so high (within the required brief period) that it acts nonlinearly to excite the chromophores at twice the frequency of the IR field. The fluorophores in the cone of the illuminating light above and below focus do not experience the two-photon effect and, therefore, are not excited or damaged. Additionally, in contrast to conventional confocal microscopy, two-photon laser scanning systems do not require an exit pinhole or an image-forming objective lens. This is because the minute two-photon excited fluorescent spot is totally isolated in space and is free of "parasitic" fluorescence in the xyplane as well as along the z-axis. Therefore, the fluorescence emission needs only to be collected by an efficient photodetector as the excitation spot is being scanned (see Chapter 28, this volume). Thus, compared to conventional confocal microscopy, twophoton microscopy permits confocal imaging of planes much deeper in the tissue, and with considerably higher light-gathering efficiency, as well as with less fluorescence bleaching and specimen damage outside of the focal plane. Denk et al. (1990) and Squirrel et al. (1999) have made extended time-lapse recordings of dividing tissue-cultured cells and mammalian embryos in twophoton microscopy. In another interesting and ingenious application of two-photon microscopy, the fluorescence excitation volume (point-spread function) has been reduced considerably below that defined by wave optics by use of two partially overlapping excitation volumes. The first excites fluorescence in the standard two-photon volume, while the second volume, concurrently generated by a somewhat longer wavelength, quenches the fluorescence (by stim-
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
ulated emission depletion) in the zone where the two volumes overlap each other. A phase plate in the depletion beam path has the effect that this beam is almost the inverse of the Airy disk and has a null at the focus. Thus, the volume of region actually fluorescing is carved smaller than the standard two-photon excitation volume, and, in fact, Hell and Wichmann (1994) report having reduced the height of the point-spread function (PSF) by as much as a factor of five (see also Klar et al., 2000, Chapter 31, this volume).
IS LASER-SCANNING CONFOCAL MICROSCOPY A CURE -ALL? With the impressively thin and clean optical sections that are obtainable, and the x-z sections and stereoscopic images that can neatly be displayed or reconstructed, one can be tempted to treat the CLSM as a cure-all. One may even think of the instrument as the single microscope that should be used for all modern cell biology or embryology. How valid is such a statement and what, in fact, are the limitations of the current instruments beyond their high costs? The fundamental limits of confocal imaging will be covered in the next chapter. Here I will comment on three topics: the speed of image or data acquisition, comparison with the depth of field in phase-dependent imaging, and some optical and mechanical factors affecting confocal microscopy.
Speed of Image or Data Acquisition Several factors affect the time needed to acquire a usable image with a confocal microscope. These include (I) the type of confocal system used; (2) the optical magnification and numerical aperture of the system; (3) the desired area covered; (4) required quality of the image (e.g., lateral and axial resolution, levels of image gray scale, degree of freedom from graininess); and (5) the amount of light reaching the sensor. Here we will survey a few general points relating to the choice of instruments, specifically as applied to biology. Among the different confocal systems, the stage-scanning type requires the longest time (-lOs) to acquire a single image because the specimen support has to be translated (vibrated) very precisely. Biological specimens are often bathed in a liquid medium, and for these, any movement presents a problem. Even if the specimen chamber is completely sealed and the gas phase excluded to minimize the inertial effects of stage scanning, specimen motion still can occur during stage scanning. The alternative lens-scanning system can encounter worse problems when oil-immersion lenses are used. Very often structures in biological specimens are moving or changing dynamically at rates incompatible with very slow scan rates. Thus, despite the many virtues of the stage-scanning system recognized by Minsky (1957) and by Wilson and Sheppard (1984), there is little chance that the stage-scanning microscope will be widely used in biology. An exception might be for large-area 3D scanning of fixed and permanently mounted specimens. Such specimens require, or can take advantage of, those virtues of the stagescanning system that cannot be duplicated by other confocal designs. In the Petnin-type TSM or the Kino-type confocal microscope, the disk can be spun rapidly enough to provide images at video rate (30 frames/s). When speed of image acquisition is of paramount importance, as in the study of moving cells, living cells at high magnification, or microtubules growing in vitro, the type of
11
speed provided by the Nipkow disk system may be indispensable. For example, at the -1 O,OOOx magnification needed for clear visualization, the Brownian motion of microtubules (even those many micrometers long) is so great that an image acquisition time of >0.1 s blurs the image beyond use. As discussed earlier, the downside of the classical Nipkow disk-type system is that the efficiency of light transmission is low, light reflected by the spinning disk reduces image contrast, and the image may suffer from intrusive scan lines. Also, observation is usually by direct viewing through the ocular, or via some photographic or video imaging device, rather than using a PMT. While video imaging does have its own advantages, video sensors other than cooled charge-coupled devices (CCDs) and special returnbeam-type pickup tubes operate over a limited dynamic range. Conventional video pickup tubes seldom respond linearly over a range of > 100: 1 (more commonly somewhat less; see Inoue and Spring, 1997), and they have relatively high measurement noise. By contrast, a PMT can have a dynamic range of ;?:106 • When exceedingly weak signals need to be detected from among strong signals, or when image photometry demands dynamic range and precision beyond those attainable with standard video cameras, an imaging system using a cooled CCD or a PMT detector may be required. Modern stage-scanning- and laser-scanning-type confocal microscopes use such detectors (Chapter 12, this volume). Nevertheless, for some applications improved versions of the Nipkow-disk-type confocal instruments may provide optical sections with better signal and image quality than with CLSMs as discussed below under "Yokogawa Disk-Scanning Confocal System." The frame-scanning rate of the CLSM falls somewhere between that of the stage- and tandem-scanning types, normally about 1 to 2 s/frames. This rate is the minimum time required by the mirror galvanometers (that are used to scan the illuminating and return beams) to produce an image of, say, 512 x 768 picture elements. This limitation in scanning speed relates to the absolute time required to scan along the fastest axis (usually the x, or horizontal, scan). The scanning speed cannot be increased without affecting image resolution or confocal discrimination (Chapters 3, 21, and 25, this volume). The x-scanning speed can be increased by using a resonance galvanometer, a spinning mirror, or an acousto-optical modulator instead of the mirror galvanometers (Chapters 3, 9, and 29, this volume). However, doing so may reduce both scan flexibility (i.e., no optical "zoom" magnification) and inefficient use of the duty cycle. Furthermore, in a scanning confocal system used for fluorescence microscopy, one cannot use the same acousto-optical device (or other diffraction-based electro-optical modulator) to both scan the exciting beam and de-scan the emitted beam because the modulator would deviate the two beams by different amounts based on their A. Of even greater importance, the image captured by a CLSM in a single, 1- to 2-s scan time is commonly too noisy because the image-forming signal is simply not made up of enough photons. The image generally must be integrated electronically over several frame times to reduce the noise, just as when one is using a highsensitivity video camera. Thus, with a CLSM, it often requires several, or many, seconds to acquire a well-resolved, high-quality fluorescence image. If, in an attempt to reduce the number of frames that must be integrated, one tries to increase the signal reaching the PMT by raising the source brightness, by opening up the exit pinhole, or by increasing the concentration of fluorochrome, each alteration introduces new problems of its own. In fact, in CLSMs used for fluorescence imaging, if anything, one wants to reduce the light
12
Chapter 1 • S. Inoue
reaching the specimen in order to avoid saturation of the fluorophores, significant bleaching, and other excitation-induced damage. There is almost an indeterminacy principle operating here: One simply cannot simultaneously achieve high temporal resolution, high spatial resolution, large pixel numbers, and a wide gray scale simultaneously. This speed limitation must be seen as a disadvantage of the CLSM. As already discussed, two-photon confocal fluorescence microscopy (Chapter 28, this volume) is a promising new approach that may reduce the effect of some of these limitations in addition to providing excellent lateral and axial resolution. However, because the time between pulses is long (10-12 s) compared to the fluorescence time of organic dyes, it only produces signal 10% to 20% of the time. This low-duty cycle exacerbates the data rate limit. While the sampling rate for obtaining whole images with the CLSM is limited, this does not imply that the temporal resolution of the detector system is inherently low. For example, one can measure relatively high-speed events with the CLSM, if one decides to sacrifice pixel numbers by reducing the size of the scanned area or even by using a single, or a few, line scan(s). In addition to the high temporal resolution, the bleaching of diffusible fluorochromes and photodynamic damage to the cell are reported to be significantly reduced when the scan is restricted to a single line (Chapter 19, this volume). Another alternative for gaining speed is to use a slit instead of a pinhole for confocal scanning. This approach, although somewhat less effective than confocal imaging with small round pinholes, is surprisingly effective in suppressing the contribution of out-of-focus features. Several manufacturers have produced laser-illuminated, slit-scanning confocal microscopes that provide video-rate or direct-view imaging systems that are quite easy to operate, at a fraction of the price of the normal CLSM. However, the rapid bleaching of fluorescent dyes encountered with the slitscanning system has been a disappointment for those hoping to gain confocal scanning speed for studies on living cells.
Yokogawa Disk-Scanning Confocal System A new confocal disk-scanning unit (CSU-lO and CSU-21) designed by Yokogawa Electric Corporation provides video-rate and faster confocal imaging with several advantages while overcoming the two major factors that had limited earlier TSM systems. The new system uses two Nipkow-type disks located one above the other with precisely aligned perforations. In place of pinholes, the first disk contains some 10,000 micro lenses, each of which focuses the collimated laser beam onto a corresponding pinhole on the second disk. The micro lenses increase the throughput of excitation laser from a scant 1% to 2% of conventional Nipkow disks to nearly 50%. At the same time, a dichromatic filter cube is placed between the two disks, so that light reflected or scattered from the initial disk no longer contributes unwanted background to the fluorescence signal received by the detector (Fig. 1.6). These confocal scanning units can be attached to any upright or inverted research-grade light microscope. The 1000 or so pinholes that scan the specimen in parallel at any instant of time are arranged in a unique geometrical pattern. The unique pattern reduces image streaking (found with conventional Nipkow disks) and provides uniform illumination of the whole field of view (Inoue and Inoue, 2002). With any multiple-pinhole- (or slit-) scanning system, some light originating from outside the focal plane is transmitted through "neighboring" pinholes, so that focal discrimination is not as effec-
LASER
!
ROTATION MICROLENS DISK
~
--......-
-
MICROLENS
~+--==
--~-
:.--
PINHOLE DISK (NIPKOW DISK)
t SAMPLE
FIGURE 1.6. Schematic of optics in the CSU-JO. The expanded and collimated laser beam illuminates the active portion of the upper Nipkow disk containing some 20,000 microlenses. Each micro lens focuses the laser beam (through the dichromatic filter cube) onto its corresponding pinhole, thus significantly raising the fraction of the illuminating beam that is transmitted by the main Nipkow disk containing the pinhole array. From the pinholes, the beams progress down to fill the aperture of the objective lens. The objective lens generates a reduced image of the pinholes in the focal plane of the specimen. Fluorescence given off by the illuminated points in the specimen is captured by the objective lens and focused back onto the Nipkow disk containing the pinhole array. Each pinhole now acts as its own confocal exit pinhole and eliminates fluorescence from out-of-focus regions, thus selectively transmitting fluorescence that originated from the specimen points illuminated by that particular pinhole. (However, for specimens with fluorescence distributed over large depths, some out-of-focus fluorescence can leak through adjacent pinholes in multiple pinhole systems such as the CSU-lO; see text.) The rays transmitted by the exit pinholes are deflected by the dichromatic beam splitter. located between the two Nipkow disks, and proceed to the image plane. (Figure courtesy of Yokogawa Electric Corporation.)
tive as, or the confocal stringency does not match that of, a pointscanning confocal system. Nevertheless, the Yokogawa system provides very effective focal discrimination and capability for providing striking optical sections in real time, either for direct observation through the eyepiece or captured on a video or photographic camera or a CCD. The residual out-of-focus contribution can be rapidly and effectively reduced by "unsharp masking" or "neighborhood deconvolution" digital processing. (Several examples of the dynamic cellular changes captured with the CSU-lO, as well as the effectiveness of postprocessing, are illustrated in Inoue and Inoue, 2002. See also Chapter 10, this volume.) In addition to the effective, real-time and faster-than-video-rate confocal fluorescence imaging in real color (which can be viewed superimposed with the brightfield or DIC image of the specimen), several observers have been impressed by the significantly slower fluorescence bleaching rate and much longer survival time for living cells observed with the CSU-lO (coupled to low-noise CCD or video cameras) compared to imaging of the same objects with point-scanning confocal systems. For example, CSU-lO imaging was found indispensable for capturing the dynamic growth, motion, and gliding of GFP-expressing microtubules in yeast cells
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
as well as speckle images of tubulin flux in pTk-l tissue cells and in the thick spindles undergoing mitosis in Xenopus egg extracts (Waterman-Storer and Salmon, 1997 ; Grego et al., 2001; Tran et al., 200 I; Maddox et al., 2002 ). The reasons for low-fluorescence bleaching and extended cell survival are discussed in Inoue and Inoue (2002 and in Chapter 38, this volume.) The advantage of the real-time, direct-view confocal system extends beyond capturing sharp images of moving or dynamically changing objects, whose images would be blurred or distorted by the slow frame-capture rate of conventional CLSMs. For example, with a CLSM it is difficult to visualize, or even to find , minute fluorescent objects that are sparsely distributed in three dimensions. With an image intensifier CCO camera coupled to a direct-view system , the signal from such sparsely distributed objects is readily found in real time as one focuses through the specimen.
Depth of Field in Phase-Dependent Imaging The z-axis resolution measured in epi-fluorescence imaging with a confocal laser scanning microscope is reported to be 1.5 11m with an NA 0.75 objective lens (Cox and Sheppard, 1993) and 0.48 11m with an NA 1.3 objective lens (Hell et al. , 1993) at a wavelength of 514 nm . Kino reports a depth of field of 0.35 11m for NA 1.4 confocal optics, when imaging point-like reflecting objects. These numbers are in good agreement with Eqs. 2 and 3, and the height of the 3D diffraction pattern of a point object discussed earlier. In addition, Stephan Hell, as described above, has achieved even shallower field depths by superimposing two 3D diffraction spots of differing wavelengths in stimulated depletion pointscanning confocal microscopy. How do these shallow depth of fields attainable with a confocal microscope compare with those obtainable in the absence of confocal imaging? While I could come up with no hard numbers for fluorescence microscopy without confocal imaging (except where 3D deconvolution is employed, see Chapters 23, 24, and 25 , this volume), it is well known that the fluorescence from out-offocus objects substantially blurs the in-focus image. On the other hand, for contrast generated by phase-dependent methods such as phase-contrast, OIC, and polarized-light microscopy, Gordon Ellis and I have obtained data that show remarkably thin optical sections in the absence of confocal imaging. Thus, using a lOOx NA 1.4 Nikon PlanApo objective lens, combined with an NA 1.35 rectified condenser whose full aperture was uniformly illuminated through a light scrambler with 546-nm light from a lOO-W high-pressure Hg arc source (as described in Ellis, 1985, and in Inoue, 1986, Appendix 3), 1 obtained depth of fi elds of ca. 0.2, 0.25, and 0 . 15I1m, respectively, for phasecontrast, rectified DIC, and rectified polarized-light microscopy. These values were obtained by examining video images of surface ridges on a tilted portion of a human buccal epithelial cell. The video signal was contrast enhanced digitally but without spatial filtration. The change in image detail that appeared with each 0.2-l1m shift of focus (brought about by incrementing a calibrated stepper motor) was inspected in the image and enlarged to -IO,OOOx on a high-resolution video monitor. As shown in Figure 1.7, the fine ridges on the cell surface are not contiguous in the succeeding images stepped 0 .211m apart in the polarizedlight and phase-contrast images, but they are just contiguous in the DIC images. From these observations, the depth of field in the rectified polarized-light image is estimated to be somewhat below, and the DIC image just above, the 0.2-l1m step height. [The phasecontrast images here should not be compared literally with the images in the two other contrast modes because the diameter of
13
the commercially available phase annulus was rather small, and out-of-focus regions intruded obtrusively into the image. With Ellis' aperture-scanning phase-contrast microscope, the illuminating rays, and the correspondingly minute phase absorber spot, scan the outermost rim of the objective lens aperture in synchrony. Therefore, essentially the full NA of the objective lens is available to transmit the waves diffracted by the specimen. Under these conditions, the z-axis resolution of the optical section in phasecontrast appears to be even higher than that of the two other contrast modes shown here (Ellis, 1988; Inoue, 1994).] For polarization microscopy of specimens with low retardances, the LC-Pol scope system devised by Olden bourg and Mei (1995; see also Oldenbourg, 1996) also provides effective optical sectioning. The LC-Pol scope generates an image (retardance map) whose pixel brightness is proportional to the retardance of the specimen at each pixel, independent of the specimen's azimuth orientation, while the algorithm used to compute the retardance map also reduces the polarization aberrations introduced by the optics (that otherwise degrade the image; Shribak et al., 2002). Thus, with the LC-Pol scope system, objective lenses with NA as high as 1.4 can be used at their full aperture to detect retardances as low as 0.03nm. The use of the high NA lenses at full aperture then provides the shallow depth of field (of less than I-11m thickness) as illustrated in Figure 1.8. In fact, the LC-Pol scope can individually resolve two flagellar axonemes that cross each other and are separated by no more than their diameter of about 0.211m (Oldenbourg et al., 1998). We do not yet quite understand why the depth of field of the non-confocal phase-dependent images should be so thin. It may well be that contrast generation in phase-dependent imaging involves partial coherence even at very high NAs, and that an effect similar to the one proposed elsewhere for half-wave masks (Inoue, 1989) is giving us increased lateral as well as axial resolution . Whatever the theoretical explanation turns out to be, our observations show that for phase-dependent imaging of relatively transparent objects, even in the absence of confocal optics, optical sections can be obtained (at video rate) that appear to be somewhat thinner than for fluorescence imaging in the presence of confocal optics. Moreover, they perform this function without requiring that energy be deposited in the specimen, i.e., without producing photodamage.
OTHER OPTICAL AND MECHANICAL FACTORS AFFECTING CONFOCAL MICROSCOPY
Lens Aberration With stage- or object-scanning confocal microscopes, we saw earlier that high NA lenses with simplified design and long working distances could be used because the confocal image points (source pinhole, illuminated specimen point, and detector pinhole) all lie exactly on the optical axis of the microscope. Thi s same principle is now used widely in the design of optical disk recorder/players. In contrast, with TSM and CLSM sharp images of the source "pinhole(s)" must be focused over a relatively large area away from the lens axis. In addition, the objective lens and the scanner must bring images of the illuminated spot(s) and the source pinhole(s) into exact register with the exit pinhole(s), and for fluorescence microscopy, do so at different wavelengths. Thus, for these systems to function efficiently, the microscope objective lens has to be exceptionally well corrected. The field must be flat over an
14
Chapter 1 • S. Inoue
FIGURE 1.7. Optical sections of surface ridges on an oral epithelial cell. These ultrathin optical sections were obtained without confocal imaging in phasecontrast (left), rectified ole (middle), and rectified polarized-light microscopy (right). The focus planes for the successive frames in each contrast mode were incremented O.2~m. Scale bar IOllm. (See text and original article for details. From Inoue, 1988.)
appreciable area, axial and off-axis aberrations must be corrected over the field used, and lateral and longitudinal chromatic aberrations must be well corrected for both the emission and illuminating wavelengths. As far as is possible, the aberrations should be corrected within the objective lens without the need to use a complimentary ocular. [For details of these subjects and design of modern lenses to overcome the aberrations, see Inoue and Oldenbourg (1994), Shimizu and Takenaka (1994), and Chapter 7, this volume.] Finally, the lens and other optical components must have good transmission over the needed wavelength range. These combined conditions place a strenuous requirement on the design of the objective lens. Fortunately, with the availability of modem glass stocks and high-speed computer-optimized
design, a series of excellent-quality, high-NA, PlanApo, and highUV-transmitting lenses have appeared from all four major microscope manufacturers (Leitz, Nikon, Olympus, Zeiss) during the past decade. Even with excellent lenses, however, the image loses its sharpness when one focuses into a transparent, live, or wet specimen by more than a few micrometers from the inside surface of the coverslip. The problem here is that oil-immersion microscope objectives are designed to be used under rather stringent optical conditions, namely homogeneous immersion of everything, including the specimen itself, in a medium of 11 = 1.52. When such a lens is used on live or wet specimens immersed in water or physiological saline solution, even with the coverslip properly oil con-
Foundations of Confocal Scanned Imaging in Light Microscopy • Chapter 1
15
FIGURE 1.B. Optical sections of meiosis-l metaphase spindle in live spermatocyte of a crane fly Nephrotoma sturalis observed with the Le-pol scope. In focus are: (A) the upper kinetochores (Ks) and their two K-fibers for the left bivalent chromosome; (B) upper Ks and fibers for the middle and right bivalent s; (e) lower Ks a nd fibers for the middle and right bivalcnts. The birefringence retardation of the K-fibers. made up of a dense bundle of microtubules , is ca. I nm greater than that due to the background array of spindle microtubules. The etfect of optical sectioning is more obvious in the mitochondrial threads (which are mue h thinner than the K-fibers and free of background birefringence) that surround the spindle. Imaged with 546-nm illumination in LC-Pol seope with Nikon 60>9 or University of Illinoi s at U rbana-Champaign, Departm ent of Physics, Laboratory for Fluorescence Dynamics, Urbana, Illino is M artin J. vandeVen • Department o f Cell Physiology, Microfluorimetry Sectio n, Biomedical Institute, H asselt Un iversity and Trans Nation al University Lim bur g, and Institute for M ateri al s Research IMO/IMOMEC, Di epenbeek, Belgi um
80
Handbook of Biological Confoca l M icroscopy , Th ird Edition , edit ed by Jam es B . Pawley, Springer Science-B usin ess Media, LLC , New York . 2006.
Laser Sources for Confocal Microscopy • Chapter 5
concentration of 1O-5M is reasonable, the optical density (OD) of a 111m path length is "" 10-4 • The number of photons absorbed is then photons absorbed = (flux per pixel) x (OD) 2 x 105/(s * pixel * mW) at SOOnm.
=
Assuming a quantum yield of 0.8 and a collection efficiency of 10%, the detector receives photons at the detector = 1.6 x 10 4 (s * pixel * mW of incident light).
1
Given the quantum efficiency for a good detector (10% at 500nm), the final detected photon flux should be about flux detected
= 1600 photons/(s * pixel * mW of light).
This flux should be compared with the dark noise of most detectors, which can vary between 10 and 100 equivalent photons/(s * pixel). In our estimation, the only quantities that can vary over a wide range are the power of the laser and the effective concentration of the probe. Lasers can have up to watts of power and the concentration of the probe can be higher than we have assumed. The efficiency of detection is usually smaller than we estimate and the noise can be larger. The purpose of our calculation is to give a rough idea of the kind of power that a laser must furnish to be usable for fluorescence detection in confocal laser scanning microscopy (CLSM). Tsien and Waggoner (Chapter 16, this volume) find an optimal power with the best signal-to-noise ratio (SIN) with respect to autofluorescence and elastic and inelastic scattering of 761lW at 488nm and 590IlW, as long as triplet formation is neglected. Therefore, a laser power of I to 2 m W spread over 106 pixels at the specimen position should be more than sufficient for most applications. Effectively, 10 to 10011W is common in confocal. Assuming a 10% optical path efficiency a laserhead output power of >-1 m W suffices. There are two different types of saturation effects. One is related to the number of molecules that can absorb light in a given area for a certain incident flux. In a given pixel, assuming a volume of 111m3, the volume is 10- 15 L. At a molar concentration of 10-5 , we should have approximately 6000 molecules/pixel. Since the number of photons absorbed per milliwatt of incident light is about 2.S x 1O+5/s on a single pixel in widefield, each molecule is excited about 40 times per second. From the photophysical point of view, the decay of fluorescein (and in general any singlet single state decay) is very fast (4 x 10-9 S), so that the ground state should be repopulated very rapidly. However, in the confocal microscope for a pixel dwell time of about IllS, the 40 x 4 ns = 160 ns dead time represents 16% of the pixel period. There are many possible photochemical processes that are either irreversible or have a cycle time of several milliseconds to seconds. In this latter case, even if the quantum yield for these effects is very low (below 0.001), and the exposure time is on the order of seconds, molecules lost to the long-lived state will severely limit the overall peak excitation intensity that can be used before the output loses its linear relationship with the input. For quantitative microscopy this is the most important limitation. Hess and Webb (2002) found that their FCS data implied a nonGaussian three-dimensional (3D) volume and distortion of the calibration of the excitation volume at a power level of 10 to 100 IlW at 488 nm, for one photon Rhodamine Green excitation and 5 to 10m W at 980 nm, for the two-photon case (Rhodamine Green or Alexa 488, Molecular Probes). Having discussed the power requirements, we continue with a concise description of the basic elements of a laser, its principle
81
of operation, and other important practical aspects, such as heat removal and mechanical and optical stability. In general, confocal microscopes work best at
,.,
(3
~ ;.
e!.
,.,0-
:::I
o
rl
.,0-
~ 'I>
~
o '" c
~
'I>
I»
r-
Wave-length (nm)
Life & Power Cons.
(h) & (W)
CW Av. Pwr (mW)
CW Av. Pwr (mW)
457.9 488 514.5 MLVIS 454.5514.5 458 488 514
>5,000h
:::l Co
~
~
-
:::l Co
~
:::l
S
!"" C'l
""
ib
"1:1
.
~
:r
!"l
t.l
>50,000
Melles Griot & Green HeNe Yellow HeNe Orange HeNe Lasos Laser! & Red HeNe
0.7
0.88 0.75 0.88
1-2
10,000
>5000 >5000
Twice that of argon ion
>20k
>IOk
5
10,000
Continuous Wave DPSS Lasers
CoboltAB & Blues 10
(n
I&>
(n
...;;-
III "tl
::r
("1
'
'"
a
;:;'
~
!!!.
0I">
::s
c
("1
~
'"
o c rlno
IJl
l"'"
~ ~
Hamamatsu' & PLP-IO
Becker-Hickl 6 & BDH series
Pulsed diode lasers
LambdaPhysik & LPXPro TuiLaser AG & Excistar'
Pulsed excimer lasers
SpectraPhysics 4 & VSL337
& Model
Manufactory
h
Lifetime J.lradiC
rms, %
90
0.8,2.4 0.9,0.5, 10 70,50, 20,20, 30,30, 30,30, 50pk power
3
~ ;:;0
~
0'
:::I
o
0' n
..
II>
I't>
~
c
o
~
I't>
II>
1,064/532
Optronics'
6
85 170
700 1,400
710-9S0 700--1,000
n.a & 0.12nJ O.22nJ n.a. & 0.4nJ 0.75nJ
150
1.5k
S00--900 1,064
1,600 & >80nJ
710-990
>400 >100-10k
SOO & >50
n.a & 1
S80
1,230-1,270
n.a. & 20nJ
800 center
>200 >650 ISO-250
±1
1Ak
240-320 350-500 800
I.3k SOO 1,400 I.3k 800
700--9S0
2-15%
>lk
lk
720-9S0
1,030
n.a & ~ ~T
VAL ' BA D TIM
LATED
E
MI SIO
FIGURE 5.11. Schematic energy diagram for a semiconductor laser. It is based on laser emission from a forward biased p-n junction when driven by a well-regulated drive current.
volume (Fig. 5.12), but to a stack of very thin layers forming multiple quantum wells (MQW), or planes of quantum wires or dots (superlattice). Due to the small confined volume a high radiative efficiency exists as well as a low lasing threshold. Most devices still operate in NIR but the trend is to develop diode lasers using wide bandgap materials that have an output below the red. Blue diode lasers based either on ZnSe or doubling 860nm light, emit at around 430 nm and are about to enter the commercial market. At higher power levels (Figueroa, 2002), direct frequencydoubling in the diode laser forms an alternative route to the blue region. For example, the DO (Direct Doubled Diode laser, Coherent) delivers IOmW at 430nm. When selecting a drive current source, it is important to select one with a low noise, good stability, and including a temperature controller on the diode laser head. Diode lasers can change their emission wavelength over a limited
- 500 )..lm
M
range (10-20 nm) by varying their drive current and junction temperature (Hodgson, 1994). The approximate tuning rate is about 0.1 nm fOe. Cooling brings the lasing wavelength down. Beam quality suffers from astigmatism that has to be optically corrected (Snyder and Cable, 1993). Edge emitters now provide high power, up to several watts CW in NIR. Semiconductor lasers are very appealing because they are small [Fig. 5.13(B)], highly efficient, easy to use, and relatively cheap. Integrated fiber-optic output is another feature available from many manufacturers [Fig. 5.l3(A)]. However, it should be stressed that this small package comes with important special requirements. The devices can be rapidly destroyed if both current transients or nanosecond current spikes at start-up and internal heating are not kept under control by the power supply electronics. Static discharges (SD) from a person or an ungrounded soldering iron, or the use of solder that is too hot or remains in contact for too long may instantly destroy the laser. A mechanical shunting device (Unger, 1994) may prevent SD damage during handling. Alternating current (AC) line filters are recommended (Hodgson, 1994). This market is strongly driven by the digital video disk (DVD) and audio compact disk (CD) industry where the goal is to increase information storage densities. Most, if not all, of the following examples are also wavelength stabilized by stacks of multi-layer coatings usually deposited at the HR side and producing bandpass filter reflectivity only for the lasing wavlength [Fig. 5.l4(B,C)]. Combining a proper OC coating with a fiber pigtail having an inscribed Bragg grating has the same stabilizing effect.
B
A
--
LA ER D IODE wrrH DBR FIBER 1'1'1 PIGTA IL
2
L
OXIDE ..._-_....
p-TYP lasing channel
n-TYPE ...............
~
T
·.. LLIPTI OUTPUT
I that for a spectral width of 200 nm one obtains a IOfs pulse at 800 nm.
DFB DI DE LA ER PUMPS
WDM CO PLER
SESAM M DE-L
KER
R E L' UL TRAFA T , MODE-La KED FIBER L S R
FIGURE 5.20. Schematic diagram for a femtosecond pulse mode-locked ultrafast diode laser. The emission of the fiber pigtailed distributed feedback (DFB) diode laser is coupled to the main laser gain medium fiber via a wavelength division multiplexer (WDM). Laser output exits from the laser resonator cavity in a similar way. The resonator consists of a lens-coupled high-reflecting mirror and a semicondcutor saturable absorber mirror (SESAM). The other end of the cavity contains the grating tuning element. For polarization, pulse width and dispersion control, extra elements are spliced in. Coupling the SESAM in a different way and lens coupling both gain medium fiber ends to the tuning grating would create a fiber-ring laser. In a flexible and compact way, diode-Iaserpumped amplifier stages, pulse-width compressors, and frequency-doubling stages can be added.
114
Chapter 5 • E. Gratton and M.J. vandeVen
New Optics, Ltd. , a company that has also developed modelocked, tunable Yb and Nd fiber lasers. Typical specifications are tuning from 1040 to 1120 nm, 3 ps pulse width and 30 to 100MHz repetition rate. With some further development, it may become a competitor for the Ti:Sa laser (Okhotnikov et al., 2003; Rusu et al., 2004). Pumped by a pigtailed diode laser, the mode-locked ultrafast Er fiber laser FemtoFiber Scientific (FFS) from Toptica Photonics AO (Table 5.3) produces pulses with a width of < 100 fs at 100 ± 10MHz, optional tunability over 1150 to 1400nm and SHO doubling to 575 to 700nm, 1550nm is standard with 200mW average and 18 kW peak power; the Femtolite family Er fiber laser and Yb-fiber based chirped pulsed amplifier FCPA JlJewel series from IMRA America Inc.
WAVElENGTH EXPANSION THROUGH NON - LINEAR TECHNIQUES The spectroscopist wishing to excite with laser light finds several gaps in the table of available wavelengths. Although dye lasers cover an extended portion of the spectrum , they require too much maintenance to be useful in standard confocal microscopy. Optically anisotropic crystals provide a good alternative towards tunable light as long as they are pumped with lasers having high (pulsed) peak power at a high repetition frequency.
Second and Higher Harmonic Generation: Second Harmonic Generation, Third Harmonic Generation, Fourth Harmonic Generation Label-Free Microscopy Non-linear optical (NLO) effects occur in certain classes of optically anisotropic crystals (Lin and Chen, 1987; Tebo, 1988). Focusing intense laser light into these crystals generates radiation at half the wavelength or double the frequency. This process, therefore, is known as "frequency doubling" or "second harmonic generation" (e.g., Huth and Kuizenga, 1987; Higgins, 1992). The intensity of the frequency-doubled light output is proportional to the square of the incoming light intensity. A Cr:forsterite laser running at 1230nm was used to create a very elegant second harmonic generation (SHO) and third harmonic generation (THO) system operating at 615 and 410 nm (respectively) with a pulse repetition rate of I 10 MHz and a pulse width of 140ps (Hogan, 2004; see also Chapter 40, this volume). When the incoming beam is very intense, even third, fourth , or fifth harmonics can be generated. However, the efficiency of harmonic generation decreases at higher orders. Most doubling crystals can be angle or temperature tuned. THO is very interesting for label-free monitoring of inhomogeneities at interfaces, for example, cell membranes (Moreau x et al., 2000). Yelin (2002) and Brocas and colleagues (2004) compare the THO performance of several systems: the Spectra Physics Tsunami-OPAL synchronously pumped OPO, which emits an 80 MHz 150 fs pulsetrain at around 1500 nm with an average power of 350mW; and the T-pulse laser (Amplitude Systemes) emitting 50MHz 200fs pulses at 1030nm with 1.1 W of average power. THO reaches 500 and 343 nm, respectively, so UV transmitting optics may be required. A not-sa-pleasant consequence of the doubling process is the doubling of the noise level. Pump lasers with excellent stability are a must. Higher peak powers and new, low-threshold materials require a careful reconsideration of design strategies (Beausoleil, 1992). CSK Optronics (Table 5.3) sells a line of compact doublers and triplers for Ti:Sa lasers. An important issue in this type of experiment is the amount
of optical damage done to living cells and tissue by these powerful femtosecond pulses (Chu et al. , 2003; and also see Chapter 38, this volume). THO peak intensities of 100 to 3000W/cm 2 illumination (i.e., nJ/pulse) are not damaging. Presently, it is also possible to combine less powerful tunable dye or semiconductor lasers to cover the wavelength range from 170 nm to 18 Jlm almost continuously.
Sum or Difference Mixing Another technique for generating different wavelengths from certain basic laser wavelengths is sum or difference mixing. When two laser beams of high and low intensity and of frequency WI and ~ , respectively, are simultaneously focused into a non-linear optical crystal, a sum signal is generated. The intensity of the sum signal, 1(0)[+0)2) is proportional to the product l(cu[) x l(ro2)' The higherintensity WI laser helps in generating enhanced UV output I (WI + ( 2) with an extended tuning range. In a similar way, difference mixing, I(WI~0)2) leads to a tunable IR laser. (For examples, see Dunning, 1978; Adhav, 1986; Herrmann and Wilhelmi, 1987; Kaiser, 1988; Marshall, 1994; Demtroder, 1996)
Optical Parametric Oscillators and Optical Parametric Amplifiers The highest repetition rate that can be obtained depends on pump energy storage. Potma and colleagues (1998) developed a cavitydumped optical parametric oscillators (OPO) with 13 nJ pulses that reached a repetition rate of 4MHz. These devices can generate a continuously tunable output by non-linear conversion of fixed wavelength high intensity CW or pulsed laser light (Butcher, 1994; Radunsky, 2002). OPOs are a reliable and easy-to-use means to cover spectral gaps left by dye or Ti:Sa lasers. Between the Ti:Sa fundamental, covering 700 to 1000nm, and the second harmonic, covering 350 to 500nm, a spectral gap also exists from 250 to 350 nm. This means that neither tryptophan, tyrosine, coumarine, or naphthalene nor the caged probes can be covered by a Ti:Sa. OPO efficiencies easily reach 40% to 60% (Anderson, 1993), and practical systems are available from many companies (Table 5.2). A continuously tunable output stretching from UV to IR is generated by non-linear conversion, that is, optical frequency mixing. Parametric frequency conversion creates two lower-energy photons (signal and idler) from each incoming photon. This process is the opposite of sum mixing. There are two methods to overcome the low efficiency of the parametric generation: Method I - Parametric amplification in a non-linear BBO crystal pumped by the second harmonic of the Ti :Sa followed by pulse compression. Parametric down-conversion is incompatible with long pixel integration times, and high pulse energies may possibly be damaging, and instrumentationally, it is quite involved. Method 2 - Cavity-dumping a synchronously pumped OPO can produce pulse repetition rates up to 100 MHz. This allows FUM imaging and reduces somewhat the average power delivered and cell damage. In addition, residual pump beams at 400 and 800nm are available. Synchronous pumping means that the cavity lengths and, therefore, the pulse round-trip times are equal in both pump source and OPO (Potma et al., 1998). Non-critical phase-matching (NCPM) in a crystal is preferred because it allows tight focu sing of the pump beam and long interaction lengths, resulting in low OPO thresholds. Efficiency can be further increased by using periodically polled (PP) waveguides, which do not diffract a tightly focused beam as does bulk material, while at the same time they allow long interaction lengths with a good mode profile. For example, green output from a Q-switched
Laser Sources for Confocal Microscopy • Chapter 5
diode-pumped, frequency-doubled Nd-YAG laser pumps a nonlinear crystal creating a broad tuning range and narrow bandwidth with 15 mW average power at around 670 nm. Typical oscillator and amplifier crystal combinations cover various spectral regions. Angle tuning a proper crystal will give a 405 to 2700nm tuning range with 45% total (signal + idler) efficiency (QWULasertechnik GmbH; Table 5.3). Continuum (Table 5.3) introduced the Sunlite OPO (Anderson, I 994a), which is currently the Sunlite EX OPO, with a very broad tuning range stretching from 222.5 to 1750 nm without a gap and providing line widths of 0.075 cm- I . It is pumped with 100 to 160m] pulses at 355nm from a Q-switched Nd: YAG. Such a system is a good replacement for pulsed dye lasers. As mentioned above, there is a trend to create easy-to-operate automated all-in-one-box solutions. Spectra-Physics provides the LBO-based femtosecond synchronously pumped OPAL. Near the peak of the OPO tuning curve, it generates 200 to 250mW and covers a wavelength range from 1100 to 2260 nm with pulse widths
130
Chapter & • A. Nolte et al.
11~-Y~--~---+------------~--------=
0.1-Jkci-~=
B
100 10
r
,~ ,)~~\}v~~J)\"j~'···-··j\-I\-···-··-·-·-,·",·,_J· .
-1- ..:.c.."'.-.-/+ ..:. -j ----
J,t' rI - _,: ,rJ'l OOW Xe II
)
.,
t (ozone free)
"'--~"--J
-
"
100W Hgl
300W Xe
0.1 200
300
800
FIGURE &.5. Wavelengths resolved irradiance of various light sources. (A) Lower line: 30W deuterium arc lamp; upper line: 7SW xenon arc lamp; dashed line: SOW halogen filament; dashed/dotted line: SOW mercury arc lamp. (B) Dashed line: 100 W xenon arc lamp; dashed/dotted drawn line: 100 W mercury arc lamp; dotted drawn line: 300W xenon arc lamp. (Spectra kindly provided by OriellNewport, Irvine, CA.)
source is almost 8x larger than the comparable area of the lOOW source, its brightness is only half as high.
Coherence A fourth important source parameter is its coherence. Although coherence is closely related to brightness, in that bright sources are likely to also be highly coherent, the term brightness is more used to describe the ability of a source to focus a large number of photons into a small area, whereas coherence is more of a measure of the ability of wave functions that describe these photons to interfere with each other, either at the focus plane (good) or between the reflections from every dust particle and imperfection in the optical system (bad). In coherent light, a large fraction of the wave functions passing a given point are in-phase with each other. Although light from a laser is extremely coherent, even here there are limitations. Laser light emerges not at a single wavelength but over a very narrow wavelength band. Consequently, even if all the photons started off in phase, after some distance (the coherence length) those with slightly longer wavelength will become out of phase with those of slightly shorter wavelength. What this means practically is that light scattered by a dust particle on one side of a piece of glass can only interfere with light scattered on the other side if the thickness of the glass is less than the coherence length. II Interference of this type is the source of
"laser speckle," the pattern of random interferences that converts what should be a uniform beam into a pattern of black and white blobs. Because in microscopy it is preferable that the contrast seen in the image represent only the interference that occurred within the specimen, light sources of low coherence are preferred. The process of light amplification by the stimulated emission of radiation that produces laser light (see Chapter 5, this volume) necessarily produces very coherent light. Although it is often assumed that coherent light can be produced in no other way, this is untrue. If one thinks of light as an oscillating electro-magnetic field that propagates in one direction, and chooses to measure this field at a particular instant and in a location of space that is smaller in dimensions than its wavelength, the wavefields of all the photons passing through this volume at any instant are added to produce a single electromagnetic wave vector; that is, the light emerging from this tiny volume is completely coherent. Were one to focus light from a filament onto a 0.2 11m diameter aperture, the light emerging from the hole would be coherent and all of it could be focused into a diffraction-limited spot.12 The problem is that it would also be a very dim spot (about 6 nW of white light by extension of the calculations noted above). As the aperture gets bigger, the coherence is reduced. Incoherent light, such as sunlight from a cloudy sky, and coherent light from a laser, are each limiting theoretical constructs. Even though, as presented in Chapter 1, using these limiting conditions simplifies the process of writing equations that describe the image formation process in the microscope, neither coherence condition can be realized in practical microscopy. In conventional microscopy, little attention is paid to the degree of coherence of the illuminating light except when considering diffraction and interference effects. Roughly speaking, light is considered to be incoherent when it does not produce speckle effects and coherent when it does. Speckle is bright if the interference of light scattered by the feature is constructive with that from the background and it is dark if destructive interference occurs [Fig. 6.6(A); Briers, 1993]. The apparent size of the scattering feature and that of the individual speckles are related to the resolution limit (or NA) of the optics. In the case of incoherent illumination [Fig. 6.6(B)], overlap between speckle patterns having different wavelengths partially cancels them out to produce a lower contrast pattern. Because speckle is an interference phenomenon, any movement of the optical system or the specimen will result in a complex change of the speckle pattern in time. Even in imaging situations that are well described by incoherent phenomena, coherent effects can often be detected if the results are examined with sufficient optical resolution (Reynolds et aI., 1989). In general microscopy, light with low coherence is desired for bright-field and reflection modes, while light with higher coherence is required for phase and interference modes. The process of fluorescent emission involves so many intermediate steps between excitation and emission that any coherence in the illuminating light is usually lost, and the light emitted from the specimen is basically incoherent. If the coherence of the illuminating light is too high, microscopy in the reflected or back scattered light (BSL) mode l3 yields images with fringes caused by interference of the coherent
Conversely, the extent to which this is NOT possible is a good measure of incoherence. " BSL is the more general term for reflected light, the term "reflected" should be reserved for the coherent scattering that occurs from smooth surfaces.
12
II
The time required for light to move one coherence length is the coherence lime.
Non-Laser Light Sources for Three-Dimensional Microscopy • Chapter 6
131
Coherent Illumination
Objective
Interference
From laser Few wavelengthsHigh-contrast speckle
Surface or refractive-index interface of specimen
A Incoherent Illumination
Many, overlapping wavelengths Low contrast speckle
To microscope
Wedge
FIGURE 6.7. Phase randomization scheme for laser use. (After Hard, et aI., 1977.)
from Born and Wolf (1980). This yields an upper limit of coherence length of approximately 111m. Points in the specimen whose optical path difference are greater than 111m cannot interfere to give rise to noticeable interference artefacts in transparent microstructures. A major function of Kohler illumination systems is to make the illumination homogeneous at the image planes and to control its coherence somewhat. However, it is essentially a coherent system and does not "scramble" light to any extent. Additional scrambling is sometimes needed to decrease spatial inhomogeneity, spatial coherence, or temporal coherence. Although most scramblers have been designed to work with highly coherent laser light. the same methods can also be used to reduce the coherence of light from other sources.
B FIGURE 6.6. Speckle patterns. High-contrast patterns result from interference between two reflection maxima in a highly coherent system. (A) Low-contrast speckle results from the summation of interference intensity patterns in a system with incoherent illumination. (B) Speckle spot size is a function of system resolution.
light reflected from any of the optical surfaces: lenses, mirrors, dust windows, and, in particular, the coverslip surface. This complex interference can appear as defined rings, but more commonly, it appears as a high-contrast granular speckle superimposed upon the image, making real image details hard to interpret. Furthermore, when the specimen is transparent and has multi-layered microstructure, the speckle spots become even more complex figures. The coherence of non-laser light sources can be modified by changing the magnification of the Kohler illumination system to reduce the effective source size. Doing so makes the light less intense but more suitable for interference microscopy. While the sun is considered an incoherent source, under the imaging conditions of high-resolution microscopy, sunlight has enough coherence to impart speckle to the image. Tungsten ribbon lamps as well as light-emitting diode (LED)-based lamps have relatively low spatial coherence because of the large size of the emitter. Arc lamps have higher coherence unless a large area of the plasma is used as the source. As the short-arc Xe/I source mentioned below uses a plasma spot that, after demagnification, is one third the area of the Airy disk, as a source for a confocal microscope, the illumination is almost fully spatially coherent (Hell et af., 1991). However, Hell and colleagues argue that because temporal coherence, associated with a bandwidth of 360 to 570nm, corresponds to a frequency bandwidth of 3.07 x IOKMHz, to the extent that the intensity spectrum of the arc approximates a continuum, it is appropriate to use coherence length
~
C/frequency bandwidth
Scrambling and Filtering the light Three methods have been proposed to reduce coherence. Hard and colleagues (1977) proposed a method to phase-randomize laser light for illuminating a conventional microscope by inserting a rotating optical-wedge-and-ground-glass combination into the light path (Fig. 6.7). Because the wedge and the ground glass rotate, any remaining temporal coherence becomes cyclical. This method requires very stable placement and accurate alignment of the rotating device and the laser relative to the microscope (Reynolds et at., 1989). The second method to reduce coherence is to focus the laser light into a flexible length of single multi-mode optical fiber (Fig. 6.8). The internal reflections in the bent fiber are constantly changing because the fiber is vibrated at up to 100 kHz (Ellis, 1979) and this makes the exit beam appear uniform in intensity over time (rather than having the Gaussian profile characteristic of lasers). The phase is scrambled by the varying path lengths of the light passing through the fiber on different trajectories. but the high radiance and monochromaticity are preserved. Technical Video Ltd. (Woods Hole, MA) markets a non-vibrating version with a fixed, single quartz fiber segment bent to a specific radius in two perpendicular planes. Applied Precision Instruments (API, Issaquah,
Zoom .:-}
-
lens to microscope
FIGURE 6.B. Phase randomization scheme for laser (after Ellis, 1979).
132
Chapter 6 • A. Nolte et al.
WA) also offers such a system as part of their Deltavision threedimensional (3D) imaging system. All these methods minimize speckle by making it change with time. Speckle is not apparent as long as the recording system records for a time period that is longer than the period of the scrambler. As the motorized system is quite slow, it is only suitable for photographic recordings. Although the fiber scrambler can be oscillated more rapidly, even this is too slow to cause a significant reduction in coherence over the 1.6 f..l.s pixel time that characterizes many beam-scanning confocal instruments. As light scramblers can be damaged if subjected to a high total light flux, provision must be made to remove heat and other unwanted wavelengths before they enter the scrambler system. Ideally, only wavelengths critical to image formation should leave the source. Dichroics that reflect only specific wavelengths should be used to separate the useful light and allow the unwanted heat to escape. In single-sided disk scanners, this would not only reduce heating but also decrease flare and scattering from the top surface of the disk. Heat-absorbing glass is the most common heat filter, but a filter consisting of a chamber filled with a salt solution chosen to screen out infrared light has much higher heat capacity. Aside from heat removal, liquid filters can also be made to function as bandpass or cut-off filters by the careful choice of the salt used . An extensive description of useful solutions is described by Loveland (1970).
TYPES OF SOURCES AND THEIR FEATURES The following section discusses the important parameters of various common non-laser light sources with respect to the goals of microscopy. In this sense "parameter" means a degree of freedom a user can change to optimize illumination for the application. Unfortunately these parameters are not independent of each other, and all components of an illumination system have to be treated together as a unit. Overall performance depends on the geometry of the source, the focal length, magnification and NA of the collecting optics, and these in tum depend on the shape and position of the mirrors and lenses. Much can be learned by removing the lamp housing and adjusting the controls used to position the source, the reflector, and the collector to project an image on the source on a distant wall. (When using arc sources, take care not to aim it in such a way that the direct light gives anyone a sunburn or strikes anyone in the eye.)
A
Structure
The Actual Source of the Light Microscope sources are optimized to produce the maximum light intensity or brightness (photons/s/cm2/steradian) from a tungsten ribbon or the arc of an electrical discharge. The filaments of halogen lamps are often bent to resemble disks or wide, flat bands to match the input aperture of the light-collection optics. Arc lamps generally generate light in a ball discharge at the tip of a pointed electrode (Fig. 6.9). The two electrodes in xenon arcs have different shapes. The anode has a bigger diameter and a flatter tip. As a result, the light will be most intense where electric flux lines are closest together near the point of the cathode [Fig. 6.9(8)]. As the pointed electrode erodes, the field at the tip decreases and the plasma ball becomes larger and therefore less intense. These sources are geometrically similar but are different in size. The brightest part of the arc in a common HBO-lOO arc lamp is about 0.3 x 0.5 mm in cross-section, the tungsten filament of a 100 W halogen lamp is about 4 x 2 mm wide. Both source dimensions are set by the manufacturer, and there is no option to vary them. Electrons passing through the depletion region of any forwardbiased semiconductor diode lose energy equal to eVg , where Vg is the bandgap of the semiconductor. This energy is converted into a photon having an energy equal to the bandgap energy. Silicon diodes emit in the near-infrared (IR) region, but diodes made of other semiconductors emit in the visible and even the near ultraviolet (UV). When such a diode is configured primarily to produce light, it is called a light-emitting diode or LED (Schubert, 2003). Recently, technological developments have increased available power levels and now LEDs are used in applications where their long life and high efficiency are important, such as traffic lights. They are of interest to microscopists because they are compact and efficient light sources that can emit a high flux of quasimonochromatic photons from a small area.
E
§.Ot---h~'"
§. ~ 1 .5 1--~h..-"':=:=5l=:::""cl"""':"'''''''':'''--l
90 70
LU Cl
o::c
LU Cl
~
o ::c
()
:::E
~
oa::
()
:::E
2 1----+\-+lrHl-:*:+\*I-++----I o
z
fE 0.5
70
~
en C5
t
I. Structure (spatial distribution, source geometry, coherence, alignment, collecting optics, etc.). 2. Available wavelengths. 3. Stability in time and wavelength. 4. Source radiance (brightness). 5. How source parameters such as intensity, color, and spatial distribution are controlled.
B
E
u.. ~
The important parameters of non-laser sources are:
I---+l-f---f-E-++-+++---;
LU
()
90
Z
~
3
C5
0
I---+-~~~~~-+---; CATHODE
2
1
0
1
2
- - DISTANCE FROM ARC AXIS (mm) - -
0.5
0
0.5
DISTANCE FROM ARC AXIS (mm)
1
FIGURE 6.9. Iso-intensity plots from the discharge of a mercury arc lamp (A) and xenon arc lamp (B). Values in ked/cm' " Note the different size scales and the fact that the Hg arc has two intense plasma balls while the xenon has only one, and this is smaller and less intense than the Hg. (Plots kindly provided by OriellNewport, Irvine, CA.)
Non-Laser Light Sources for Three-Dimensional Microscopy • Chapter 6
AI
In Ga p
~ . In, ,~a,N • .' 'v _ Y
AI , :.?a,~s_
:. ..
A
GaA,!" p.
AI-bonded n-terminal
I•
0,8
Topview
Sideview
1,0~~r'I~~ r --:----;-~~~~A:::::;f"I~f\~ 0 ,9 II
133
280 11m
/ MOW active
0,7
Sokier layer
-;; 0,6
.cl.
?:- 0.5
'iii
.s 0.4
.!:
~,
0.3
0,2 0,1
p-terminal
Submount
B Standard
ThinGaN
NOTA
ATON
\
t
'I.
GaAs
~
~
350 400 450 500 550 600 650 700 750 800 850 900 950 1000
t
wavelength [nm]
FIGURE 6.10. Spectral distributions of monochrome LEOs. The compounds used for the LED dies are shown above. The black line with a sharp peak at 460nm and a smaller, wider peak at about 575 nm , is the emission curve of a so-called "white" LED - a blue emitter surrounded with a phosphor that absorbs at 575nm and emits at longer wavelengths.
An LED source consists of an area of semiconductor crystal approximately 0.3 mm x 0.3 mm in size called a die. The most common crystals used are Gal _x Al,Asl _yPy, GaN, and ZnSe and each emits in a different waveband (Fig. 6.10). Normally one or more dies are embedded in a larger LED structure for protection, light collection, and electrical and thermal handling. The advantage of LED technology is that one can combine these small units to build up a light source of the shape best suited to the needs of the application. Possible source geometries are limited only by heat dissipation and the permitted package density of the surface mount device (SMD) technology used to integrate a number of dies onto the printed circuit board. Very dense, bright, custom-designed sources can be fabricated in this way. Figure 6.11 shows the general structure of an LED and how they have changed over recent years (Steigerwald, 2002). Die dimensions up to 1.5 x 2 mm are now available.
2001 22x flux
FIGURE 6.11. CA) General structure of an LED with the different layers marked. The mirror is used to reflect the back-emitted light to the right direction. The metal bond pad on the top is for the electrical contact. (B ) Changes in the geometry of LED dies in recent years have led to larger dies and have increased the efficiency with which light is emitted. (With kind permission from OSRAM Optical Semiconductors, Regensburg, Germany.)
3000k ~------------~--------------'-----~------'
.. ..
How to Cope with the Heat? The most important aspect of any microscopy light source is an efficient heat sink. Incandescent and the arc-based lamps produce a lot of heat because of their low optical conversion efficiency (5 %-10% ). The holders and hou sings of these lamps are made of a material resistant to high temperature and designed to dissipate -100 W of heat. As a result, they cannot be mounted inside a microscope system. Although present LEDs have similar efficiency, all the photons produced are emitted over a narrow range of wavelengths (see below) and they operate at a much lower temperature. This means that less electrical power is needed for the same optical output and they can be more compact; for instance, they can be bonded to a small metal heat sink, cooled by a small, computer-controlled fan . This technology makes it possible to mount LED sources inside the mi croscope system, closer to the specimen, and avoid loss of light intensity in transit. Despite this flexibility, it is important to remember that LED-based sources do need an efficient heat sink becau se operating much above room temperature causes lower lifetime and a loss of optical output efficiency (Perduijn et ai., 2004 ; Fig. 9. I 2).
1998 15x flux
1994 3x flux
250%
.
.
..
.
.
-~-----~------- ~ ---------------- ~ ------
:
:
200%
.
150%
,
~ ----------------~-----,
,,
100%
,
50%
.
,
,
------- i ------- ~ -------~------- ~--------~-----: :
.
.
..,
.,,
20
40
60
80
0% _ _.....:-: - 20
0
100
Temperature [0C] FIGURE 6.12. The dependence of li ght flux on LED temperature, standardized to 70°C for dies made using different types of LED crystals. Because dies emi tting at longer wavelengths have a smaller bandgap energy, they are more affected by changes in temperature. Formulas for the compounds, note above.
134
Chapter 6 • A. Nolte et al.
Heat sink
Collector Reflector mirror -...
FIGURE 6.13. Diagram of lamp housi ng. The discharge of the arc is located at the focal points of the rear reftector and the first condenser lens, both of which can be moved in three directions to permit alignment with the optical system.
Focus lever
Adjustment screw Arc lamp
Ceramic insulator
Table moves in X, Y, Z
Reflecting and Collecting the Light: Source Optics As incandescent and arc lamps are almost isotropic emitters that radiate equally in all directions, a spherical reflector can be placed behind the source to create the image of the source beside the actual arc. A collector lens is mounted so its focal point coincides with the center of the source; it catches a large fraction of the light and focuses it at the BFP of the condenser (Fig. 6.13). Due to the small source size and the desire for a short compact structure, collector lenses are usually aspheric and have high chromatic aberration. LED-based sources use three different principles to reflect and collect the light generated inside the die. The first approach was to use the clear, molded plastic LED package itself to collect and focus the light. Although suitable for low-level bright-field applications, this method is rarely used in microscopy because of the limited optical power available from a single LED (e.g., the white light LED NEPW500, manufactured by Nichia with a 0.3 mm 2 die). The second approach is to work without a collector and print the LED dies directly onto the printed circuit board. The density of the packaging achievable in this way is only limited by the need to bond the single dies with the connecting wires and the ability to dissipate heat. The disadvantage is one loses any light that the LED die emits to the side (e.g., the monochrome arrays available from Laser2000 with 88 single 0.3 mm2 dies packaged on a ceramic plate). The third possibility is to place the LED die into a mirrored well that acts as a reflector (e.g., the "Golden Dragon" manufactured by ORAM) and then arrange these units onto a
printed circuit board. Because the reflectors are bigger than the dies, this results in a lower pac king density. As every die is a separate source, when an LED array is built up of several dies, one needs a different method to combine the light from them than one would use for a conventional lamp. The most efficient way to collect the emitted light is to use a microlens array placed the proper distance in front of the LED array. This array can be made of molded plastic or of glass and must be designed so that every LED die has its own collecting lens (Fig. 6.14). The main goal of this configuration is to catch as much light as possible and deliver it into the acceptance angle of the microscope illumination optics so that it fills the condenser aperture diaphragm with axial, parallel light as homogeneously as possible.
Source Alignment The rule of thumb says the smaller the source, the more important it is to align it properly. For arc lamps and incandescent lamps, one must align the reflector and collector in such a way the source and its image from the reflector lie side by side and are centered in the aperture diaphragm of the illumination path. Only in this way will one be able to fill the aperture diaphragm approximately homogeneously (Fig. 6.15). Modern systems usually either incorporate special viewers so that one can check alignment or use selfaligning sources such as the self-aligning HBO-lOO source available from Carl Zeiss (Fig. 6.16). The heart of the self-aligning source is a photodiode with four quadrants. Some of the light emitted by the arc in the direction towards the reflector passes through a small hole onto the optical axis where it is captured by
Non-Laser Light Sources for Three-Dimensional Microscopy • Chapter 6
Electrodes w ith discharge arc
Array of single sources "Macroscopic optics"
135
Photodiode with four quadrants
a
c
II
Optical axis
III
t
Collector lens Side view
Reflecting mirror
Top view
Array of microlenses FIGURE 6.14. Principle of a micro-lens array. CAl Each mircro-lens efficiently cou ples the light from a single LED source to the macroscopic optics of the microscope. As a result, the macroscopic collector optics can have a lower aperture, and therefore, a larger focal length at a reasonable diameter. In addition , such optics have lower spherical and chromatic aberrations. (8) Hexagonal micro- le ns array made of quartz glass. Thi s example is normally used in telecommu nication to couple two bundl es of optical fibers together.
the quadrants of the photodiode. A microcontroller measures the intensity of all four quadrants and moves the burner on an xy-stage until there is no difference between them. As the right reflector position depends on the particular optics of the microscope, the zposition of the reflector still has to be aligned manually. For LED-based sources, alignment is simpler. In bright-field microscopy the main goal is to fill the aperture diaphragm as homogeneously as possible. This can be accomplished by mounting a large (5-IOmm) LED array just behind the aperture diaphragm. The problem in this case shifts to the alignment between the microlens array and the LED sources. As the focal length of the microlenses has to be relatively short and the LED sources are very small, the tolerance to a misalignment is also small.
Wavelength Hi storically, many fluorophores were selected and used because they are excited by the intense lines of Hg arc lamps. This was not only because of the increased brightness but also because the narrow bandwidth made it more easy to design effective dichroic
FIGURE 6.16. Components for a self-aligning source. Light passing through small hole in the reflector strikes a quadrant detector. Signals from this detector are used to move the arc in y and z until the output of all four diodes are eq ual.
emission filters and beam-splitters. In addition, microscope objectives are often designed to give optimal correction at these wavelengths (Herman, 1998). As the laser wavelengths available seldom match these wavelengths, the microscopist using a laser confocal microscope must sometimes exchange familiar fluorochromes for new, less fa miliar ones that can be excited at laser wavelengths. Because histochemistry is complex enough without having to try entirely new stains and chemistry, it is important to have a light source flexible enough to excite the most suitable dye for each application. The sun has very high brightness and the continuous spectrum of a black-body radiator with a surface temperature of 5800 K (Fig. 6.1). A clock-driven heliostat was used by Petran and colleagues (1985) to track it as a light source for confocal microscopy. The sun 's broad continuous spectrum allows easy selection of wavelengths for difficult specimens. Aside from the sun, only a synchrotron can provide as bright and continuous a spectrum (van der Oord, 1992; Gerritsen er al., 1992, 1994). Although using the sun or a synchrotron as a light source might be ideal in many ways, the nuisance of having to depend on geography, season, cloudiness, and time of day led to the development of portable sources. Incandescent lamps also emit essentially black-body radiation, the spectral shape of which depends on the temperature (Fig. 6. 17). However, because the filament is only 3200K or 3400K, the light is less intense and has more red li ght than sunlight. Although it is possible to increase the blue component by raising the tempera-
FIGURE 6.15. Image of the filament of a 100 W halogen source. (A) Image of fil ament and its reflection superimposed; (8) filament and its image side-by-side, right (C) after insertion of ground glass to randomize image of filament. The illumination optics do not magnify the image of the filament quite enough to fill the BFP completely. Even with the ground glass in place one can see a visibl e drop of intensity at the outer edges or the BFP.
136
E c: N
E
Chapter 6 • A. Nolte et al.
9
8 7
3!: g 6
10,000 K ,OOOK~~~~
4,000 K
Q)
0
c:
.l!l
'0
x
W
~ ti Q) a.
en
:i -1 8' -2
--l
_3L-----L---~~~--~~--~~-L~--~-L----~
100
1,000
10,000
Wavelength (nm)
FIGURE 6.17. Spectral distribution of black-body radiation at various temperatures. Note that the vertical scale is logarithmic. The total luminosity of a black body varies with the fourth power of the temperature.
ture. doing so increases the rate at which the tungsten sublimes. The tungsten vapor condenses on the inside surface of the quartz envelope where it absorbs light from the source and heats the envelope. In addition, the filament becomes thinner until it fails. The halogen gas in a quartz-halogen source interrupts this process by first reacting with the evaporating tungsten to form tungsten-halide compounds. These then decompose when they strike the hot filament, effectively returning the tungsten to the hottest (and thinnest) part of the filament. This permits the filament to be operated at a higher temperature with little darkening and a longer time before failure. Higher temperatures may also deform the filament structure, causing it to move from its correct location in the optical system. The optimal wavelengths for exciting classic fluorescent dyes (436nm, 546nm, 579nm, 365nm, 405 nm) coincide with Hg arc emissions lines [Fig. 6.5(A)]. The carbon arc (not shown) has a single very intense line at 400nm. Mixed-gas Hg-Xe arcs have many intense, useful spectral lines in the UV and visible. Xenon and zirconium arcs have spectral lines in the near IR. The emission spectra of arc sources can be classified in three ways: (1) continuous, (2) line spectra, and (3) mix oflines and continuous spectra. Figure 6.5(A) shows the spectra of arcs using mercury and xenon, as well as a tungsten halogen incandescent source, plotted with the same horizontal scale. The super-pressure xenon arc provides intense broadband illumination without prominent spectral lines in either the UV or the visible because the high pressure broadens the xenon spectral lines. Because the intensity of the continuum of a 75 W high-pressure xenon lamp is 2x higher than the continuum of a 100 W Hg arc lamp, the xenon lamp is better suited for low efficiency fluorophores not excited well by the prominent Hg lines. The availability of broadened spectral lines sometimes allows the simultaneous activation of several fluorochromes with differing emission wavelengths, including those in the Uv. The only disadvantage of the xenon lamp is the high pressure in the bulb (approximately 30 bar at room temperature and approximately lOx that at operating temperature). This makes it necessary to use protective gloves, goggles, and a shield to shelter the chest when changing a bulb in case it should explode. The LED technology makes it possible to supply the right excitation wavelength for each fluorophore. Most wavelengths are now
available from UV (365 nm) up to IR (>800nm; Fig. 6.10) with intensities sufficient for widefield fluorescence. The FWHM of a quasi-monochrome LED varies from 20 to 40nm, which is similar to the width of the excitation band of many fluorophores. Compared to laser light, the wider bandwidth of the LEDs make it easier to excite a variety of dyes, and compared to the continuous spectrum of an arc lamp they are cooler, smaller, and provide an easier way to choose the wavelength one wants and to do so rapidly. That said, one still needs to use excitation filters to remove the tails of their emission wavelength distribution. Using conventional fluorescence microscopy as many as five different fluorescent labels have been imaged simultaneously in the same living cell (De Basio et aI., 1987).14 This certainly could have been done as well using a disk-scanning confocal microscope having an efficiently configured arc source. When imaging living cells, success requires using minimum light intensity at the specimen plane and choosing illumination wavelengths that interfere as little as possible with the life process under study. Using a continuous spectrum source, one can often make small wavelength changes that reduce specimen mortality. Because fluorophore excitation maxima are altered by their molecular environment, and because this is especially true for a fluorophore attached to a functioning macromolecule in a living cell, fine tuning the excitation wavelength can result in increased emission. Using strip- or wheel-type, continuously graded interference filters, such as those manufactured by Ocean Optics (Dunedin, FL), the microscopist can often precisely adjust the illuminating wavelength to minimize interference with the process under study and to maximize the excitation of a fluorophore in a particular cellular environment.
Stability in Time and Wavelength Ramp-Up and Short-Time Stability Everyone who has used an arc lamp knows the buzzing noise that occurs when the lamp is switched on and the plasma arc discharge is building up between the two electrodes. All sources based on plasma-discharge or incandescence need a considerable time to reach thermal stability. Figure 6.18 shows typical intensity ramp-up curves for the various sources. 15 All lamps that produce significant heat show a dependence of the emission on the source temperature. This even applies to LED-based sources. It can take up to I h until the sourcc is sufficiently stable to make reproducible measurements or to make a good time-lapse movie. Once operating temperature has been reached, the halogen lamp is the most stable source over time periods of a few milliseconds because of the high thermal inertia of the tungsten filament. LED-based sources react very fast (in a few microseconds) and, therefore, they are affected by any highfrequency instability in the power supply. In general, the most unstable source is the arc lamp. Not only is the arc itself a chaotic, flickering discharge but its light output can also be affected by ambient electromagnetic fields or an unstable power supply. Stability can be increased by using the signal from a light sensor as a feedback signal to control the excitation power. The Oriel 68950 power supply controller (Newport, Inc., Irvine, CA) De Basio and colleagues measured five cell parameters: (fluorescent dye) nuclei (Hoechst 33342), mitochondria (diIC-[5]), endosomes (lissamine rhodamine), B-dextran, actin (fluorescein). cell volume (Cy7-dextran). 15 These curves were measured with an upright microscope, using a source mounted in a common epi-illumination lamp housing. A salsa beam-splitter was used and a radiospectrometer placed at the BFP of the condenser. 14
Non-Laser Light Sources for Three-Dimensional Microscopy • Chapter 6
~
§
6.3r - - - - - - - - --;=======iI 6.2 ...-:;;:-- - - - - - - - - - --1
.ci
-
HB0103W/2 LED Halogen 100W
-
~ 6 .1
~6.0 ~~::~~~~::::::::~::::::::~:;::::~
. c
Q)
- -=---==i
C 5 .9 t--~""-----""Ia.!==~==:::::j;::::-......
5.8 .....- - - _ _ . - - - -......- - - -......-------4
o
50
100
200
150
Time from start (min.) FIGURE 6.18. Ramp-up of brightness and long-term stability for various microscope light sources.
137
time to keep the electrodes cool, as is the case for the electronic flash used for photography. Alternatively, the position of the arc plasma can be stabilized either by a periodic magnetic field imposed by a rotating permanent magnet or by the superimposition of a small, high frequency AC current on the main DC excitation current (Woodlee el al., 1989) . As the photoelectric effect is fully reversible, it is not surprising that the LED has the lowest operating temperature and, as a result, is the most stable source. In addition, as long as it is operated at the proper voltage/current, the LED has a much longer lifetime than all other sources. One usually has to change an arc about every 200h and incandescent lamps about every 500h, but the LED sources have lifetimes in the range of a few thousands hours without significant loss of intensity. Some manufacturers promise a lifetime of 100,000h before the source intensity drops to 70%.
Stability in Wavelength improves the stability of arc and halogen sources from 0.4% to 0.0 I % (Fig. 6.19). However, these figures are for total light output, and they cannot prevent local flickering in the particular region of the plasma that happens to be focused onto your specimen. Some suppliers of deconvolution systems, such as Applied Precision Instruments (Issaquah, WA), constantly monitor arc output during the CCO exposure time. They then use this information to normalize the CCO output for each plane of the z-stack.
Long-Time Stability, Degradation When warmed up and powered by a regulated power supply, tungsten halogen sources are suitable for making photometric measurements. Generally speaking, arcs are less stable than filament lamps because the points of the electrodes slow ly erode. The larger radius of curvature that results reduces the concentration of current flow (and brightness) near the tip and also increases the power level needed to sustain the arc . Eventually the arc will not ignite. The intensity of the xenon arc can be deeply and rapidly modulated in
1.010
Radiance
1.008
i!'
'iii
cQ)
1.006
Without 68950 control
C Q)
>
1.004
.~
Q)
a:
In the discharge of an arc lamp, the gas pressure affects the composition of the spectrum. As the pressure rises, the prominent lines broaden and the continuum rises [e.g. , compare the continuum brightness of a mercury and a high-pressure xenon lamp in Fig 6.5(A)]. On the other hand, the peak wavelengths of prominent lines shift just a few nanometers «5). The continuous spectrum of an incandescent lamp depends only on the temperature of the filament (Planck's law) and on the gas present in the bulb (Fig. 6.17). At a fixed current, a change in color temperature can only occur if evaporated tungsten condenses on the glass bulb. A change in the current to an LED causes a shift of the emission peak similar in magnitude to that seen in lines of the arc lamps. This may be due to the LED die not being perfectly homogeneous. The size of the shift depends on the type and quality of the crystal material used. As the shifts are small compared to the width of the lines, one can neglect the effect for arc lamps. For LEOs, one either has to calibrate the dependence of the wavelength on the operating current or operate the die in a different way (see sections on Radiance and Control) .
1.002
With 68950 control 1.000 0
20
40
60
80
Time (hours) FIGURE 6.19. Stability plot of halogen source, with and without opticalfeedback power stabilizing. Although quartz halogen sources are very stable, temperature rise increases output, and air-conditioning and other factors can produce transients. These can be compensated for by using the output from a monitor photodiode as the input to a negative feedback system controlling the power supplied to the lamp. (The plot was kindly provided by OriellNewport, Irvine, CA and reflects the performance of their 68950 power-supply controller.)
Arc lamps are several orders of magnitude more radiant than tungsten filament lamps [Fig. 6.5(A)]. The HBO-IOO (l00 W highpressure mercury arc lamp) is the most radiant of the commonly used lamps, whatever the wattage, because it has a very small source size (compared to the HBO-200, for example). Because of the optics rule mentioned at the beginning of this chapter, the larger arcs are only useful to illuminate larger areas of the specimen rather than to illuminate a single spot with maximum intensity. The main limitation on arc radiance is that the electrode tips erode or even melt as the power level increases (Fig. 6.5). New arc sources of very high radiance have been described by Steen and Sorensen (1993). In these sources, commercial Hg, Xe, or Hg/Xe arc lamps have been modified to permit direct water cooling of the electrodes and the superposition of large-amplitude short-duration (20 /ls) cun'ent pulses on the DC operating current. Increases in output of up to 10-fo ld were observed during each pulse. Hell and colleagues (1991) describe a new generation of short arc lamps with extremely short electrode distances (0.5 mm) using a XciI fi II and tungsten carbide electrodes in a quartz bulb. The tungsten iodide TIJ dopant gives rise to a radiating plasma spot only 150/lm in diameter. Over the 450 to 550 nm band, the radiance exceeded that of the conventional Xe short arc by a factor of 3 to 5 and that of an HBO-IOO mercury lamp by 12. The new arc lamp
138
Chapter 6 • A. Nolte et al.
has useful radiance from 360 to 570 nm. These developments push the arc source radiance closer to the realm once thought available only from lasers. Tables of the radiance of various arcs and lasers can be found in the catalogues of Melles Griot (Rochester, NY) and Oriel Corp. (Stratford, CT, now a division of Newport Inc.). When driven by a 120 Hz square wave, the modulated (Hg-I) arc lamp introduced by LTM Corp. (Sun Valley, CA) produces a very useful spectrum with an efficiency of 110 lumenslW, compared to 301umensIW for a normal xenon arc. The deep squarewave modulation of the Hg-I arc reduces average heat production, allowing the lamp housing to be compact enough for placement near the microscope. As the bulk depletion region inside the LED die is an isotropic emitter (lambert), one might assume that the light leaving the front surface of the die would also be isotropic in all directions. However, as the light generated in the volume of the crystal must pass through the crystal/air interface, any rays that strike this surface at less than the critical angle will be reflected and reabsorbed by the crystal. Approximately 50% of the light generated internally is lost in this way and less light is emitted at bigger angles. The radiance of the high-brightness LEDs available today is still far less than that of the prominent lines of the arc lamps (Fig. 6.20). In the continuous operation mode, the brightest 2 x 2 mm LED die (Luxeon, 5 W Emitter, Fa. Lumileds, San Jose, CA) today delivers around 50% of the continuum radiance of a 75 W XBO at the same wavelength, and is bright enough to get an acceptable fluorescent signal from a well-stained specimen (Braun and Merrin, 2003). Unfortunately these high-power dies are not yet very stable and because of the high thermal load they degenerate very fast. More to the point, they are still not available in all wavelengths. At present, better results are obtained using smaller emitters (I W emitter, Luxeon) to build up a light source. With a proper heat sink these are now very stable. In pulsed-mode operation (see subsection under Control) , the available radiance can be a factor of 20 higher than for the same unit used in continuous mode. When grabbing fluorescence pictures quickly, this mode is the most suitable one, as one can trigger the camera with the light pulse to ensure efficient usage of the emitted light. This mode of operation is also now being used in the machine vision area of industry where LEDs have become a long-life substitute for xenon flash lamps for illuminating moving objects (see, e.g., available light sources at http://www.laser2000.de). For other manufacturers, see the links in Table 6.1.
i::'
.~
Q.)
~
._-
-- ....
...
10.00 .--~.ot:..,r---......~-:-,-r-r::;-'--'r-:-:-:~-::p~........- - '
,
InGaN
AllnGa
,
I,~~+------+----~~-----r---- A/GsAs 1.00~
\
" ........
SIC
Low performance , commodity LEDs
0 . 01~----~~--~-----L------L-----~----~
400
450
500
550
600
650
700
Wavelength (nm) Blue
FIGURE 6.20. LED luminous intensity for various colors and dye compounds. Efficiency is measured in lumens/watt and refers to the efficiency with which electric power is converted into optical output.
TABLE 6.1. Useful Links Microscopes, optics, and light sources http://www.zeiss.coml http://www.lot-oriel.com/ http://www.mellesgriot.com http://www.chroma.coml http://www.edmundoptics.com http://www.optics.org/ http://www.wahl-optoparts.de/ http://www.oceanoptics.com LEOs http://www.luxeon.comlproducts/family.cfm?familyld=l http://www.osram-os.coml http://www.nichia.com/ http://we.home.agilent.comIUSeng/nav/-11143.0/home.html http://www.led.com/ http://www.optotech.com/ http://www.toyoda-gosei.com/led/index.html http://www.stockeryale.com/ http://ledmuseum.home.att.netiledleft.htm http://www.laminaceramics.com Basics and history http://micro.magnet.fsu.edu/primer/ http://inventors.about.comllibrarylinventorslbllight.htm
Control The quartz-halogen lamp is simply driven by a stabilized DC power supply converting the plug voltage into an adjustable voltage of 2 to 12 volts. Varying the voltage controls the temperature of the filament and thereby the spectral properties and intensity of the light. Arc lamps are usually also driven by a current-stabilized power supply. The current can be decreased to 70% to lower the optical output and conserve the electrodes if one does not need the full optical power. Below 70% the plasma becomes unstable. Because of the decreasing temperature of the discharge, the vapor pressure drops and the discharge stops. Modern arc lamps have a heating filament wrapped around the bulb. This filament heats the bulb, restoring the vapor pressure, and allowing the current to be decreased down to 30% without stopping the discharge. Neither arcs nor halogen lamps can be switched on rapidly. To change the emitting wavelength or intensity quickly one has to use mechanical shutters and filter wheels and switching times are usually longer than lOOms. The stabilizing circuitry of the arc power supply can stabilize the voltage, the current, or the total power (voltage x current). If the voltage is stabilized, the current (and the brightness) will slowly decrease as the electrodes become worn round. If the current is stabilized, the brightness will stay fairly constant l6 until the electrodes become too rounded for the arc to "strike." However, because an ever-higher voltage is required to maintain the fixed current, as the electrodes wear the power sent to the arc slowly increases. As a result, it can overheat and sometimes explode. Although power supplies that stabilize the total power level will avoid overheating, the light output will slowly drop with the current as the voltages needed to maintain the arc increases. All this suggests that it is best not to run an arc too long and that it might be a good idea to monitor the voltage across the arc to detect warning signs.
16
The total light will stay about the same but it will be less concentrated at the tip of the electrode.
Non-laser Light Sources for Three-Dimensional Microscopy • Chapter 6
139
FIGURE 6.21. What an actual, functioning LED microscope source looks like. A close-packed array with four monochrome (470, 525, 590, 620nm) colors on one matrix.
As each electron passing the depletion region emits one photon, LEDs can be controlled by any current-stabilized electrical source.17 Depending on what is needed, the LED configuration as well as the control circuit can be easily changed. If only a single LED device is used (e.g., a "white" LED, usually one that couples a blue primary emission with red and green light from blue-excited phosphors), only a single-channel current source is needed and the intensity is controlled by changing the current flowing through the LED. It is more common to use more complex LED structures combining LED dies with different emission wavelengths to obtain either narrow-band light for multi fluorescence or "white light" in bright-field microscopy (Fig. 6.21). Such devices are controlled by a multi-channel current source. By rapidly switching these currents on and off, it is possible to change intensity or emission wavelength on a microsecond or even nanosecond scale. This is a very important feature for short-time scale methods such as fluorescence lifetime imaging measurements (FUM; Hermann et aI., 200 I). The switching on this timescale is called pulse mode. Because the peak emission of a given type of a LED can be shifted by changing the current level, it is often more suitable to operate multi-LEDs in the pulse mode. One sets the peak current to produce the desired output wavelength and then changes the average source brightness by varying the pulse width at a fixed peak current. Although, compared to continuous mode, more total light is available in this way, using a higher current for more than a short pulse will lead to thermal damage. The "damage threshold" current pulse width must be evaluated for each LED. The spectral output of the LED can be controlled very precisely in this way. The optical output follows the current pulse without significant delay. Pulse-modulation frequencies up to megahertz are possible.
MEASURING WHAT COMES THROUGH THE ILLUMINATION SYSTEM The procedures for measuring the light throughput of any microscope with a photometer are thoroughly described in a step-by-step manner in the book Photomicrography (Loveland, 1970). He even describes making a photometer for such a purpose, but a useful substitute can be made by attaching almost any small photodiode (or even a small "solar cell") to an inexpensive digital voltmeter set on a sensitive current range. Using this, one can measure light in the
17
On the other hand, use of a voltage-stabilized source will almost certainly damage these devices. At a fixed voltage the current increases with temperature, causing thermal runaway.
most obscure locations within the microscope. "Photon bookkeeping" based on such measurements is the only way to pinpoint those parts of the light path where preventable loss is occurring. Microscopists not interested in building a photometer can obtain one of the commercial units. World Precision Instruments (Sarasota, FL) markets a fiber-optic monochromator and photomultiplier subsystem that can be used to examine light at the intensities present at any location in a confocal optical system. A 50 11m fiber is standard with this system and other vendors are listed in Table 6.3. Young (1989) described the use of a feedback-controlled LED to generate known amounts of light from small (5-50 11m) sources. Using this system in either the source plane or the image plane, he was able to calibrate the input-output characteristics of a microscope system over four orders of magnitude. For a measurement of radiance resolved by angle and wavelength simultaneously one must use a professional radiospectrometer such as the CS-lOOO made by Minolta. Such a device can provide very detailed information on the quality of an illumination source and the illumination path. Selective light loss can occur anywhere along the optical path, heat filters, tilted interference filters, and dust windows, as well as obvious lens elements rconsider that every lens surface causes the loss of at least I % of the incident light despite anti-reflection (AR) coatings]. In the past, the transmission characteristics of objectives were seldom displayed in manuals, and even the general characteristics of UV versus non-UV lenses are still often hard to obtain (some figures are listed in Chapters 7, 27, and 29, this volume). Popular photography magazines often feature the color bias of various camera lenses, and these show that the color effect of a given AR coating is different for the large NA rays than for those near the optical axis, due to the quality and the optical properties of the lens coating. The AR coating on dust windows may block the transmission of UV or IR illumination. Epi-illumination requires broad transmission in both the illumination and the viewing direction.
The Bare Minimum Even if one isn't inclined to be a full-time photon sleuth, it is wise at least to monitor the performance of the illumination system under a few commonly used standard conditions. For example, one should monitor how much light emerges from a favorite high-NA objective when it is set up for Kohler illumination with a particular filter cube and with the field diaphragm set to just illuminate the full field of this objective. This can be measured with a I cm 2 photometer paddle held in front of the objective. Assuming that one does not want to oil-couple the sensor to the objective, make
140
Chapter 6 • A. Nolte et al.
sure that the front of it is free from oil so that the fraction of the light escaping into the air is set only by total internal reflection from the flat front element (not an ideal situation but at least one that is repeatable). Then do the same with the most commonly used low magnification lens and maybe a couple of different filter cubes. Changes in these numbers will warn of misalignment, dust, aging arc bulbs, damaged filters, or help one determine the final resting place of that bit of paper that fell down inside. An alternative to the photometer paddle is an Ulbricht sphere (also called an integrating sphere). Using an Ulbricht sphere, there is no problem with rays being reflected by the surface of the detec-
tor because the entrance of the detector is just a hole. Using this device, it is easy to measure what fraction of light reaches the specimen. On an upright microscope, remove the condenser and the xystage and mount the Ulbricht sphere below the objective with the entrance hole at the specimen plane. It may help to use the halogen lamp with the bright-field contrast in reflection to align the parts to each other. Mount the lamp of interest at the reflection port for illumination. All the light passing the objective is captured by the sphere. Table 6.2 shows the optical power delivered to the specimen plane measured in this way for various light sources and for two objectives with different fields of view.
TABLE 6.2. Optical Power of Different Light Sources in the Specimen Plane Filterset No. Used
Excitation Wavelength (pass width) in nm
Examples for Fluorescence Dyes
HBOlO3W2
XB075
LEDs
HALlOO (at 12V)
Optical Power [mWl
Objective PlanNeoFluar 40xlO.75, field of view 0.625mm diameter #2
365 (50)
DAPI, Hoechst33342
30.8
0.1
#47
436 (20)
CFP, ChromomycinA
10.5
3.5
#9
450-490
GFP, Fluorescein
6.4
12.7
#46
500 (20)
Calcium Green, YFP
1.6
4.4
#14
510-560
RhodamineB
20.2
12.7
#20 #26
546 (12) 575-625
Cy3. Rhodamine Cy5
11.1 125.0
2.8 9.7
4.4 Nichia 3 W 365 nm 1.4 Luxeon 3 W 450 nm 1.9 Luxeon 1W 470nm 4.8 Luxeon 3W 470nm 10.0 Luxeon 3 W 450 nm 0.4 Luxeon I W 505 nm 1.5 Luxeon 3 W 505 nm 1.3 Luxeon I W 530 nm 2.4 Luxeon 3 W 530nm 0.1 Luxeon I W 505 nm
Q 0.5 Luxeon 1 W 590 nm
0 0.2 3.2
1.5
7.1
1.4 8.3
Objective tluar 10xlO.5, field of view 2.5mm diameter #2
365 (50)
61.4
3.4
#47
436 (20)
43.9
10.5
#9
450-490
27.2
33.8
#46
500 (20)
8.5
11.7
#14
510-560
79.0
36.7
#20 #26
546 (12) 575-625
44.2 243.1
8.3 30.1
8.5 Nichia 3 W 365 nm 3.4 Luxeon 3 W 450 nm 4.8 Luxeon I W 470nm 11.1 Luxeon 3 W 470nm 23.0 Luxeon 3 W 450 nm 1.5 Luxeon I W 505 nm 4.3 Luxeon 3 W 505 nm 2.8 Luxeon I W 530 nm 4.0 Luxeon 3 W 530nm 0.9 Luxeon I W 505 nm
Q 2.7 Luxeon I W 590 nm
0 1.8 11.1
4.5
22.6
4.2 30.1
Non-laser Light Sources for Three-Dimensional Microscopy • Chapter &
141
The different filter sets were chosen to show a representative profile of the optical power of the light sources at different wavelengths. The bandpasses of filter sets #9 and #46 lie between two prominent lines of the HB0103W2. The XB075 is much more radiant. The optical power of the HALlOO is between one third and one half of the continuum of the HB0103W2. This is sufficient to excite the brightest dyes. We have made preliminary measurements of two examples for LED-based sources. In each case, a single I W Luxeon emitter was used with a single collector lens with no further alignment. This means that the values for the LEDs in Table 6.2 represent a minimum for the optical power at the specimen plane using only a single, 1 W die. With a more radiant emitter, or more individual dies and proper collecting optics, the optical power at the specimen plane can easily be increased by a factor of 4 to 8. This would put the LED radiance between the HAL 100 and the arc sources.
mirrors and beam-splitters permits the illuminating and sensing apertures to be distinct; thereby preventing light reflected by the solid part of the disk on the illumination side from reaching the imaging side of the system. The light-source optics must fulfill two functions. They must illuminate the area of the disk that will be utilized to form the final image (usually I to 2.5 cm diameter). In addition, this light must leave the disk with the correct angle of divergence to fill the BFP of the objective lens. On disks with very small holes, diffraction at the holes will usually ensure that the second condition is satisfied, so the problem becomes how to get the maximum amount of light incident on the active area of the disk. This is important because, as only 1% to 2% of the disk is open, the system is very wasteful of light. Careful matching of a large-NA collector lens and an optimized condenser is needed to ensure that sufficient light reaches the specimen to form an image in a reasonable time.
TYPES OF CONFOCAL MICROSCOPES THAT CAN USE NON - LASER LIGHT SOURCES
Single-Sided Disk Scanning: Basic Description
The notion that confocal microscopes must use laser illumination is widespread because most commercial confocal microscopes are single-beam instruments (Leica, Nikon Real Time, Olympus, Zeiss), and these currently use only laser illumination. In fact, none of these microscopes even make a provision for the user to connect a non-laser light source for use in the confocal mode. Nevertheless, it is not true that single-beam confocal microscopes require laser light. Minsky (the inventor of the first confocal microscope) used a zirconium arc illuminator in the functional prototype stage-scanning microscope he built in the 1950s (Minsky, 1988). Many current commercial disk-scanning confocal microscopes come only with non-laser sources because only such sources can provide the broad beam needed to simultaneously illuminate the many confocal apertures in the field of view. Using a tandem-scanning confocal microscope with a high-NA water-immersion objective, transparent ciliate protozoa such as paramecium and vorticella can easily be viewed by eye swimming in water in the confocal BSL mode. BSL images are formed using the light that is scattered by the index of refraction difference between organelles and water. By carefully adjusting the rotation speed of the aperture disk, one can view the rapidly beating cilia with stroboscopic illumination. A field of beating cilia viewed en face appears as dots (cross-sections of the cilia) slowly moving in circles. Though easily viewed by eye, this motion is difficult to capture electronically or photographically because the brain is able to extract meaning out of successive images with the slight trailing-edge blur that renders electronically captured single images meaningless.
Tandem Scanning: Basic Description The tandem-scanning mechanism consists of a symmetrical, spinning Nipkow aperture disk at the intermediate image plane of the objective. Thousands of apertures arranged in spirals both send beams to be focused on the object and sample the light returning to form the intermediate image. The double-sided optical system developed by Petran uniformly illuminates the excitation area of the disk that is to be imaged onto the object. Through a series of mirrors and beam-splitters, the image returning from the spots in the specimen is focused onto the lower surface of the opposite side of the disk where the in-focus light passes through a mirror image conjugate set of holes (Fig. 10.4, this volume). This series of
In a single-sided, disk-scanning optical system, the spinning Nipkow aperture disk is again located at the intermediate image plane of the objective but, because the beam-splitter is above the disk, the same apertures now serve as both sources and pinholes (Fig. lOA , this volume). The aperture disk is tilted and covered with "black chrome" to reduce reflections of the source from reaching the eyepiece. Furthermore, a polarizer placed in the illumination path, a quarter waveplate above the objective, and an analyzer at the eyepiece form an "anti-flex" system to further reduce the effect of disk reflections. IS Because it lacks any mirrors between the disk and the objective, the single-sided system is self-aligning. Boyde and Petran (1990) directly tested the light budget in the two types of disk confocal systems and found little difference in light efficiency, though the tandem system appeared to have better contrast.
EXPOSURE TIME AND SOURCE BRIGHTNESS In conventional microscopy, data for every point in the image is collected in parallel. This leads to a short capture time compared to any scanning process. The capture time is only limited by the time needed to fill the pixels in the CCD camera. Depending on the magnification, the fluorescent dye concentration, and the quality of the CCD itself, recording times vary from a few milliseconds to several seconds. Living cell observations with moving specimens are no problem. Figure 6.22 shows comparison pictures of cells taken with an HBO arc lamp and an LED-based source. Although the lower brightness of present LEDs required a recording time about four times longer than that needed with the HBO source, the quality of the images is comparable. In scanning microscopy, the image is formed by scanning a point or a group of points over the surface to be imaged, and this scanning process takes a finite amount of time. If a raster scan is used, the image is completed when the raster is finished, and so the confocal microscope is a sampling system in both time and space. To view specimens that move or change accurately, the scan time must be short compared to the expected rate of change. This requires not just a fast scanning system but also a light source
" While at the same time, reducing light throughput by about two thirds.
142
Chapter 6 • A. Nolte et al.
FIGURE 6.22. Fluorescent actin filaments imaged using HBO and LED sources. CA) HB0103WI2, exposure 600ms, (Bl Luxeon Star LED, I W, exposure 2 s. In the enlarged insets, one can see that some of the broadband light from the HBO gets past the excitation filter to excite the red MitoTracker Red dye as well as the Bodipy FL phallacidin, green dye, while the narrowband light from the LED does not do so. (Specimens are fixed BPAE cells stained on a Molecular Probes FluoCells prepared #1 slide. Imaged with an Axioplan 2, 63x 11.4 oil objective, Axiocam HrC, Filterset #9.)
bright enough to elicit from the specimen sufficient signal to make a usable image during the available scan time. In other words, shorter scan times need brighter sources. In practice, it has been the inability of arcs and LED-based sources to match the brightness of the laser that has prevented disk-scanning instruments from seriously challenging laser instruments for viewing low-intensity fluorescent specimens. Because the rate at which signal can be derived from a fluorescent specimen depends fundamentally on the rate at which excitation photons impinge on the imaged area, some idea of the relative merits of the two approaches can be gained by measuring this quantity. When comparing disk- and laser-scanning data rates, it is important to normalize for the size of the area illuminated on the specimen from which data are being recorded because the former collects data in parallel while the latter collects only from one point at a time. Given comparable pinhole sizes and optical efficiencies, the crucial factor for a disk-scanning microscope is the rate at which the narrow-band, excitatory radiation strikes the area of the specimen that can be imaged by a high-quality imaging detector, such as a cooled CCD. A 512 x 512 CCD operating at the total magnification needed for proper Nyquist sampling of 0.25 /.tm resolution data (i.e., 0.1 /.tm pixels at the specimen; Chapter 4, this volume), images an area about 50 x 50/.tm2 on the specimen. In 1990, the illumination systems of the commercial, doublesided disk-scanning confocal microscopes could concentrate only 2 to 3 /.tW of narrow-band light into a 50 x 50/.tm area, while the single-sided instruments could produce 6 /.t W (personal communication, V. Cejna, Technical Instruments, San Jose, CA). By contrast, the laser sources on the confocal laser-scanning microscope (CLSM) can easily deliver 100x more power (without producing
significant singlet-state fluorescence saturation!) and can consequently produce data from (and bleaching of!) the specimen at a proportionally higher rate. This improvement is only partially offset by the fact that the CCD detector is about 3 to 6 times more quantum efficient than the photomultiplier tube used in the laser systems. On the other hand, because disk-scanning instruments use many simultaneous apertures, the absolute limit on data acquisition presented by singlet-state fluorescence saturation (Chapters 2, 16, and 21, this volume) is far less of a limitation on these instruments. As a result, if higher radiance non-laser sources are developed and low-read-noise, electron-multiplier CCDs (EM-CCDs) replace the conventional CCDs now commonly used, the diskscanning approach could eventually produce even higher useful frame rates than single-beam laser instruments and do so with the same number of photons from the specimen. Another way to speed up the data rate is to use larger CCD sensors (1000 2). These would permit parallel detection of data from a larger area of the specimen while still maintaining the same pixel size. Although this strategy would increase the effective data acquisition rate by an amount proportional to the number of sensors in the detector, it would do so only at the price of viewing ever larger fields of the specimen. In other words, while it can be useful to survey larger fields in the same amount of time, it is not the same thing as increasing the source brightness or optical efficiency to permit imaging a particular cell more rapidly. As current disk-scanning systems utilizing better illumination sources and improved optics claim power levels at the specimen that are 5 to 10 times higher than those measured in 1990, the balance may soon shift to the disk scanners.
Non-laser light Sources for Three-Dimensional Microscopy • Chapter 6
FUTURE TRENDS
100
Viewed in one way, the arc lamps produce enough light right now to saturate or bleach common fluorescent dyes. The weak point of these sources is the difficulty of rapidly controlling operating parameters such as emitted intensity or spectral distribution. As these sources have been optimized over a period of many years, a quantum leap in performance is unlikely. On the other hand, arcs using different gas mixtures and electrode materials are constantly being developed and small improvements are likely to continue. The same is true for tungsten-halogen sources. Viewed in another way, 100 years after the first use of Hg arcs in microscopy, and 40 years after the advent of the LED, microscopy has a new light source with many exciting possibilities. LEDs have all the features that arc lamps lack and they will soon be efficient enough to be run on a small battery. Although their weak point is still their marginal intensity, if one looks at the LED optical efficiency line over the last few years (Fig. 6.23), one notices a very interesting trend. LED brightness is projected to increase by about a factor of 3 in the next 5 years. Efforts are under way to use different growth mechanisms to produce LED die crystals with a geometry that decreases the loss of light through internal reflection. If this effort is successful, LEDs should be able to succeed in all fluorescence applications. Improvements in the coupling between the LEDs and the microlens array and diffractive optical elements will increase the efficiency of the collection optics. Molded plastic technology is becoming ever more flexible in being formed to satisfy the requirements of this application. The brightness of a multi-color source mounted as a single flat array, such as that shown in Figure 6.21, could be increased by 3 or 4x by coupling together light from three to four planar sources, each emitting at a different wavelength, using an arrangement of dichroics and prisms similar to that used to separate R, G, and B light in a three-chip color CCD sensor. This would allow each LED color to be emitted at every location in the BFP. Three times as many dies would increase total light output by 300%. Because LEDs have no emission at either 1/2 or 2x the design wavelength, it should be possible to develop more inexpensive and efficient light-source/filter sets for particular fluorescent dyes. Somewhat farther over the horizon is the promise of organic LEDs (OLEDs). First developed in the early 1990s, OLEDs are built of organic molecules and light is generated by exciting the molecular orbitals of chromophore groups. Organic molecules
143
Fluorescenlla p
Incandescent I mp (unfiltered)
1975
1980
1985
1990
1995
2000
Time (year) FIGURE 6.23. Development of the optical efficiency of LEDs in the last 40 years.
sitting in a plastic lattice can be formed into almost any shape one can imagine. The brightness of these sources is now similar to that of conventional LEDs 20 years ago (Zhou, 2001). Although the operating lifetime of these sources is still very short, this technology is important as a potential area-display for consumer goods, such as video cellular phones, so substantial improvements may occur. Significant improvements in arc sources (-lOx) could be realized by using a modified elliptical collector (Luthjens et aI., 1990) with one of the newer Xe/I sources and by optimizing the magnification of the illumination optics to concentrate more light from the brightest part of this source into the 50 I-lm field covered by a 512 x 512 CCD. Another development that should be mentioned is the emergence of a number of companies providing stand-alone light sources for use in microscopy (Table 6.3). Although initially these tended to be fiber-optic illuminators suitable for use with dissecting microscopes, more recently sources suitable for use in high-performance epifluorescence microscopy have been offered. Besides utilizing optimized arcs in elliptical collector mirrors and high-speed filters wheels for rapidly shifting the output wavelength, they also provide fiber-optic light scrambling.
TABLE 6.3. Companies Making Stand-Alone Microscopy Light Sources Rapid Company
A Shifting?
Wavelength range, nm
Model
Type of Source
API Dolan-Jenner
Uniform light source Cold light sources
Yes No
320-700 UV-VIS
Sold only w/system Optical fiber
http://www.api.com/ http://www.dolanjenner.com
EXFO Illumination Technologies Schott
X-Cite 120PC Cold light sources
HBO I 00/2 Arcs, e.g., metal halide l20W metal halide Halogen
No No
UV-VIS VIS
Custom coupling optics Optical fiber
http://www.exfolifesciences.com http://www.illuminationtech.com
No
UV-VIS
Optical fiber
http://www.schott.com
StockerYale Sutter lnst Till Photonics Technical Video Ltd
Cold light sources Lambda DG4 Polycrome V Fiber optic light scrambler
Arcs, e.g., metal halide Halogen 175 W XBO based ISO W XBO based HBOlOOWI2
No Yes, 1.2ms Yes,400nm/ms No
VIS 300-700 320-6RO 320-800
Optical Optical Optical Optical
http://www.stockeryale.com http://www.sutter.com/ http://www.till-photonics.de http://www.technicalvideo.com
Cold light sources
Interfaces to ...
fiber fiber __ flexible fiber __ flexible fiber __ All
URL
144
Chapter 6 • A. Nolte et al.
1100 N
E
~
•
g
'II'
::
~ 10
,It
'-- 6297
f ~: :~ l 200 W EmArc™ ~'" "..:!. ::.: . \",.,,\ j'i."_'\~~' .. r .. /·. " ,..,/ --'-#... ,
c
ct!
'0
~
~~~
::: :::
#
,-
-
-
... '
... . -
.. _ - - ,
- 6247 75WXe
,
0. 1 ~'~~~~~~~~~~77~~~~~77~~~~~~
200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400
Wavelength (nm) FIGURE 6.24. Spectral distribution of a 200W EmArc, mercury-halide arc source and a 75 W xenon arc lamp (dashed line) 200 to 2400nm. (Spectra kindly provided by OriellNewport, Irvine, CA.)
In a way, the really good news is that there is a great demand for high brightness light sources in areas of the economy that have considerably more market power than microscopy. Perusal of the Internet will show a wide variety of ingenious sources responding to both the general need to get more light for fewer watts of electric power and the more specific need to keep the source size small. The digital projectors used at scientific meetings are just the early scouts of a digital revolution about to overtake the movie industry. These projectors will need bright sources. Figure 6.24 shows the performance of one of the early candidates. a 200 W enhanced metal arc source that uses a mixture of gasses to produce an order of magnitude increase in light output in the visible, compared to a short arc 75W Xe arc (USHIO, Inc. Cypress, CA). Although, at over US$2000 each, it is unlikely that many microscopists will use such sources immediately, there is every reason to think that eventually prices will come down.
ACKNOWLEDGMENTS The authors thank Nancy Fernandez and Roger Milvid at the Oriel Division of Newport, Inc., for providing us with the spectra and diagrams used in Figures 6.1, 6.4, 6.5, 6.9, 6.13, 6.17, and 6.23; Volker Haerle at OSRAM Optical Semiconductors for permission to show Figure 6.11, and Anton Moffat at Carl Zeiss Jena, GmbH, for his support.
REFERENCES Born, M., and Wolf, E., 1980, Principles of Optics, 6th ed., Pergamon Press, Oxford, England. Boyde, A., and Petritti, M., 1990, Light budgets, light and heavy losses: One-or-two sided tandem scanning (real time, direct view, confocal microscopy), J. Microsc. 160:335-342. Braun, D., and Merrin, J., 2003, Exciting brightfield fluorescence with power LEOs, Center for Studies in Physics and Biology, Rockefeller University, 1230 York Avenue, New York, NY, 10021, p. 22. Briers, J.D., 1993, Speckle fluctuations and biomedical optics: Implications and applications, Optical Eng. 32:277-421.
De Basio, R, Bright, G.R, Ernst, L.A., Waggoner, A.S., and Taylor, D.L., 1987, Five-parameter fluorescence imaging: Wound healing of living Swiss 3T3 cells, J. Cell Bioi. 105:1613-1622. Ellis, G. W., 1979, A fiber-optic phase-randomizer laser for microscopy by laser, J. Cell Bioi. 83:303a. Gerritsen H.C., van der Oord, c., Levine, Y, Munro, I., Myring, W., Shaw, D., and Rommerts, E, 1992, Synchrotron radiation as a light source in confocal microscopy of biological processes. In: Time-Resolved Spectroscopy for Biochemistry III, Proc. SPIE 1640 (J.R. Lakowicz, ed.), SPIE-The International Society for Optical Engineering Press, Bellingham, Washington, pp. 754-760. Hard, R, Zeh, R, and Allen, R.D., 1977, Phase-randomizer laser illuminator for microscopy, J. Cell Sci. 23:335. Hell, S., Witting, S., Schickfus, M.V., Wijnaendts van Resandt, RW., Hunklinger, S., Smolka, E., and Neiger, M., 1991, A confocal beam scanning white-light microscope, J. Microsc. 163:179-187. Herman, B., 1998, Fluorescence Microscopy, 2nd ed., BIOS Scientific Publishers, Oxford, 1998. Hermann, P., Maliwal, B.P., Lin, HJ., and Lakowicz, J.R., 2001, Frequencydomain fluorescence microscopy with LED as a light source, 1. Microsc. 203: 176-181. Hiraoka, Y., Sedat, J.W., and Agard, D.A., 1990, Determination of the threedimensional imaging properties of an optical microscope system: partial confocal behavior in epi-fluorescence microscopy. Biophys. J. 57:325333. Inoue, S., 1997, Videomicroscopy, The Fundamentals, 2nd ed., Plenum Press, New York. Kam, Z., Jones, M.O., Chen, H., Agard, D., and Sedat, J.W., 1993, Design and construction of an optimal illumination system for quantitative widefield multi-dimensional microscopy. Bioimaging 1:71-81. Loveland, L., 1970, Photomicrography, John Wiley and Sons, New York. Luthjens, L.H., Hom, M.L., and Vermealen, MJ.W., 1990, Improved metal compound ellipsoidal-spherical mirror condenser for xenon short-arc lamp, Rev. Sci. Inst. 61 :33-35. Minsky, M., 1988, Memoir on inventing the confocal scanning microscopy, Scanning 10: 128-138. Perduijn, A., de Krijger, S., Claessens, J., Kaito, N., Yagi, T., Hsu , S.T., Sakakibara, M., Ito, T., Okada, S., 2004, Light output feedback solution for RGB LED backlight applications. (Application note available on the Lumiled home page.) Petri'u1, M., Hadravsky, M., and Boyde, A., 1985, The tandem scanning reflected light microscope, Scanning 7 :97-108. Piller, H., 1977, Microscope Photometry, Springer-Verlag, New York, p. 253. Reynolds, G.O. , DeValis, J.B., Parrent, G.B. Jr., and Thompson, B.J., 1989, Parametric design of a conceptual high-resolution optical lithographic printer, In: The New Physical Optics Notebook, SPIE, The International Society for Optical Engineering Press, Bellingham, Washington, pp. 549-564. Schubert, E, 2003, Light-Emitting Diodes, Cambridge University Press, New York. Steen, H.B., and Sorensen, 0.1., 1993, Pulse modulation of the excitation light source boosts the sensitivity of an arc lamp-based flow cytometer, Cytometry 14:115-122. Steigerwald, D.A., Bhat, J.c., Collins, D., Fletcher, RM., Holcomb, M.O., Ludowise, MJ., Martin, P.S., Rudaz, S.L., 2002, IEEE Journal of Selected Topics in Quantum Electronics, Lumileds Lighting, San Jose, California, 8:310-320. van der Oord, C.J.R, 1992, Synchrotron radiation as a light source in confocal microscopy, Rev. Sci. Instrum. 63:632-633. Woodlee, R.L., Fuh, M-RS., Patonay, G., and Warner, I.M., 1989, Enhanced DC arc-lamp performance for spectroscopic application, Rev. Sci. Instrum. 60:3640--3642. Young, I.T., 1989, Image fidelity: Characterizing the image transfer function, Methods Cell. Bioi. 30: 1-45. Zhou, X., Pfeiffer, M., Huang, J.S., Blochwitz-Nimoth, J., Qin, D.S., Werner, A., Drechsel, J., Maennig, B., Leo, K., 2002, Low-voltage inverted transparent vacuum deposited organic light-emitting diodes using electrical doping, Appl. Phys. Lett. 81 :922-924.
7
Objective Lenses for Confocal Microscopy H. Ernst Keller
INTRODUCTION No other component of the microscope is as instrumental in det ermining the information content of an image as the objective. Th e resol ved det ail, the contrast at which this detail is presented , the depth throu gh the object from which useful information can be der ived, and the diameter of the useful field are all limited by the perfo rmance of the objective. All ot her imag ing co mpo nents, such as relay optics, Telan sys tems , tub e len ses and eye pieces, o r projectives may have so me corrective function but otherwise serve only to prese nt the im age generated by the objective to the detector in such a way that most of its informa tion content ca n be reco rde d without degradation. Whil e thi s is true for any co nve ntio nal micro scope, it is particul arly true for confocal scanning. where the objective becom es the co ndense r as well and need s to co mbine a high degree of optica l co rrec tio n with goo d throu ghput and a minimum of internal stray light or photon no ise generation. In ge nera l, the de mand s on the performance of the objective for co nfoc al sca nning are ide ntical to the needs for dem anding video mic roscopy. ph otomicroscop y, densitom et ry, ph otom etry, spec tro photo me try, and morph om et ry. However, thi s does not mean that confocal micro scop y wi ll not eventually call for special new len ses in which certai n corr ections may be sacrificed to enhance specific capabilities. In biological applications inv ol vin g living ce lls, high photon effici ency is so important that it may be worth accepting a reduction in field size and chromatic correcti on in orde r to ac hieve the highest possibl e tran smittance at a reasonable wor king distance by using a mini mum number of len s e lement s. Ano ther problem is the loss of co rrection for spherica l aber ratio n as an oil-immersio n len s is foc use d deep into an aqueo us speci me n. High num eri cal ape rture (N A) wa terimmersion objec tiv es with co rrec tio n co lla rs for cov erslip thickness variations . refract ive ind ex variatio ns in the medi um. or tem perature-dependent inde x cha nge s have become the len ses of cho ice for live-cell studies . Because of the diffi culty manu all y rotating the correc tion co llar while ac tua lly obse rving livin g specimen s, co rrec ting spherical aberratio n may require the addition of ei the r deformable mirror corr ector s or motor-driven correcti on optics mounted later in the o ptica l path (see also Chapter 20, this volume). Becau se the cr itical dem and s of light mic ros copy and co nfoca l scanning micro scopy ha ve increasingl y forced the performance of objectives to approach their theoret ical lim its. a brief refr esher on aberra tio ns, design co nce pts, materials, etc., may be in ord er. An overview of optical aberration s in refr act ive systems - both
H. Ern st Kell er . Carl Zeiss, lnc. , O ne Z ei ss
Drive, Thornwood,
inherent and induced by imp roper use of the mi croscope - and the basic performance ch aracteristics of the different generic type s of objectives will be pre sented. The basic design con cepts of microscope optics - finite ver sus infinite image distance, compensating versus fully correc ted sys tems - need to be understood to properly match optical compo nents and their properties for spec ific applications. Optical materials, ce ments , and anti-reflec tion coatings all influence the performance of the object ive. Immersion liquids, the cover slip , and the mounting medium are all part of the optical train and ca n strongly affect the qu ality of the final image. We will try to put all of this into qualitati ve perspective , part icularly as it pertain s to co nfoc al scanning. Add itio na l information on the optics fo r microscop y can be found in a rev iew paper by Inoue and Oldenbourg (1995) and in a review article on objective len s de sign by Shim izu and Takenaka (1994 ). A detailed , qu ant itative co mpariso n of the performance of different micro scope objec tives mu st be based on accepted cri teria and precisel y defined testing meth od s. Unfortunately, although the major microscop e makers have developed their own proprietary meth od s, no inde pe nde nt, fully "objec tive" test pro cedure is readily access ible to users to qu ant ify all performance data of an objective. (Juskaitis describes a sophisticated me ans for interfero metric testing of objectives in Ch apter II, this volume.) How, then , should the user of a confocal microscope j udge the pe rformance of an obj ective? Me asurement and analysi s of the real, not just theoretical , point spread function (PSF) of an objective, or better, of the complete imagin g system, is critical for threedimen sional (3D ) deconvolu tion . Ob servations of subreso lution pinh ole s in an evaporated silver or aluminum coating are adequate to j udge spherical aberratio n, astig matis m, co ma, and flatne ss in tran sm itted light but do not wo rk we ll in the epi-mode. Flu oresce nt bead s in the 0.1 J.1m range are suitable replacem ent s, but the fluorescen ce fades. Diatom s have long been a standard because of their pre cise and regul ar spacing and because the y can be viewe d in the back scattered (BSL, so metimes referred to as reflected ) light mod e or after embedme nt in fluorescent dye (Chapter 35 , this volume). For example, how do we determine, at lea st qu alit ati vely. how an image is degraded - by focu sing deep into a spec ime n or by pairing optical compon ent s that are not matched? These are all cha llenges that are not yet fully reso lved. They po int to a need for detailed testin g pro cedures that cover all aspects of the optics from so urce to detector. Still , with our ability today to ray-trace len ses for their geometr ical optical performan ce and to calculate wa ve- front aberra tion s, PSF, and inten sity ratios through the Ai ry di sk , most
N ew Yor k 10 594
Handb ook of Biolog ical Conf ocal Microscopy , Thi rd Edition, edited by Jame s B. Pawley, Springer Science+Bu siness Med ia, LLC , New York, 2006.
145
146
Chapter 7 • H.E. Keller
objectives now offered are close to diffraction limited, at least in the center of the image field. However, field size and optical performance at the periphery of the field are also especially important in beam-scanning confocal microscopy. Long-term mechanical, thermal, and chemical stability of objectives used with lasers are a function of manufacturing tolerances and materials chosen. Submicron tolerances for the centration and spacing of lens elements in sophisticated, high-power objectives call for careful, gentle treatment by the user. A minor mechanical shock may generate enough stress on a lens element to seriously reduce the objective's performance in polarized light, in differential interference contrast (DIC), or in critical confocal scanning.
1.22A 2NA or the radius of the Airy disk, where 11 sina is the NA. This pointto-point resolution for a given objective in turn determines the magnification required to enable any given detector to record all the resolvable details. Taking the spacing of the rods in the human retina as setting the limit, the total required magnification for visual observation - the so-called useful magnification - becomes about 500 to 1000 times the NA of the objective, while for digital recording we enlarge the Airy disk to 4 to 5 times the pixel dimension ofthe detector (Nyquist sampling, see Chapter 4, this volume).
Defocusing ABERRATIONS OF REFRACTIVE SYSTEMS The ideal "diffraction-limited" objective generates a 3D PSF from an infinitely small object point. A cross-section perpendicular to the optical axis through the center of the PSF is the Airy disk, as shown in Figure 7.1 (A). The diameter dAiry of the first dark ring, generated by destructively interfering, diffracted wavefronts [Fig. 7.1(B)] is
d. = 1.22A Airy 11 sina where A is the wavelength of light, 11 is the refractive index of medium between the object and the objective, and a is the halfangle of the collected rays from the object point. The Rayleigh (or Abbe) criterion sets the limit for the smallest resolvable distance d between two points at one-half this diameter or,
A
Defocusing will change the size and intensity distribution of the unit image point (Fig. 7.2). Because defocusing and depth of field are closely related, let us take a look at the 3D PSF or "image body" of the "diffraction limited" objective (Chapter I, this volume). Figure 7.3 again shows a cross-section perpendicular to the optical axis in the optimally focused image/object plane, which is, of course, the intensity distribution of the Airy disk, while Figure 7.4 represents a section along the optical axis and its intensity distribution (actually, the log of the intensity, to make it more visible). Defocusing results in alternating bright and dark spots along the axis of the Airy disk (Fig. 7.2). The extension of the central bright body along the axis is 4AJ(NA)2, but we can detect a change in the image with a defocus of only ±AJ(NA)2 (the Rayleigh/Abbe unit in the z-direction). We call this the wave-optical depth of field (Figs. 7.3 and 7.4).
B
~~_." Objektiv
Objektebene
FIGURE 7.1. (A) Airy disk and its intensity distribution. (8) Generation diagram and profile of the Airy disk or unit image.
Objective lenses for Confocal Microscopy • Chapter 7
147
.0" FIGURE 7.2. Changes in intensity distribution with focus changes.
Deviations from the "diffraction-limited" point image caused by lens aberrations can be grouped into either wavelengthindependent (monochromatic) or chromatic aberrations.
Monochromatic Aberrations Spherical Aberration This axial aberration is generated by non-spherical wavefronts produced by the objective itself or by improper use of the objective,
FIGURES 7.3 and 7.4. Horizontal (focal plane) and vertical (optical axis) cross-section through unit image body.
in particular, failure to use the correct coverglass thickness or maintain the designated tube length or the presence of substances between the objective and the focus plane having the wrong 11 (Chapter 20, this volume). Spherical aberration has the effect that paraxial rays have a different focal length from peripheral rays, and a blurring of the image body produces an asymmetrical intensity change when defocusing by ±LlZ (Fig. 7.5 and Fig. 20.3, this volume). Spherically ground and polished lenses have a shorter focal distance for peripheral rays than for paraxial rays. Spherical aberration can be optimally corrected only for accurately specified object and image distances. It can, therefore, be easily induced by improper tube length caused by introduction of optical elements into the converging beam path of finitely designed systems or by the use of improper "windows," such as non-standard coverslips or immersion oil of non-specified refractive index between object and objective. Figure 7.6 shows the changes in size and intensity distribution through the image point with increasing penetration into an object in watery medium with a planapochromat 63x, NA 1.4 oil. The effect of this induced spherical aberration on the image point needs to be considered and either corrected for or at leas t understood before confocal microscopy can be applied optimally to 3D reconstruction (see Chapter 24, this volume). More specifically, let us consider how the image quality of another objective designed for diffractionlimited performance (the Plan-Neofluar 40x, NA 1.3 oil) deteriorates when focusing lO!-tm into an aqueous medium. At the water-coverslip interface, spherical aberration is generated, which shifts the focus of peripheral rays above the optimal focus of paraxial rays [Fig. 7 .7(A,B)]. Figure 7.8 illustrates the change in intensity of a fluorescing object detected with a pinhole equal to the diameter of the Airy disk (1 Airy Unit) at 543 nm as a function of different penetration depths or focus changes, indicated in both micrometers and Rayleigh units (RE) into water. For depths of 0 !-tm, 5 !-tm, 10 !-tm, and 20!-tm, thc signal intensity at the pinhole detector drops dramatically along with deteriorating depth discrimination . Minor changes in refractive index are generated by changes in salt concentration, temperature, and differences in molecular structure. All these can induce positive or negative spherical aberration. With increasing NA , changes in the thickness or the refractive index of the "window" between the object and the objective becomes critical, particularly with "dry" objectives. In the lowpower, low-NA objective with relatively higher NA;mage,; ue where
148
Chapter 7 • H.E. Keller
FIGURE 7.5. Nonsymmetrical change in the intensity distribution with focus change above (left) and below (right) best focus in a system limited by spherical aberration.
NA
. lmageSIde
=
NAobjectside
Magnification
small changes in tube length quickly lead to inferior images. While the spherical aberration can be corrected to less-thanperceptible limits for visual observation for all types of objectives, this holds true only if all optical specifications for a given lens are fulfilled. For oil-immersion lenses with high NA, this usually means using a coverslip of 0.17 mm thickness, and 11 = 1.518 at 546 nm and 589 nm and an immersion oil with 11 = l.5180 ± 0.0004 at 546nm or 11 = l.515 at 589nm, a condition complicated by the fact that, in all materials, 11 is a function of 'A and temperature. If the exact properties of the coverslip and the oil are specified, then the manufacturer can correct spherical aberration for several values of 'A: Zeiss achromats are corrected for 2 'As, neofluars for 3 'As, and planapochromats for 4 'As.
With high NA dry or water-immersion objectives, the thickness of the coverslip, standardized throughout the industry as 0.17 mm (# l.5), is particularly important. Figure 7.9 shows the changes in the half-width of the intensity distribution curve with changes in coverslip thickness. With tolerances of ± 10 11m for top-quality coverslips, the half-width can change by more than a factor of 2. With increasing NA (>0.5), particularly with dry and water-immersion lenses, selection of coverslips for correct thickness is particularly important. Even oil-immersion lenses such as the planapochromat
Oil
Oil
Oil
t FIGURE 7.6. Change in the intensity distribution with increasing penetration into a watery medium with a planapochromat. Penetration depth from 0 to 4 flm. Note: In these plots, peak intensity has been normalized. In practice, it should decrease dramatically as the base of the intensity plot widens.
Focus with full spherical correction
Focus with spherical aberration
FIGURE 7.7. Ray diagram of an NA = 1.3 oil-immersion objective, focusing into oil (left) versus focus 10 flm in water (right).
Objective lenses for Confocal Microscopy • Chapter 7
149
Pinhole energy [%) in pinhole 0 = 0 Airy (543nm) = 20 11m
100.0
z
=
o11m
z = 5 11m 50.0
z = IOl1m
z = 20 11m
-60.0
1\ -25.0
-20.0
-15.0
-10.0
-5.0
0.0
-50.0
-40.0
-30.0
-20.0
-10.0
0.0
10.0
[n x RE J at Z
= 0 um
FIGURE 7.B. Pinhole energy measured through a 40x, NA 1.3 oil objective; as function of defocus and depth of penetration (Z) into water. As the thickness of the water layer increases from 0 and 50 ~m, the resolution is reduced by about 3x and the intensity by a similar amount.
63x, NA 1.4 perform optimally only with a coverslip thickness of 0.17 mm (Chapter 8, this volume). Fortunately, electro-mechanical micrometers capable of reading coverslip thickness to an accuracy of ± 1 f-lm are now available at relatively low cost. The spherical aberration induced by non-specified coverslip thickness leads to loss of energy at the pinhole, reduced depth discrimination, and an axial shift of the best focus. This is shown in Figure 7.10 for a diffraction-limited, water-immersion 40x, NA 1.2 objective. These examples underline the importance of maintaining the specified and computed optimal conditions for microscope objectives in order to achieve the full benefits of confocal microscopy.
FWHM [~ml
2.5 2.0
1.5 1.0
0.5 ~--~---+--~~--+---~.~
120
140 160 180 200 220 glass plate thickness [~ml
FIGURE 7.9. Changes in the half-width of the intensity distribution with changing coverslip thickness. Plan-Neofluar 63x, NA 1.2 water.
On high-NA, dry objectives or on multi-immersion objectives, eliminating induced spherical aberration requires that the correction collar be set exactly (±2 f-lm of glass-replaced-by-water at NA 1.2, see Figs. 7.9 and 20.3). From the above, it becomes obvious that a water-immersion objective is the best choice to minimize induced spherical aberration when penetrating aqueous media. Unfortunately, the collection of 3D data requires moving the specimen with respect to the lens. If this is not to produce motion of the specimen, a coverslip must be used. Experiments have shown, however, that precise correction for coverslip thickness or even the use of "Cytop" (a new coverslip material with a refractive index of 1.34, which is almost that of water, developed by Olympus) will not suffice. The refractive differences in physiological media and in biological materials still require systems that permit adjustable correction for the spherical aberration induced by the specimen itself (see Figs. 2.3-2.5, this volume). With the introduction of high-NA water-immersion objectives designed for coverglass use, some with working distances of 0.24 mm, microscopy on live cells or tissue has been greatly facilitated. Their correction collars not only allow one to compensate for variations in the coverglass thickness (after it has been measured with a micrometer or via the confocal software) but also allows one to compensate for refractive index changes due to temperature or concentration changes in the medium. Although these water-immersion objectives have nominally lower NAs than comparable oil lenses, keep in mind that the effective NA of the oil objective, when looking through water, is dictated by the refractive index of the water or of the medium of lowest refractive index between object and objective. For most aqueous specimens the water-immersion lens is clearly the preferred choice. Not all experiments permit optimal imaging conditions, and the question arises whether objectives with lower NA or with a built-in iris would not sometimes yield better results. Figure 7.11 again plots
150
Chapter 7 • H.E. Keller
Pinhole energy in pinhole 0 = 0 Airy (543nm) = 23 pm
100.0
d=0.17
50.0
[pm Water I
O.-r-----.----,-----,-----.----.-----.----.-----.-----.----. -5.0
-3.0
-4.0
-2.0
-1.0
0.0
1.0
2.0
3.0
4.0
5.0 [n x REI
Defocus relative to best focus at corresponding coverglass thickness d
FIGURE 7.10. Pinhole energy as a function of defocus and coverslip thickness (d). Objective 40x, NA 1.2 water.
Pinhole energy in pinhole 0 = 0 Airy (543nm,NA= 1.3) =20J.lm
100.0
NA=0.92
50.0
-1.0
0.0
1.0
[J.lm I
2.0
0.0
2.0
[n x RE I
O.-r-----.-----.-----.----.-----,-----,-----,-----,-----,----, -6.0
-4.0
FIGURE 7.11. Pinhole energy as a function of defocus and NA: 40x, NA 1.3 objective with iris (variable aperture) penetrating IOllm into water.
Objective lenses for Confocal Microscopy • Chapter 7
151
FIGURE 7.12. Intensity distribution in unit image with coma present: slight (left); serious (right).
detected pinhole energy as a function of focus change and NA for a 40x oil-immersion objective with variable NA. At the full NA of 1.3, penetrating 10Ilm into a watery medium reduces pinhole energy to
,
Photometer
,
, ,
56 , 52
,
,,
~.
,, ,> , , , ,, ,
, ,
,
, ,
-7 .- ~: L....'
~
n = 1.5, 4% reflection per surface reflected light d ecreases image contrast
- -- -
Integrating Sphere with Detector
~~iti-Tayer coated Objective
...................................... .~.i.ng.I~~.I.
80
o m ----1
Iris Diaphragm
I
f--
60 Monochromator
40
n = 1.8 uncoated
20
Light Source 0 4---~~--~--,--,---,--,--,--,
o
2
4
10 12 6 8 air-glass surfaces
14
16
18
FIGURE 7.26. Reflections o n surfaces of n = 1.5 (above) versus n = 1.8 (below). At n = 1.5 , the eight elements with their 16 surfaces, each reflecting 4%, result in a throughput of only 52%. At n = 1.8, the 16 uncoated surfaces would pass on ly 26%. Single-layer anti -reflection coating increases total transmission to 85%, and multi-layer coating to 94.6%. This increase in throughput, and corresponding reduction in in ternal scatter and "noise," substantially enhances the contrast of the image because it both makes bright features brighter and dark features darkcr.
A typical arrangement to measure the spectral transmission of an objective is shown in Figure 7 .27 : light source (tungsten halogen and/or xenon) with monochromator and iris diaphragm fill the objective's back-aperture. A parallel light beam equal in diameter to the pupil of the objective strikes the rear of the objective and emerges through the front lens where it is collected by an integrating sphere and measured by a PMT with photometer read-out. The result is compared to a second measurement made with the objective removed.' Table 7.4 shows the relative spectral transmission of several representative objectives that have been optimized for high transmi ssio n in the near-Uv' With confoca l fluorescence microscopy extending both towards the near UV as well as towards the IR , the
FIGURE 7.27. Setup to measure an objecti ve's spectral transmission.
spectral transmission of a total system needs to be looked at. The transmission curves for a number of objectives by Nikon, Leica, and Olympus are included in the Appendix at the end of this chapter. Keep in mind that these transmission curves are typical only for a given lens. Tolerances, particularly in the AR coating, can introduce significant variations, particularly at the cut-on and cut-off ends of the transmission curve. Depending on the application, objectives can be chosen with high throughput in the near-UV or with extended IR transmission . Having good transmission in both the near-UV and the near-IR is not possible. A number of new objectives, designed specifically for confocal microscopy, are either already available or in preparation. More detailed information on their performance and spectral throughput is best obtained directly from the manufacturers.
TABLE 7.4. Relative Transmission of Some Typical Objectives Wavelength (nm) Manufactuerer Nikon Olympus
' The high-NA rays from an oil-immersion objective can only escape from the glass into the sphere if a small, planoconvex lens is attached to the front element with oil.
Zeiss
Objective
320
350
400
500
600
Fluor 40xll.3 oil UVSLM 40x/0.9 water UVSLM l oox/I. l water Fluar 40xll.3 oil Planapo 40x/I.2 water Achropam lOOx/I.O water
16 %
66% 56% 60% 79% 54% 605
80%
90% 88% 90% 95 % 89% 94 %
9 1%
29% 20%
88% 86% 90%
99% 92 % 90%
160
Chapter 7 • H.E. Keller
CONCLUSION Many modem microscope objectives are well suited for confocal scanning as long as they are used within their design specifications. With the ever-wider use of confocal scanning microscopy for 3D live-cell imaging and with new microscope techniques emerging, new and improved objectives continue to appear. Often they are optimized for a specific application. This trend will continue. Just a few examples are the new highest-NA 60x/1.45 and 100xll.45 oil-immersion lenses or the special objectives Olympus introduced for epi-illuminated total internal reflection fluorescence (TIRF) microscopy with an NA of 1.45 or 1.65. It follows from the discussion above that the latter NA requires a special immersion medium and rather costly high-index coverglasses. In addition to the call for objectives with higher NAs, longer working distances, and maximum transmission in the visible and UV, this may be the place to "dream" about other exotic lens features. We might imagine a tunable objective, a lens whose chromatic correction can be "tuned" to the specific excitation and emission wavelengths actually in use to produce the best image quality with only a few elements and high transmission. If such tunability is impossible, perhaps special achromatic lenses will be designed for use with specific fluorophorellaser combinations, such as the "blue" objectives recently offered by Leica. An even more exotic dream (included in the second edition of this volume) for a substantial increase in signal intensity realized by using two, matched lenses above and below the specimen, has now been commercially realized in the Leica 4Pi microscope (Chapter 30, this volume). On optically homogeneous specimens, equipping the "far" objective with a reflector to return the spot into itself, might not only increase photon efficiency but also might permit combining reflected and transmitted confocal scanning. As mentioned earlier, two-photon excitation, will probably also spawn the development of special objectives, not only in terms of their spectral characteristics but with spherical and chromatic correction optimized for the IR, such as the recent "IR" objectives from Olympus. For the study of live organisms, one needs water-immersion objectives that combine the long working distance needed for deep
penetration, with the highest possible NA for better resolution and improved photon collection efficiency. These are just a few thoughts on possible future developments in this area. No doubt, specialists in optical design will quickly return us to reality and make us do with less than our dreams.
ACKNOWLEDGMENTS Many thanks to a number of scientists at Carl Zeiss, Germany, for their help in preparing this paper and for many of the illustrations. Special thanks to Mr. Franz Muchel, head of the mathematics group at Carl Zeiss, whose paper (Zeiss Information #100, 1/89, 20-27) on ICS optics has been particularly helpful, to Dr. Bernd Faltermeier for much support, and to Willi Ulrich, who provided many of the computer simulations and other valuable suggestions.
REFERENCES Inoue, S., and Oldenbourg, R., 1995, Optical instruments: Microscopes, In: Handbook of Optics (M. Bass, ed.), 2nd ed., McGraw-Hill, New York, pp. 17.1-17.52. Boyde, A., and Jones, S.S., 1995, Mapping and measuring surfaces using reflection confocal microscopy, In: Handbook of Biological Confocal Microscopy, 2nd ed., Plenum, New York, pp. 255-266. Maly, M., and Boyde, A., 1994, Real-time stereoscopic confocal reflection confocal microscopy using objective lenses with linear longitudinal chromatic dispersion, Scanning 16:187-193. Pawley, J.B., Amos, W.B., Dixon, A., and Brelje, T.e., 1993, Simultaneous, non-interfering collection of optimal fluorescent and backscattered light signals on the MRC 500/600. In: Proceedings olthe 51st Annual Meeting of the Microscopy Society of America, San Francisco Press Inc., San Francisco, pp. 156-157. Petrall, W.M., Cavanaugh, H.D., and Jester, J.Y., 1993, Three dimensional imaging of corneal cells using in vivo confocal microscopy, J. Microsc. 170:213-219. Shimizu, Y., and Takenaka, H., 1994, Microscope Objective Design in Advances in Optical and Electron Microscopy (e. Shepard and T. Mulvey, eds.) Academic Press, San Diego, pp. 249-334.
APPENDIX: LIGHT TRANSMISSION SPECIFICATIONS FOR A NUMBER OF MODERN OBJECTIVES MADE BY DIFFERENT MANUFACTURERS.4
leica objectives Objective HC PLAPO 1OX/0.40CS APO L 20X/0.50W UVI HC PL APO 20x/0.70, multi-imm, collar HCX PL APO 63x/1.40-0.60 oil HCX PL APO 63x/1.30 glycerin collar HCX PL APO 63x/1.2 W Corr
Mag
NA
WD(Il)
T(350)%
T(546)%
T(900)%
10 20 20 63 63 63
0.40 0.50 0.70 1.40 1.30 1.20
2200 3500 260 100 280 220
40 60 40 20 18 16
90 91 87 82 87 85
75 75 71 63 64 65
Blue
Immersion dry water water/glycerol/oil oil glycerol: water, 80 :20 water
Blue = Blue variant available with color correction optimized in blue range for UV applications and GFPs with 405 or 430 excitation
4
The specifications printed in this chapter are as supplied to the author by the manufacturers at the time that this book went to press (May 2005). The author and the editor have endeavored to insure that the printed versions of these specifications accurately reflect the reports that they received or that were available on the Internet. We make no other claim and readers are advised that improvements are to be expected and will doubtless be announced on the Internet.
Objective lenses for Confocal Microscopy • Chapter 7
161
Carl Zeiss objectives 100 90
'if. 60 c:
o
50
--
,-
~
!It!oooo..
~
..,
70
'E en
~
~ V; ~ -""'" Ill( ,IY
80
'ill
...
,r ...
-
~~
~
~
20 10
o
,
~~
"
11111
"
rIJJ
300
400
500
600
700
800
40xll ,2 C-Apo 1OOxll .45 alpha-Plan-Fluar 1OOxll .4 Plan-Apo
--
~~ ~ ~ -... .....
....--
~ ,~ ~ ~
c: 40
30
--
25x10.9 LCI Plan-Neofluar
- - 63x1l .3 Objective LCI Plan-Neofluar 63x1l ,2 C - Apo
-- ~ -
~ """~..... ~
"- r=-..
--
"
.","
-
10xlO,50 Objective Fluar 10xlO.3 EC Plan-Neolluar
65x11 .4 Plan-Apo
--
65x11 .0 Objective W Plan-Apo
900 1000 1100 1200 1300 1400 1500 A (nm)
Nikon objectives Objectives
N,A,
Coverglass (mm)
Working Distance (mm)
Trans % @350nm*
Trans % @550 nm*
Tran s % @900nm*
CF I PLAN ACHROMAT 100x OIL CFI PLAN A PO 20x CF r PLAN APO 40x w/collar. CFI PLAN APO 60x A OIL IR CFI PLAN APO 60x WI w/collar CFI PLAN APO VC 60x WI CFI PLAN FLUOR 20x mult-imm CFI PLAN FLUOR 40x CFI PLAN FLUOR 100x OIL CF I SUPER FLUOR 40x w/collar CFI SUPER FLUOR 40x OIL CFI W FLUOR 60x dipping CFI PLAN 100x dipping CFI Plan Apo TIRF 60xH w/collar CFI Plan Apo TIRF 100xH
1.25 0.75 0.95 1.40 1.20 1.20 0,75 0,75 1.30 0,90 1.30 1.00 1.10 1.45 1.45
0,17 0,17 0,17 0,17 0.15- 0.18 0,13-0.19 0,17 0.17 0,17 0,17 0,17 n/a n/a 0.17 0.17
0,20 1.00 .12- ,16 0.21 0,22 0.27 ,033-,035 0.72 0.20 0.30 0.22 2.00 2.50 0.13 0,13
B E E E E E B A B B B B E E E
A A A A A A A A A A A A A A A
A B B B B B A A A B A A A B B
Key : For transmission percentages: A = > 7 1%, B = 51 - 70%, C = 31 - 50%, 0
= 16- 30%,
Olympus objectives
100~------------------------~~----~----------r-------------------~ - UPLFLN40XD. NA1 .3. WD 0.2
90
1------------::1l====iE:;:;::=---------4
-
UPLSAP060XW. NA1 ,2, WD 0.28,
I ------------~.J~~~~~~~~~~~~~----------~ --.. -
80 I-
-PLAPON60XD, NA1.42, WD 0.15 CG
- UPLSAP0100XD, NA1A . WD 0,12
;,i 60 .I---------;~~~--~--------=:-------:::::-~~:_..--"~----t -
LUMPLFL40XWYP2, NAO,8, WD
U
c:
o
'ill
50
'E en c: 40
~
-
LUMPLFL60XWYP2, NAO.9, WD 2 LUMPLFLl OOXW, NA 1, WO 1,5
~----~~----~--------------------------------~~
30 ~--~~--~~--------------------------------__i
- - XLUMPLFL20XW, NAO.95, WD 2 XLUMPLFL20XWIR-SP, NAO.95,
20+--~~--~---------------------~
WD2 LUMFL60XW, NA1 .1, WD 1.5, CG
_ 10 ·~~~--~--------------------------------------~
f!
PLAPON80XOTFFM, NA1 .45, WD
o ~~--~__--------------------------------------~~--0~.-1---------~ 325
350
375
400
500
600
700
Wavelength (nm)
800
900
1000
8
The Contrast Formation in Optical Microscopy Pi ng-Ch in Cheng
INTRODUCTION In any for m of microscop y, one need s not only an imaging sys tem with enough resolution to deli neate the fine details of the spec imen but also a suitable contrast mec hanism by which to "se e" the shape of the structures of interest. Contrast is the difference between the signal in one pixel and that in another that conveys to the viewer information about the shape of the specimen. It is the difference between a blank screen and an imag e. In photographi c terms, contras t is the change in brightn ess of a negative or print. In other words, co ntras t is the differen ce in sig nal strength between vario us parts of an imag e or between details of interest and "bac kground" (see also Chapt er 4, this volume) . The contrast (y) is proportional to the intensity difference (Ill) between two ima ge areas, divided by the average image brigh tness i.
M y=-
I
In optical microscopy, co ntras t der ives from differenc es in the way the vario us sub-vo lumes of the speci men (voxe ls) interact with the illumination. Thi s interaction may include linear and nonlinear abso rption, single- and mult iple-photon fluorescence, Ram an em ission , fluore sce nce spectral shift, fluorescence lifetime, refraction , reflection, pha se shift, scattering, changes in polarization , harmonic generation, etc. A co ntras t mechanism can thus be co nsidered to be a spe cial " filter" function by which specific types of spatial or temporal signals are selected to form a two dimen sional (2D) or three-dimensional (3D) image. Th is chapter will provide an introduction to the contrast characteristics of those modalities that have been well inve stigated in co nfoca l and nonlinear microscop y, including fluore scence and sca tteri ng co ntrast. It will also co nsi der the del eterious influence on the co nfoca l and multi-ph oton image of the absorptive, refractive, and reflective propert ies of struc tures that are bet ween the plane-o f-focus and the objec tive lens, of these topi cs are relevant to the operation of all epi-illuminated micro scopes and some also app ly to sig nals dete cted in transm ission , particularly as used to de tect seco nd harmonic (SHG) and third harmonic (THG) signals. In addition, this chapter will introdu ce ways in which the co ntrast present in the raw data from the microscope can be digitally modi fied before bein g presen ted in the final ima ge. At various stag es in this sequence, the signal may be ratioed, filtered , and co rrupt ed and the contras t reduced by the addition of noise. In all these areas, this chapter serves as an introduction to other chapters in which indiv idual co ntrast mech ani sm s are discussed in more de pth.
Image contrast arises fro m the interactio n of an incide nt light beam with the specimen. Various physical and digital "filters" ca n be used to select spec ific signals. For example, one ca n discrim inate specific wavelengths using dichroic beam-splitters and barr ier filters; the effective numerical aperture (NA) of the objec tive len s ca n produce topographic co ntras t from the geometric shape of spec imen surfaces; polarized light can be used to obtain co ntras t ca used by specimen bire frin genc e; and fluorescence signals from ion- specific dyes in two different spec tral channels can be ratioed to detect the concentration of ions such as [Ca'"] and [H+J. Th e co ntrast that forms a microscopic image is determined by the number of physical, chemical, and biol ogic al phenom ena. Cont rast mechani sms ca n be subdivi ded into (a) opti cal cont rast, (b) geo metric cont rast, (c) biological and chemica l co ntras t, and (d) synthetic contrast. Fro m the point of view of the speci men, the co ntras t mech ani sm can be intrin sic or ext rin sic in nature. Although each of these co ntras t mechani sm s will be discussed separa tely, it is not uncommon for more than one to be act ive at the sa me time and care must be taken to choose experime ntal parameters that emphasize the co ntras t that highli ght s the most informative of these interact ions. Th e interaction of an incident light beam with a sample is a co mplex eve nt. Figure 8. 1 shows a simplified version of such an inter act ion as well as some of the effects produced by the voxels above and below the voxe l bei ng sampled. These interacti on s give rise to the optical phenom ena and the photochemical and biochemical effects that pro vide the bases of all the contrast. When a beam of light with intensity 10 is incident on a specimen, a number of phy sical phe nomena may occur. Th ese include the scattering of light due to Rayleigh, Mie, and Raman scatteri ng (I Sj (, I s_M, and Is_Raman)' Rayleigh scattering is caused by interactions with very sm all particle s in the specimen (from the size of molecules up to - 10% of the wavelength ) and its strength (IS_R) is stro ngly wav elength depend ent . It also depends on direct ion : sca ttering at right angles to the illumination is only half the forwa rd inten sity (I u). Rayleigh scattering is elastic scattering becau se the scattered photon s have the same energy as the incident photons. By contrast, scattering in which the scattered photons have either a slightly higher or lower photon energy is ca lled Ram an sca ttering (IS_Raman)' Thi s energy change usuall y involves eit her the exc itation of some vibra tional mode of the molecule (giving the sca ttered photon a lower energy), or the scattering of the photon off an excited vibrational state of a molecule (which add s its vibr ational ene rgy to the incident photon ). Cell organelles and other refracti ve structures larger than the wave length of the illuminat ion co ntribute to Mie sca tteri ng (I S,M)'
P.c. Cheng . State University of N ew York at Buffalo, Buffalo, N ew Yo rk 14 2 6 0 and National University of Singapo re, Singapore 162
Ha ndbook of Biologi ca l Confoca l M icro scop y , Thi rd Editio n, ed ited by J ames B . Pawley, Springer Science-B usiness M edi a, LLC , New York , 2006.
The Contrast Formation in Optical Microscopy • Chapter 8
Mie scattering has a sharper and more intense forward lobe for larger particles than Rayleigh scattering (Fig. 8.2), is not wavelength dependent and is responsible for the almost white appearance of adipose and brain tissue. Such a scattering center can act as a secondary light source within the specimen. In addition to these scattering events, refractive index (RI or 11) differences between various biological structures can cause the incident ray to deviate from its original path, producing defocusing or sampling errors (i.e., imaging the signal from the "wrong" voxel). Reflection occurs at any interface separating regions of different RI. The incident light can also be attenuated by absorption. Photon absorption occurs when the quantum energy of the photon matches the energy gap between the initial and final states of some electron in the specimen. If no pair of energy states exists such that the photon energy can elevate an electron from a lower to an upper state, then the matter is transparent to this radiation. In the absorption process, some of the absorbed energy can be re-emitted in the form of a fluorescent or phosphorescent photon. Biological molecules have specific absorption characteristics and the low effective absorbance of most tissue provides a relatively transparent " window" in the visible and near-infrared range (NIR). This window provides the working spectrum for all the optical microscopy (see Figs. 21.1 and 40.2, this volume). When the
163
FIGURE 8.2. Rayleigh scattering (left) and Mie scattering (right). In Mie scattering, the forward lobe becomes larger as particles become larger. This is the reaso n that dark-field illumination produces more signal in transmission than in ep i mode.
incident light intensity is very high, nonlinear optical phenomena such as nonlinear absorption, multi-photon fluorescence, and harmonic generation of incident light become prevalent (Fig. 8. l) . It is well known that different RIs can be associated with different crystallographic orientations of crystaJline materials. For example, calcite crystals have indices of refraction for the 0- and e-rays of 1.6584 and 1.4864, respectively. Mineral crystals showing two distinct indices of refraction are referred to as birefri ngent materials. Birefringence has to do with anisotropy in the binding forces between the atoms forming a crystal. A large number of quasi-crystalline biological materials also exhibit birefringence (Fig. 8.3).
SOURCES OF CONTRAST
Absorption Contrast If the light-specimen interaction is predominantly photon absorption (as it is in the prepared tissue sections, stained with absorbing dyes), and the specimen has a uniform thickness [Fig. 8.4(A)], and if the incident light is 10 , then the transmitted intensity, II and 12 through structures I and 2 in the voxel being sampled is: FIGURE 8.1. Interaction of light with a voxel of the specimen. The hypothetical specimen consists of four layers with different Rls (n2 , n3 , n4, n5) and the surface of the top layer is slanted with respect to the optical axis. The illuminating beam (10) is refracted by the surface of the top layer (pink) and subsequent layers with the result that the focal spot is not on the original optical axis (red line). Various signals, including flu orescence (lFLU) , scattering (Is), reflection (lR, n), Raman (TR,",,"), second harmonic generation in both forward and backward direction (IB.SHG and h .SHal, and third harmonic generation Un,,;) may be produced when conditions are favorable. The pink spot at the focus indicates the volume where nonlinear phenomena prevail at high illuminati on intensity. Transmission intensity, 1,;,,,.
and /2=
/oe-~r
where J.l.I and J.l.2 are the absorption coefficients of structures 1 and 2, respectively, and x is the length of the absorption path (the thickness of the voxel in this case). The contrast due to absorption (Yab.,)
164
Chapter 8 • P.-c. Cheng
A
B
10
10
111
1 1
FIGURE 8.4. Hypothetical thin (A) and thick (B) specimens composed of a light-absorbing matrix with an absorption coefficient of III and containing small structures with an absorption coefficient of 112. The transmission intensities of the matrix and structure are I" 12 , and 13 , as marked.
However, in real specimens, the structure of interest is generally much smaller than the thickness of the specimen and therefore, it resembles the situation in Figure 8.4(B), the effective absorption coefficient flT for a given light path is:
a term that describes an absorbing specimen composed of m different subunits, each having an absorption coefficient (fl;), where flT is the absorption coefficient associated with the light attenuation along the specific light path. For a sample with thickness x, and the effective absorption coefficients of an area of interest (flTl and flT2), the absorption contrast (Yabs) is:
Y
abs
= b..! = II l -=.12 =1 (e- flTjX _e-flT2X)/I I OB
n
L.Ioe-fl,x
where
I = ~k-=,]~_ _
I Band n is the number of n n pixels in the image. The absorption contrast (Yab.,) between pixel and 2 shown in Figure 8.4(B) is:
The Rose criterion (Rose, 1948) relates contrast and noise to visibility. It states that, to be visible, the contrast between a feature and its surroundings must be 5 times the noise in the surroundings [signal-to-noise ratio (SIN) >5: 1], e(fl2-~j)X
2:: 5
In5 In5 f.l 2 -fl l 2::x
(fl 2 -fl])x 2::
FIGURE 8.3. Birefringence of artificially reconstituted collagen fibers. (Upper) A retardance image of collagen fibers. (Middle) Orientation of the collagen fibers indicated by small arrows. (Lower) Orientation coded as color. [Images courtesy of Hanry Yu, Department of Physiology, National University of Singapore and made using a Polscope (Cambridge Research Inc., Cambridge, MA).]
Apart from the fact that the optical transfer function implies that small features will be rendered in the image with less contrast than large ones and that the Poisson noise associated with recording any image adds uncertainty to the measurement of contrast, large features will also have low contrast if the tissue is only lightly stained, and small, intensely stained structures will have less contrast when visualized inside a thick, stained region of the sample. Finally, unstained biological tissues contain few molecules that absorb in the visible and our ability to detect small changes in
The Contrast Formation in Optical Microscopy • Chapter 8
FIGURE 8.5. Nonlinear absorption of the upconverting dye, APSS, and of an ethanol extract of yellow petal of Canna, as a function of excitation light intensity.
8rr================='-~ ,. Canna yellow
I-APssl
1.0
7 6
~0.8
I
ct)J:
+ EOX(2)
E(t)2
+ EOX(3)
E(t)3
+ ...
== p(l)(t) + p(2)(t) + p(3\t) + ... where X(l) is the linear susceptibility, i 2 ) is the second-order nonlinear susceptibility, and X(3) is the third-order nonlinear susceptibility. The effective nonlinear absorption cross-section /-l(lO)eff is a function of the incident intensity. The nonlinear transmission intensity,
FIGURE 8.6. (Top) When the objective is focused into a particle-filled clear matrix, the scattered and reflected rays from the structure of interest are modulated by the particles located between the objective and focal plane. (Bottom) Diagram of self-shadowing.
r/'/~ ,./
o 100 200 300 400 500 600 700 800
E(t)
/10
%3
0.0
EoX(l)
A
///~•
"S 4
.0
p(t) =
y = 0.05 +0.96x y=0.05+0.96x-1.37'10'-2'x'2// R'2 = 99.9%, SD = 0.04 /
8. 5
.0
130fps Full serial and paralic I (vertical) 100% 220,000e 12 bits @ 10MHz >15 bits @ I MHz 14116 bits Masked w/full frame transfer; C-mount
Back Andor E2V CCO 60 128 x 128 24[.lm 3.1 x 3.1 mm 92% @ 575nm 10, 5, 3, I MHz -90°C xlOOO 0.004e/p/s 60e @ 10MHz 500fps > 1300 fps Full serial and parallel (vertical) 100% 200,000e 12 bits @ 10MHz >15 bits @ IMHz 14116 bits Masked w/full frame transfer; C-mount
Front Andor TI TC285 1000 x 1000 8[.lm 8.0 x 8.0mm >65% @ 600nm 35,27,13,5MHz -90°C x 1000 0.001 e/p/s 25e @ 35MHz 31 fps >100fps Full serial and parallel (vertical) 100% 40,000e 12 bits @ 35MHz 13 bits @ 5MHz 14 bits Masked w/full frame transfer; C-mount
"EM-CCO cameras are sold by several companies including Andor (shown here), Hamamatsu, Roper, PCO, and others. Many use identical chips from E2Y or Texas Instruments, however, other properties such as chip cooling temperature, vacuum and sealing, amplifier, hardware (c.g., TTL integration), software control and integration, internal shutters, etc., can vary considerably between manufactures and can strongly atlect camera and versatility. Data kindly provided by Colin Coates at Andor Tech. (Belfast, Northern Ireland). All specifications based on manufacturers claims. "This is a popular high readout speed, scientific CCO and chosen as a respective point of reference. There are hundreds of CCOs to choose from and specifications will vary considerably and are occasionally optimistic. 'Cameras with IS [.lm pixels will require the use of a 2x phototube to produce Nyquist-sampled images with 40x and 60x high resolution objectives. d Manufacture specifications; note often some of this bit depth is empty and the true dynamic range. which is determined by the ratio of the full well depth over the noise floor of the camera, is considerably smaller (e.g., if in the Orca ER 18,000e/Se = 2250, while 2" = 4096).
end of the spectrum. However, as good red performance often goes along with high dark current unless the photocathode is cooled, many intensifiers use the S-20 photocathode material commonly used in PMTs and having a QE at least 2x lower. In addition, as only about 50% of the photoelectrons reaching the micro-channel plate (MCP) are actually multiplied at all, and as the amount by which each of the remainder are actually amplified varies greatly (creating multiplicative noise), the effective QE is about half that of the photocathode itself. Finally, because both the MCP and the process of accelerating the charge onto the phosphor to make light, involve loss of spatial precision, the resolution of the entire camera is substantially less than would be indicated by the array size of the CCD. On the other hand, the intensified CCO is the only common image sensor that can be gated on and off in nanoseconds,21 permitting the extremely short acquisition bursts needed in specialty applications such as fluorescent lifetime imaging (FUM). Intensifier Advantages (+) and Disadvantages (-): (+) Noise level very low; able to "see" signal from a single
photoelectron.
21
The camera readout time is much longer (e.g., < lOOfps; Kindzelskii and Petty, 2003).
(+) Fast gating possible. (±) Effective QE of 10% to 25%.
(-) Pixel blooming of bright signals can occur. (-) Scintillation or hot pixels can occur. (-) Photocathode can be permanently damaged by brief overexposurc to light.
On-Chip Electron Multiplying Charge-Coupled Device Recently, the problem of increasing the signal from a single photoelectron above the noise level of the CCO-read amplifier has led to the perfection of on-chip electron multiplication (EM). Although high-gain (I O~x), avalanche electron multiplication in a reverse-biased semiconductor photodiode has been used for some time to turn a single photoelectron into a current pulse large enough to be easily measured by electronics, efforts to apply this technique to the output of a CCD failed until recently. The main distinction between an EM-CCO and a normal, scientific CCD is the addition of a second horizontal register, called the gain register, between the chip and the read amplifier (see Fig. 10.11, top). In the EM gain register, a higher voltage on one of the three sets of the charge-transfer electrodes sets up electric fields to produce, not the high gain impact multiplication found in the avalanche photodiode, but very low gain. 22 By then repeating the
234
Chapter 10 • D. Toomre and J.B. Pawley
•ii • 1I11-G:a=in~I~0uIpl" .
:unplilie
-
-BV
80 70
-
0
c:
CD
'13
60
IE W
50
E :::J
40
CIS
30
C
:::J
0
VP
rately, the spatial resolution (MTF) of the EM-CCO is identical to that of the same CCO chip with the electron-multiplier turned off. There is, of course, a price to pay and the price is increased multiplicative noise. While in a normal CCO, each photoelectron is counted the same, because of the statistical rules governing the impact ionization process in the EM-CCO, some electrons are amplified much more than others. As the noise term describing this process is of the same form as Poisson noise, and as such terms are "added" as the square root of the sum of the squares, the overall effect is to increase the "Poisson" noise level to lAx what it should be based on the number of photoelectrons actually produced. As the only way to reduce Poisson noise lAx is to count twice as many photoelectrons, it is perhaps easiest to think of the EM-CCO as having essentially no read noise but operating as though the effective QE were only 50% as high as it would be were the same CCD used without the electron multiplier. This fact must be remembered when viewing raw EM-CCO QE specifications.
Electron Multiplication Charge-Coupled Devices and Disk Scanners 24
I CY3 ,
~ ICYSI
20 10 0 200
300
400
500 600 700 800 Wavelength (nm)
900
1000 1100
FIGURE 10.11. (Top) Anatomy of an electron multiplying CCD (EM-CCD) chip. Illumination of the imaging register (orange). read register (pink), and horizontal, readout register is to the same as any frame-transfer CCD chip. The main addition is an electron multiplying (EM) gain register (in red) before the output amplifier. (Bottom) Raw, spectral response curves of backilluminated visible (BV) and front-illuminated visible (FV) EM-CCD chips. The emission wavelengths of some common dyes are indicated. (Data kindly provided by Cnlin Coates at Andor Tech.) When these chips are used with the EM register activated, the effective QE is only 50% of that shown.
process 500 or 1000 times over a line of many pixels, a gain of more than 1000 can be obtained. This makes the signal from even one photoelectron easily visible above the noise of the CCO read amplifier, even one operating at a very high read speed.23 With read noise effectively absent, the main remaining noise sources are dark current and clock-induced charge (CrC). Oark current can be substantially reduced by cooling the chip with a multi-stage Peltier cooler. At -80°C spurious dark current and crc signals are reduced to about one count in 250 pixels. crc is caused when lattice electrons are pulled into the valence band by the charge transfer process. Unmeasurable before the advent of the EM-CCO, crc is now an important noise source that can be reduced somewhat by carefully shaping the charge transfer control pulses or made slightly worse by using back-thinned chips. Amplifying the signal "on-chip" is appealing as it reduces the number of components and one can take advantage of the good QE of the CCO. As the signal from each pixel is amplified sepa-
"The gain depends on the temperature and the exact voltage applied to a special set of charge transfer electrodes in the gain register, but is commonly about I %/transfer, that is, a charge packet of 100 electrons would, on average, become 101 electrons after one transfer, or again on average, a single electron would become two electrons after about 100 transfers. '-' The new chips will operate up to 35 MHz or 140 fps for a 512 x 512 array.
How do EM-CCOs improve disk-scanning confocal imaging? Good resolution and QE are important, but single-photon sensitivity means that the EM-CCO can detect very low signals when a normal CCO would only read noise. This is important not only because disk scanners without microlenses are light starved (e.g., it is like using a 95%-98% neutral density filter on your WF fluorescence system), but also because in a selectively stained specimen, by far the most common voxel intensity is zero and the EM-CCO measures zero very wel1. 25 Figure 10.12 shows that even using a disk scanner and a backthinned 90% primary QE EM-CCD, the exposure time for adequate SIN is significant (150ms). Were a normal high quality CCO to be used instead, -3- to 5-fold longer exposures would be needed for similar SIN. Even if one has enough light to use a normal CCO, using a more sensitive detector allows one to use less excitation and produce less photobleaching. Currently, EM-CCO chips are manufactured by E2V (Enfield, UK) and Texas Instruments (Houston, TX) and are used in cameras from several companies: Andor, Hamamatsu, Roper, PCO, Red Shirt, and others. Several different EM-CCO chips exist and one should match the QE, pixel clock rate, pixel size, well depth, and total chip area to the biological process being investigated (see Table 10.3). For fast imaging, a smaller chip (e.g., 256 x 256 pixels) will provide increased frame rate. At present, all the available chips use full-frame transfer. While good for QE, the fact that the charge pattern moves past the image as it proceeds to the read register can cause some streaking and moire effects with the scan lines of disk scanners even when fast vertical transfer is used (Chong et al., 2004). This can be avoided ifEM-CCO chips incorporating interline transfer appear because such chips permit electronic shuttering at very high speed. However, unless such chips are also fitted with micro-lens arrays, these will pay a price in fillfactor and at least exhibit 2x lower effective QE. Our tests of the Olympus OSU disk-scanning system have shown that EM-CCOs can make the difference between getting an acceptable image and seeing nothing because the signal level is below the noise threshold of our normal CCO. In addition, it 24 25
See discussion at http://www.emccd.com. However, because of the reduction in effective QE the normal CCD begins to outperform the EM-CCD when the signal level gets much above about 100 photons/pixel. The exact crossover point depends on the read speed as this seriously changes the read noise of conventional CCDs.
Disk-Scanning Confocal Microscopy • Chapter 10
235
identically), there is considerably less background fluorescence with the scanning disk in place.
Fast 30/40 Imaging
FIGURE 10.12. Disk-scanning slit confocal improves optical sectioning and SIN compared to epi-fluorescent, but requires longer exposure times. Lamprey neurons were microinjected with fluorescent-tagged phalloidin to label axons (ring-like structures). A single 15 ms exposure epi-fluorescent image was taken using an Olympus BX-Sl microscope (1.1 NA 60x water-immersion objective) and an Andor 887, back-illuminated EM-CCD camera (top). After inserting the DSU disk #3 (10% transmission) a ISOms image was acquired (same gain, bottom). In both images. the same number of photons struck the specimen. (Images kindly provided by Dr. Jennifer Morgan, Yale University Medical School.)
increased the number of high-quality images that could be acquired before photobleaching. 26 EM-CCD Advantages (+) and Disadvantages (-): (+) Like CCOs, EM-CCDs are photon efficient (good QE), have good contrast (MTF), are mechanically robust, and are not harmed by exposure to bright light. (+) They have single-photon sensitivity. (+) Ergo, they can greatly improve the speed or number of images acquired before photobleaching. (-) No fast gating. (-) Higher cost than normal CCO (roughly equivalent to an intensified CCO).
APPLICATIONS AND EXAMPLES OF CONFOCAL DISK-SCANNING MICROSCOPES Comparison with Epi-Fluorescence Imaging As discussed above, a simple disk-scanning confocal should have better optical sectioning but lower light throughput than a WF epifluorescence microscope. To demonstrate this we took images of the same specimen with both an epi-fluorescence setup and with an Olympus OSU spinning-slit system (10% transmission) using an Andor 512 x 512 BI EM-CCO camera. Ten times longer exposures were used with the disk in to compensate for reduced illumination striking the specimen. As seen in Figure 10.12 (processed
The micro-lens-assisted spinning disks in the Yokogawa design transmit more illumination to the specimen, increasing the maximum imaging speed. Even with a traditional CCO such as the Hamamatsu Orca ER, one can record high resolution multi-color confocal stacks, as demonstrated in Figure 10.13. In -20s we acquired a 3D stack with 200 sections; a 3D reconstruction is shown in the lower panel. Using a single-beam confocal such as the Zeiss LSM51O, acquisition of a data stack of similar size and quality would take -5 min+. In the xz -views. one can see that z-resolution appears roughly on par with that of a good confocal LSM. The ability to acquire confocal images rapidly allows fast 40 imaging. An example of such an application is shown in Figure 10.14 where 3D stacks of -20 layers were each acquired in under 2.5 s and repeated over time. The scientific advantage of 40 imaging is that one gets a complete view of the process. In this example, c1athrin dynamics can be seen on the upper cell surface, near the perinuclear Golgi region and also at the bottom of the cell; areas where dynamic changes occur are circled. The lower panel shows a single. 3D time point from a 40 stack of a double-labeled cell projected in space at two angles. As it is hard to convey in a static medium (this book) the visual power and the high level of biological contextual information available from looking at a 3D image over time from any angle. A movie made from this 3D data stack will be available on the Web site associated with this book http://www.springer.comJO-387-2592I-X. The challenge of how best to process and visualize large datasets will likely be an increasingly significant hurdle. It is worth commenting that while photobleaching is still a real limiting factor with a normal CCO, ultrasensitive cameras allow us to push this envelope as we enter the realm of the real-time 40 confocal imaging.
Blazingly Fast Confocal Imaging Some biological processes can only be revealed (and understood) with high-speed imaging. Calcium sparks, flashes, and waves are examples. The 3D diffusion is so rapid that even alms exposure may not be fast enough and the process is definitely obscured in real-time (30fps) or slower imaging. 27 Optical sectioning is needed both to improve SIN and also to sample a slice through the middle of a cell because such an image is less subject to volumetric changes in signal than is a normal epi-fluorescence image. Maximum image acquisition speed is determined by four factors: the excitation power delivered to the sample, photophysical limits such as saturation, the disk scan speed, and the camera pixel clock and sensitivity. Because the read noise of a normal CCD rises rapidly with read speed while that of the EM-CCO does not, an ultrasensitive camera is essential. In Figure 10.15, the combination of a Yokogawa disk-scanning confocal and EM-CCO camera has permitted -II Ofps imaging of calcium sparks (top panel) or calcium waves induced by electrophysiological depolarization (bottom). These images are truly just the start of seeing life in the (ultra)fast lane. 27
,(, The same holds for image intensifiers. but lower QE and reduced resolution mitigate this benefit.
Even current EM-CCDs may not be fast enough to fully visualize this process, as experiments with 50 ns time resolution made using a gated intensified camera (but -30ms repeat rates) show very fast initiation of the calcium spark (Kindzelskii and Petty, 2003).
236
Chapter 10 • D. Toomre and
J.B.
Pawley
Phalloidin (F-actin)
FM 4-64 (membranes)
FIGURE 10.13. Fast, multi-color 3D optical sectioning (and reconstruction) of a single axon within an intact lamprey spinal cord recorded using a Yokogawa CSU-22 disk-scanning confocal microscope mounted on an Olympus IX-71 with a Hamamatsu Orca ER camera. (Left) Alexa-488 phalloidin and FM 4-64 were microinjected into a single axon of a lamprey spinal cord. 60 x 1.2NA objective. Merged image clearly reveals that F-actin surrounds the synaptic vesicle clusters. (Right) A 3D stack of -200 zsections was acquired at 100nm steps (-lOOms exposures); the total acquisition time was only -20s (much less than for 200 sections with a LSM confocal). Reconstruction made using Volocity (Improvision) software. (Images kindly provided by Dr. Jennifer Morgan, Yale University Medical School.)
FUTURE DEVELOPMENTS? Although forecasting future development can be risky, it is always useful to consider what could be done better. One of the most challenging aspects of disk-scanning confocals is the low light budget
1=0
1=305
1=605
for fluorescent excitation. The Yokogawa system avoids this at the cost of laser operation (limited wavelengths, high cost, high power/cooling. and large footprint). Two technological developments may mitigate this problem. Small, efficient solid-state lasers will doubtless become less expensive and even more con-
t=905
1.4
2.4
~.
.."
0
3.4
4.2
FIGURE 10.14. 30/40 visualization of membrane trafficking with a disk-scanning confocal. (Left) 40 movie showing clathrin-GFP dynamics in PtK, cells imaged on an Ultraview PerkinElmer Yokogawa spinning disk system using a 60 x 1.2NA water objective, Hamamatsu Orca ER CCO. Stacks of 21 z-slices were collected at 200 nm spacing, with exposure times of 100 ms/slice or 0.4, using high spatial frequency gratings for in-plane resolution enhancement, or >0.8, when using lower spatial frequency gratings for optical-sectioning applications. On the other hand, if one uses an arc source to project a pattern large enough so that its image is only marginally affected by diffraction, one can use an illumination system that utilizes light from a larger area of the source and the photon flux becomes sufficient for rapid optical sectioning using patterned illumination. Indeed, the Zeiss ApoTome is such a device. In the future, it may be possible to use an array of high-power, light-emitting diodes (LEDs) as a bright, incoherent source of patterned illumination. The choice of the illumination pattern density very much depends on the type of sample as well as on the intended method of data processing. A comparably thick sample andlor a sample with volume-like staining is more difficult to process when using densely patterned illumination because the large amount of out-offocus fluorescence will dominate the small amount of modulated fluorescence stemming from the focal plane (see Appendix, this chapter). Although using a sparse pattern reduces this signal-tonoise problem, it always requires the acquisition of more raw data images to produce a single set. This implies a longer image acquisition time. An elegant way to achieve sufficient flexibility and to optimize this trade-off between relatively noise-free sectioning and acquisition speed is to generate the pattern using a programmable array (Verveer et aI., 1998; Hanley et aI., 1999; Heintzmann et aI., 2001; Fukano and Miyawaki, 2003). Out-of-focus-light can be reduced by closing the field diaphragm (Hiraoka et aI., 1990) and this technique may be very valuable. However, when using non-local reconstruction methods (e.g., the Fourier-space based approaches) the illumination pattern is assumed to be periodic at all out-of-focus positions. A small field diaphragm violates this assumption and may lead to problems during Fourier-based reconstruction. The illumination pattern can be displaced relative to the sample by either translating the mask by a well-defined distance (Neil et aI., 1997; Gustafsson, 2000, 2005), by reprogramming the
268
Chapter 13 • R. Heintzmann
pattern of a programmable diffraction (amplitude or phase) mask (Verveer et aI., 1998; Hanley et aI., 1999; Heintzmann et aI., 2001; Fukano and Miyawaki, 2003), by altering the relative phases of interfering beams (e.g., by splitting the beam and repositioning a piezo-actuated mirror in one of the beam paths) (Lanni et at., 1993; Frohn et aI., 2000; Failla et aI., 2002), or by translating the sample past a fixed pattern, followed by a computational correction to account for this sample movement (Heintzmann and Cremer, 1999a). In the latter case appropriate interpolation kernels or Fourier-space based resampling approaches (leading to a sinc kernel) must be used to reduce interpolation-induced artifacts (Yaroslavsky, 2003) that occur as features on the object move across the pixels of the CCD. Promising attempts to replace the confocal pinhole by a multiple element detector (Bertero et aI., 1984; Barth and Stelzer, 1994) with fast readout (Sheppard and Cogswell, 1990; Pawley et aI., 1996) also belong to the same category of structured illumination with somewhat widefield-like detection. s The data can thus be treated with methods similar to the descriptions given below, as long as the descanning is accounted for.
COMPUTING OPTICAL SECTIONS FROM STRUCTURED-ILLUMINATION DATA The data acquired with a setup similar to the one shown in Figure 13.1 consists of a z-series of sets of images. Each set is taken at one focus position and each member of a set is taken at a different position of the illumination light pattern. The data from each set is first processed to yield an optically sectioned image using one of the methods described below. Most methods for deriving an optical section from a set of structured-illumination images try to estimate the degree of modulation at each pixel. For simplicity, in the case of a onedimensional grid [e.g., Fig. 13.2(C), top row in Fig. 13.7, p. 272, Fig. 13.11, p. 278] it is assumed that N images are acquired, each with the pattern shifted by liN with respect to the replicative unit cell. 6 In the top row of Figure 13.7, the interleaved positions of the three illumination patterns are indicated on the left side of each image for N = 3. As displayed in Figure 13.11, those parts of the sample that are out of focus are more homogeneously illuminated. Light emitted from these parts will undergo an additional blurring when imaged onto a detector conjugate to the in-focus plane. Thus, in contrast to the in-focus light, light from out-of-focus areas will exhibit very little modulation upon variation of the x-positions of the excitation pattern. Computing the degree of modulation over the multiple images, in a pixel-by-pixel fashion, will permit the discrimination of the non-modulated, out-of-focus from the modulated, in-focus information. The degree of modulation can be calculated locally, considering only the modulation at a single pixel position over time, by various approaches: 1. Dodt (Dodt, 1990; Dodt and Becker, 2003) suggests emulating "synthetic pinholes" by summing up the thresholded images
For speed reasons, the suggested detector arrays are usually relatively small (e.g., 5 x 5 elements), yielding a confocal operation "bias" for thick specimens. 6 The unit cell is defined as the smallest (vectorial) translation that can be applied to the pattern to reproduce its structure (e.g., the distance D indicated in C, Fig. 13.2).
of a series of data acquired by transmission imaging with structured illumination. This technique has been successfully applied to the transmission infrared (IR) imaging of onion skin (Dodt and Becker, 2003) and 300l1m unstained, freshly prepared thick rat hippocampal slices in which single spines on dendrites could be resolved throughout the sample (40x, NA 0.8 objective at 780nm, Dodt et aI., 2001). The advantage of this approach is that, while the optical aberrations induced by such samples render the operation of a standard confocal microscope in transmission close to impossible, the data processing strategy employed is very adaptive to optical aberrations and thus is able to yield a useful image containing information mainly from the focal slice. 2. Benedetti and co-workers (Benedetti et aI., 1996) determined the reconstructed slice I ree by calculating the difference between maximum and minimum measured intensity Ii in each pixel i: [Eq. 1, see also Fig. 13.7(B)]. This approach7 is robust with respect to various artifacts that often arise in imaging such as readout noise and especially fixed-pattern noise from the CCD. This noise, which usually does not vary systematically over time but rather randomly from pixel to pixel, is efficiently eliminated, as is scattered light from out-of-focus planes and from anywhere in the optical path. It is observed that this approach generates spurious patterns if the number of scanning steps is too low for the size of the illumination pattern as is seen by the residual horizontal stripes visible in Figure 13.7(B). However, it should be noted that the selected width for this pattern was extraordinarily wide and only three steps were chosen. For a denser pattern or an increased number of scanning steps, the visible performance of this method is similar to approach 3 [Fig. 13.7(C)], which is discussed below. If the illumination pattern is sufficiently sparse [i.e., the spotto-spot or line-to-line distance D in Figure 13.2(C) is large enough or, more generally, if the ratio of open to opaque area, the so-called mark/space ratio, is low], simply computing the maximum (Eq. 2) yields fairly good optical sections (Bendedetti et aI., 1996), albeit without the background suppression advantages of Eq. 1. These approaches have been applied to epi-f1uorescence microscopy (Benedetti et al., 1996) and transmission IR imaging (Dodt and Becker, 2003). By Eq. 3 (termed super-confocal), a further increase in optical sectioning quality is obtained. However, Eqs. 2 and 3 yield satisfying results only for data acquired using sparse illumination. 3. Other ways of determining the degree of modulation (Eqs. 4, 5) are described by Neil (Neil et aI., 1997). Equation 4 is based on square-law detection [Fig. 13.7(C)] and Eq. 5 emulates a homodyne detection scheme. Another possible approach is the computation of the absolute magnitude of the pixel-by-pixel Fourier-transform over each set of images (Ben-Levy and Peleg, 1995). For the case of three such images, this method is identical to Eq. 5. It should be noted that both of these methods would not be able to reconstruct the high xy-spatial frequency given in the moire example discussed later [Fig. 13.5(A)], in which the Fouriertransform of the illumination grid has its peaks just outside the range of detectable spatial frequencies. This is the case when the highest possible fluorescent excitation spatial frequency is passed through the objective. Such light will consist of two beams passing through the edges of the back-focal plane. Because of the Stokes shift of the excited fluorescence light, this spatial frequency will
5
7
A system (ViCo) based on this and related concepts is marketed by Biomedica Mangoni s.n.c., Pisa, Italy.
Structured Illumination Methods • Chapter 13
269
not be imaged by the objective on the return trip. A similar situation occurs when employing grazing incidence illumination as in Frohn and colleagues (2000). Most strategies assume equally distributed phases between the images, but this is not a strict requirement. Fukano and Miyawaki (Fukano and Miyawaki, 2003) use a three-phase scheme with a relative phase of 21t/4 between the three images, and a modified Eq. 4 for sectioning. With unequal phase approaches, care has to be taken (e.g., by shifting the O-phase between series of images) to avoid bleaching illumination structure into the sample over time. 4. Scaled subtraction of the background is yet another way to process the raw images. Assuming that the nominal position and width of the illumination pattern is known, a pixel at the plane of focus is either illuminated or not in each of the images. First
1.
I
2.
I rei(x,y)
= Max I n (X,y) n=O ... N
3.
I"Jx,y)
=Max In(x,y)+Min I n(x,y)-2Avg I,,(x,y)
ree
(x,y)=MaXI (x,y)-MinI (x,y) n=-O ... N
n
n=-D ... N
n=O ... N
n
n=-D ... N
"",0.
N
N
4.
FIGURE 13.4. Confocal slices computed from structured-illumination data.
I,,,(x,y)= L(in(x,y)-I(n+Nd'"2}rnodN(x,y))2 n=O
N
N
n=-O
n=-O
6a. I"Jx,y) = LI,.(x,y)Mask~n(x,y)-y Lln(x,y)Mask~ff(x,y)
(A) Virtual pinhole diameter chosen to be half the spot-to-spot distance. (B) Virtual pinhole selected as 0.15 of the spot-to-spot distance. (C) Scaled subtraction method with fixed Il = 1 (see Eq. 6 in boxed list of equations) and illumination spot of Gaussian shape with width as in (B). [The experimental data for this figure was kindly provided by Pier Alberto Benedetti (pollen grain taken at 1.3 NA, -450nm excitation, -550nm emission, 6 x 6 = 36 patterned images acquired for this single slice. 200nm pixel pitch in sample. spot-to-spot distance, 2.1 ~m in sample).]
1
y=---l MAR N
Lln(x,y)Mask~"(x,y) 6b.
I",(x,y)=~
N
Lln(x,y)Mask,~ff(x,y)
.!!n"'-o'---cN--------.!!n"'-o'---"N------
LMask~"(x,y)
LMask~ff(x,y)
n:=O
N
n=-O
[ N
13= N?uMask~n(x,y)- ?uMask~n(x,y) N
[ N
]2
N~(Mask~n(x,y))2 - ~Mask~n(x,y)
]2
n: index of the acquired images (the grid position varies with n) N: number of acquired images, x,y: in-plane positions in the image i: imaginary constant, I: indexing the components (separated order) ~~x, ~i?, ~v~: respective maximum, minimum or average value,
computed at each pixel over all N successively acquired images. Mask~" and Mask~ff = I - Mask~n define which pixel was illuminated or not illuminated in a given frame n MAR is the mark/area ratio describing the fraction of pixels considered to be illuminated
the value of each pixel is averaged over multiple images, each recorded at one of the positions of the illumination structure, that illuminates the in-focus parts of the sample in this pixel (ON). A second average is taken of all the pixel values where the pixel is nominally not irradiated (OFF) and subtracted from the first ON average. This technique estimates and removes the pinhole-topinhole background fluorescence, assuming that in-focus parts of the object do not yield fluorescence when not irradiated and outof-focus parts fluoresce with equal brightness independent of the position of the illumination structure. It is termed "scaled subtraction" because only a scaled fraction of the sum of the pixel intensities in the non-illuminating frames has to be subtracted from the sum in the illuminating frames. Such a technique is commonly used for data where only these sums are acquired (conjugate and non-conjugate light) as in the programmable array microscope (Hanley et ai., 1999; Heintzmann et ai., 2001). The method and the scaling factor y that is applied to the non-illuminated sum is covered by Eq. 6. Theoretically, such a scaled subtraction may yield negative results even in the totally noise-free limit, but these are so small that they can be neglected in practice. Note that even though MaskoN and MaskoFF in Eq. 6 were initially thought of as being binary masks, this technique of scaled subtraction can also be used with smooth masks [see Eq. 6(b)], in which case the binary mask is replaced by the spatially varying excitation probability in the computation. Equation 6(b) has been constructed in such a way that small spatial variations in mask intensity (e.g., due to moire effects) are accounted for. If this is not an issue, ~ can be set to 1. The effect of Eq. 6(b) with ~ fixed to 1, is shown in Figure 13.4(C).
270
Chapter 13 • R. Heintzmann
Note that, in contrast to filtering approaches, which consider information in neighboring pixels, this processing considers the multiframe data pixel by pixel. The optical sectioning visible in Figure 13.4 can thus not be explained as being the effect of a high-pass filter applied to the image data. It is a genuine method of optical sectioning obtained by processing images made at multiple illumination positions. The scaled subtraction approach has also been used in a slightly different context of aperture correlation microscopy (Wilson et af., 1996). In the first image, a mask with holes at random positions in it is scanned over the object and the image though the same mask is acquired. The second image consists of a, usually shorter, exposure with widefield illumination. Scaled subtraction of the two images serves to remove the remaining widefield information, which is present in the image acquired with illumination and detection through the random mask. A similar, yet more signal-to-noise effective approach, obviating the need for a separate widefield image, is to image the light rejected at the mask onto a separate detector (Hanley et af., 1999; Heintzmann et aI., 2002) and also process the data by scaled subtraction. All of the above techniques process the data locally by considering only a single pixel position in all the members of an image set for its own reconstruction and ignoring the intensities of neighboring pixels. Some of the above methods (Eqs. 1-5) are nonlinear in the sense that, during reconstruction, they employ at least one nonlinear operation such as thresholding, squaring, computing the absolute magnitude, or the maximum. In the absence of noise, they do remain linear with respect to the emitted light intensity. As long as the signal-to-noise ratio is high, the deviations from a linear treatment remains negligible and the resulting images can be used for quantification. However, in low-signal situations (e.g ., high out-of-focus background) the deviation of the output from the true intensities in the sample will be severe. The nonlinear steps (absolute value, square root, maximum, minimum) statistically bias the result due to the influence of the noise, for example, when the true degree of modulation in a pixel should be zero but a high background is present in each of the images, methods based on taking the absolute value (Eqs. 4, 5) will yield a positive result just from the noise. They will not yield zero on average, whereas a linear method would be expected to do so. The above evaluation techniques have the advantage of being fast and easy to compute. They do not require any knowledge of the absolute pattern position nor do they need to estimate this information from the measured data. Furthermore, they often show an inherent robustness to inexact pattern positioning. However, the results they produce are generally inferior to those achievable by approaches for the linear processing of structured-illumination data as outlined below.
RESOLUTION IMPROVEMENT BY STRUCTURED ILLUMINATION In addition to achieving optical sectioning (Frohn et af. , 2001), structured illumination can also yield improved lateral resolution. The reason is that there is a moinS effect between the structured illumination pattern and the structure of the object, such that previously inaccessible spatial frequencies of the sample become detectable (Fig. 13.5). However, to yield a useful reconstruction, the illumination pattern must first be disentangled from the detected moire fringes. This method was first conceived by Lucosz for the case of a rather dense line-grid illumination [Fig. 13.2(C)] and detection masks (Lucosz and Marchand, 1963) (the unit-cell
"'--~~Transformed
into Fourier Space
FIGURE 13.5. Moire effect. A microscope can only detect information up to a maximal spatial frequency. In other words , a minimum distance between the maxima of a grid-like feature in the sample is required. The left column shows real-space features. whereas the right column shows the corresponding situation in Fourier-space. The circle indicates the limiting frequency (pass-band of the transfer function) up to which the microscope can detect information. In (A), a grid feature of the sample is shown that cannot be resolved with evenly distributed illumination, yielding equal fluorescence everywhere in the detected image. By illumination with another dense spatial grid (B) , an aliased grid (C) is generated (moire effect) that can then be partially detected (D). However, this detected grid (D) has "incorrect" spacing. Detailed knowledge about the moire effect generated by the illumination pattern (B) can be used to reassign the detected spatial frequency to the correct place, thereby reconstructing (A).
distance in the sample coordinate system was on the order of the size of the point spread function) and also for two-dimensional grid [Fig. 13.2(A)] patterns (Lucosz, 1967). These arguments are based heavily on Fourier-space considerations and form the basis for all the computational unmixing systems used for data from widefield detectors recording images excited by structured illumination (as in Fig. 13.1).
Structured Illumination Methods • Chapter 13
FOURIER-SPACE - AN INTRODUCTION Fourier-space is extremely useful in discussing optical imaging because the process of imaging can be modeled by simply multiplying the Fouriertransform of the sample distribution by the Fourier-transform of the point spread function of the microscope (the so-called optical transfer function); a process that is carried out in Fourier-space. To turn this result into a simulation of the image, this product must be "inverse-Fourier transformed" to bring it back into real space. The concept of Fourier-space is based on the idea that any function in one-, two-, or three-dimensional space (e.g., a f1uorophore distribution) can be represented as a sum of sinusoids varying in spatial frequency, direction, strength, and phase (position). The spatial frequency (or wave number) of a sine wave describes the number of bright maxima per distance (e.g., with the unit of meter-'). Its component along the x-axis is indicated by kx while the number of maxima counted along the y-axis is termed k,. The right column of Figure 13.5 shows the magnitude of the Fouriertransformation of simple structures consisting of one (A, B, D) or only a small number of sine waves [Fig. 13.5(C)]. The displayed brightness is proportional to the magnitude of the transform at each location in Fourierspace (also called reciprocal space, frequency space, or k space). Such a plot shows only the strength of the sine wave, its direction, and spatial frequency. The exact phase (defining the position of the first maximum) is not displayed. The center in Fourier-space is located in the middle of each image. This position of zero spatial frequency corresponds to a uniform brightness in real space. The intensities away from the origin represent smaller and smaller spacings the farther out that they are. As is apparent from Figure 13.5(A,B,D), a sinusoid shifted to allpositive values (by adding a constant), for example, the emitted sample intensity, yields three peaks in its Fourier-transformation. The central, zero-frequency peak represents the added constant and the two other peaks, taken together form the remaining sine wave in real space. Each individual peak in Fourier-space actually forms a complex-valued wave exp(i2rrkr) with the respective k vector in positive/negative directions. As a sum these constitute the sine in real space: 2sin(2rrkr). For a preliminary understanding of Fourier-space, it is sufficient to know that the values at two opposing positions in k space are complex conjugates to each other and when combined always form a sine wave in real space. Note that the small features of a sample are represented by sums of sine waves with high spatial frequency (small distance between successive wave maxima), such as is indicated in Figure 13.5(A). A coarser sinusoid [e.g., Fig. 13.5(0) as compared to Fig. 13.5(A)] has a k vector closer to the origin of Fourier-space. The process of imaging is represented as a modification of the Fouriertransform of the sample distribution caused by mUltiplying it by the optical transfer function in Fourier-space. This function decays smoothly and amounts to zero everywhere beyond a certain maximum spatial frequency. This position, beyond which no information can be transferred, is indicated by the white circles in Figure 13.5. The decaying optical transfer function of a widefield microscope is also indicated by the dotted line in Figure 13.6. For a widefield microscope this in-plane cut-otf frequency, when translated into a peak-to-peak distance of a sinusoid corresponds to the equation d = A/(2NA), with the vacuum wavelength A and the numerical aperture NA. To properly measure a sample feature (sinusoid) of this frequency. the observer needs to measure at a pixels-to-pixel spacing of below half this distance to avoid misinterpretation (aliasing) of the result. This required maximal pixel-to-pixel distance is called the Nyquist distance d Ny = A/(4NA). For a different introduction to Fourier space, see the Appendix to Chapter 24, this volume.
Imaging can be treated elegantly in Fourier-space (see box "Fourier-Space - An Introduction") because a microscope essentially acts as a Fourier-filter. The periodic pattern introduced in the illumination path of the microscope yields a modification of the incident and thus of the emitted light. For fluorescence or reflection-type microscopy, the emitted light can be described as a multiplication of the sample structure times the illumination inten-
271
Magnitude Excitation Structure
A
Detectio;-;;ass-band
B
,,-----
----~./
Widefield'" Detection Pass-band
FIGURE 13.6. Scheme of the linear image reconstruction. (A) Fouriertransform of the structure of the emitted light. The optical transfer function defining the range of detectable special frequencies is indicated by the dotted line. (B) Detection sensitivity for the various reconstructed orders with their zero-frequency relocated to the origin.
sity or amplitude structure, respectively. This multiplication in real-space translates into a convolution in Fourier-space. Due to the periodicity of the illumination distribution, its Fouriertransform is a number of (delta) peaks at the reciprocal grid positions. The pattern of emitted light is a multiplication (in real space) of the illumination intensity distribution with the Fouriertransformed object, thus its Fourier-transformation is a sum of multiple Fourier-transformed objects (termed object components). These object components have their zero-frequency displaced to align with the reciprocal grating of the illumination distribution [Fig. 13.6(A), Fig. 13.8(C)]. The position and shape of the illumination structure in real space determines the individual position, strength, and phase of the multiple overlapping object components in Fourier-space. The imaging of this emission intensity distribution is then described in Fourier-space by a multiplication with the optical transfer function. It is possible to computationally unmix the sum of the displaced copies of the object [Fig. l3.8(D)] by inversion (or pseudo-inverse) of the mixing matrix M, which mathematically describes the linear superposition of displaced object components and their relative phases [Eq. 7(b); see also Gustafsson, 2000; Heintzmann, 2003] and to shift the displaced position of their zero object frequency back to the real zero frequency. This shifting makes it obvious that, in comparison to flat-illuminated widefield microscopy, the pass-band of the microscope has then been increased by this moire effect [Figs. 13.6(B), 13.8(E)]. Note that as opposed to the previously described methods (paragraphs 1-4), such resolution increase is even possible when the spatial modulation frequency falls outside the pass-band and is thus not imaged. Acquiring a set of images (similar to the top row in Fig. 13.7), makes it possible to unmix the overlapping displaced object components [Fig. 13.8(C)]. An equation system for this unmixing can be constructed [Eq. 7(b)] by using the unique dependence of the complex phase of each object component on the position of the illumination structure. In a last step of image reconstruction, the individual unmixed object components [e.g., Fig. 13.8(D)] are shifted to their proper positions in Fourier-space, and multiple components that are present at the same frequency are averaged with frequency dependent weights (e.g., with the inverse variances
FIGURE 13.7. Reconstruction results obtained by different strategies. The top row shows the three individual raw data images taken at different illumination pattern positions (as indicated by the white lines at the left side of each image) with a Zeiss ApoTome setup (Axiovert, 40 x NA 1.3 objective, -3.4 11m pattern pitch in the sample, Axiocam 6.7 11m pixel pitch). The middle row shows the results obtained by (A) The sum (B) max-min, Eq. I, (C) quadrature method, Eq. 4. Panel (D) shows an xz-cut obtained from the ApoTome software at the approximate slicing position indicated by the dashed line in panel (A). The threedimensional sectioning capability discriminating between layers of cells is seen nicely in (B), (C), and (D).
A
B
c
FIGURE 13.8. Various steps in the image reconstruction process. (A) An example raw data image. (B) The widefield image computed by summing over all partial images. The region of interest (ROI) used for images (G) to (K) is indicated as a black rectangle. (C) Magnitude of the Fourier transformation of (A), displaying the multiple overlapping components. (D) One (k = 1.0) object component separated from multiple measurements similar to (C) taken at varying illumination mask positions [one such position is shown in panel (A)]. The region of support in detection is indicated by the dotted white circle. (E) All components shifted, averaged, and apodized (border indicated by white circle). (F) Final reconstruction result in the linear case (total of 7 orders). (G) ROI extracted from raw data shown in panel (A). (H) ROI from widefield-like image as shown in (B). (I) Image (H) contrast-enhanced. (J) Contrast-enhanced ROI from confocal processing as shown in Figure 13.4(B). (K) Equivalent zoom of image (F). [The raw data for this figure was kindly provided by Pier Alberto Benedetti (for acquisition parameters see Figure 13.4).]
Structured Illumination Methods • Chapter 13
273
FIGURE 13.9. Demonstration of resolution extension by structured-illumination microscopy. The sample is a squash preparation of a polytene chromosome from a salivary gland of the fruitfly Drosophila melanogaster, stained for DNA with Oligreen to show the characteristic banding pattern. Many band features that are undetectable in the conventional microscope image (A) are clearly visible in the structured-illumination microscopy reconstruction (B). The structured illumination consisted of a one-directional pattern of parallel lines with a period of 0.20 /lm, at a wavelength of 457 nm. This pattern frequency, about 80% of the resolution limit, is much higher that those used in the rest of this chapter. Nine raw data images were used: three images with a relative phase shift of 21[/3, for each of three pattern orientations spaced 1200 apart. The reconstruction used a Fourier-based algorithm that includes compensation for the detection OTF, as described in Gustafsson (2000). [These images were kindly provided by Mats Gustafsson (unpublished data) and the sample was prepared by Harry Saumweber.J
[Eq. 7(a); Gustafsson, 2000; Heintzmann, 2003; Fig. 13.8(E)]. The weights of this averaging in Fourier-space are adjusted as inverse variances of the noise such that the best quality reconstruction at each frequency is preferred over more noisy reconstructions. Finally, there is an inverse Fourier transformation back to real space [Figs. 13.8(F)]. The resolution improvement obtained can be seen in Figure 13.8(K) compared to what is obtained by just adding the widefield detected set of images [Fig. 13.8(H)] even after contrast enhancement [Fig. 13.8(1)] or by processing the same data to simulate confocal imaging [Fig. 13.8(1)]. In Figure 13.9 the resolution improvement is even more obvious. This figure was generated by illumination with a grating close to the highest transmittable spatial frequency, which leads to more prominent highfrequency components with less noise after their extraction (kindly provided by M. G. L. Gustafsson). In some of the figures [especially in Fig. 13.8(1)] a residual patterning can be observed. This can be attributed to having selected too few pattern positions (here 6 x 6 == 36) in the set. It is interesting to note that the sampling of the raw data images need only satisfy the Nyquist limit of the widefield microscope even though the resolution finally achieved extends beyond this limit. During the relocation process of the object components in the reconstruction, the discrete frequency space can be "extended," essentially resampling the data onto a different grid. This reconstruction does not contradict information theory because many images, each with different high-frequency sample information (but downshifted into the detection pass-band) are used to construct a single image containing information outside the detection passband. Although negative intensities can in principle result
during the process of image reconstruction, this does usually not pose a practical problem. With this approach, the resolution of the microscope can theoretically be enhanced by a factor of about 2 in-plane (xy) as well as along the optic axis (z) compared to the standard epifluorescence widefield microscope. Although the fundamental pass-band limit obtained by this method is not larger than the passband of a confocal microscope, a substantial practical improvement over standard confocal microscopy is achieved. The high spatial frequencies of the object are detected much more efficiently because the moire effect of the illumination grid shifts them into a region of the pass-band that is more efficiently detected. Lucosz's formulation (Lucosz and Marchand, 1963; Lucosz, 1967) makes it possible to understand both the confocal microscope and multiple dot-scanning systems in the nomenclature of computational reconstruction. The moving detection mask of these systems, in combination with the integration of partial images on the detector, permits the required unmixing to be achieved automatically, the positions of the zero frequencies of the unmixed components to be correctly adjusted and integrated (which means essentially summed) in the detector with component-dependent weights. Thought of in this way, the shape and size of the detection pinhole defines the relative weights of individual object components. As opposed to Lucosz' approach or scanning disk systems, in which the decoding is achieved by a detection mask, computational reconstruction is far more flexible. Assume as a gedanken experiment that we illuminate the specimen with an array of bright points and then scan these points over the field of view so that
274
Chapter 13 • R. Heintzmann
FIGURE 13.10. Performance simulation of dense structured illumination for thick samples. (A) Simulated object. (8) Widefield image (1000 expected photons in maximum). (C) Confocal image (pinhole size 0.5 Airy units, detector efficiency 20% that of panel 8). (D) yz-Section computed using Eq. I from dense structured illumination. three phases. Further reconstructions were based on (E) Eq. 4, and (F) Eq. 6(b). (G) Widefield image with a background fluorescence added corresponding to approximately five layers of this object when densely packed. (H) Reconstruction (Eq. 4) of a structured-illumination dataset under similar sample conditions. (I) Reconstruction (Eq. 4) of data with the spacing between the illuminating bars increased 3-fold (total illumination dosage was kept constant for all simulations). Simulation parameters as described in the text.
eventually it is all covered. If we record a separate widefield image of the specimen for each location of the excitation array, we now have a stack of data that contains all the information we need to construct an optical section of the focus plane merely by summing the signal collected at virtual "pinholes" near the locations of the excitation points in each image of the stack (similar to the approach chosen for Fig. 13.4). Assuming the same illuminating power, the same scan time, and the same detector performance,s this optical section will be in every way identical to one made with a diskscanning confocal microscope. However, this admittedly rather tedious approach has the added advantage that we could have used any of the seven methods (Eqs. 1-7) for computing optical
8
Including pixels small enough to accurately delineate the image of each round pinhole in the image of the disk.
sections discussed above and in addition we could compute results to reflect any detector pinhole size after the acquisition of the data. Although this approaches pure confocal operation (by using very sparse structured illumination), it would be extremely timeconsuming to take the many individual images. In addition, taking so many high-speed individual images will increase the readout noise with obvious consequences. Other problems would be caused by the massive amount of data (a full image acquired for each scan position, of which there could be a million). Thus, despite these theoretical advantages, practical limitations still render standard confocal microscopy more useful for imaging thick specimens (for a detailed example, see box "Imaging of Thick Specimens"). Approaches such as the programmable array microscope (Verveer et aI., 1998; Hanley et aI., 1999; Heintzmann et aI., 2001)
Structured Illumination Methods • Chapter 13
IMAGING OF THICK SPECIMENS In Figure 13.10, a sample consisting of a yz-oriented star-shaped stack ("wagon-wheel pasta," see also Chapter 24, this volume) and its images were simulated [diameter of object: 41lm, voxel size 100 x 100 x 100 nm, theoretical point spread function of 1.2 NA water immersion (n = 1.33) objective. Pattern spacing in sample: (B-H) 900 nm, (I) 2.7Ilm, pattern width 300nm.] The total dosage delivered to the sample for the acquisition of all of the raw data necessary for panels (B) to (I) was kept constant. This yielded a total of 1.2 x 10" detected photons for the central slice (B, O-F), 28 x 106 photons (G-I), and 36 x 103 photons for the confocal image (C). The optical sectioning advantage of the confocal (C) over the widefield image (B) is clear even though the signal-to-noise ratio in the confocal image is lower due to rejection of light at the pinhole and lower detection quantum efficiency (QE). Note that in the widefield image only -17% of the photons in the central slice stem from the in-focus region (3 voxels deep). Images obtained from processing the structured-illumination data (O-F) show only minor differences in quality and are comparable also to the confocal image. However, when additional background is added (as in G, H) stemming from a distant out-of-focus region in the form of five additional densely packed layers of overlying wagon-wheel pasta, the reconstructions from dense three-phase structured-illumination data are of poor quality (H), whereas the confocal image quality (C) would be unaffected. For this amount of background less than 5% of the collected photons stem from the i n-focus area. However, by adjusting the spacing between the illumination bars to yield a more sparse illumination (I), a better sectioning is achieved and the quality of the reconstruction becomes more acceptable. The price of this improvement is that data acquisition takes longer (if limited by the instrument) and the raw data file is larger. By going to even mores sparse illumination and a two-dimensional pattern, confocal quality will be reached, but the long acquisition times may be unacceptable. Note that each final reconstruction (including the widefield and confocal cases) assumed that the same total number of photons was emitted by the sample. The confocal detector (photomultiplier tube assumed) was assumed to have a quantum efficiency 20% as high as the detectors used in the widefield detection mode (a good quality CCO assumed). Read noise of the CCO was not accounted for because it would be of marginal influence in these bright images. For a further discussion on the effect of imaging thick specimens see the Appendix of this chapter.
that can rapidly change "pinhole patterns" using a digital mirror device for both illumination and detection, can make use of normally rejected light by sending it to a second detector (Heintzmann et ai., 2001) and have obvious advantages in speed and reduction of readout noise in comparison to acquiring all the individual images of a set. However, when integrating over a set of pinhole positions, these devices do not permit the widefield detection of each of the images of a set of structured-illumination data (e.g., changing the pinhole size retrospectively as in Fig. 13.4 is usually not possible). Computational reconstruction of structured-illumination data is in some ways related to achieving resolution improvement by analysis of a series of intensity distributions at the pinhole plane based on a singular functional decomposition (Bertero et ai., 1989), the major difference being that aperture-modifying filters influence the light amplitude and not only the intensity as does reconstruction. The minimum number of images required is defined by the number of object components that need to be separated [see Figs. 13.6(A) and 13.8(0)]. For a single optical slice, it ranges from three images when using a one-dimensional diffraction grid for improved sectioning (Neil et al., 1997), a low number of images
275
(typically 7-9) for in-plane resolution improvement with successive rotation of the illumination grid (Heintzmann and Cremer, 1999; Gustafsson, 2000) to many tens of images (Benedetti et aI., 1996) for use with relatively sparse two-dimensional patterns (pinhole distance ~PSF size). However, looking at this issue from the perspective of information theory indicates that a reduction of this number should be possible (Cox and Sheppard, 1986). The weighted averaging in the reconstruction process also hints that much of the information has been acquired multiple times. The key to a major reduction in the minimum number of images required is to modify the reconstruction process so that both order separation and weighted averaging are combined into a single processing step (Heintzmann, 2003). When this idea is extended to three-dimensional data, another substantial reduction in the number of images per slice can be expected. For a z-sectioning image stack using dense grids, this could reduce the minimum to less than three images per slice on average. When acquiring a focus series (a z-stack of image sets), the plane of focus for the illumination pattern usually coincides with the plane of focus of the detection path. In this situation, the theory of the imaging process becomes slightly more complex than outlined above. In the description above, the patterns were assumed as being "part of the object," that is, as its multiplicative modification. However, in z-direction, patterning has to be treated as being "part of the point spread function" because the illumination pattern usually stays aligned with the plane of best focus as it steps though the sample, and is not fixed to the object. The standing-wavefield microscope (Bailey et ai., 1993; Lanni et aI., 1993; Krishnamurthi et aI., 1996; So et ai., 2001) and the incoherent illumination image interference imaging (1 5 M) microscope (Gustafsson et al., 1995, 1999) generate patterned illumination with either coherent laser light or an incoherent light source, respectively, and illuminate with alternating bright and dark xy-planes stacked along the optical axis. In 15M, the maximum of the illumination pattern coincides with the plane of best focus in detection. The illumination pattern stays in a fixed spatial relation to the detection point spread function and can thus be treated as part of it. This simplifies the reconstruction, as no computational unmixing is required. However, standing-wavefield microscopy with illumination solely along both directions of the optic axis suffers from a large region of missing intermediate z-frequencies, essentially rendering three-dimensional reconstruction of large features impossible. (In real space this corresponds to the ambiguity problem between lobes along the z-axis of the PSF.) In 15 M, the situation is much improved, such that it enables image reconstruction (Gustafsson et ai., 1999). The effect of noise on 15M and 4Pi (which is essentially a point-scanning technique, described in Chapter 30, this volume) were compared by Nagorni and Hell (2001 a, 2001 b) indicating the superior performance of the 4Pi approach, especially along the axial direction. However, a combination of 15M with additional patterning along the in-plane directions can be expected to yield an additional increase in resolution. This may also have the potential to overcome some of the signal-to-noise difficulties of 15M in comparison to 4Pi microscopy. All of the reconstruction techniques described above suffer to some extent from photobleaching of the fluorophores. Because bleaching is caused by structured illumination, patterns may be bleached into the sample and have to be compensated for during reconstruction. Efforts to compensate for patterned bleaching are treated in detail by Schaefer and colleagues (2004). Although the three-dimensional imaging techniques described above have usually been developed for fluorescence, they have
276
Chapter 13 • R. Heintzmann
also been applied to incoherent reflection (Neil et aI., 1997) and transmission (Dodt et aI., 2003). For the coherent case, approaches similar in spirit to the Fourier-techniques developed for synthetic aperture radar are used (Mermelstein, 1999; Schwarz et aI., 2003; Nellist et at., 1995).
Nonlinear Structured Illumination As described by Heintzmann and colleagues (2002), the methods of structured illumination can be extended to the nonlinear regime of the experiment in a straightforward way, and doing so yields another substantial resolution improvement over linear structured illumination. If any kind of nonlinearity exists between the illumination intensity and the emission intensity finally measured, further peaks in the Fourier-transfonnation of the effective excitation distribution will arise. In the absence of noise, this allows for any number of components to separate and thus, theoretically, for infinite resolution in the reconstruction. In practice, signal-to-noise issues and the type of nonlinearity limit the achievable resolution, even if grid quality, positioning accuracy, and detector linearity are perfect. One of the nonlinear effects we have discussed (Heintzmann et al., 2002) is fluorescence saturation (Sandison et aI. , 1995). In this case, the sample is irradiated with structured illumination over the full field of view. The intensity required to achieve saturation is extremely high, but nevertheless possible by using pulse lasers and illuminating only a few nanoseconds. Recently, this approach has been practically demonstrated by Gustafsson (2005) who claimed an in-plane resolution of -50nm. This fluorescence saturation idea is related to the saturation of the stimulated emission employed in stimulated emission depletion (STED) microscopy (Klar et aI., 2000; see also Chapter 31, this volume). It should be mentioned that multi-focal, two-photon microscopy yields inherent sectioning due to the two-photon effect and thus does not require a detection mask (see also Chapter 29, this volume; Andresen et aI., 200 I). As these systems that illuminate with a pattern of light could allow the detection of the image at every pattern position, the reconstruction technique described above should also be applicable, with its potential for flexibility and resolution improvement. If one neglects the extra readout noise associated with recording all those images, a reconstruction based on such a set of structured-illumination images would increase the signal-to-noise ratio especially for high frequencies in comparison to just acquiring the summed images, as would be done in the standard system. However, because the excitation probability using the two-photon effect is proportional to the square of the incident intensity, analyzing the data as a case of structured illumination would contribute a factor of 2 increase in resolution from the excitation side of the scheme. This would compensate for the longer excitation wavelength used for two-photon excitation. Thus structured illumination based on the nonlinear two-photon effect leads to no major resolution increase compared to single-photon structured illumination. Other nonlinear effects, such as saturation phenomena, do have higher orders, which can also be utilized for substantial resolution improvements (such as in stimulated emission depletion microscopy). As any type of nonlinearity can be used in combination with this concept, the nonlinearities considered in Hell and Kroug (1995) and Schonle and co-workers (1999) are also promising approaches for the concept described by Heintzmann and Cremer (l999b) and Heintzmann and colleagues (2002). Other extremely promising candidates are dyes (Corrie et at., 2001; Giordano et ai., 2002) or proteins (Ando et at., 2004) that can be converted
between two (or multiple) states under wavelength selective illumination. These compounds constitute multi-level systems in which saturation characteristics can be utilized without requiring excessive illumination intensities (Hell, 2004).
SUMMARY • Structured illumination in combination with widefield detection typically requires the acquisition of a large amount of data in comparison to standard confocal or Nipkow-type disk systems. • One need not select a "pinhole size" during data acquisition. The data acquired can be processed in different ways to emphasize different contrast and resolution/noise trade-offs. Depending on the nature of the pattern employed, a practical resolution improvement of a factor slightly above 2 in each direction of space can be achieved compared to standard widefield microscopy. Using nonlinear approaches this factor can be made substantially bigger. • Practical speed limits are imposed by the current camera and readout technology, but, more important, by the absence of bright light sources suitable for incoherent full-field illumination.
ACKNOWLEDGMENTS Drs. James Pawley, Stefan Hoppner, David Richards, Stefan Hell, Thomas Jovin, Anje Sporbert, Verena Hafner, and Keith Lidke are thanked for their help in improving this manuscript. I am very thankful to Dr. Pier Alberto Benedetti, who contributed experimental image data (Fig. 13.8) as did Drs. Bauch, Kempe, Schadwinkel, and Schaffer (Fig. 13.7) from Carl Zeiss, Gottingen, Germany. Furthermore, Drs. Thomas lovin, Christoph Cremer, Mats Gustafsson, Stefan Hell, Jan Boese, and Quentin Hanley are thanked for many fruitful discussions regarding patterned illumination.
REFERENCES Andresen, V., Egner, A., and Hell , S.W., 2001, Time-multiplexed multifocal multiphoton microscope, Opt. Lett. 26:75-77. Ando, R. , Mizuno, H. , and Miyawaki , A. , 2004, Regulated fast nucleocytoplasmic shuttling observed by reversible protein highlighting, Science 306:1370-1373. Barth, M. , and Stelzer, E .H.K. , 1994, Boosting the optical transfer-function with a spatially resolving detector in a high numerical aperture confocal reflection microscope, Optik 96:53-58. Bailey, B., Farkas, D.L., Taylor, D.L. , and Lanni, F., 1993, Enhancement of axial resolution in fluorescence microscopy by standing wave excitation, Nature 366:44-48. Ben-Levy, M., and Peleg, E ., August 1995, WO 97/06509, US Patent 5,867,604. Benedetti, P.A. , Evangelista, v., Guidarini , D. , and Vestri, S., 1996, US Patent 6,016,367. Bertero, M. , Boccacci, P. , Defrise, M ., De Mol , c., and Pike, E.R., 1989, Superresolution in confocal scanning microscopy: n. The incoherent case, Inverse Pmblems 5:441-461. Bertera, M ., De Mol, c., Pike, E.R, and Walzer, J.G., 1984, Resolutio n in diffraction-limited imaging, a singular value analysis. IV. The case of uncertain localization or non-uniform illumination of the object. Opt. Acta 31 :923-946. Bewersdorf, J., Pick, R., and Hell, S.W., 1998, Multifocal multiphoton microscopy, Opt. Lett. 23:655-657.
Structured Illumination Methods • Chapter 13
Corrie, J.E., Davis, c.T., and Eccleston, J.E, 2001, Chemistry of sulforhodamine-amine conjugates, Bioconjug. Chem. 12:186-194. Cox, U., and Sheppard, CJ.R, 1986, Information capacity and resolution in an optical system, J. Opt. Soc. Am. A 3:1152-1158. Dodt, H.-D., 1990, German Patent DE 4023650 AI. Dodt, H.-D., and Becker, K., 2003, Confocal microscopy in transmitted light, Proc. SPIE 5139:79-87. Dodt, H.-D., Becker, K., Eder, M., and Zieglgansberger, W., 2001, Confocal microscopy of unstained neurons in brain slices, Schloessmann Seminar 2001 Abstractbook, Max-Planck Society, Schloss Elman, Oberbayern, DE, pp.22-23. Egner, A., Jakobs, S., and Hell, S.W., 2002, Fast 100-nm resolution 30microscope reveals structural plasticity of mitochondria in live yeast, Proc. Natl. Acad. Sci. USA 99:3370-3375. Failla, AV, Spoeri, D., Albrecht, B., Kroll, A., and Cremer, c., 2002, Nanosizing of fluorescent objects by spatially modulated illuminated microscopy, Appl. Opt. 41:7275-7283. Frohn, J.T., Knapp, H.E, and Stemmer, A., 2000, True optical resolution beyond the Rayleigh limit achieved by standing wave illumination, Proc. Nat!. Acad. Sci. USA 97:7232-7236. Frohn, J.T., Knapp, H.E, and Stemmer, A., 2001, Three-dimensional resolution enhancement in fluorescence microscopy by harmonic excitation, Opt. Lett. 26:828-830. Fukano, T., and Miyawaki, A., 2003, Whole-field fluorescence microscope with digital micromirror device: Imaging of biological samples, App!. Opt. 42:4119--4124. Giordano, L., Jovin, T.M., Irie, M., and Jares-Erijman, E.A., 2002, Diheteroarylethenes as thermally stable photoswitchable acceptors in photochromic fluorescence resonance energy transfer (pcFRET) J. Am. Chem. Soc. 124:7481-7489. Gustafsson, M.G.L., 2000, Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy, J. Microsc. 198:82-87. Gustafsson, M.G.L., 2005, Non-linear structured-illumination microscopy: widefield fluorescence imaging with theoretically unlimited resolution, Pro. Nat. Acad. Sci. USA 102:13081-13086. Gustafsson, M.G.L., Agard, D.A., and Sedat, J.W, 1999, 15 M: 3D widefield light microscopy with better than 100nm axial resolution, J. Microsc. 195: 10-16. Gustafsson, M.G.L., Agard, D.A., and Sedat, J.W, 2000, Doubling the lateral resolution of widefield fluorescence microscopy using structured illumination microscopy, Proc. SPIE 3919: 141-150. Gustafsson, M.G.L., Sedat, J.W., and Agard, D.A., DS Patent 5,671,085. Hanley, Q.S., Verveer, PJ., Gemkov, M.J., Arndt-Jovin, D., and Jovin, T.M., 1999, An optical sectioning programmable array microscope implemented with a digital micromirror device, J. Microsc. 196:317-331. Heintzmann. R., 2003, Saturated patterned excitation microscopy with twodimensional excitation patterns, Micron 34:283-291. Heintzmann, R., and Cremer, c., 1999a, Laterally modulated excitation microscopy: improvement of resolution by using a diffraction grating, Proc. SPIE 3568:185-196. Heintzmann, R., and Cremer, c., March 1999b, Patent WO 0052512. Heintzmann, R., Jovin, T.M., and Cremer, c., 2002, Saturated patterned excitation microscopy - a concept for optical resolution improvement, J. Opt. Soc. Am. A 19:1599-1609. Heintzmann, R., Hanley, Q.S., Arndt-Jovin, D., and Jovin, T.M., 2001, A dual path programmable array microscope (PAM): Simultaneous acquisition of conjugate and non-conjugate images, J. Microsc., 204:119-137. Hiraoka, Y., Sedat, J.W., and Agard, D.A., 1990, Determination of 3-dimensional imaging properties of a light-microscope system - partial confocal behaviour in epifluorescence microscopy, Biophys. J. 57:325333. Hell, S.W., 2004, Strategy for far-field optical imaging and writing without diffraction limit, Phys. Lett. A 326:140-145. Hell, S.W., and Kroug, M., 1995, Ground-state depletion fluorescence microscopy, a concept for breaking the diffraction resolution limit, Appl. Phys. B 60:495--497.
277
Ichihara, A., Tanaami, T., Isozaki, K., Sugiyama, Y., Kosugi, Y., Mikuriya, K., Abe, M., and Demura, I., 1996, High-speed confocal fluorescence microscopy using a Nipkow scanner with microlenses for 3-D imaging of single fluorescent molecules in real time, Bioimages 4:57-62. Klar, T.A., Jakobs, S., Dyba, M., and Hell, S.W, 2000, Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission, Proc. Natl. Acad. Sci. USA 97:8206-8210. Krishnamurthi, V, Bailey, B., and Lanni, E, 1996, Image processing in 3D standing-wave fluorescence microscopy, Proc. SPIE 2655: 18-25. Lanni, E, Bailey, B., Farkas, D.L., and Taylor, D.L., 1993, Excitation field synthesis as a means for obtaining enhanced axial resolution in fluorescence microscopes, Bioimaging I: 187-196. Lanni, E, Taylor, D.L., and Waggoner, A.S., 1986, DS Patent 4,621,911. Lukosz, W., 1967, Optical systems with resolving powers exceeding the classicallimit. II, J. Opt. Soc. Am. 57:932-941. Lukosz, W., and Marchand, M., 1963, Optischen Abbildung unter Uberschreitung der beugungsbedingten Auflosungsgrenze [in German], Opt. Acta 10:241-255. Majoul, I., Straub, M., Duden, R., Hell, S. W, and SOling, H. D., 2002, Fluorescence resonance energy transfer analysis of protein-protein interactions in single living cells by multifocal multi photon microscopy, Rev. Mol. Biotechnol. 82:267-277. Mermelstein, M.S., 1999, Synthetic aperture microscopy, Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, MA, June 1999. Nagorni, M., and Hell, S.W., 2001a, Coherent use of opposing lenses for axial resolution increase in fluorescence microscopy. I. Comparative study of concepts, J. Opt. Soc. Am. A 18:36--48. Nagorni, M., and Hell, S.W., 2001b, Coherent use of opposing lenses for axial resolution increase. II. Power and limitation for nonlinear image restoration, J. Opt. Soc. Am. A 18:49-54. Neil, M.A.A., Juskaitis, R., and Wilson, T., 1997, Method of obtaining optical sectioning by using structured light in a conventional microscope, Opt. Lett. 22: 1905-1907. Nellist, P.O., McCallum, B.C., and Rodenburg, J.M., 1995, Resolution beyond the "information limit" in transmission electron microscopy, Nature 374:630-632. Pawley, J., B1ouke, M., and Jenesick, J., 1996, The CCDiode: An optimal detector for laser confocal microscopes, Proc. SPIE 2655: 125-129. Petrilli, M., Hadravsky, M., Egger, M.D., and Galambos, R., 1968, Tandemscanning reflected-light microscope, J. Opt. Soc. Am. 58:661. Sandison, D.R., Williams, R.M., Wells, K.S., Strickler, J., and Webb, W.W., 1995, Quantitative fluorescence confocal laser scanning microscopy (CLSM), In: Handbook of Biological Confocal Microscopy, 2nd ed. (J.B. Pawley, ed.), Plenum Press, New York, pp. 267-268. Schaefer, L.H., Schuster, D., and Schaffer, J., 2004, Structured illumination microscopy: artefact analysis and reduction utilizing a parameter optimization approach, J. Microsc. 216:165-174. Schonle, A., Hanninen, P.E., and Hell, S.W., 1999, Nonlinear fluorescence through intermolecular energy transfer and resolution increase in fluorescence microscopy, Ann. Phys. (Leipzig) 8:115-133. Schwarz, C.J., Kuznetsova, Y., and Brueck, S.RJ., 2003, Imaging interferometric microscopy, Opt. Lett. 28: 1424-1426. Sheppard, C.J.R., and Cogswell, c.J., 1990, Confocal microscopy with detector arrays, J. Modern Opt. 37:267-279. So, P.T.c., Kwon, H-S., and Dong, c.Y., 2001, Resolution enhancement in standing-wave total internal reflection microscopy: a pointspread-function engineering approach, J. Opt. Soc. Am. A 18:28332845. Verveer, P.J., Hanley, Q.S., Verbeek, P.W., van Vliet, L.J., and Jovin, T.M., 1998, Theory of confocal fluorescence imaging in the programmable array microscope (PAM), J. Microsc. 189: 192-198. Wilson, T., Juskaitis, R., Neil, M.A.A., and Kozubek, M., 1996, Confocal microscopy by aperture correlation Opt. Lett. 21: 1879-1881. Yaroslavsky, L., 2003, Boundary effect free and adaptive discrete signal sincinterpolation algorithms for signal and image resampling, Appl. Opt. 42:4166--4175.
278
Chapter 13 • R. Heintzmann
APPENDIX: IMAGING THICK SPECIMEN WITH STRUCTURED IllUMINATION What illumination pattern do we choose for what specimen? If we neglect issues of resolution, it is apparent that a very thin sample (e.g., single molecules diffusing in a flat biomembrane) can be imaged with full-field illumination and widefield detection, whereas a very thick sample, containing a lot of fluorophores (e.g., a fish embryo expressing green fluorescent protein (GFP) throughout all the cells) requires confocal imaging. Widefield imaging has the advantage of being relatively fast and efficient in signal detection. Confocal imaging allows for the rejection of the "bad," that is, unemployable out-of-focus light that would dominate the signal in full-field illumination. Structured illumination with a user-defined illumination pattern covers the entire range from full-field illumination through dense illumination patterns to sparse illumination, approximating single-spot confocal illumination conditions. The crucial task is to obtain an estimate of the noise level that we should expect in the final data with different illumination patterns. If we first assume full-field illumination with the fraction of the illumination area, the mark/area ratio MAR = I, each horizontal plane will be illuminated with the same intensity, independent of its axial position. Depending on the structure of the sample, a defined thickness Zcq will yield an equal amount of foreground signal and out-of-focus haze. Structured-illumination techniques and even widefield deconvolution aim at computationally removing this out-of-focus haze. However, although background removal (e.g., by subtraction) can be achieved, the noise contribution of the background to the signal will still be present. Thus, specimens considerably thicker than Zeq (e.g., more than 10 times as thick) will run into signal-to-noise problems (see discussion below). Consider now an illumination pattern of horizontal bars of a thickness of d = 5/lm with a pitch (distance between the beginning of one bar and the beginning of the next bar) of D = 50/lm.
d
In the discussion above, it is assumed that out-of-focus light only influences the result as if it were detected as an additional, local uniform background (with its associated Poisson noise), and that a uniform illumination applies even to fluorophores situated directly behind or in front of the plane of focus. This is clearly NOT true for the patterned illumination geometry given in Figure 13.II(B). The parameters d and D define three illumination/detection regimes depending on whether the fluorescent structure is in focus, close, or distant from the focus plane. These regimes are characterized as follows:
• In focus: In this refined model, diffraction effects will blur the thin-line illumination, reducing its intensity. As a result, "background" contributions will arise from neighboring regions where illumination was not intended. These in-plane "spillover" effects are approximated in an ad hoc way by the following in-focus contrast reduction factor of the foreground: contrast = dre/(dre! + E), where d re ! = d/dmin is the width of a bar compared to the smallest transmittable spatial wavelength of the objective at the emission wavelength d min = Aem/(2NA). The number, E = 1.828 was fitted from a simulation of the image of a bar with variable width at NA = 1.4.
1
The mark/area ratio is defined as MAR = D = 10 which is related
d 1 to the mark/space ratio by MSR = - - = . If these D-d _1__ 1 MAR bars are sufficiently wider than the diffraction limit, such that diffraction effects can be neglected, structured illumination allows the acquisition of the foreground in full brightness (knowing which parts of the in-focus slice are illuminated) while reducing the outof-focus haze by the mark/area ratio. When an in-focus structure of interest is not illuminated, which in this case happens in 90% of the images acquired in each set of frames, the background described above is still present. However, it can be assumed that the reconstruction algorithm can account for this foreground region not being illuminated because it knows the current illumination structure and uses this information merely for a precise estimate 9 of the amount of background. This estimate is then successively removed from the "foreground" data, acquired when the particle was illuminated (e.g., using Eqs. 6a and 6b).
9
The quality of this estimate is better by a factor of signal, thus exhibiting only 113 of the noise.
~ M~R - 1 = 3 than the
FIGURE 13.11. Background in thick samples. The sample is assumed to resemble a "sea" of fluorescence. The image depicts the illumination brightness distribution in the sample. viewed from the side. The fraction of emitted light from a specific plane falling on the illuminated grid-pattern geometric positions is also proportional to this distribution, which has to be integrated over a full unit-cell (here distance D). The region very close to the illuminating bars or squares should start with a widefield-like behavior. However, to keep the model simple the whole region from the bars to the distant region is denoted close in which a linear decay in the background stemming from an out-of-focus plane, detected at the nominally illuminated in-focus positions. Note the difference for dense (A) and sparse illumination (B) corresponding to the simulations shown in Figure 13.10.
Structured Illumination Methods • Chapter 13
• Close [0 < z : $ 15,000.
choice is voxel rendering, as it avoids potential artifacts of segmentation at this early stage. This catch-all algorithm could have interactive parameter entry in order to explore the new structure if the processing were fast enough. Contrast control and careful data thresholding (to remove background only) would normally be used with this quick-look approach. More specific voxel segmentation (removing data values outside a given brightness band, intensity gradient, or other parameter range) should be used with caution during the identification of a new structure. Artifactual boundaries (surfaces) or apparently connected structures (e.g., filaments) can always be found with the right segmentation and contrast settings. In subsequent refinement stages, a case can usually be made for a more specific segmentation model. For example, maximum
intensity segmentation can be used to visualize a topological reflection image of a surface. The prerequisites for such a choice can only be confirmed by inspection of the entire image data. Finally, visualization models involving more complex segmentation, absorption and lighting effects, whether artificial or based on a priori knowledge, must be introduced in stages after the basic distribution of image intensities has been established (Fig. 14.2). Computer graphics research is beginning to offer techniques for automated or computer-assisted refinement of the visualization algorithm to automatically tune it for the particular supplied data (He et ai., 1996; Marks, 1997; Kindlmann and Durkin, 1998). Some useful user-interface tools, such the Visual Network Editor of AVS, assist in the design stages of more complex multi-step or interactive visualization procedures.
FIGURE 14.1. Viewing multidimensional LSM data. In order to make maximum use of imaging resources, multidimensional CLSM images are routinely viewed away from the microscope. "Thick" 2D oblique sections (A , B) can be extracted at moderate rates by many software packages. 2D orthogonal sections (C- E) can be viewed at video rate. Reconstructed 3D views (F, G) require more extensive processing, now common in all commercial systems. (A-D, F, G) are refl ec ti on images of Golgi-stained nerve cells. (E) Multiple xy views (e.g. , from an an imation) of fluorescently stained nerve cells.
284
Chapter 14 • N.S. White
FIGURE 14.2. Identifying unknown structures. It is important to make as few assumptions as possible about the imaged structures during the exploratory phase of 3D visualization. Average (or summation) projection (A), though simple, often gives low-contrast views due to the low weight given to small structures . Maximum brightness (B) gives higher weight to small bright structures but when used in isolation provides no z-discrimination. (C) Background thresholding (selling to zero below a base line) is simple, easy to interpret and increases contrast in the view. (D) Re-orientating the 3D volume (even by a few degrees) can show details not seen in a "top down view," and coupled with animation (see text), this is a powerful exploratory visualization tool. (A-C) processed by simple z-axis projections, (D) "Maximum intensity" using the Lasersharp software. Lucifer Yellow stained nerve cell supplied by S. Roberts , Zoology Department, Oxford University.
Highlighting Previously Elucidated Structures Having ascertained the importance of a particular feature, the next step is to enhance the appearance for presentation and measurement (Fig. 14.3). Connectivity between voxels in, for example, a filament or a positively stained volume, may be selectively enhanced (the extracted structures may even be modeled as graphical tubes or solid objects; see SoftWorx from API and Imaris packages for examples). A threshold segmentation band can be interactively set to remove intensities outside the particular structure. 3D fill routines, 3D gradient, dilation, and other rank filters are the basis for structural object segmentation. Opacity (reciprocal to transparency) is possibly the most used visualization parameter to highlight structures segmented by intensity bands. This parameter controls the extent to which an object in the foreground obscures features situated behind it. Consequently, it artificially opposes the intrinsic transparency of biological specimens. Artificial lighting is
applied during the final stage. Artificial material properties (such as opacity, reflectivity, shininess, absorption, color, and fluorescence emission) are all used to simulate real or macroscopic objects with shadows, surface shading, and hidden features.
Visualization for Multi-Dimensional Measurements Often, the final requirement of objective visualization is the ability to extract quantitative measurements. These can be made on the original image, using the reconstructed views as aids, or made directly on the display views. The success of either of these methods depends on the choice of reconstruction algorithm and the objective control of the rendering parameters. Table 14.2 gives an overview of visualization tools that might be useful for objectively exploring the image data (see also Chapter 15, this volume).
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
285
FIGURE 14.3. Enhancing and extracting objects. Having elucidated a particular structure within the volume, filtering and segmentation permit selective enhancement. (A) "Maximum intensity" view [as in Fig. 14.2(B)] after two cycles of alternate high-pass and noise-reduction filters on the original 2D xy sections (using 3 x 3 high-pass and low pass Gaussian filters). (B) More extreme threshold segmentation to extract the enhanced details during the projection. (C,D) two rotated and tilted views (Lasersharp software) using a "local average" (see text) to bring some "solidity" to the view. This example shows the principle danger of segmentation, that of losing fine details excluded from the intensity band.
TABLE 14.2. Overview of Visualization Parameters Desirable for Visualizing Multi-Dimensional Biological Microscopy Data Processing step "3D algorithms
Parameter General modes Quick modes
hControlling the reconstruction process
Visualization parameters
Pre-processing tools
'Interactive controls
Visualization parameters Measurements on image Measurements on views Simultaneous measures on image & views
Minimum required
Xy, XZ, yz orthogonal slices Z-weighted "projections" Fast xy, xz slices Maximum projection Projection angles z-stretch Animation controls Sequence (movie) mode 2D & 3D image edit 2D image rank filter threshold/background contrast Slice positions rotation angles data threshold animation controls 2D measures on slices Multiple measurements on screen
Desirable enhancements Arbitrary slices, Voxel a-blending Surface rendering Hardware acceleration, Sub-sampling data Viewing angle, z-fill, Data threshold, Voxels/surfaces, Shading control, Lighting controls, Material properties, Opacity, SFP/"special" modes, Perspective, Batch processing, Post-lighting n-D image edit, n-D filters, image restoration z-correction, morphological filters, math operations All render parameters, Data/view angles, View Zoom & pan, "Real time" control 3D measures on slices, n-D measurements 3D measures on views, n-D measurements Measures auto, Tracked in both displays
"The range of 3D algorithms indicates diversity of modes for tackling various kinds of image data, while the quick-look modes include general sectioning and the fastest voxel algorithms. Simple projections and section movies are always faster than more sophisticated reconstruction modes. hControl of the visualization process suggests the range of parameters that the user can modify directly to influence the resulting views. More controls give more flexibility but also more complexity in use. e A very rough idea of the level of interactive control (i.e., rapid changes of the result in response to mouse clicks, etc.) for visualization and for measurements on source image data andlor reconstructed views. n-D = any number of multiple dimensions.
286
Chapter 14 • N.S. White
WHAT CONFOCAL LASER SCANNING MICROSCOPY IMAGES CAN THE VISUALIZATION SYSTEM HANDLE? Image Data: How Are Image Values Represented in the Program? Storing the Image Values All digital microscopes produce image data values from analogto-digital (A/D) converters. These will be packed into one or more 8-bit byte values, each of which may have an uncalibrated intensity range of 0 to 255. Greater collection precision is possible by integration, averaging, or other high-dynamic-range filtering operations. However, improvements in electronics and detectors now make possible the direct acquisition of l2-bit, l6-bit, or even higher precision digital data. Single-byte storage is more efficient, and is adequate, for the majority of purposes, particularly for results from low photon-flux imaging ofliving cells. It is supported by all packages. Some instruments allow 16-bit storage (a range of 0 to 65,535). Intensity data digitized by 12-bit AID converters (standard in most current LSMs) is usually packed and unpacked into 16-bit words by the loss-less conversion: Il6
=
Il2
x 16 + 15
This slightly cumbersome conversion is necessary to correctly rescale values in the range 0 to (2" - 1) without any rounding errors. The 16- to l2-bit operation is rapidly achieved by bit-shifting the binary l2-bit values towards the high byte and filling the additional 4 bits with 1 s. This operation can be precisely reversed for any integer value in the range. Sixteen-bit processing of original 8-bit or l2-bit data may also be desirable for derived images such as those from some fluorescence ratio experiments. However, this is excessive for the majority of confocal fluorescence images, which seldom record more than a few hundred photons/pixel and therefore have 10% to 20% shot noise (see Chapter 2, this volume). Microscopy image pixel values and views are economically represented by integer values. Permanent floating-point storage is rarely supported. Floating-point calibrations of integer data are discussed in the following section. A distinction should be made between storage and display precision. Historically, some digital microscopy systems have used the display memory as both a recording and a view buffer with a built-in image or signal processor. Current approaches use a virtual display in main computer memory, which is copied to the display view, allowing decoupling of data and display view and greater storage precision than video memory if required. This is necessary, for example, for storing and displaying intermediate results in image restoration programs. Image processing systems developed for cross-platform compatibility (see Amira, Image/volumeJ, and AVS examples of packages running over four platforms) have always used virtual displays allowing arbitrary precision images and views to be manipulated with as little as 5- or 6-bits of display depth per primary color. The price of this flexibility used to be a significant reduction in interactive visualization and display speed, caused by the loss of direct processor access to the video memory. One solution is to isolate platform-specific accelerations and link them to core routines when porting to high-performance workstations with non-standard hardware. Although this approach allows the rapid introduction of new proprietary hardware, it has now been almost universally superseded by the use of agreed platform-independent hardware standards with a defined software interface. Of the
several contenders for a universal graphics standard the clearly adopted winner is the OpenGL scheme [see http://www. OpenGL.org and Woo (1999) for details of the OpenGL software programming interface]. This evolving scheme adds definitions for the handling of new technologies as they are introduced into each newly released OpenGL compatible display card or system.
Calibrating the Image Data Values Multi-dimensional microscopy instruments provide the means for obtaining accurate and repeatable quantitative measurements. All parameters including calibration must be linked to the corresponding image for inclusion in subsequent processing stages. A discussion of file formats follows the section on image dimensions. Software packages normally use their own internal calibration structures because most of the general or so-called standard image formats do not support all the parameters necessary to fully describe multi-dimensional microscopy data. It might be thought desirable to store directly calibrated real number data values. A fixed-precision mantissa and exponent would certainly provide consistent relative accuracy, regardless of the magnitude of the data values. Constant precision could, however, be maintained by using a logarithmic digitization (or detector) response. This is consistent with the fact that the presence of shot noise means that, if gray levels are separated by one standard deviation they must become wider as the signal level increases. More bits would then be assigned to low intensities and less to brighter values. A fixed precision (log) calibration could then be attached to the 8- or 16-bit integer data values. The minimum requirement is a floating-point offset (the real value of a 0 pixel), an increment (the real increment/pixel value), and at least a text label or key for the linear parameter represented [e.g., log (intensity), concentration, pH, etc.]. Nonlinear changes require a look-up table (LUT) for calibrations. Multiple-channel images require separate calibrations for each component. Ion imaging data need at least a fixed precision calibration and often a sigmoidal scale (defined by R min , Rmaxo and K; e.g., Bolsover et al., 1995). Table 14.3 summarizes data value calibration, arithmetic, and measurement requirements for a multi-dimensional visualization system.
What Dimensions Can the Images and Views Have? Programmable scanning capabilities of all LSM instruments, motorized focus and/or xy-stage control of any microscope, and spectral or time-lapse capabilities yield images with a number of spatial, temporal, and other dimensions. Point-scanning LSM instruments normally acquire a temporal (sequential) and spatial (line) scan in the x-axis, repeated at further time points and optionally at progressive y- and/or z-axis positions. Hence, spatial and temporal sampling dimensions are simultaneously generated. In this way, xy, xt, xz, etc., 2D sections and xyz, zyt, xzy, xzt, etc., 3D volume images are collected. Time-lapsed volumetric (e.g., xyzt, etc.) or multi-channel spectral (e.g., xyzc, xyct, etc.) are examples of 4D images. Once considered no more than a curiosity by biologists, new dimensions of data are becoming routine. The possible five-dimensional (5D) (x, y, Z, t, c) imaging space can now be augmented with xy- (stage) position (ox, oy, oz), spatial rotation (9, , v), lifetime ('t), polarization angle (P), polarization anisotropy (r). This makes 3D to 6D data (from 12 or more possible dimensions available on a given system) a routine target for data management. Visualization systems need to support multi-channel images (Tables 14.3, 14.4, 14.6). Although ultimately, each channel is processed separately and the results merged together for display,
Visualization Systems for Multi-Dimensional Microscopy Images. Chapter 14
287
TABLE 14.3. Overview of Image Data Handling Features for Visualizing Multidimensional Biological Microscopy Data Data handling feature
Parameter
Minimum required
Desirable additional enhancements
'Data storage
Types Bits Channels
byte 8,24 (3 x 8) R,G,B, included, Merge function Linear, Offset, Range 20 Point, 2D Line 2D Arbitrary area Summed area volume 2D histogram ASCII file output +,-J,* logical, Contrast/gamma mapping, Manual z-weighting
Integer, fp, real 12116,24 (3 x 8), 36/48 (3 x 12116), n x 8 Arbitrary no. of channels n-channel merge Non-linear, Log, Sigmoidal, arbitrary 3D point, 3D line, trace, 3D area (surface), Arbitrary 3D volume, Results histograms, DDE to Excel Trig functions, Log Auto z-weighting
hCalibration of intensities Intensity measures (distribution of pixel values)
Math operations
'All systems support 8-bit (byte) data types. A few allow higher precision. This is useful for high dynamic range images. The use of 8-bit indexed or l6-bit "hi-color" modes for multi-channel data is now less common than 24-bit RGB support. Most scientific CCO cameras and LSMs now support 12-bit data (usually packed into 16-bit words) but few packages support these data types for visualization. hIt is important to clearly distinguish calibration of the intensity data values from the image dimensions (Table 14.4). Calibrated intensities also allow real values of pH, Ca2+, etc. and other concentrations to be visualized.
visualization packages must now manage these parallel operations seamlessly in order to show multi-channel changes interactively. This is particularly important where interaction between the values across channels is required by the chosen algorithm (e.g., the Imaris SFP algorithm allows transparency in one channel to alter the simulated light emission from another fluorochrome channel). Image editing is required to extract (1) subregions of a large data set or (2) a structure from the complexity of surrounding features. Sub-region editing should be available through each of the many possible dimensions of the data. All these dimensions must be appropriately treated, for example, correctly calibrated, if the results are to have quantitative validity.
Image Size Maximum image dimensions should support the full resolution of the instrument (see Table 14.4). In extreme cases, several adjacent sections or even volumes may be co-aligned (by correlation and warping) and tiled together to form a single giant data set (e.g., Oldmixon and Carlsson, 1993). Generally, total image size should be limited only by the available memory. Virtual memory management provides transparent swapping of programs and data between RAM and disk. This increases run-time significantly but can enable very large data sets to be processed. Many software developers prefer to implement a proprietary mechanism of image caching or data swapping between RAM and disk, even with the built-in capabilities of the Windows family of operating systems. The best way to minimize these overheads is by careful crafting of the visualization algorithm. The plummeting price of RAM makes the use of ever more memory irresistible by the programmer, and thus inevitable by the end user.
Anisotropic Sampling Most multi-dimensional microscopes are operated with different sampling steps in two or more axes. Visualization software must produce views with correctly proportioned dimensions and preferably have the ability to expand or contract each individually (Table 14.4), for example, artificially expanding the z-dimension of an image through a thin preparation (such as a biofilm or a stratified tissue) to highlight the morphology in each layer. The most concise way of specifying this aspect ratio information is to apply a cor-
rection factor to the appropriate axis calibration. This should be done interactively so that some imaging distortions can be corrected (e.g., for a specimen such as skin with layers of different refractive index). This does not change the data values in any way and is preferable to resampling the entire data volume, which would tend to use up precious memory. When the data is subsequently processed or displayed, a floating-point z-stretch parameter (and equivalents for x, y, etc.) would correctly specify the spacing of each plane. An integer z-fill parameter represents the number of equally spaced sections to optionally add between each of the repositioned planes. These extra data values are derived by interpolation, by pixel replication or linear, cubic, or higher polynomial spline. An obvious question arises here: How real are the extra data points? A priori knowledge of the specimen and imaging system is required for an informed choice. On-the-fly data expansion during processing will conserve storage space but requires more computations. Pre-expansion, for example, during the image loading cycle, will optimize processing speed at the expense of memory. A good compromise is rapid expansion during the computation using precalculated linear geometric LUTs.
Calibrating the Image Space To make measurements, image and view dimensions must have the correct calibrations (Table 14.4). These must be updated during any resampling, zooming, and image editing. Minimum requirements for each dimension are again floating-point values for offset and increment, and an axis label. Warping conveniently handles nonlinear dimensions by resampling onto a rectilinear grid. Correction of acquisition errors should ideally be incorporated into a single intensity and sampling interpolation. These errors include: • Spherical aberration caused by mismatch between the refractive index of the immersion medium, the imbibing medium and the design of the objective. • Axial chromatic aberration (a focus shift seen with all objectives). • Lateral effects, such as chromatic magnification error. • Photometric signal attenuation and correction of geometric distortions from refraction within the sample (e.g., Carlsson, 1991; Visser et al., 1991, 1992; Fricker and White, 1992; Hell et ai., 1993) are desirable preprocessing tools.
288
Chapter 14 • N.S. White
TABLE 14.4. Overview of Desirable Image and View Dimension Parameters for Visualizing Multi-Dimensional Biological Microscopy Data Feature 'Image dimensions
Parameter
Minimum required
Desirable additional enhancements
Single plane
Full un-edited image (from camera, LSM etc), Held in RAM
Total image size
Fully sampled 3D image, 3D image in RAM 2D,3D x,y,z time 2D sub-area, 3D sub-volume, edit on slices Background normalization z-atttenuation Integer value linear Integer for large angles 3D diagonal of image View movie in RAM JPEG compression 120 views (360 x 3degs)
Unlimited Display independent Multiple images in RAM Unlimited, Display independent, Multiple 3D images in RAM, noD images in RAM, efficient caching noD, View angles, Rotation angles, Stage position, Polarization/ anisotropy, Lifetime 3D arbitrary sub volume, Edit in view, 3D cut-away, n-D ROJ
bSupported dimensions Editing the dimensions (geometric operations)
Sub regions (ROI) 'Data corrections
Z-geometry
z-stretch
View dimensions
z-fill Single view
Number of views
dCalibration of dimensions Dimension measures
Channels in view Image View On image
R.G.B x,y,z,t X,y,z,t,angle 2D Point, 2D Line, 2D histogram, 2D Arbitrary area, ASCII file output, Summed area volume,
On view
Non-linear corrections, Photobleaching, Flat field, n-D corrections, Optical corrections, Image restoration Real value Non-linear (e.g., cubic, etc.) Adaptive for chosen angles Unlimited, Display independent Multiple views in RAM Efficient caching, Efficient compression Unlimited, Display independent Multiple movies in RAM Arbitrary no. of channels All dimensions All dimensions 3D point, 3D line, trace, 3D area (surface), Arbitrary 3D volume, Results histograms, DDE to Excel 3D point, 3D line, Trace, 3D area (surface), Arbitrary 3D volume, Results histograms DDE to Excel
'In most cases tbe image data space is limited only by the amount of disk and/or memory available. The operating system and/or the application program may provide virtual memory management and disk caching. bMost packages can handle time, spectral, etc., data as for a "z-stack" but few can directly interpret time or wavelength calibrations with any meaning. 'Complex corrections usually involve sample-specific data and some pre-processing. dCalibration of dimensions should be clearly distinguished from tbose of tbe data intensity values (Table 14.3). n-D = any number of multiple dimensions.
Standard File Formats for Calibration and Interpretation While there are many standard formats, there is no universal standard currently adopted for microscope images. However, there are established imaging formats (e.g., DICOM, see http://medical.nema.org/) that are routinely used by visualization packages such as the Cedara, Vital Images, and Analyze software that fully describe multi-dimensional volume data from medical scanners. As LSM and other research microscopes become more routinely used as screening instruments and for clinical applications, it is hoped that such standards will become routine from these suppliers as well. A catch-all image input facility such as the RAW options offered by many programs allows any packed binary file with a fixed-size header to be read in. Microscope instrument manufacturers have taken one of two options: (1) developed a completely proprietary structure and made this available to other developers and users, or (2) taken an existing extendable format (such as the Tagged Image File Format [or TIFF]) and added system-specific components (e.g., for TIFF, licensed specific tags) to store the extra acquisition parameters. A problem with this second approach is the necessary proliferation of a number of variants or compliance classes of such formats. Any third-party reader program must recognize (and provide updates for) several different versions. A widely adopted alternative is to use a proprietary
structure and to provide conversion tools to import/export data via standard formats. Unsupported parameters are transferred into the program by an ad hoc semi-manual process. A flexible, industry-standard approach to image-related details is to use a conventional database to store preprogrammed fields of information (sometimes a third-party software product is used with the visualization tool - such as the ImageAccess database used by Imaris - which can manage all image and image-related files). Two types of information must be stored and linked with each image: (1) instrument-specific details describing the instrument settings used for the collection and (2) image-specific information describing the dimensions, calibration, and experimental details. The database can hold both sets of details, together with a pointer to the image data (or even the image data itself for small images). Alternatively, the database may hold just the system configurations used as stored settings or methods (e.g., the Bio-Rad Lasersharp program stores all the instrument and user settings in a Microsoft Access database). This latter approach requires a pointer to the relevant settings to be saved with the image data, separately from the database. Table 14.5 summarizes some important image and view data parameters.
Processing Image Data Less obvious than the storage representation are the data type and precision used during computations. Floating-point representations
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
289
TABLE 14.5. Overview of Desirable File Format and Image Information Features for Visualizing Multi-Dimensional Biological Microscopy Data Feature
Parameter
"Image file format
Proprietary standard
View file format
Proprietary standard
hGeneral params. stored
Size Calibration Annotation Microscope
Special view parameters stored
Minimum required
Desirable additional enhancements
Fully defined open source Multi-file TIFF, AVI (for series) Fully defined open source Multi-file TIFFIBMP, lPEG, AVI All dimensions x,y,z,t dimensions ROIs Data specific parameters Stored in image format, or ASCII file
Fully defined open source, Full range of conversions, A Universal standard! Fully defined open source, Full range of conversions, Efficient compression, A Universal standard! All dimensions All dimensions ROIs, Text, pointers All instrument parameters, Experimental parameters, Stored in image format and/or database Userlexp notes, Informatics Image filename, Links in database etc., Visualization history/log Rotation angles, etc. Algorithm name, Algorithm parameters, Display options
Notes Source image Orientation etc. Visualization parameters
"Proprietory file formats are used by most systems. "Standard formats" such as TIFF may also cause confusion as there are many different compliance classes of TIFF, so only a sub-set of the TIFF tags in a particular file may be recognized by a given reader. hSome parameters may be stored with the image data in the same file, in a separate (e.g., ASCII) file or in a database. It is important to know the whereabouts of this information if the image is to be taken to another program with the associated data intact. ROI = region of interest.
reduce rounding errors during geometric transformation interpolations. Even this requirement for floating-point representation can be partly avoided by either (I) combining several interpolation steps into a single, composite geometric and photometric transformation, or (2) increasing the sampling by a factor of at least 2 for each subsequent interpolation. This second approach is somewhat extravagant in terms of storage and will not help if the sampling is already at the Nyquist limit. The processor architecture is an important factor in determining the processing speed. Fast multi-word and floating point arithmetic is now standard in microprocessors. Despite this, some instruments, notably the Zeiss range of LSMs, use specialized, programmable digital signal processors.
Processor Performance: How Fast Will My Computer Process Images? Personal computer (PC) performance for image manipulation is constantly improving, making the specification of system performance in a text such as this somewhat pointless. However, the principal components of the computer system required can be described in terms of their relative importance to performance. At the time of writing, the Pentium PC processors are the norm, running at around 4GHz with bus speeds around 1 GHz. These are very approximately 30 times faster than 10 years ago, representing a doubling of speed every 2 years. Non-Intel processors with alternative combinations of price/performance through low power consumption, higher capacity of on-chip memory for data caching, and other enhancements appear from time to time with advantages in different applications. Alternative Intel processors, such as the Xeon, also compete in these areas and offer improved workstation and multi-processor performance. Provided the software is correctly designed, transfer bottlenecks can be reduced with a processor having at least 512 kB of level 2 memory cache. Apple Macintosh machines have undergone something of a renaissance in recent years; the current G5 is broadly equivalent to the latest Pentium devices, and still have competitive and equivalent components for efficient numerical performance and a highly optimized bus for image transfers. The current Macintosh OS X
operating systems have been significantly updated and based on Unix technology in order to take advantage of the large software developer base. Unix workstations are still a costly alternative to ever-improving PC platforms. Improving processor performance alone is still reflected in the voxel rendering performance (in voxels/second or vps) for visualization of multi-dimensional microscopy data. Improvements in other areas of the PC have been either necessary to keep pace with the processor speeds or provided enhanced capabilities directly. Hard disk drive data transfer speed can limit the speed of animations (movies) and 3D visualization when applied to large data sets. At the time of writing, so-called ultra-fast, ultra-wide SCSI interfaced devices, with capacities up to 250Gb per disk still have higher performance than IDE devices and tend to be more robust and easier to upgrade, although plugand-play technology makes this last issue less important in the latest PCs. The latest PCs and Macintosh computers can access 4 to 8 Gb of RAM. This is adequate for most multi-dimensional data sets but will inevitably still limit performance if many large data sets are opened simultaneously, especially if the software is designed to read in the entire data set and/or the system caching or swap file is inefficiently configured for the ratio of disk to RAM. Computer video display subsystems have, over the last 10 years or so, taken on more and more graphics processing operations, allowing greater optimization and relinquishing the general purpose CPU for other tasks. The display system has become a dedicated graphics processing unit (GPU) and up to 256Mb of dedicated display memory. Some entry-level PCs may still use portions of main memory, set aside for access by the display, but these are not recommended for multi-dimensional visualization. Many functions are carried out by dedicated hardware and/or firmware running on the GPU and associated devices. Depending on the sophistication of the chosen graphics system, supported operations may include: • Rapid display of 2D views for animations, etc. • Rendering of surfaces composed of triangle or other polygon primitives.
290
Chapter 14 • N.S. White
Voxel Rendering Speed 1.E+06 r--~--------=~~~------
HOW WILL THE SYSTEM GENERATE THE RECONSTRUCTED VIEWS?
Assessing the Four Basic Steps in the Generation of Reconstructed Views
l.E+04
l.E+03 SGI SGI SGI Pentium Pentium Pentium Pentium Speciat R4400 R4400 R4400 PC PC PC PC hardware (150M Hz) (l50MHz)(l50MHz) (2OOMHz) (200MHz)(200MHz) (1.5GHz)
FIGURE 14.4. Voxel rendering speed: dependence on hardware performance. When assessing a visualization system, many factors need to be taken into account (see text). Parameters for individual performance figures should be assessed with care. On standard platforms, processor speed is still an important factor. From a simple ratio of voxels processed per second per MHz processor speed (with a basic algorithm) a figure of merit can be obtained. Running the same a-blended. voxel-rendering software (Voxblast) on multiple Pentium processors gives broadly the same figure of merit as single processors when normalized for the total processor speed. The same is found for multiprocessor graphics workstations. Running on special hardware (this example is the VolumePro board) sacrifices portability across platforms but (as has historically always been the case) gives vastly improved performance.
• • • • • • •
Geometric manipulations (e.g., warping of data vectors). Rendering of voxel objects. Transparency/opacity of graphics and voxel objects. Artificial lighting and shading. Rapid manipulation of color or grayscale. Panning and zooming of the display. Texture mapping (used to rapidly render image layers onto a growing display view).
Figure 14.4 shows some approximate voxel rendering speeds that might be expected over a range of processing platforms (data corresponds to rendering speeds of Voxblast from Vaytek Inc.). Optimization of the GPU functions is controlled largely by the supplied driver software and contributes to the major differences between various hardware configurations. CPU operations are coded by the application programmer and this is also reflected in the performance of the software. The relative efficiency of these two aspects can have an important influence on the effectiveness of the package as a whole. An effective visualization algorithm for multi-dimensional data must include optimized numerical loops (particularly nested sequences) and use fast indexing into preprocessed parameter LUTs for frequently used values. The following section describes some of the optimizations that are responsible for the performance range seen between different programs.
(I) The image (or a subregion) is loaded into the data space (an area of computer memory) . Preliminary image editing, preprocessing, and/or analysis is used to define calibrated image values with known dimensions. This constitutes the input to the visualization process. Packages such as Analyze have some useful preprocessing capabilities (Robb, 1990). Alternatively, a more general program such as Metamorph, ImageJ, and Image Pro, etc., can be run alongside the visualization package for image preprocessing (or the visualization component can be added to the 2D package). (2) A view must be chosen (subject to the available reconstruction algorithms) that will produce the most flexible and appropriate representation of the image data. An intelligent choice of view can minimize the number of reprocessing cycles (see also He et aI., 1996; Marks, 1997; Kindlmann and Durkin, 1998, for attempts at computer-assisted choice of visualization algorithm). The visualization step consists of two transformations: First, a geometric orientational mapping of the image space into the reduced dimensions of the view, and, second, a photometric mapping (sometimes called a transfer function) whereby the image intensities are processed to determine the brightness of a pixel at each position within that view. (3) The display step consists of a second geometric and photometric mapping that together constitute output or matching of the multi-dimensional view into the available physical display. In practice this may consist of scaling and copying or a more complex operation (e.g., animation or stereoscopic presentation). These presentation or output options, dictated by the display space capabilities, will determine the most efficient use of screen resolution, color, animation , etc. In tum, these will influence the choice of appropriate reconstruction algorithm. (4) Dimensional loss during the visualization processing is partly restored by the display step and partly by a priori knowledge used to interpret the 2D display view (e.g., depth cues, animation, stereo, etc.). Inspection and analysis of the display view is the last step of the visualization process.
The next sections describe more details of these processes that are important for microscopy images using techniques found in the systems listed in Table 14.1.
Loading the Image Subregion Any image processing program must first open the image file, read and interpret the file format information, and then load some or all of the data values. Interpretation of the file format largely determines the flexibility of subsequent manipulations. An on-screen view of (I) image sizes (in pixels and calibrated units), (2) intensity data calibrations, (3) associated notes and (4) a quick-look reconstruction will aid selecting the required image or subregion (Analyze and Imaris, e.g., show image volumes as thumbnail pictures with a grid showing the image dimensions, etc.). The file format information will determine whether the data needs resampIing to produce correct image proportions, and interactive adjustment is essential for z-stretchlfill (see above) . Interactive photometric rescaling or segmentation (in combination with a simple volume representation) are essential to remove (e.g., set to zero or ignore) background values that would unnecessarily
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
slow down computations. Multiple passes through the data will be time consuming for large images. A single composite geometric and photometric resampling (using fast LUTs) should be combined with loading the data from disk. Data expansion options (e.g., z-fill) may be temporarily restricted to speed up computations during the exploratory stages. Flexibility and control at this stage must be balanced against the efficiency of a single processing stage. Some image preparation options are detailed in Table 14.2.
Choosing a View: The 5D Image Display Space As introduced above, the efficient use of all the available display space greatly increases the flexibility of visualization algorithms.
The 20 Pixel Display Space Pixel resolution must accommodate a reconstructed view at least as big as the longest diagonal through the 3D volume. This means that, for example, to reconstruct a 768 x 512 x 16 frame 3D image with z-stretch of 4 and arbitrary rotation, the display view that will be generated in memory could be up to 950 pixels square. For the same computation and display of 1024 x 1024 pixel frames, 1450 pixels square are required to avoid clipping. As it is advisable to display the views without resampling, these values represent the minimum display window in pixels. Processing time may dictate that only a subregion can be processed but, in general, display of a single full resolution frame should be considered a minimum. Although the display pixel range may be only 8 bits/channel, the ability to generate projected views with a higher intensity range (perhaps 16 bits) means that, for example, average projections of images with dimensions greater that 256 pixels can be made with no loss of detail. It is now standard practice to produce an output view that is independently rescaled according to the size of the display window as defined by the user. This is transparently handled by the software and display driver. However, rescaling can always give rise to unwanted aliasing effects so it is wise to restrict the display zoom to integral multiples of whole pixels. The application program, perhaps using features of the display driver, may allow an image of greater size than the physical display to be rapidly panned, giving access to display views of almost unlimited size. High-resolution non-interlaced displays are now standard for all computers. Although a single display may deliver resolutions up to 2650 pixels, it is now standard practice to provide multiple screen outputs from the best display systems. In this way, desktops of perhaps 2560 x 1024, 3200 x 1200 or more can be spread seamlessly across a pair of similar monitors placed together. The display drivers will automatically handle traversing of the mouse and program windows between the physical screens or even allow large images to straddle the entire display space. (A commonly recommended supplier of multi-view compatible OpenGL graphics cards for PCs is n Vidia, http://www.nvia.com. while on Macintosh's they are built in.)
The Color Display Space Color is a valuable resource for coding multi-dimensional information on a 2D device. It is important to ascertain the number of different colors that can be simultaneously displayed, as distinct from the number of possible colors present in the data or display view. The author prefers the generic term palette to represent the subset of simultaneously displayable colors and color resolution to indicate the full set of possible colors present in the data. Although many different graphics display resolutions have been used by imaging systems in the past, the plummeting cost of hard-
291
ware have made the use of all 8-bit displays redundant. Useful display systems will be encountered that have either 16,24, or 32 bits per pixel of display depth. It should be carefully noted that a display system may have additional memory associated with each pixel for storing other values important for controlling the display process, but here we are concerned only with the color and intensity information. A standard 24-bit display has 8 bits of memory storing 1 of 256 possible values each of red, green, and blue intensity. Additional display panes are available in 32 bit modes. With 224 possible display colors (over 16 million), a 4000 x 4000 pixel image could display each pixel with a different and unique color. There is, therefore, significantly more contrast available in a color image than in a monochrome (e.g., grayscale) representation. However, it must always be remembered that the actual red, green, and blue colors corresponding to each component are fixed by the spectral characteristics of the physical screen material. These will, in general, never coincide with filter spectra of the microscope, or even with the nominal red, green, and blue characteristics of a color camera. When more than 24 bits of data are stored for each pixel of the display view, such as for a 3-channel l6-bit color image (a 48-bit data set), the visualization software must resample this down to the available color space of the display system. An extreme example is the option in the Imaris program of having color-space computations carried out with 96-bit precision as opposed to 24bit precision, improving accuracy at a 4-fold cost in extra memory. However, at present cathode-ray tube (CRT) and liquid-crystal displays (LCD) are capable of displaying little more than 8 or 9 bits of data in each color channel.
Pseudo Color Pseudo or indexed color was used by older 8-bit displays. It is now only really important when a display system of very high resolution is implemented, for cost purposes, with 16 bits of high color pixel depth. Each of 65,536 entries (for a 16-bit mode) or 256 entries (for 8-bit modes) in a color look-up table (CO-LUT) is assigned a unique composite RGB value. This CO-LUT is an indexed palette and can be rapidly updated or modified to change the displayed colors without altering the view data values. It is the CO-LUT value pointed to by each data value that determines what RGB intensities appear on the monitor. Visual perception of color is far more acute than for intensity in bright VDU displays because cone density in the eye is highest in the fovea (Perry and Cowey, 1985). Pseudo color is, therefore, useful for presentation of calibrated 16-bit maps of image intensities, for example, from ion indicators. In general, it adds contrast to subtle changes in (1) brightness, (2) temporal, or (3) coordinate (typically z-relief) views.
True Color True color means any display element can have independent values for each of its red, green, and blue components. The simplest way of representing these in a byte-structured format is a 24-bit (3 x 8 bit) RGB voxel. Other color-coding schemes are possible. Hue, saturation, and brightness (HSB) are useful for intensityindependent color transformations. Process color space (using cyan, magenta, yellow and black ~ CMYK) is the form used for hard copy and publication. RGB values map directly into a 24-bit display with no intermediate processing. Color manipulations can be carried out by modifying the component color data directly. Alternatively, each 8-bit channel may be driven by a monochrome (R, G, or B) 8-bit indexed CO-LUT (palette).
292
Chapter 14 • N.S. White
Multiple-Channel Color Display Either color or brightness resolution (or both) can be traded for extra channels. A hue (preferably non-primary, i.e., magenta, cyan, yellow, orange, etc.) is defined by a unique RGB ratio for each channel. An intensity (brightness) scale in each hue then represents the indexed data values of that channel. All channels are combined into one true-color RGB view, adjusted to fit the display resolution, that is, each channel using a segment of the available palette. Figure 14.5 includes examples of the use of color for 3D multichannel and stereo display.
Animations As with pixels and color, the temporal display space can also be effectively used for visualization. The simplest temporal mapping is the sequential display of view sequences. Temporal range is determined by the number of frames that can be stored for rapid animation. Time resolution has two components: the frame rate (the time between successive views) and the refresh rate (how fast
a single view is updated) (Table 14.6). The refresh rate contributes significantly to the perception of smooth animation. Retinal persistence results in a screen refresh rate of ~:;;I/18th of a second being perceived as virtually instantaneous. For through-focus sequences, fading between frames is advantageous. A long-persistence display phosphor (such as on older video-rate monitors) assists this fading process for low framing rates ~4 Hz but contributes degrading blur at higher speeds. Perception of smooth motion requires high lateral resolution and visual acuity. Therefore, smooth rotation animations require (1) fine rotation steps, (2) a short persistencelhigh refresh rate display, and (3) an animation frame rate of above 10 Hz (with ~O.05 s data refresh). Hardware and software compression/expansion (see Chapter 48, this volume) are built into some display systems, allowing suitable data to be animated at up to video rate with reduced storage requirements. Both RAM-based and hard-disk-based systems can now easily provide full-color video-rate animation. Screen update is improved by a double-buffered display comprised of both a visible and a
FIGURE 14.5. Efficient use of the SD display space. The upper four images are two "maximum intensity" stereo-pairs (Lasers harp software). They were generated with -8 deg of "rotation" around the y (vertical axis) giving binocular stereoscopy through the x display dimension and similar "tilts" around x to give motion parallax and temporal interpolation depth cues using the y display axis. In these static views, the top pair tilt forwards and the lower pair tilt backwards. With few exceptions, the most efficient use of the color space is for the display of multi-channel views. Data (courtesy of Bio-Rad Microscience) is from a large confocal series from lung tissue. (A) shows a triple-stained fluorescently labelled pancreatic islet, with each channel "maximum projected" and combined into a 24-bit Windows bit-map view (data collected by T.J. Brelje, University of Minnesota, Minneapolis, MN). (B) shows a dividing shrimp embryo stained with rhodamine-conjugated beta-tubulin antibody (27 five-micron optical sections) rendered using VoxelViewlUltra. Cell bodies and membranes (blue) are assigned low opacity so that microtubule structures (orange-red) can be viewed distinctly in their cellular context. Analytical geometric data (lines and surfaces) can also be inserted, not overlaid, using the VoxelViewlUltra Embedded Geometry tool. Embedded lines highlight the axes of division and centers of polarity of the dividing cells and help indicate directions of movement during mitosis. (Data courtesy of Dr. W. Clark, Jr., University of California, Bodega Marine Laboratory, CAl. Intensity gray levels can also code for z-depth (see text). (C) shows a Lasersharp "height" view of a living fibroblast imaged by fluorescence optical sectioning of an excluded FITC-dextran-containing medium (see Shotton and White, 1989). (D) Binary (single-bit) line graphics make inefficient use of display space. However, this "YMOD" or height profile of BCECF-stained living chondrocytes from pig articular cartilage (data by RJ. Errington, Physiology Dept., Oxford University) does show relief more clearly than would an intensity view alone.
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
293
TABLE 14.6. Overview of Image and View Display Options Desirable Visualizing Multi-Dimensional Biological Microscopy Data Feature "Display 20 pixel size Color Screen
Movie
Parameter
Display mode "Multiple-channels Display refresh (Hz) 'Size (diagonal inches) dFPS (Hz) Recording
Minimum Required
Desirable additional enhancements
To show full image I : I without clipping (typically needs 1280 x 1024 or more) 24-bit ROB Standard ROB 70Hz CRT: 19 Flat: 17 -10 at full image resolution Variable fps Store movies as digital file
As large as possible, Multiple-screens, Seamless desktop
Hardware assisted Stereoscopic
Modes
Manual stereo pairs Movie pairs Anaglyph display
Animation
2 color anaglyph
24-bit of higher + overlays Arbitrary number of channels -100Hz CRT: multi 21, Flat: multi 19, Wide-aspect & Specialist Mega-pixel screens Up to about 30fps, Motion compression, Variable fps Digital video store, Digital Compression, DVD recorder, DVDRAM z-buffer video system, Compression AV quality disk system Side-by-side, LCD viewers, Shuttered screen, Projection, Non-viewers (e.g., lenticular) Standard hardware (e.g., OpenOL) Full color Auto stereo from 6deg sequence
"Most operating systems and application programs decouple the image size from the display size. If the image must be sub-sampled by the system to display in the chosen window, aliasing may occur (some graphics systems have anti-aliasing devices built in). Usually, not all of the screen area is available for the image, so the screen needs to be bigger than the largest image. Some graphics systems allow an image larger than the screen to be "displayed" and the visible part is panned around with the mouse. "With color-space mixing, it is possible to "merge" an arbitrary number of channels into a standard 3-color, RGB display. 'CRT sizes are usually given as the tube diagonal, not the visible screen size, flat panel displays indicate the actual viewing area. "The animation software must be highly optimized to deliver the fastest, smoothest framing rates. This can be achieved in software at the expense of standardization and so the fastest systems may only work with a limited range of display hardware. FPS = frames per second. LCD = liquid crystal display.
second (non-displayed) buffer. The second buffer is updated while the first is read out to the monitor and pointers to the two buffers are switched between successive frames. New graphics standards, especially OpenGL, now fully support double buffering on standard displays. Double buffering obviously needs twice the memory on the display card.
Stereoscopic Display Depth perception by stereo parallax provides a third spatial display dimension. Some of the geometric z-information can be mapped into this stereo space. Like the color/intensity range compromise, the stereoscopic depth requires some reduction in bandwidth of another component. The two halves of a stereo-pair can be displayed using screen space components normally reserved for (1) 2D pixel area, (2) color, or (3) sequences (temporal space). Each of these methods requires an appropriate viewing aid to ensure that each view is presented only to the appropriate eye. The detailed implementation of stereo presentations are described in a later section. Table 14.6 shows some image and view display options for visualization systems.
Optimal Use of the 5D Display Space Several conflicting factors must be balanced: (1) visual acuity in a particular display dimension, (2) efficient use of display resources, and (3) minimizing processing time. Because intensity variations are difficult to interpret in a low-contrast image, it is sometimes tempting to use y-mod geometric plots and other display tricks to represent, for example, a fluorescence ion concentration. Using the geometric space rather inefficiently in this way for a simple intensity display may allow us to view inherently planar information with greater accuracy. A contrasting example might be when a depth profile surface can be defined from a 3D
object: (1) the geometric space in x, y, and z (e.g., stereo) could be used to show the object's surface relief, (2) color used for material properties (fluorescence, reflectivity, etc.), and (3) intensity (grayscale) employed to reinforce z-cues by depth weighting. Automontage is an interesting application (from Syncroscopy, Ltd.) where surfaces are extracted by ray-casting projections of widefield z-focus series. The resultant 3D data is then visualized using z-profile plots or stereoscopic views. The best display space for a particular component will depend on the available resolution and the range of the data. In general, multi-channel images are best shown as different colors, in which case depth information must be coded using stereo or motion parallax or some lighting/shading mechanism. Binocular stereoscopy works for views rotated around the y-(vertical) axis when looking at the screen, that is, with parallax shifts in the horizontal x-direction (Frisby and Pollard, 1991; Poggio and Poggio, 1984) (for an upright observer!), while motion parallax is perceivable around any axis within the xy-plane. The perception of depth by so-called motion parallax is actually a subconscious interpolation of the images between each view to fill in the path of features presented at discrete loci along a simple trajectory (Nakayama, 1985; Fahle and de Luca, 1994). Therefore, a sequence of side-by-side stereopairs at increasing tilt angles only will give stereo depth in x and animation depth in y with minimum processing. Rotations are interpreted more readily by most observers than other temporal sequences, such as through focus z-series, largely because of our acute perception of parallax. Detailed assessment of particular presentation modes requires an in-depth know ledge of the physics and physiology of visual perception (e.g., Braddick and Sleigh, 1983; Murch, 1984; Landy and Movshom, 1991). Alternative temporal coding strategies, such as color-coded tixels (Kriete and Pepping, 1992) thus aid the presentation of non-rotated time series, partic-
294
Chapter 14 • N.S. White
ularly if interactive measurements are to be made. Figure 14.5 shows examples of the efficient use of display space for multidimensional display.
Mapping the Image Space into the Display Space Having chosen both the image dimensions and display space, a suitable mapping of image space to the output view dimensions must be found. Choices for this geometry processing are intimately linked to the implementation of the transfer algorithm for combining the data intensities, however, it is more useful to consider these components separately in order to make the most of available resources ; geometry and intensity processing software and hardware may reside in different subsystems of the visualization workstation. For a general multi-dimensional image I(xi, Yi, Zj, tj, Ci, ... ) and view V(xv> Yv> Zv> tv> Cv> ... ) we can define an overall reconstruction function (R) such thaI (1)
where x, y, Z are the spatial coordinates, t is time, c is color channels, etc. R has two components: a geometric transform (G) converting image to view dimensions and a compositing (sometimes called transfer function) or projection operation (P) performed on intensities through the view space (Fig. 14.6). These components of R are thus related by
v = P(xv, Yv, Zv. tv> Cy, ... ) where Xv = Gx(Xi, Yi, Zi. Ii, Ci, ... ), Yv = Gy(Xi, .. .) etc.
(2)
The following sections describe various G functions used in visualization systems (listed in Table 14.7). Projections are described later (and listed in Table 14.8).
Simple Visualization: Reducing the Geometric Dimensions This involves discarding all but two of the voxel coordinates and mapping the remaining dimensions to screen xy-positions. A non-rotated orthoscopic (non-perspective) view of a 3D (Xi, Yi, zJ volume (viewed along the Zv-axis) is a simple geometric mapping of serial sections that can be projected (e.g., by summing, maximum intensity, etc.) onto a 2D orthogonal (XY) v display (Fig. 14.6). The G function is defined by Xv
= Xi, Yv = Yi,
Zv
= Zi. i.e.,
V(x" Yv)
= P(Xi, Yi,
Zi)
For a 3D time series (Xi, Yi, t;) to be viewed edge-on, (x;, transformation is equally simple: Xv = Xi, Yv
= tj,
Zv
= Yii.e.
V(x" Yv)
= P(Xi, t i, Yi)
(3) ti)'
the (4)
These are the basis of the familiar three-pane orthogonal section views found in many visualization programs.
Rotations A single, orthoscopic, x-axis rotation (9) requires a geometric mapping given by Xv
= Xi,
Yv
= Yicos9 + zisin9,
Zv
= -Yisin9 + Zicos9
(5)
then The projection is then performed in the (re-oriented) view coordinates. Offset coordinates (i.e., Xi - x o , etc.) are used for rotations around a point (x o, Yo, zo) other than the image center. Rotations are non-commutative, that is, the order of X-, yo, and z-axes matters. The observer 's coordinate system (Xv. y" zv) is static, and all orientations are given in this frame of reference. To generate a view of the image rotated simultaneously about the three image axes (i.e., a tumbling algorithm), three consecutive transformations are required. The first [e.g., a tilt around Xi. Eq. (5)] is obviously
.\'r;r-. '+ ~ · l·
:>
f-=
equence of 2 0 pseudo-color or RGB di play views V (X .. y ,.I.)
= P(x'!', y,!" 2'!').
+ Se] + (6) b
2
)]
+ gcn-.!ml rotation
+ y [be( l
- C) - Sa]
ruus
+
.h,c,
where S is the sin(twist angle), C is the cos (twist angle) around an axis (a, b, e), and a is the COSe,wiSl, b is the COSct>,wiSl' e = CO S'f',wi", e , ct>, and 'f' are the view space polar coordinates of the twist axis (Fig. 14.7). This whole transform must be repeated for each of the three re-orientation axes (a, b, e)xo (a , b , e\., and (a , b, e),. A viewing transform should ideally be added to observe the rotated structure from different viewing positions. However, by fixing the view direction, this additional step is avoided. The efficiency (and ultimately, the speed) of the geometric algorithm is determined by the degree to which the general form (6) can be simplified. If all rotations are specified around the observer 's axes (xv> Yv, zv) the direction cosines (a, b, e) become either zero or unity. It is also easier for the user to anticipate the final orientation when the rotation axes do not change between each component twist. For a fixed viewing axis along 2" a tilt 9 (around x v) is obtained by a, b, c = I , 0, 0 giving
(7) where S is sin9 and C is cose. When combined with a subsequent q> rotate (around Yv) this becomes (8)
o
y,
A 3- D ROlation Mound a general axis
y.
B )-D Til! around x, = x,
'.
] · 0 ' Rotale' around )',
FIGURE 14.7. Rotation transformations. 3D rotations can be performed in the image-space around a general axis of rotation (A ). General re-orientation requ ires rotation around three axes x" y, and z,. A simpler (and thus faster) scheme is to use a "tilt" (B) fo llowed by a "rotate" (C) around x, and y, in the view-space. If viewing along z", no additional information is obtained by a z, "turn ."
296
Chapter 14 • N.S. White
If the twists are "tilt first, followed by rotate" and the viewing axis is oriented along x" Yv, or zv, the twist around that axis can (optionally) be omitted because no more structure will be revealed (Fig. 14.7). By thus omitting the tum (9) and projecting (and viewing) along zv, the resultant (xv, Yv, zv) vectors only change as a function of Xv and Zv (i.e., dyjdx i = 0) while traversing a row of image data. This can lead to extremely rapid projection algorithms. The above scheme is used for the 3D visualization routines in Lasersharp, which are executed entirely on the CPU without any involvement of the graphics hardware. Other tricks can be used to optimize the rendering geometry.
Working in Image or Space View One implementation approach is to progress through the image space (I space) coordinates (in Xi, Yi, Zi), voxel by voxel, transforming each to view space (V space) coordinates (in Xv, Yv, zv) before projecting each intensity into the final 2D view at (xv, Yv). For an image of llXi x nYi x nzi voxels, the orthoscopic transformation proceeds via N serial planes or sections of data, normal to Zi in forward (n = 1, nzi) or reverse (n = nzi, 1) order. Alternatively the same G function can be implemented in V space. There are now N planes normal to Zv (and thus cut obliquely through the image volume). For each V(xv, Yv, zv) the contributing I(xi, Yi, Zi) are found and the computation can again proceed in forward (n = 1, nzv) or reverse (n = nzv, 1) order. The V space implementation makes the G function more difficult but facilitates straightforward projections (P) (Fig. 14.8). The I space implementation is the reverse. The V space method has many more advantages for geometric polygon data than for rastered voxels. Hybrid implementations are also possible, combining both I and V space components. In a hardware-accelerated OpenGI implementation, a favorite approach is to take each image plane (i.e., to progress through I space) and warp it into its projected (rotated) aspect in view space. The warped frame is then mapped onto the view layer by layer. Sophisticated graphics cards have hardware for texture mapping that assists the painting of images onto objects using this technique by altering the view pixels according to the color of the image. This texture mapping hardware can be used to paint the warped frames into the growing view modifying the output by opacity values controlled by the voxel intensities. Hardware texture mapping is used extensively by Volocity and Amira packages. In all these examples the reconstruction process has resulted in a loss of Zv dimensional information. Some of this may be automatically retained by the P function (discussed below), but an efficient G function can further optimize the dimensional content of the view.
x,
y,
A I-space projection via serial x;y; sections
x,
y,
B V-space projection via serial xs, sections
FIGURE 14.8. V- and I-space projections. I-space voxel reconstructions CA) proceed via serial X;,y; sections that are oblique to the viewing axis. This is the most efficient way of processing voxels in "object order" and is used by most voxel renderers. V-space algorithms (B) process x"y, sections that are normal to the viewing axis. This z,-ordered reconstruction is more useful for constructing polygon objects than for voxels.
D = 0.05 - 0.07m, d = 0.25 * x/D m, s = 2 * arctan[(0.5 * Did)]
for example, if D = 0.06, x = 0.06, then d = 0.25 and s = 13.6°. A simpler alternative to the computationally intensive rotations is to shift each section horizontally (in xv) by a constant factor of the Zv coordinate during a top-down compo siting projection (3). This corresponds to a stereo pixel-shift G function Zv = Zi
How Do 3D Visualizations Retain the z-Information? True 3D visualization requires that Zv depth information is retained or at least partially restored when the image is mapped into a 2D view.
Stereoscopic Views Some impressive methods of coding Zv information use stereoscopic techniques (Tables 14.6 and 14.7; Figs. 14.9 and 14.10). These require that an observer be simultaneously presented with a left-eye and right-eye view differing by a small rotation. The author has found the following geometry to give acceptable results: for views of width x, angular difference " viewed at a distance d with an interoccular spacing D
(10)
(11)
where () Xleft = tan(0.5 *s) and ()Xnght = -()Xleft. From Eq. (11) and the viewing conditions of Eq. (10), ()x is ± tan (7.8°) or about 0.14 times the z-spacing (Fig. 14.9). A slightly different equation is described by Cheng and colleagues (1992) and its derivation is attributed to Hudson and Makin (1970). It can be used to derive an optimal pixel-shift. Using a notation consistent with the above 8p = 2*nZcalib * M * sin(arctan(8x*nz/nzcalib)
(12)
where 8p is the parallax shift between equivalent points in two views (ideally kept around 3-5 mm); nZealib is the the calibrated zsize of the data (thickness of the original specimen), and M is the magnification. The pixel-shift method is only an approximation. In particular, the result is stretched out or warped in the x-direction by a small
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
" y,
(i) CCW (+ve) rotation around y, for left eye view
(ii) Right eye view with 'Tilt' only (around x,)
A Stereo pairs from rotated views
·vc pix~1 shifl
lor len eye \'il!w
x. Y.
B Stereo pairs from pixel-shifted views
FIGURE 14.9. Pixel-shift and rotation stereo. Pairs of views differing by typically around ±7-8deg (depending on pixel dimensions and inter-plane spacing) of y, rotation , (A) can be extracted from a sequence and displayed using anaglyph , switched-view or other stereo viewers (see text). The pixelshift approximation to these rotations (B) results in only trivial distortion and is much faster to implement for small angles about the Zv axis. See Figures 14.5 and 14.10 for example images and the section on 3D depth information for details of stereo generation algorithms.
factor of l/cos(0.5 * s), but for small angles this can be neglected. This shearing and warping algorithm is the basis for fast texturemapped voxel projections (e.g., Cabral et al., 1994; Lacroute and Levoy, 1994; Elisa et aI. , 2000). The two views, or stereo-pair, must be fused into a 3D representation by positioning each in the display space so as to occupy the entire field of view (or at least the central region) for each corresponding eye. The observer then perceives the combined information as a single 3D view located near the viewing distance d, depending on the origin of the Zj-coordinates used in the above equation. For zj-coordinates centered in the middle of the image volume, the view depth is centered about d. A few high-contrast features must be present in the field in order to successfully fuse the reconstruction. This makes the choice of algorithm and associated parameters critical for successful stereo presentations. The stereo scene geometry is efficiently accomplished by placing the views in two halves of a display, bisected in x or y, and using a viewing aid. The viewer must make the two views appear to be coming from the center of the field while keeping the observer's ocular convergence/divergence near to parallel (i.e., relaxed as
297
though viewing at a distance). By carefully editing a vertical subregion (e.g., 384 x 512 pixels) of interest, side-by-side pairs can be viewed in full color on a horizontal display format (e.g., 768 x 512, 1024 x 768, etc.) without subsampling. Side-by-side stereopairs are easily viewed with horizontal prismatic viewers. An alternative is to fix one's binocular convergence point at infinity and refocus each eye onto the respective view (using the lens and cornea only). The left and right views may also be swapped and the eyes crossed or converged to a point between the observer and screen. Some seasoned stereoscopists can fuse stereopairs using these methods without additional aids, although prolonged viewing can give rise to eye strain and headaches. Above-and-below pairs can be seen through vertical prismatic viewers, but this geometry cannot be so easily fused by the unaided observer! Partitioning the color space into two distinct regions allows full size anaglyph stereo-pairs to be observed in monochrome (gray levels only). Both views occupy the entire pixel display area and are transmitted to the corresponding eyes by RG or RB spectacles. Due to the spectral characteristics of some monitor phosphors and the availability of low-cost filter materials, some bleed-through of signal between RG channels often occurs. Red/cyan viewers often have improved extinction. The optimal intensity balance between the component views must be individually detemlined. Anaglyph stereo-pairs can be fused by observers irrespective of their capacity to differentiate colors. Even rare red/green color blindness is no obstacle provided sufficient intensity levels can be differentiated, although it is usually found that around 10% of observers fail to perceive 3D effects from stereo cues alone (Richards, 1970). Because the anaglyph views occupy the same physical area, the eyes are drawn into a convergence naturally. The monitor should be as large as possible to increase the distance from the viewer and decrease eye strain, that is, so the convergence angle is as small as possible but each eye is focused at a distance. Health and safety recommendations usually specify at least 18" comfortable viewing distance for VDUs (more for extended viewing) with alternative work breaks every hour. In order to maximize simultaneously spatial and color resolution, the temporal display space can be partitioned to display the component stereo-pairs. This requires more sophisticated hardware than the previous methods at increased cost. The left and right component views are displayed alternately on the video monitor in rapid succession. Observers watch through a viewing device that synchronously blanks and unblanks the visual field of each eye while each corresponding view is displayed. These alternate (or switched) display stereo systems are characterized by differences in the format of the video signal, the switched viewing hardware, and the method of synchronization. Older video stereo systems display images in two alternate and interlaced fields of each video frame. This method gives half the temporal and y-axis resolution of a normal video signal, an obtrusive (25/30 Hz) display flicker, as well as the low intensity associated with interlaced displays. Computer displays are exclusively non-interlaced but alternative frames may be used to show stereo components rapidly in succession . An interlaced implementation will require the two component images to be interleaved line-by-line in the display memory. A more convenient organization is to have sequential buffers for the two fields which can be updated independently. Non-interlaced computer displays give a brighter view with a flicker-free refresh rate of 60 to 90Hz. This would allow alternate frame non-interlaced stereo, but would still exhibit noticeable flickering around 30 to 45 Hz. A continuous stereo display requires at least 120 Hz and preferably a double-speed scan of 180 Hz.
298
Chapter 14 • N.S. White
FIGURE 14.10. Example stereo images. (A) Anaglyph (two color, red/green) stereo-pair of Oolgi-stained nerve cells, generated by the pixel-shift method (see text). The alternative method of extracting two images from a rotation series [shown immediately below (A) as a side-by-side pair] produces a virtually identical result to (A) for small angles about the z-axis. (B) Anaglyph stereo-pair of fluorescent Paramecium. [Data for (A) and (B) courtesy of Bio-Rad Microscience.] The third pair of panels shows a stereo-pair of voxel-rendered images representing Feulgen-stained chromosomes from the plant Milium. (Original data from Bennet el aI., 1990.) Voxel gradient lighting is used with a high-opacity a-blend algorithm to give an effect of "solid" structure. The final panels show a triplestained thick section of skin showing extra cellular matrix proteins and vascular structures. Simple maximum projections of each channel z-series are combined into the 24-bit ROB view. While the anaglyph stereos can span the entire display window but require the color space for the stereo effect, these full color sideby-side stereo pairs (as for monochrome pairs) can span only half the available display resolution. These alternative stereo display methods illustrate the way that display resources can be "traded" for improved (x,y) resolution, z-depth or multi-channel (color) rendition. Side-by-side pairs in this figure should be viewed by divergent eyes or viewing aids (i.e., with eyes relaxed as when viewing distant objects but focused on the page). This is best achieved by focusing on a distant object and bringing the page into view while keeping the eyes relaxed but refocused on the page. If the eyes are crossed or a convergent viewing aid is used, the 3-D effect will appear "back-to-front." This can be problematic with some maximum projections (e.g., the Oolgi-stained neurons above) where some brightest features are towards the "back" of the rotated object in some views. This is often found when there is attenuation with depth in the sample. Maximum intensity can incorrectly bring these features to the "front" of the view. Samples with more isotropic intensity distributions with depth (e.g., the skin images above) tend not to show these anomalies.
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
Some high-resolution stereo displays still use an interlaced storage mode (at half the y-resolution) for stereo presentation. All video display systems suffer to some extent from flicker arising from an interference with mains-powered room lights and should be run in reduced ambient lighting. Synchronizing the display to the switched viewing system can be accomplished in a number of ways. A bright intensity marker in each video frame can identify the left and right views, and this has been used to trigger a photo switch placed over the comer of the video monitor. This switch controls synchronous LCD shuttered viewers. The synchronizing pulse can be generated directly from the field or frame synch of the display signal and passed to the viewers by cable or optical (infra-red) link. A more convenient alternative to switched viewers is the polarizing LCD shuttered screen. This is a large LCD shutter (often constructed from two parallel units for faster switching) that toggles between two polarization states. The left- and right-eye views are thus differently polarized and can be viewed through inexpensive glasses. Planepolarized shutters (and glasses) give the best extinction, but clockwise and counterclockwise rotary polarization allows for more latitude in the orientation of the observer's head (important when a large group is viewing a single monitor!). Example stereo views are shown in Figure 14.10 (and also Figs. 14.5 and 14.24). The Imaris package has a number of selectable stereo modes depending on the type of viewers to be used (these include Raw Stereo - using the OpenGL functions, alternate image (interlaced modes), and three combinations of two-color anaglyphs). The 3D-constructor plug-in for ImagePro Plus (Media Cybernetics Inc.) also supports OpenGL stereo.
299
vp,
vPo
y
(i) Single vanishing point (vPa) for view-space
(ii) Two vanishing points (vPo, vp,) for image-space
A True perspective geometries flX,
"'ID 7 XY section
XZ section
B Isometric geometry
FIGURE 14.11. Non-orthosopic views. True perspective views (A) give the most "realistic" reconstruction geometries. A simple approximation can be implemented in the view-space (I) with one "vanishing poinr." x" Yv coordinates are reduced by a linear function of Zv between Zo and z,(see text). A general perspective G-function for image-space implementation (2) with rotation around Yv requires two vanishing points. Isometric reconstructions (B) retain the physical dimensions of the original image axes (rather than the trigonometrically projected axis lengths). 3D distance measurements can thus be made directly on the views.
Non-Orthoscopic Views Visual perception of depth makes extensive use of nonstereoscopic cues, and these can be coded by appropriate algorithms into a corresponding part of the display space. A series of G functions exist that code depth information with unmodified intensities. These are the non-orthoscopic transformations and include both perspective and non-perspective geometries (Fig. 14.11). The most straightforward of these algorithms require that the corresponding view coordinates for each data voxel are modified by the image z-coordinates. The isometric G function involves a constant shift in screen Xv and/or Yv coordinates as the I space renderer traverses each dimension of the data Xv
+ XiCOSqJ, Yv = Yi + YisinqJ + Zv = Zi and usually qJ = 60°
= Xi
Zi,
(13)
This geometry (as the name suggests) gives rotated X" Yv and axes in the same proportions as the original image axes (for rendered examples, see Wilson, 1990). Direct Xi, Yi, Zi measurements can be made from the 2D view screen. True perspective views attempt to visualize the data as a large real-world or macroscopic object with Xv and Yv converging to a vanishing point at a large Zv distance. This point is on the horizon (usually in the center of the field). The perspective G function decreases the dimensions as a linear factor of the z-coordinate. So after the data have been rotated, the projection is accomplished through a new perspective space (xp, Yp). True perspective can be approximated in the G function by Z"
(14)
where a is a factor reflecting the object-to-observer distance. Perspective views can be readily interpreted from objects with well-defined structures. This is because we assume such views
arise from connected features such as real-world surfaces and edges. Popular optical illusions indicate this is not always true! Connectivity between neighboring voxels in a confocal laser scanning microscopy (CLSM) fluorescence image can rarely be assumed without substantial a priori knowledge and then only after careful control of noise. Non-orthoscopic depth coding is useful for removing ambiguity from stereoscopic views. Volocity and Imaris produce both orthoscopic and perspective 3D (and 4D) views.
Temporal Coding and z Depth Time axes can often be treated in exactly the same manner as depth or z-coordinates. This is a useful observation because many visualization packages do not specifically recognize time points. Thus t-coordinates can be directly mapped by animation, coded in the intensity or color space, and even represented in non-orthoscopic or stereoscopic views. The importance of time points as a component sampling space in 4D imaging is encapsulated within the socalled 4D tixel object. Significant efforts have been invested in the efficient visualization of tixel images within the 5D display space (Kriete and Pepping, 1992). z-Position can be inferred from the display temporal space by animating serial sections as a throughfocus sequence or by motion parallax (Fig. 14.12) from a rotation series (Wallen et al., 1992). Since xy-acuity improves smooth motion perception, large rotated views require reduced angular increments as well as higher refresh and framing rates compared to those from smaller volumes.
300
Chapter 14 • N.S. White
lected by the microscope objective. Digital processing to more closely realize this goal for CLSM imaging requires a knowledge of the microscope transfer function (e.g., Sheppard and Gu, 1992, 1993) and the use of digital restoration methods (e.g., Holmes and Liu, 1992; Shaw and Rawlins, 1991; see also Chapters 23 and 24, this volume). For now, the 3D image is considered as the physical or macroscsopic object, itself being imaged, as if by a lens or eye, to a 2D view. Light rays from each point voxel in the image would contribute to an equivalent (but not unique) point in that view. This mapping of object points x image voxels x display view pixels is the crude model. This is refined and made more elaborate by (1) segmenting the multi-dimensional image data into a set of objects, (2) considering other objects in the path of each simulated light ray, and (3) adding additional, artificial lighting and surface effects. These aims are met using physical or material properties attached to the objects.
B
y,
Z-Depth perception by temporal interpolation and motion parallax FIGURE 14.12. Z-depth by temporal interpolation and motion parallax. If a similar feature is seen in two sequential views of a rotation series viewed along z, (e.g., the dark shape appearing at "1" in the first view and at "2" in the second), two processes contribute to the perception of its depth within the view. (I) The direction of motion, either left or right across the field of view will determine whether it is perceived to be behind (A), the center (C), or in front (B) of the screen. (2) provided the views are at small angles apart and shown in a "smooth" sequence (which may need up to 18 fps), the details are mentally interpolated in time (taking into account the perceived motion) to "fill in" the missing information (dotted features). For a fully transparent view with no other depth cues, front and back may be arbitrarily reversed by different observers.
Mapping the Data Values into the Display As with the G function, data values must also be transformed to the display by a well-defined operation. This is the combination, compositing or projection function P, described in Figure 14.6 (examples in Table 14.8 and Fig. 14.14).
The Visualization Model It is useful to consider even the simplest P function in terms of a lighting model. In a gross simplification, each image voxel is considered to have a brightness equivalent to the amount of light emanating from a corresponding point in the specimen and col-
Choosing the Data Objects So far, we have considered the image as an array of voxel (3D) or tixel (4D x, y, z, t) objects with no implied connectivity. Computeraided macroscopic imaging and display systems have fostered the growth of visualization models based on the grouping of voxel samples into geometric objects. The consequence is a growth in special hardware and software for efficient treatment of vectorized geometries. Microscope systems invariably produce rastered image arrays. The array dimensions (nx, ny, nz, etc.) imply the arrangement of rows, columns, sections, etc. Vector objects are non-ordered lists of geometric figures, each of which is defined by a very few parameters. Thus, a simple voxel might be five values (x, y, z, intensity, opacity, etc.) but a triangle comprising many tens or hundreds of voxels may be specified with only 10 to 15 parameters (three sets of coordinates plus some material properties). If the material properties vary across the polygon, values at each vertex only need be stored. Each vectoral component will need at least 16-bit integer precision and preferably floating-point. A compact form is to store all vertices in one list or table. Each geometric object is then a set of polygons, specified by indices pointing into the vertex table. While early workstation programs stored voxels as vectors, the availability of optimized hardware and software for G functions and 3D rendering of rastered byte images has assisted the devel-
TABLE 14.8. Overview of Image-to-View Data Mapping (Projection) Options for Visualizing Multi-Dimensional Biological Microscopy Data Feature
Parameter
'Data objects
types
Voxels
storage
3 D ras tered array Basic filters, Volume cut-away Region of interest/edit, Intensity segment, Color channels
Selective enhancement of features
bprojection algorithms 'Special graphics resources
Minimum required
Maximum Average VGA graphics
Desirable additional enhancements noD array of voxels, Object list, Surface (triangles), Surface (polygon net) Objects "embedded" within voxels noD rastered array, object vector list Gradient segmentation, Image masking, Opacity/transparency Material properties , Morphological filters, "Seed/flood fill" Object modelling Front, Z-co-ordinate, a-blend, SFP, Iso-intensity surface, "Local" projection OpenGL with acceleration, Texture mapping, Hardware shading/ lighting etc., a-buffer
Data objects are usually voxels or lists of vertices defining geometric surfaces. Other surface descriptions are possible. AVS allows for a particularly large range of geometries, while in most object reconstructions, the options are usually restricted to triangulated surfaces - primarily to make use of efficient accelerations in the graphics system. h Some geometric object renderers use Front or a-blend P-functions in the same way as for voxel objects. Specific "surface" algorithms may also be implemented that capitalize on the connectivity of the geometries. There is always a trade-off between the portability of a program across platforms (or over new versions of platforms) and the use of non-standard "special" hardware. The OpenGL standard largely avoids these problems by setting the interface requirements and providing standard support for new hardware. SFP = simulated fluorescence process. a
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
opment of voxel visualization algorithms. The most efficient way to process rastered voxels is by whole-image operations. I space implementations effectively warp successive orthogonal (xy, xz, or tZ)i sections, using a transformation geometry engine, to their projected shape in the final view. Rows and columns of data are traversed in forward or reverse order to preserve the forwards or backwards compositing direction. The only drawback of this method, as opposed to a z-ordered list of polygons, is the throughput to the display. Because the entire image data must be streamed into the display, bandwidth is critical. This requires highly efficient pipelined operations to achieve a display rate sufficient to service the output from an optimized geometry engine. The highest specification systems use multiple display devices (often video projectors) to simultaneously render multiple views for immersive reality installations. So-called multi-pipe versions of these visualization programs are used to massively increase the data throughput to the display. The vectoral representation is more efficient for triangles encompassing more than about 20 equivalent voxels, provided the segmentation algorithm is justified for the particular data. Computer models allow entire surfaces and bounded volumes to be described by single parametric equations using, for example, Bezier coefficients (see Watt, 1989). This type of simple object definition is generally not practical for confocal fluorescence data due to the low signal-to-noise and discontinuities in antibody or other stains. As a result, many vertices must be individually stored. Vectored object lists can be traversed in I or V space in object order. Thus, complex figures can be rotated and projected individually. Refinements in geometric object technologies may appear to place voxel rendering at a disadvantage due to its ordered processing of many more data points. Choice of data objects is largely determined by the amount of information known about the specimen, the sophistication or realism required in each view and the availability of appropriate hardware and software on the chosen platform. It should now be apparent that there is no fundamental distinction between any of the data objects discussed. A trade-off between processing speed and transfer rate must be weighed against the problems or bias associated with segmentation of the image data into parametric structures. In any case, voxel rendering speeds are now equivalent to the speed at which vector graphics objects could be drawn just a few years ago. All of the compositing or P function rules described below can be applied to any data objects. Sophisticated graphics systems are now relatively inexpensive and are supplied with even modest-specification PCs. Some include hardware accelerated geometry-processing engines, with standardized OpenGI interfaces, intimately linked with the GPU and display memory together with additional buffers, LUTs, etc. Thus many of the above operations are now readily available to any software developer writing for a standard hardware platform.
301
which the image is rapidly mapped. Morphological segmentation requires a selection based on some connectivity rule such as the geometric surface objects extracted by the Marching Cubes (c.f. VoxBlast) or Delaunay Triangulation (c.f. Visilog) algorithms. Three-dimensional intensity gradient filters can also find or enhance boundaries between geometric objects. Cheng and coworkers describe the use of such a filter to extract a gradient image, which is then blended with the original to enhance edges. Using the earlier notation lout(Xi, Yi, Zi,)
= lin(Xi, Yi,
Zi,) [k
+ (I
- k) grad(arctan(Iin(xi, Yi, Zi,»] (15)
where the arctan is an (optional) approximation to a sigmoidal background suppressor and grad is the unsigned 3D gradient grad(xi, Yi, Zi) '= I[(dl/dxii + (dIldyJ2 + (dIldzJ 2 ] This 3D gradient is simply the resultant of the unsigned component gradients along each axis. These individual components are conveniently approximated (for low noise data) by subtracting the values on each side of the voxel whose gradient is required, that is, grad = (1«I(Xi + I ,Yi, Zi) - I (Xi - I, Yi, Zi»2 + (I(Xi, Yi + I, Zi) I(Xi, Yi - 1, zi)f + (I(Xi' Yi, Zi + 1) - I(Xi' Yi, Zi - 1»2»/2. Gradient filters are also extensively used to provide data for the realistic artificial lighting effects (discussed later). Other background filters are based on patch convolutions with kernel weights given by smoothing functions. Forsgren and colleagues (1990) use a 3D Gaussian filter where the mask weights are given by: F(x, y, z) = 1I(cr'2nl(2n» exp (- (X2
+ l + Z2 )/cr 2 )
(16)
different strengths are specified by the width term cr. This filter is separable and is readily implemented as a sequence of ID filters, allowing for asymmetric sampling in the multi-dimensional image. A general approach is to use a 3D filter (or ideally a PSF deconvolution) with a cut-off at the Nyquist frequency (see also Chapters 4 and 25, this volume), followed by a threshold segmentation of the filtered output. The segmented output is then used in a logical test (or mask) to segment the original image. Segmentation (Fig. 14.13) may simply exclude background voxels or the included voxels can then be grouped into polygon objects. Voxel objects can use a material look-up table (M-LUT) for each property. It is possible to specify all material properties for any object types as indices in one or more material look-up tables (M-LUTs) and to use these to separate different materials within the volume. The Analyze package has a good example of a dedicated segmentation menu that brings together image edit, morphology operations, and object/surface extraction functions. The Surpass optional module in Imaris groups together the object segmentation capabilities of that package.
Segmenting the Data Objects
Scan Conversion
Segmentation is a process by which objects are extracted from the rastered voxel image data. Voxels are selected according to their brightness and/or by some property linking them to other voxels. A lower and/or upper threshold may be applied producing a segmented band containing all the voxels of a particular object. Most systems allow this intensity segmentation to remove background or to isolate a homogeneously stained structure. Threshold values are readily chosen by masking a histogram and observing this simultaneously with an interactive, color-banded screen display of the segmented voxels. Histogram peaks indicate subpopulations of background voxels, possible structural features, etc. This edited histogram produces a photometric look-up table (P-LUT) through
After geometric rotation, etc., the data objects are drawn into the final view. For polygons, surface nets, etc., this requires a scan conversion whereby the vectors are turned into rastered lines of pixels (Watt, 1989). A pixel view of each polygon patch is then composited into the final view as for rastered voxels. A simple approach is to draw a closed outline and to use a fast-fill algorithm, modified by any shading required. It is often more efficient to generate just the ends of each scan line through the polygon, filling the gaps by simple line drawing. For two vertices S(x" y,) and E(xe , Ye) defining the start and end of a polygon edge, there are (ye - Ys) intermediate scan lines. The view coordinates for the start of the jth scan line are
302
Chapter 14 • N.S. White
(a) Orig ina l
(e)
(b) High pass threshold
(d ) Mid threshold slIIface extraCt
(g) Gradient magnitude
Xv
(e) High threshold surface
Yv
id ·i ntensilY band
(I) Image edit or cut·away
(i) Product of gradient magnitude 'lnd original
(h) High pa s threshold and sced fi ll
= (int)(xs + jOx),
= Ys + j
(17)
where Ox = (xe - xJ/(ye - Ys). The ends of each scan line are obtained by a similar calculation. Scan conversion or rasterization is a well-defined component of the OpenGL graphics pipeline and this affords a convenient point in the reconstruction process to combine or embed graphical objects into a voxel or previously generated pixel view. This is a powerful technique used to good effect by the major rendering packages to show high contrast segmented objects within the context of a general volume view (Amira, for example, allows for an arbitrary number of datasets, image modalities and/or visualization modes within the same display view).
Projection Rules As each geometric object or image plane is compo sited into the view, incoming data either replaces or is blended with the accumulating result. The algorithm used is usually a function of the z-coordinate, the intensity value, and the associated material properties (Table 14.8). This P function has two components: an arithmetic or compositing (C) function (which may be just a simple assignment) and an optional logical test (T). For the nth (of N) frames or objects compo sited into the view V, a general form is if {T [I(x" y" zv)no V(xv, Yv, Zv)n _ d} then V(x" Yv)n = C [V(x" y" zv)no I(x" Yv, Zv)nl
(IS)
Simple compositing functions (Fig. 14.14) include:
= I(xv,
Maximum intensity projection, or MIP (e.g., Sakas et aI., 1995), is now the most widely used quick visualization function (e.g., see Analyze, Cedara, Imaris, FIRender, Voxblast etc) and can be efficiently implemented as a stand-alone mode or in combination with other algorithms.
• Average Intensity (no test) V(x" Yv)n = V(x v, Yv, Zv)n + I(xv, y" Zv)iN (20)
Less common than MIP, mean intensity is found in the LSM programs and Analyze. It is useful when the data volume is very dense, which would tend to give a very saturated view with MIP. Because it effectively averages along the projected ray, this function reduces noise but produces lower contrast. This can be partly overcome with an appropriate display LUT.
• First or Front Intensity 2': t if {[I(xv, Yv, Zv)n > tl & [(Zv.n-I) < (Zv.n)]} then V(x" Yv)n = I(x v, Yv, Zv)n (21)
This is a quick way of exploring a surface or boundary in a voxel representation, particularly as an aid to determining parameters for more complex modes (e.g., see Lasersharp projection modes).
• Alpha Blend (no test) V(xv, Yv)n = (1 - a) V(xy, Yy, Zv)n-I + (a) I(x" Yv, Zv)n
• Maximum Intensity if [l(x" y" Zv)n 2': V(x" y" zv)n-d then V(x" Yv)n
FIGURE 14.13. Object segmentation. Voxels may be segmented from the image volume only to remove background and unwanted features, or they may additionally be grouped into larger geometric objects. (A-I) show 20 slices from a small 3D volume before and after segmenting with various algorithms. High-pass (B) and mid-intensity band (C) are the simplest operations and are often used on their own, or in conjunction with other operators. Surface extractions (D,E) (e.g., Marching cubes, c.f. VoxBlast or Delaughney Triangulation, c.f. Visilog) produce bounded structures obscuring internal details. Manual or semiautomatic image editing or reconstruction cut-aways (F) reveal internal details without corrupting the data values. Gradient magnitude filters (G) highlight voxel boundaries and edges. Thresholded seed fill (H) (c.f. Visilog, ImageSpace) gives a solid object for simple volume estimates, etc. Complex modes such as the gradient magnitude blended with the image (I) (see text) can give reasonable enhancement of edges without artifactual "halos" (c.f. Prism II1DeltaVision).
Yv, zv)n (19)
(22)
This is the standard mode used for object visualization and works particularly well for voxel objects and is implemented in virtually
FIGURE 14.14. Projection combination or compositing rules. 3D visualizations of a stack of 130 confocal optical sections through a portion of Arabidopsis plant root using Lasersharp (A-C) and lmaris (J-L). Autofluorescence (and a little non-specific fluorescent thiol-staining) shows vascular tissue, cell walls and nuclei and lateral root buds. A small attached fungus is seen attacking the plant along the lower edge towards the center of this portion of root. (Maximum intensity projections as rotations at -30deg (A), Odeg (B) and +30deg (C). Equivalent average projections (D-F) show the characteristic "fuzzy X-ray" character of images made with this algorithm. The first voxel (nearest the viewer) along each cast ray is recorded in (G-l) showing a more solid appearance while retaining original intensity values but at the expense of losing many details. a-rendering at the same angles (J-L) clearly represents the density of the structure as well as fine details but introduces some spurious high intensities where the structure is thin (e.g., at the end of the lateral root tips).
304
Chapter 14 • N.S. White
all of the packages in Table 14.1. Although the added complexity of having to set the a parameter sometimes makes it harder to use than MIP for LSM fluorescence data, this parameter is the basis for all opacity/transparency effects (see below). The tests in Eqs. 19 and 21 may be reversed to obtain minimum and below-threshold versions. The alpha term is a general function that can partly define the physical properties of the rendered object. This blending factor can be dynamically modified as a function of Zv or n during the projection. A Kalman version of Eq. 20 may be implemented in this way; the number of composited frames N need not be known in advance, and the process can be stopped when the required result is reached. • Kalman Average V(xv , Yv)n = (l - lin) V(xv , Yv, Zv)n-I + (l/n)I(xvo Yv, Zv)n
(23)
This is a dynamically modified a = lin, and the projection may proceed in positive or negative Zv order. These define the front-toback and back-to-front view space rendering geometries. It should be readily obvious that these two processes afford a simple means of modeling the lighting of voxels from front-to-back or the emission of light from voxels running from back-to-front with respect to the viewer. The latter rendering order also shows the view building up towards the viewer that can reveal internal details of an object during the visualization process, provided the screen is updated during the computation. A display equipped with a-plane hardware (with associated firmware or software) automatically blends the incoming data using factors stored and updated in an alpha memory plane. a-blending may also be implemented entirely in software. This is a standard OpenGL feature. The a value is an eighth parameter for vectors (after the coordinates and RGB intensity) or a fifth value for voxels (with implied coordinates) specifying blending properties for the intensities of that object. Visibility of an incoming voxel or data object can thus be made dependent on object opacity or transparency and therefore depth (zv) in the rotated view (see below). Earlier we discussed ways in which the geometric transformation could be modified to encode additional z-information into the display geometry. In addition to z-related blending operations, the compositing algorithm can retain z-information within the displayed intensities using alternative algorithms.
How Can Intensities Be Used to Retain z-I nformation
if {[I(xv, Yv, Zv)n-I > t] & [(Zv,n-I) < (Zv,n)]} then V(x v, Yv)n = etc. (25)
These z-coordinate modes are used by simple 3D topology programs (e.g., Automontage) as well as some renderers including Analyze, Volume], Lasersharp. etc. • Iso-Intensity Surface (etc.)
(26)
Although related to the previous projections, the iso-surface routine is usually implemented by a recursive 3D algorithm such as the marching cubes (Lorensen and Cline, 1987) and is the basis of many surface object segmentation algorithms (e.g., Voxblast and Analyze). Through-focus images from widefield microscopy have been processed using a maximum test applied to a prefiltered version of each frame. For example, the maximum local variance can produce an auto-focus intensity or coordinate view (c.f., Automontage). The depth discrimination of the confocal microscope allows a simpler auto-focus routine using just maximum brightness (for a single-surface object). This requires that the z-response for the particular specimen (point, plane, etc.) has no significant side lobes (the confocal PSF is investigated by Shaw and Rawlins, 1991). Instead of replacing the brightness with zv, intensities can be modified (weighted) by their z-coordinate giving a so-called depthweighted projection (Fig. 14.16). The simplest form is a linear weighting from 1 (nearest the observer) to 0 (furthest away). • Linear Depth-Weighted Projection (27)
where Z = (Zback - Zv,n)/(Zback - Zfront), that is, a normalized coordinate. A more sophisticated form is
z-
• Exponential Depth Weighting (28)
where f is a constant ~1. This algorithm (which can be implemented in any z-order) is often described as an attempt to model the absorption of light from a source directed along the z-axis of the data from behind (transmitted) or the attenuation of emitted fluorescence. Not surprisingly it turns out that this result is identical to that achieved by an ordered recursive algorithm traversing the data from front to back using a constant a-blending factor (~1): • Recursive Exponential Weighting
z-Coordinate or Height Views The height, range, topological, relief or (the author's preference) z-coordinate view (e.g., Boyde, 1987; Freire and Boyde, 1990; Forsgren et al., 1990; Odgaard et al., 1990) technique has been used for many years to directly record the Zv-depth in the intensity or color space (or both) of the display. The rule includes a test that selects a particular voxel along the observer's line of sight. The Zvcoordinate is then assigned [after an offset (zo) and scaling (Zf)] to a value in the view. A range of z-coordinate tests, similar to the intensity tests (Eqs. 19,20) above, are found (Fig. 14.15):
The commonest form of Eqs. 27 to 29 is a linear average or summation, but any compo siting function can, in principle, have a depth-weighted component. By making the factor equal to f= ZsJ(Nn),
(30)
where N is the number of serial planes in the projection, a nonlinear projection of strength equal to Zstr is obtained. If Zstr = N, the result is identical to a Kalman average. For Zstr = 1, the front or maximum height voxels completely dominate the view. Other values give intermediate results [Fig. 14. 16(A-C)].
• Coordinate of Maximum Intensity if [ I(xv, Yv, Zv)n ;::: V(xv, Yv, zv)n-d then
V(xv, Yv)n = Zo + ZfZv,n (24)
• Coordinate of First Intensity;::: (t) (i.e., Maximum Height or Nearest)
Hidden-Object Removal Z-Buffering z-Ordered compositing simplifies the implementation of zalgorithms. Front-to-back projections using the first-object test
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
305
FIGURE 14.15. Z-coordinate views. (A) Maximum intensity projection of serial confocal optical sections through pollen grains in a partially dissected plant anther. Some sections are missing from the data stack (or perhaps the specimen is incomplete) but it cannot be elucidated whether the loss is from the top or bottom of the set from this projection, which carries no z-information. (8) shows a height or distance projection where the brightness is coded by axial distance from the viewer (dark is furthest away) recording the first voxel above a background threshold. It can now be seen that the lost sections are at the top of the stack which was projected top-to-bottom. (e) shows the height of each voxel chosen by maximum projection in (A), confirming the finding from (8). (D) shows the same algorithm as in (8) projecting bottom-to-top through the stack.
merely check if a non-zero intensity has been encountered. Hidden voxels are never processed. Conversely, if a back-to-front pass is used, the last voxel will be the one displayed and no test is required. The entire volume is now traversed and all voxels are processed. This can be very informative if the rendering process is visible on an interactive display. For non-z-ordered objects, a more intensive logical comparison of intensities and/or z-coordinates is needed, such as a front-object z-test; this is known as z-buffering and is a fundamentally important OpenGL-controlled process in modern graphics systems. Implementation can be at various stages of the rendering process. Voxel data is z-buffered efficiently as a whole section operation after warping. Polygon-ordered rendering uses standard pixel z-buffering but does not use information about adjacent vertices efficiently. Scan line z-buffering is more economical. It is intimately associated with the scan conversion of polygon edges to pixels line-by-line without needing a full 2D z-buffer. Spanning scan line z-buffering is highly optimized to extract visible object( s) from all structures that intersect the screen line that is currently being drawn. Modern graphics hardware can encompass intensity, z-buffer, and a-planes to program all of the computations described above into the display logic, releasing the processor for other computations.
Local Projections The compositing functions described above can be used in combination for more control over the rendered objects. A conflict exists between hidden-object removal and significantly modifying the image intensities (by excessive use of a factors and lighting terms). A novel solution is implemented in the Lasersharp visualization program: the trick is to use coordinate or z-buffer algorithms to derive a segmentation reference (R z) for projected pixels in the final view. R z defines the z-coordinate of a surface [e.g., by maximum intensity (Eq. 24)] or boundary [e.g., by maximum height above threshold (Eq. 25)], etc. R z (for each view pixel) can then become the center for a z-banded or local projection. This technique is another example of an intelligent z-buffer. The range of the local projection is defined by Zfront and Zback given as z-offsets from the reference. Useful local projections include: (I) Reference Height Above Threshold + Local Maximum Intensity, e.g., Maximum Height;::: t (Reference). if {[I(x" y" zv),,-I > t] & [(Z,,-I) < (z,,)]} then RzCx" Yv) = followed by Local maximum intensity
Zn
306
Chapter 14 • N.S. White
FIGURE 14.16. Depth weighting. (A-C) MPL "non·linear average" depth-weighted projections (see text) with "strengths" of I (A), 3 (8) and 9 (C) (number of sections = 50). MicroVoxel "depth-weighted" views (D) summation (average). (E,F) "first mode" renderings of nerve cells.
Adding Realism to the View where [R,(x" yJ + Zfmn'] range).
~ Zr ~
[R,(x" Yv) + Zback], (i.e., the local
Similarly one might use (2)Height at Maximum Intensity + Local Kalman Average. (3)Height at First Intensity ~ t1 + Offset Local Height at Intensity ~ t2
(4)Height at Maximum Intensity + Offset Local Maximum Intensity e.g., with
Zback
>
ZffOn'
> R,(x"
yV>.
Local projections 3 and 4 use a range that is offset from, that is, does not span, R z• This is an objective way to segment a second object or surface within a given range of a more dominant primary feature (which is used for the reference). Thus, a plant cell wall or animal cell membrane may be found by a reference segmentation, and then structures within ZffOn' voxels outside or Zback voxels inside the cell can then be projected in isolation. Comparative results of some local projections from Lasersharp are shown in Figure 14.17.
The algorithms discussed so far use test and compositing rules to project multi-dimensional images into the view space. Views of macroscopic objects contain depth cues (similar to those described above) along with textural cues arising from the position and properties of light sources. These can be used to add realism to reconstructed views in microscopy by (l) mimicking artificial macroscopic lighting effects or (2) developing a more objective visualization model incorporating a priori knowledge concerning the optical properties of the sample. Advanced algorithm options are listed in Table 14.9 and described in Figures 14.18, 14.19, 14.21, and 14.22 with example results in Figures 14.17 and 14.20.
Artificial Lighting: Reflection Models The ambient lighting of the model assumed above has no directional components. Artificial lights have intensity (photometric), directionality (geometric), and color (chromatic) properties. These characteristics interact with the material properties of data objects to modulate the rendering process. Local lights are near or within the data volume and infinity sources are parallel rays coming from infinity. Ambient lighting is a general level diffused by multiple reflections off all objects within the volume, as distinct from light coming direct from the source to a given object. Reflections from
Visualization Systems for Multi-Dimensional Microscopy Images • Chapter 14
307
FIGURE 14.17. Local projections. These 3D views are made from the same data set shown in Figure 14.14. (A,B) show local projections where the maximum intensity is found through about one third of the depth of the sample above and below each voxel found by a previous application of the "above threshold" rule. (A) is from above and (B) below. (C,F) show the same operations applied through a depth of about one tenth of the sample thickness from the reference voxel. More structures towards the outer surface of the root are apparent compared to regular maximum projections, masking underlying features. (D,E) are local average projections corresponding to (A,B). Normalizing these intensities to fit the dynamic range of the display leads to lower contrast views than in (A,B) but shows some weak thiol-staining more clearly against the vascular autofluorescence. These views are more amenable to direct quantification than z or b.
an object (Fig. 14.18) can be approximated by a function (L) composed of (I) ambient reflections (La), (2) diffuse reflections (Ld), and (3) specular highlights (Ls) (strongest in the source direction). L where
zobi
= La + (Ld + Ls)/(Zobj + constant)
is the distance (in
Zv
(31)
of the object from the observer).
The denominator in Eg. 19 is an approximation to the z~ denominator from the diffuse reflection law of Lambert (Born and Wolf, 1991, p. 183) for cases where Zobs is large compared to the object size. Each component Lm is the product of the light source intensity Sm, reflectivity Rm (a material property), and a direction term Dm (m = a, d, or s). The color of a voxel is determined by the
TABLE 14.9. Overview of "Realistic" Visualization Techniques for Multi-Dimensional Biological Microscopy Data Feature "Visualization models Material properties
Surface shading models Artificial lighting
Parameter color opacity reflection
Minimum Required Voxel render RGB channels
emission Hidden-objects
Simulated fluor. Software z-buffer
Color Lightable objects Lighting models
RGB Voxels Ambient/diffuse (brightness)
Desirable additional enhancements Shaded surface. Voxel gradient. Lighting models. SFP. Embedded objects Arbitrary colors/channels a-channel, channel dependent Diffuse, Specular ("shiny"), Interactive control Hardware texture mapping Hardware z-buffer Flat, Incremental, Gourard, Phong model Arbitrary colors & sources Voxels, Surface normal Voxel gradient, Smoothed voxel "surface," Phong surface normal, Fast mode, Precision mode
"Realistic visualization modes are often used to promote particular packages. and often with carefully chosen data! Control of the object material properties and a clear understanding of each parameter are essential. The final reconstructed view should always be studied along with a record of all the steps and variables and preferably alongside the original image sections. h Material properties and artificial lighting should be standardized for each view if intensity information is to be reliably compared between results (particularly for SFP and gradient-lit voxel modes). Hardware acceleration has made possible more interactive 3D views and "real-time" processing of modest data sets on desktop PCs.
308
Chapter 14 • N.S. White
h
n
FIGURE 14.18. The artificial lighting model. Ambient lighting contributes only to the overall image brightness and has no directional components. An artificial light source is directed along I with parallel rays. Diffuse reflections from the object surface (solid panel) emanate in all directions (Ld ). Some of Ld may be seen by the viewer along v. Specular reflections (L,) occur in a narrow cone around the "reflection" direction r (9 nl away from the surface normal n). These are observed along v if 7. To convert this to the optical cross-section per molecule, one must multiply £ by (l000cm31L) . (In 10)/(6.023 x 10 23 molecules/mole) = 3.82 x 10-21 cm3 • mole· L -I • molecule-I, giving a cross-section s of 3.06 x 10- 16 cm2/molecule. In a beam of 1.25 x 1024 photons· cm- 2 s- l , each ground state molecule will be excited with a rate constant ka = aI, or 3.8 x 10 8 S-I in this example. The excited state lifetime 'tr of fluorescein in free aqueous solution is known to be about 4.5 ns (Bailey and Rollefson, 1953), which means that molecules in the cxcited state return to the ground state with a rate constant kr of 2.2 x 108 S-I. Note that kr is defined here as the composite rate constant for all means of depopulating the singlet excited state, the sum of the rate constants for fluorescence emission, radiation less internal conversion, intersystem crossing to the triplet, etc. Because of the Stokes shift between excitation and emission AS, kr is not significantly enhanced by stimulated emission effects. If x is the fraction of molecules in the excited state and (l - x) is the fraction in the ground state, at steady-state ktx = ka(l - x). Solving for x yields the equation x = k.l(ka + kr), which shows that in this example 63% of the fluorescein molecules would be in the excited state and only 37% in the ground state. Obviously the emission is nearly saturated, and further increase in excitation intensity could hardly increase the output. The actual rate of photon output per molecule is Qekrx = Qekrk)(ka + kr), where Qe is the emission quantum efficiency, about 0.9 for free fluorescein dianion (but usually less for fluorescein bound to proteins, see below).
extremely intense bursts of excitation as the focused laser beam
In this example each molecule would be emitting at an average
sweeps past. If the laser-scanned image consists of n pixels (typically n > 105 ), anyone pixel is illuminated for lin of the total time,
rate of l.3 x 108 photons/s, close to the absolute maximum of Qekr of about 2 x 108 photons/so In current typical scanning confocal
Roger Y. Tsien • School of Medicine, University of California, San Diego, California 92093 Lauren Ernst and Alan Waggoner· Carnegie Mellon University, Pittsburgh, Pennsylvania 15213
338
Handbook of Biological Confocal Microscopy. Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
Fluorophores for Confocal Microscopy: Photo physics and Photochemistry • Chapter 16
"0
0
Q)
Q)~
~
I I!)
C\I
c,
"0
0
~
I« v
C\l1'vI!) vv
C\I I!)- (0
(')(')
(')
~ co co v
~>'=
Triplet State Saturation
Q)
z
~(')
~ co (0
I!) I!)
I!)
vC\l
Q)~
I~
(')1'(')v (0(0
250000
200000 C Q)
·u
:E Q)
150000
0
u
c 0
U 100000 .s
x
L.U
50000
400
500
600
339
700
Wavelength (nm)
FIGURE 16.1. Absorption spectra from left to right of Hoechst 33342+DNA, and Fluorescein, CY3, TRITC, Lissamine Rhodamine B, Texas Red, and CY5 conjugated to antibodies. Extinction coefficients are given on a per-dye basis. Common laser emission AS are presented at the top of the figure. AS in bold are for the lower power, less expensive lasers, which provide sufficient excitation intensity for most fluorochromes imaged with laser-scanning microscopes with higher power objectives.
microscopes, the beam dwells on each pixel for 1 to lOlls, so one would expect each molecule to produce several hundred photons. However, only a minority enter the microscope objective, only a fraction of these manage to pass all the way through the scanner optics, and only 10% to 20% of these create photoelectrons in the photomultiplier cathode, so that each molecule probably contributes on the order of only one photoelectronlpixellsweep. Because of fluorescence saturation, increasing the laser power will not significantly increase the signal amplitude. In reality, it is difficult to accurately predict the laser power at which a f1uorophore will saturate. Accurate knowledge of the extinction coefficient at the A of excitation and the excited state lifetime of the f1uorophore is essential but not always easy to obtain. Coumarins, for example (Table 16.1, see p. 344), have extinction coefficients approximately 2 to 5 times smaller than the fluoresceins, rhodamines, and cyanines, and would be expected on this basis alone to require a 2- to 5-fold increase in laser power before saturation. However, the emission lifetimes of coumarins are intrinsically longer than those of the fluoresceins, rhodamines, and cyanines, so that the factor of 2 to 5 is not realized in practice. To make calculations more difficult, the extinction coefficients and the excited state lifetimes of most probes depend on the environment of the f1uorophore. For example, increasing the f1uorochrome-to-protein ratio of a labeled antibody from 2 to 5 can reduce the average excited state lifetime of the bound f1uorochromes several-fold. For these reasons, the power saturation values for the probes listed in Table 16.1 are not given. However, as a general rule, f1uorochromes with extinction coefficients and quantum efficiencies similar to fluorescein will also saturate under similar conditions.
The above calculation considers only the ground state and lowest excited singlet state. Saturation of emission could occur at even lower excitation intensities if a significant population of f1uorophores becomes trapped in a relatively long-lived triplet state. This would take place if a significant quantum yield exists for singlet-to-triplet conversion, or intersystem crossing. For example, if ground-state fluorescein molecules each absorb 3.8 x 108 photons· S-I, have an excited singlet lifetime of 4.5 ns, cross to the triplet state with a quantum efficiency QISC of 0.03 (Gandin et aI., 1983) and reside in the triplet state for a mean time 'tT of lO-6 s, the triplet state would contain 81 % of the f1uorophore population at steady state, which would be attained with a time constant of about ['tT- 1 + QIscktkal(ka + kf)r l or about 190ns in this case. Only 12% and 7% would be left in the first excited singlet and ground states, respectively, at steady state, so that the fluorescence emission would be weakened about 5-fold compared to its initial value just after the illumination began but before significant triplet occupancy had built up. Therefore, if the dwell time/pixel is comparable to or greater than the triplet lifetime, then a severe reduction in output intensity may be expected beyond that due simply to the finite rate of emission from the singlet state. In the above calculation, the most uncertain figure is that for the triplet lifetime 'tT; in very thoroughly deoxygenated solution, 'tT for the fluorescein dianion is 20 ms (Lindqvist, 1960), but oxygen is expected to shorten 'tT down to the 0.1 to Ills range. The rate at which different f1uorophore environments in a sample quench triplet states and reduce 'tT will of course affect the extent of triplet-state saturation and the apparent brightness of each pixel at these high illumination levels. When 'tT is long, triplet-state saturation is easily attained even without laser illumination (Lewis et at., 1941).
Contaminating Background Signals Rayleigh and Raman Scattering Meanwhile, there may be unwanted signals, such as Rayleigh scattering, due either to excitation AS leaking through the dichroic mirror and barrier filter or to imperfect monochromaticity of the excitation source, for example, if the laser is being run in multi-line mode with only an interference filter to select one line. Even if all the A filtering is perfect, Raman scattering will contribute a fluorescence-like signal, for example, at a A of P"exc- I - 3380cm- 1r l due to the characteristic 3380cm- 1 Raman band of water. For an exciting Acxc of 488nm, the Raman peak would appear at 584nm. At high concentrations of protein or embedding media, additional Raman bands closer to the excitation wavelength might appear. Both Rayleigh and Raman scattering are directly proportional to the laser power and will not saturate as the desired fluorescence does, so that excessive laser power diminishes the contrast between fluorescence and such scattering signals.
Autofluorescence from Endogenous Fluorophores Another major source of unwanted background is autofluorescence from endogenous f1uorophores. Flavins and f1avoproteins absorb strongly at 488 nm and emit in the same spectral region as fluorescein. Reduced pyridine nucleotides (NADH, NADPH) and lipofuscin pigments absorb light from ultraviolet (UV) laser lines. These f1uorophores usually have lower extinction coefficients or shorter fluorescence lifetimes than most exogenous f1uorophores. For example, FMN (flavin mononucleotide) and FAD (flavin
340
Chapter 1 & • R.Y. Tsien et al.
adenine dinucleotide) have the extinction coefficients of 1.1 to 1.2 x lO4 M- 1 em- I at 445 to 450 nm (Koziol, 1971) and fluorescence lifetimes of about 4.9 and 3.4/0.12ns, respectively (Lakowicz, 1989), whereas NADH has an extinction coefficient of 6.2 x 103 M- 1 em- I at its 340nm peak (Kaplan, 1960) and a lifetime of about O.4ns (Lakowicz, 1983). Therefore autofluorescence from these molecules will be more difficult to saturate than the fluorescence from most of the common probes. See Chapter 27, this volume, for more on confocal lifetime.
What Is the Optimal Intensity? The above discussion shows that if laser power is increased to nearly saturate the desired fluorophores, autofluorescence as well as Rayleigh and Raman scattering will increase background levels and decrease the overall signal-to-noise ratio (SIN). What is the optimal intensity? Assuming the system is limited by photoncounting statistics, the irreducible noise level N is proportional to the square root of the background signal level B. Both B and the absorption rate constant ka of the desired probe are directly proportional to the illumination intensity I and to each other. Therefore N is directly proportional to k!l2. If triplet state saturation can be ignored, for example, if the dwell time/pixel is short compared to the time for the triplet state to build up, then the desired signal is Qek fk.l(k. + kr) as derived above, so that SIN is proportional to k~I2/(k. + kr). Regardless of the proportionality constant, this expression is maximal when k. = k f • This is a remarkably simple but important result, for which we are grateful to Prof. R. Mathies (University of California-Berkeley). At the other extreme, once the ground state and excited singlet and triplet have all come to an equilibrium steady state, the desired signal is readily calculated to be Qek.l[1 + k. (k f- I + QISC"tT)]
so that the SIN is proportional to Qek!I2/[1 + k. (kl + QISC"tT)]'
This reaches its maximum when ka('tr + QISC'tT) = 1, where
Lf
= kr- I is the lifetime of the excited singlet. In this approximation,
appropriate for slow scans in which the dwell time/pixel is long compared to the time for triplet state equilibration, the laser power P in photons/s should be optimal at about Itw 2 /[(3.82 x lO-21 cm 3 . M) £ ('tr + QISC'tT)]' For the present values of w = 2.5 X lO-scm , £ = 8 X 104M- 1 cm- 1, 'tr = 4.5 X 10-9 sec, Q1 SC = 0.03, and 'tT = 10-6 s, this expression gives an optimal P of 1.86 x 10 14 photons/s, or about 76 J.l.W at 488 nm. If triplet formation could be neglected, the optimal P would be 590J.l.W, slightly less than the 1 mW initially postulated to be the input.
PHOTODESTRUCTION OF FlUOROPHORES AND BIOLOGICAL SPECIMENS One obvious way to increase the total signal is to integrate for a longer time, either by slowing the scan, or by averaging for many scans. Given that image processors are now relatively cheap, the latter alternative is likely to be the easier to implement, and it has at least two major advantages: repetitive scans give time for triplet states to decay between each scan, and the user can watch the SIN gradually improve and choose when to stop accumulating. However, irreversible photochemical side effects such as bleaching of the fluorophore or damage to the specimen set limits on the useful duration of observation or total photon dose allowable.
Photochemical damage is one of the most important yet least understood aspects of the use of fluorescence in biology; in this discussion we can do little more than define our ignorance. At intensities of up to 3 X 1022 (Hirschfeld, 1976) or 4.4 x 23 10 photons· cm-2. S- 1 (Mathies and Stryer, 1986), fluorescein is known to bleach with a quantum efficiency Qb of about 3 x 10-5 . If this value continues to hold at the somewhat higher intensity of the above example, the molecules would be bleaching with a rate constant of Q~fk.l(ka + kr), or about 4.2 x lO3 S-I. This would mean that lie or 37% of the molecules would be left after 240 J.l.s of actual irradiation. The corresponding number of scans would be 240 J.l.s divided by the dwell time that the beam actually spends on each pixel. The average number of photons emitted by a fluorophore before it bleaches is the ratio of emission quantum efficiency to bleaching quantum efficiency, or QelQb; this is generally true regardless of whether the illumination is steady or pulsed. For fluorescein under ideal conditions, the above Qe/Qb works out to 30,000 to 40,000 photons ultimately emitted/dye molecule (Hirschfeld, 1976; Mathies and Stryer, 1986). The number of photons detected from each molecule will, of course, be considerably less. Thus, in obtaining an image at a single plane by averaging say l6 scans, one would expect from 6% to 50% bleaching of a fluorescein signal (in the absence of anti fade reagents), resulting from exp[-(l6 sweeps) x (1-IOJ.l.s dwell time/sweep)/(240J.l.s lifetime)]. In an optical sectioning experiment, the cone of light illuminating the sample above and below each plane of data being acquired is causing photobleaching even though under confocal conditions no signal is being recorded from these regions. This means that if 16 optical sections are obtained with only one sweep each, the last image will have been bleached 10% to 50% by the preceding sweeps. If the signals are large, it may be possible for software to adjust the intensities of sequential images to compensate for bleaching. Of course, for quantitative fluorescence measurements it would be most desirable to use the most bleach-resistant fluorophores available.
Dependency on Intensity or Its Time Integral? Theory One major uncertainty is whether the bleaching quantum efficiency Qb really does remain constant even at such high instantaneous excitation intensities. Theoretically, Qb could rise at high intensities due to multi-photon absorptions. For example, the normal excited singlet or triplet state could itself absorb one or more additional photons to extra-high energy states, which will probably have picosecond lifetimes. If their main reaction pathway were back to the lowest excited state, then such higher-order states would be innocuous, but if bond dissociation competes with decay to the lowest excited state, then photodestruction will rise steeply as intensities reach levels that significantly deplete the ground state (as in the previous example). One can also imagine the opposite dependency: bleaching mechanisms whose quantum efficiency might decrease when the excitation was bunched into brief intense pulses. For example, suppose the dye is bleached by a 1: I reaction of its excited state with a molecule of oxygen, and that the overall dye concentration exceeds the oxygen concentration. Then within the zone of intense illumination, the first few excited dye molecules might react with all the locally available O 2 , and the resulting anoxic environment would protect the rest of the excited molecules. Oxygen would be diffusing in from the surrounding non-illuminated environment, but the time required would be on the order of the spot radius squared
Fluorophores for Confocal Microscopy: Photophysics and Photochemistry • Chapter 16
341
divided by the diffusion constant, that is, (0.25 x 10-4 cm)2/(3 x 10- 5 cm 2 . S-I) or 20lls, that is, considerably longer than the time for the beam to move to the next pixel. By contrast, with low-intensity illumination, no local anoxia would develop, and each fluorophore would tak~ its chances with the full ambient oxygen concentration. Such a mechanism would be the photochemical equivalent of predator-prey interactions in ecology, where it is often advantageous for the prey to school together or breed in synchrony in order to minimize losses to predation (Wilson, 1975).
was better because it allowed repair mechanisms to operate during the dark intervals. It should be noted that in all the above biological examples, the illumination intensity was well below that expected to be necessary to reach saturation of excited state dye populations (see also Chapter 38, this volume).
Experiment
light Collection Efficiency
Theory is all well and good, but empirically how does the photodestruction quantum yield depend on intensity? White and Stryer (1987) tested R-phycoerythrin in a flow system and found that the rate of photodestruction was indeed directly proportional to laser power so that the quantum yield was constant. Unfortunately, the range of intensities tested only went up to about 1020 photons . cm- 2 's- l , so that they were still 3 orders of magnitude below saturation of the phycoerythrin. Peck and colleagues (1989) examined B-phycoerythrin and found that photodestruction saturated with increasing input intensity in just the same way as fluorescence emission, so that the photodestruction quantum yield was roughly constant even when the phycobiliproteins were heavily driven into saturation. However, there was some indirect evidence that simpler fluorophores may undergo a nonlinear acceleration of bleaching under such conditions. White et al. (1987) reported that bleaching by their scanning confocal microscope seemed to be greatest at the plane of focus. Because the plane of focus receives about the same time-averaged photon flux but much higher peak intensities than out-of-focus planes do, preferential bleaching at the plane of focus would imply that a given number of photons are more injurious when they are bunched, that is, that the photodestruction quantum yield increases at high intensities. Obviously, further testing of this possibility and improvement in photon collection efficiency will be of great importance in confocal microscopy. A number of workers have reported that damage to biological structures (as distinct from bleaching of the fluorophore) can sometimes be reduced if the given total number of photons is delivered with high intensities for short times rather than low intensities for long times. An early report of the advantage of pulsed illumination was by Sheetz and Koppel (1979) studying the crosslinkage of spectrin by illumination of fluoresceinated concanavalin A on erythrocyte membranes. Bloom and Webb (1984) found similar results for lysis of XRITC-labeled air-equilibrated erythrocytes, whereas well-deoxygenated cells were much more resistant but lysed after a constant total photon dose regardless of whether delivered quickly or slowly. Recently Vigers and colleagues (1988) showed that increasing the illumination intensity up to about I OJ W/cm 2 decreased the time required for fluorescein-labeled microtubules to dissolve, as one might expect. Surprisingly, intensities above this threshold actually stabilized the microtubules again st dissolution, so that the dissolution time became a linearly increasing function of intensity. This paradoxical stabilization at high intensities was attributed to local heating based on the assumption that diffusion of heat was negligible. Because this assumption needs to be checked (see Axelrod, 1977; Bloom and Webb, 1984) and because microtubu Ie stabilization and dissolution are not well-defined molecular events, the local heating hypothesis should be viewed with caution. Bonhoeffer and Staiger (1988) have reported that photodynamic damage to rat hippocampal cells was reduced if the light is delivered at 100 exposures each 200 ms long separated by 30 s dark periods rather than continuously for 20 s. They speculated that intermittent illumination
STRATEGIES FOR SIGNAL OPTIMIZATION IN THE FACE OF PHOTOBlEACHING
The above discussion has shown that to increase signal amplitude and SIN in laser-scanning confocal microscopy, increasing laser power helps only until the onset of saturation, and increasing observation time is limited by photodestruction. What other measures can be tried? Obviously any increase in light-collection efficiency (i.e., higher numerical aperture of the objective), transmission efficiency through the scanner and A filters, and quantum efficiency of photodetection is extremely valuable. Despite the importance of these factors , newcomers to low-lightlevel microscopy often use low numerical aperture (NA) objectives, excessively narrow emission bandpass filters , inefficient optical couplings, and photomultipliers of less-than-optimal quantum efficiencies. Nearly all the fluorophores that fluoresce strongly in aqueous solution with visible As of excitation are characterized by small Stokes shifts, or difference between absorption and emission peak AS. It may then be difficult to find or fabricate filters and dichroic mirrors that efficiently separate the two A bands. In that case, it would usually be preferable to displace the excitation A to shorter AS away from the peak of the excitation spectrum, so that the emission filters can accept as much of the entire output as possible. Although the excitation is less efficient, thi s can be made up by increased laser power as long as the photobleaching is reduced by the same factor. By contrast, if the excitation is at the peak A and the emission acceptance band is pushed to longer As that exclude much of the emission spectrum, then emitted photons are wasted, while excess scattered photons are collected. The SIN ratio is thus lowered. This is often a severe problem with rhodamine excited using the 514nm line of the argon-ion laser.
Spatial Resolution Another tactic to increase signal is to increase the effective size of the confocal apertures, that is, decrease the spatial resolution. If the illuminating and detecting aperture diameters are doubled, the pixel area quadruples and the volume sampled will increase 8-fold. Assuming the total laser power is increased to maintain the same intensity in photons· cm-I's- I and that the ftuorophore concentration is uniform in the increased volume (as might be true for an ion indicator distributed in the cytosol), the signal should increase 8-fold, though at the price of degraded spatial resolution.
Protective Agents As mentioned above, light-induced damage to both the fluorophore and to the biological specimen is often dependent on the presence of molecular oxygen, which reacts with the triplet excited states of many dyes to produce highly reactive singlet oxygen. Reduction of the partial pressure or concentration of oxygen often greatly increases the longevity of both the fluorophore and the specimen. In dead, fixed samples, it has become common to add antioxidants
342
Chapter 16 • R.Y. Tsien et al.
such as propyl gallate (Giloh and Sedat, 1982), hydroquinone, pphenylenediamine, etc., to the mounting medium. The preservative effects of these agents may go beyond removing oxygen because White and Stryer (1987) found propyl gallate to be more effective than thorough deoxygenation at protecting phycoerythrin in vitro. One might speculate that polyphenols like propyl gallate might quench dye triplet states and other free radicals, which could prevent forms of photodegradation other than singlet oxygen formation . Protection of GFP and fluorophores from photobleaching in fixed cells has been discussed recently by Bernas and colleagues (2004). The problem of protecting living cells from oxygen-dependent photodynamic damage is more difficult. The above antioxidants would not be attractive because they would be expected to have strong pharmacological effects at the high concentrations generally employed on fixed tissue. If the tissue can tolerate hypoxia or anoxia, one would probably prefer to remove O 2 by bubbling the medium with N2 or Ar rather than using chemical reductants. Biological oxygen-scavenging systems such as submitochondrial particles or glucose oxidase + glucose are often helpful (Bloom and Webb, 1984). If one cannot reduce the O 2 concentration, the next best tactic may be to use singlet oxygen quenchers. The most attractive here are those already chosen by natural selection, namely carotenoids. Their effectiveness is shown by classic experiments in which carotenoid biosynthesis was blocked by mutation; the resulting mutants were rapidly killed by normal illumination levels at which the wild type thrived (Matthews and Sistrom, 1959). A watersoluble carotenoid would be easier to administer acutely than the usual extremely hydrophobic carotenoids such as carotene itself. The most accessible and promising candidate is crocetin, which is the chromophore that gives saffron its color, and which consists of seven conjugated C = C units with a carboxylate at each end. Crocetin quenches aqueous singlet oxygen with a bimolecular rate constant of 5.5 x \09 M- ' ·s- ' , which is almost diffusion controlled; of this rate, about 95% represents catalytic quenching and about 5% represents bleaching or consumption of the crocetin (Manitto et aI., 1987; somewhat morc pessimistic rate constants are reported by Matheson and Rodgers, 1982). Longer-chain carotenoids are supposed to be even more efficient at destroying singlet oxygen without damage to themselves, but have not yet been tested in aqueous media. There is some evidence that 50 11M of either crocetin or etretinate (a synthetic aromatic retinoid) can protect cultured cells (L cells and WI-38 fibroblasts) from hematoporphyrin-induced photodynamic damage (Reyftmann et ai. , 1986). Other water-soluble agents that might be considered as sacrificial reactants with reactive oxygen metabolites include ascorbate (e.g., Vigers et ai., 1988), imidazole, histidine, cysteamine, reduced glutathione (Sheetz and Koppel, 1979), uric acid, and Trolox, a vitamin E analog (Glazer, 1988); these would have the advantage over carotenoids of being colorless and non-fluorescent, but would have to be used in much higher concentrations (probably many millimolar) because their bimolecular reaction rates with oxygen metabolites are not as high and they are consumed by the reaction rather than being catalytic. It should also be remembered that thorough deoxygenation may increase the triplet state lifetime, and worsen the problem of excited-state saturation.
'T
Fluorophore Concentration Increasing the concentration of fluorophore molecules will only increase the signal as long as they do not get too close together. When mUltiple ftuorophores are attached within a few nanometers
of each other on a macromolecule, they usually begin to quench each other. For example, the relatively high quantum yield for free fluorescein in aqueous solution at pH 7 is reduced to near 0.25 when an average of 5 fluorophores are bound to each IgG antibody (Southwick et ai., 1990). Charge-transfer interactions with tryptophanes are yet another mechanism for quenching fluoresceins bound to a protein, for example, anti-fluorescein antibody (Watt and Voss, 1977). Most rhodamines and certain cyanines also show a striking reduction in average quantum yield on conjugation, which appears to arise from dye interactions on the protein surface. Absorption spectra of labeled antibodies clearly show evidence of dimers, and fluorescence excitation spectra demonstrate that these dimers are not fluorescent. The propensity of relatively nonpolar, planar rhodamines to interact with one another is not surprising. Even if the fluorophores do not form ground-state dimers, they can also rapidly transfer energy from one to another until a quencher such as O 2 is encountered. In other words, proximity-induced energy transfer between f1uorophores multiplies the efficacy of quenchers. Perhaps fortunately it also shortens the excited-state lifetime, so that a higher intensity of laser excitation can be applied for a given degree of saturation (Hirschfeld, 1976). If such increased intensity is available, much of the emission intensity lost by f1uorophore proximity can be regained, but at the cost of increased background signal from Rayleigh and Raman scattering and other non-saturated f1uorophores.
Choice of Fluorophore Perhaps the most drastic alteration is to change to a different f1uorophore altogether. Unfortunately, there is not a wide selection of fluorescent physiological indicator probes that respond selectively to a cellular parameter and can be excited at appropriate AS. However, with fluorescent labeling reagents, it is sometimes possible to choose an optimal f1uorophore within a defined A range. A selection of fluorescent labels for confocal microscopy is described below.
flUORESCENT LABELS FOR ANTIBODIES, OTHER PROTEINS, AND DNA PROBES
Fluorescent Organic Dyes Certain fluorescent reagents carry reactive groups for covalent attachment to target biomolecules and show minimal spectral sensitivity to environmental changes (Table 16. 1). These reagents are primarily used to quantify the presence and distribution of probes and targets. The most well-known dyes of this type are derivatives of fluorescein and rhodamine (e.g., tetramethyl rhodamine, lissamine rhodamine, and sulforhodamine 101). A variety ofreactive groups have been incorporated into these dyes permitting coupling of the dyes to different functionalities. Isothiocyanates, succinimidyl esters, and pentafluorophenyl esters couple with amino groups of target molecules, while haloacetamides, maleimides, and vinyl sulfones react with sulfhydryl groups. Dichlorotriazinyl (DCT) groups can couple effectively with both amines and alcohols, depending upon reaction pH and temperature. Reactivity and selectivity of the conjugation reactions can be manipulated by the reactive groups chosen and the conditions of the coupling reaction (e.g., reaction pH, solvent polarity, and temperature). The cyanine dyes (Mujumdar et ai., 1993), the boratedipyrromethene (BODIPY) complexes (Wories et ai., 1985; Kang et ai., 1988), and the AlexaFluor dyes (Panchuk-Voloshina
Fluorophores for Confocal Microscopy: Photophysics and Photochemistry • Chapter 16
et ai., 1999) have been developed to complement the traditional fluorescein and rhodamine reagents. The sensitivity of detecting fluorescent conjugates is determined by the spectral properties of the fluorescent dye (its molar extinction coefficient and fluorescence quantum yield) and the quality of the dye-target conjugate (i.e., its tendency to precipitate and its degree of fluorescence quenching). The cyanine reagents, and later the AlexaFluor dyes, were engineered to have very high water solubility as well as cxcellent spectral properties. Galbraith and colleagues (1989) and Mujumdar and colleagues (1993) have shown that, at least in the case of cyanine dye labeling agents, appropriate placement of charged sulfonate groups on the f1uorophore can reduce dye interactions and increase the brightness of relatively heavily labeled antibodies. Particularly useful in this regard are the indopentamethine-cyanines, CY 5 dyes (excitation, 630-650 nm; emission 670nm) and indotrimethine-eyanines, CY3 dyes (excitation, 530-550nm; emission 575nm). Antibodies labeled with these dyes have a brightness comparable to or brighter than f1uoresceinlabeled antibodies and have little tendency to precipitate from solution even when labeled with as many as 10 dye molecules/antibody (Wessendorf and Brelje, 1992). CY5 is somewhat more photostable than fluorescein, and CY3 is significantly more stable. CY5 can be optimally excited with a red HeINe laser, whereas CY3 can be excited fairly efficiently with the 514nm line and marginally well with the 488 nm line of the argon-ion laser. These f1uorophores are useful for nucleic acid labeling and have found wide use in in situ hybridization and gene expression assays. Fluorescent reagents, structurally related to the cyanines, with a squaric acid group replacing part of the polymethine chain have been described (Oswald et ai., 1999). Also, novel polymethine dyes developed by Czerney's group were introduced recently as fluorescent reagents (Czerney et af., 2001). These new fluorescent reagents in combination with fluorescein, the phycobiliproteins, or other f1uorophores make it possible to do multi-color fluorescence imaging with laser-scanner microscopes equipped with an argon and a HeiNe laser or with the argonlkrypton "white light" laser, which has lines at 488, 568, and 647 nm. Brelje and colleagues (1993) have described the use of the Ar/Kr laser to excite samples stained with fluorescein (488 excitation); Texas Red, Lissamine, or CY3 (568nm); and CY5 (647). Fluorescent labeling reagents available commercially are described on the following Web sites: http://www.amershambiosciences.com; http://www.probes.com; http://www.dyomics.com; and http://www.mobitec.de.
Phycobiliproteins Phycobiliproteins (Oi et ai., 1982) currently hold the record for the highest extinction coefficients and largest number of photons emitted before bleaching (QJQb; Mathies and Stryer, 1986), partly because each macromolecule simply contains a large number of component f1uorophores, partly because the proteins have been engineered by natural selection to protect the tetrapyrrole f1uorophores from quenching processes (Glazer, 1989; Sun et af., 2003). Suitable optimization of laser power, optics, flow rate, and detection permits the detection of fluorescence pulses from single phycobiliprotein molecules in flowing systems (Peck et aI., 1989; see also Nguyen et ai., 1987). Phycoerythrin (PE) in combination with fluorescein has been valuable for immunofluorescence determination of cell-surface markers by single laser (488 nm excitation) flow cytometry. This approach can also be useful in confocal microscopy, provided that the emission signal is split into a 530nm (fluorescein) and a 575 nm component (PE) and two photomultipliers are used for detection. Detection of a third color is possible
343
using another photosynthetic protein, PerCP (Rechtenwald, 1989), or the PE tandem conjugates formed by coupling energy acceptor f1uorophores that emit at long A-S to PE. PE conjugates with Texas Red or to CY 5 (Waggoner et af., 1993) have become popular for cell surface measurements with flow cytometers using a 488 nm laser. Phycobiliproteins have also been coupled with CY7 dyes providing additional choices of fluorescence colors (Gerstner et ai., 2002; Roederer et ai., 1996; Beavis and Pennline, 1996). The use of phycobiliproteins is likely to prove less useful for intracellular antigens because the size of a PE-antibody conjugate, around 410 kD, restricts penetration into denser regions of fixed cells and tissues. Efforts to develop low-molecular-weight analogs of PE that have similarly large Stokes shifts and excitation A-S have not yet been successful, but work in this area continues. If fluorophores with large Stokes shifts could be found for both 488 nm and 633 nm excitation, a two-laser microscope could obtain fourcolor immunofluorescence images. Other probes are listed in Table 16.1.
DNA Probes An application of fluorescent labels that has attracted a number of investigators is fluorescence in situ hybridization, or FISH (Trask, 1991). Initially, DNA to be used to probe genetic sequences in chromosomes and interphase nuclei was labeled by nick translation using biotin or digoxigenin-tagged deoxynucleotide triphosphates (dNTPs). Fluorescent secondary reagents were applied after hybridization to detect the binding of the DNA probe. The use of dNTPs attached by linker arms to f1uorophores to form directly labeled fluorescent DNA probes is replacing methods involving fluorescent secondary reagents. Directly labeled DNA probes are simpler to use, often give less background and provide easier access to multi-color-multi-sequence detection (Ballard and Ward, 1993). However, for detection of short genetic sequences, use of multiple fluorescent secondary reagents may be required for sufficient sensitivity. Directly labeled oligonucleotides provide alternatives when there are numerous copies of the target sequence, as in the case of histone mRNA (Yu et at., 1992). For non-specific, but stable labeling of DNA, the bis-intercalating reagents, TOTO, YOYO, DRAQ5, etc., provide an attractive solution, and the reagents are available in several fluorescent colors (Rye et ai., 1992). The anthraquinone derivative, DRAQ5, is membrane permeant, is highly sclective for nuclear DNA, and can be used with two-photon excitation from 800 nm to beyond 1000 nm. Studies of Ii ving cells with DRAQ5 must be done with care because of its high cytotoxicity (Smith et at., 2000; Errington et ai., 2005). Snyder (2003) suggests that some caution may be needed when combining DRAQ5 with other probes. There appears to be decreased uptake of bodipylabelled compounds in the presence of the nuclear stain, DRAQ5.
Luminescent Nanocrystals Fluorescent nanocrystals, or quantum dots, exhibit interesting properties for biological labeling reagents. These nanometer-sized inorganic crystalloid structures have broad absorption (and excitation) spectra with molar absorptivities of more than six million at 450 nm. The emission band, determined by particle composition and dimensions, can be very narrow (ca. 25 nm FWHM) with quantum yields approaching unity. Also, quantum dot luminescence is very resistant to photobleaching. Selection of uniformly sized quantum dots gives preparations with very sharp fluorescence bands. The challenge for biological applications is to provide biocompatible surfaces for the nanocrystals that maintain their fluorescence in
344
Chapter 16 • R.Y. Tsien et al.
TABLE 16.1. Spectroscopic Properties of Selected Probes Parameter
Probe"
Absorption Maximum b
Covalent labeling reagents
Fluorescein-amines, sulfhydryl Tetramethylrhodamine-amines X -rhodamine-amines Texas Red®-amines CY3 CY5 CY7 BODIPY® FL BODIPY 5811591 BODIPY 630/650 Cascade Blue® AlexaFluor® 430 AlexaFluor 488 AlexaFluor 532 AlexaFluor 594 NBD-amine
490 554 582 596 554 652 755 502 581 625 378, 399 430 494 530 590 478
67 85 79 85 130 200 200 80 136 101 26 15 73 81 92 24.6
520 573 601 620 568 672 778 510 591 640 423 545 517 555 617 520-550
0.36/0.21
NBD-S-CH2CH20H
425
12.1
531
0.002
Coumarin-phalloidin DY-555
387 555
100
470 580
DY-63 I
637
185
658
480-565 650 488 365 & higher 340 340 383 434 489 514 558 340 350 646 510 536 480
1960 700
578 660 680 400-850+ 615 545 445 477 508 527 583 450 470 681 595 623 520
Nanocrystals Lanthanide chelates Expressible labels
DNA-RNA content'
Membrane potential
Phycoerythrin-R Allophycocyanine PerCP (quantum dots) Europium chelate Terbium chelate EBFP ECFP EGFP EYFP DsRed Hoechst 33342 DAPI DRAQ5 Ethidium Bromide Propidium Iodide Acridine Orange
Extinction Maximum'
-6000@400nm Determined by ligand 31 26 55 84 72.5 120 21 3.2 6.4
Emission Maximumb
Pyronine Y
440-470 549-561
67-84
650 567-574
Thiazole Orange TOTO-I YOYO-3 diO-Cn-(3) diI-Cn-(5) diBA-Isopr-(3) diBA-C4-(5) Rhodamine 123
560-562 497 453 514 612 485 646 493 590 511
70-90 42 26 112 115 149 200 130 176 85
565-574 563 480 533 631 505 668 517 620 534
Quantum Yield 0,71 0,28 0,26 0.51 0.14 0.181 0.02"
0.68 0.68
See FN i See FNj 0.25 0.40 0.60 0.61 0.68 0.83
0.09
0.04-0.26
0.05-0.21 Low 0.08
0.05 0.4 0.03 0.9
Measurement Conditions pH 7, PBS pH 7, PBS pH 7, PBS pH 7, PBS pH 7, PBS pH 7, PBS pH 7, PBS MeOH MeOH MeOH Water pH 7 pH 7 pH 7 pH 7 EtOHlMeOH
pH 7.5,10% glycerol Water
pH 7, PBS pH 7, PBS pH 7, PBS PBS Water pH 7, PBS pH 7, PBS pH 7, PBS pH 7, PBS pH 7, PBS +DNA (excess) +DNA (excess) PBS +DNA (excess) +DNA (excess) +DNA +RNA +ds DNA'
+ds RNA' +ss RNA RNA MeOH MeOH MeOH MeOH MeOH EtOH EtOH
Referencesd Haugland (1983), W, MP Haugland (1983), MP W Titus et al. (1982), W, MP Mujumdar et al. (1993) Mujumdar et al. (1993) Mujumdar et al. (1993) MP MP MP MP MP MP MP MP Kenner & Aboderin (1971), Allen & Lowe (1973), Bratcher (1979)
MP MoBiTec GmbH (http://www.mobitec.com) MoBiTec GmbH (http://www.mobitec.com) Oi et al. (1982) Oi et al. (1982) Rechtenwald (1989) Paul (2002) Paul (2002) Patterson et al. (200 I) Patterson et al. (2001) Patterson et al. (200 I) Patterson et al. (200 I) Patterson et al. (200 I) W W Smith, Pl et al. (2000) Pohl et al. (1972) W Kapuscinski et al. (1982), Shapiro (1985) Darzynkiewicz et al. (1987), Kapuscinski & Darzynkriwicz (1987)
Lee et al. (1986) MP, Rye et al. (1992) MP, Rye et al. (1992) Sims et al. (1974), W Sims et al. (1974), W Sims et al. (1974), W W EK, Kubin & Fletcher (1983)
Fluorophores for Confocal Microscopy: Photophysics and Photochemistry • Chapter 16
345
TABLE 16.1. (Continued) Parameter pH
Membrane location and fluidity
Calcium"
Probe"
Absorption Maximum b
505 460 SNARF-I (pKa = 7.5) 518-548 574 DCDHB 340-360 340-360 Diphenylhexatriene (DPH) 330, 351, 370 diI-CI8-(3) 546 484 DiO 491 DiA NBD phosphatidylethanolamine 450 361,381 Anthroyl stearate Pyrene-sulfonamidoalkyls 350 Fura2 335 360 Indo I 330 350 Fluo-3 506
Extinction Maximum'
BCECF
506
Emission Maximumb
Quantum Yield
530
77 (35Inm) 126 149 52
24g 8.4,7.5 30 33 27 34 34 83
78
365/450
587 636 500-580 420-440 430 565 501 613 530 446 380-400 512-518 505-510 390-410 482-485 526
526 590
0, sensor
Ru[dpp(S0 3Na)2h
cAMP Enzyme substrates
FIERhR Rhodamine-di-arg-CBZ Product of rxn. (rhodamine) Coumarin-glucoside substr.
490 495 495 316
67 13
520, 573 532 523 395
Rxn. product (hydroxy coumarin) Fluorescein digalactosidase product Resorufin galactosidase product ELF97 phosphatase product
370
17
450
490
67
520
0.07
1: 1:
571 345
70
58 Precipitate
585 530
Measurement Conditions High pH Low pH pH 5.5 pH 10.0 High pH Low pH Hexane MeOH MeOH MeOH Lipid MeOH
0.23 0.49 0.56 0.38 0.183
Low calcium High calcium High calcium Low calcium High calcium
0.0051
Low calcium No oxygen
= 3.7~s = 0.93~s 0.09 0.91
0.71
References d MP MP MP MP Valet et al. (1981), MP MP W MP MP Struck et al. (1981) Waggoner & Stryer (1970) MP Grynkiewicz et al. (1985) Grynkiewicz et al. (1985) Minata et al. (1989) & Mukkala et al. (1993), MP Castellano & Lakowicz (1981)
Air pH 7, cAMP Hepes pH 7.5 +15% EtOH pH5.5+1% Lubrol pHIO+ 1% Lubrol pH 7, PBS
W
pH 9 pH 8
MP MP
Adams et al. (1991) Leytus et al. (1983)
W MP
"Abbreviations: NBD, 7-nitrobenz-2-oxa-I,3-diazole; DAPI, 4',6-diamidino-2-phenYlindole; DCDHB, dicyano-dihydroxybenzene. b Measured in nanometers. 'Multiply value listed by 1000 to get liters/mol· em. dEK, Eastman Kodak Chemical Catalog; MP, Molecular Probes, Inc catalog; W, Waggoner laboratory determination. 'See Table III in Amdt-Jovin & Jovin (1989) for additional DNA content probes. t Base-pair dependent. 'Value for NBD-ethanolamine in MeOH which has an abs.max at 470nm and an emission max at 550nm [Barak & Yocum (1981)]. h See Tsien (1989) for additional details and other ion indicators. 'See corporate Web sites: Quanturm Dot Corp. (www.qdots.com), Evident Technologies (www.evidenttech.com), BioCrystal, Ltd. (www.biocrystal.com) and Crystalplex Corp. (www.crystalplex.com). I Time resolved detection. Extinction times quantum yield approx. 2100.
aqueous media. Promising results have been obtained using quantum dots (laiswal et at., 2003; Hoshino et at., 2004; Voura et at., 2004). However, there is ample opportunity for significant improvements before these reagents can be used routinely for fluorescent labeling. Their relatively large size and high mass limit their use in applications requiring high diffusional mobility.
Fluorescent Lanthanide Chelates Lanthanide chelates are another group of fluorescent reagents with special spectral properties. These reagents have microsecond fluorescence lifetimes that are readily distinguished from typical nanosecond autofluorescence background (Soini et at., 1988; Seveus et at., 1994; Vereb et aI., 1998). Temporal separation of
probe fluorescence from background signal can give a very high SIN and highly sensitive probe detection even though the brightness of these reagents is only modest. The main advantage of these compounds is that, as their fluorescent properties are dependent on electron energy levels in an atom rather than those of a molecule, they are very resistant to photodamage (and perhaps also to phototoxicity). On the other hand, it is possible that the excitation light may break the bond with the chelator causing the "dye atom" to become a non-specific stain. The current versions of these reagents require enhanced antennary ligands for more efficient lanthanide excitation and improved biological stability and compatibility. On the other hand, the long decay times give rise to problems when they are used in scanning microscopes. When the decay time is equal to the pixel dwell time,
346
Chapter 16 • R. Y. Tsien et al.
each molecule can be excited no more than one time and signal levels are very low. As a result, these dyes will probably only be used with widefield imaging where the exposure time seen by each molecule is long compared to its fluorescent lifetime and discrimination from the fast decay of the autofluorescence is still useful.
FLUORESCENT INDICATORS FOR DYNAMIC I NTRACELLULAR PARAMETERS Membrane Potentials The use of confocal microscopy to measure dynamic properties of living cells such as membrane potentials (Gross and Loew, 1989) or ion concentrations (Tsien, 1988, 1989a,b) deserves some special comment. Preliminary attempts to use fast-responding, nonredistributive voltage-sensitive dyes in neuronal tissues were unsuccessful (Fine et ai., 1988); the dye could be seen, but the SIN was inadequate to observe voltage-dependent changes, which would have been at most only a few percent of the resting intensity. Lasers are inherently noisy light sources; even with optical negative feedback, their fluctuations are greater than the stabilized tungsten filament lamps conventionally used to see the small changes in fluorescent output that characterize fast voltagesensitive dyes (Cohen and Lesher, 1986). "Slow" redistributive dyes, which accumulate in cells according to the Nernst equilibrium (Ehrenberg et ai., 1988), would seem to be more suitable for present-day confocal microscopes because their signals are much bigger and the slowness of their response (seconds to minutes) is actually a better match to the rather slow scan times of the current instrumentation. Confocal optical sectioning should work well using such accumulative dyes because in principle one could directly compare the internal concentrations of dye accumulated without having to correct for the greater path length of a thicker cell or for extracellular dye above and below the plane of focus. Freely diffusing anionic oxonol dyes have been paired with dyes anchored at the cell surface permitting fast ratiometric detection of membrane potential changes in single cells by fluorescence resonance energy transfer (Gonzalez and Tsien, 1995, 1997; Gonzalez and Maher, 2002) (see also Figure 8.45, this volume).
Ion Concentrations Wavelength Ratioing Some indicators of ion concentrations respond not just with changes in fluorescence amplitude but also with A shifts of the excitation or emission spectrum or both. Such shifts permit ratioing between signals obtained at two or more As (Tsien, 1989a,b). Ratioing is highly valuable because it cancels out differences in dye concentration and path length as well as fluctuations in overall illumination intensity (Tsien and Poenie, 1986; Bright et al., 1987). Emission ratioing is the most valuable because, with a single excitation A, the emission can be passed through a dichroic mirror to split it into two bands that can be monitored absolutely simultaneously. Such ratioing would give the best possible cancellation of laser noise or specimen movement. Emission ratioing is particularly easy to do with a laser-scanning system, because one can simply add a dichroic mirror and an extra photodetector after the scanning system. Whereas geometrical registration of all the corresponding pixels in two separate low-light-level video cameras is quite difficult (Jericevic et ai., 1989), the registration problem is trivial in a laser-scanning system assuming that the deflection is
achromatic, which it must be in order to get excitation and even one emission in register. Disk-scanning confocal microscopes use charge-coupled device (CCD) or electron multiplying CCD (EMCCD) cameras as detectors and often lack this elegant compatibility with emission ratio scanning. Excitation ratioing of images requires sequential illumination with the two excitation As. Intensity fluctuations of the source and movement of the specimen are canceled out only if they are much slower than the rate of alternation. Excitation ratioing is most applicable to tandem-scanning systems where conventional systems for alternating two grating monochromators or interference filters could be used. Alternating between two laser lines is more convenient now that acousto-optical deflectors are common, but it is still less flexible in choice of A pairs.
pH Indicators A number of ratiometric pH indicators were reviewed by Tsien (l989b) and more recently by Yip and Kurtz (2002). The most popular excitation-ratioing indicator is probably the modified fluorescein, BCECF, whose pH-sensitive and insensitive As are around 490nm and 439nm, respectively. Several emissionshifting probes, 3,6-dihydroxyphthalonitrile (also known as 2,3-dicyanohydroquinone; Kurtz and Balaban, 1985; Kurtz and Emmons, 1993), and various naphthofluorescein derivatives (SNAFs and SNARFs; Haugland, 1989) are also available.
Ca 2+ Indicators Three currently available Ca2+ indicators have different sets of advantages and disadvantages for confocal microscopy (Tsien, 1988, 1989a,b). Fura-2, the dye most used in conventional microscopic imaging, shows a good excitation shift with Ca2 +, typically ratioed between 340 to 350 nm and 380 to 385 nm, but hardly any emission shift, so it would be most effectively used with a UVenhanced disk-scanning instrument. Considerable re-engineering would be necessary for those early designs of tandem-scanning confocal microscope designed mainly for reflectance rather than fluorescence. The beam-splitting pellicles used are inefficient because they are partially reflective but not dichroic; also, they are sometimes made from UV-blocking material in which the excitation has to pass through the pellicle whereas the emission would have to reflect off the pellicle. Even if the pellicle were replaced by a dichroic, this choice of beam geometry is unfortunate, as it is much easier to make good broadband dichroics in which the shorter A reflects and the longer A transmits than vice versa. Disk systems in which the same area of the disk is used for both source and detector may be more flexible (see Chapter 10, this volume). Indo- I, the dye most used in laser flow cytometry for [Cart] determination, shows a fine emission shift from 485 to 405 nm with increasing Ca2+ and is preferred for ratiometric laser scanning. However, either a UV laser (e.g., a high-power argon-ion or krypton-ion system) or a titanium-sapphire two-photon system is required for excitation in the 350 to 365 nm region. Also, Indo- I fluorescence has As similar to those of reduced pyridine nucleotides, so autofluorescence could be a problem, and Indo-l also bleaches much more quickly than Fura-2. Fluo-3 and its less-tested rhodamine analogs are the only Ca2+ indicators currently available with visible As suitable for lowpower visible lasers (Minta et ai., 1989; Kao et ai., 1989). Therefore, it has been the first to be exploited in confocal microscopy (e.g., Hernandez-Cruz et ai., 1989), even though it lacks either an
Fluorophores for Confocal Microscopy: Photophysics and Photochemistry • Chapter 16
excitation or emlSSlon shift and is restricted to simple intensity measurements that are relatively difficult to calibrate in terms of absolute [Ca2+li units. Of course, the ideal would be an indicator excitable at 488 nm with a large emission shift, high quantum efficiency, and strong resistance to bleaching, but this goal is a difficult challenge in molecular engineering. In general, strong fluorescence in aqueous media is much easier to obtain using shorter excitation As because fluorescence demands planarity and molecular rigidity, which is obviously easier to achieve in small molecules that absorb short As than in the larger molecules with longer chromophores. Most of the known chromophores that combine large size, long As, and rigidity are essentially insoluble in water. Even if solubilizing groups are added on the periphery, the huge expanse of hydrophobic surface still promotes the formation of non-f1uorcscent aggregates. Finally, the quantum mechanics of absorption and fluorescence predict that the intrinsic radiative lifetime of a chromophore is proportional to the cube of the A if other factors remain constant (Strickler and Berg, 1962). Short radiative lifetimes mean that fluorescence emission competes more successfully with nonproductive forms of deactivation and, therefore, correlates with high quantum yields of fluorescence.
Oxygen Sensor Much effort has been spent developing ruthenium chelates as luminescent reagents. Common ligands for the central ruthenium ion include substituted bipyridines and I,IO-phenanthrolines. Currently, the most promising applications for these complexes are for monitoring the oxygen tension of solutions (Li et aI., 1997; Castellano and Lakowicz, 1998; Ji et aI., 2002). As with the lanthanide complexes, the modest excitation efficiencies are dependent on energy transfer from the ligands to the metal ion. The emission spectra are relatively weak and broad. However, timeresolved detection of the long-lived ruthenium fluorescence yields excellent SIN even from these low intensities. The presence of oxygen, even at the levels found in air, decreases the fluorescence lifetimes of ruthenium chelates by a factor of 4, a factor that should make them suitable for fluorescence lifetime imaging (FUM; see Chapter 27, this volume), now that the equipment is commercially available. Improved ligands with enhanced absorbance at visible wavelengths, more efficient energy transfer to the metal, and better biocompatibility are needed before these reagents find broader applications.
cAMP Indicators The important intracellular messenger cAMP (cyclic adenosine 3,5-monophosphate) can now be imaged with a fluorescent indicator made from cAMP-dependent protein kinase labeled on its C and R subunits with fluorescein and tetramethylrhodamine respectively (Adams et aI., 1991). In the holoenzyme complex, the fluorescein and rhodamine are close enough for moderately efficient fluorescence resonance energy transfer, so that excitation of the fluorescein with blue-green light gives a significant amount of orange emission from the rhodamine. Binding of cAMP dissociates the subunits and eliminates energy transfer, increasing the emission of green light directly from the fluorescein and decreasing the amplitude in the rhodamine band. This change in emission ratio and the As employed are ideal for dual-channel detection by confocal microscopy (Bacskai et ai., 1993). The strategy of labeling an important endogenous sensor protein gives both advantages and potential problems. Because careful derivatization of the
347
kinase does not change its cAMP affinity and phosphorylating activity, the indicator is inherently tuned to the physiologically relevant concentration range (a few nanomoles to a few examoles), and molecules of cAMP that bind to the indicator can still have a biological etfect. An indicator that was not a physiological effector molecule would have a greater tendency to competitively inhibit or buffer the pathway under study. Furthermore, after elevation of cAMP, the interesting trafficking of the Rand C subunits can be separately observed by standard dual-label imaging (e.g. Harootunian et aI., 1993). However, scrambling of subunits with unlabeled endogenous kinase is a potential problem, so far rarely serious (for discussion of this point and a more extensive review of the entire technique, see Adams et aI., 1993).
Fatty Acid Indicator Fatty acids are of considerable importance in nutrition, membrane structure, protein modification, eicosanoid formation, and modulation of cell signaling. Recently a group led by Alan Kleinfeld has developed an emission-ratioing fluorescent probe for free fatty acid levels (Richieri et aI., 1992) by labeling recombinant intestinal fatty-acid-binding protein with acrylodan, an environmentally sensitive f1uorophore. The acrylodan reacts with surprising specificity for Lys27 of the protein and probably resides in the fattyacid binding pocket. Binding of fatty acids shifts the emission peak from 432 nm to 505 nm, probably by displacing the acrylodan into an aqueous environment. The 505 nml432 nm ratio thereby increases by up to 25-fold, which would be an ideal signal for confocal microscopy if the necessary excitation at 386nm or 400nm were available. The labeled protein (dubbed ADIFAB) binds all common long-chain fatty acids with approximately micromolar dissociation constants. In a cuvet, ADIFAB can detect free fatty acid concentrations as low as a few nanomolar. It has already proven highly useful in measuring the release of free fatty acids from stimulated rat basophilic leukemia (RBL) cells (Richieri et aI., 1992) and the binding constants of fatty acids to albumin (Richieri et aI., 1993) and to cells (Anel et ai., 1993).
Other Forms of Ratioing Because ratioing is so desirable for quantitative measurements, but appropriate A shifts are often unavailable, several alternatives to A ratioing have been proposed. The easiest is simply to ratio poststimulus image intensities against a prestimulus image. An example is shown by Smith and Augustine (1988). This method has the advantage of minimal hardware requirements and high time resolution, though it only cancels out variations in dye loading and path length, not shape change or dye bleaching, and by itself cannot yield an absolute calibration of the analyte, for example, [Ca2+1. Another approach would be to link the fluorescent indicator covalently to a separate reference f1uorophore. This approach would ideally generate a composite molecule in which the ratio of the indicator fluorescence to the reference fluorescence would signal the analyte concentration. Potential disadvantages would be the requirement for significant skill in organic synthesis, the likelihood that the conjugate would be too large for loading by ester hydrolysis, and the possibility that the two f1uorophores would bleach at different rates, so that the operation of ratioing would fail to correct for bleaching. Yet a third mode of ratioing could be based on temporal dissection of excited-state lifetimes, as first shown for quin-2 by Wages and colleagues (1987). If the free and bound forms of the indicator have sufficiently different fluorescence lifetimes, their relative contributions to the (ideally) biexponential decay might be separated by nanosecond or high frequency mod-
348
Chapter 16 • R. Y. Tsien et al.
ulation techniques (see Chapter 27, this volume). However, even when the instrumentation challenge of combining lifetime kinetics with imaging has been solved, the problem remains that probes like fura-2 and Indo-I, which are fairly strongly fluorescent both when free and when bound to Ca2+, have almost the same lifetimes in those two states. For example, fura-2 with and without Ca2+ has lifetimes of 1.8 ns and 1.3 ns, respectively, at 25°C; for Indo-l the corresponding numbers are 1.7 ns and 1.3 ns at 20°C (Wages et al., 1987). In order to have a significant difference in lifetimes between Ca2+-bound and free indicator, as in quin-2 (10.1 ns and 1.3 ns, respectively, at 25°C), one of the species has to be much more dimly fluorescent than the other. As a result, the weaker and faster component will be hard to measure accurately and to distinguish from autofluorescence background. Ratiometric measurements can also be applied to other parameters such as probe polarization and local viscosity (Tinoco et aI., 1987; Axelrod, 1989; Dix and Verkman, 1989), proximity between macromolecules by fluorescence energy transfer (Uster and Pagano, 1986; Herman, 1989), and even water permeability (Kuwahara and Verkman, 1988; Kuwahara et al., 1988).
GENETICALLY EXPRESSED INTRACELLULAR FLUORESCENT INDICATORS
Green Fluorescent Protein Fluorescent proteins from jellyfish and corals have revolutionized biological optical microscopy because they provide genetic encoding of strong visible fluorescence of a wide range of colors. Entire books (Chalfie and Kain, 1998; Sullivan and Kay, 1999; Hicks, 2002) have been devoted to many aspects of the prototypical fluorescent protein, the green fluorescent protein (GFP) from the jellyfish Aequorea victoria. A shorter, relatively self-contained introduction to GFP may be found in Tsien (1998). Other general reviews on applications of GFP and other members of the fluorescent protein superfamily include (Cubitt et aI., 1995, 1999; Hassler, 1995; Niswender et aI., 1995; Rizzuto et aI., 1995; Steams, 1995; Kahana and Silver, 1996; Misteli and Spector, 1997; Patterson et al., 1997; Tsien and Miyawaki, 1998; Ellenberg et al., 1999; Heim, 1999; Lippincott-Schwartz et aI., 1999; Phillips, 1999; Piston et al., 1999; Chamberlain and Hahn, 2000; Miyawaki and Tsien, 2000; Sacchetti et aI., 2000; Zaccolo and Pozzan, 2000; Zacharias et aI., 2000; Blab et aI., 2001; Chiesa et al., 2001; Harms et aI., 2001; Lippincott-Schwartz et al., 2001; Patterson et aI., 2001; Reits and Neefjes, 2001; Wahlfors et aI., 2001; Labas et aI., 2002; Matz et al., 2002; Miyawaki, 2002; van Roessel and Brand, 2002; Zacharias, 2002; Zhang et al., 2002; Zimmer, 2002; Choy et aI., 2003; Ehrhardt, 2003; Hadjantonakis et al., 2003; Lippincott-Schwartz and Patterson, 2003; Lippincott-Schwartz et aI., 2003; March et aI., 2003; Meyer and Teruel, 2003; Miyawaki, 2003; Tsien, 2003; Viallet and Vo-Dinh, 2003; Weijer, 2003; Verkhusha and Lukyanov, 2004). Space does not permit listing the huge number of reviews and primary papers describing more specialized uses of fluorescent proteins.
ligand-Binding Modules Genetic manipulations have also been used to incorporate into cells expressible modules that bind fluorescent ligands. One type of module contains a tetracysteine motif, which binds biarsenical ligands (Griffin et al., 1998; Adams et aI., 2002; Nakanishi et aI., 2004). These small, membrane-permeant ligands bind with high
affinity and specificity to the four sulfhydryls arranged in an alpha helix domain. Biarsenical fluorescein (FlAsH), tetramethyl rhodamine (TrAsH), and a few additional dyes have been reported. In a different approach, cells transfected with the sequence encoding a single-chain antibody (scFv) expressed a module that tightly bound cell-permeant hapten-fluorophore conjugates (Farinas and Verkman, 1999). Incorporation of several scFvs would enable multiple ligands to be detected simultaneously. Like the expressible GFPs, these modules can be directed to specific intracellular compartments by including the appropriate localization sequences.
Ion Indicators Mutants of GFP have been identified that show pH-dependent fluorescence properties (Kneen et al., 1998; Llopis et aI., 1998). Reversible absorbance and fluorescence emission changes were observed with apparent pKa values ranging from 4.8 to 7.1. Combining the GFP sequences with specific targeting signals permitted the acidification of specific organelles to be followed noninvasively. Genetically encoded Ca2+ indicators, dubbed "chameleons," utilize fluorescence resonance energy transfer (FRET) between different emitting GFPs attached to calmodulin and a calmodulin-binding peptide, M13 (Miyawaki et aI., 1997, 1999). Binding of Ca 2+ to calmodulin increases the interaction between the GFPs. Optimization of the relative orientation of the two GFPs in the chimeras has expanded the dynamic range of Ca2+ detection (Nagai et al., 2001, 2004).
FUTURE DEVElOPMENTS Speculation on future directions in fluorophore designs is difficult because the small number of laboratories working on fluorophore chemistry makes progress a much noisier function of time than advances in instrumentation or computers. One major advance seen since publication of the previous edition of this book is a recognition of the optimal characteristics for fluorescent reagents used in biology. A thought prompted by preparation of this review is that, for present purposes, the excited triplet state of the fluorophore is a major villain without any redeeming virtues. It is responsible for a pernicious form of output saturation, for singlet oxygen production, and for nearly all covalent photochemistry such as bleaching. Similar problems have been encountered in laser dyes; a proposed solution (Liphardt et al., 1982, 1983; Schafer, 1983) is to attach triplet-state quenchers to each fluorophore. Such a construction is reminiscent of the way that evolution has assembled photosynthetic complexes and may be an area where biology can repay its debt to synthetic chemistry. Little progress has been reported in this area during the two decades since these approaches were proposed, but the potential gains from triplet-state relaxation maintain this as an attractive area for study. The development of fluorescent inorganic nanocrystals and chelates may provide the needed photostable biological tagging reagents, however. Another area ripe for development is signal amplification schemes for detecting low copy numbers of cell receptors and genetic sequences. Use of enzymes that produce fluorescent precipitates in a localized area offers promise (Haugland, 1989); again reported progress is limited. Finally, there is always need for additional sensitivity. Timeresolved fluorescence detection, where the fluorescence signal is collected against a dark background after pulse excitation, offers a method to circumvent autofluorescence and Raman light scat-
Fluorophores for Confocal Microscopy: Photophysics and Photochemistry • Chapter 16
tering that is becoming much more widely available (Lakowicz et al., 1983; Marriott et aI., 1991; Chapters 27 and 31, this volume). This approach is now realized with sophisticated excitation/detection components and will work with any fluorophore with an excited state lifetime in the nanosecond time range. Quantum dots, with fluorescence lifetimes in the range of 20 ns, may prove particularly useful, but their early-stage development as biological labeling reagents has not provided sufficient incentives for the required detection system modifications. Another ingenious approach would involve excitation of an extended lifetime fluorophore with a scanning laser followed by detection of the signal with an array detector, such as a cooled EM-CCD camera, instead of a single photomultiplier tube. The signal could be integrated on the cleared pixels of the CCD chip, even milliseconds after the excitation beam has passed the corresponding region of the sample. The extended lifetime labels could be lanthanide complexes with millisecond lifetimes that have been developed as protein-labeling reagents by Hemmila and others (Soini et al., 1988; Mukkala, 1993). In situations with heavy autofluorescence, the lanthanide complex labels have demonstrated sensitivity over fluorescein labels by factors of hundreds (Seveus et aI., 1994). Phosphor particles developed by Beverloo and colleagues (1992) could be used in a similar fashion. Surface chemistries used to improve the biocompatibility of quantum dots may also be effective in enhancing the fluorescence properties of these phosphors in biological applications. Chemistry will continue to play a major role in furthering the power of confocal microscopy.
ACKNOWLEDGMENTS We thank Profs. Richard Mathies and Alex Glazer for helpful discussions and criticisms of the manuscript and permission to cite their unpublished data.
REFERENCES Adams, S.R, Campbell, RE., Gross, L.A., Martin, B.R, Walkup, G.K., Yao, Y, L1opis, J., and Tsien RY, 2002, New biarsenicalligands and tetracysteine motifs for protein labeling in vitro and in vivo: synthesis and biological applications, J. Am. Chem. Soc. 124:6063-6076. Adams, S.R, Harootunian, A.T., Buechler, YJ., Taylor, S.S., and Tsien, R.Y, 1991, Fluorescence ratio imaging of cyclic AMP in single cells, Nature 349:694-697. Adams, S.R, and Tsien, R.Y., 1993, Controlling cell chemistry with caged compounds, Annu. Rev. Physiol. 55:755-784. Anel, A., Richieri, G.V., and Kleinfeld, A.M., 1993, Membrane partition of fatty acids and inhibition of T cell function, Biochemistry 32:530-536. Axelrod, D. , 1977, Cell surface heating during fluorescence photobleaching recovery experiments, Biophys. J. 18:129-131. Axelrod, D., 1989, Fluorescence polarization microscopy, Methods Cell Bioi. 30:333-352. Bacskai, B.J., Hochner, B., Mahaut-Smith, M., Adams, S.R., Kaang, B.-K., Kandel, E.R, and Tsien, RY, 1993, Spatially resolved dynamics of cAMP and protein kinase A subunits in Aplysia sensory neurons, Science 260:222-226. Bailey, E.A. Jr., and Rollefson, G.K., 1953, The determination of the fluorescence lifetimes of dissolved substances by a phase shift method, J. Chem. Phys.21:1315-1322. Ballard, S.G., and Ward, D.C., 1993, Fluorescence in situ hybridization using digital imaging microscopy, J. Histochem. Cytochem. 12:1755-1759. Beavis, A.J., and Pennline, K.J., 1996, AIl0-7: a new fluorescent tandem dye for use in flow cytometry, Cytometry 24:390-395.
349
Bernas, T., Zarebski, M., Cook, RR., and Dobrucki, J.w., 2004, Minimizing photobleaching during confocal microscopy of fluorescent probes bound to chromatin: Role of anoxia and photon flux, J. Microsc. 5:281-296. Beverloo, H.B., van Schadewijk, A., Bonnet, J., van der Geest, R., Runia, R, Verwoerd, N.P., Vrolijk, J., Ploem, J.S., and Tanke, H.J., 1992, Preparation and microscopic visualization of multicolor luminescent immunophsphors, Cytometry 13:561070. Blab, G.A., Lommerse, P.H.M., Cognet, L., Harms, G.S., and Schmidt, T., 2001, Two-photon excitation action cross-sections of the autofluorescent proteins, Chem. Phys. Lett. 350:71-77. Bloom, J.A., and Webb, W.W., 1984, Photodamage to intact erythrocyte membranes at high laser intensities: methods of assay and suppression, J. Histochem. Cytochem. 32:608-616. Bonhoeffer, T., and Staiger, v., 1988, Optical recording with single cell resolution from monolayered slice cultures of rat hippocampus, Neurosci. Lett. 92:259-264. Brelje, T.e., Wessendorf, M.W., and Sorenson, R.L. , 1993, Multicolor laser scanning confocal immunofluorescence microscopy: practical applications and limitations. In: Methods in Cell Biology 38: Cell Biological Applications of Confocal Microscopy (B. Matsumoto, ed.), Academic Press, San Diego, California. Bright, G.R, Fisher, G.W., Rogowska, J., and Taylor, D.L., 1987, Fluorescence ratio imaging microscopy: Temporal and spatial measurements of cytoplasmic pH, J. Cell Bioi. 104:1019-1033. Castellano, EN., and Lakowicz, J.R, 1998, A water-soluble luminescence oxygen sensor, Photochem. Photobiol. 67:179-183. Chalfie, M., and Kain, S., 1998, Green Fluorescent Protein: Properties, Applications, and Protocols, Wiley-Liss, New York. Chalfie, M., Tu, Y, Euskirchen, G., Ward, W.W., and Prasher, D.e., 1994, Green fluorescent protein as a marker for green expression, Science 263: 802-805. Chamberlain, C., and Hahn, K.M., 2000, Watching proteins in the wild: fluorescence methods to study protein dynamics in living cells, Traffic 1:755-762. Chen, RE, and Scott, e.H., 1985, Atlas of fluorescence spectra and lifetimes of dyes attached to protein, Analyt. Lett. 18:393-421. Chiesa, A., Rapizzi, E., Tosello, v., Pinton, P., de Virgilio, M., Fogarty, K.E., and Rizzuto, R, 2001, Recombinant aequorin and green fluorescent protein as valuable tools in the study of cell signaling, Biochem. J. 355:1-12. Choy, G., Choyke, P., and Libutti, S.K., 2003, Current advances in molecular imaging: noninvasive in vivo bioluminescent and fluorescent optical imaging in cancer research, Mol. Imaging 2:303-312. Cohen, L.B., and Lesher, S., 1986, Optical monitoring of membrane potential: Methods of multi site optical measurement, In: Optical Methods in Cell Physiology (P. De Weer, B.M. Salzberg, eds.), John Wiley and Sons, New York, pp. 71-99. Cubitt, A.B., Heim, R, Adams, S.R, Boyd, A.E., Gross, L.A., and Tsien, RY., 1995, Understanding, using and improving green fluorescent protein, Trends Biochem. Sci. 20:448-455. Cubitt, A.B., Woollenweber, L.A., and Heim, R., 1999, Understanding structure-function relationships in the Aequorea victoria green fluorescent protein, Methods Cell Bioi. 58: 19-30. Czerney, P., Lehmann, E, Wenzel, M., Buschmann, V., Dietrich, A., and Mohr, GJ., 2001, Tailor-made dyes for fluorescence correlation spectroscopy (FCS), BioI. Chem. 382:495-498. Dix, J.A., and Verkman, A.S., 1989, Spatially resolved anisotropy images of fluorescent probes incorporated into living cells, Biophys. J. 55:189a. Ehrenberg, B., Montana, v., Wei, M.D., Wuskell, J.P., and Loew, L.M., 1988, Membrane potentials can be determined in individual cells from the Nernstian distribution of cationic dyes, Biophys. J. 53:785-794. Ehrhardt, D., 2003, GFP technology for live cell imaging, Curr. Opin. Plant Bioi. 6:622-628. Ellenberg, J., Lippincott-Schwartz, J., and Presley, J.E, 1999, Dual-colour imaging with GFP variants, Trends Cell Bioi. 9:52-56. Errington, R.J., Ameer-Beg, S.M., Vojnovic, B., Patterson, L.H., Zloh, M., Smith, P.J., 2005, Advanced microscopy solutions for monitoring the kinetics and dynamics of drug-DNA targeting in living cells, Adv. Drug Deliv. Rev. 57(1):153-167.
350
Chapter 16 • R. Y. Tsien et al.
Farinas, J., and Verkman, A.S., 1999, Receptor-mediated targeting of fluorescent probes in living cells, J BioI. Chem. 274:7603-7606. Fine, A., Amos, W.B., Durbin, R.M., and McNaughton, P.A., 1988, Confocal microscopy: Applications in neurobiology, Trends Neurosci. II :346351. Galbraith, W., Ernst, L.A., Taylor, D.L., and Waggoner, A.S., 1989, Multiparameter fluorescence and the selection of optimal filter sets: Mathematics and computer program, Proc. Soc. Photo. Opt. lnstrum. Eng. 1063:74122. Gandin, E., Lion, Y., and Van de Vorst, A., 1983, Quantum yields of singlet oxygen production by xanthene derivatives, Photochem. Photobiol. 37:271-278. Gerstner, A.O., Lenz, D., Laffers, W., Hoffman, R.A., Steinbrecher, M., Bootz, E, and Tarnok, A., 2002, Near-infrared dyes for six-color immunophenotyping by laser scanning cytometry, Cytometry 48:115-123. Giloh, H., and Sedat, J.W., 1982, Fluorescence microscopy: Reduced photobleaching of rhodamine and fluorescein protein conjugates by n-propyl gallate, Science 217: 1252-1255. Glazer, A.N., 1988, Fluorescence-based assay for reactive oxygen species: A protective role for creatinine, FASEB J. 2:2487-2491. Glazer, A.N., 1989, Light guides. Directional energy transfer in a photosynthetic antenna, J. BioI. Chem. 264: 1-4. Gonnilez, J.E., and Maher, M.P., 2002, Cellular fluorescent indicators and voltage/ion probe reader (VIPR(tm»: Tools for ion channel and receptor drug discovery, Receptors Channels 8:283-295. Gonzalez, J.E., and Tsien, R.Y., 1995, Voltage sensing by fluorescence resonance energy transfer in single cells, Biophys. J. 69: 1272-1280. Gonzalez, J.E., and Tsien, R.Y., 1997, Improved indicators of cell membrane potential that use fluorescence resonance energy transfer, Chem. Bioi. 4:269-277. Griffin, B.A., Adams, S.R., and Tsien, R.Y., 1998, Specific covalent labeling of recombinant protein molecules inside live cells, Science 281 :269272.
Gross, D., and Loew, L.M., 1989, Fluorescent indicators of membrane potential: Microspectrofluorometry and imaging, Methods Cell Bioi. 30: 193218. Hadjantonakis, A.K., Dickinson, M.E., Fraser, S.E., and Papaioannou, V.E., 2003, Technicolour transgenics: Imaging tools for functional genomics in the mouse, Nat. Rev. Genet. 4:613-625. Harms, G.S., Cognet, L., Lommerse, P.H., Blab, G.A., and Schmidt, T., 2001, Autofluorescent proteins in single-molecule research: Applications to live cell imaging microscopy, Biophys. J. 80:2396-2408. Harootunian, A.T., Adams, S.R., Wen, W., Meinkoth, J.L., Taylor, S.S., and Tsien, R. Y., 1993, Movement of the free catalytic subunit of cAMP dependent protein kinase into and out of the nucleus can be explained by diffusion, Mol. BioI. Cell 4:993-1002. Hasan, M.T., Friedrich, R.W., Euler, T., Larkum, M.E., Giese, G., Both, M., Duebel, J., Waters, J., Bujard, H., Griesbeck, 0., Tsien, R.Y., Nagai, T., Miyawaki, A., and Denk, W., 2004, Functional fluorescent Ca2+ indicator proteins in transgenic mice under TET control, PLoS Bioi. 2:763-775. Hassler, S., 1995, Green fluorescent protein - the next generation, BiofTechnology 13:103. Haugland, R.P., 1989, Molecular Probes: Handbook of Fluorescent Probes and Research Chemicals, Molecular Probes Inc., Eugene, Oregon, pp. 86-94. Heim, R., 1999, Green fluorescent protein forms for energy transfer, Methods Enzymol. 302:408-423. Hernandez-Cruz, A., Sala, E, and Adams, P.R., 1989, Subcellular dynamics of [Ca2+J. monitored with laser scanned confocal microscopy in a single voltage-clamped vertebrate neuron, Biophys. J. 55:216a. Herman, B., 1989, Resonance energy transfer microscopy, Methods Cell Bioi. 30:219-243. Hicks, B.W., 2002, Green Fluorescent Protein: Applications and Protocols, Humana Press, Totowa, New Jersey. Hirschfeld, T., 1976, Quantum efficiency independence of the time integrated emission from a fluorescent molecule, Appl. Opt. 15:3135-3139. Hoshino, A., Hanaki, K., Suzuki, K., and Yamamoto, K., 2004, Applications of T-Iymphoma labeled with fluorescent quantum dots to cell tracing markers in mouse body, Biochem. Biophys. Res. Commun. 314:46-53.
Jaiswal, J.K., Mattoussi, H., Mauro, J.M., and Simon, S.M., 2003, Long-term multiple color imaging of live cells using quantum dot bioconjugates, Nat. Biotechnol. 21 :47-51. Jericevic, Z., Wiese, B., Bryan, J., and Smith, L.c., 1989, Validation of an imaging system: Steps to evaluate and validate a microscope imaging system for quantitative studies, Methods Cell Bioi. 30:47-83. Ji, J., Rosenzweig, N., Jones, I., and Rosenzweig, Z., 2002, Novel fluorescent oxygen indicator for intracellular oxygen measurements, J. Biomed. Opt. 7:404-409. Kahana, J., and Silver, P., 1996, Use of the A. victoria green fluorescent protein to study protein dynamics in vivo, In: Current Protocols in Molecular Biology (EM. Ausabel, R. Brent, R.E. Kingston, D.O. Moore, J.G. Seidman, J.A. Smith, K. Struhl, eds.), John Wiley and Sons, New York, pp. 9.7.22-9.7.28. Kang, H.C., Fisher, PJ., Prendergast, EG., and Haugland, R.P., 1988, Bodipy: A novel fluorescein and NBD substitute, J. Cell BioI. 107:34a. Kao, J.P.y', Harootunian, A.T., and Tsien, R.Y., 1989, Photochemically generated cytosolic calcium pulses and their detection by fluo-3. J. BioI. Chem. 264:8179-8184. Kaplan, N.O., 1960, The pyridine coenzymes, In: The Enzymes, 2nd ed. (P.O. Boyer, H. Lardy, K. Myrback, eds.), Academic Press, New York, pp. 105-169. Kneen, M., Farinas, J., Li, Y., and Verkman,A.S., 1998, Green fluorescent protein as a noninvasive intracellular pH indicator, Biophys. J. 74: 1591-1599. Koziol, J., 1971, Fluorometric analyses of riboflavin and its coenzymes, Methods Enzymol. 18B:253-285. Kurtz, I., and Balaban, R.S., 1985, Fluorescence emission spectroscopy of 1,4dihydroxyphthalonitrile: A method for determining intracellular pH in cultured cells, Biophys. J. 48:499-508. Kurtz, I., and Emmons, c., 1993, Measurement of intracellular pH with a laser scanning confocal microscope, In: Methods in Cell Biology 38: Cell Biological Applications of Confocal Microscopy (B. Matsumoto, ed.), Academic Press, San Diego, California. Kuwahara, M., and Verkman, A.S., 1988, Direct fluorescence measurement of diffusional water permeability in the vasopressin-sensitive kidney collecting tubule, Biophys. J. 54:587-593. Kuwahara, M., Berry, C.A., and Verkman, A.S., 1988, Rapid development of vasopressin-induced hydroosmosis in kidney collecting tubules measured by a new fluorescence technique, Biophys. 1. 54:595-602. Labas, Y.A., Gurskaya, N.G., Yanushevich, y'G., Fradkov, A.E, Lukyanov, K.A., Lukyanov, S.A., and Matz, M.V., 2002, Diversity and evolution of the green fluorescent protein family, Proc. Natl. Acad. Sci. USA 99:4256-4261. Lakowicz, J.R., 1983, Fluorescence lifetime imaging, Anal. Biochem. 202:316-330. Lakowicz, J.R., 1989, Principles of Fluorescence Spectroscopy, 2nd ed., Kluwer Academic, Plenum Press, New York. Lewis, G.N., Lipkin, D., and Magel, T.T., 1941, Reversible photochemical processes in rigid media. A study of the phosphorescent state, J. Am. Chem. Soc. 63:3005-3018. Li, L., Szmacinski, H., and Lakowicz, J.R., 1997, Synthesis and luminescence spectral characterization of long-lifetime lipid metal-ligand probes, Anal. Biochem. 244:80-85. Lindqvist, L., 1960, A flash photolysis study of fluorescein, Arkiv. Kemi. 16:79-138. Liphardt, B., Liphardt, B., and Liittke, w., 1982, Laserfarbstoffe mit intramolekularer Triplettliischung, Chemische Berichte 115:2997-3010. Liphardt, B., Liphardt, B., and Liittke, W., 1983, Laser dyes III: Concepts to increase the photostability of laser dyes, Opt. Commun. 48: 129133. Lippincott-Schwartz, J., and Patterson, G.H., 2003, Development and use of fluorescent protein markers in living cells, Science 300:87-91. Lippincott-Schwartz, J., Altan-Bonnet, N., and Patterson, G.H., 2003, Photobleaching and photoactivation: following protein dynamics in living cells, Nat. Cell BioI. Suppl. S7-S14. Lippincott-Schwartz, J., Presley, J.E, Zaal, KJ., Hirschberg, K., Miller, C.D., and Ellenberg, J., 1999, Monitoring the dynamics and mobility of membrane proteins tagged with green fluorescent protein, Methods Cell Bioi. 58:261-281.
Fluorophores for Confocal Microscopy: Photophysics and Photochemistry • Chapter 16 Lippincott-Schwartz, J., Snapp, E., and Kenworthy, A., 200 l, Studying protein dynamics in living cells, Nat. Rev. Mol. Cell BioI. 2:444-456. Llopis. J.• McCaffery, J.M., Miyawaki, A., Farquhar, M.G., and Tsien, R.Y., 1998, Measurement of cytosolic, mitochondrial, and Golgi pH in single living cells with green fluorescent proteins, Proc. Natl. Acad. Sci. USA 95:6803-6808. Loew, L.M., 1993, Confocal microscopy of potentiometric fluorescent dyes, In: Methods in Cell Biology 38:Cell Biological Applications (B. Matsumoto, ed.), Academic Press, San Diego, California, pp. 195-209. Manitto, P., Speranza, G., Monti, D., and Gramatica, P., 1987, Singlet oxygen reactions in aqueous solution. Physical and chemical quenching rate constants of crocin and related carotenoids, Tetrahedron Lett. 28: 422l-4224. March, J.c., Rao, G., and Bentley, W.E., 2003, Biotechnological applications of green fluorescent protein, Appl. Microbiol. Biotechnol. 62:303-315. Marriott, G., Clegg, R.M., Arndt-Jovin, DJ., and Jovin, T.M., 1991, Timeresolved imaging studies. Phosphorescence and delayed fluorescence imaging, Biophys. 1. 60:1374-1387. Matheson, I.B.C., and Rodgers, M.AJ., 1982, Crocetin, a water soluble carotenoid monitor for singlet molecular oxygen, Photochern. Photobiol. 36:1-4. Mathies, R.A., and Stryer, L., 1986, Single-molecule fluorescence detection: A feasibility study using phycoerythrin, In: Applications of Fluorescence in the Biomedical Sciences (D.L. Taylor. A.S. Waggoner, R.E Murphy, E Lanni, RR. Birge, eds.), Alan R Liss, New York, pp. l29-140. Matthews, M.M., and Sistrom, W.R., 1959, Function of carotenoid pigments in non photosynthetic bacteria, Nature (Lond.) 184:1892-1893. Matz, M.Y., Lukyanov, K.A., and Lukyanov, S.A., 2002, Family of the green fluorescent protein: Journey to the end of the rainbow, BioEssays 24:953-959. Meyer, T, and Teruel, M.N .• 2003, Fluorescence imaging of signaling networks, Trends Cell BioI. 13:101-106. Minta, A., Kao, J.P.y', and Tsien, RY., 1989, Fluorescent indicators for cytosolie calcium based on rhodamine and fluorescein chromophores, f. Bioi. Chern. 264:8171-8178. Misteli, T, and Spector, D.L., 1997, Applications of the green fluorescent protein in cell biology and biotechnology, Nat. Biotechnol. 15:961-964. Miyawaki, A., 2002, Green fluorescent protein-like proteins in reef Anthozoa animals, Cell Struct. Funct. 27:343-347. Miyawaki, A., 2003, Fluorescence imaging of physiological activity in complex systems using GFP-based probes, Curr. Opin. Neurobiol. l3:591-596. Miyawaki, A., and Tsien, R.Y., 2000, Monitoring protein conformations and interactions by fluorescence resonance energy transfer between mutants of green fluorescent protein, Methods Enzyrnol. 327:472-500. Miyawaki, A., Griesbeck, 0., Heim, R., and Tsien, R.Y., 1999, Dynamic and quantitative Ca2+ measurements using improved chameleons, Proc. Natl. Acad. Sci. USA 96:2135-2140. Miyawaki, A., Llopis, J., Heim, R., et aI., 1997, Fluorescent indicators for Ca'+ based on green fluorescent proteins and calmodulin, Nature 388:882-887. Miyawaki, A., Sawano, A., and Kogure, T, 2003, Lighting up cells: Labelling proteins with fluorophores. Nat. Cell BioI. 5(Suppl):SI-S7. Mujumdar, RB., Ernst, L.A., Mujumdar, S.R., Lewis, c.J., and Waggoner, A.S., 1993, Cyanine dye labeling reagents: Sulfoindocyanine succinimidyl esters, Bioconjug. Chern. 4: I 05-lll . Mukkala, Y.M., 1993, Development of stable, photoluminescent Europium (III) and Terbium (III) chelates suitable as markers in bioaffinity assays: Their synthesis and luminescence properties, PhD thesis, University of Turku, Turku, Finland. Nagai, T, Sawano, A., Park, E.S., and Miyawaki, A., 200 I, Circularly permuted green fluorescent proteins engineered to sense Ca 2+, Proc. Natl. Acad. Sci. USA 98:3197-3202. Nagai, T, Yamada, S., Tominaga, T, Ichikawa, M., and Miyawaki, A., 2004, Expanded dynamic range of fluorescent indicators for Ca(2+) by circularly permuted yellow fluorescent proteins, Proc. Natl. Acad. Sci. USA 101: 10554-10559. Nakanishi, l., Maeda, M., and Umezawa, Y., 2004, A new protein conformation indicator based on biarsenical fluorescein with an extended benzoic acid moiety, Anal. Sci. 20:273-278.
351
Nguyen, D.C., Keller, RA., Jett, J.H., and Martin, J.C., 1987, Detection of single molecules of phycoerythrin in hydrodynamically focused flows by laser-induced fluorescence, Anal. Chern. 59:2158-2161. Niswender, K.D., Blackman, S.M., Rohde, L., Magnuson, M.A., and Piston, D.W., 1995, Quantitative imaging of green fluorescent protein in cultured cells: Comparison of microscopic techniques, use in fusion proteins and detection limits, f. Microsc. 180: I 09-1 J 6. Oi, Y., Glazer, A.N., and Stryer, L., 1982, Fluorescent phycobiliprotein conjugates for analyses of cells and molecules, f. Cell Bioi. 93:981-986. Oswald, B., Duschl, l., Patsenker, L., Wolfbeis, O.S., and Terpetschnig, E., 1999, Synthesis, spectral properties, and detection limits of reactive squaraine dyes, a new class of diode laser compatible fluorescent protein labels, Bioconjug. Chern. 10:925-931. Panchuk-Yoloshina, N., Haugland, RP., Bishop-Stewart, l., BhaJgat, M.K., Millard, P.J., Mao, E, Leung, W.-Y., and Haugland, RP., 1999, Alexa dyes, a series of new fluorescent dyes that yield exceptionally bright, photostable conjugates, f. Histochem. Cytochern. 47: 1179-1188. Patterson, G., Day, RN., and Piston, D., 2001, Fluorescent protein spectra, f. Cell Sci. 114:837-838. Patterson, G.H., Knobel, S.M., Sharif, W.D., Kain, S.R, and Piston, D.W., 1997, Use of the green-fluorescent protein (GFP) and its mutants in quantitative fluorescence microscopy, Biophys. f. 73:2782-2790. Peck, K., Stryer, L., Glazer, A.N., and Mathies, RA., 1989, Single molecule fluorescence detection: Autocorrelation criterion and experimental realization with phycoerythrin, Proc. Natl. Acad. Sci. USA, 86:4087-4091. Phillips, G.N., 1999, Structure and dynamics of green fluorescent protein, Curr. Opin. Struc. BioI. 7:821-827. Piston, D.W., Patterson, G.H., and Knobel, S.M., 1999, Quantitative imaging of the green fluorescent protein (GFP), Methods Cell Bioi. 58:31-47. Prasher, D.C., Eckenrode, Y.K., Ward, W.W., Prendergast, EG., and Cormier, M.J., 1992, Primary structure of the Aequorea victoria green-fluorescent protein, Gene 111:229-233. Rechtenwald, DJ., 1989, US Patent 4,876,190. Reits, E.A., and Neefjes, J.J., 2001, From fixed to FRAP: Measuring protein mobility and activity in living cells, Nat. Cell Bioi. 3:EI45-EI47. Reyftmann, J.P., Kohen, E., Morliere, P., Santus, R., Kohen, C., Mangel, W.E, Dubertret, L., and Hirschberg, J.G., 1986, A microspectrofluorometric study of porphyrin-photosensitized single living cells - I. Membrane alterations, Photochern. Photobiol. 44:461-469. Richieri, G.Y., Ogata, R.T., and Kleinfeld, A.M., 1992, A fiuorescently labeled intestinal fatty acid binding protein, f. BioI. Chern. 267:23495-23501. Richieri, G.Y., Anel, A., and Kleinfeld, A.M., 1993, Interactions of long-chain fatty acids and albumin: Determination of free fatty acid levels using the fluorescent probe ADIFAB, Biochemistry 32:7574-7580. Rizzuto, R, Brini, M., Pizzo, P., Murgia, M., and Pozzan, T, 1995, Chimeric green fluorescent protein (GFP): A new tool for visualizing subcellular organelles in living cells, Curr. Bioi. 5:635-642. Roederer, M., Kantor, A.B., Parks, D.R, and Herzenberg, L.A., 1996, Cy7PE and Cy7 APC: Bright new probes for immunofluorescence, Cytometry 24: 191-197. Rye, H.S., Yue, S., Wemmer, D.E., Quesada, M.A., Haugland, R.P., Mathies, R.A.. and Glazer, RN., 1992, Stable fluorescent complexes of doublestranded DNAs with bis-intercalating asymmetrical cyanine dyes: Properties and applications, Nucleic Acids Res. 20:2803-2812. Sacchetti, A., Ciccocioppo, R., and Alberti, S., 2000, The molecular determinants of the efficiency of green fluorescent protein mutants, Histol. Histopathol.15:101-107. Schafer, EP.. 1983, New developments in laser dyes, Laser Chern. 3:265-278. Schneider, M.B., and Webb, W.W., 1981, Measurement of submicron laser beam radii, Appl. Opt. 20:1382-1388. Seveus, L.. Yaisala, M., Hemmlia, I., Kojola, H., Roomans, G.M., and Soini, E., 1994, Use of fluorescent Europium chelates as labels in microscopy allows glutaraldehyde fixation and permanent mounting and leads to reduced autofluorescence and good long-term stability, Microsc. Res. Techn.28:149-154. Sheetz, M.P., and Koppel, D.E., 1979, Membrane damage caused by irradiation of fluorescent concanavalin A, Proc. Natl. Acad. Sci. USA 76:3314-3317. Smith, S.l., and Augustine, G.l., 1988, Calcium ions, active zones and synaptic transmitter release, Trends Neurosci. II :458-464.
352
Chapter 16 • R. Y. Tsien et al.
Smith, P.J., Blunt, N., Wiltshire, M., Hoy, T., Teesdale-Spittle, P., Craven, M.R, Watson, J.V., Amos, W.B., Errington, R.1., and Patterson, L.H., 2000, Characteristics of a novel deep red/infrared fluorescent cell-permeant DNA probe, DRAQ5, in intact human cells analyzed by flow cytometry, confocal and multiphoton microscopy, Cytometry 40(4):280-291. Snyder, D.S., Garon, C.P., 2003, Decreased uptake of bodipy-Iabelled compounds in the presence of the nuclear stain, DRAQ5, 1. Mierosc. 211(Pt 3):208-211. Soini, E.J., Pelliniemi, L.1., Hemmila, LA., Mukkala, Y.-M., Kankare, J.1., and Froidman, K., 1988, Lanthanide chelates as new fluorochrome labels for cytometry, 1. Histochem. Cytochem. 36: 1449-1451. Southwick, P.L., Ernst, L.A., Tauriello, E.W., Parker, S.R., Mujumdar, RB., Mujumdar, S.R., Clever, H.A., and Waggoner, A.S., 1990, Cyanine dye labeling reagents - Carboxymethylindocyanine succinimidyl esters, Cytometry II :418-430. Steams, T., 1995, Green fluorescent protein - The green revolution, Curro Bioi. 5:262-264. Strickler, S.J., and Berg, R.A., 1962, Relationship between absorption intensity and fluorescence lifetime of molecules, 1. Chem. Phys. 37:814822. Sullivan, K.P., and Kay, S.A., 1999, Green Fluorescent Proteins, Academic Press, San Diego, California. Sun, C., Yang, J., Li, L., Wu, X., Liu, Y., and Liu, S., 2004, Advances in the study of luminescence probes for proteins, 1. Chromat. B: Anal. Technol. Biomed. Life Sci. 803:173-190. Sun, L., Wang, S., Chen, L., and Gong, X., 2003, Promising fluorescent probes from phycobiliproteins, IEEE I Select. Topics Quant. Electron. 9:177-188. Tinoco, I., Mickols, W., Maestre, M.P., and Bustamante, C., 1987, Absorption, scattering, and imaging of biomolecular structures with polarized light, Annu. Rev. Biophys. Biophys. Chem. 16:319-349. Trask, B.J., 1991, DNA sequence localization in metaphase and interphase cells by fluorescence in situ hybridization, Methods Cell Bioi. 35:1-35. Tsien, RY., 1988, Fluorescence measurement and photochemical manipulation of cytosolic free calcium, Trends Neurosci. II :419-424. Tsien, RY., 1989a, Fluorescent probes of cell signaling, Annu. Rev. Neurosci. 12:227-253. Tsien, R.Y., 1989b, Fluorescent indicators of ion concentrations, Methods Cell Bioi. 30: 127-156. Tsien, R.Y., 1998, The green fluorescent protein, Annu. Rev. Biochem. 67:509-544. Tsien, RY., 2003, Imagining imaging's future, Nat. Rev. Mol. Cell Bioi. 4(Suppl):SS 16-SS21. Tsien, RY., and Miyawaki, A., 1998, Seeing the machinery of live cells, Science 280:1954-1955. Tsien, RY., and Poenie, M., 1986, Fluorescence ratio imaging: A new window into intracellular ionic signaling, Trends Biochem. Sci. 11:450-455. Uster, P.S., and Pagano, R.E., 1986, Resonance energy transfer microscopy: Observations of membrane-bound fluorescent probes in model membranes and in living cells, 1. Cell Bioi. 103:122-124. van Roessel, P., and Brand, A.H., 2002, Imaging into the future: visualizing gene expression and protein interactions with fluorescent proteins, Nat. Cell Bioi. 4:EI5-E20. Vereb, G., Jares-Erijman, E., Selvin, P.R., and Jovin, T.M., 1998, Temporally and spectrally resolved imaging microscopy of lanthanide chelates, Biophys. 1. 74:2210-2222. Verkhusha, Y.Y., and Lukyanov, K.A., 2004, The molecular properties and applications of Anthozoa fluorescent proteins and chromoproteins, Nat. Biotechnol. 22:289-296.
Viallet, P.M., and Vo-Dinh, T., 2003, Monitoring intracellular proteins using fluorescence techniques: From protein synthesis and localization to activity, Curro Protein Peptide Sci. 4:375-388. Vigers, G.P.A., Coue, M., and McIntosh, J.R, 1988, Fluorescent microtubules break up under illumination, 1. Cell Bioi. 107:1011-1024. Voura, E.B., Jaiswal, J.K., Mattoussi, H., and Simon, S.M., 2004, Tracking metastatic tumor cell extravasation with quantum dot nanocrystals and fluorescence emission-scanning microscopy, Nat. Med. 10:993-998. Wages, J., Packard, B., Edidin, M., and Brand, L., 1987, Time-resolved fluorescence of intracellular quin-2, Biophys. 1. 51:284a. Waggoner, A., DeBiasio, R, Conrad, P., Bright, G.R, Ernst, L., Ryan, K., Nederlof, M., and Taylor, D., 1989, Multiple spectral parameter imaging, Methods Cell Bioi. 30:449-478. Waggoner, A.S., Ernst, L.A., Chen, C.-H., and Rechtenwald, D.J., 1993, A new fluorescent antibody label for three-color flow cytometry with a single laser, Ann. N.y. Acad. Sci. 677:185-193. Wahlfors, J., Loimas, S., Pasanen, T., and Hakkarainen, T., 2001, Green fluorescent protein (GFP) fusion constructs in gene therapy research, Histochem. Cell BioI. 115:59-65. Ward, W.W., Cody, e.W., and Hart, Re., 1980, Spectrophotometric identity of the energy transfer chromophores in Renilla and Aequorea greenfluorescent proteins, Photochem. Photobiol. 31:611-615. Watt, RM., and Voss, E.W. Jr., 1977, Mechanism of quenching of fluorescein by anti-fluorescein IgG antibodies, Immunochemistry 14:533-541. Weijer, e.J., 2003, Visualizing signals moving in cells, Science 300:96-100. Wessendorf, M.W., and Brelje, T.C., 1992, Which fluorophore is brightest? A comparison of the staining obtained using fluorescein, tetramethylrhodamine, lissamine rhodamine, Texas Red, and cyanine 3.18, Histochemistry 98:81-85. White, J.e., and Stryer, L., 1987, Photostability studies of phycobiliprotein fluorescent labels, Anal. Biochem. 161 :442-452. White, J.G., Amos, W.B., and Fordham, M., 1987, An evaluation of confocal versus conventional imaging of biological structures by fluorescence light microscopy, 1. Cell Bioi. 105:41-48. Wilson, E.O., 1975, Sociobiology, Harvard University Press, Cambridge, Massachusetts, pp. 38-43. Wories, H.J., Koek, J.H., Lodder, G., Lugtenburg, J., Fokkens, R, Driessen, 0., and Mohn, G.R, 1985, A novel water-soluble fluorescent probe: synthesis, luminescence and biological properties of the sodium salt of the 4sulfonato-3,3',5,5' -tetramethyl-2,2'-pyrromethen-l, I' -BF, complex, Reel. Trav. Chim. Pays-Bas 104:288-291. Yip, K.-P., and Kurtz, I., 2002, Confocal fluorescence microscopy measurements of pH and calcium in living cells, Methods Cell Bioi. 70:417427. Yu, H., Ernst, L.A., Wagner, M., and Waggoner, A.S., 1992, Sensitive detection of RNAs in single cells by flow cytometry, Nucleic Acids Res. 20:83-88. Zaccolo, M., and Pozzan, T., 2000, Imaging signal transduction in living cells with GFP-based probes. IUBMB, Life 49:375-379. Zacharias, D.A., 2002, Sticky caveats in an otherwise glowing report: Oligomerizing fluorescent proteins and their use in cell biology, Sci. STKE 2002:E23. Zacharias, D.A., Baird, G.S., and Tsien, RY., 2000, Recent advances in technology for measuring and manipulating cell signals, Curro Opin. Neurobioi. 10:416-421. Zhang, J., Campbell, RE., Ting, A.Y., and Tsien, RY., 2002, Creating new fluorescent probes for cell biology, Nat. Rev. Mol. Cell Bioi. 3:906-918. Zimmer, M., 2002, Green fluorescent protein (GFP): Applications, structure, and related photophysical behavior, Chem. Rev. 102:759-781.
17
Practical Considerations in the Selection and Application of Fluorescent Probes lain D. Johnson
INTRODUCTION Due to its sensitivity, multiplexing capacity, and applicability to live specimens, fluorescence is the dominant contrast mechanism used in three-dimensional (3D) biological microscopy. Use of fluorescence detection generally requires specimens to be labeled with extrinsic probes. This is because most biological molecules and structures of interest are not intrinsically fluorescent in spectral ranges that are useful for detection, and even those that are cannot usually be discriminated from each other on the basis of their intrinsic fluorescence . Extrinsic labeling circumvents these problems at the expense of introducing others. Extrinsic probes must be delivered to the region of interest and remain there for long enough to acquire the experimental data. Once in situ, the probe should ideally be a passive reporter that does not induce significant perturbations of the biological structure or function that we wish to study. Furthermore, the detection process itself may have deleterious side effects in the form of photobleaching and photoxicity, resulting from the interaction of excitation light with the probe and the specimen. Based on the premise that understanding probe behavior is a key component in evaluating the information content of images obtained using fluorescence microscopy, this chapter reviews the practical considerations involved in probe selection and use. In Table 17.1, fourteen key characteristics of probes and specimens are listed in relation to their impact on labeling and detection processes. The ordering of topics in Table 17.1 reflects the sequence of discussion in subsequent sections of this chapter.
SELECTION CRITERIA FOR DYES AND PROBES A fluorescent dye (or, synonymously, a fluorophore) is a fluorescent molecule that does not associate with any particular biological target. A fluorescent probe is a dye that has been modified in some way to detect specific biological targets (Fig. 17.1). Targets include specific groups of cells in a tissue, organelles, proteins, nucleic acids, ions (Ca2+, Mg2+, H+, Na+, etc.) and enzymes. From this perspective, fluorescein is a dye whereas fluorescein-labeled proteins and peptides are probes. Similarly, the green fluorescent protein (GFP) can be considered to be a "dye" and GFP fusion proteins are probes. In some cases, the structural characteristics of the dye itself are sufficient to confer biospecificity. For example, cationic dyes such as MitoTracker Red CMXRos, JC-l, and rhodamine 123 stain mitochondria driven by the internally negative
membrane potential. Dyes and probes have both biochemical and spectroscopic properties. Biochemical properties determine the molecular association, transport, and metabolic fate of the probe. Examples include water solubility, membrane permeability, receptor binding affinity, and enzymatic conversion rates. Spectroscopic properties primarily determine the number and energy distribution of photons available for detection. They include excitation and emission spectra, molar absorptivity (extinction coefficient), fluorescence quantum yield, and photobleaching rate.
Organic Dyes Fluorescein and its derivatives (Fig. 17.1) have been the most widely used class of organic dyes used as fluorescent probes. Their utility is derived in part from the fact that fluorescein is efficiently excited by the 488 nm argon-ion laser line. Coumarins and rhodamines have been the primary blue (-450nm) and orange (-580 nm) emitting dyes used alongside fluorescein (green emission, -520nm) in two-color labeling applications. Each class has some significant disadvantages. Fluoresceins are pH sensitive and highly susceptible to photobleaching. Rhodamines have a tendency to aggregate in aqueous solutions resulting in self-quenching of fluorescence (see below). Coumarins have relatively low excitation efficiencies - Em,x - 20,000M- 1 cm- I compared to Ema, - 100,000 M- 1 cm- I for fluoresceins and rhodamines. Coumarins also have other drawbacks associated with ultraviolet excitation, namely, phototoxicity and a requirement for expensive quartz optical components. These latter problems can be circumvented by use of two-photon excitation. The deficiencies of fluorescein, rhodamine, and coumarin dyes have spurred the development of new dye classes that have been deliberately optimized for biomolecular detection applications. Three of these classes will be discussed here (in alphabetical order) - AlexaFluor dyes, BOPIDY dyes, and cyanine (Cy) dyes (Table 17.2). The AlexaFluor dye series has 19 members with excitation maxima matched to principal laser output lines between 350 and 750nm (Fig. 17.2). The AlexaFluor dyes are more water-soluble than their fluorescein and rhodamine counterparts, resulting directly or indirectly in reduced levels of self-quenching upon coupling to proteins and improved photostability (Panchuk-Voloshina et al., 1999; Berlier et al., 2003). The Cy dye series (Mujumdar et al. , 1993) has fewer excitation wavelength variants but contains similar design elements to the AlexaFluor dyes series - sulfonic acid substituents to increase aqueous solubility and the use of N-hydroxysuccinimidyl (NHS) ester reactive chemistry for cou-
lain D. Johnson. Molecular Probes, Inc. , Eugene, Oregon 97402
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC. New York. 2006.
353
354
Chapter 17 • 1.0. Johnson
TABLE 17.1. Key Characteristics of Probes and Specimens Significance
Property I. Excitation spectrum
Should be wavele ngth-matched with instrument source output for optimum fluorescence excitation efficiency (see Fig. 17.2). Overlap with donor emission spectrum required for FRET. Determining factor in fluorescence output per dye. Ability to resolve probe signal from autofluorescence. Selection of probes for simultaneous imaging of multiple targets. a Overlap with acceptor excitation spectrum required for FRET. Determining factor in fluorescence output per dye. Often environment dependent. Impacts the proportionality of dye fluorescence to concentratio n. Impacts choice of loading method. Impacts choice of loading method. Determines total fluorescence output. May compel use of signal amplification techniques" or GFP overexpression. Impacts selection of excitation/emission wavelength ranges and level of labeling required. Localized accumulations are more readily detectable than the diffuse distributions. Determines stability of labeling. Impacts specimen viability. May require reduced loading concentration and/or selecting a different probe. May necessitate attenuation of excitation power. Determines ability to conduct time-lapse experiments. Impacts specimen viability.
2. Extinction coefficient (e; units M- ' em-I) 3. Emission spectrum
4. 5. 6. 7. 8. 9. 10. II. 12.
Fluorescence quantum yield (QY) Environment sensitivity Probe size/permeability/solubility Specimen type (single cell, cell population, tissue) Target abundance Autofluorescence Probe localization Probe metabolism and retention Probe-mediated cytotoxicity
13. Photobleaching 14. Phototoxicity
"Dyes with narrow emission bandwidths arc preferred for minimizing spillover of the signal into adjacent detection channels (see Fig. 17.2). Probes that exhibit environment-dependent spectral shifts , such as the JC-l and BOPIDY FL ceramide are difficult to employ in these applications for thi s reason. b See Wang and colleagues (1999) for example.
piing to free amine groups on proteins and other biomolecules [Fig. 17.3(A)]. In contrast to the AlexaFluor and Cy dye series, BOPIDY dyes are non-polar and relatively insoluble in water [Fig. 17.3(B)]. Instead of protein labeling, they are primarily utilized in fluorescent lipid analogs (Pagano e l at., 2000; Farber et at. , 2001) and analogs of receptor ligands such as nucleotides, steroids, alkaloids, and peptides (Daly and McGrath, 2003).
o
C12 Fluorescein HO
Environmental factors including pH, solvent polarity, binding to proteins, and dye-dye interactions can exert strong influences on dye fluorescence. Susceptibility to environment varies widely among dye classes. Dyes that are designed primarily for covalent labeling of proteins and nucleic acids such as AlexaFluor dyes and Cy dyes are highly fluorescent in water and retain similar levels of fluorescence after coupling. In histochemical and cytochemical
Fluorescein
0
CI
HO
...
NH
9=0 Calcium Green-1
~
I
c =o
OJ OJ
u
40
en
0
'E
20 0 300
20
00605
350 400 450 500 550
600
650 700
0
::>
u:::
0 750
Wavelength (nm)
FIGURE 17.4. Fluorescence excitation and emission spectra of QD605-streptavidin and QD655-streptavidin (QD = quantum dot). The continuous and almost identical excitation spectra and the sharp, well-differentiated emission spectra contrast markedly with the corresponding characteristics of organic dyes [Fig. 17.2(A,Bl].
plane (Zipfel et aI., 2003a). Probes that were previously of limited utility in confocal microscopy due to their requirements for ultraviolet excitation, such as fura-2 (Ca2+), SBFI (Na+), monochlorobimane (glutathione), and DAPI (nuclear DNA), have acquired a new lease of life (Rose et ai., 1999; Fricker and Meyer, 2001; Rubart et at., 2003). Furthermore, in situ imaging of small endogenous fluorophores such as serotonin and NADH, that are almost inaccessible to one-photon excitation, has now become practicable (Zipfel et ai., 2003b; Rocheleau et at., 2004). With these exceptions, the collection of dyes and probes currently in use for two-photon excitation microscopy is largely the same as that employed in confocal and widefield fluorescence imaging. There are several published collections of two-photon excitation spectra and cross-sections that provide guidance on compatibility of dyes and probes with excitation sources (Xu and Webb, 1996; Xu et al., 1996; Bestvater et ai., 2002; Dickinson et ai., 2003). If two-photon spectral data is not available, the corresponding one-photon spectrum plotted on a doubled wavelength axis can be used as a first approximation. However two-photon excitation spectra differ from their one-photon counterparts to an extent that depends on the molecular orbital symmetry of the fluorophore (greater difference for higher symmetry fluorophores; Zipfel et al.,2003a). Consequently, most two-photon excitation spectra are blue shifted and broader compared to the corresponding one-photon spectra. Simply stated, a dye with a one-photon excitation peak at 500nm will probably have a two-photon excitation maximum at 1000). 10. Whole-cell patch pipette delivery: Similar to microinjection in applicability with the additional capacity for simultaneous imaging and electrophysiological measurements (Eilers and Konnerth, 2000). Inward diffusion of solutes from the patch pipette allows precise control of the intracellular environment but outward diffusion of cytoplasmic contents may be deleterious. A selection of probes with proven utility for general characterization of living specimens using simple and rapid labeling protocols is listed in Table 17.3.
Tissues Scientific imperatives allied to technical developments such as two-photon excitation and confocal endoscopy (Helmchen, 2002) mean that fluorescence microscopy is increasingly being applied to tissues and entire organisms. Most of the techniques described above can be adapted for application to tissue specimens, while GFP expression can be confined to specific cell types by coupling to tissue-specific promoters (Hara et at., 2003). Water-soluble dyes
such as FM 1-43 and lucifer yellow CH generally show deeper penetration into tissues than non-polar molecules such as AM esters, which tend to accumulate in the superficial cell layers (Takahashi et at., 2002). A modification of the AM ester technique, referred to as multi-cell bolus loading (MCBL), utilizes localized ejection of small volumes (-O.4nL) of dye loading solution from a micropipette to label populations of neurons in brain tissue (Stosiek et at., 2003). For imaging tissues in situ, probe delivery is typically accomplished via the internal pathways of the digestive (Farber et aI., 2001), respiratory (Lombry et al., 2002), circulatory (Ballou et at., 2004), or nervous (Grutzendler et aI., 2003) systems.
Target Abundance and Autofluorescence Considerations We have considered spectroscopic properties relating to the fluorescence output of labels such as excitation wavelength, exctinction coefficent, photobleaching, and fluorescence quantum yield. However, factors such as the abundance and spatial distribution of the target and the levels of background autofluorescence often have more impact on the contrast and resolution of the final image. The abundance and degree of localization of the target within the specimen are critical determinants in probe selection. For example, it is much easier to image DNA localized in the nucleus than receptors distributed on the plasma membrane surface. The DNA content of a typical mammalian cell is about 7 pg, corresponding to about 6 x 109 base pairs. This amount of DNA can accommodate the binding of up to 1.2 X 109 intercalating dyes (I dye: 5 base pairs). Consequently, nuclear stains such as propidium iodide (PI) are easily detectable despite the fact that the fluorescence intensity per dye is relatively modest (lOmax - 5000M- 1cm- 1 and QY - 0.1). In contrast, detection of cell-surface EGF receptors, present at -10,000 copies/cell, by confocal microscopy may require the use of fluorophores such as R-phycoerythrin (lOmax 1,960,000M- 1 cm- I and QY - 0.82) to generate sufficient signal (Good et at., 1992). Similar considerations apply when simultaneously imaging two targets using a single excitation wavelength. Typically it is quite difficult to find two dyes that can be excited with equal efficiency at the selected wavelength and also have
TABLE 17.3. Dyes for Rapid Assessment of Living Cells by Fluorescence Microscopy Labeling Target Hoechst 33342 Ca1ceinAM PM 4-64 Propidium iodide LysoTracker Red DND-99 MitoTracker Red CMXRos AlexaFluor 594 hydrazide
Incubation Concentration"
ExlEm b
Laser Lines"
GFP Compatibled
2~M*
350/460 494/515
5~M*
506175('/
351-364nm or 405nm 488nm 488 nm, 514 nm, or 568 nm
Y N Y
5~M*
I~M*
Nucleus Cytosol' Plasma membrane, endosomes Nucleus (dead cellsY Lysosomes
50nM
535/620 575/590
488nm or 514nm 568nm
Y Y
Mitochondria
50nM
578/600
568nm
Y
Water (fluid phase tracer)
IOmMh
588/615
568nm
Y
"These dyes can generally be used to stain live eukaryotic cells by incubation for IS to 30min at the indicated concentration. Dyes marked with an asterisk (*) can be imaged directly in the dye incubation medium, without a subsequent wash step. b Fluorescence excitation/emission maxima in nanometers. 'Laser lines suitable for excitation. "Y, fluorescence emission is spectrally well resolved from that of GFP. 'Fluorescence is dependent on cytosolic esterase activity and is therefore positively correlated with cell viability. fThe fluorescence emission of PM 4-64 has a wide spectral bandwidth and can be detected anywhere from 625 to 800nm. g Propidium iodide is imperrneant to live cells; fluorescence is therefore inversely correlated. with cell viability. . .. .. h Used to fill cells (typically neurons) via microinjection of a IOmM aqueous solulion. LUCIfer yellow CH eXCIted at 405 nm IS a WIdely used alternatIve but IS not spectrally well resolved from GFP.
Practical Considerations in the Selection and Application of Fluorescent Probes • Chapter 17
emission spectra that are sufficiently well separated to be discriminated without resorting to spectral unmixing (Fig. 17.2) or the use of quantum dot labels. In this situation, the less efficiently excited dye should be used to detect the more abundant target, thereby equalizing the two emission signals. Bulk loading procedures, such as AM ester loading, generate intracellular concentrations of up to 100 11M, corresponding to about I x 10 8 molecules in the cytoplasm of a typical mammalian culture cell with a total volume of 4000 11m3 (of which about SO% is occupied by organelles). Perhaps of more immediate concern to the experimentalist is the cell-to-cell concentration uniformity. A confocal imaging study of neuroblastoma cells loaded with fura2, AM by Fink and colleagues (1998) found a mean intracellular dye concentration of 38 11M in 123 cells, with individual cell values ranging from 10 to 90 11M. Self-referencing, ratiometric measurements are often used to correct for variability of dye concentration when making cell-to-cell comparisons of fluorescence intensity. Similar intracellular concentrations are achieved in typical microinjection protocols in which lOmM dye solution equi valent to about 1% of cell volume is injected. Much higher concentrations can be attained in situations where probes are sequestered in subcellular compartments. For example, potentialdriven uptake of cationic dyes in mitochondria can result in concentrations that are up to 1000-fold higher than in the cytosol (Nicholls and Ward, 2000). These probes should therefore be applied at very low external concentrations ( I 00 11m) where non-descanned detection increases signal substantially, but on thinner specimens, the actual damage/excitation may be greater for MP than for single-photon confocal imaging (Tauer, 2002; see also Chapter 38, this volume). Moreover, the cost differential is such that one could have two to three graduate students working away on disk scanners or simpler beam-scanning confocal units for every one on a MP unit. At any rate, most of the topics covered in this chapter are relevant for both single-photon confocal and MP excitation.
Michael E. Dailey. University of Iowa, Iowa City, Iowa 52242 Erik Manders. Swammerdam Institute for Life Sciences, University of Amsterdam, Amsterdam, The Netherl ands David R. 5011 • University of Iowa, Iowa City, Iowa 52242 Mark Terasa ki • University of Connecticut Health Center, Farmington, Connecti cut 06032 Handbook of Biolordcal Confocal Micmsconv. Third Edition. edited
bv James B. Pawley. Sorin!!er Science+Busine ss Media. LLC. New York. 2006.
381
382
Chapter 19 • M.E. Dailey et al.
OVERVIEW OF LIVING-CELL CONFOCAL IMAGING TECHNIQUES Although live-cell imaging often involves time-lapse microscopy to monitor cell movements, modem approaches are extending these observations well beyond simply making movies of cell structure. Increasingly, time-lapse imaging is being integrated with specialized techniques for monitoring, measuring, and perturbing dynamic activities of cells and subcellular structures. Below we summarize some major techniques available for studying the dynamic organization of molecules and cells in live biological specimen. These techniques are summarized in Table 19.1.
Time-Lapse Fluorescence Imaging Time-lapse fluorescence imaging involves repeated imaging of a labeled specimen at defined time points, thereby permitting studies on the dynamic distribution of fluorescently labeled components in living systems. Imaging can be performed in one, two, or three spatial dimensions: one-dimensional (lD) imaging involves rapid and repeated imaging of single scan lines; two-dimensional (2D) imaging involves repeated imaging of single focal planes; and three-dimensional (3D) imaging involves repeated imaging of multiple focal planes in a thick specimen. The time intervals for sequential image collection can range from sub-second to days or even months (e.g., Gan et aI., 2003). Many small molecule, vital fluorescent probes that give highly specific cellular or subcellular patterns of labeling are now available (see below and Chapters 16 and 17, this volume). In addition, GFP or GFP-related proteins are now routinely fused to other proteins of interest, and the inherent brightness and photostability of many of these fluorescent proteins make them well suited for the repeated imaging needed for time-lapse studies. Together, these fluorescent probes are affording a seemingly limitless array of possibilities for imaging molecular components in live cells.
Multi-Channel Time-Lapse Fluorescence Imaging The plethora of excellent vital fluorescent labels with varying spectral characteristics (including spectral variants of GFP) allows multi-label experiments to visualize the relative distribution of several different cell or tissue components simultaneously. Advances in imaging technology have facilitated automated collection of more than one fluorescent channel (either sequentially or simultaneously) with improved ability to maximize signal collection and to separate partially overlapping signals. In addition to studies using multiple fluorescent tags, multichannel data collection permits ratio metric imaging of single probes whose spectral properties (absorption or emission) change depending on ionic conditions, such as the Ca2+ sensitive physiological indicator, indo-l (see Chapter 42, this volume).
Spectral Imaging and linear Unmixing Increasingly, experiments are incorporating multiple fluorescent probes within single cells or tissues to define the differential distribution of more than one labeled structure or molecular species. Such multi-color or multi-spectral imaging experiments require adequate separation of the fluorescent emissions, and this is especially problematic when the spectra are substantially overlapping. Spectral imaging utilizes hardware to separate the emitted light into its spectral components. Linear unmixing is a computational
process related to deconvolution that uses the spectra of each dye as though it were a point-spread function at a fixed location to "unmix " the component signals (Tsurui et aI. , 2000; Lansford et aI., 200 I; Hiraoka et aI., 2002). Although together these analytical tools can be used to discriminate distinct fluorophores with highly overlapping spectra (Zimmermann et aI., 2003), they do so at the cost of requiring that significantly more photons be detected from each pixel.
Fluorescence Recovery After Photobleaching Fluorescence recovery after photobleaching (FRAP), also known as fluorescence photobleaching recovery (FPR), is a technique for defining the diffusion propelties of a population of fluorescently labeled molecules (Axelrod et aI. , 1976; Koppel et aI., 1976; for review, see Lippincott-Schwarz et aI., 2003). Typically, a spot or line of intense illumination is used to bleach a portion of a fluorescent cell, and the recovery of fluorescent signal back into the bleached area from adjacent areas is monitored over time (usually seconds to minutes). Although this technique can yield quantitative information on the diffusion coefficient, mobile fraction, and binding/dissociation of a protein, care has to be taken not to use so much power in the bleach beam that the cellular stnlcture is disrupted (Bloom and Webb, 1984; Flock et al., 1998; see also Figures 49 .10 through 49.14, this volume). Quantitative assessments of FRAP data, which can be confounded by uncertainties in the experimental and biological parameters in living cells, may benefit from computer simulations (Weiss, 2004).
Fluorescence loss in Photo bleaching This technique utilizes repeated photobleaching in an attempt to bleach all fluorophores within a given cellular compartment (Lippincott-Schwarz et aI., 2001). Thus, fluorescence loss in photobleaching (FUP) can be used to assess the continuity of membrane bounded compartments (e.g., endoplasmic reticulum [ER] or Golgi apparatus) and to define the diffusional properties of components within, or on the surface of, these compartments.
Fluorescence Resonance Energy Transfer Fluorescence resonance energy transfer (FRET) is a technique for defining interactions between two molecular species tagged with different fluorophores (Stryer, 1978; Sekar and Periasamy, 2003). It takes advantage of the fact that the emission energy of a fluorescent "donor" can be absorbed by (i.e., transferred to) an "acceptor" fluorophore when these fluorophores are in nanometer proximity and have overlapping spectra (see Chapters 16, 27, and 45 , this volume).
Fluorescence lifetime Imaging Thi s technique measures the lifetime of the excited state of a fluorophore (Lakowicz et aI. , 1992 and Chapter 27, this volume). Each fluorescent dye has a characteristic "lifetime" in the excited state (usually 1-20ns), and detection of this lifetime can be used to distinguish different dyes in samples labeled with multiple dyes. Fluorescence lifetime imaging (FUM) can be utilized in conjunction with FRET analysis because the lifetime of the donor fluorophore is shortened by FRET. In fact, FUM can improve the measurement during FRET analysis because the fluorescence lifetime is independent of the fluorophore concentration and excitation energy (Bastiaens and Squire, 1999; Elangovan et aI., 2002;
Confocal Microscopy of Living Cells • Chapter 19
383
Chen et aI., 2003; Chapter 27, this volume). However, the lifetime can be modulated by environmental considerations (e.g., pH, ion concentration), and this change can be used to measure changes in the concentration of certain ions (Lin et aI., 2003).
and photoactivation are complementary to FRAP and can be used in conjunction with time-lapse imaging to mark and follow a population of molecules in order to study their kinetic properties within living cells.
Fluorescence Correlation Spectroscopy
Optical Tweezers/Laser Trapping
Fluorescence correlation spectroscopy (FCS) measures spontaneous fluorescence intensity fluctuations in a stationary microscopic detection volume (about I fL) (Magde et aI., 1974). Such intensity fluctuations represent changes in the number or quantum yield of fluorescent molecules in the detection volume. By analyzing these fluctuations statistically, FCS can provide information on equilibrium concentrations, reaction kinetics, and diffusion rates of fluorescently tagged molecules (Elson, 2001). An advantage of this approach is the ability to measure the mobility of molecules down to the single molecule level and to do so using a light dose orders of magnitude lower than used for FRAP.
Optical tweezers, or single beam laser traps, use the "radiation pressure" of a stream of photons emitted from an infrared laser to "trap" small objects (often a protein-coated bead) and to move them around (Sheetz, 1998; Kuo, 2001; Chapters 5 and 9, this volume). This technique has been especially useful for quantifying forces generated by motor protein movement (Ashkin et at., 1990; Block et at., 1990; Kuo and Sheetz, 1993) or the strength of adhesions mediated by cell adhesion molecules (e.g., Schmidt et aI., 1993; Baumgartner et aI., 2003). Although "laser tweezers" often are used in widefield imaging systems, they also have been incorporated into confocal (Visscher and Brakenhoff, 1991) and MP (Goksor et al., 2004) imaging systems.
Fluorescence Speckle Microscopy The dynamic growth and movement of fluorescently labeled structures can be difficult to analyze when these structures are densely packed and overlapping within living cells. Fluorescent speckle microscopy (FSM) is a technique compatible with widefield or confocal microscopy (Adams et aI., 2003) that uses a very low concentration of fluorescently labeled subunits to reduce out-offocus fluorescence and improve visibility of labeled structures and their dynamics in thick regions of living cells (Waterman-Storer et aI., 1998). This is accomplished by labeling only a fraction of the entire structure of interest. In that sense, it is akin to performing FCS over an entire field of view, albeit with more focus on spatial patterns than on quantitative temporal analysis. FSM has been especially useful for defining the movement and polymerization/depolymerization of polymeric cytoskeletal elements, such as actin and microtubules, in motile cells (Salmon et aI., 2002).
Photo-Uncaging/Photoactivation Photo-uncaging is a light-induced process of releasing a "caged" molecule from a caging group to produce an active molecule (Politz, 1999; Dorman and Prestwich, 2000). A variety of caged molecules have been synthesized and used experimentally, but in some examples cages have been used to mask a fluorophore, inducing a non-fluorescent state. Excitation light of -350nm is used to break photolabile bonds between the caging group and fluorophore, thereby uncaging the fluorophore and yielding a fluorescent molecule. A related technique utilizes genetically encoded, photoactivatable fluorescent proteins, of which there are currently about a dozen (for review, see Patterson and Lippincott-Schwartz, 2004). Two examples include a photoactivatable (PA) form of GFP, called PA-GFP (ex/em; 504/517), which shows a 100-fold increasc in fluorescence following irradiation at 413 nm (Patterson and Lippincott-Schwartz, 2002), and Kaede (ex/em: 572/582), which shows a 2000-fold increase following irradiation at 405 nm (Ando et at., 2002; see also Figure 8.37, this volume). An extension of the photoactivation approach, termed reversible protein highlighting, has been developed (Ando et aI., 2004). This involves reversible, light-induced conversion of a coral protein, Dronpa, between fluorescent and non-fluorescent states. One study used this approach to monitor fast protein dynamics in and out of cell nuclei (Ando et al., 2004). Thus, photo-uncaging
Physiological Fluorescence Imaging The availability of fluorescent physiological indicators extends live-cell confocal and MP imaging studies beyond structural aspects to study cell and tissue physiology (Niggli and Egger, 2004; Rubart, 2004; Wang et aI., 2004). Calcium indicators have been the most commonly used physiological probes because calcium is a central signal transduction molecule and in many cell preparations the calcium-sensitive probes give robust signals. These signals often are temporally resolvable in full field scans as calcium transients that persist for several seconds. Fast scanning systcms, or line-scanning mode in laser scanning systems, have been used to resolve more rapid calcium events (e.g., Fan et at., 1999; Wang et at., 2004). Although non-ratiometric, visible wavelength calcium indicators (e.g., fluo-3, calcium green) have been more widely used in confocal applications, some studies have employed ultraviolet (UV) excited ratiometric calcium indicators, such as indo-l (e.g., Pasti et aI., 2001). In addition to calcium indicators, other fluorescent physiological probes are useful for reporting various ions including sodium, magnesium, potassium, and chloride, pH, heavy metals such as zinc, and membrane potential, to name a few (see Chapter 42, this volume). Although many of these probes are small molecules, genetic (GFP-based) probes have been developed (see Miyawaki, 2003) and are being incorporated into transgenic animals (e.g., Hasan et al., 2004). In combination with state-of-the-art confocal and MP imaging systems, these probes will increasingly permit detailed spatio-temporal analyses of physiological processes in intact tissues and organisms (Ashworth, 2004).
Combining Fluorescence and Other Imaging Modalities Although advancements in fluorescence imaging technology coupled with the availability of a multitude of vital fluorescent probes have combined to make fluorescence the method of choice for most high resolution studies of living cells, it is sometimes advantageous to combine fluorescence imaging with other imaging modalities. For example, differential interference contrast (DIC) microscopy can be used in conjunction with laser-scanning confocal microscopy to simultaneously monitor the whole cell in DIC mode while imaging the phagocytic uptake of fluorescent microspheres (Hook and Odeyvale, 1989) or the distribution of fluores-
8 Fluorescence localization after photobleaching (FLAP). 9 Fluorescence resonance energy transfer (FRET).
7 Fluorescence loss in photobleaching (FLIP).
after photobleaching (FRAP).
6 Fluorescence recovery
4 Three-dimensional multi-channel (5D) timelapse fluorescence imaging. 5 Spectral imaging and linear unmixing.
I Time-lapse fluorescence imaging. 2 Multi-channel or ratiometric time-lapse fluorescence imaging. 3 Three-dimensional time-lapse (4D) imaging.
Technique
Measures recovery of fluorescence after bleaching of a portion of the specimen. Recovery may be due to protein diffusion, binding/dissociation or transport processes. Repeated photobleaching used to determine continuity of cell compartments and mobility of fluorescent proteins within these compartments. Method for localized photo-labeling and subsequent tracking of specific molecules bearing two different fluorophores within living cells. Non-radiative energy transfer from a donor to an acceptor fluorophore with overlapping emission and excitation spectra. Useful for measuring interactions between two fluorescentiy tagged proteins.
• Used FLAP to show that actin is rapidly delivered to the leading edge of protruding cells (Zicha et aI., 2003) • FRET used to study activation of small G proteins during phagocytosis (Hoppe and Swanson, 2004) • FRET analysis shows that GTP-Rac coupling to effectors is locally enhanced in lamellipodia (Del Pozo et aI., 2002)
• Dunn et al., 2002
• Sekar and Periasamy, 2003 • Wouters et aI., 200 I
• Used FRAP to study integrin turnover at focal adhesions (Ballestrem et al., 2001)
• Resolve multiple fluorescent proteins in vertebrate cells by multiphoton imaging spectroscopy (Lansford et aI., 2001)
• Dynamics and retention of correctly folded and misfolded proteins were compared in native ER membranes (Nehls et aI., 2000)
Berg, 2004 Zimmermann et aI., 2003 Seyfried et aI., 2003 Hiraoka et aI., 2002 Dickenson et aI., 200 I Lippincott-Schwarz et aI., 2003 Meyvis et aI., 1999
• Imaged neuronal dendritic spines in brain slice cultures (Marrs et aI., 2001) • Imaged mitosis and migration of developing cortical neurons (Noctor et al., 2004) • Imaged T-cell-dendritic cell interactions in lymph nodes (Stoll et aI., 2002) • Microglial phagocytosis in brain slices (Petersen & Dailey, 2004) • Unmix spectrally similar fluorophores in plant cells (Berg, 2004)
• Imaged dynamic changes in fluorescently labeled Golgi membranes (Cooper et al., 1990) • Monitored sorting of CFP- and YFP-tagged proteins through the Golgi Apparatus (Keller et aI., 2001)
Selected Examples/References
• Lippincott-Schwartz et aI., 2001
• • • • • • •
Method for discriminating distinct fluorophores with strongly overlapping emission spectra.
Gerlich & Ellenberg, 2003 Bement et aI., 2003 Hammond & Glick, 2000 Thomas & White, 1998 Andrews et aI., 2002 Gerlich et aI., 200 I
• Stricker, 2004 • Ellenberg et aI., 1999 • • • • • •
Repeated collection of z-stacks in two or more fluorescent channels over time.
Review Article( s) • Cooper et aI., 1999
Repeated collection of z-series stacks of images over time.
Repeated imaging of a field of view (single optical section) in live specimen over time. Simultaneous or sequential imaging in two or more fluorescent channels over time.
Description
TABLE 19.1. Overview of Live Cell Fluorescence Confocal Imaging Techniques
Measures spontaneous fluorescence intensity fluctuations in a microscopic detection volume. Provides information on equilibrium concentrations, reaction kinetics, and diffusion rates of molecules. Uses very low concentration of fluorescent subunits to reduce out-of-focus fluorescence and improve visibility of fluorescentiy labeled structures and their dynamics in thick regions of living cells. Photo-induced activation of an inert molecule to an active state (e.g., release of a caging group from a "caged" compound), or activation of a photoactivatable fluorescent protein (e.g., PA-GFP, Kaede).
Uses the "radiation pressure" of a stream of photons emitted from an infrared laser to "trap" small objects and molecules. Rapid, repeated collection of single scan lines or 2D images of specimen labeled with physiological indicators.
Repeated simultaneous collection of one or more fluorescent channels and a transmitted light channel (e.g., DIC).
11 Fluorescence correlation spectroscopy (FCS).
14 Optical tweezers/laser trapping. 15 Fast physiological imaging • Full field • Line-scanning
16 Combined fluorescence and transmitted light imaging.
13 Photo-uncaging/ Photoactivation.
12 Fluorescence speckle microscopy.
Method to investigate molecular interactions, metabolic reactions, and energy transfer in cells and subcellular structures.
10 Fluorescence lifetime imaging (FUM). Peter and Ameer-Beg, 2004 Periasamy et aI., 2002 Bastiaens and Squire, 1999 Pepperkok et al., 1999
• Photo-release of caged Ca'+ in brain astrocytes regulates vascular constriction (Mulligan and MacVicar, 2004) • Used a reversible photoactivatable fluorescent protein to study nuclear import and export of ERKI and importin (Ando et aI., 2004) • Studied strength of cadherin adhesions in endothelial cells (Baumgartner et aI., 2003) • Imaged Ca'+ sparks in muscle fibers (Hollingworth et at., 2000; Brum et aI., 2000) • Ca'+ imaging in neuronal dendritic spines (Pologruto et aI., 2004)
• Patterson & LippincottSchwartz, 2004 • Park et al., 2002 • Dorman and Prestwich, 2000 • Politz, 1999 • Kuo,2001 • Schwarzbauer, 1997 • Rubart, 2004 • Wang et aI., 2004 • Ashworth, 2004 • Niggli & Egger, 2004 • Miyawaki, 2003 • Cogswell & Sheppard, 1991, 1992
• Imaged chromatin dynamics during the formation of the interphase nucleus (Manders et aI., 2003) • Imaged E-cadherin-GFP accumulation at cell adhesions in epithelial cells (Adams et aI., 1998).
• Studied coupling of microtubule and actin movements in migrating cells (Salmon et aI., 2002)
• Used FUM to study interaction between CD44 and ezrin (Legg et al., 2002) • Quantified dimerization of transcription factor CAATT/enhancer binding protein alpha in living pituitary cells (Elangovan et al., 2002) • Compared mobility and molecular interactions between CaM and CaMKII in solution and in living cells (Kim et aI., 2004)
• Adams et aI., 2003 • Waterman-Storer et aI., 1998
• Bacia & Schwille, 2003 • Hess et al., 2002 • Elson, 2001
• • • •
386
Chapter 19 • M.E. Dailey et al.
cently tagged proteins and molecules (e.g., Adams et ai., 1998) within these cells. Although it is difficult to perform ole and epi-fluorescence imaging both simultaneously and optimally in widefield microscopy, it is somewhat easier to ensure that the fluorescence signal is not subjected to the light loss that occurs in the analyzer used as part of the DIe system if one uses a singlebeam confocal. Thus, the Ole image can be collected from a fluorescently labeled specimen using transmitted light that would otherwise be wasted. Recently, differential phase contrast (DPC) has been implemented in a scanning laser microscope system (Amos et at., 2003), and this may offer additional capabilities where DIe optics are unsuitable. Notably for live-cell imaging, ope reportedly needs 20 times less laser power at the specimen than Ole.
GENERAL CONSIDERATIONS FOR CONFOCAL MICROSCOPY OF LIVING CELLS What factors must be considered when performing a live-cell confocal imaging experiment or observation? The major factors are to (1) label the preparation in order to clearly visualize the biological component of interest, (2) maintain the preparation in a condition that will support normal cell or tissue health, and (3) image the specimen with sufficient spatial and temporal resolution in a way that does not perturb or compromise it. Table 19.2 outlines several of the most important experimental considerations for Jjvecell imaging, including the most common problems and some potential solutions.
TABLE 19.2. Experimental Considerations for Live Cell Imaging Consideration
Problem
I Temperature
Many biological phe nomena are temperature sensitive.
2 Oxygenati on
Most live biological specimens require 0 , (and removal of CO, ) to remain healthy. Oxygen may become depleted in closed chambers.
3 pH
Metabolism of live biological tissues can induce severe pH changes in chamber media over time.
4 Humidity
Stage heating (especially with forced air) may cause evaporation from an open chamber, leading to dramatic changes in sa linity and pH. Weakly fluorescent probes or low concentration of probes can yield weak signals that produce images with low signal-to-noise ratio.
5 Fluorescence signal strength
6 Channel bleedthrough or cross-talk
In biological specimens labeled with multiple fluorescent probes, signals from one channel may be detected in other channels.
7 Photobleachi ng
Fluorescent probes bleach with repeated illumination. Some fluorescent probes bleach quickly.
8 Spatial resolution
Some observations require very high spatial resolution in x-y or Z.
9 Temporal resolution
Some biological phenomena are rapid relative to the rate of im age collection (especially problematic with laser scanning confocal systems).
10 Focus drift
Live biological specimens on heated microscope stages, or features within li ve specimens (e.g., mitotic cells), can move relative to a fixed focal plane.
Potential solution(s) • Use stage heaters; inlinc perfusion heaters; objective lens heaters; environmental boxes. • Take precautions against stage drift: - increase thermal mass, - use open-loop controls • Use a perfusion chamber. • Exchange used chamber media with oxygenated media intermittently or continuously. • Increase volume of chamber to promote health. • Monitor chamber pH. • Use HEPES (l0-25mM)-buffered media. • Exchange chamber media intermittently or continuously (perfusion). • Use media without phenol red pH indicator. • Use closed chamber configuration (perfusion chamber). • Use humidified environmental box. • Use auto-fill system for open chambers. • Increase pixel dwell time. • Open confocal pinhole aperture (e.g., to >2 Airy disks). • Maximize throughput of emission pathway (e.g., in spectral imaging systems with variable spectral filters). • Use line or frame averag ing to improve signal-to-noise ratio. • Adjust illumination (filling) of back aperture of objective lens. • [mage separate fluorescence channels sequentially (either line-byline or frame-by- frame in scanning systems). • Use spectral imaging and linear unmixing algorithms. • Use modern , hard-coated interference filters and dichroics. • Reduce incident illumination. Then reduce it againl • Use fade resistant dyes. • Open confocal pinhole aperture. • Maximize throughput of emission pathway (e.g., in spectral imaging systems with variable spectral filters). • Reduce pixel dwell time (in scanning systems). • Reduce frequency of image capture. • Blank laser beam during fl yback (in scanning systems). • Only scan specimen when actually collecting data. • To improve SIN, always deconvolve 3D data before view ing. • Use high NA objectives. Reduce size of confocal pinhole aperture (to -I Airy disk). • Increase spatial sampling frequen cy (guided by Nyquist theorem). • Increase electronic zoom (but avoid empty magnification). • Decrease step size in z-stacks. • Use water-immersion objective lenses to reduce spherical aberration. • Deconvolve the images. • Reduce field of view (e.g., collcct fewer horizontal lines). • Reduce pixel dwell time (e.g., increase scan speed). • Reduce spatial sampling frequ ency (e.g., reduce pixel array from 1024 to 512). • Collect a z-stack of images, and reconstruct these images following the observation. • Manual focus adjustments may be required periodically. • Auto-focus methodology may be employed in some cases.
Confocal Microscopy of living Cells • Chapter 19
Maintenance of Living Cells and Tissue Preparations
In Vitro Preparations Specimen maintenance is a very important part of any live imaging study and usually requires both mechanical ingenuity and insight into the biology of the cell or tissue under study. The specimen chamber must keep the cells or tissues healthy and functioning normally for the duration of the experiment while allowing access to the microscope objective. This can be particularly difficult when high-numerical-aperture (NA) oil- or water-immersion lenses are used. In many cases, there must also be a controlled and efficient way to introduce a reagent to perturb a particular cellular process. Other important factors are simplicity, reliability, and low cost. It is advisable to monitor the conditions within the imaging chamber carefully. It may be helpful to use microprobes that can detect pH, O 2 , and CO 2 (e.g., Lazar Research Laboratories, Inc., Los Angeles, CA). The early closed perfusion chambers designed by Dvorak and Stotler (1971) and later by Vesely and colleagues (1982) were inexpensive and permitted high-resolution transmitted light observation. They relied on an external heater that warmed the entire stage area for temperature control. Setups for different cells vary widely. Mammalian cells probably pose the greatest problems. McKenna and Wang's article (1986) is a general introduction to the problems associated with keeping such cells alive and functioning on the microscope stage. This article discusses culture chamber design as well as strategies for controlling pH, osmolarity, and temperature. The authors describe their own chamber, in which temperature is controlled by heating the air in a box surrounding the stage area, and mention earlier designs such as the resistively heated Lieden Culture System first described by Ince and colleagues (1983) and later improved by Forsythe (1991). Strange and Spring (1986) describe their setup for imaging renal tubule cells where temperature, pH, and CO 2 are controlled. They provide a detailed account of the problems of establishing laminar flow perfusion systems, temperature regulation, and maintenance of pH by CO 2 buffering. Somewhat later, Delbridge and colleagues (1990) describe a sophisticated, open-chamber superfusion system permitting programmed changes of media, precision control of media surface height, and temperature regulation between 4°C and 70°C using a Peltier device to control the perfusate temperature. Myrdal and Foster (1994) used a temperaturestabilized liquid passing through a small coil suspended in media filling a plastic Nunc chamber to provide temperature control for confocal observations of the penetration of fluorescent antibodies into solid tumor spheroids. An automatic system maintained fluid level and bathed the area in CO 2 but special precautions were required to prevent drift of the confocal focus plane during long time-lapse sequences. Methods for observing microglial cell movements in mammalian brain slices are described in detail in a later section of this chapter. Chambers have even been built for the microscopic observation of cells as they are being either frozen or thawed in the presence of media that could be changed during the process (e.g., Walcerz and Diller, 1991). In this case, computer-controlled pumps deliver temperature-controlled nitrogen gas at between -120°C and 100°C to special ports connected to a temperature cell (-55°C to 60°C) that forms the upper boundary of the perfusion chamber. Other ports carry either the perfusate or a separate nucleating agent to the cell chamber itself. More recently, a specialized in vitro cell culture system has been developed to maintain mammalian neuronal cells for over a year (Potter and DeMarse, 2001)!
387
There are several companies that provide ready-made microscope stage chambers, temperature-control units, automated perfusion systems, and a variety of related accessories. These are summarized in Table 19.3.
In Vivo Preparations The ultimate goal of many research programs is to understand the normal (or abnormal) structure and function of molecules, cells, and tissues in vivo, that is, in the living organism functioning within its native environment (Frosting, 2002; Megason and Fraser, 2003). There has been some remarkable progress recently on extending high resolution confocal and MP imaging in this direction, especially in preparations that are essentially translucent. Several model organisms, including zebrafish (Cooper et ai., 1999), frog (Fraser and O'Rourke, 1990; Robb and Wylie, 1999), fruit fly (Paddock, 2002), leech (Baker et ai., 2003), and worm (Crittenden and Kimble, 1999), have emerged as excellent preparations for cellular and molecular imaging studies spanning a variety of biological questions. As an example, studies in the zebrafish have been carried out on the structural development of vasculature (Lawson and Weinstein, 2002; Isogai et ai., 2003), cell division (Gong et aI., 2004; Das et ai., 2003), neuronal migration (Koster and Fraser, 2001), axonal pathfinding (Dynes and Ngai, 1998), synapse formation (Jontes et at., 2000; Niell et al., 2004), and synaptic plasticity (Gleason et ai., 2003), to name a few. Physiological studies in zebrafish have included, for example, imaging intracellular calcium during gastrulation (Gilland et al., 1999), in the intact spinal cord (O'Malley et al., 1996; Gahtan et al., 2002), and in brain (Brustein et aI., 2003). Other examples related to studies of embryos are covered in Chapter 43. Each of these biological preparations embodies its own unique set of specimen mounting and maintenance challenges. Indeed, it is sometimes necessary to anesthetize the preparation to prevent it from crawling or swimming away during the imaging session! Perhaps the most difficult conditions involve imaging in a living mammal, an undertaking for which the confocal or MP microscope enjoys the twin advantages of epi-illumination and optical sectioning that make it possible to view solid tissues without mechanical disruption. Confocal microscopy has long been an important tool for in vivo imaging of eye tissues non-invasively (Petran et al., 1986; Jester et al., 1991, 1992; Masters, 1992; Petroll et al., 1992, 1993; Poole et al., 1993). In terms of imaging interior tissues, early studies described methods for examining microcirculation of the brain cortex in anesthetized rats (Dirnagl et al., 1992) or changes in kidney tubules during ischemia (Andrews et at., 1991). Confocal microscopy also has been used to image leukocyte-endothelium interactions during infections through closed cranial windows (Lorenzi et ai., 1993). More recently, MP has been used to image live mammalian brain tissues in vivo, either through a cranial window (Svoboda et at., 1997; Trachtenberg et ai., 2002), fiberoptic coupled devices (Mehta et al., 2004), or directly through the intact but thinned skull (Yoder and Kleinfeld, 2002; Zhang et at., 2005). Dual-channel MP imaging also has been used to image other tissues in vivo, including lymphoid organs (e.g., Miller et al., 2002). It is generally accepted that MPimaging is superior to single-photon confocal for these in vivo imaging studies (Cahalan et aI., 2002).
Fluorescent Probes Except in those cases where an adequate image can be derived from either the backscattered-light signal or from autofluorescence, confocal microscopy of living cells is dependent on the properties and availability of suitable fluorescent probes. In addition to binding specifically to what one is interested in studying,
TABLE 19.3. Commercially Available Chambers for Live Cell Imaging Source 20/20 Technology, Inc. Bldg. 2, Unit A 311 Judges Road Wilmington, NC 28405 USA ALA Scientific Instruments Inc. 1100 Shames Dr. Westbury, NY 11590 USA ASI / Applied Scientific Instrumentation Inc. 29391 W. Enid Rd. Eugene, OR 97402 USA AutoMate Scientific, Inc. 336 Baden Street San Francisco, California 94131 USA Bellco Glass, Inc. 340 Edrudo Road, Vineland, NJ 08360 USA BioCrystal Ltd OptiCell 575 McCorkle Blvd. Westerville, OH 43082 USA Bioptechs, Inc. 3560 Beck Road Butler, PA 16002 USA Bioscience Tools - CB Consulting Inc., 4527 52nd Street, San Diego, CA 92115 USA C&L Instruments, Inc. 314 Scout Lane Hummelstown, PA 17036 USA CellBiology Trading Hinsbeker Berg 28a Hamburg, 22399 Germany Dagan Corporation 2855 Park Avenue, Minneapolis, Minnesota 55407 USA Digitimer Ltd 37 Hydeway Welwyn Garden City Hertfordshire, AL7 3BE, England Grace Bio-Labs, Inc. P.O. Box 228 Bend, OR 97709 USA
Harvard Apparatus Inc. 84 October Hill Rd. Holliston, MA 01746 USA Integrated BioDiagnostics Schellingstrasse 4 80799 MUnchen, Germany Intracel, Ltd. Unit 4 Station Road Shepreth, Royston Herts, SG8 6PZ England In Vitro Systems & Services GmbH Rudolf-Wissell-Str. 28 37079 Gottingen, Germany Life Imaging Services Kaegenstrasse 17 CH-4153 Reinach, Switzerland MatTek Corporation 200 Homer Avenue Ashland, MA 01721 USA Molecular Probes, Inc. 29851 Willow Creek Road Eugene, OR 97402 USA
DescriptioniFeatures
Contact Info
Heating, cooling, atmosphere control instrumentation for microscopy.
TEL: 1-910-791-9226 WEB: http://20-20tech.com/
Microincubators and temperature control; Peltier heating & cooling pre-stage; recording chambers; inline perfusion heating tube. Supplier for So lent and Bioptechs incubation chambers.
TEL: 516-997-5780 WEB: www.alascience.com EMAIL:
[email protected] TEL: 541-461-8181 WEB: http://www.asiimaging.com/ EMAIL:
[email protected] Programmable controlled perfusion systems, temperature control, valves and fittings, oocyte perfusion chamber, Petri dish perfusion chamber, sub-millisecond switching, submerged and interface tissue and brain slice chambers. Sykes-Moore culture chambers; used with stationary culture when medium is changed intermittently.
TEL: 415-239-6080 WEB: http://www.autom8.com/ EMAIL:
[email protected] OptiCell is a sterile, sealed cell culture environment between two optically clear gas-permeable growth surfaces in a standard microtiter plate-sized plastic frame with ports for access to the contents. Live cell microscopy environmental control systems. Thermal regulation of specimen and objective, electronic control and integration of temperature and perfusion. Glass bottom Petri dishes; ultra-thin imaging chambers; temperature control; perfusion systems; small volume delivery systems; ultra-fast temperature/solution switch. Fluorometers and fluorometer components for steady-state fluorescence measurements; complete fluorescence systems for photometry and fluorescence imaging. Microinjection and incubation; EMBL live cell observation chamber. Microscope stage temperature controller; perfusion controller.
AutoMate Scientific, Medical Systems and Scientific Systems Design incubators, chambers, and perfusion systems.
Manufactures 3-D microporous coatings on microscope slides, and a variety of "press to seal" enclosures for microarrays, cell culture, and high throughput cytochemistry, hybridization, cytogenetics, and fluorescent imaging applications. Variety of valve controlled perfusion systems.
The Il-slide family of live cell imaging flow chambers; suited for optical studies of hydrodynamic shear stress on biofilms or adhesion studies on cell layers. WillCo glass bottomed dishes; Bioptechs micro-environmental control systems.
Gas-permeable plastic foil (bioFOLIE 25); sterile tissue culture dish (petriPERM); Petri dish with gas-permeable base; two-compartment system. Ludin imaging chamber; microscope temperature control system Glass bottom culture dishes
Attofluor cell chamber designed for viewing live-cell specimens on upright or inverted microscopes. Chamber gaskets for imaging, perfusion, and incubation.
TEL: 1-800-257-7043 WEB: http://www.bellcoglass.com/ EMAIL:
[email protected] TEL: 614-8 18-0019 WEB: http://www.opticell.com EMAIL:
[email protected] TEL: 724-282-71 45 WEB: hltp:llwww.bioptechs.coml EMAIL:
[email protected] TEL: 1-877-853-9755 WEB: http://biosciencetools.com/ EMAIL:
[email protected] TEL: 1-717-564-9491 WEB: http://www.fluorescence.com/ EMAIL:
[email protected] TEL: 49-0-40-53889432 WEB: http://cellbiology-trading.com/ EMAIL:
[email protected] TEL: 612-827-5959 WEB: http://www.dagan.com/ EMAIL:
[email protected] TEL: +44 (0) 1707328347 WEB: http://www.digitimer.com/ EMAIL:
[email protected] TEL: 1-800-813-7339 WEB: http://www.gracebio.com/ EMAIL:
[email protected] TEL: 508-893-8999 WEB: http://www.harvardapparatus.com EMAIL:
[email protected] TEL: +49 (0)89 I 2180 64 19 WEB: http://www.ibidi.de/ EMAIL:
[email protected] TEL: 01763 262680 WEB: http://www.intracel.co.uk/ EMAIL:
[email protected] TEL: ++49 551 50097-0 WEB: http://www.ivss.de/ EMAIL:
[email protected] TEL: ++41 (0)61 7116461 WEB: http://www.lis.chi EMAIL:
[email protected] TEL: 1-800-634-9018 WEB: http://www.glass-bottom-dishes.com/ EMAIL:
[email protected] TEL: 1-541-465-8300 WEB: http://www.probes.com/ EMAIL:
[email protected] Confocal Microscopy of living Cells • Chapter 19
389
TABLE 19.3. (Continued)
PeCon GmbH ZiegeleistraBe 50 89155 Erbach Germany Physitemp Instruments, Inc. 154 Huron Avenue Clifton, New Jersey 07013 USA SDR Clinical Technology 213 Eastern Valley Way Middle Cove, NSW 2068 Australia Solent Scientific Limited 14 Matrix Park, Talbot Road, Segensworth POlS SAP, UK Stratech Scientific, Ltd. Unit 4 Northfield Business Park, Northfield Road, Soham, Cambridgshire CB7 5UE UK Warner Instruments, Inc. 1125 Dixwell Avenue, Hamden, CT 06514, USA WiliCo Wells BV WG Plein 287 1054 SE Amsterdam, The Netherlands World Precision Instruments, Inc. 175 Sarasota Center Boulevard Sarasota, Florida 34240 USA
Contact Info
DescriptioniFeatures
Source
Live cell imaging solutions including stage heating and cooling, CO, and 0, regulation, and evaporation reduction.
TEL: 0049 (0) 7305 95666-0 WEB: http://www.pe-con.de/peconlindex.htm EMAIL:
[email protected] Heating & cooling stages (Peltier) for microscopes (-20° to + 100°C); custom thermal stages.
TEL: 1-973-779-5577 WEB: http://www.physitemp.com/ EMAIL:
[email protected] TEL: +61-2-9958-2688 WEB: http://www.sdr.com.au/ EMAIL:
[email protected] TEL: +44 (0)870 774 7140 WEB: http://www.solentsci.com/ EMAIL:
[email protected] TEL: +44 (0)1353 722500 WEB: http://www.stratech.co.uk/
Physiological recording chambers & accessories for use on the microscope stage; bath perfusion; temperature control. Manufacturers of full enclosure incubation chambers for research inverted microscopes, confocal microscopes and multi-photon microscopes. CoverWell imaging chambers are designed to stabilize and support thick and free-floating specimens for confocal microscopy and imaging applications. Full range of recording, imaging, and perfusion chambers; perfusion and valve control. WiliCo-dish glass bottom dishes.
FluoroDish glass-bottom culture dish; Air-Therm ATX Air Heater Controller; programmable automated multi-channel perfusion system.
the fluorescent probe should produce a strong signal and be both slow to bleach and non-toxic. Chapters 16 and 17 discuss fluorescent dyes that have been used in published work with confocal microscopy in detail. Many dyes are useful when introduced to the medium surrounding cells to be labeled. Some of the classic and most commonly used cell stains include DiI for labeling the plasma membrane (Honig and Hume, 1986; Baker and Reese, 1993), DiOC 6(3) for labeling internal membranes (Terasaki et ai., 1984), NBD-ceramide and bodipy-ceramide which label the Golgi apparatus (Pagano et ai., 1991), rhodamine 123 which labels mitochondria (Johnson et ai., 1980), potential sensitive dyes such as DISBAC 2 (3) (see Fig. 8.65, this volume) (Loew, 1993), and FM 1-43 (Betz et al., 1992) which is used to follow plasma membrane turnover and vesicular release. Fluorescent ion indicators such as Fluo-3 (Minta et al., 1989) can either be microinjected or added to the media in a cell-permeant acetoxymethylester form that becomes trapped inside the cell after being cleaved by intracellular esterases (see Chapter 42, this volume).
Minimizing Photodynamic Damage Once the cells are labeled and on the microscope, one is faced with the challenge of collecting data without compromising the cell or bleaching the label. In practice, the major problem is light-induced damage. Fluorescent molecules in their excited state react with molecular oxygen to produce free radicals that can then damage cellular components and compromise cell health (Dixit and Cyr, 2003; see also Chapters 38 and 39, this volume). In addition, several studies suggest that components of standard culture media might also contribute to adverse light-induced effects on cultured cells (see Siegel and Pritchett, 2000). Some early studies (Spierenburg et ai., 1984; Zigler et ai., 1985, 1991; Lepe-Zuniga et ai., 1987) indicated a phototoxic effect of N-2-
TEL: 1-800-599-4203 WEB: http://www.warneronline.com/ EMAIL:
[email protected] TEL: ++31(0)206850171 WEB: http://www.willcowells.com EMAIL:
[email protected] TEL: 941-371-1003 WEB: http://www.wpiinc.com/
hydroxyethylpiperazine-N' -2-ethanesulfonic acid (HEPES) containing media on cells under some circumstances. It seems possible that this effect might be more directly related to inadequate levels of bicarbonate (Cechin et ai., 2002). Other studies suggest that riboflavin/vitamin B2 (Zigler et ai., 1985; Lucius et ai., 1998) and the essential amino acid tryptophan (Griffin et ai., 1981; Silva et ai., 1991; Silva and Godoy, 1994; Edwards et ai., 1994) may mediate phototoxic effects. Whether these effects occur under typical confocal imaging conditions is unknown, but many of the photoeffects are reduced by antioxidants, so it seems advisable to maintain antioxidants (and some bicarbonate, as well) in the specimen chamber (see below) and to use photons with great efficiency.
Improving Photon Efficiency There are several strategies to minimize the amount of excitation light required to collect data (see Chapters 2 and 9, this volume, for more details). Briefly, higher-NA objective lenses collect more of the fluorescent emission. For a given lens, there is also a theoretical optimal setting of the zoom magnification that best matches the resolution required to the allowable dose (see Chapter 4, this volume). When the focus plane is more than 51lm from the coverslip, water-immersion lenses should be used to avoid the signal loss caused by spherical aberration when using an oil lens (see Chapters 7 and 20, this volume). Another way to reduce light damage is to minimize the duration of the light exposure during the experimental setup. For example, one should try to focus as rapidly as possible and tum off the light source as soon as the focus range has been chosen. In addition, in single-beam scanning systems, make sure that your scanner is set up to blank the laser beam during scan retrace. Otherwise, areas on both sides of the imaged area will receive a very high light exposure as the beam slows down to change direction. Finally, photon efficiency can be maximized by using the best mirrors, the correct pinhole size for the resolution required (in x,
390
Chapter 19 • M.E. Dailey et al.
y, and z), and photodetectors that yield the highest quantum efficiency at the wavelength of the signal.
Antioxidants As noted above, one can also reduce photodynamic damage by adding antioxidants to the medium. Oxyrase (Oxyrase Inc., http://www.oxyrase.com) is an enzyme additive used to deplete oxygen in order to grow anaerobic bacteria. It has been used at 0.3 unitlmL to reduce photodynamic damage during observations of mitosis (Waterman-Storer et aI., 1993). Another approach is to include ascorbic acid in the medium. This reducing agent is typically used at 0.1 to 1.0 mg/mL but has been used at up to 3 mg/mL. A recent confocal study of calcium transients in isolated chondrocytes reported a relationship between laser intensity and the frequency of Ca2+ oscillations and cell viability: Ca2+ events were more frequent and cell viability was decreased with higher laser intensity (Knight et al., 2003). Treatment with ascorbic acid reduced the Ca2+ events and improved cell viability (see also Chapters 16 and 17, this volume).
THE ONLINE CONFOCAL COMMUN ITY Confocal microscopy of living cells is an area of active research where individuals are constantly developing new techniques and approaches. One way to keep up with current practice is to join about 1600 others who subscribe to the Confocal e-maillistserver. This can be done by registering at the listserver Web site, located at http://listserv.buffalo.edu/archives/confocal.html. You will then begin to receive messages from other microscopists. Recent topics have included discussions on such diverse issues as autofluorescence problems, glass-bottomed culture chambers, damage to live cells during FRAP experiments, and announcements of confocal workshops. The listserver also has an extensive, searchable archive dating back to 1991, and this is freely accessible.
A CONVENIENT TEST SPECIMEN Knebel et al. (1990) showed that onion epithelium (Allium cepa) is a simple preparation that can be used as a convenient test specimen for confocal microscopy of living cells. Figure 19.1 shows
FIGURE 19.1. A convenient test specimen for confocal microscopy of cells in a living tissue: onion epithelium. This drawing shows how to obtain onion epithelium. As described in the text, the epithelium is stained with DiOC 6 (3), and the ER and mitochondria within the cells provide a bright and motile specimen.
how to prepare onion epithelium. First, a small square of a layer is cut out using a razor blade. A forceps is used to peel off the thin epithelium on the inner surface of the onion layer. The epithelium is then put onto a microscope slide, covered with a drop or two of staining solution containing DiOC 6 (3), a marker of mitochondria and endoplasmic reticulum, and coverslipped. The stock solution of DiOC 6 (3) (0.5 mg/mL in ethanol) can be kept indefinitely if protected from light in a scintillation vial. The staining solution is a 1 : 1000 dilution in water on the day of the experiment. The center of these cells is usually occupied by a large vacuole, and the ER and mitochondria are located in a thin cytoplasmic region near the plasma membrane. Motion of the ER is relatively quick and easily detected in consecutive J-s scans.
SPECIFIC EXAMPLE I: VISUALIZING CHROMATIN DYNAMICS USING VERY lOW LIGHT LEVELS It is clear from the discussion above that microscopy of living cells has become a technique of major importance in cell biology: it can be used to tell us where molecules are located, when they become localized, how fast they are moving, with which molecules they are interacting, and how long they stay attached to these molecules. All these properties can be observed in the natural environment of the living cell. The major limiting factor in live-cell imaging is phototoxic effect of light used for the observation of the cell. Here we will address some practical issues of phototoxicity based on our experience in imaging chromatin dynamics in living cells (Manders et aI., 1996; 1999; 2003; Verschure et aI., 1999; Mone et al., 2004).
Phototoxicity A large number of photochemical reactions are responsible for the photo toxic effect of light. Light can be absorbed by cellular components and induce chemical alterations in their molecular structure. For example, UV light is absorbed by DNA (absorption peak at 280nm), directly inducing DNA damage. Here we assume that, working with visible light, the direct photodamage is negligible. In fluorescently labeled cells, the main source of photodamage is the production of reactive oxygen species (ROS) including singlet oxygen e02), superoxide (·On, hydroxyl radical CHO·), and various peroxides. These activated oxygen species react with a large variety of easily oxidizable cellular components, such as proteins, nucleic acids, and membrane lipids. Singlet oxygen is responsible for much of the physiological damage caused by reactive oxygen species. For the production of singlet oxygen, the fluorescent label acts as a photo synthesizer in a photochemical reaction where dioxygen C02) converts into singlet oxygen e02)' Singlet oxygen mainly modifies nucleic acid through the selecti ve oxidation of deoxyguanosine into 8-oxo-7 ,8-dihydro-2'deoxyguanonine. Proteins and lipids also will be damaged by ROS. Phototoxicity likely depends on several variables:
• The photochemical properties of the fluorescent molecule. Some molecules induce more phototoxicity than others, depending on the lifetime of their triplet state. For photodynamic therapy (PDT), dedicated molecules called photos ensitizers have been designed in order to induce a maximum damage in tissue for the treatment of cancer (e.g., halogenated fluorescein is much more toxic than fluorescein). Another property that influences the phototoxicity of a molecule is the
Confocal Microscopy of living Cells • Chapter 19
local environment of the molecule. The active fluorophore of a GFP molecule is positioned on the inside of the protein, within the barrel structure (the "13-can"). Probably this hydrophobic protein environment contributes to the relatively low phototoxicity of GFP compared with naked fluorophores such as fluorescein or rhodamine. • The subcellular location of the fluorescent molecule. When fluorescent molecules are situated close to DNA, the damaging effect of singlet oxygen is more pronounced. Despite several DNA-repair mechanisms, the cell will not continue its cell cycle (arrest) and may even die if there is too much DNA damage. Therefore, fluorophores in the cytoplasm seem to induce less phototoxicity than fluorophores in the nucleus. • The concentration of fluorophore. It is clear that there is a relationship between the local concentration of fluorophore and the level of phototoxicity. We assume a linear relationship between fluorophore concentration and toxicity, although this has not been assessed directly and is complicated by the fact that if there is more dye, one need use less excitation. • The excitation intensity. Fluorescent cells in a dark incubator are quite happy for weeks as long they are maintained in the dark. As the word phototoxicity implies, photons are needed to induce toxicity in a fluorescently labeled specimen. We usually assume a linear relationship between excitation light dose and toxicity, although the temporal regimen of the excitation may be important to how cells handle the accumulation of phototoxic biproducts. Phototoxicity is dependent on the wavelength of light in the sense that the wavelength of the toxic excitation light matches the excitation curve of the fluorophore. In other words, it is the excited fluorophore that is toxic. Koenig also found that, with two-photon excitation, the damage is proportional to the number of molecular excitations (see Chapter 38, this volume). There is no clear evidence for differences in phototoxicity between green, red, or far-red fluorophores. In principle, excited CyS can be as toxic as excited FITC. However, the wavelength of excitation light can be a factor when imaging in thick specimen because stronger incident illumination is needed for comparable excitation of shorter wavelength fluorophores due to increased tissue scatter at shorter wavelengths.
Reduction of Phototoxicity For many researchers, phototoxicity is a serious (and annoying!) limitation of their observations of living cells. When you do not look at a cell it is alive, but the moment you start to observe how it lives, it is killed by the light used to observe it. In experiments so far, we have succeeded in obtaining acceptable time series of living cells by carefully optimizing all steps in the imaging process in an effort to achieve (i) maximal signal-to-noise ratio (SIN), (ii) maximal spatial and temporal resolution, and (iii) minimal phototoxic effects. Specifically, phototoxicity has been minimized by (i) using radical scavengers (e.g., trolox) in the culture medium and (ii) using culture medium without phenol-red. Most important of all, however, is minimizing the total excitation light dose. The excitation light dose is the product of the light intensity and the exposure time. Decreasing either the excitation intensity or the excitation dose implies a loss of fluorescent signal. It is inevitable
391
Improving Image Quality in Low-Dose Microscopy Figure 19.2(A) shows a single time frame projection from a 3D time-series of a HeLa cell expressing the fluorescent histone fusion protein, H2B-GFP. In the time series shown in Figure 19.2(B), the cell is in late telophase at the start of the imaging and proceeds into interphase during the movie. This movie shows data from a study on the dynamics of chromatin during decondensation (Manders et at., 2003).1 In these experiments, the excitation light intensity was kept below lS0nW and the total exposure time 2 of a cell that was 3D imaged for 3 h was not more than 70 s. Under these conditions the total light dose was approximately 10 1 cm-2• In experiments where we used a higher dose of light we observed phototoxic effects, such as cell cycle arrest and cell death. Reducing the total light dose during an experiment requires that the number of 3D images in the sequence (temporal sampling rate) be low. Because of this limited sampling rate, live-cell movies are usually under-sampled in time according to the Nyquist criterion. As a result, such movies often show cells that nervously move from one place to another and sometimes suddenly rotate. We have applied an image processing procedure to correct for all the movements (translation and rotation) of the cell. For each 3D image of the time sequence, a translation and rotation transform vector was calculated in order to obtain a best fit with the previous image in the sequence. After a series of such transformations, a new movie was produced showing a stable cell that does not move or rotate. Only internal movements are visible. After this correction procedure, we applied a simple Gaussian spatial filter to reduce noise in the image [Fig. 19.2(C,D)]. We also applied a temporal filter by adding to each voxel of the 3D image at each time-point the value for that voxel in the previous and subsequent image multiplied by an intensity factor of O.S. Our experience is that temporal filtering makes the movie easier to interpret.
Low-Dose Imaging Conclusion The success of live-cell microscopy is very much dependent on minimizing or avoiding any toxic effect of light on the biological system under observation. A certain dose of light may induce serious DNA damage that may arrest the cell cycle, whereas the diffusion coefficient of a certain protein is not influenced at all at the same dose. In the experimental example shown here [Fig. 19.2(B)], we used only ISOnW of incident beam power. This dose was found to be phototoxic in other experiments using fluorescein instead of GFP, and it was found necessary to drop the laser power to SOnW [Fig. 19.2(E)]. These power levels are farlower than (i.e., 20 h) at relatively short time intervals (-5 min).
A
B
Call 3
FIGURE 19.4. Use of a Dynamic Image Analysis System (DIAS; Soll, 1995, 1999) to characterize microglial motility behaviors in time-lapse imaging experiments. (A) Boundaries (red lines) of three FITC-IB4-labeled microglial cells were defined by automated, computer-assisted edge detection. (B) Tracings of the three cells show motility behaviors over a 2h period. The cell centroid (black dots) were computed and plotted for each time-point. Note that all three cells are motile, but cell 2 does not locomote. (C) Areas of new protrusion (green) and resorption (red) are shown for cell 1 at 4min intervals. The new centroid (open dot) is shown in relation to the former centroid (filled dot).
Confocal Microscopy of Living Cells • Chapter 19
397
B
~2
FIGURE 19.5. Two-channel time-lapse imaging of microglia (green) and cell nuclei (red). (A) Time lapse sequence shows capability of simultaneously imaging IB 4 -labeled microglia and nuclei of cells using a cell membrane permeant DNA binding dye (Syto61). Note the nuclei (arrows) in two different migrating microglial cells. The sequence spans l56min. Only a small portion of the original field of view is shown. (8) Automated tracing of the movement of cell nuclei shows paths taken by nuclei of cells in the experiment show in (A). Nuclei were detected by DIAS software (Soli, 1995). Only a select subset of nuclei are shown. Note differences in the movement among different cells.
398
Chapter 19 • M.E. Dailey et al.
Microglia
Dead cell nuclei
Merged
C C
CIS
£
U I
N
FIGURE 19.6. Two- and three-dimensional, two-channel time-lapse imaging of microglial movements and phagocytosis of dead cell nuclei in live rat brain tissue slice cultures (P6+2DIV). (A) Microglia are stained with a fluorescent lectin, FITe-IB4' (B) Dead cell nuclei are stained with To-Pro-3. (e) Merged image of (A) and (B). The image represents a projection of nine optical sections spanning 40llm in z-depth. Images were captured using a 20xJO.7 dry objective. (D) Time-lapse sequence of a portion of the field of view above showing phagocytic clearance of a dead cell nucleus (arrowhead) by a locomotory microglial cell (arrow). Note that the microglial cell maintains a rapid rate of locomotion as it sweeps over and picks up the dead cell nucleus. (E-G) 3D stereo projections of images in A through e above. Use red-green glasses (red over left eye) to view depth. (H) 3D stereo time-lapse sequence showing mitosis of a microglial cell (arrow) near the surface of the brain slice culture. The small, round objects represent the condensed nuclei of dead cells. Time is shown in minutes. See the supplemental video movies at http://www.springer.comJO-387-25921-X (Adapted from Petersen and Dailey, 2004.)
Multi-channel (5D) imaging can provide information on the dynamic relationship of different cell or tissue components. Future developm~nts should address constraints on high-resolution imaging deep (>50/J.m) within tissue. Improvements will likely be achieved by using water-immersion lenses and external automatic spherical aberration correctors (see Chapter 20, this volume) and by employing longer wavelength dyes to reduce light scatter by the tissue and to minimize phototoxic effects.
FUTURE DIRECTIONS For confocal microscopy of living cells, the most important characteristic of the instrument is its efficiency in collecting and detecting the fluorescence emission light from the specimen (Chapter 2, this volume). Any improvement in this efficiency reduces the amount of light damage and allows the gathering of more data. The increased data can either be in the form of more images, images
Confocal Microscopy of Living Cells • Chapter 19
with less statistical noise, or images obtained with greater spatial or temporal resolution. Newer models of existing commercial confocal microscopes have substantially improved photon efficiency. In addition, there have been technological improvements in the ability to separate the excitation and fluorescence emission of f1uorophores, providing greater flexibility for multi-channel imaging and quantitative image analysis in live cells and tissues. Finally, the advantages of either Gaussian-filtering 2D data or deconvolving 3D data to reduce the effects of Poisson noise are now widely appreciated. Routine application of this approach can reduce the light load to the specimen by a factor of from LO to LOO while still producing images with the same apparent resolution and signal-tonoise ratio. Technological and conceptual advancements are also likely to push the spatial and temporal resolution and other modes of fluorescence microscopy (e.g., Gustafsson, 1999; Hell et ai., 2004; Chapters 13,30, and 31, this volume) . Some of these approaches (e.g., 4Pi-microscopy) look promising for live cells (Gugel et ai., 2004) , but their potential for widespread use in biological applications has yet to be established, and there are limitations on the sample thickness (Gustafsson, 1999 and Chapter 21 , this volume). In addition , higher resolution implies smaller pixels and therefore more photons/square micrometer and more bleaching and toxicity. Undoubtedly, there will be more improvements and wider applications along these lines in the future. Although it is difficult to predict the future of confocal microscopy of living cells, as confocal microscopy (and its richer cousin, multi-photon microscopy) are in all probability the optimal methods for studying the 3D structure of living cells, the future seems sure to be bright!
ACKNOWLEDGMENTS The authors thank Glen MacDonald for helpful information on autofluorescence and light sensitivity of culture media, and for bringing relevant literature to our attention. MED thanks the staff of the W.M. Keck Motion Analysis Facility in the Department of Biological Sciences, University of Iowa, for their assistance in obtaining the data shown in Figures 19.4 and 19.5. Data for Figures 19.4 and 19.5 were kindly provided by Marc Waite. Data for Figure 19.6 were kindly provided by Mark Petersen.
REFERENCES Adams, C.L., Chen, YT. , Smith, SJ. , and Nelson, W.J., 1998 , Mechani sms of epithelial cell-cell adhesion and cell compaction revealed by hi ghresolution tracking of E-cadherin-green flu orescent protein, J. Cell Bioi. 142: 1105-1119. Adams, M.e., Salmon, W.C., Gupton, S.L., Cohan, e.S., Wittmann, T. , Prigozhina, N., and Waterman-Storer, C.M. , 2003, A high-speed multispectral spinning-disk confocal microscope system for fluorescent speckle microscopy of living cells, Meth ods 29:29-41. Amos, W.B. , Reichelt, S., Cattermole, D.M ., and Laufer, 1., 2003 , Reevaluation of differential phase contrast (DPC) in a scanning laser microscope using a split detector as an alternative to differential interference contrast (DlC) optics, J. Mierose. 210:166-175. Ando , R., Hama, H. , Yamamoto-Hino, M. , Mizuno, H ., and Mi yawaki, A. , 2002, An optical marker based on the UV-induced green-to-red photoconversion of a fluorescent protein, Proe. Natl. A cad. Sci. USA 99: 12651-12656. Ando, R., Mizuno , H., and Miyawaki, A., 2004, Regulated fast nucleocytoplasmic shuttling observed by revers ible protein highlighting, Science 306: 1370-1373.
399
Andrews, P.M., Petroll, W.M., Cavanagh, H.D., and Jester, J.V., 1991 , Tandem scanning confocal microscopy (TSCM) of normal and ischemic living kidneys , Am. 1. Anat. 191 :95-102. Andrews, P.O., Harper, I.S. , and Swedlow, J.R, 2002, To 50 and beyond: Quantitative fluorescence microscopy in the postgenomic era, Traffic 3:29-36. Ashkin, A., Schutze, K., Dziedzic, J.M. , Euteneuer, U., and Schliwa, M. , 1990, Force generation of organelle transport measured in vivo by an infrared laser trap, Nature 348:346-348. Ashworth, R , 2004, Approaches to measuring calcium in zebrafi sh: Focus on neuronal development , Cell Calcium 35:393-402. Axelrod, D. , Koppel , D.E., Schlessinger, J. , Elson, E., and Webb, W.w., 1976, Mobility measurement by analysis of fluorescence photobleaching recovery kinetics , Biophys. J. 16:1055- 1069. Bacia, K. , and Schwille, P., 2003 , A dynamic view of cellular processes by in vivo fluorescence auto- an d cross-correlation spectroscopy, Methods 29:74-85. Baker, G.E., and Reese, B.E. , 1993, Using confocal laser scanning microscopy to investigate the organi zation and development of neuronal projections labeled with Oil, Methods Cell BioI. 38:325-344. Baker, M.W. , Kauffman B., Macagno, E.R, and Zipser, B., 2003, In vivo dynamics of CNS sensory arbor formation: a time-lapse study in the embryonic leech, J. Neurobiol . 56:41-53. Ballestrem, e. , Hinz, B., Imhof, B.A. , Wehrle-Haller, B. , 2001 , Marching at the front and dragging behind: differential alphaVbeta3-integrin turnover regulates focal adhesion behavior, J . Cell BioI. 155(7):1319-1332. Barber, R.P. , Phelps, P.E., and Vaughn, J.E., 1993, Preganglionic autonomic motor neurons display normal translocation patterns in slice cultures of embryonic rat spinal cord, 1. Neurosci. 13:4898-4907. Bastiaens, P.I., and Pepperkok, R., 2000, Observing proteins in their natural habitat: The living cell, Trends Biochem. Sci. 25:631-637. Bastiaens, P.l., and Squire, A. 1999, Fluorescence lifetime imaging microscopy: Spatial resolution of biochemical processes in the cell, Trends Cell Bioi. 9:48-52. Baumgartner, W., Schutz, GJ. , Wiegand, J .. Golenhofen, N., and Drenckhahn, D. , 2003, Cadherin function probed by laser tweezer and single molecule fluorescence in vascular endothelial cells, 1. Cell Sci. 116: 100 I-I 011 . Bement, W.M. , Sokac, A.M., and Mandato, C.A., 2003, Four-dimensional imaging of cytoskeletal dynamics in Xenopus oocytes and eggs, DifferenTiation 71:518-527. Benedi ktsson, A.M., Schachtele, SJ., Green, S.H. , and Dailey. M.E., 2005, Ballistic labeling and dynamic imaging of astrocytes in organotypic hippocampal slice cultures. 1. Neurosci. Methods 141:41-53. Berg. RH. , 2004, Evaluation of spectral imaging for plant cell analysis, J. Micro.I'c. 214(Pt 2):174-181. Betz, W.J. , Mao, E , and Bewick, G.S. , 1992, Activity-dependent fluorescent staining and destaining of living vertebrate motor nerve terminals, J. New·osei . 12:363-375. Block, S.M ., Goldstein , L.S. , and Schnapp, B.J. , 1990, Bead movement by si ngle kinesin molecules studied with optical tweezers, Nature 348:348-352. Bloom, J.A. , and Webb, W.W. , 1984, Photodamage to intact erythrocyte membranes at high laser intensities: methods of assay and suppression, 1. Histochem. Cytochem. 32:608- 6 16. Brum, G., Gonzalez, A., Rengifo, J. , Shirokova, N .• and Rios, E., 2000, Fast imaging in two dimension s resolves extensive sources of Ca2• sparks in frog skeletal muscle, J . Physiol. 528:419-433. Brustein, E., Marandi, N., Kovalchuk, Y , Drapeau, P., and Konnerth, A., 2003. "In vivo" monitoring of neuronal network activity in zebrafish by twophoton Ca(2+) imaging, Pflugers A rch. 446(6):766-773. Cahalan , M.D., Parker, I. , Wei , S.H. , and Miller, MJ. , 2002, Two-photon tissue imaging: Seeing the immune system in a fresh light, Na t. Rev. lmmunol. 2:872-880. Chalfie, M. , Tu, Y , Euskirchen , G., Ward, W.W. , and Prasher, D.C., 1994, Green fluorescent protein as a marker for gene expression, Science 263:802-805. Chen, Y., Mills, J.D., and Periasamy, A. , 2003, Protein localization in living cells and tissues using FRET and FUM, Differentiation 71 :528-541. Cechin, S.R., Gottfried, e. , Prestes, e.e., Andrighetti, L., Wofchuk, S.T., and Rodnight, R., 2002, Astrocyte stellation in saline media lacking bicarbon-
400
Chapter 19 • M.E. Dailey et al.
ate: possible relation to intracellular pH and tyrosine phosphorylation, Brain Res. 946:12-23. Cleemann, L., Wang, W., and Morad, M., 1998, Two-dimensional confocal images of organization, density and gating of focal Ca'+ release sites in rat cardiac myocytes, Proc. Nat!. Acad. Sci. USA 95:10984-10989. Cogswell, C.l., and Sheppard, C.l.R, 1991, Visualization of 3-D phase structure in confocal and conventional microscopy, Proc. SPIE. 1450:323-328. Cogswell, C.1., and Sheppard, C.l.R, 1992, Confocal differential interference contrast (DIC) microscopy: Including a theoretical analysis of conventional and confocal DIC imaging, l. Microsc. 165:81-101. Cooper, M.S., Cornell-Bell, A.H., Chernjavsky, A., Dani, 1.W., and Smith, S.1., 1990, Tubulovesicular processes emerge from trans-Golgi cisternae, extend along microtubules, and interlink adjacent trans-Golgi elements into a reticulum, Cell 61 :135-145. Cooper, M.S., D' Amico, L.A., and Henry, C.A., 1999, Confocal microscopic analysis of morphogenetic movements, Methods Cell Bioi. 59: 179-204. Cornell-Bell, A.H., Finkbeiner, S.M., Cooper, M.S., and Smith, S.l., 1990, Glutamate induces calcium waves in cultured astrocytes: Long-range glial signaling, Science 247:470--473. Crittenden, S.L., and Kimble 1., 1999, Confocal methods for Caenorhabditis elegans, Methods Mol Biol. 122:141-151. Dailey, M.E., and Smith, S.l., 1993, Confocal imaging of mossy fiber growth in live hippocampal slices, lpn. l. Physiol. 43:S183-S192. Dailey, M.E., and Smith, S.l., 1994, Spontaneous Ca2+ transients in developing hippocampal pyramidal cells, l. Neurobiol. 25:243-251. Dailey, M.E., and Smith, S.l., 1996, The dynamics of dendritic structure in developing hippocampal slices, l. Neurosci. 16:2983-2994. Dailey, M.E., Buchanan, 1., Bergles, D.E., and Smith, S.l., 1994, Mossy fiber growth and synaptogenesis in rat hippocampal slices in vitro, l. Neurosci. 14:1060-1078. Dailey, M.E., and Waite, M., 1999, Confocal imaging of microglial cell dynamics in hippocampal slice cultures, Methods 18:222-230. Dailey, M.E., Marrs, G.S., and Kurpius, D., 2005, Maintaining live cells and tissue slices in the imaging setup, In: Imaging in Neuroscience and Development (R Yuste and A. Konnerth, eds.), Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York. Das, T., Payer, B., Cayouette, M., and Harris, W.A., 2003, In vivo time-lapse imaging of cell divisions during neurogenesis in the developing zebrafish retina, Neuron 37:597-609. Delbridge, L.M., Harris, P.l., Pringle, 1.T., Dally, L.l., and Morgan, T.O., 1990, A superfusion bath for single-cell recording with high-precision optical depth control, temperature regulation, and rapid solution switching, Pfiuegers Arch. 416:94-97. Del Pozo, M.A., Kiosses, W.B., Alderson, N.B., Meller, N., Hahn, K.M., Schwartz, M.A., 2002, Integrins regulate GTP-Rac localized effector interactions through dissociation of Rho-GDI, Nat. Cell BioI. 4(3):232-239. Dickinson, M.E., Bearman, G., Tille, S., Lansford, R, Fraser, S.E., 2001, Multispectral imaging and linear unmixing add a whole new dimension to laser scanning fluorescence microscopy, Biotechniques 31(6):1272, 1274-1276, 1278. Dirnagl, U., Villringer, A., and Einhaupi, K.M., 1992, In vivo confocal scanning laser microscopy of the cerebral microcirculation, l. Microsc. 165:147-157. Dixit, R., and Cyr, R., 2003, Cell damage and reactive oxygen species production induced by fluorescence microscopy: Effect on mitosis and guidelines for non-invasive fluorescence microscopy, Plant l. 36:280-290. Dorman, G., and Prestwich, G.D., 2000, Using photolabile ligands in drug discovery and development, Trends Biotechnol. 18:64-77. Dunn, G.A., Dobbie, I.M., Monypenny, 1., Holt, M.R., and Zicha, D., 2002, Fluorescence localization after photobleaching (FLAP): A new method for studying protein dynamics in living cells, l. Microsc. 205: 109-112. Dunwiddie, T.V., 1981, Age-related differences in the in vitro rat hippocampus: Development of inhibition and the effects of hypoxia, Dev. Neurosci. 4:165-175. Dvorak, 1.A., and Stotler, W.F., 1971, A controlled-environment culture system for high resolution light microscopy, Exp. Cell Res. 68:269-275. Dynes, 1.L., and Ngai, 1., 1998, Pathfinding of olfactory neuron axons to stereotyped glomerular targets revealed by dynamic imaging in living zebrafish embryos, Neuron 20: 108 1-109 l.
Edwards, A.M., Silva, E., lofre, B., Becker, M.l., and De Ioannes, A.E., 1994, Visible light effects on tumoral cells in a culture medium enriched with tryptophan and riboflavin, l. Photochem. Photobiol. B 24:179-186. Elangovan, M., Day, R.N., and Periasamy, A., 2002, Nanosecond fluorescence resonance energy transfer-fluorescence lifetime imaging microscopy to localize the protein interactions in a single living cell, l. Microsc. 205:3-14. Ellenberg, 1., Lippincott-Schwartz, J., and Presley, 1.F., 1999, Dual-colour imaging with GFP variants, Trends Cell Bioi. 9:52-56. Elson, E.L., 2001, Fluorescence correlation spectroscopy measures molecular transport in cells, Traffic 2:789-796. Fan, G.y', Fujisaki, H., Miyawaki, A., Tsay, RK., Tsien, RY., and Ellisman, M.H., 1999, Video-rate scanning two-photon excitation fluorescence microscopy and ratio imaging with chameleons, Biophys. l. 76:2412-2120. Firestone, L., Cook, K., Culp, K., Talsania, N., and Preston, K. lr., 1991, Comparison of autofocus methods for automated microscopy, Cytometry 12: 195-206. Flock, A., Flock, B., and Scarfone, E., 1998, Laser scanning confocal microscopy of the hearing organ: fluorochrome-dependent cellular damage is seen after overexposure, l. Neurocytol. 27:507-516. Forsythe, l.W., 1991, Microincubator for regulating temperature and superfusion of tissue-cultured neurons during electrophysiological or optical studies, Methods Neurosci. 4:301-318. Fraser, S.E., and O'Rourke, N.A., 1990, In situ analysis of neuronal dynamics and positional cues in the patterning of nerve connections, l. Exp. Bioi. 153:61-70. Frostig, R, ed., 2002, In Vivo Optical Imaging of Brain Function, CRC Press, Boca Raton, Florida. Gahtan, E., Sankrithi, N., Campos, 1.B., and O'Malley, D.M., 2002, Evidence for a widespread brain stem escape network in larval zebrafish, l. Neurophysiol. 87(1):608-614. Gahwiler, B.H., 1984, Development of the hippocampus in vitro: Cell types, synapses, and receptors, Neuroscience 11 :751-760. Gahwiler B.H., Capogna M., Debanne D., McKinney RA., and Thompson S.M., 1997, Organotypic slice cultures: A technique has come of age, Trends Neurosci. 20:471-477. Gan, W.B., Kwon, E., Feng, G., Sanes, 1.R, and Lichtman, 1.W., 2003, Synaptic dynamism measured over minutes to months: Age-dependent decline in an autonomic ganglion, Nat. Neurosci. 6:956-960. Gerlich, D., and Ellenberg, 1., 2003, 4D imaging to assay complex dynamics in live specimens, Nat. Cell Bioi. 5:S14-S19. Gerlich, D., Beaudouin, J., Gebhard, M., Ellenberg, J., and Eils, R, 2001, Fourdimensional imaging and quantitative reconstruction to analyse complex spatiotemporal processes in live cells, Nat. Cell Bioi. 3:852. Gilland, E., Miller, A.L., Karplus, E., Baker. R, and Webb, S.E., 1999, Imaging of multicellular large-scale rhythmic calcium waves during zebrafish gastrulation, Proc. Nat!. Acad. Sci. USA 96: 157-161. Gleason, M.R, Higashijima, S., Dallman, J., Liu, K., Mandel, G., and Fetcho, 1.R., 2003, Translocation of CaM kinase II to synaptic sites in vivo, Nat. Neurosci.6:217-218. Goksor, M., Enger, 1., and Hanstorp, D., 2004, Optical manipulation in combination with multiphoton microscopy for single-cell studies, App!. Opt. 43:4831-4837. Gong, Y., Mo, C., and Fraser, S.E., 2004, Planar cell polarity signalling controls cell division orientation during zebrafish gastrulation, Nature 430:689-693. Griffin, F.M., Ashland, G., and Capizzi, RL., 1981, Kinetics of phototoxicity of Fischer's medium for L5178Y leukemic cells, Cancer Res. 41:22412248. Grossmann, R, Stence, N., Carr, 1., Fuller, L., Waite, M., and Dailey, M.E., 2002, Juxtavascular microglia migrate along brain capillaries following activation during early postnatal development, Glia 37:229-240. Gugel, H., Bewersdorf, 1., lakobs, S., Engelhardt, 1., Storz, R., and Hell, S.W., 2004, Combining 4Pi excitation and detection delivers seven-fold sharper sections in confocal imaging of live cells, Biophys. l. 87(6):4146-4152. Gustafsson, M., 1999, Extended resolution fluorescence microscopy, Curro Opin. Struct. Bioi. 9:627-634. Hammond, A.T., and Glick, B.S., 2000, Raising the speed limits for 4D fluorescence microscopy, Traffic 1:935-40.
Confocal Microscopy of living Cells • Chapter 19 Hasan, M.T., Friedrich, R.W., Euler, T., Larkum, M.E., Giese, G., Both, M., Duebel, J., Waters, J., Bujard, H., Griesbeck, 0., Tsien, R.Y., Nagai, T., Miyawaki, A., and Denk, W., 2004, Functional fluorescent Ca2 + indicator proteins in transgenic mice under TET control, PLoS Bioi. 2:eI63. Epub 2004 Jun 15. Heid, P.J., Voss, E., and Soil, D.R., 2002, 3D-DIASemb: A computer-assisted system for reconstructing and motion analyzing in 4D every cell and nucleus in a developing embryo, Dev. Bioi. 245:329-347. Hell, S.W., Dyba, M., and Jakobs, S., 2004, Concepts for nanoscale resolution in fluorescence microscopy, Curro Opin. Neurobiol. 14:599-609. Hess, S.T., Huang, S., Heikal, A.A., and Webb, WW, 2002, Biological and chemical applications of fluorescence correlation spectroscopy: A review, Biochemistry 41:697-705. Hiraoka, Y, Shimi, T., and Haraguchi, T., 2002, Multispectral imaging fluorescence microscopy for living cells, Cell Struct. Funct, 27:367-374. Hollingworth, S., Soeller, c., Baylor, S.M., and Cannell, M.B., 2000, Sarcomeric Ca2+ gradients during activation of frog skeletal muscle fibres imaged with confocal and two-photon microscopy, J. Physiol. 526:551-560. Honig, M.G., and Hume, R.I., 1986, Fluorescent carbocyanine dyes allow living neurons of identified origin to be studied in long-term cultures, J. Cell Bioi. 103:171-187. Hook, G.R., and Odeyale, C.O., 1989, Confocal scanning fluorescence microscopy: A new method for phagocytosis research, J. Leukoc. Bioi. 45:277-282. Hoppe, A.D., Swanson, J.A., 2004, Cdc42, Racl, and Rac2 display distinct patterns of activation during phagocytosis, Mol. BioI. Cell 15(8):3509-3519. Ince, c., Ypey, D.L., Diesselhoff-Den Dulk, M.M., Visser, J.A., De Vos, A., and Van Furth, R., 1983, Micro-CO,-incubator for use on a microscope, J. Immun. Methods 60:269-275. Isogai, S., Lawson, N.D., Torrealday, S., Horiguchi, M., and Weinstein, B.M., 2003, Angiogenic network formation in the developing vertebrate trunk, Development 130:5281-5290. Jester, J.y', Andrews, P.M., Petroll, W.M., Lemp, M.A., and Cavanagh, H.D., 199 I, In vivo, real-time confocal imaging, J. Electron Microsc. Tech. 18:50-60. Jester, J.y', Petroll, W.M., Garana, R.M.R., Lemp, M.A., and Cavanagh, H.D., 1992, Comparison of in vivo and ex vivo cellular structure in rabbit eyes by tandem scanning microscopy, J. Microsc. 165:169-181. Johnson, L.y., Walsh, M.L., and Chen, L.B., 1980, Localization of mitochondria in living cells with rhodamine 123, Proc. Natl. Acad. Sci. USA 77:990-994. Jontes, J.D., Buchanan, J., and Smith, S.J., 2000, Growth cone and dendrite dynamics in zebrafish embryos: early events in synaptogenesis imaged in vivo, Nat. Neurosci. 3:231-237. Kim, S.A., Heinze, K.G., Waxham, M.N., and Schwille, P., 2004, Intracellular calmodulin availability accessed with two-photon cross-correlation, Proc. Natl. Acad. Sci. USA 101: 105-110. Keller, P, Toomre, D., Dlaz, E., White, J., and Simons, K., 2001, Multicolour imaging of post-Golgi sorting and trafficking in live cells, Nat. Cell Bioi. 3:140-149. Knebel, W., Quader, H., and Schnepf, E., 1990, Mobile and immobile endoplasmic reticulum in onion bulb epidermis cells: short- and long-term observations with a confocal laser scanning microscope, Eur. J. Cell Bioi. 52:328-340. Knight, M.M., Roberts, S.R., Lee, D.A., and Bader, D.L., 2003, Live-cell imaging using confocal microscopy induces intracellular calcium transients and cell death, Am. J. Physiol. Cell Physiol. 284:CI083CI0839. Koppel, D.E., Axelrod, D., Schlessinger, J., Elson, E.L., and Webb, WW, 1976, Dynamics of fluorescence marker concentration as a probe of mobility, Biophys. J. 16:1315-1329. Koster, R.W., and Fraser, S.E., 2001, Direct imaging of in vivo neuronal migration in the developing cerebellum, Curro Bioi. 11:1858-1863. Kuo, S.c., 2001, Review: Using optics to measure biological forces and mechanics, Traffic 2:757-763. Kuo, S.c., and Sheetz, M.P., 1993, Force of single kinesin molecules measured with optical tweezers, Science 260:232-234. Kurpius, D., and Dailey, M.E., 2005, Imaging microglia in live brain slices and slice cultures, In: Imaging in Neuroscience and Development (R. Yuste and
401
A. Konnerth, eds.), Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York. Lakowicz, J.R., Szmacinski, H., Nowaczyk, K., Berndt, K.W., and Johnson, M., 1992, Fluorescence lifetime imaging, Anal. Biochem. 202:316-330. Lansford, R., Bearman, G., and Fraser, S.E., 2001, Resolution of multiple green fluorescent protein color variants and dyes using two-photon microscopy and imaging spectroscopy, J. Biomed. Opt. 6:311-318. Lawson, N.D., and Weinstein, B.M., 2002, In vivo imaging of embryonic vascular development using transgenic zebrafish, Dev. Bioi. 248:307318. Legg, J.W., Lewis, c.A., Parsons, M., Ng, T., Isacke, C.M., 2002, A novel PKCregulated mechanism controls CD44 ezrin association and directional cell motility, Nat. Cell BioI. 4(6):399-407. Lepe-Zuniga, J.L., Zigler, J.S. Jr., and Gery, I., 1987, Toxicity of light-exposed Hepes media, J. Immunol. Methods 103:145. Lin, H.J., Herman, P., and Lakowicz, 1.R., 2003, Fluorescence lifetimeresolved pH imaging of living cells, Cytometry 52A:77-89. Lippincott-Schwartz, J., Snapp, E., and Kenworthy, A., 2001, Studying protein dynamics in living cells, Nat. Rev. Mol. Cell. Bioi. 2:444-456. Lippincott-Schwartz, J., Altan-Bonnet, N., and Patterson, G.H., 2003, Review: Photobleaching and photoactivation: Following protein dynamics in living cells, Nat. Cell Bioi. 5:S7-S 14. Loew, L.M., 1993, Confocal microscopy of potentiometric fluorescent dyes, Methods Cell BioI. 38: 195-209. Lorenzi, S., Koedel, U., Dimagl, U., Ruckdeschel, G., and Pfister, H.W., 1993, Imaging of leukocyte-endothelium interaction using in vivo confocal laser scanning microscopy during the early phase of experimental pneumococcal meningitis. J. Infect. Dis. 168:927-933. Lucius, R., Mentlein, R., and Sievers, J., 1998, Riboflavin-mediated axonal degeneration of postnatal retinal ganglion cells in vitro is related to the formation of free radicals, Free Radic. BioI. Med. 24:798-808. Magde, D., Elson, E.L., and Webb, W.W., 1974, Fluorescence correlation spectroscopy. II. An experimental realization, Biopolymers 13:29-61. Manders, E.M.M., Stap, J., Strackee, J., Van Driel, R., and Aten, J.A., 1996, Dynamic behavior of DNA replication domains, Exp. Cell Res. 226:328-335. Manders, E.M.M., Kimura, H., and Cook, P.R., 1999, Direct imaging of DNA in living cells reveals the dynamics of chromosome formation, J. Cell BioI. 144:813-821. Manders, E.M.M., Visser, A.E., Koppen, A., de Leeuw, W.c., van Liere, R., Brakenhoff, G.J., and van Driel, R., 2003, Four-dimensional imaging of chromatin dynamics during the assembly of the interphase nucleus, Chromosome Res. 11:537-547. Marrs, G.S., Green, S.H., and Dailey, M.E., 2001, Rapid formation and remodeling of postsynaptic densities in developing dendrites, Nat. Neurosci. 4:1006-1013. Masters, B.R., 1992, Confocal microscopy of the in-situ crystalline lens, 1. Microsc. 165: 159-167. McKenna, N., and Wang, YL., 1986, Culturing cells on the microscope stage, Methods Cell Bioi. 29: 195-205. Megason, S.G., and Fraser, S.E., 2003, Digitizing life at the level of the cell: High-performance laser-scanning microscopy and image analysis for in toto imaging of development, Meehan. Dev. 120:1407-1420. Mehta, A.D., Jung, J.c., Blusberg, B.A., and Schnitzer, M.J., 2004, Fiber-optic in vivo imaging in the mammalian nervous system, Curro Opin. Neurobiol. 14:617-628. Meyvis, T.K., De Smedt, S.c., Van Oostveldt, P., and Demeester, J., 1999, Fluorescence recovery after photobleaching: A versatile tool for mobility and interaction measurements in pharmaceutical research, Pharm. Res. 16: 1153-1162. Miller, M.J., Wei, S.H., Parker, I., and Cahalan, M.D., 2002, Two-photon imaging of lymphocyte motility and antigen response in intact lymph node, Science 296: 1869-1973. Miller, M.J., Safrina, 0., Parker, I., and Cahalan, M.D., 2004, Imaging the single cell dynamics of CD4+ T cell activation by dendritic cells in lymph nodes, J. Exp. Med. 200:847-856. Minta, A., Kao, J., and Tsien, R., 1989, Fluorescent indicators for cytosolic calcium based on rhodamine and fluorescein chromophores, J. Bioi Chem. 264:8171-8178.
402
Chapter 19 • M.E. Dailey et al.
Miyawaki, A" 2003, Fluorescence imaging of physiological activity in complex systems using GFP-based probes, Curr. Opin. Neurobiol. 13:591-596. Mizrahi, A., Crowley, J.e., Shtoyerman, E., and Katz, L.e., 2004, Highresolution in vivo imaging of hippocampal dendrites and spines, J. Neurosci. 24:3147-3151. Mone, M.J., Bernas, T., Dinant, e., Goedvree, F.A., Manders, E.M.M., Volker, M., Houtsmuller, A.B., Hoeijmakers, J.H.J., Vermeulen, W., and van Driel, R, 2004, In vivo dynamics of chromatin-associated complex formation in mammalian nucleotide excision repair, Proc. Natl. Acad. Sci. USA 101: 15933-15937. Montoya, M.e., Sancho, D., Bonello, G., Collette, P., Langlet, C., He, H.T., Aparicio, P., Alcover, A., Olive, D., and Sanchez-Madrid, F., 2002, Role of ICAM-3 in the initial interaction of T lymphocytes and APCs, Nat. Immunol.3:159-168. Mulligan, S.J., MacVicar, B.A., 2004, Calcium transients in astrocyte endfeet cause cerebrovascular constrictions, Nature 431 (7005): 195-199. Myrdal, S., and Foster, M., 1994, Time-resolved confocal analysis of antibody penetration into living, solid tumor spheroids, Scanning 16:155-167. Nehls, S., Snapp, E.L., Cole, N.B., Zaal, K.J., Kenworthy, A.K., Robert, T.H., Ellenberg, J., Presley, J.F., Siggia, E., and Lippincott-Schwartz, J., 2000, Dynamics and retention of misfolded proteins in native ER membranes, Nat. Cell BioI. 2:288-295. Niell, C.M., Meyer, M.P., and Smith, S.J., 2004, In vivo imaging of synapse formation on a growing dendritic arbor, Nat. Neurosci. 7:254-260. Niggli, E., and Egger, M., 2004, Applications of multi-photon microscopy in cell physiology, Front Biosci. 9:1598-1610. Noctor, S.C., Martinez-Cerdeno, V., Ivic, L., and Kriegstein, A.R., 2004, Cortical neurons arise in symmetric and asymmetric division zones and migrate through specific phases, Nat. Neurosci. 7:136-144. O'Malley, D.M., Kao, Y.H., and Fetcho, J.R., 1996, Imaging the functional organization of zebrafish hindbrain segments during escape behaviors, Neuron 17:1145-1155. O'Rourke, N.A., and Fraser, S.E., 1990, Dynamic changes in optic fiber terminal arbors lead to retinotopic map formation: An in vivo confocal microscopic study, Neuron 5:159-171. O'Rourke, N.A., Dailey, M.E., Smith, S.J., and McConnell, S.K., 1992, Diverse migratory pathways in the developing cerebral cortex, Science 258:299-302. Paddock, S.W., 2002, Confocal imaging of Drosophila embryos, Methods Cell Bioi. 70:361-378. Pagano, R.E., Martin, O.e., Kang, H.e., and Haugland, RP., 1991, A novel fluorescent ceramide analogue for studying membrane traffic in animal cells: Accumulation at the Golgi apparatus results in altered spectral properties of the sphingoid precursor, J. Cell Bioi. 11:1267-1279. Park, M.K., Tepikin, A.Y., and Petersen, O.H., 2002, What can we learn about cell signalling by combining optical imaging and patch clamp techniques? Pjiugers Arch. 444:305-316. Pasti, L., Zonta, M., Pozzan, T., Vicini, S., and Carmignoto, G., 2001, Cytosolic calcium oscillations in astrocytes may regulate exocytotic release of glutamate, J. Neurosci. 21:477--484. Patterson, G.H., and Lippincott-Schwartz, J., 2002, A photoactivatable GFP for selective photolabeling of proteins and cells, Science 297:1873-1877. Patterson, G.H., and Lippincott-Schwartz, J., 2004, Selective photolabeling of proteins using photoactivatable GFP, Methods 32:445--450. Pepperkok, R, Squire, A., Geley, S., and Bastiaens, P.l., 1999, Simultaneous detection of multiple green fluorescent proteins in live cells by fluorescence lifetime imaging microscopy, Curr. Bioi. 9:269-272. Periasamy, A., and Day, RN., 1999, Visualizing protein interactions in living cells using digitized GFP imaging and FRET microscopy, Methods Cell Bioi. 58:293-314. Periasarny, A., Elangovan, M., Elliott, E., and Brautigan, D.L., 2002, Fluorescence lifetime imaging (FUM) of green fluorescent fusion proteins in living cells, Methods Mol. Bioi. 183:89-100. Peter, M., Ameer-Beg, S.M., 2004, Imaging molecular interactions by multiphoton FUM, BioI. Cell. 96(3):231-236. Petersen, M., and Dailey, M.E., 2004, Diverse microglial motility behaviors during clearance of dead cells in hippocampal slices, Glia 46: 195-206. Petran, M., Hadravsky, M., Benes, J., and Boyde, A., 1986, In vivo microscopy using the tandem scanning microscope, Ann. N. Y. Acad. Sci. 483:440--447.
Petroll, W.M., Cavanagh, H.D., Lemp, M.A., Andrews, P.M., and Jester, J.Y., 1992, Digital image acquisition in in vivo confocal microscopy, J. Microsc. 165:61-69. Petroll, W.M., Cavanagh, H.D., and Jester, J.Y., 1993, Three-dimensional imaging of corneal cells using in vivo confocal microscopy, J. Microsc. 170:213-219. Politz, J.C., 1999, Use of caged fluorochromes to track macromolecular movement in living cells, Trends Cell Bioi. 9:284-287. Pologruto, T.A., Yasuda, R, and Svoboda, K., 2004, Monitoring neural activity and [Ca'+] with genetically encoded Ca'+ indicators, J. Neurosci. 24:9572-9579. Poole, e.A., Brookes, N.H., and Clover, G.M., 1993, Keratocyte networks visualized in the living cornea using vital dyes, J. Cell Sci. 106:685692. Potter, S. M., 2004. Two-photon microscopy for 4D imaging of living neurons, In: Imaging in Neuroscience and Development: A Laboratory Manual, 2nd ed., (R Yuste and A. Konnerth, eds.), Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, pp. 8.1-8.12. Potter, S.M., and DeMarse, T.B., 2001, A new approach to neural cell culture for long-term studies, J. Neurosci. Methods 110:17-24. Reits, E.A., and Neefjes, J.J., 2001, From fixed to FRAP: measuring protein mobility and activity in living cells. Nat. Cell Bioi. 3(6):EI45-EI47. Rios, E., Shirokova, N., Kirsch, W.G., Pizarro, G., Stern, M.D., Cheng, H., and Gonzalez, A., 2001, A preferred amplitude of calcium sparks in skeletal muscle, Biophys. J. 80:169-183. Robb, D.L., and Wylie, C., 1999, Confocal microscopy on Xenopus laevis oocytes and embryos, Methods Mol. Bioi. 122:173-183. Rubart, M., 2004, Two-photon microscopy of cells and tissue, Circ. Res. 95: 1154-1166. Salmon, W.e., Adams, M.e., and Waterman-Storer, e.M., 2002, Dual-wavelength fluorescent speckle microscopy reveaJs coupling of microtubule and actin movements in migrating cells, J. Cell Bioi. 158:31-37. Schmidt, e.E., Horwitz, A.F., Lauffenburger, D.A., and Sheetz, M.P., 1993, Integrin-cytoskeleta1 interactions in migrating fibroblasts are dynamic, asymmetric, and regulated, J. Cell Bioi. 123:977-991. Schwarzbauer, J.E., 1997, Cell migration: may the force be with you, Curr. Bioi. 7(5):R292-R294. Sekar, R.B., and Periasamy, A., 2003, Fluorescence resonance energy transfer (FRET) microscopy imaging of live-cell protein localizations, J. Cell Bioi. 160:629-633. Seyfried, Y., Birk, H., Storz, R, and Ulrich, H., 2003, Advances in multispectral confocal imaging, Progress in Biomedical Optics and Imaging 5139:146-157. Sheetz, M.P., ed., 1998, Laser Tweezers in Cell Biology, Academic Press, San Diego, California. Siegel, W.H., and Pritchett, T., 2000, Tutorial: Examining the relationship between media and light, Biopharmaceuticals 13:65-66. Silva, E., Salim-Hanna, M., Edwards, A.M., Becker, M.l., and De Ioannes, A.E., 1991, A light-induced tryptophan-riboflavin binding: biological implications, Adv. Exp. Med. Bioi. 289:33--48. Silva, E., and Godoy, J., 1994, Riboflavin sensitized photooxidation of tyrosine, Int. J. Vitam. Nutr. Res. 64:253-256. Smith, S.J., Cooper, M., and Waxman, A., 1990, Laser microscopy of subcellular structure in living neocortex: Can one see dendritic spines twitch? In: XYIII Symposia Medica Hoechst, Biology of Memory (L. Squire and E. Lindenlaub, eds.), Schattauer, Stuttgart, Germany, pp.49-71. Soli, D., 1995, The use of computers in understanding how animal cells crawl, Int. Rev. Cytol. 163:43-104. Soil, D.R., 1999, Computer-assisted three-dimensional reconstruction and motion analysis of living, crawling cells, Comput. Med. Imaging Graphics 23:3-14. Spierenburg, G.T., Oeriemans, F.T., van Laarhoven, J.P., and de Bruyn, e.H., 1984, Phototoxicity of N-2-hydroxyethylpiperazine-N'-2-ethanesulfonic acid-buffered culture media for human leukemic cell lines, Cancer Res. 44:2253-2254. Stence, N., Waite, M., and Dailey, M.E., 2001, Dynamics of microglial activation: A confocaJ time-lapse analysis in hippocampal slices, Glia 33:256-266.
Confocal Microscopy of Living Cells • Chapter 19
Stoll, S., Delon, J., Brotz, T.M., and Germain, R.N., 2002, Dynamic imaging of T cell-dendritic cell interactions in lymph nodes, Science 296: 18731876. Stoppini, L., Buchs, P'A., and Muller, D., 1991, A simple method for organotypic cultures of nervous tissue, 1. Neurosci. Methods 37: 173-182. Strange, K., and Spring, K.R., 1986, Methods for imaging renal tubule cells, Kidney Int. 30: 192-200. Streit, WJ., and Kreutzberg, G.W., 1987, Lectin binding by resting and reactive microglia, J. Neuroeytol. 16:249-260. Stricker, S.A., 2004, Dual-channel confocal ratioing of calcium dynamics in living eggs and oocytes, Methods Mol. Bioi. 254:137-148. Stryer, L., 1978, Fluorescence energy transfer as a spectroscopic ruler, Annu. Rev. Biochem. 47:819-846. Sullivan, W., Daily, D.R., Fogarty, P., Yook, KJ., and Pimpinelli, S., 1993, Delays in anaphase initiation occur in individual nuclei of the syncytial Drosophila embryo, Mol. Bioi. Cell 4:885-896. Svoboda, K., Denk, W., Kleinfeld, D., and Tank, D., 1997, In vivo dendritic calcium dynamics in neocortical pyramidal neurons, Nature 385:161165. Tauer, U., 2002, Advantages and risks of multiphoton microscopy in physiology, Exp. Physiol. 87:709-714. Terasaki, M., Song, J., Wong, J.R., Weiss, M.J., and Chen, L.B., 1984, Localization of endoplasmic reticulum in living and glutaraldehyde fixed cells with fluorescent dyes, Cell 8:101-108. Thomas, C.F.. and White, J.G., 1998, Four-dimensional imaging: The exploration of space and time, Trends Biotechnol. 16: 175-182. Thompson, N.L., 1991, Fluorescence correlation spectroscopy, In: Topics in Fluorescence Spectroscopy, Vol. I (J. R. Lakowicz, ed.), Plenum Press, New York, pp. 337-378. Trachtenberg, 1.T, Chen, B.E., Knott, G.W., Feng, G., Sanes, J.R., Welker, E., and Svoboda, K., 2002, Long-term in vivo imaging of experiencedependent synaptic plasticity in adult cortex, Nature 420:788-794. Tsurui, H., Nishimura, H., Hattori, S., Hirose, S., Okumura, K., and Shirai, T, 2000, Seven-color fluorescence imaging of tissue samples based on Fourier spectroscopy and singular value decomposition, J. Histochem. Cytochem. 48:653-662. Verschure, P.J., van der Kraan, 1., Manders, E.M., and van Driel, R., 1999, Spatial relationship between transcription sites and chromosome territories, J. Cell Bioi. 147: 13-24. Vesely, P., Maly, J., Cumpelik, J., Pluta, M., and Tuma V., 1982, Improved spatial and temporal resolution in an apparatus for time-lapse, phase contrast cine light micrography of cells in vitro, J. Mierose. 125:6776. Visscher, K., and Brakenhoff, GJ., 1991, Single beam optical trapping integrated in a confocal microscope for biological applications, Cytometry 12:486-491. Walcerz. D.B., and Diller, K.R., 1991, Quantitative light microscopy of combined perfusion and freezing processes, J. Mierose. 161 :297-311, and US Patent 5,257,128.
403
Wang, S.Q., Wei, c., Zhao, G., Brochet, D.X., Shen, J., Song, L.S., Wang, w., Yang, D., and Cheng, H., 2004, Imaging microdomain Ca'+ in muscle cells, Cire. Res. 94:101I-1022. Waterman-Storer, C.M., Sanger, J.W., and Sanger, 1.M., 1993, Dynamics of organelles in the mitotic spindles of living cells: Membrane and microtubule interactions, Cell Mati!. Cytosketelon 26: 19-39. Waterman-Storer, C.M., Desai, A., Bulinski, J.c., and Salmon, E.D., 1998, Fluorescent speckle microscopy, a method to visualize the dynamics of protein assemblies in living cells, Curro Bioi. 8:1227-1230. Weiss, M., 2004, Challenges and artifacts in quantitative photobleaching experiments, Traffic 5:662-671. Wiseman, P.W., and Petersen, N.O., 1999, Image correlation spectroscopy. II. Optimization for ultrasensitive detection of preexisting platelet-derived growth factor-beta receptor oligomers on intact cells, Biophys. J. 76:963-977. Wiseman, P.W.. Squier, J.A., Ellisman, M.H., and Wilson, K.R., 2000, Twophoton image correlation spectroscopy and image cross-correlation spectroscopy, J. Microse. 200: 14-25. Wood, c., Kabat, E.A., Murphy, L.A., and Goldstein, U., 1979, Immunochemical studies of the combining sites of the two isolectins, A4 and B4, isolated from Bandeiraea simplicifolia, Arch. Biochem. Biophys. 198: I-II. Wouters, F.S., Verveer, P.J., and Bastiaens, P.I.H., 2001, Imaging biochemistry inside cells, Trends Cell BioI. II :203-211. Yoder, E.J., and Kleinfeld, D., 2002, Cortical imaging through the intact mouse skull using two-photon excitation laser scanning microscopy, Microse. Res. Teehn. 56:304-305. Zhang, S., Boyd, J., Delaney, K., Murphy, T.H., 2005, Rapid reversible changes in dendritic spine structure in vivo gated by the degree of ischemia, J. Neurosei. 25(22):5333-5338. Zicha, D., Dobbie, I.M., Holt, M.R., Monypenny, J., Soong, D.Y., Gray, c., and Dunn, G.A., 2003, Rapid actin transport during cell protrusion, Science 300:142-145. Zieger, M.A., Glofcheski, D.J., Lepock, J.R., and Kruuv, J., 1991, Factors influencing survival of mammalian cells exposed to hypothermia. Y. Effects of hepes, free radicals, and H,O, under light and dark conditions, Cryobiol0RY 28:8-17. Zigler, J.S. Jr., Lepe-Zuniga, J.L., Vistica, B., and Gery, I., 1985, Analysis of the cytotoxic effects of light-exposed HEPES-containing culture medium, In Vitro Cell Dev. Bioi. 21:282-287. Zimmermann, T., Rietdorf, J., Girod, A., Georget, Y., and Pepperkok, R., 2002, Spectral imaging and linear un-mixing enables improved FRET efficiency with a novel GFP2-YFP FRET pair, FEBS Lett. 531 :245-249. Zimmermann, T, Rietdorf, J., and Pepperkok, R., 2003, Spectral imaging and its applications in live-ceJl microscopy, FEBS Lett. 546:87-92. Zuo, Y., Lubischer, J.L., Kang, H., Tian, L., Mikesh, M., Marks, A., Scofield, Y.L., Maika, S., Newman, c., Krieg, P., and Thompson, W.J., 2004. Fluorescent proteins expressed in mouse transgenic lines mark subsets of glia, neurons, macrophages, and dendritic cells for vital examination, J. Neurosci. 24: 10999-11 009.
20
Aberrations in Confocal and Multi-Photon Fluorescence Microscopy Induced by Refractive Index Mismatch Alexander Egner and Stefan W. Hell
INTRODUCTION Modem optical microscopes are so good that many scientists forget that these instruments only provide their optimal performance if they are used under certain operating conditions. Typical users may be unaware of the very existence of such limitations either because they may unwittingly work within the limits or because they fail to recognize their effects. It is probably also correct to assume that the manufacturer does not intend to discourage purchase by emphasizing the pitfalls that unavoidably arise from the physics of imaging. However, advanced microscopists tend to use their instruments at the limits of their performance. They wish to use lasers from the ultraviolet (UV) to the infrared (IR), special emission lines from very large arc-lamps, charge-coupled device (CCD) cameras with a dynamic range of up to 16 bits, photon-counting avalanche photodiodes with unmatched sensitivity, and several fast photomultiplier tubes. They want to observe two or more dyes simultaneously. They expect the stage to remain in a stable position for several hours, possibly while going through heating and cooling cycles, and sometimes they want to record low-level fluorescence emissions from rather thick specimens mounted in an aqueous medium with high numerical aperture (NA) oil-immersion lenses. Is this possible? Can one expect an off-the-shelf product to perform well under all these circumstances? The answer is: Within certain limits, yes, you can. The issue is to specify and recognize these limits. This chapter describes the problems that occur when observing specimens that are mounted in a medium whose refractive index is different from that of the immersion liquid. Classic examples are live cells kept in a physiological buffer solution or even fixed cells kept in a glycerol-based mountant that are imaged by an oil-immersion lens of large numerical aperture. This chapter first outlines the physics of the situation, both for confocal and multi-photon microscopy, then presents the results of a theoretical investigation, compares them with a series of experiments, and finally draws conclusions that are particularly relevant to the quantitative observation of (living) biological specimens.
THE SITUATION Figure 20.1 describes a common situation encountered in microscopy. The sample is mounted between a coverslip and a glass slide, which in fact can be another coverslip, and is immersed in a special mounting medium, such as an aqueous buffer or a more viscous solution based on glycerol. Coverslip glass has a refractive index (RI) n == 1.518. The immersion oil between the cover-
slip and the objective lens is assumed to have the same n. The n of the mounting medium around the sample will usually be different from that of the glass and of the immersion oil. Water has an index of n == 1.33 and glycerol has n == 1.47. The sample itself will have an n that is not much different from that of the mounting medium and slightly higher than that of water (see Chapter 18, this volume, for the RIs of common mounting media). A light ray emerging from an oil-immersion objective lens that is coupled to the coverslip with the appropriate oil will not be refracted until it passes the interface from the coverslip into the mounting medium. The light ray is usually only slightly affected by the sample itself and is assumed to carry on straight towards the focal region once it has passed the interface between the mounting medium and the coverslip. The discussion can therefore be restricted to the effects caused by the change in n at the glassmedium boundary and to the distance from this interface to the focus point somewhere inside the sample. What effects can be expected? • A light ray is refracted at the glass-medium interface. The angle of the ray is changed; therefore, the different rays focus at different positions along the z-axis than they would in a perfectly matched optical system. In microscopy, nl is usually larger than n2, and the focus is, therefore, closer to the coverslip than under ideal conditions. The position of an object will then appear to be further away from the coverslip. If n l were smaller than n2, the focus would be further from the coverslip than it should be and the object would then appear to be closer to the coverslip. • Whenever light is refracted, some light is also reflected (Born and Wolf, 2002). As refraction occurs only when the angle of incidence is lower than the angle of total internal reflection, the NA of the immersion system is effectively reduced. • Perfect imaging is only possible if the wavefront remains spherical. Any deviation from sphericity results in a larger spread of the focus and hence in a reduction in both spatial resolution and peak intensity. • This spreading of the focus means that the image of the focal spot focused back towards the confocal pinhole is also spread. This second defocus effect means that less light penetrates the pinhole, and the observed intensity decreases still more.
THEORY The calculations are performed in a vectorial theory following Hell and colleagues (1992). The sample object is a layer of fluorophore immersed in the mounting medium. The immersion oil and the
Alexander Egner and Stefan W. Hell· Department of NanoBiophotonics; Max Planck-Institute for Biophysical Chemistry, Am Fassberg 11; D-37077 Goettingen, Germany 404
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
Aberrations in Confocal and Multi-Photon Fluorescence Microscopy Induced by Refractive Index Mismatch • Chapter 20
E~
objective
_ __
(1)
!
AFP
\,a, 1 ,
I
\r
,/
\1/
•
p ~ (x. y, z)
~ z-axis
FIGURE 20.1. Terminology used for the calculation of point spread functions in optically mismatched systems. A spherical wavefront emerges from the objective lens. The wavefront is described by its electric field strength if and its direction k, while the objective lens is described by its focal length and its numerical aperture (NA) = n, sin(a). The sample consists of two coverslips, each with a refractive index of n, and the mounting medium with a homogeneous refractive index of n2. The immersion medium between the upper coverslip and the objective lens has the same n as the coverslip. In a perfectly matched system, n, is equal to n2 and the geometrical focus is a distance NFP (nominal focal position) away from the glass/mounting medium interface. In a mismatched system, n, is not equal to no. The focus suffers from a focal shift and is found at AFP (actual focal position). The theory describes the calculation of the electric field strength if in a point P(x,y,z) close to both AFP and NFP.
coverslip have n = nl and the mounting medium n = n2' Focusing into the sample is achieved by mechanically varying the distance between the objective lens and the bottom of the coverslip. The distance between this surface and the geometrical focus in a perfectly matched system is referred to as the nominal focal position (NFP).I The difference between the NFP and the actual focal position (AFP) is referred to as the focal shift in the optically mismatched system. We wish to calculate the AFP and the intensity at a point P(x,y,z) in the vicinity of the AFP. The optical system is described by the wavelength of the incident light ray (A.), the NA of the objective lens, and the diameter of the aperture in the lens. In a modern optically perfectly matched microscope, the so-called infinity-corrected lens is assumed to accept a perfectly planar incoming-wavefront and produce a perfectly spherical outgoing wavefront that produces an aberration free point spread function (PSF) at the focal point. We note that our considerations apply to any point within the field of view specified for the lens. In a confocal microscope, a point source is used to define the extent and the position of illumination, whereas a point detector discriminates against any light emitted outside a certain region. In physical terms, the effective PSF (Hell et aI., 1992) of the confocal fluorescence microscope is given by the product of the illumination intensity and detection PSF: , It should be noticed that the NFP is the actual distance of a feature in the
object from the surface of the coverslip.
405
where hill denotes the amplitude of the illumination light in the focal region and hdet is the amplitude distribution for the detection, which is similar to hill, but is calculated for the wavelength of fluorescence emission. So, while hill is proportional to the light field used for illumination, and Ihii to its intensity, the effective confocal PSF h,! is proportional to the probability that a given focal coordinate contributes to the signal at the confocal detector. If the excitation and emission wavelengths are rather similar, it follows that Ih ill12 '" Ihdet l2 • In this case, h,jis proportional to the fourth power of the illumination amplitude and hence to the square of the illumination intensity (Wilson and Sheppard, 1984), as indicated on the right-hand side of Eq. I. This quasi-quadratic signal dependence of the recorded intensity on the illumination intensity causes a drop of the detected fluorescence from points away from the geometrical focus and is the actual physical reason why a confocal setup defines a confined recording volume in three-dimensional (3D) space. The z-response fez) to infinitely thin fluorescent planes and the response fedge(z) to half-volumes in the z-direction are of practical importance as well because they quantify the ability of a microscope to distinguish planes that are stacked in the z-direction: z
fez)
=
f f hcf(r)dx dy and fedge(z) = f f(z')dz'
(2)
A volume confinement can also be achieved by multi-photon excitation of the fluorophores (Denk et aI., 1990; Sheppard and Gu, 1992a, 1992b) (Chapters 28 and 37, this volume). The probability of two photons being simultaneously absorbed by the same molecule is proportional to the square of the local intensity (Kaiser and Garret, 1961). Therefore, the effective PSF of a microscope based on two-photon absorption is described by the square of the illumination PSF, that is, the fourth power of the field amplitude for illumination:
(3) The similarity to Eq. I indicates that the illumination process defines a volume in a manner similar to the combined illumination and detection processes in a confocal single-photon excitation fluorescence microscope (Hell and Stelzer, 1992; Sheppard and Gu, 1992b; Stelzer et aI., 1994). This means that all the problems discussed for single-photon confocal microscopy are also encountered in two-photon fluorescence microscopes. The only real difference is that the latter requires wavelength doubling. Additional confocalization of the system means that the effective PSF is given by h rj2h " = Ihilll4 X Ihdet l2 and for an n-photon confocalized system we obviously have hcf2hv
- 12n x 1-hctct 12 • = 1hill
(4)
In an aberration-free system, the calculation of the fields hill and hdet is rather straightforward because these functions solely depend on the wavelength and the NA. By contrast, when focusing through RI interfaces, the evaluation of hill and hctct is complicated by the fact that, loosely speaking, it has to be calculated once for medium nl and then for medium n2 (Hell et aI., 1993). The latter publication treated the problem with specific regard to confocal microscopy and quantitatively predicted all the effects
406
Chapter 20 • A. Egner and S.W. Hell
encountered with refractive index mismatched samples . Therefore, in this chapter we will follow the argument presented in that publication. In the meantime, significant advancements in the formulation of this theory have been made. These are considered later in this chapter (Torok et ai., 1995; Egner and Hell, 1999). Starting from simple terms, the calculation basically requires the solution of a variational problem in optics, that is, the application of Fermat's principle from a point of the converging spherical wavefront to the point of the focal region in question [Fig. 20.2(A»). According to the Huygens-Fresnel construction, each point on the spherical wavefront is a source of secondary spherical wavelets (Hopkins, 1943; Li and Wolf, 1981). In a matched medium, we have
her) = cf f A(F).! K(x)exp(iks)dF F
(5)
s
where A(F) denotes the wavefront amplitude over the surface F of the spherical wavefront and dF is the surface element (see Fig. 20.2); s is the distance between the origin of the wavelet q and the point P, and X is the angle of inclination between the normal at q and the direction from q to P.
1 K(X) = - 2A (1 + cos(X»
2
a2n
h(F)=cf f 00
COS(S) + (1- cos(S»sinZ( q» A(F) = Ai.1cOS(S) (1- cos(S»cos(q> )sin( q» [ sin(S)sin(q»
1
exp(iks)sin(S)dq> dS
(8)
The final problem is the determination of s, which, as is pointed out above, is a variational problem and a careful analysis of the light transition at the interface between n, and nz, to which one solution has been provided (Hell et ai., 1993). If the NFP is small compared to the focal length of the objective lens, an assumption which is well met in confocal microscopy, the variational problem can be solved analytically for the region near the NFP (Egner and Hell, 1999). Torok and colleagues derived an identical solution (Torok et ai., 1995) using a plane wave expansion of the light field, that is, the Debye approximation instead of the Huygens-Fresnel construction. In both cases the solution for h(F) is given by:
(9)
(6)
In an aplanatic objective lens, the wavefront A(F) is given for in the x-direction polarized light of Amplitude Ai by Richards and Wolf (1959):
[COS(S)+(1-COS(S»Sin (q» .Jcos(S) (l-cos(S»cos(q»sin(q» sin(S)sin(q»
Where the diffraction integrals In are defined by:
10 := 1.Jcos(S1)Sin(S1)('t, +'tpcos(S2))lo(kn1.1x 2 + y 2 sin(S,))
1
o
exp(ik(ct>(NFP) + n2zcos(Sl)))dS,
(7)
For the numerical calculation of her), the term K(X) can be neglected because it varies by only a small amount. Because the calculation is restricted to the volume close to the geometrical focu s, 1/s remains a constant and can therefore also be neglected. The function her) can be simplified to
a
f
/ 1:= .Jcos(S,) sin(S1)( 't p sin(S2))1, (kn,.1 Xl + yl sin(S,)) o
ex p(ik(ct>(NFP) + n2Z COS(S2)))dS, 12 : =
1,.1 COS(S1) sin(S,)( 't
s -
't p COS(S2) )12(kn1 .J x 2 + y2 sin(S, ))
o
exp(ik(ct>(NFP) + n2 zcos(SJ))dS I
(10)
B
A
wavefront
exit aperture
interface x
n, z-axis
E
FIGURE 20.2. (A) The variational approach to the calculation of the electric field strength at a point P(x,y,z) close to both the AFP and the NFP. The geometry is basically the same as described in Figure 20.1. The problem is to minimize the distance s,(x) + s,(x) . It can be shown that this sum depends on the angle and on the origin of the beam in the primary spherical wavefront. The calculation, therefore, has to search for those beams that contribute to the field at position P according to Fermat's principle. For each point P(x ,y,z) in the object, the contribution is found by integrating the complex electric fie ld across the whole exit aperture. This makes the calculations somewhat tedious. (B) If the NFP is small compared to the focal length of the objective lens, which is always the case in microscopy, the calculation becomes much easier as only the transmission and refraction for a beam incident upon the boundary with an angle 9, has to be known.
Aberrations in Confocal and Multi-Photon Fluorescence Microscopy Induced by Refractive Index Mismatch • Chapter 20
I n are Bessel functions of the first kind and nth order and "Cp,s are Fresnel transmission coefficients for s- and p-polarized light (Born and Wolf, 2002), The aberration function CP(NFP)
=
-NFP(nJ
cos(SJ) -
n2
COS(S2»
(11)
depends on the nominal focusing position, the azimuth angle Sand therefore on the aperture angle a and the difference of the refractive indices nJ and n2' The differences between the calculations of the illumination and the detection PSFs are the wavelength and the term -v'cos(S) which is omitted,
RESULTS OF THEORETICAL CALCULATIONS The theory described above does not result in an analytical description of the PSE The PSFs have to be calculated numerically as a function of the NA, excitation and emission wavelengths, NFP, nJ, and n2' In order to illustrate how focusing into a mismatched medium affects the PSF, Figure 20,3 shows xz-images of calculated PSFs (logarithmic scale) and the corresponding experimental through-focus series (linear scale) for focusing with a water-immersion lens either lO!lm deep into water [Fig, 20.3(A,B)] or into immersion oil [Fig, 20.3(C,D)], In the mismatched case, the main maximum is shifted, becomes relatively broader, and drops in peak intensity (Fig, 20A), In addition, the PSF loses its axial symmetry with respect to the main maximum whereby focusing above the AFP leads to a different image from focusing below the AFP [Fig, 20,3(D)], an effect that is not present in the matched case [Fig, 20,3(B)],
407
The results of several calculations are summarized in Figures 20A, 20,6, 20,7, 20,8, 20,9, and 20,10 and in Tables 20,1 and 20,2 for water and for glycerol mounting media, Figure 20A shows the integrated intensity for various NFPs using water as the mounting medium, The first image indicates the ideal situation encountered with a fluorophore mounted in immersion oiL The following images show again that the integrated intensity is smeared along the optical axis, and an additional peak appears below the main maximum, The main maximum itself is shifted, drops in peak intensity, and becomes relatively broader, These values can be evaluated to obtain the focal shift [(NFP) - (AFP)], the drop in peak intensity, and the full-width half-maximum (FWHM) of the main peak, While, because of its convoluted shape (Fig, 20,5), it is difficult to specify a simple metric that quantifies the sharpness of a spherically aberrated focal spot, the FWHM of the main peak is relatively simple to measure and hence has become the most common measure of the xy or axial resolution, Calculations for some numbers encountered in real situations have been combined with the experimental results in Figures 20,6, 20,7, and 20,8, Figure 20,9 shows the focal shift of the excitation PSF of a 0,6 oil-immersion objective lens directly as a function of NFP for various refractive indices, Figure 20, I a also plots the focal shift but for an NA I A oil-immersion objective lens, The AFP is regarded as the position of the global maximum of the PSF along the optical axis rather than the center of gravity or some other measure of this complex shape, An important result of these calculations is that the effects of spherical aberration increase rapidly with (n! - n2), NA, and distance of the object from the coverslip (NFP), As long as the
water ... water A
NFP
water ",oil C
NFP
FIGURE 20.3. Influence of a mismatched medium on the PSE (A) and (C) show xz-images of calculated PSFs in logarithmic scale for focusing with a waterimmersion lens (n = 1.334) 10Ilm deep into water and immersion oil (n = 1.518), respectively. The inlets in the lower left corners show the central part of the PSFs in linear scale, (8) and (D) show corresponding experimental through-focus series in a linear scale, These series represent exactly what one would see when focusing through a point-like object. Focusing into a mismatched medium causes a shift and a broadening of the main maximum. The aberrated PSF can be clearly identified in the through-focus series as focusing above the AFP leads to a different image from focusing below the AFP. The image series was made of a mirror specimen by], Pawley using a Zeiss Axioskpp 50, with a 40xINAI.2 C-Apo objective, and recorded with a Sony TRV-900 camcorder using the zoom lens to provide the high magnification.
Chapter 20
408
.
A. Egner and S. W. Hell
A 100
B
80 ~ 60 <Jl c OJ 40
C
100
1\
100
80
80
>60 c
'iii 60
Zc
''iii
OJ
240
E
E
.£
20
20
0 -3
0 -2
-1
0
1
0
-4
3
-3
NFP[flm]
-2
-1
0
5
2
D
E
80
E 60 c
~ c
80
'Ec 60
20
20
240
.£
20
~
0 -2
-1
-7
0
NFP[flm]
1
F
60
~ 40
2
100
80
~ 40
-3
3
NFP[flm]
100
-4
-4
NFP[flm]
100
0 -5
40 20
-6
-5
-4
-3
NFP[flm]
-2 -1
0 7
-6
5
-4
3
2
1
FIGURE 20.4. Integrated z-responses for various penetration depths (NFP) in water for an NA = 1.3 oil-immersion objective lens. The excitation and emission wavelengths were 514nm and 590nm, respectively. (A) Ideal situation in oil, penetration depths of (B) 5 11m, (e) 10 11m, (D) 15 11m, (E) 20 11m, and (F) 25 11m. All curves are normalized to the ideal situation encountered with immersion oil as the mounting medium. The point spread function is obviously not confined to the minimal volume but instead continues to spread the larger NFP becomes. This causes a decrease of the maximal intensity, an increase of the full-width half-maximum, and a focal shift. The intensity is distributed among several axial peaks of which at least two are clearly visible. Please note that the NFP axis has been offset.
NFP[flm]
I
n,=1,518
coverslips
n2 fluorophore in mounting medium
A
n, = 1,518
J
I -
B
FIGURE 20.5. Sample used for measuring the edge response in a confocal laser-scanning microscope. (A) The sample consists of two coverslips or a coverslip and a glass slide with a fluorophore dissolved in the mounting medium (water, glycerol, immersion oil) in between. The sample thickness (upperllower interface distance) is at most IOOllm. In this sample, the concentration of the fluorophore is zero inside the glass and abruptly reaches a high concentration when moving the probe along the optical axis. (B) During the experiment, the point spread function penetrates the mounting medium through a coverslip. The measured fluorescence intensity signal depends on the penetration depth. The fluorescence emission is maximal when the point spread function is completely inside the sample. Because the point spread function has a finite size, it has a response curve with a finite steepness. The slopes close to the interfaces can, therefore, be used to determine the extent of the point spread function along the optical axis, and this is the axial resolution of the instrument.
Aberrations in Confocal and Multi-Photon Fluorescence Microscopy Induced by Refractive Index Mismatch • Chapter 20
TABLE 20.1. Result of Calculations for Glycerola,b
NFP (Mm)
TABLE 20.2. Result of Calculations for Watera
Axial Edge
Axial PSF
Lateral PSF
Focal Shift (Mm)
Normalized Intensity
FWHM (Mm)
FWHM (Mm)
FWHM (Mm)
0 -0.28 -0.55 -0.83 -1.10 -1.33 -1.54 -2.30
100 95 91 78 62 50 40 31
0.53 0.53 0.53 0.555 0.65 0.81 0.97 1.00
0.47 0.47 0.47 0.47 0.50 0.57 0.77 0.72
0.16 0.16 0.16 0.16 0.18 0.18 0.20 0.20
0 5 10 15 20 25 30 50
409
NFP (Mm)
Axial Edge
Axial PSF
Lateral PSF
Focal Shift (Mm)
Normalized Intensity
FWHM (Mm)
FWHM (Mm)
FWHM (Mm)
0 -1.0 -1.83 -2.57 -3.30 -4.02 -4.72 -7.57
100 60 39 28.50 23 19 16.60 11
0.53 0.68 0.90 1.08 1.22 1.37 1.45 1.79
0.47 0.58 0.70 0.88 0.93 1.00 1.11 1.37
0.16 0.20 0.20 0.23 0.24 0.24 0.24 0.24
0 5 10 15 20 25 30 50
For various nominal focal positions (NFP), the focal shift, normalized intensity, axial edge response, axial width of the PSF, and lateral width of the PSF have been calculated for an NA = 1.3 oil-immersion objective lens and excitation and emission wavelengths of 5l4nm and 590nm, respectively. The values for an NFP = 0 are the ideal values if immersion oil is used as the mounting medium.
"For various nominal focal positions (NFP), the focal shift, the normalized intensity, the axial edge response, the axial width of the PSF, and the lateral width of the PSF have been calculated for an NA = 1.3 oil-immersion objective lens and excitation and emission wavelengths of 514 nm and 590 nm, respectively. hThe values for an NFP= 0 are the ideal values if immersion oil is used as the mounting medium.
a
spherical aberrations are below a certain threshold, the focal shift depends linearly on the NFP which can be used to correct the apparent thickness of a sample. If the spherical aberrations exceed the threshold, the relation between the NFP and the focal shift becomes nonlinear which leads to image distortions along the zaxis and a rapid drop in intensity as the excitation and emission PSF will be substantially displaced. This effect is most prominent with high NAs and very low for oil-immersion lenses having an NA of ill
2
'\j
The overall behavior of the detector for a given number of incident photons depends on both the quantum efficiency and the sensor noise: for a large number of photons, SIN is proportional to &:, whereas, for a very small number of photons, it is proportional to the ratio QElnn • We can consider the information content, b, in the image, in bits, given by Shannon (1949)
(5) In terms of the number of discernible gray levels g, we can write
b = log2(g),
(6)
so that the number of gray levels is simply
g=l+S/N=l+
Qn E
P
.jQEnl' +n~ .
(7)
Thus, if SIN is zero, we can only perceive a single gray level, that is, the image is featureless. For SIN = 1, we see two gray levels, corresponding to a binary (black and white) image. For large values of SIN, the number of gray levels increases linearly with SIN. Note that this derivation of the number of gray levels is based on the information content in the image, which is an objective property. Although it does require that the specimen be capable of producing the measured signal levels (i.e., having a certain voxel staining ratio), it does not rely on the subjective perception of an observer, as do some investigations of the number of discernible gray levels. The disadvantage of the latter approach is that the number of perceived gray levels is altered by changing the properties of the display (e.g., by altering the lookup table), so that the results are no longer a property of the image data alone. In Figure 22.1, the gray-level behavior of a photomultiplier tube (PMT, S20 photocathode, QE = 13% and nn = 0), predicted by Eq. 7, is compared with that of a cooled frame-transfer chargecoupled device (CCD) detector (QE = 70%, n" = 9 RMS electrons/pixel). The sensor noise of the PMT has been neglected because of its very high gain.! The behavior of an ideal detector (QE = 100%, nn = 0) is also shown for comparison. For all detectors the number of gray levels increases from unity with the number of photons incident on each pixel. The behavior of the PMT is limited by the poor quantum efficiency. It is clear that for greater than about 25 photons incident per pixel the CCD exhibits superior performance, whereas for a smaller number of photons the photomultiplier tube is better. In the region where the photomultiplier is superior, the number of gray levels is less than 1 + ..)13 % x 25, or about three, so that the image is of poor quality.
I
Actually. the PMT has considerable multiplicative noise but, in this discussion, this noise has been accounted for by assuming a fairly low QE.
20
40
80 60 photons/pixel
100
FIGURE 22.1. The number of gray levels discernible in an image formed using various detectors under different illumination conditions.
Background Noise Next, we can incorporate noise emanating from the background (stray light) in the measured signal. A number of different models for stray light have been considered. Here we consider two different models. In the first, which we shall term Nl, the stray light is assumed constant in intensity over the detector plane. In fluorescence this would be an appropriate model for noise that originates from autofluorescence in the cements and coatings of the optical elements of the microscope system, particularly the objective lens, as in single molecule fluorescence studies. It is also a good model for BSL imaging into a scattering medium. This background n,J3Nl can be assumed to be distributed uniformly over the detector pinhole, so that the intensity that passes through the pinhole is proportional to its area and can be written (8)
where a is a constant, discussed later, and detector in normalized coordinates
Vd
is the radius of the
(9) where rd is the true radius of the pinhole and sin ad is the numerical aperture (N A) of the system in detector space (= n sin aiM, where n sin a is the NA of the objective in object space, and M is the lateral magnification from object plane to detector). A value of Vd = 3.83 corresponds to the first zero of the Airy disk, that is, to 1 Airy unit. For simplicity, the A of the fluorescent radiation is assumed equal to that of the incident radiation throughout this chapter, that is, there is no Stokes shift. As the increase in A for common dyes, such as fluoroscein isothiocyanate fluorescein, rhodamine, Texas Red, or allophycocyanine, is usually in the range 3% to 8%, its effect of the behavior of the system is unlikely to be very significant. Model Nl has been found useful (Sheppard, 1991) for describing the behavior of confocal BSL microscopy, in which case the background can come from specular reflection from the optical elements and light scattered from within the microscope body. Of course, in fluorescence microscopy, both these contributions are substantially removed by the barrier filters. Wells and colleagues (1990) showed in their experiments on fluorescence microscopy that N I was not a good model. This may be an indication that autofluorescence was the primary cause of the background. Webb and colleagues (1990, pp. 73-108) have considered a further model in which the signal was considered to originate from within the resolution element while detected light not from this source is considered as background. This model has the disadvantages that the
444
Chapter 22 • C.J.R. Sheppard et al.
boundary between signal and background is arbitrary and that the size of the resolution element varies with the parameters of the microscope. It also assumes an infinite field together with a finite object thickness (infinite field with an infinite object thickness gives infinite background for a WF microscope). We choose to assume a finite field and an infinite object thickness in order to stress the fact that reduction of the field size is an important factor in improving noise performance. Our second model (N2) assumes that the background originates uniformly from everywhere within an infinitely large fluorescent object. This model would be appropriate for the case of observing fluorescence labeling in the presence of background object autofluorescence. In this case, the intensity that passes through the pinhole n,}3N2(v) is a function of the pinhole size, as will be described in the next section. Both these background contributions are detected by the photodetector and the statistical variations of this value give rise to noise in the image. Thus, the overall SIN is given by
S/N=
Q£npF(vd) ~Q£np[F(vd)+avU4+BN2(vd)]+n~ ,
6
8
10
FIGURE 22.2. The signal level from a point object as a function of detector aperture shape and size.
(10)
where F(Vd) is the fraction of signal light incident on the pinhole that passes through to the detector.
SIGNAL LEVEL IN CONFOCAL MICROSCOPES Consider a single point object situated at the focal point of the microscope. For a confocal system with a point source, the intensity in the detector plane is an Airy disk (11)
where In(v) is a Bessel function of order n and the factor 1I4n; normalizes the total intensity in the detector plane to unity. The normalized signal for a circular detector radius Vd is (Cox and Sheppard, 1986)
These functions are illustrated in Figure 22.2. It is seen that in each case the normalized signal rises from zero until it approaches unity for large pinhole diameters. For a circular pinhole the normalized signal reaches about 85% at Vd = 4, and thereafter there is little increase in signal level. In practice, if the pinhole diameter is increased further the signal collected from a thick object increases as the optical sectioning effect is weaker. However, this increased signal is not useful as it represents light collected from out-offocus regions of the specimen. As is apparent from Figure 22.2, slit apertures can give increased signal strength, but this is associated with poorer optical sectioning performance compared with that from a circular aperture. Figure 22.3 shows the effect of pinhole or slit width on various imaging properties of a confocal fluorescence microscope. The transverse resolution is measured as the reciprocal of the full-width at half-maximum (FWHM) of the image of a point object, normalized by that for a point detector. There is a small improvement in transverse resolution for small
(12)
Circular aperture Slit aperture
For small pinhole radii, we have
F;(v d ),,"vJj4.
(13)
For a slit detector, width 2Vd, placed in front of the detector (i.e., an incoherent slit detector) and a point source, the normalized signal we denote F 2 (Vd)' Then F 2 (Vd) is
F2 (V d ) =
~ n;
1"1 0
(;v) dv v
(14)
""
16vd/3n;2,
(15)
so that the signal is proportional to the width of the slit. For a square detector, side 2Vd, as is the case for a single CCD element, the signal can be evaluated by integrating over the Airy disk. As the signal for a small detector is proportional to the area of the detector, for small detectors we have (16)
__ - - - - - - - - Signal (plane)
0
~~ O:!::::: C/l
~
N=cO ro c
E
Z
0.8
C/l
C
OJ "Oc OJ .-
0
where HI is a Struve function. For long slits, the signal from a point object is independent of the length of the slit. For small Vd the normalized signal is (Sheppard and Mao, 1988) F2(V d )
c
0.6 0.4
.~ C/l
0
0.2 0
~~~2~__~~4____~6____~8~__~10 ~
1
2
Diameter or width (Airy units) FIGURE 22.3. The effect of pinhole or slit size on various imaging properties of a confocal fluorescence microscope. The transverse resolution for a pinhole aperture is defined as the reciprocal of the full-width at half-maximum (FWHM) of the image of a point object. normalized by that for a point detector. The axial resolution, for pinhole or slit apertures, is defined as the reciprocal of the FWHM of the image of a thin fluorescent planar object, normalized by the FWHM of a planar object in a confocal BSL microscope with point detector. The signal level from a planar fluorescent object is also shown.
Signal-to-Noise Ratio in Confocal Microscopes • Chapter 22
pinhole sizes, but the pinhole radius,
Vd
has to be very small
1
(Vd
.01
< 2) for this to be appreciable. The behavior is identical for fluorescence or BSL systems. The axial resolution is measured as the reciprocal of the FWHM of the image of a thin fluorescent planar object, normalized by the FWHM of a planar object in a confocal BSL microscope with point detector. The axial resolution steadily degrades as the pinhole diameter is increased, until there is no z-resolution for a large detector, equivalent to a WF microscope. Even for a point detector, the axial resolution is poorer than for a confocal BSL microscope. The signal level from a planar fluorescent object is also shown. The behavior is qualitatively similar to that for a point object, as shown in Figure 22.2. As the pinhole radius is increased, the signal level increases, but the resolution also decreases, so that there is an optimum behavior for some intermediate pinhole radius. Turning now to the slit aperture, although the signal level is greater than for a circular pinhole of the same width, the z-resolution is worse. Overall, there is little to be gained from using a slit aperture from the point of view of image quality. Experimental results (Sheppard et al., 1991) have confirmed that the theoretically predicted behavior is observed in practice for pinhole radii up to a value Vd of about 100. For very large pinholes, optical behavior will eventually be limited by other apertures present in the optical system.
SIGNAl-TO-NOISE RATIO FOR CONFOCAL MICROSCOPES
QE, N1, and Stain level Signal-to-noise ratios for a confocal microscope can be calculated from Eq. 10. In the following we shall assume for simplicity that the sensor noise nn is zero so that
In the rest of this chapter SIN is normalized by the constant factor ~QEnp. Let us consider first the case of noise model Nl so that BN2 is taken as zero, and after normalization (Sheppard et at., 1991)
SIN =
F(v d ) ~F(vd)+avJ/.,f4
(18)
For a circular pinhole and small values of Vd, using Eq. 13 we have
~N=
~
2"'-"1+a'
(rn
so that a represents the magnitude of the background from noise model Nl relative to the signal for a small circular pinhole. Thus, a is smaller for stronger staining. The behavior of SIN with pinhole radius for different strengths of background is shown in Figure 22.4. For a particular value of a, SIN increases as the pinhole radius increases, reaches a maximum value for an optimum pinhole size, and then decreases as l/Vd as the pinhole is made larger. For a wide range of values of a, the optimum pinhole radius corresponds to values of Vd between about 2 and 3. Simply, a larger pinhole lets in more background but does not produce a significantly stronger signal. Interestingly, as the strength of the signal relative to the background decreases (i.e., a increases), the optimum pinhole radius decreases so that, contrary to expected behavior, the pinhole should actually be closed down for very weak signals. For a value of a of unity, for example, the optimum pinhole radius has decreased to Vd = 2.3. For SIN to be better than
445
SIN
o
2 Diameter (Airy units)
FIGURE 22.4. Variation in SIN with pinhole size for different strengths of background (noise model Nl: background from autofluorescence from optical components).
50% of that achieved at the optimum pinhole radius, Vd must be kept in the range between 0.8 and 5.6. Using a larger pinhole radius further decreases SIN. Thus, using a pinhole radius that is too large, as well as not giving an increase in useful signal, actually results in a decrease in SIN from weakly fluorescent objects. On the other hand, it should be noted that for small values of a the maximum in SIN is broad, so that choice of an appropriate pinhole diameter is not critical from the point of view of SIN. Thus, if the background is weak, the pinhole can be opened up to increase the number of detected photons, and hence SIN. Although one would expect this to reduce the z-resolution, it is very common for the z-resolution to be limited more stringently by low signal level (and the resulting inability to measure pixel brightness accurately) than by purely optical considerations. In this case, the extra signal may enable one to make a better estimate of the true structure of the dye in the specimen in spite of the large PSF. For instance, if we are imaging a small, bright, spherical object say 20 voxels in diameter, surrounded by a low intensity background (-1 % of the peak), the signal obtained from a voxel in the center of this object will be substantially greater and the uncertainty of the measurement will be less if the pinhole is opened from I to perhaps 2 or 3 Airy units. Then the difference between this signal and that from a nearby background voxellocated above the object will be greater but the absolute value of this background measurement will also be more because the reduced z-resolution means that the sensed volume is larger and contains more dye. Although one can retrieve some of this lost resolution if one dec onvolves the data recorded, using a PSF measured at the same pinhole size, this will be decreasingly effective as the pinhole gets bigger. Specifically, the SIN of the imaging system will begin to decrease as the size of the detection PSF exceeds the size of the object. From this discussion, one can see that the optimal pinhole size becomes highly dependent on the details of the 3D stain distribution in the specimen (both signal and background), and that deconvolving the recorded data will always help. However, even though one can calculate the SIN for some limiting conditions (as is done below) in the end, one must often determine the optimal pinhole size for a given type of specimen experimentally. For large values of a, corresponding to weak: signals, SIN tends to a behavior independent of the value of a. In this case, again renormalizing SIN, we obtain for a circular pinhole, a slit detector, and a square detector
446
Chapter 22 • C.J.R. Sheppard et al.
I(v)
+_--:.2=--_.,.::4~---,r:6~---....!o:8~---..!.!:10 V 2
~____T2____-r4____-r6____-r8____~10 ~
1
A
2
Airy units
Diameter (Airy units)
FIGURE 22.6. The variation in the intensity of the fluorescence arriving at the plane of the detector when a thick featureless volume is illuminated by an axial focused laser beam. The horizontal axis is calibrated in Airy units, with zero being the optical axis.
ing over the detector aperture. For a circular pinhole radius vd we obtain 0.1 ~____.2____-.4 ____-.~6____~8____~10
1
B
Vd
2
Diameter (Airy units)
FIGURE 22.5. Variation in SIN with aperture shape and size for weak signal (noise model Nl): (A) circular or square detector; (B) slit detector with length 2v, = 2nvd'
Figure 22.7 shows the variation in background with pinhole radius for a circular pinhole. For small pinhole sizes the background is vJ/4 (as in Eq. 13), that is, proportional to the area of the pinhole, but for large pinholes it is proportional to Vd' This is because for large Vd the thickness of the optical section is proportional to Vd' This is an idealized model: the situation is more complex when the background is not featureless but contains structures in out-offocus planes. We have for the normalized SIN for weak signals
(20)
(23)
respectively, where 2vs is the length of the slit. These are shown in Figure 22.5. The peak value for SIN is slightly higher for a circular, rather than a square, aperture. For the circular pinhole the peak SIN occurs at about 0.52 Airy units (Vd"" 2). For the slit aperture the optimum SIN occurs at 0.48 Airy units (Vd = 1.84). The SIN is decreased relative to the confocal case with a circular detector by a factor of about .,;n12, where n = V/Vd'
We have termed this property the detectability (Gan and Sheppard, 1993) of the system: it describes the ability with which a weakly fluorescent point object can be detected in a background of a uniformly fluorescent background. Actually, other forms of detectability have also been defined (Gan and Sheppard, 1993). These describe the ease with which a point object can be detected within a background uniform over a plane [the two-dimensional (2D) detectability], and the ease with which a planar object can be detected within a background volume (the axial detectability). Here we shall only discuss the 3D detectability, appropriate for detecting a point in a background volume.
N2 and Detectability Turning now to the second noise model, N2, the background EN2 resulting from a featureless volume can also be calculated for these various systems (Gu and Sheppard, 1991). The intensity in the detector region produced by a volume object is given by the convolution of the intensity point spread function of the illumination lens with that of the collection lens. Assuming these are identical, this is conveniently calculated as the 3D Fourier transform of the square of the optical transfer function (OTF). Then the normalized intensity in the plane of the detector aperture is
1 2 l(v) = - - (2 -1)\ 4 + l)Jo(vl)dl.
12
f
6
5 BN2
4
3 2
(21)
0
This is illustrated in Figure 22.6. The intensity decays monotonically with distance from the axis, for large distances decaying as lIv3 • The background detected can then be calculated by integrat-
2 Diameter (Airy units) FIGURE 22.7. The background from a thick featureless volume as a function of pinhole radius.
Signal-to-Noise Ratio in Confocal Microscopes • Chapter 22
447
SIN 0.25
SIN
0.2
0.25
0.15 0.2
0.1
0.15
0.05
2
4
6
8
10
o
3
4
B
A
FIGURE 22.S. Variation in SIN with aperture size for (A) circular and (B) slit apertures, for weak signal (noise model N2: background from autofluorescence from specimen bulk).
The SIN for N2 for a confocal microscope with a circular or slit detector is shown in Figure 22.8. The behavior is broadly the same as for the case of Nl in that SIN reaches a maximum value for an optimum aperture dimension, which is Vd "" 2.4 for the circular pinhole, and vd "" 1 for the slit aperture. For the case of noise model N2, for large circular pinholes SIN decays as 1/.JV:.
for two-photon and three-photon fluorescence, respectively. These results are quantitatively in agreement with experimental observations (Wang and Fraser, 1997).
Multi-Photon Fluorescence Microscopy
The basic form of the confocal microscope uses illumination from a point source, coupled with a photodetector spatially limited by a pinhole. Scanning can be achieved by either moving the specimen stage relative to a stationary light spot, or by scanning the beam. The most common method of beam scanning uses galvomirrors, which, as is also the case for stage scanning, is limited to comparatively slow scan rates, up to a few frames per second. Resonant scanners can achieve faster scan rates, and acousto-optic deflection is even faster but results have other problems for fluorescence imaging, notably the fact that the fluorescent signal cannot be descanned in a simple manner because of the Stokes shift. A solution to this problem is to use a slit-shaped detector aperture and to descan only in one direction (Draaijer and Houpt, 1988). An alternative approach for increasing scan speed is to scan simultaneously with an array of points. The earliest example of this method is the tandem-scanning microscope, in which a rotating disk with an array of holes is used (Petnii'i et al., 1968). It is also possible to use an array of holes, a slit, or an array of slits. In the line-illumination microscope, such as the Zeiss 5 Live, a line of the specimen is illuminated (Koester, 1980; Sheppard and Mao, 1988; Benedetti et al., 1992; Brakenhoff and Visscher, 1992) and the image of the whole line can be recorded using a linear CCD array. Figure 22.10 shows schematic diagrams for three different systems using line-illumination and/or linear CCD detectors. In Figure 22.1 O(A) a scanning spot is deflected over the specimen and partially descanned onto a CCD detector. In Figure 22.1 O(B) the scanning spot is again partially descanned, but on to a single point detector with a slit aperture. In Figure 22.1 O(C), the object is illuminated by a line of light, and the image along the line recorded by a CCD detector. By including a coupled display y-scan, the 2D image can be observed directly or recorded using a 2D CCD detector. These various systems have been compared by Awamura and Ode (1992), in which they claim that arrangement (A) is superior
The SIN can be calculated in a similar way for multi-photon fluorescence microscopy (Gauderon and Sheppard, 1999). The SIN for model N2 for a multi-photon microscope with a circular detector is shown in Figure 22.9. In this case, for large pinholes the SIN does not tend to zero as in the single-photon fluorescence microscope, so the SIN can be normalized to unity for large pinhole diameters (i.e., non-descanned detection). It is found that the SIN exhibits an optimum value for a pinhole radius of 0.63 Airy units (Vd = 2.42) for two-photon fluorescence, and 0.64 Airy units (Vd = 2.44) for three-photon fluorescence. The value of SIN is then a factor of 1.6 or 1.9 better than for a system with a large pinhole,
2
SIN
2
4
6
8
2 Diameter (Airy units) FIGURE 22.9. Signal-to-noise ratio for two-photon and three-photon fluorescence microscopes (noise model N2: background from autofluorescence from specimen bulk).
DESIGNS OF CONFOCAL MICROSCOPES
448
Chapter 22 • C.1.R. Sheppard et al.
/ Y"4) I-u-I
x scan
--.3_
If the number stored as the intensity value for a given pixel is pro-
portional to the total number of photons striking a specific area of the CCD, then the resolution of the image data is degraded somewhat compared with that produced by digitizing a continuous, bandwidth-limited signal at equally spaced time points (i.e., integral vs. point sampling). The resultant point spread function is the one without sampling, convolved with the detector sensitivity distribution for a single picture element. In the absence of noise, the more frequent the sampling, the closer the behavior tends to the ideal response.
linea, CCO
A
YSC:JJ
x scan
/ I'-cr-I
COMPARATIVE PERFORMANCE OF FLUORESCENCE MICROSCOPES
_J-'''.PM
Bleaching-limited Performance
B
line
__S..;.j;_t__....,J.F-/_ _ _Y_S--f~_m~ratlon -
-
linear CCD
C FIGURE 22.10. Schematic diagrams of various optical arrangements using line-illumination and/or linear CCD detectors.
in resolution to either (B) or (C). Arrangement (A) also has the major advantage for metrological work that geometrical distortion is absent. In (C), speckle noise can be a problem in BSL imaging unless the spatial coherence of the illumination is destroyed. However, (C) has the major advantage for fluorescence imaging that a complete line of infonnation can be recorded simultaneously with the result that data can be acquired rapidly and with less danger of fluorescence saturation.
SAMPLING For WF fluorescence imaging, the Nyquist sampling spacing is
VN
=n12. As the first zero of the Airy disk occurs at v = 3.83, corresponding to 1 Airy unit (AU), the Nyquist sampling spacing is VN =n/(2 x 3.83) = 0.41 AU = 1 AU12.44. If the image is sampled at the Nyquist spacing, or more closely, in the absence of noise, the exact image profile can be recovered. When sampling further apart than the Nyquist spacing, aliasing occurs. Note that for fluorescence imaging with sampling at the Nyquist spacing, the image can be recovered, but for bright-field imaging sampling must be closer than, rather than equal to, the Nyquist spacing. For a detector array with elements each of length 2Vd, to achieve Nyquist sampling we have Vd = n/4 (in optical units). With line illumination and a linear detector array, imaging along the array has the same spatial frequency cut-off as for WF imaging so that sampling at the Nyquist rate is appropriate. If there are n samples in a one-dimensional (lD) image, the field of view is
2vs = nn/2.
(24)
The comparative performance with regard to SIN of various designs of the fluorescence microscope can be considered by calculating the appropriate signal and background levels. In this section we consider the case when SIN is limited by bleaching of the fluorophore. Thus, a finite number of fluorescent photons can be emitted before the fluorophore bleaches. (This assumes that photobleaching rate is independent of illumination intensity.) We tenn this regime Rl. We have considered a wide range of different configurations including confocal microscopes with circular, square, or slit detector apertures, and point source or finite-sized source; line illumination with coherent or incoherent lines having vanishingly small width, or finite width; disk-scanning systems with arrays of apertures of various geometries; and WF microscopy with a finite source and a detector array. Figure 22.11 shows five different configurations of microscope that we designate as follows: Cl. Confocal microscope with a point source and a finite-sized detector radius Vd. C2. Line illumination with a coherent line source length 2v, and a linear array of square detector elements, each with dimensions 2Vd. C3. Confocal microscope with a point source and a slit detector length 2vs and breadth 2Vd. C4. Disk-scanning microscope with a square array of incoherent sources, separation Vh, radius Vd and a similar array of detectors. CS. WF microscope with a square source, sides 2v" and a 2D array of square incoherent detectors dimensions 2Vd. In order to model these different configurations of microscope we need to know the relative values of signal in the various systems.
Cl
C2
C3
I1
Source
888 CiliJ1 -- -
2V
s
vh
Detector
0
[]
radius vd
2Vd
~1
2v s
2vd
CS
C4
0
radius vd
2V s
D1 2v s
-[]
2vd
FIGURE 22.11. Five ditl'erent configurations of microscope.
Signal-to-Noise Ratio in Confocal Microscopes • Chapter 22
relative background is thus 1tV~: 4VdVs: 4v~ for circular, slit, and square detectors, respectively. The total energy that illuminates the sample in forming a complete image in a confocal microscope of a section is proportional to n2 • For the line-scanning microscope, a whole line of the specimen is illuminated at a time. Thus, the background is increased by a factor n. Thus for 13 and a square detector (C2), instead of Eq.8,
1
41i;-
11
1 4~2
12
449
d
14
~J-8 ..
(28)
Similarly, for a conventional microscope a whole field is illuminated at one time and the background is increased by a factor n2 • For 12 and a square detector (C5)
.. 2vd
(29)
FIGURE 22.12. Four different illumination models.
We consider the illumination models shown in Figure 22.12. Great care must be taken in obtaining the correct normalization for the various different microscope configurations. The appropriate assumption is that the total power incident on the specimen in forming a complete 2D image is kept constant:
For the disk-scanning microscope, the distance between the samples is taken as Vh' If ns is the number of samples needed to form a line image ns = Vh12Vd. We also introduce the number of points in a line illuminated simultaneously nh = 2V/Vh so that n = nsnh' Again the total energy incident in forming an image of a section is n2 • The background for the disk-scanning microscope is (30)
For line illumination, C2, we have
11. A point source. Normalizing the total power in the focal plane to unity, the image of a point source in the object plane is an Airy disk
I =~(2Jl(V))2 41t
(25)
V
so that the intensity at the focal point is 1I41t. 12. The pixel is illuminated uniformly with intensity 1I4vJ This normalization is taken so that the total exposure in recording a complete section is the same as for 11, as a conventional microscope images the whole section simultaneously. 13. The image of a long line source gives in the focal plane.
I=_1_(SinV)2. 21tvd
v
SjN=F3(Vd )
(31)
vJ.JYi
In Figure 22.13, the dashed line is a plot of this value against Vd for N = 512. Unlike the confocal microscope with a slit detector (C3) as shown in Figure 22.5, the SIN decreases monotonically for increasing detector element size. This is because the length of the detector array is proportional to Vd if the detector elements are square. For Vd = 1t/4, satisfying the Nyquist criterion the SIN is 0.012. For the disk-scanning microscope, C4, the SIN for large Vd, using 14, is
SjN= 4F;(vd ) ..Jiiv~nh (26)
The line illumination microscope images a whole line at a time.
(32)
On the other hand, for illumination by an array of points, which is a good approximation for small values of Vd, (33)
14. The image of a circular source radius Vd, for Vd large enough that diffraction effects can be neglected. This will be approximately true for Vd larger than about 3. For the alternative illumination schemes, the signal is scaled by the appropriate intensity at the focal point. In particular, for 12 and a square detector (C5), for 13 and a square detector (C2), and for 14 and a circular detector (C4)
respectively. The comparative behavior has been investigated for background originating either from the optical components (noise model Nl), or from the 3D bulk of the specimen (noise model N2). We take as the worst case a weak signal compared with the background (a > 1), so that the noise from the signal fluctuations is negligible compared with that of the background. When using noise model Nl, the background is proportional to the area of the detector aperture. For the confocal system, the
SIN
0.01
o
2 Diameter or width (Airy units)
FIGURE 22.13. Signal-to-noise ratio for line illumination C2, noise model Nl (dotted), the disk scanning microscope C4, noise model Nl (lower), and the conventional microscope CS, noise model Nl (upper).
450
Chapter 22 • C.I.R. Sheppard et al.
are imaged simultaneously. For line illumination n pixels are imaged in parallel, so that for a given illumination power the background in forming a complete image is reduced by a factor n, so that SIN is increased by a factor ..fii compared with a point source. Similarly, for a WF microscope, n 2 pixels are imaged in parallel, and SIN is multiplied by n. The relative SIN, assuming all are operated at the onset of saturation, are shown in Table 22.1. Under these conditions, the line-illumination microscope's behavior is superior to the confocal microscope with a circular pinhole.
SIN
2.5
5
7.5
Effects of Scanning Speed
10
FIGURE 22.14. The variation with detector element size in SIN for incoherent line illumination (noise model N2, width of line equal to detector element, 512 samples).
Let us assume that
V h,
the hole spacing, is 40, corresponding to Nh
= 20.1 for N = 512 and sampling at the Nyquist rate. Then the variation in SIN is shown as the lower line in Figure 22.13. The optimum SIN of -0.006 occurs at about 0.52 Airy units (Vd - 2). The SIN decays quickly with increasing hole diameter as the source hole illuminates a larger region of the object as well as the detector hole collecting more background. For the WF microscope, C5,
SIN= 1tF;(Vd)
(34)
2v~n
which is plotted as the upper line in Figure 22.13. Again, the SIN decays quickly with Vd as the area of illumination, that is, the field of view, increases as the area of a detector element is made larger. At the Nyquist sampling rate the signal to noise ratio is -0.001. The thickness of the object is assumed infinite, so that for noise model N2, finite dimensions for the length of line illumination, for the finite source in WF microscopy, or for the array in the diskscanning microscope are assumed. These are assumed to be just large enough for Nyquist sampling of an n x n image (Eq. 23). As an example of SIN for noise model N2 for a line-illumination system, with an incoherent line source of width 2Vd and of length 2nvd and a detector element that is square with sides 2Vd, is shown in Figure 22.14 for n = 512. The SIN reaches an optimum value at about 0.29 Airy units (Vd = 1.1). For a very narrow line illumination the optimum width of detector element is slightly larger and SIN falls off more slowly with Vd' The value for the optimum SIN is much the same in either case. Table 22.1 summarizes results for SIN according to noise models Nl and N2 for the various forms of microscope. Roughly, compared with a confocal microscope with optimum pinhole diameter, line-illumination or slit-detection systems are a factor of 2 worse for noise model N2, while conventional microscopes are about an order of magnitude worse.
Saturation-Limited Performance If the behavior of the various systems is limited by saturation, rather than bleaching, their relative merits are very different. In this case, which we call regime R2, the limit to the number of photons detected per pixel is set not by the total energy incident per pixel, but by the illumination intensity. For a given illumination power striking the specimen, the line-illumination, diskscanning, or WF systems have the advantage that numerous pixels
If our specimen is unchanging with time, we can in principle take as much time as we like in forming an image. Thus, we can always scan slowly enough that saturation effects are insignificant. The SIN behavior is then limited by bleaching (regime Rl), and the confocal microscope with a point source and a circular pinhole of optimum diameter is clearly the best performer. If we choose to speed up the scan so that a complete 2D image is recorded in some particular finite time, then the confocal microscope with a point source will reach saturation of the fluorophore before the other systems because the dwell time on each pixel must be shorter. We can define a characteristic time Tc as the time per pixel needed to generate a 2D image while just avoiding saturation. The characteristic time Tc is a property of the particular fluorophore that depends critically on its environment. It is thus the energy per pixel that can be incident on the pixel before photobleaching of the fluorophore (saturation to the l/e point) divided by the maximum power that can be incident on the pixel without significant saturation (the power needed to give an output fluorescent intensity that is half its saturated value). The frame scan time for onset of satufor a line-scanning system and ration is reduced by a factor by N for a WF microscope. In the regime R2, where behavior is limited by saturation, SIN is proportional to the square root of the imaging time. The SIN behavior is thus as shown in Figure 22.15. Let us consider a particular example. Suppose the fluorophore is such that an incident power of 1 m W is just enough to cause saturation, and that bleaching occurs after 10 frames when scanning at 3 s/frame in a confocal microscope with point source. Then the best SIN is achieved if all 10 frames are averaged to collect the maximum possible number of photons. The dwell time/pixel/frame is about lOlls, so that the total dwell time (per pixel) for 10 frames is 100 Ils. In this case, the characteristic time Tc for the fluorophore in the confocal microscope is also 100 Ils. For collection of 10 frames at 3s/frame (dwell time 100Ils), the point-source confocal microscope is clearly the best performer.
m
TABLE 22.1. Signal-to-Noise Ratio Noise model
Regime Rl Confocal Line illumination Slit detector WF Regime R2 Confocal Line illumination Slit detector WF
WF, widefield.
NI
N2
1.0 0.07 0.07 0.006
1.0 0.5 0.4 0.08
1.0 1.6 0.Q7 3.0
1.0 11.0 0.4 40.0
Signal-to-Noise Ratio in Confocal Microscopes • Chapter 22
confocal
SIN ________ ling-illl!..mlngtign
.1
.01
WF -------------------------------
N1 .001 '------'----'----~------' 10 .01 .1 .001 dwell time I T c
A R1
SIN .1
.01 N2 .001 .001
.01
.1
451
algorithm (Cox and Sheppard, 1983). These two methods result in substantially similar images, with the mean-projection method giving lower apparent noise levels, but lower levels of contrast as well, which makes it particularly sensitive to the correct setting of the zero signal level. Generation of two projections in different directions gives a stereo-pair. The process of generating a projection improves its noise performance compared with a single section (Roy and Sheppard, 1993). This is the case for both mean and peak projections, although the mean-projection method is more effective at reducing noise. The mean-projection method increases SIN as the square root of the number of sections processed that have significant signal: only four such sections are required to increase SIN by a factor of 2.2 The SIN behavior of the peak-projection method depends on the statistical properties of the signal and the display lookup table. For large numbers of photons per pixel, the noise level is reduced by a factor of 2 by processing 24 sections, while for just two photons per pixel, thousands of sections are required to double the SIN. In general, in order to produce a projection with a given SIN, it is not necessary for each section to have that SIN . The meanprojection algorithm reduces noise as a direct result of the averaging inherent in the method, no averaging of the individual sections being necessary. There will still then be noise on the depth information in a stereo-pair presentation, but this is a second-order effect and is expected to be much less important perceptually.
SUMMARY 10
• Care is necessary in selecting the various system parameters, such as pinhole diameter, type of detector, pixel size, illumination intensity, and scan rate, in order to optimize the performance of the microscope system. • Optimum operation will also depend on the shape of the stain in the object, how the image is to be processed, and what information is eventually required from the image. • Usually a relatively small setting for pinhole diameter results in improved SIN, and a circular confocal aperture of about 0.63 times the diameter of the first dark ring of the Airy disk is a good benchmark for optimum performance. • Even if a single frame looks noisy, the noise may be reduced by the generation of projections, especially using the meanprojection algorithm and an even greater improvement can be gained by deconvolving the dataset in 3D. • The SIN performance of a point-source confocal microscope is most noticeably better than other forms in the presence of background from autofluorescence from the cements and coatings of the optical elements. • Point-source confocal microscopes are preferable for generation of high-quality 3D data sets when there is no limit on the time available for a single frame. • Line-scanning systems, or even WF microscopes with cooledCCD detector arrays, can give superior SIN performance for fast scanning, particularly in the presence of background consisting of autofluorescence from around the stained object. • In both these cases, the illumination field should be reduced so that it is not much larger than the observed region to increase SIN.
dwell time I Tc
B FIGURE 22.15. Effect of pixel dwell time on SIN: (A) Noise model N I; (8) noise model N2.
As discussed in a later section, these 10 frames can be 10 different axial sections, and the noise still tends to average out when forming a projection as the total number of photons per display pixel is increased. When forming a single image at 3 slframe, the line-illumination system achieves slightly better SIN according to noise model N2 [Fig. 22.1O(B)] , whereas the WF microscope still exhibits inferior SIN . For ten frames at television rates, the lineillumination performs best with respect to SIN, while for a single television-rate frame both the line-illumination and the WF microscope have better SIN than for the point source confocal microscope. (Note that it is not possible to read out the cooled-CCD detector we have assumed for the WF case at video rates without degrading its performance.) However, for noise model Nl, where the background comes from autofluorescence of the optical components, the confocal microscope maintains its superior SIN performance even for high scan speeds.
3D IMAGING Once a 3D data set, representing the intensity vanatlOns in the image, is stored in the computer imaging system, in many cases it is necessary to generate a projection through the 3D object rather than a single section. There are a number of ways of doing this, but the most usual are the mean-projection algorithm (Wilson and Hamilton, 1982; Sheppard et at., 1983), and the peak-projection
2
As the process of calculating the mean value at each pixel reduces the brightness, the resulting image may be hard to see unless the contrast is expanded using a simple lookup table. Doing so does not decrease the SIN.
452
Chapter 22 • C.1.R. Sheppard et al.
REFERENCES Awamura, D., and Ode, T., 1992, Optical Properties of Type 1 - Type 2 Microscopes, New Trends in Scanning Optical Microscopy, Okinawa, lapan. Benedetti, D.A., Evangelista, v., et aI., 1992, Confocal line microscopy, 1. Microsc. 165:119-129. Brakenhoff, GJ., and Visscher, K., 1992, Confocal imaging with bilateral scanning and array detectors, 1. Microsc. 165:139-146. Cogswell, C.l., and Sheppard, CJ.R., 1990, Confocal brightfield imaging techniques using an on-axis scanning optical microscope, In: Confocal Microscopy (T. Wilson, ed.), Academic Press, London, pp. 213-243. Cox, LJ., and Sheppard, C.l.R., 1983, Digital image processing of confocal images, Image Vis. Comput. 1:52-56. Draaijer, A., and Houpt, P.M., 1988, A standard video-rate confocal laserscanning reflection and fluorescence microscope, Scanning 10: 139145. Dunn, A.K., Smithpeter, C., et aI., 1996, Sources of contrast in confocal reflectance imaging, Appl. Opt. 35:3441-3446. Gan, X.S., and Sheppard, CJ.R., 1993, Detectability: A new criterion for evaluation of the confocal microscope, Scanning 15:187-192. Gu, M., and Sheppard, C.l.R., 1991, Effects of finite-sized detector on the OTF of confocal fluorescent microscopy, Optik 89:65-69. Huang, D., Swanson, E.A., et aI., 1991, Optical coherence tomography, Science 254:1178-1181. Koester, C.l., 1980, A scanning mirror microscope with optical sectioning characteristics: Applications in ophthalmology, Appl. Opt. 19: 17491757. Petral!, M., Hadravsky, M., et aI., 1968, Tandem scanning reflected light microscope,l. Opt. Soc. Am. 58:661-664.
Roy, M., and Sheppard, C.l.R., 1993, Effect of image processing on the noise properties of confocal imaging, Micron 24:623-635. Shannon, C.E., 1949, Communications in the presence of noise, Proc. Inst. Radio Eng. 37:10-21. Sheppard, C.l.R., 1991, Stray light and noise in confocal microscopy, Micron Microsc. Acta 22:239-243. Sheppard, C.J.R., and Cogswell, C.l., 1990, Three-dimensional imaging in confocal microscopy, In: Confocal Microscopy (T. Wilson, ed.), Academic Press, London, pp. 143-169. Sheppard, C.l.R., Cogswell, C.l., et aI., 1991, Signal strength and noise in confocal microscopy: Factors influencing selection of an optimum detector aperture, Scanning 13:233-240. Sheppard, C.l.R., Gu, M., et aI., 1992, Signal to noise ratio in confocal microscope systems, 1. Microsc. 168:209-218. Sheppard, C.l.R., Hamilton, D.K., et aI., 1983, Optical microscopy with extended depth of field, Proc. Roy. Soc. Lond. A 387: 171-186. Sheppard, CJ.R., and Mao, X., 1988, Confocal microscopes with slit apertures, 1. Mod. Opt. 35:1169-1185. Slusher, R.E., Hollberg, L.W., et aI., 1985, Observation of squeezed states generated by four-wave mixing in an optical cavity, Phys. Rev. Lett. 55:2409-2412. Webb, W.W., Wells, K.S., et aI., 1990, Optical Microscopy for Biology, WileyLiss, New York. Wells, K.S., Sandison, D.R., et aI., 1990, Quantitative confocal fluorescence with laser scanning confocal microscopy, In: Handbook of Biological Confocal Microscopy (1. B. Pawley, ed.), Plenum Press, New York, pp. 23-35. Wilson, T., and Hamilton, D.K., 1982, Dynamic focussing in the confocal microscope, 1. Microsc. 128:139-148.
23
Comparison of Widefield/Deconvolution and Confocal Microscopy for Three-Dimensional Imaging Peter
J.
Shaw
INTRODUCTION The biggest limitation inherent in optical microscopy is its lateral spatial resolution, which is determined by the wavelength of the light used and the numerical aperture (NA) of the objective lens. Another important limitation is the resolution in the direction of the optical axis, conventionally called z, which is related to the depth of field. The presence of a finite aperture gives rise to undesirable and rather complicated characteristics in the image. In essence, the depth of field depends on the size of structure or spatial frequency being imaged. Fine image detail, which is generally of most interest, has a small depth of field, and only features within a small distance of the focal plane contribute to the image. On the other hand, large structures - low spatial frequency components - have a relatively large depth of field, and contribute to the detected image seen at distant focal planes. This is very noticeable in dark-field imaging modes, such as epi-fluorescence, and means that the fine image detail may be swamped by low resolution "out-of-focus" light and thus either lost, or visualized with very much reduced contrast. The principal advantage of confocal microscopy for biological imaging is that the optical arrangement has the effect of eliminating much of the out-of-focus light from detection, therefore improving the fidelity of focal sectioning (and hence the threedimensional imaging properties), and increasing the contrast of the fine image detail. But the rejection of the out-of-focus light necessarily means that a proportion of the light emitted by the specimen is intentionally excluded from measurement. All illumination of the specimen has deleterious effects - bleaching of the fluorochrome or phototoxicity to living cells. These specimendependent factors are the ultimate limitation to the quality of the image, and inevitably confocal imaging does not detect much of the emitted light. An alternative way of removing the out-of-focus light involves recording images at a series of focal planes using a conventional microscope, often called widefield (WF) to distinguish it from confocal imaging, and then using a detailed knowledge of the imaging process to correct for it by computer image processing. This procedure is called deconvolution, and its application to biological problems actually preceded the widespread introduction of biological confocal microscopes (Castleman, 1979; Agard and Sedat, 1983; Agard et al., 1989). In contrast to confocal imaging, up to 30% of the total fluorescent light emitted by the specimen can be recorded (i.e., all the light that can be collected by a single, high-NA objective). This chapter examines the question: Is it better to record all the light emitted and process the WF images to
redistribute the out-of-focus light to produce a more accurate threedimensional (3D) image, or to exclude the out-of-focus light from measurement in the first place by confocal optics and then deconvolve the confocal data?
THE POINT SPREAD FUNCTION: IMAGING AS A CONVOLUTION In order to derive a soundly based description of the degradation introduced by an optical microscope, especially if any attempt is to be made to reverse this degradation, it is necessary to be able to describe the relation between the specimen and its optical image in mathematical terms. We shall give here a very condensed explanation - the interested reader is referred elsewhere for more rigorous mathematical derivations (Agard et ai., 1989; Shaw, 1993; Young, 1989; and Chapters 20, 21, 22, 24, and 25, this volume). Within some quite general l~mitations, the object (specimen) and image are related by an operation known as convolution. In a convolution, each point of the object is replaced by a blurred image of the point having a relative brightness proportional to that of the object point. The final image is the sum of all these blurred point images. The way each individual point is blurred is described by the point spread function (PSF), which is simply the image of a single point. This is illustrated diagrammatically in Figure 23.1. The conditions that must be met for an imaging process to be described as a convolution are that it should be linear and shift invariant (Young, 1989). Imagine cutting the specimen into two parts and imaging each part separately with the microscope. If adding these two subimages together produces the same result as imaging the whole specimen, and does this irrespective of how the specimen is cut up, then the imaging is said to be linear. If the imaging is indeed linear, then the specimen can be imagined cut up into smaller and smaller pieces, until the size of each piece is well below the resolution limit, and can be considered to be simply a point. The image is then the sum of the images of each of the points, each multiplied by a function corresponding to the amount of light coming from that point. The multiplication and summing is represented mathematically by an operation called convolution. Shift invariance simply means that the imaging characteristics and thus the PSF are the same over the whole field of view, and knowing one PSF is enough to characterize the imaging properties of the microscope (Agard et ai., 1989; Shaw and Rawlins, 1991a). Although imaging modes such as phase contrast and differential interference contrast (DIC) are not linear, because their contrast depends on differences of refractive index within the object,
Peter J. Shaw· John Innes Centre, Colney, Norwich NR4 7UH, United Kingdom Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
453
454
Chapter 23 • P.J. Shaw
object
image
([f'(r)]
=
I[I
f(s)o(r - S)dS]
+00
=
A
f f(s)[o(r-s)]ds
+00
B
,
=
.. ." .. . ...-............ 30,000 Depends on CCD readout (>0.05) IOnW Hg arc 350-650nm
Single Spot Confocal
Scanning-Disk Confocal
3%-12% (PMT) 0.05)
3%-12% (PMT) 7) and 3.0ns at low pH (pH < 3). In addition to the fluorescence lifetime effect, the quantum yield of carboxy-fluorescein went down to 5% at pH 3, rclative to that at pH 7. At an excitation wavelength of 800 nm, the biofilms show some autofluorescence with an average fluorescence lifetime of
FIGURE 27.13. Intensity image ofbiofilm (A) with individual bacteria visible. Fluorescence lifetime image (B) shows a homogeneous pH of 6.2 ± 0.3.
less than 1 ns. The difference in fluorescence lifetime between the autofluorescence and the carboxy-fluorescein was exploited to suppress the autofluorescence in the pH measurements. This was achieved by opening the first gate I ns after the excitation pulse. This suppresses 80% of the autofluorescence and only 20% of the carboxy-fluorescein signal. In addition, a comparatively high probe concentration of 50 to 100/-lM was used in the pH imaging experiments. The autofluorescence was less than 1% of the total fluorescence intensity in all the measurements, even at low pH. In Figure 27.12, the pH calibration of the microscope is shown. Intensity and pH xy-images (30 x 30/-lm2 , z = 60/-lm) of carboxy-fluorescein stained biofilm are shown in Figure 27.13. The intensity image [Fig. 27.13(A)] shows individual bacteria, while the fluorescence lifetime image [Fig. 27.13(B)] shows an almost homogeneous pH of pH 6.2 ± 0.3. At a constant pH the quantum efficiency of the carboxy-fluorescein is constant. Therefore, the heterogeneous fluorescence intensity distribution can be attributed to a non-homogeneous probe distribution. Binding of some of the probe to the bacteria may cause this. A pH gradient was induced by overlaying the specimen with a 14mM sucrose solution. The fermentation of sucrose lowers the pH outside the bacteria. In Figure 27.14, the intensity [Fig. 27.14(A)] and fluorescence lifetime [Fig. 27.14(B)] images 94min after the addition of the sucrose are shown. These images were recorded at the same position as that of Figure 27.13. The fluorescence intensity image shows some brightly fluorescing areas. At the same position in the fluorescence lifetime image, a pH is found of less than 3. This observation is somewhat unexpected because the quantum efficiency of the probe goes down with pH. It suggests that a high local probe concentration is present at the bright spots. This effect may be caused by probe precipitation and makes the measurements at these locations less reliable. Therefore, the bright spots were excluded from the pH analysis. The average pH of this image is 5.2 ± 0.4, one pH unit lower than the average pH of the reference image.
Probes for Fluorescence lifetime Microscopy A literature study and some measurements on probes in our own laboratory revealed a large list of probes which can potentially be used for fluorescence lifetime imaging, this list is summarized in Table 27.3.
FIGURE 27.14. Intensity (A) and lifetime (B) images after overlaying the biofilm specimen with a 14mM sucrose solution. The average pH is 5.2 ± 0.4.
Fluorescence lifetime Imaging in Scanning Microscopy • Chapter 27
531
TABLE 27.3. Ion Sensitive Probes PH Probes Probe SNAFL-I Carboxy-SNAFL-I Carboxy SNAFL-2 Carboxy SNAFL-3 Carboxy SNARF-5 SNARF-6 Carboxy SNARF-6 SNARF-I SNARF-2 Carboxy SNARF-I Carboxy SNARF-I Carboxy SNARF-2 2-Naphtol Acridine Virginiamycin S' BCECF DM-NERF Carboxy-fluorescein d CI-NERF Acridine Orange
A"JA,m (nm)
'ooid"
(ns)
'h"," (ns)
Reference Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et al. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Whitaker et at. (1991) Laws and Brand (1979)
295"/500b 295"/500b 295"/500" 295"/500b 295"/500"
3.58 3.44 4.19 3.63 4.21
1.14 0.95 0.87 1.16 0.73
295"/500b 295"/500b 295"/500b 295"/500" 295"/500b 295"/500b 313/360
4.50 0.46 0.30 0.60 0.60 0.27 0.80 (0.35) 7.30(1.91) 9.6 0.48 3.0 2.3 3.0 1.3 1.8
1.04 1.54 1.75 1.32 1.32 1.67 4.82
348/560 330/420 490/520 490/520 490/520 490/520 490/520
31.1 1.9 3.8 4.0 4.0 4.0 5.3
Gafni and Brand (1978) Clays et at. (1991) This laboratory This laboratory This laboratory This laboratory This laboratory
Ca 2+Probes Probe
Am/A,m (nm)
'no Ca++" (ns)
'C'++" (ns)
Fura-2
340/420 380/420
0.77(0.49) 1.5(0.51)
1.77
Keating et al. (1989)
488/520
1.6
3.5
Lakowicz et at. (1994) This laboratory
488/520
2.1
3.6
This laboratory
488/520
0.5
3.4
This laboratory
488/520
1.5
3.7
This laboratory
Reference
Quin-2 Calcium green Bapta I Calcium green Bapta 2 Calcium green Bapta 5N Oregon Green Bapta 1,2 Oregon Green Bapta 5N BTC Indo I
488/520
0.5
2.8
This laboratory
340/400-475
0.7 1.4
1.2 1.7
Indo 5F
340/400-475
1.4
1.4
This laboratory Lakowicz and Szmacinski (1993) This laboratory
CI Probes Probe SPQ/CI-
A"JAcm (nm)
'tnoct (ns)
'CI" (ns)
-/-
26.0
2.0
Reference Dix and Verkman (1990)
DNA or RNA Selective Probes Lifetime Differences with Free Probe and Probe Bound to DNA
A"JAcm (nm)
Ar"," (ns)
'[bound
Proflavine, acriflavine, acridine yellow (208C)
-/-
4.5
4 (0.7)
9-Aminoacridine
-/-
Rivanol
-/-
6.5
Quinacrine
-/-
4
Probe
Ethidium bromide Acridine orange, DNA RNA RNA
532/488/520 488/520 488/>630
15
1.7
"
7 (0.3) 10.5 (0.7) 31 (0.3) 5.5 (0.7) 13 (0.3) 3.5 (0.7) 19 (0.3) 24.2 1.7-1.9 1.7-1.9 5-20
Reference Duportail et at. (1977)
Duportail et al. (1977) Duportail et al. (1977) Duportail et al. (1977) Atherton and Beaumont (1984) Marriott et al. (l991b) Marriott et al. (1991 b) Marriott et al. (1991 b) continued
532
Chapter 27 • H.C. Gerritsen et al.
TABLE 27.3. Ion Sensitive Probes (Continued) Lifetime Differences with Different Amounts of A + T
"-900nm), we have to consider heating due to increased absorption by liquid water, which is not a problem at visible and near-UV AS where water is very transparent (e.g., see Fig. 3 in Svoboda and Block, 1994 or Fig. 23.3, this volume). For IPE, an upper-bound estimate of the temperature rise can be made using a 2D approximation because absorption occurs all along the beam path. The calculation sketched here is the same as was used to analyze thermal lens effects (Whinnery, 1974; Kliger, 1983, and references therein). For the temperature rise T at the center of the beam (r = 0) as a function of the time t after switching on the beam one obtains
T 2D (t,r = 0) = uP Inl2t + IJ 4nkT
(5)
't,
where a is the absorption coefficient, P the laser power, kT the thermal conductivity, and 't, = ffii'l(4K) the thermal time constant, which is a measure of how fast steady-state conditions are ap-
539
pro ached and which depends on ffio, the Gaussian beam parameter and is equivalent to the beam radius (lk intensity) in the focal plane, and on the thermal diffusivity K = krlp where p is the volume heat capacity. For a diffraction-limited beam at high-NA (w" = 200 nm), 't, z 70ns in water (using kw = 0.6 WK-' m-', Kw = 1.44 X 10-7 m's-I) and for absorption by pure water, the pre-factor in Eq. 4 is 0.013, 0.21, and 0.66K at AS of 700, 1000, and 1300nm, using the absorption coefficients for water of 0.02, 0.32, and 1.0cm-', respectively; the laser power was assumed to be 50mW (approximately the saturating intensity; Denk et at., 1990). Slightly lower values, still logarithmically diverging with time, are found if axial heat transport is taken into account (Schonle and Hell, 1998). Due to the small beam diameter, 't, is rather short (z70ns) for high-NA objectives. For video-rate scanning microscopes (Goldstein et at., 1992; Fan et at., 1999; see Chapter 29, this volume) this results in a temperature rise of only 1.55 times the pre-factor but at 10 /1s dwell-time (typical for non-resonant mirrorscanned instruments), the temperature rise is 5 times the prefactor. For an infinite sample no steady-state value for the temperature would ever be reached. In practice, the temperature rise will eventually be limited, by the finite sample size and by convection or bath perfusion, which remove heat at a rate much faster than heat conduction alone. For stationary applications or when continuously scanning a small area, rather large logarithmic factors can occur, however. Therefore, water absorption may have to be taken seriously, particularly at high illumination powers and long wavelengths or when attempting multi-photon excitation with CW lasers (Hanninen et al., 1994, 1996; Booth and Hell, 1998; Hell et at., 1998). Fast scan rates, rapid bath perfusion, thin sample cells, and, of course, maximizing the two-photon advantage using the shortest pulses possible are remedies to reduce high peak temperatures. To assess the IP effects of infrared (IR) beams on biological specimens, we can also exploit the experience gained with optical tweezers, which are routinely used on living cells at comparable or higher power densities (Ashkin et at., 1987; Svoboda and Block, 1994) and for which damage has been assessed for most of the wavelength regime used in 2PM (Neuman et aI., 1999). Heating due to 2PA is restricted to the focal region. A 3D model is, therefore, appropriate. Because we are interested in the case of high-NA, we can use the approximation that the release of heat occurs uniformly within a sphere with radius ffio centered at the focus. The relationship one gets for the temperature rise is:
T
-
3D -
Pahs
ffio4nkT
[1-~] V~
(6)
where Pub., is the total absorbed power (see below). For large t, when the square root goes to zero, T1J), unlike Tw, approaches an asymptotic value, given by the factor in front of the square brackets. For high energy (m]) pulses at low repetition rate, the local temperature rise during a single pulse can easily be large enough to cause damage, but we know little about how damage might be exacerbated for the case of pulsed light at high repetition rates if. zlOOMHz) compared to the CW case with the same average power. Because 't, is longer than the interpulse interval (l!fR), the incremental temperature rise during a single pulse is smaller than the steady-state temperature rise Tw (t = 00) roughly by a factor of 't,F,.
We conclude that heating during high-repetition-rate pulsed illumination can largely be treated like CW illumination and is often negligible at practical 2PM parameters. Attention has to be paid to situations where high local concentrations of chromophore
540
Chapter 28 • W. Denk et al.
occur, as, for example, for DNA stains, which can bind at a concentration of one per base pair or where equilibration of the molecular temperature with its environment cannot be automatically assumed (Akaaboune et al., 2002). Because of the localization of excitation to the focal volume, total photobleaching in MPM is generally much reduced compared to 1PE microscopy. However, it has been shown that an increased photobleaching rate from within the focal volume can occur by a mechanism where the fluorophore is initially excited by simultaneous 2PA, and then one or more photons interact with the excited molecule, possibly via higher-order resonance absorption (Patterson and Piston, 2000). This effect can be quite pronounced for readily photobleachable dyes, such as fluorescein, where the difference between one- and two-photon photobleaching rates can be a factor of 10 at the power levels that are typically used in biological imaging (lOO~W CW for single-photon excitation, and 3 mW 150fs pulses for MPM). However, for more stable dyes, such as the green fluorescent proteins (GFPs), carbocyanines, and AlexaFluors, the photobleaching rate is in our experience often too small for the difference to be measurable at the usual imaging intensities. While in some cases direct higher-order absorption (three or more photons) may be relevant, several studies (Koester et aI., 1999; Konig et aI., 1999) have found that longer pulses (which reduce higher-order absorption) do not reduce the damage done per excitation event.
options for detection. In this section we will discuss laser sources suitable for MPE, the advantages and drawbacks of the various methods of detecting fluorescence and other contrast signals, and specific problems that occur with non-mechanical (e.g., acoustooptical) beam power control and deflection. We assume that the reader of this chapter is familiar with the principles of IPCM (other chapters, this volume). Short shrift will, therefore, be given to those aspects such as mechanical beam scanning, data collection, storage, and display that are largely identical for IPCM and MPM instruments. The potential user should also be aware that MPMs are relatively easy to set up and are now available as integrated systems from several manufacturers. MPM systems are still expensive with the price of the laser system (>$150,000) being between one third to one half of the total system cost. With a mode-locked laser, one has, however, also acquired the light source necessary to do time-resolved fluorescence measurements (Piston et aI., 1992; Zhang et aI., 2002; and Chapter 27, this volume).
lasers and the Choice of Excitation Wavelengths CPM laser The first 2P images (Denk et aI., 1990) and 2P photochemical microcopy images (Denk, 1994) were recorded using collidingpulse mode-locked (CPM) lasers (Valdmanis and Fork, 1986) at 615 nm excitation wavelength. Today this laser type is of only historical interest.
INSTRUMENTATION
Hybrid Mode-locked Dye laser Setup (Fig. 28.3) and operation of a MPM system are very similar to those of a lP laser-scanning microscope. The main differences lie in the type of excitation lasers and in the increased number of
Another early type of ultrashort pulse dye laser system is the hybrid mode-locked dye laser. These systems use an actively mode-locked argon-ion or a frequency-doubled neodymium: YAG dichroic mirror
scan
mode-locked laser
eyepiece
filter -- ~, ~
time scales
pinhole
(PMT')
pulse repeat
~ (fluor:~::ncl pulsewidth
-
~.,
10-8 s
objective lens
emission
FIGURE 28.3. Schematic diagram of a two-photon laser-scanning microscope illustrating various detection possibilities. The stream of incoming laser light pulses is raster scanned (xy scanner, only one axis is shown here) in a way that is identical to the single-photon LSM. For fluorescence microscopy, several detection possibilities are indicated: (I) external: fluorescence light bypasses objective lens; (2) whole-area: fluorescence light passes objective lens and is then deflected by a dichroic mirror to be focused onto the detector by a transfer lens; (3) descanned: as in the lPLSM, the fluorescence light is reflected off the scanning mirrors, allowing confocal detection (see text). Not shown, but possible and occasionally used, isfocal-array detection, where, after deflection by a dichroic mirror, fluorescence light is detected by an array detector located in an image plane. Yet another possibility is non-optical detection using, for example, an electrically recorded signal from the sample. Time scales are indicated in the left inset.
Multi-Photon Molecular Excitation in Laser-Scanning Microscopy • Chapter 28
laser to pump a dye laser that also contains an intracavity saturable absorber jet. Such systems are rather expensive and difficult to operate and are therefore rarely used for multi-photon imaging. The remaining advantage over the titanium: sapphire laser (see below) is the access to the range 550nm < A < 700nm, which is desirable for some uncaging experiments but has been virtually abandoned for imaging due to photodamage problems (Kiskin et at., 2002).
Titanium: Sapphire Laser For most applications, the light source of choice for MPM currently is the self-mode-Iocked titanium: sapphire (Ti: Sa) laser (Spence et al., 1991), nowadays pumped by a frequency-doubled diodepumped Nd: Vanadate laser rather than a power- and cooling-water hungry argon-ion laser. The Ti: Sa laser provides a large tuning range (from slightly below 700nm to slightly above 1050nm) with pulse lengths shorter than 100fs and sufficient power (2 W average at the peak of the tuning curve, down to a few hundreds of milliwatts at the edges when pumped with lOW) to permit saturating excitation (see Physical Principles) of most fluorophores with a high-NA objective over much of the laser's tuning range. The tuning range of Ti: Sa is now covered by a single set of cavity mirrors, with optics changes only required to reach wavelength above 1000nm or below 700nm. Currently, several manufacturers offer turnkey laser systems that contain the pump source and Ti: Sa laser inside a single housing, are computer controlled, and no longer require any mechanical adjustments by the operator.
Other Light Sources If losses in the excitation path are too large, it is sometimes not possible to achieve the desired excitation rates with the multiphoton advantage factors available for a laser oscillator alone. A reduction in repetition rate while maintaining average power can then increase the excitation efficiency substantially (Beaurepaire et al., 2001). This can be achieved by increasing the cavity length, cavity dumping, or regenerative amplification. The last approach has recently been shown to allow imaging down to the surfacegenerated-background limit (Theer et aI., 2003). Direct use or frequency doubling of femtosecond pulses from optical parametric oscillators (OPO) (Cheung and Liu, 1991; Fu et aI., 1992; Powers et aI., 1994; Keller, 1996) may provide an almost universal, if expensive, solution to cover almost all of the desired wavelength range. One factor limiting multi-photon microscopy is the cost of the laser source, which, in spite of early hopes, has not come down significantly with the introduction of diode pumping (for a review, see Keller, 1994). One reason is that gain materials that can be directly diode-pumped (Keller, 1996) have insufficient tuning ranges and/or unfavorable thermal characteristics. In niche applications, other sources (Wokosin et at., 1996a) have been used, partly within, for example, the Cr: LiSaF laser (Svoboda et al., 1996a) or outside, for example the Cr: Forsterite laser (Liu et at., 2001), the Ti: Sa tuning range.
Excitation Wavelengths One reason for the success of the Ti: Sa laser for MPM is that the range 700 nm < ~n < 1050 nm (corresponding to 350 nm < Aln < 525 nm) covers the range of excited state energies for many commonly used fluorophores (see below). Much shorter wavelengths, in particular ~n < 640nm, are likely to cause photodamage due to intrinsic absorption, for example, by tryptophan rich proteins (Rehms and Callis, 1993). To minimize scattering one might
541
lengthen the excitation wavelengths and take advantage of a dip in the absorption spectrum of water around 1040 nm, which is well known from optical trapping experiments (Svoboda and Block, 1994). Fortunately, a large selection of microscope lenses has become available with excellent transmission and optical correction in the near IR (Chapter 7, this volume). The use of older lenses that were not designed for the infrared can be problematic, particularly in the case of highly corrected lenses (Neuman et aI., 1999), where non-optimal performance of the antireflective multi-layer coatings on each of the numerous internal surfaces can reduce overall transmission catastrophically.
Beam Delivery and Power Requirements In general, the laser is mounted on the same vibration isolation platform as the microscope because delivery of ultrashort pulses through standard, single-mode fibers, which is possible in principle (Wolleschensky et al., 1998) (see also Chapter 26, this volume), requires substantial technical efforts to prevent unacceptable pulse broadening at the laser powers routinely required. Development of special optical fibers, such as photonic band gap fibers or large cross-section single-mode fibers (Helmchen et aI., 2002; Ouzounov et aI., 2002), may facilitate MPM applications where fiber delivery is essential (Helmchen et al., 2001). In non-absorbing, non-scattering samples saturating pulse energies (corresponding to several tens of milliwatts of average power) can easily be reached over most of the Ti: Sa tuning range even with 5 W of green pump power. However, one rarely has more than the desirable power in scattering samples such as in brain slices or the intact brain. Power availability may also be limiting when attempting to optimize resolution by overfilling the objective back aperture.
Detection As discussed above, excellent 3D localization is accomplished in MPM by excitation alone. This allows more flexibility in the optical design and, as a consequence, considerable improvements of fluorescence collection efficiency are possible compared to the IPCM. Figure 28.3 depicts the various options. The positions of non-imaging detectors are designated PMT because photomultiplier tubes are usually the detectors of choice for MPM. In general, considerations as to which detector type to use in MPM are quite similar to those for 1PCM, and the reader is referred to Chapter 12 of this book. Among non-imaging schemes (one or a few detector elements), the main distinction is whether the emitted light passes back through the scanning mirrors (descanned detection) or whether the detector is sensitive to emitted light from the whole image area at all times (whole-area detection). A variant of the latter is external detection, where detected light does not pass through the objective lens.
Whole-Area and External Detection Whole-area detection (WAD) (Piston et at., 1992, 1994) is now the detection mode of choice in the majority of MPM applications. The WAD pathway uses a dichroic mirror somewhere between the scanner and the objective to separate excitation and fluorescence (alternatively the excitation light can be coupled in by reflection from a dichroic), preferably after a minimum number of optical surfaces to maximize detection efficiency. The signal is then passed through the collection optics, which needs to avoid vignetting. If the back aperture of the objective is conjugate to the photocathode of the PMT the effect of spatial heterogeneities in the photocathode sensitivity is reduced. One of the main advantages of WAD is the ability to efficiently collect fluorescence from
542
Chapter 28 • W. Denk et al.
specimens that scatter light at '",m so strongly that only a very small fraction can be refocused for confocal detection (Denk et al., 1994; Denk and Svoboda, 1997; Beaurepaire and Mertz, 2002). WAD is as vulnerable to contamination from ambient room light as is widefield imaging with highly sensitive cameras. One thus loses a convenient but rarely essential advantage of confocal imaging. While WAD through the excitation lens is usually the most convenient and efficient mode, external detection, where the detected light bypasses the objective lens, can be necessary when light needs to be detected that cannot (e.g., because it is of too short a wavelength) or did not (e.g., because it went off in the wrong direction) pass the objective. Combining through-the-lens collecting with collecting the light passing through the condenser has been used successfully to increase the signal-to-noise ratio in embryo (Denk et al., 1997) and in brain slice imaging (Koester and Sakrnann, 1998; Mainen et al., 1999a, 1999b). Another disadvantage of WAD is that detectors with a large "phase-space volume" (given by the product of detector area and acceptance angle) are needed thus ruling out the use of small-area photon-counting avalanche photodiodes (Tan et al., 1999) or of spectrometers (Lansford et ai., 2001).
Descanned Detection When converting a confocal microscope to multi-photon operation (Denk et ai., 1990) descanned detection naturally results. While this mode is less efficient than WAD, even for clear specimens, descanned detection does allow the use of detectors with small apertures such as avalanche photodiodes or spectrometers. A pinhole that is several times larger than the optimal confocal size can be useful for excluding room light contamination from the detected signal while still being near optimal for signal collection. The use of a confocal pinhole as a tight spatial filter in addition to multi-photon excitation (Stelzer et aI., 1994) is rarely used because it is fraught with several drawbacks: (1) A pinhole small enough to produce any substantial increase in resolution causes a large drop in detection efficiency due to the fact that the diffraction-limited volume at the '"em is smaller than the excitation volume determined by the ,"2ex because '"em < ,"2ex. Such a loss of detection efficiency is particularly serious because fluorescence imaging of living specimens is often limited by photobleaching and photodamage. A technicalIy complex yet feasible solution to this problem might be to use a small array of detectors together with the appropriate deconvolution algorithms (Sheppard and Cogswell, 1990). Chromatic aberration, already a problem in IPCM, is exacerbated in the confocal operation of MPM because the typical shift between and '",m is much larger (50nm < ,",m - ,"lex < 200 for IPE, 200nm < ,",m < 500 in 2PE, and further increasing with >2 PE).
'"20
'"20 -
Non-Optical Detection A number of non-optical detection schemes have become very promising owing to the high degree of spatial localization achieved during excitation alone. Two-photon scanning photochemical microscopy (Denk, 1994; Furuta et ai., 1999; Matsuzaki et ai., 2001) generates images of receptor distributions by locally releasing agonists such as neurotransmitters from "caged" precursors and detecting the agonist-induced ionic current in voltage-clamped cells. In fact this concept was one of the motivating factors for the initial development of MPM. Opto-acoustic detection, which has been used to measure two-photon absorption coefficients (Patel and Tam, 1981; Bindhu et ai., 1998) could be used to measure
spatially resolved absorption that is not accompanied by fluorescence or induced current, but has to date not been tried as a contrast mechanism in MPM.
Focal-Plane Array Detection A rather different strategy, which does not rely on scansynchronized detection to build up the image, is the use of an imaging detector. As in conventional fluorescence microscopy, the fluorescence is refocused to an image plane, and the image is generated by spatially sorting the fluorescence photons into the pixels of an array detector such as a charge-coupled device (CCD). The lateral resolution is then determined solely by '"m" which is considerably shorter than ,"2ex. The optical sectioning effect due to two-photon excitation is, however, retained and provides discrimination and resolution in z-direction. This method is the equivalent of widefield fluorescence microscopy with only a thin focal slice rather than the whole thickness of the sample excited. Focal array detection is particularly useful in connection with mUlti-point illumination, where it allows the acceleration of image acquisition (Straub and Hell, 1998; Egner and Hell, 2000; Andresen et ai., 200 I; Fittinghoff et ai., 2001; HelI and Andresen, 2001; Nielsen et ai., 2001; Egner et ai., 2002). The main disadvantage of focal array detection is that, different from the case of single-point scanning, with whole-area detection, scattering of fluorescence light leads to an immediate degradation of image contrast and resolution.
Optical Aberrations Aberrations inherent in the microscope and spherical aberration introduced by focusing through refractile layers such as the coverslip, immersion oil, and sample (Sheppard and Cogswell, 1991; HelI et ai., 1993) broaden the focus, shift the apparent focal point (Visser et ai., 1991), and reduce the peak excitation intensity. Due to the mathematical equivalence between the optical transfer function of the non-confocal 2PM and that of the confocal 1PM (Sheppard and Gu, 1990; for a minor modification, see Visser et ai., 1991), the effects of monochromatic aberrations, such as spherical aberration and astigmatism, on the amplitude and resolution of the detected signal are the same in both cases. In the two-photon case, however, the number of molecular excitations is actually reduced due to the smeared-out focus spot and the intensity-squared dependence of the excitation probability (see above). When photobleaching or photodamage are the limiting factors, this can provide a significant advantage of MPM over the 1PCM case, where the same number of excitations occur, but the fraction of the emitted light that reaches the detector is reduced. Nonetheless, one must take the same precautions with MPM as with 1PCM when interpreting absolute light levels as a function of focusing depth. Note also, that most aberrations become rapidly less severe as NA is reduced. The best way to avoid spherical aberration in aqueous specimens, even at high-NA, is the use of water-immersion objective lenses, which are now widely available corrected even for the IR range (Chapter 7, this volume). A significant motivation for the development of 2PM was the circumvention of the poor chromatic correction then found for most microscope lenses in the uv. Chromatic aberration problems playa role in (non-confocal) 2PM only in connection with the broad," spectrum of ultrashort pulses (see above). However, this spread is generalIy smaller than a typical Stokes shift and chromatic correction is easier in the IR where glass dispersion flattens out.
Multi-Photon Molecular Excitation in Laser-Scanning Microscopy • Chapter 28
Pulse Spreading Due to Group Delay Dispersion As discussed above, the optical materials comprising the microscope optics cause the excitation pulses to spread in time and thus become less efficient in exciting multi-photon transitions. The group delay dispersion (GDD) has been measured for some objectives (Guild et aI., 1997; Squier et aI., 1998). As mentioned above, the optical effort needed to generate the prechirping necessary to compensate for the GDD has to be weighed against the improvements expected. As a general rule, GDD compensation will be helpful or even essential when laser power is limiting, such as for deep tissue imaging, or when single-photon absorption contributes to damage. If coherent control techniques are used, complete dispersion control is, of course, essential but then the optics used to tailor phases can be employed for dispersion control as well.
Control of laser Power For slow control of the laser power, mechanically actuated devices such as filter wheels, graded neutral density filters, or rotating halfwave-plate/polarizer combinations (Denk, 200l) can be used. Faster shuttering (e.g., in order to blank the beam during retrace) or modulation requires non-mechanical devices such as acoustooptical (AO) or electro-optical (EO) modulators (i.e., Pockels cells), which can respond on the microsecond and even nanosecond timescale (Chapter 3, this volume). EOMs achieve high throughput but often incomplete extinction, while AOMs are lossy, and due to limited diffraction efficiency, their extinction is excellent. A few problems arise specifically when ultrashort pulses are used together with such devices: (1) In AOM, AOD, or acoustooptic tuning filter (AOTF) devices, an acoustic wave diffracts the incoming beam by an angle that is dependent on A,X" For ultrashort pulses, which are spectrally broad, the focus, therefore, becomes distorted and diffraction efficiency is reduced. (2) Both EOMs and AOMs use dispersive materials, which spread the laser pulse temporally (see above). While the temporal spread can be easily compensated (in a few cases multi-photon microscopy setups already contain GDD compensators), it is much more difficult but not impossible to compensate for the angular spread in an AOM (Lechleiter et al., 2002). Limited extinction from the EOMs is often not a serious problem because the quadratic intensity dependence of two-photon excitation allows even a moderate power reduction ratio to translate into almost complete elimination of unwanted excitation.
Resonance and Non-Mechanical Scanning The time resolution of closed loop galvanometer scanners is sufficient for most applications, in particular if a limited number of measurement points can be selected. However, because the time per line cannot be reduced significantly below the about I ms with closed-loop scanners, scan times for large areas can become too long for the time resolution desired. One solution to this problem is the use of resonant galvanometer scanner (Fan et aI., 1999; see also Chapters 3 and 29, this volume) which provide a fixed line rate about 10 times faster, albeit at some loss of flexibility. Acoustooptical scanning (Art and Goodman, 1993) requires correction of the diffractive spread of the wavelengths comprising short-pulse light (Lechleiter et aI., 2002), but has the advantage of more rapid access (still limited by the acoustic transit time across the diffraction medium) and allows both scanning and intensity control.
543
CHROMOPHORES (FlUOROPHORES AND CAGED COMPOUNDS) The criteria for choosing, or designing, fluorophores for MPM are essentially the same as for any other fluorescence microscopy technique: large absorption cross-section at convenient A"s, high quantum yield, low rate of photobleaching, and minimal chemical or photochemical toxicity to living cells. In the early days of MPM, a heuristic approach prevailed and fluorophores were selected that had proven useful in widefield fluorescence microscopy or I PCM. In most cases, two-photon excitation was found whenever there is single-photon absorption at a A corresponding to twice the energy of the excitation photons. Most MPM imaging still uses conventional fluorophores, and we now have two-photon spectra for many of these (Xu and Webb, 1996; Xu et al., 1996; Zipfel et al., 2003). On the other hand, there is a considerable effort to generate chromophores tailored to MP excitation using a donor-acceptor-donor or acceptor-donor-acceptor strategy. These molecules maximize the electrical dipole transition by electron transfer over relatively long distances from donor to acceptor. By this approach, molecules can be created with two-photon excitation cross-sections about 10fold greater than conventional fluorophores (Albota et aI., 1998b; Ventelon et aI., 1999). Nanoparticles, also called quantum dots (Bruchez et aI., 1998; Han et al., 2001), which offer broad excitation spectra, but very narrow emission spectra, have the largest measured two-photon cross-sections seen to date. This allows their detection at very low concentrations, even in vivo (Larson et al., 2003). Another notable development is the movement to longer wavelengths. While in the early days of MPM the emphasis was on UVexcited dyes that were 2P-excited by red lasers, the emphasis now is on fluorophores normally excited by visible light and 2P-excited by IR light. This trend is mainly driven by the desire for lower background fluorescence and deeper penetration into scattering tissue.
Two-Photon Absorption Cross-Sections Differences between one- and the two-photon excitation spectra have been exploited in molecular spectroscopy because they provide additional information about the structure of excited states. These differences can be quite significant, see, for example, the case of Bis-MSB (Kennedy and Lytle, 1986) or the aromatic amino acids tyrosine and phenylalanine (Rehms and Callis, 1993), but note the spectral similarities for tryptophan. As a rule of thumb, in symmetrical molecules one expects A2ex < 2A". Calculations of two-photon cross-sections are difficult to perform for complex molecules. Direct experimental measurements of multi-photon absorption are equally difficult because even under optimal conditions, the fraction of the incident power that is absorbed is rather small (using Eq. 4 we find, e.g., P'h/P = 3 x 10-' for a chromophore with a cross-section of 10-'0 m 4 s photon-', at a concentration of 10mM and a laser power of 100mW with a two-photon advantage of 10'). While thermal lensing or acousto-optical techniques have been used to measure two-photon absorption (Kliger, 1983), these techniques are much more complicated than single-photon spectrophotometry. For fluorescent molecules, the shape of the two-photon excitation spectrum can be determined by detecting the intensity of fluorescence emission as a function of excitation wavelength. In order to determine the action spectrum, the incident average laser power (P,), the probability of detecting fluorescence photons, and the two-
544
Chapter 28 • W. Denk et al.
photon advantage S (Eq. 2) need to be known (Xu et ai., 1995). The absolute value of the two-photon absorption cross-section can then be calculated using the fluorescence quantum yield. Quite a number of measured spectra are now available in the literature (Xu et aI., 1996; Albota et aI., 1998b also includes URL.) While precise calculations of two-photon absorption crosssections are difficult, several new fluorophores with particularly large two-photon absorption cross-sections have been designed using theoretical considerations (He et ai., 1995, 1997; Marder et aI., 1997; Albota et aI., 1998a; Ventelon et aI., 1999, 2002; Adronov et ai., 2000; Kim et aI., 2000; Zojer et ai., 2002). Before such fluorophores can come into common use, however, problems with water solubility, derivatization, etc., will have to be solved. For the fluorophores studied so far, the spectra of the emitted fluorescence were found to be essentially independent of whether excitation occurs via single- or two-photon excitation (Curley et aI., 1992). This is not surprising because the molecular relaxation process (on the picosecond scale) almost always occurs to the same state (the lowest excited singlet state) prior to the emission (on the nanosecond scale) and therefore erases the memory of the excitation pathway and energy.
Caged Compounds Two-photon absorption spectra for caged compounds are more difficult to measure than those for fluorophores because the amount of uncaged material generated is too small to be easily measured with most analytical techniques. In some cases, uncaging can be detected when fluorescence assays for the released agonist exist, such as for caged ATP (Denk et aI., 1990), when the product itself is fluorescent, as it is with caged fluorescein (Svoboda et al., 1996b), or when biological effects can be detected, such as the opening of ion channels by the two-photon-induced release of caged neurotransmitters (Denk, 1994; Matsuzaki et aI., 2001; Kasai et aI. , 2002). Photochemical reactions are often much slower than fluorescence emission and their speed can strongly depend on the chemical environment such as pH and ionic strength (Milburn et al., 1989; Corrie and Trentham, 1993; Kao and Adams, 1993). The speed of release is important for at least two reasons: (1) The pixel dwell-time must be at least as long as the duration of the signal used to generate image contrast, which at best is as fast as the photochemical reaction rate; (2) diffusion of the released agonist tends to blur the image and thus prevents high-resolution mapping. A delay of lOms, for example, allows the released agonist, typically a small organic molecule with a diffusion constant of 5 x 10-9 m2 S - I , to diffuse a distance of about 31.1m (Kiskin et al. , 2002).
CEll VIABILITY DURING IMAGING The survival of the biological sample while it is being imaged is one of the most important constraints on the usefulness of any vital microscopy technique. While one of the reasons for pursuing MPM as a new technique was the expectation of greatly reduced photodamage (Denk et aI., 1990), it has to be kept in mind that in the focal plane, for a given excitation rate the damage is expected to be at least as large for 2P as it is for 1P excitation. This is because any effect due to reactions initiated from the excited state of the chromophore are independent of the mode of excitation. Furthermore, it cannot be ruled out that some endogenous biological molecules have unusually large two-photon cross-sections (such as bacteriorhodopsin; Birge and Zhang, 1990) and are, therefore, particularly susceptible to damage. Another concern is the possibility
of excited state absorption, particularly at excitation rates near saturation. Considerable work has been performed in this area since the first edition of this book. Two-photon excitation, particularly when using wavelengths below 800nm (Konig et ai., 1996; Oehring et al., 2000) (see Chapter 38, this volume) can, not surprisingly, generate reactive oxygen species, which are implicated frequently in photodamage (Tirlapur et al., 2001). On the other hand, when using longer wavelengths (1064 nm), generation of reactive oxygen species by flavin-containing proteins seems to be greatly reduced compared to single-photon excitation (Hockberger et aI. , 1999). At higher excitation levels, a steeper than quadratic power dependence is often found both for cellular photodamage (Koester et al., 1999; Oehring et al., 2000; Hopt and Neher, 2001) and for photobleaching (Eggeling et aI., 1998; Patterson and Piston, 2000). It appears, however, that the damage nonlinearity is not instantaneous (i.e., three- or four-photon excitation) because for the same mean two-photon excitation rate no change in the damage is seen with pulse width (Koester et aI., 1999; Konig et aI., 1999). There is virtually no experimental indication that heating by water absorption (discussed in Physical Principles) is a limiting factor in multi-photon microscopy. Heating may yet become an issue as substantially longer wavelengths are beginning to be used for the excitation of long wavelength fluorophores. A number of explicit examples show an actual and significant reduction of photodamage when using two-photon rather than single-photon imaging in biological specimens such as cultured cells (Hockberger et al., 1999), cardiac myocytes (Niggli et at., 1994b; Piston et aI., 1994), and mammalian (Squirrell et aI., 1999) and invertebrate embryos (Summers et aI., 1993). The experience of many a microscopist is that live-cell imaging can often be performed by reducing the excitation light intensity to the lowest possible level, using efficient optics and sensitive detectors (Chapters 17, 19, and 29, this volume). The experience in 2PM is similar, but the range of imageable specimens is larger. For example, in both the sea urchin (Piston et aI., 1993) and hamster embryos (Squirrell et aI., 1999), two-photon excitation allows extended observation of embryonic development, under conditions where single-photon excitation is unsuccessful. In another case, as part of a direct comparison of scanned laser UV and two-photon excitation (Niggli et aI., 1994a; Piston et aI., 1994), it was found that two-photon excitation allowed imaging of the calcium indicator dye lndo-l continuously for 5 min without compromising cell viability. Equivalent single-photon scanning with UV light resulted in considerable photobleaching, and over 80% cell death (Piston et aI., 1994). Those studies indicate that, even though damage is less than with conventional UV illumination, cultured-animal-cell viability can be compromised by two-photon excitation. Particularly worrying, and as yet unresolved, is the observation that at high illumination levels the two-photon photobleaching rate can increase much faster than the excitation rate (Patterson and Piston, 2000), even though it is not known whether there is a corresponding increase in phototoxicity and whether these highly nonlinear bleaching phenomena are limited to certain narrow classes of dyes, such as the xanthene dyes. A question that often arises is how to determine the mechanism of damage. Important information is provided by its power dependence (Neuman et aI., 1999; Hopt and Neher, 2001). For example, two-photon photochemical damage should be proportional to the square of the incident power. While a linear power dependence all but rules out two-photon effects, a superlinear dependence on the average excitation power could result from single-photon absorp-
Multi-Photon Molecular Excitation in Laser-Scanning Microscopy • Chapter 28
tion coupled with a nonlinear mediator for damage. Thermally induced damage can have a rather sharp temperature threshold due to cooperative phenomena such as protein denaturation. A definitive distinction between single- and multi-photon absorption is their dependence on pulse length; if the pulse length is varied by introducing a variable degree of GDD (see above), the spectrum, and hence the amount of linear (single-photon) absorption, remains completely unchanged while 2PA drops. Knowing the mechanism of damage is, of course, crucial for choosing the optimal excitation strategy. For example, to reduce single-photon, dose-rate-independent damage, a reduction of Fp might seem appropriate in order to increase the two-photon advantage but the peak temperature during each pulse increases as F/~ " and can become larger than the thennal time constant. Unpleasant surprises could also arise from additional absorption by molecules already in the excited state (something that is more likely to occur when operating closer to saturation) or from proximity effects mediated by free radicals (Konig et al., 1996; Hockberger et ai., 1999; Koester et ai., 1999; Konig et a!., 1999; Oehring et ai., 2000; Hopt and Neher, 2001; Tirlapur et al., 2001).
APPLICATIONS MPM has been used to address questions in quite a few areas of biology. Particularly the imaging of intact tissue has benefited from the properties of the multi-photon (predominantly two-photon) microscope.
Calcium Imaging Intracellular messenger dynamics, such as calcium ion concentration has been measured in single cells (Piston et ai., 1994), but the particular advantages of MPM over single-photon techniques come to bear most in scattering tissue such as brain slices (Denk et al., 1995, 1996; Yuste and Denk, 1995; Mainen et ai., 1999b; Sabatini and Svoboda, 2000; Wang et ai., 2000; Oertner et ai., 2002), the stomatogastric ganglion (Kloppenburg et ai., 2000), and in vivo (Svoboda et al., 1997, 1999; Debarbieux et aI., 2003). In isolated retina 2PM allowed the recording of dendritic calcium signals during visual stimulation (Denk and Detwiler, 1999; Euler et aI., 2002).
Uncaging and Photobleaching Multi-photon photochemistry has been used to map receptor sensitivities in single cells (Denk, 1994) and inside neural tissue (Matsuzaki et ai., 2001; Kasai et al., 2(02).
Autofluorescence Because MPM easily reaches into UV transition energies, it has increasingly been used to study biological autofluorescence such as from NADH (Piston et ai. , 1995 ; Piston and Knobel, 1999), serotonin in living cells (Maiti et ai. , 1997), skin (Masters et ai. , 1997), muscle cells (Schilders and Gu, 1999), glutathione in arabidopsis (Meyer and Fricker, 2000), mast cell secretion using 3P excitation of serotonin (Williams et ai., 1999), arctic fungus (Arcangeli et ai., 2(00), collagen (Agarwal et aI., 2001), biofilm (Neu et al., 2(02), tryptophan in proteins (Lippitz et ai., 2(02). and flavoproteins (Huang et ai., 2002). Recently, the sources of autofluorescence from living tissue have been analyzed in more detail (Zipfel et ai., 2003) (see also Chapter 21, this volume).
545
Developmental Biology Because of the superior depth penetration and the localized excitation associated with MPM. this approach has proven useful in many developmental biological applications. Lineage tracing has been perfonned using two-photon photorelease of caged fluorophores in sea urchin embryos (Summers et al., 1996; Piston et ai., 1998). Cellular and subcellular dynamics have been imaged and measured using MPM during development of sea urchin embryos (Summers et al., 1993, 1996), cell fusion in C. elegans (Mohler et aI. , 1998; Periasamy et ai., 1999), mammalian embryos (Squirrell et ai., 1999), zebrafish (Huang et al.. 200 I), and birds (Dickinson et al. , 2(02).
In Vivo (Intact Animal) Imaging In intact animals, the need for tissue penetration is maximal. High resolution optical imaging inside living whole animals has therefore become the almost exclusive domain of two-photon microscopy not only for functional calcium imaging (Svoboda et ai., 1997, 1999), but also to image blood flow in the fine capillaries (Kleinfeld et ai. , 1998; Chaigneau et ai., 2003), gene expression and angiogenesis (Brown et aI., 200 I), and even the dynamics of Alzheimer's disease pathologies (Christie et aI., 1998, 1999, 2001; Backskai et al. , 200 I) and, having previously been applied to observe changes in dendrite structure in brain slices (Engert and Bonhoeffer, 1999; Maletic-Savatic et ai., 1999). two-photon microscopy has most recently been used to study the long-term dynamics of neuronal fine structure (Grutzendler et at., 2002; Trachtenberg et ai., 2002) in living animals.
OUTLOOK Multi-photon excitation microscopy has extended the range of laser scanning fluorescence microscopy especially where dynamic imaging in living specimen s is needed. Much progress has been made in solving many of the technical impediments that existed in the early days of MPM. Still, only a few of the many potential contrast mechanisms established for nonlinear optical spectroscopy have been used for imaging purposes. This is mainly due to the fact that often only a small number of photons can be collected from each volume element in the small amount of time that the beam dwells on each location. Increasing use is being made, however, of second hannonic generation (Moreaux et ai., 200) and Raman scattering (Zumbusch et al., 1999; Potma et ai., 2002; Volkmer et aI., 2002) (see also Chapter 33, this volume.) Phototoxicity in cells is still not well understood in general and for ultrashort pulse illumination in particular. But the main limitation to even more widespread use of multi-photon excitation is not due to fundamental physical, chemical, or biological problems, but to the price and complexity of the instrumentation.
ACKNOWLEDGMENTS The authors' research underlying this chapter was sponsored by the Developmental Resource for Biophysical Imaging and Optoelectronics at Cornell University (NIH-9 P41 EBOOl976 and NSF-DIR-8800278), the Material Science Center Computing Facility (NSF-DMR-9l21564), other grants from NIH (DK53434, CA86283) and NSF (BIR-98-71063), Lucent Technologies, Bell Labs, and the Max-Planck Society.
546
Chapter 28 • W. Denk et al.
REFERENCES Adronov, A., Frechet, J.M.J. , He, G.S., Kim, K.S., Chung, S.1., Swiatkiewicz, J., and Prasad, P.N., 2000, Novel two-photon absorbing dendritic structures, Chem. Mater. 12:2838- 284l. Agarwal, A., Coleno, M.L., Wallace, VP., Wu, W.Y, Sun, C.H., Tromberg, B.1., and George, S.c., 2001 , Two-photon laser scanning microscopy of epithelial cell-modulated collagen density in engineered human lung tissue, Tissue Eng. 7:191-202. Akaaboune, M., Grady, R.M., Turney, S., Sanes, J.R., and Lichtman, J.W. , 2002, Neurotransmitter receptor dynamics studied in vivo by reversible photo-unbinding of fluorescent ligands, Neuron 34:865-876. Albota, M. , Beljonne, D ., Bredas, J.L., Ehrlich, J.E. , Fu, J.Y., Heikal, A.A. , Hess , S.E., Kogej, T., Levin, M.D. , Marder, S.R., McCord-Maughon, D. , Perry, J.W., Rockel, H., Rumi, M., Subramaniam, c., Webb, W.W., Wu, X.L., and Xu, c., 1998a, Design of organic molecules with large twophoton absorption cross sections, Science 281: 1653-1656. Albota, M.A., Xu, C., and Webb, W.W., 1998b, Two-photon fluorescen ce excitation cross sections of biomolecular probes from 690 to 960nm, Appl. Opt. 37:7352-7356. http://www.drbio.comell .edulInfrastructure/ FluorescentProbes_WWW/CommerciallyAvailable.htm; http://www. drbio.comell .edu/InfrastructurelN on li nearM icroscopies_WWW/vit.htm; http://www.drbio.comell.edu/lnfrastructure/NonlinearMicroscopies_ WWW/3P/3PE.htm). Andresen, V , Egner, A., and Hell, S.W. , 2001 , Time-multiplexed multifocal multiphoton microscope, Opt. Lett. 26:75-77 . Arcangeli, c., Yu, W., Cannistraro, S., and Gratton, E., 2000, Two-photon autofluorescence microscopy and spectroscopy of antarctic fungus: New approach for studying effects of UV-B irradiation, Biopolymers 57: 218-225. Art, J.J ., and Goodman, M.B., 1993, Rapid-scanning confocal microscopy, Methods Cell Bioi. 38:47-77 . Ashkin, A. , Dziedzic, J.M ., and Yamane, T, 1987, Optical trapping and manipulation of single cells using infrared laser beams, Nature 330:769771. Backskai, B.1. , Kajdasz, S.T, Christie, RH. , Carter, c. , Garnes, D. , Seubert, P., Schenk, D. , and Hyman, B.T. , 2001 , Imaging of amyloid-beta deposits in brains of living mice permits direct observation of clearance of plaques with immunotherapy, Nat. Med. 7:369-372. Beaurepaire, E., and Mertz, J., 2002, Epifluorescence collection in two-photon microscopy, Appl. Opt. 41:5376-5382. Beaurepaire, E., Oheim, M., and Mertz, J., 200 I, Ultra-deep two-photon fluorescence excitation in turbid media, Opt. Commun. 188:25-29. Bindhu, C.V., Harilal, S.S. , Kurian, A., Nampoori, VP.N., and Vallabhan, c. P.G., 1998, Two and three photon absorption in rhodamine 6G methanol soluti ons using pulsed thermal lens technique, 1. Nonlinear Opt. Phys. Maler. 7:531-538. Birge, R.R., 1979, A theoretical analysis of the two-photon properties of linear polyenes and the visual chromophores, 1. Chem. Phys. 70: 165-177. Birge, RR. , 1986, Two-photon spectroscopy of protein-bound chromophores, Acc. Chem. Res. 19:138-146. Birge, R.R., and Zhang, c.-F., 1990, Two-photon double resonance spectroscopy of bacteriorhodopsin. Assignment of the electronic and dipolar properties of the low-lying 'A': -like and 'B'"+like n,n' states, 1. Chem. Phys.92:7178-7195. Booth, M.1. , and Hell, S.W., 1998, Continuous wave excitation two-photon fluorescence microscopy exemplified with the 647-nm ArKr laser line, 1. Microsc. 190:298-304. Brakenhoff, G.J. , Squier, J. , Norris, T , Bliton, A.C., Wade, M.H., and Athey, B., 1996, Real-time two-photon confocal microscopy using a femtosec· o nd, amplified Ti : sapphire system, 1. Microsc. 181 :253-259. Brown, E.B., Campbell, R.B. , Tsuzuki, Y , Xu , L. , Carmeliet, P., Fukumura, D., and Jain , R.K., 2001, In vivo measurement of gene expression, angiogenes is and physiological function in tumors using multi photon laser scanning microscopy, Nat. Med. 7:866- 870. Bruchez, M. Jr. , Moronne, M. , Gin, P., Weiss, S., and Alivisatos, A.P., 1998, Semiconductor nanocrystals as fluorescent biological labels, Science 281:2013-2016.
Chaigneau, E., Oheim, M., Audinat, E., and Charpak, S., 2003, Two-photon imaging of capillary blood fl ow in olfactory bulb glomeruli, Proc. Na tl. Acad. Sci. USA 100:13081-13086. Cheung, E.C., and Liu, J.M. , 1991, Efficient generation of ultrashort, wavelenght-tunable infrared pulses, J. Opt. Soc. Am. B 8:1491 - 1506. Christie, R.H., Bacskai, B.1., Zipfel, WR, Williams, RM., Kajdasz, S.T, Webb, W.W., and Hyman , B.T , 2001, Growth arrest of individual senile plaques in a model of Alzheimer 's disease observed by in vivo multi photon microscopy, 1. Neurosci. 21 :858- 864. Christie, RH., Zipfel, WR., Williams, R.M., Webb, WW , and Hyman, B.T, 1998, Multiphoton imaging of Alzheimer 's disease neuropathology, J. Neuropathol. Exp. Neurol. 57:145. Christie, R.H., Zipfel, WR., Williams, R.M., Webb, WW, and Hyman, B.T, 1999 , In vivo multiphoton imaging of amyloid deposition in transgenic mice, 1. Neuropathol. Exp. NeU/vl. 58:204. Corrie, J.E.T, and Trentham, D.R., 1993, Caged nucleotides and neurotransmitters, Bioorgan. PhOiocilem. 2:243-305. Curley, P.F., Ferguson, A.I. , White, J.G., and Amos, W.B., 1992, Application of a femtosecond self-sustaining mode-locked Ti-sapphire laser to the field of laser scanning confocal microscopy, Opt. Quantum Electron. 24:851-859. Davidovits, P. , and Egger, M.D. , 1969, Scanning laser microscope, Nature 233:83 1. Debarbieux, F., Audinat, E., and Charpak, S., 2003, Action potential propagation in dendrites of rat mitral cell s in vivo, J. Neurosci. 23:55535560. Denk, W., 1994, Two-photon scanning photochemical microscopy - Mapping ligand-gated ion-channel distributions, Proc. Natl. Acad. Sci. USA 91:6629-6633. Denk, W., 1996, Two-photon excitation in functional biological imaging, 1. Biomed. Opt. 1:296- 304. Denk, W. , 2001, Optical beam power controller using a tiltable birefringent plate, US Patent no. 6249379. Denk, W., and Detwiler, P.B. , 1999, Optical recording of light-evoked calcium signal s in the functionally intact retina, Proc. Natl. Acad, Sci. USA 96:7035-7040. Denk, W., and Svoboda, K., 1997, Photon upmanship: Why multiphoton imaging is more than a gimmick, Neuron 18:35 1-357. Denk, W., Aksay, E., Baker, R. , and Tank, D.W. , 1997, Long term imaging of [Ca++j in zebrafish embryos using two-photon microscopy, Soc. Neurosci. Abstr.646. Denk, W., Delaney, K.R., Gelperin, A., Kleinfeld, D., Strowbridge, B.W., Tank, D.W., and Yuste, R, 1994, Anatomical and functional imaging of neurons using two-photon laser scanning microscopy, 1. Neurosci. Methods 54: 151-162. Denk, W., Strickler, J.H., and Webb, W.W., 1990, Two-photon laser scanning fluorescence microscopy, Science 248:73- 76. Denk, W., Sugimori , M., and Llinas. R., 1995, Two types of calcium response limited to single spines in cerebellar Purkinje cells, Proc. Na tl. Acad. Sci. USA 92:8279-8282. Denk, W., Yuste, R., Svoboda, K.. and Tank, D.W, 1996, Imaging calcium dynamics in dendritic spines, Cu rr. Opin. Neurobiol. 6:372- 378. Dickinson, M.E., Murray, B.A. , Haynes, S.M., Waters, C.W., and Longmuir, K.J. , 2002, Using electroporation and lipid-mediated transfection of GFPexpressing plasmids to label embryonic avian cells for vital confocal and two-photon microscopy, Differentiation 70: 172-180. Dyba, M., and Hell, S.W., 2002, Focal spots of size lambdal23 open up farfield fluorescence microscopy at 33 nm axial resolution, Phys. Rev. Lell. 88: 163901. Eggeling, c., Widengren, J., Rigler, R., and Seidel, C.A.M., 1998, Photobleaching of fluorescent dyes under conditions used for single-molecu le detection: Evidence of two-step photolysis, Anal. Chem. 70:2651 - 2659. Egner, A., and Hell, S.W., 2000, Time multiplexing and paralleli zation in mul· tifocal multiphoton microscopy, 1. Opl. Soc. A 17:1192- 120 1. Egner, A., Jakobs, S., and Hell, S.W., 2002, Fast 100-nm resolution lhreedimensional microscope reveals structural plasticity of mitochondria in live yeast, Proc. Nat!. Acad. Sci, USA 99:3370-3375. Engert, F., and Bonhoeffer, T , 1999, Dendritic spine changes associated with hippocampal long-term synaptic plasticity, Nature 399:66- 70.
Multi-Photon Molecular Excitation in Laser-Scanning Microscopy • Chapter 28
Euler, T, Detwiler, P,B" and Denk, W" 2002, Directionally selective calcium signals in dendrites of starburst amacrine cells, Nature 418:845-852. Fan, G.y', Fujisaki, H., Miyawaki, A, Tsay, R.K., Tsien, R.Y., and Ellisman, M.H., 1999, Video-rate scanning two-photon excitation fluorescence microscopy and ratio imaging with cameleons, Biophys. 1. 76:24122420. Fittinghoff, D.N., Schaffer, CB., Mazur, E .. and Squier, J.A., 2001, Timedecorrelated multifocal micromachining and trapping, IEEE l. Select Topics Quantum Electron. 7:559-566. Fork, R.L., Martinez, O.E., and Gordon, J.P., 1984, Negative dispersion using pairs of prisms, Opt. Lett. 9: 150-152. Friedrich, D.M., and McClain, W.M., 1980, Two-photon molecular spectroscopy, Ann. Rev. Phys. Chern. 31 :559-577. Fu, Q., Mak, G., and van Oriel, H.M., 1992, High-power, 62fs infrared optical parametric oscillator synchronously pumped by a 76-MHz Ti:sapphire laser, Opt. Lett. 17:1006-1010. Furuta, T., Wang, S.S.H., Dantzker, J.L., Dore, T.M., Bybee, W.1., Callaway, E.M., Denk, W., and Tsien, R.Y., 1999, Brominated 7-hydroxycoumarin4-ylmethyls: Photolabile protecting groups with biologically useful crosssections for two photon photolysis, Proc. Natl. Acad. Sci. USA 96: 1193-1200. Gauderon, R., and Sheppard, C.1.R., 1999, Effect of a finite-size pinhole on noise performance in single-, two-, and three-photon confocal fluorescence microscopy, Appl. Opt. 38:3562-3565. Goeppert-Mayer, M., 1931, Ueber Elementarakte mit zwei Quantenspruengen, Ann. Phys. 9:273. Goldstein, S.R., Hubin, T, and Smith, TG., 1992, An improved no-movingparts video-rate confocal microscope, Microsc. Microanal. 23:437-446. Gosnell, TR., and Taylor, A.1., eds., 1991, Selected Papers on Ultrafast Laser Technology, SPTE, Bellingham. Grutzendler, J., Kasthuri, N., and Gan, W.B., 2002, Long-term dendritic spine stability in the adult cortex, Nature 420:812-816. Guild, J.B., Xu, C., and Webb, WW., 1997, Measurement of group delay dispersion of high numerical aperture objective lenses using two-photon excited fluorescence, Appl. Opt. 36:397-401. Han, M.Y., Gao, X.H., Su, J.z., and Nie, S., 2001, Quantum-dot-tagged microbeads for multiplexed optical coding of biomolecules, Nat. Biotechno/. 19:631-635. Hanninen, P.E., Lehtela, L., and Hell, S.W., 1996, Two- and multi photon excitation of conjugate-dyes using a continuous wave laser, Opt. Commun. 130:29-33. Hanninen, P.E., Soini, E., and Hell, S.W., 1994, Continuous-wave excitation two-photon fluorescence microscopy, l. Microsc. 176:222-225. He, G.S., Xu, G.C, Prasad, P.N., Reinhardt, B.A., Bhatt, J.C., and Dillard, A.G., 1995, Two-photon absorption and optical-limiting properties of novel organic-compounds, Opt. Lett. 20:435-437. He, G.S., Yuan, L.X., Cheng, N., Bhawalkar, J.D., Prasad, P.N., Brott, L.L., Clarson, S.J., and Reinhardt, B.A., 1997, Nonlinear optical properties of a new chromophore, l. Opt. Soc. Am. B 14: 1079-1087. Hell, S., and Stelzer, E.K.H., 1992, Fundamental improvement of resolution with a 4pi-confocal fluorescence microscope using two-photon excitation, Opt. Commun. 93:277-282. Hell, S., Reiner, G., Cremer, C, and Stelzer, E.K.H., 1993, Aberrations in confocal fluorescence microscopy induced by mismatches in refractive index, 1. Microsc. 169:391-405. Hell, S.W., and Andresen, Y, 2001, Space-multiplexed multifocal nonlinear microscopy, l. Microsc. 202:457-463. Hell, S.W., Booth, M., Wilms, S., Schnetter, CM., Kirsch, AK., Arndt-Jovin, D.1., and Jovin, TM., 1998, Two-photon near- and far-field fluorescence microscopy with continuous-wave excitation, Opt. Lett. 23:1238-1240. Hellwarth, R., and Christiansen, P., 1974, Nonlinear optical microscopy examination of structure in polycrystalline ZnSe, Opt. Commun. 12:318322. Helmchen, F., Fee, M.S., Tank, D.W., and Denk, W., 2001, A miniature headmounted two-photon microscope. High-resolution brain imaging in freely moving animals, Neuron 3\:903-912. Helmchen, F., Tank, D.W., and Denk, W., 2002, Enhanced two-photon excitation through optical fiber by single-mode propagation in a large core, Appl. Opt. 41 :2930-2934.
547
Hockberger, P.E., Skimina, TA., Centonze, YE., Lavin, C, Chu, S., Dadras, S., Reddy, J.K., and White, J.G., 1999, Activation of flavin-containing oxidases underlies light-induced production of H 20, in mammalian cells, Proc. Natl. Acad. Sci. USA 96:6255-6260. Hopt, A., and Neher, E., 2001, Highly nonlinear photodamage in two-photon fluorescence microscopy, Biophys. l. 80:2029-2036. Huang, H., Vogel, S.S., Liu, N., Melton, D.A., and Lin, S., 2001, Analysis of pancreatic development in living transgenic zebrafish embryos, Mol. Cell Endocrinol. 177: 117-124. Huang, S.H., Heikal, A.A., and Webb, W.W., 2002, Two-photon fluorescence spectroscopy and microscopy of NAD(P)H and flavoprotein, Biophys. l. 82:2811-2825. Kaiser, W., and Garrett, CB.G., 1961, Two-photon excitation in CaF2 : Eu'+, Phys. Rev. Lett. 7:229-231. Kao, J.P.K., and Adams, S.R., 1993, Photosensitive caged compounds, In: Optical Microscopy, Emerging Methods and Applications (B. Herman andJ.J. Lemasters, eds.),Academic Press, San Diego, California, pp. 27-85. Kasai, H., Matsuzaki, M., and Ellis-Davies, G.CR., 2002, Two-photon mapping of functional glutamate receptors in dendritic spines of hippocampal CAl pyramidal neurons, lpn. l. Pharmacol. 88:S22. Keller, u., 1994, Ultrafast all-solid-state laser technology, Appl. Phys. B Lasers Opl. 58:347-363. Keller, U., 1996, Materials and new approaches for ultrashort pulse lasers, Curro Opin. Solid State Mater. Sci. 1:218-224. Kennedy, S.M., and Lytle, F.E., 1986, p-Bis(i-methylstyryl)benzene as a powersquared sensor for two-photon absorption measurements between 537 and 694nm, Anal. Chern. 58:2643-2647. Kim, O.K., Lee, K.S., Woo, H.Y., Kim, K.S., He, G.S., Swiatkiewicz, J., and Prasad, P.N., 2000, New class of two-photon-absorbing chromophores based on dithienothiophene, Chern. Mater. 12:284. Kiskin, N.l., Chillingworth, R., McCray, J.A, Piston, D., and Ogden, D., 2002, The efficiency of two-photon photolysis of a "caged" fluorophore, 0-1-(2nitrophenyl)ethylpyranine, in relation to photodamage of synaptic terminals, Eur. Biophys. 1. Biophys. Lett. 30:588-604. Kleinfeld, D., Mitra, P.P., Helmchen, F., and Denk, W., 1998, Fluctuations and stimulus-induced changes in blood flow observed in individual capillaries in layers 2 through 4 of rat neocortex, Proc. Natl. A cad. Sci. USA 95:15741-15746. Kliger, D.S.E., 1983, Ultrasensitive Laser Spectroscopy, Academic Press, New York. Kloppenburg, P., Zipfel, WR., Webb, WW, and Harris-Warrick, R.M., 2000, Highly localized Ca2+ accumulation revealed by multiphoton microscopy in an identified motoneuron and its modulation by dopamine, l. Neurosci. 20:2523-2533. Koester, H.1., and Sakmann, B., 1998, Calcium dynamics in single spines during coincident pre- and postsynaptic activity depend on relative timing of back-propagating action potentials and subthreshold excitatory postsynaptic potentials, Proc. Nat!. A cad. Sci. USA 95:9596-9601. Koester, H.J., Baur, D., Uhl, R., and Hell, S.W., 1999, Ca2+ fluorescence imaging with pico- and femtosecond two-photon excitation: Signal and photodamage, Biophys. l. 77:2226-2236. Konig, K., Becker, TW, Fischer, P., Riemann, 1., and Halbhuber, K.J., 1999, Pulse-length dependence of cellular response to intense near-infrared laser pulses in multiphoton microscopes, Opt. Lett. 24: 113-115. Konig, K., Kimel, S., and Berns, M.W., 1996, Photodynamic effects on human and chicken erythrocytes studied with microirradiation and confocal laser scanning microscopy, Lasers Surg. Med. 19:284-298. Lansford, R., Bearman, G., and Fraser, S.E., 2001, Resolution of multiple green fluorescent protein color variants and dyes using two-photon microscopy and imaging spectroscopy, 1. Biomed. Opt. 6:311-318. Larson, D.R., Zipfel, WR., Williams, R.M., Clark, S.w., Bruchez, M.P., Wise, F.W., and Webb, W.W., 2003, Water-soluble quantum dots for multiphoton fluorescence imaging in vivo, Science 300:1434--1436. Lechleiter, J.D., Lin, D.T, and Sieneart, 1., 2002, Multi-photon laser scanning microscopy using an acoustic optical deflector, Biophys. l. 83:22922299. Lippitz, M., Erker, W., Decker, H., van Holde, K.E., and Basche, T, 2002, Twophoton excitation microscopy of tryptophan-containing proteins, Proc. Natl. A cad. Sci. USA 99:2772-2777.
548
Chapter 28 • W. Denk et al.
Liu, T.M., Chu, S.w., Sun, e.K., Lin, B.L., Cheng, P.e., and Johnson, 1., 2001, Multiphoton confocal microscopy using a femtosecond Cr:forsterite laser, Scanning 23:249-254. Loudon, R., 1983, The Quantum Theory of Light, Oxford University Press, New York. Lozovoy, v.v., Pastirk, 1., Walowicz, K.A., and Dantus, M., 2003, Multiphoton intrapulse interference. II. Control of two- and three-photon laser induced fluorescence with shaped pulses, 1. Chern. Phys. 118:3187-3196. Maiman, T.H., 1960, Stimulated optical radiation in ruby, Nature 187:493-494. Mainen, Z.F., Maletic-Savatic, M., Shi, S.H., Hayashi, Y., Malinow, R., and Svoboda, K., 1999a, Two-photon imaging in living brain slices, Methods 18:231-239. Mainen, Z.F., Malinow, R., and Svoboda, K., 1999b, Synaptic calcium transients in single spines indicate that NMDA receptors are not saturated, Nature 399:151-155. Maiti, S., Shear, J.B., Williams, R.M., Zipfel, w.R., and Webb, w.w., 1997, Measuring serotonin distribution in live cells with three-photon excitation, Science 275:530-532. Maletic-Savatic, M., Malinow, R., and Svoboda, K., 1999, Rapid dendritic morphogenesis in CAl hippocampal dendrites induced by synaptic activity, Science 283:1923-1927. Marder, S.R., Torruellas, W.E., BlanchardDesce, M., Ricci, v., Stegeman, GJ., Gilmour, S., Bredas, J.L., Li, J., Bublitz, G.U., and Boxer, S.G., 1997, Large molecular third-order optical nonlinearities in polarized carotenoids, Science 276: 1233-1236. Masters, B.R., ed., 2003, Selected Papers on Multi-Photon Excitation Microscopy, SPIE, Bellingham. Masters, B.R., So, P.T.e., and Gratton, E., 1997, Multiphoton excitation fluorescence microscopy and spectroscopy of in vivo human skin, Biophys. 1. 72:2405-2412. Matsuzaki, M., Ellis-Davies, G.e.R., Nemoto, T., Miyashita, Y., lino, M., and Kasai, H., 2001, Dendritic spine geometry is critical for AMPA receptor expression in hippocampal CAl pyramidal neurons, Nat. Neurosci. 4: 1086-1092. McClain, W.M., 1971, Excited state symmetry assignment through polarized two-photon absorption studies of fluids, 1. Chern. Phys. 55:2789-2796. Meyer, A.J., and Fricker, M.D., 2000, Direct measurement of glutathione in epidermal cells of intact Arabidopsis roots by two-photon laser scanning microscopy, 1. Microsc. 198:174-181. Milburn, T., Matsubara, N., Billington, A.P., Udgaonkar, J.B., Walker, J.W., Carpenter, B.K., Webb, w.w., Marque, J., Denk, W., McCray, J.A., and Hess, G.P., 1989, Synthesis, photochemistry, and biological-activity of a caged photolabile acetylcholine-receptor ligand, Biochemistry 28:4955. Mohler, W.A., Simske, J.S., Williams-Masson, E.M., Hardin, J.D., and White, J.G., 1998, Dynamics and ultrastructure of developmental cell fusions in the Caenorhabditis elegans hypodermis, Curr. BioI. 8: 1087-1090. Moreaux, L., Sandre, 0., Charpak, S., Blanchard-Desce, M., and Mertz, J., 2001, Coherent scattering in multi-harmonic light microscopy, Biophys. 1.80:1568-1574. Mortensen, O.S., and Svendsen, E.N., 1981, Initial and final molecular states as "virtual" states in two-photon processes, 1. Chern. Phys. 74:3185-3189. Muller, M., Squier, J., Wolleschensky, R., Simon, U., and Brakenhotf, G.J., 1998, Dispersion pre-compensation of 15 femtosecond optical pulses for high-numerical-aperture objectives, 1. Microsc. 191:141-150. Nakamura, 0., 1993, Three-dimensional imaging characteristics of laser scan fluorescence microscopy - two-photon excitation vs single-photon excitation, Optik 93:39-42. Neu, T.R., Kuhlicke, U., and Lawrence, J.R., 2002, Assessment of fluorochromes for two-photon laser scanning microscopy of biofilms, App!. Environ. Microbiol. 68:901-909. Neuman, K.e., Chadd, E., Liou, G.F., Brau, A., Bergman, K., and Block, S.M., 1999, Characterization of photodamage induced by optical traps, Biophys. 1. 76:A96. Nielsen, T., Frick, M., Hellweg., D., and Andresen, P., 2001, High efficiency beam splitter for multifocal multiphoton microscopy, 1. Microsc. 201 :368-376. Niggli, E., Piston, D.W., Kirby, M.S., Cheng, H., Sandison, D.R., Webb, w.w., and Lederer, W.J., 1994, A confocal laser scanning microscope designed
for indicators with ultraviolet excitation wavelengths, Am. 1. Physiol. 266:C303-C310. Oehring, H., Riemann, 1., Fischer, P., Halbhuber, K.1., and Konig, K., 2000, Ultrastructure and reproduction behaviour of single CHO-Kl cells exposed to near-infrared femtosecond laser pulses, Scanning 22:263270. Oertner, T.G., Sabatini, B.L., Nimchinsky, E.A., and Svoboda, K., 2002, Facilitation at single synapses probed with optical quantal analysis, Nat. Neurosci. 10: 1O. Ouzounov, D.G., Moll, K.D., Foster, M.A., Zipfel, W.R., Webb, W.W., and Gaeta, A.L., 2002, Delivery of nanojoule femtosecond pulses through large-core microstructured fibers, Opt. Lett. 27:1513-1515. Patel, e.K.N., and Tam, A.C., 1981, Pulsed optoacoustic spectroscopy of condensed matter, Rev. Mod. Phys. 53:517-550. Patterson, G.H., and Piston, D.W., 2000, Photobleaching in two-photon excitation microscopy, Biophys. 1. 78:2159-2162. Periasamy, A., Skoglund, P., Noakes, C., and Keller, R., 1999, An evaluation of two-photon excitation versus confocal and digital deconvolution fluorescence microscopy imaging in Xenopus morphogenesis, Microsc. Res. Technol.47:172-181. Piston, D.W., and Knobel, S.M., 1999, Real-time analysis of glucose metabolism by microscopy, Trends Endocrino!. Metab. 10:413-417. Piston, D.W., Kirby, M.S., Cheng, H.P., Lederer, W.J., and Webb, w.w., 1994, Two-photon-excitation fluorescence imaging of 3-dimensional calciumion activity, App!. Opt. 33:662-669. Piston, D.W., Masters, B.R., and Webb, W.W., 1995, 3-dimensionally resolved Nad(P)H cellular metabolic redox imaging of the in-situ cornea with twophoton excitation laser-scanning microscopy, 1. Microsc. 178:20-27. Piston, D.W., Sandison, D.R., and Webb, W.W., 1992, Time-resolved fluorescence imaging and background rejection by two-photon excitation in laser scanning microscopy, Proc. SPIE. 1640:379-389. Piston, D.W., Summers, R.G., and Webb, W.W., 1993, Observation of nuclear division in living sea urchin embryos by two-photon fluorescence microscopy, Biophys. 1. 63:AIIO. Piston, D.W., Summers, R.G., Knobel, S.M., and Morrill, J.B., 1998, Characterization of involution during sea urchin gastrulation using two-photon excited photorelease and confocal microscopy, Microsc. Microanal. 4:404-414. Potma, E.O., Jones, D.1., Cheng, J.X., Xie, X.S., and Ye, J., 2002, Highsensitivity coherent anti-Stokes Raman scattering microscopy with two tightly synchronized picosecond laser, Opt. Lett. 27:1168-1170. Powers, P.E., Tang, e.L., and Cheng, L.K., 1994, High-repetition-rate femtosecond optical parametric oscillator based on Cstioaso4, Opt. Lett. 19: 37-39. Rehms, A.A., and Callis, P.R., 1993, Two-photon fluorescence excitationspectra of aromatic-amino-acids, Chern. Phys. Lett. 208:276-282. Sabatini, B.L., and Svoboda, K., 2000, Analysis of calcium channels in single spines using optical fluctuation analysis, Nature 408:589-593. Sandison, D.R., and Webb, W.W., 1994, Background rejection and signal-tonoise optimization in confocal and alternative fluorescence microscopes, Appl. Opt. 33:603-615. Schilders, S.P., and Gu, M., 1999, Three-dimensional autofluorescence spectroscopy of rat skeletal muscle tissue under two-photon excitation, Appl. Opt. 38:720-723. Schonle, A., and Hell, S.W., 1998, Heating by absorption in the focus of an objective lens, Opt. Lett. 23:325-327. Sheppard, C.1.R., and Cogswell, e.J., 1990, Confocal microscopy with detector arrays, 1. Mod. Opt. 37:267-279. Sheppard, C.J .R., and Cogswell, C,J., 1991, Effects of aberrating layers and tube length on confocal imaging properties, Optik 87:34-38. Sheppard, C.1.R., and Gu, M., 1990, Image-formation in two-photon fluorescence microscopy, Optik 86: 104-106. Sheppard, e.J.R., and Gu, M., 1994, Imaging performance of confocal fluorescence microscopes with finite-sized source, 1. Mod. Opt. 41: 1521-1530. Sheppard, e.J.R., and Kompfner, R., 1978, Resonant scanning optical microscope, Appl. Opt. 17:2879-2882. Singh, S., and Bradley, L.T., 1964, Three-photon absorption in napthalene crystals by laser excitation, Phys. Rev. Lett. 12:612. Spence, D.E., Kean, P.N., and Sibbett, W., 1991, 60-fsec pulse generation from a self-mode-Iocked Ti:sapphire laser, Opt. Lett. 16:42-44.
Multi-Photon Molecular Excitation in Laser-Scanning Microscopy • Chapter 28
Squier, J.A., Fittinghoff, D.N., Barty, CP.J., Wilson, K.R, Muller, M., and Brakenhoff, CJ., 1998, Characterization of femtosecond pulses focused with high numerical aperture optics using interferometric surface-thirdharmonic generation, Opt. Commun. 147: 153-156. Squirrell, J.M., Wokosin, D.L., White, J.G., and Bavister, B.D., 1999, Longterm two-photon fluorescence imaging of mammalian embryos without compromising viability, Nat. Biotechnol. 17:763-767. Stelzer, E.H.K., Hell, S., and Lindek, S., 1994, Nonlinear absorption extends confocal fluorescence microscopy into the ultra-violet regime and confines the illumination volume, Opt. Commun. 104:223-228. Straub, M., and Hell, S.W., 1998, Multifocal multi photon microscopy: A fast and efficient tool for 3-D fluorescence imaging, Bioimaging 6:177-185. Summers, R.G., Morrill, 1.B., Leith, A., Marko, M., Piston, D.W., and Stonebraker, A.T, 1993, A stereometric analysis of karyokinesis, cytokinesis and cell arrangements during and following 4th cleavage period in the sea-urchin, Lytechinus variegatus, Dev. Growth Diff. 35:41-57. Summers, R.G., Piston, D.W., Harris, K.M., and Morrill, J.B., 1996, The orientation of first cleavage in the sea urchin embryo, Lytechinus variegatus, does not specify the axes of bilateral symmetry, Dev. Bioi. 175: 177-183. Svoboda, K., and Block, S.M., 1994, Biological applications of optical forces, Ann. Rev. Biophys. Biomol. Struct. 23:247-285. Svoboda, K., Denk, W., Kleinfeld, D., and Tank, D.W., 1997, In vivo dendritic calcium dynamics in neocortical pyramidal neurons, Nature 385: 161-165. Svoboda, K., Denk, W., Knox, W., and Tsuda, S., 1996a, Two-photon laser scanning fluorescence microscopy of living neurons using a diode-pumped Cr: LiSrAIFllaser mode-locked with a saturable Bragg reflector, Opt. Lett. 21: 1411-1413. Svoboda, K., Helmchen, F., Denk, W., and Tank, D.W., 1999, Spread of dendritic excitation in layer 2/3 pyramidal neurons in rat barrel cortex in vivo, Nat. Neurosci. 2:65-73. Svoboda, K., Tank, D.W., and Denk, W., I 996b, Direct measurement of coupling between dendritic spines and shafts, Science 272:716-719. Tan, YP., Llano, I., Hopt, A., Wurriehausen, F., and Neher, E., 1999, Fast scanning and efficient photodetection in a simple two-photon microscope, 1. Neurosci. Methods 92:123-135. Theer, P., Hasan, M.T., and Denk, W., 2003, Two-photon imaging to a depth of 1000/lm in living brains by use of a Ti: Al,O, regenerative amplifier, Opt. Lett. 28: 1022-1024. Tirlapur, U.K., Konig, K., Peuckert, C., Krieg, R., and Halbhuber, KJ., 2001, Femtosecond near-infrared laser pulses elicit generation of reactive oxygen species in mammalian cells leading to apoptosis-like death, Exp. Cell Res. 263:88-97. Trachtenberg, J.T, Chen, B.E., Knott, G.W., Feng, G., Sanes, J.R., Welker, E., and Svoboda, K., 2002, Long-term in vivo imaging of experiencedependent synaptic plasticity in adult cortex, Nature 420:788-794. Valdmanis, 1.A., and Fork, R.L., 1986, Design considerations for a femtosecond pulse laser balancing self phase modulation, group-veloci ty dispersion, saturable absorption, and saturable gain, IEEE 1. Quant. Electron. 22:112-118. Ventelon, L., Moreaux, L., Mertz, J., and Blanchard-Desce, M., 1999, New quadrupolar fluorophores with high two-photon excited fluorescence, Chem. Commun. 20:2055-2056. Ventelon, L., Moreaux, L., Mertz, J., and Blanchard-Desce, M., 2002, Optimization of quadrupolar chromophores for molecular two-photon absorption, Synth. Met. 127: 17-21. Visser, TD., Brakenhoff, GJ., and Groen, F.CA., 1991, The one-point fluorescence response in confocal microscopy, Optik 87:39-40.
549
Volkmer, A., Book, L.D., and Xie, X.S., 2002, Time-resolved coherent antiStokes Raman scattering microscopy: Imaging based on Raman free induction decay, Appl. Phys. Lett. 80:1505-1507. Walowicz, K.A., Pastirk, I., Lozovoy, v.v., and Dantus, M., 2002, Multiphoton intrapulse interference. I. Control of multi photon processes in condensed phases, J. Phys. Chem. A 106:9369-9373. Wang, S.S.H., Denk, W., and Hausser, M., 2000, Coincidence detection in single dendritic spines mediated by calcium release, Nat. Neurosci. 3:1266-1273. Whinnery, J.R., 1974, Laser measurement of optical absorption in liquids, Acc. Chem. Res. 7:225-231. Williams, R.M., Shear, J.B., Zipfel, WR., Maiti, S., and Webb, WW, 1999, Mucosal mast cell secretion processes imaged using three-photon microscopy of 5-hydroxytryptamine autofluorescence, Biophys. J. 76:1835-1846. Williams, S.A., and Callis, P.R., 1990, Two-photon electronic spectra of nucleotides, Proc. SPIE 1204:332-343. Wilson, T, and Sheppard, C, 1984, Theory and Practice of Scanning Optical Microscopy, Academic Press, New York. Wokosin, D.L., Centonze, v., White, J.G., Armstrong, D., Robertson, G., and Ferguson, A.I., 1996a, All-solid-state ultrafast lasers facilitate multiphoton excitation fluorescence imaging, IEEE J. Quant. Electron. 2: 1051-1065. Wokosin, D.L., Centonze, V.E., Crittenden, S., and White, J., 1996b, Threephoton excitation fluorescence imaging of biological specimens using an all-solid-state laser, Bioimaging 4: 1-7. Wolleschensky, R., Feurer, T, Sauerbrey, R., and Simon, I., 1998, Characterization and optimization of a laser-scanning microscope in the femtosecond regime, Appl. Phys. B Lasers Opt. 67:87-94. Xu, C, and Denk, W., 1997, Two-photon optical beam induced current imaging through the backside of integrated circuits, Appl. Phys. Lett. 71: 2578-2580. Xu, C, and Denk, W., 1999, Comparison of one- and two-photon optical beaminduced current imaging, J. Appl. Phys. 86:2226-2231. Xu, C, and Webb, W.W., 1996, Measurement of two-photon excitation cross sections of molecular fluorophores with data from 690 to 1050nm, J. Opt. Soc. Am. B 13:481-491. Xu, C, Guild, J., Webb, W.W., and Denk, W., 1995, Determination of absolute two-photon excitation cross-sections by in-situ 2nd-order autocorrelation, Opt. Lett. 20:2372-2374. Xu, C., Zipfel, W., Shear, J., Williams, R., and Webb, W, 1996, Multiphoton fluorescence excitation: New spectral windows for biological nonlinear microscopy, Proc. Natl. Acad. Sci. USA 93: 10763-10768. Yuste, R., and Denk, W., 1995, Dendritic spines as basic functional units of neuronal integration, Nature 375:682-684. Zhang, Q., Piston, D.W., and Goodman, R.H., 2002, Regulation of corepressor function by nuclear NADH, Science 295:1895-1897. Zipfel, W.R., Williams, R.M., Christie, R., Nikitin, A.Y, Hyman, B.T., and Webb, W.W., 2003, Live tissue intrinsic emission microscopy using multiphoton-excited native fluorescence and second harmonic generation, Proc. Natl. Acad. Sci. USA 100:7075-7080. Zojer, E., Beljonne, D., Kogej, T, Vogel, H., Marder, S.R., Perry, J.W., and Bredas, J.L., 2002, Tuning the two-photon absorption response of quadrupolar organic molecules, J. Chem. Phys. 116:3646-3658. Zumbusch, A., Holtom, G.R., and Xie, X.S., 1999, Three-dimensional vibrational imaging by coherent anti-Stokes Raman scattering, Phys. Rev. Lett. 82:4142-4145.
29
Multifocal Multi-Photon Microscopy Jorg Bewersdorf, Alexander Egner, and Stefan W. Hell
INTRODUCTION Multi-photon processes relying on the cooperative action of two or more photons can broadly be divided into two families that are distinguished by the fact that the photons are either absorbed or scattered (Shen, 1984). Whereas the scattering events relevant to microscopy are second and third harmonic generation (SHG, THG), as well as coherent anti-Stokes Raman scattering (CARS), the useful multi-photon absorption events are two- and threephoton excitation (2PE, 3PE). The first multi-photon phenomenon that entered microscopy was SHG (Hell warth and Christensen, 1974; Gannaway, 1978), followed by CARS (Duncan et ai., 1982), 2PE (Denk et al., 1990; Curley et ai. , 1992), 3PE (Hell et al., 1996; Maiti et ai., 1997), and THG (Barad et ai., 1997; Muller et al., 1998). Timed with the advent of more accessible pulsed lasers, the seminal work by Denk and colleagues (1990) on 2PE microscopy opened a new epoch of research and application with multi-photon processes in microscopy (Guo et ai., 1997; Gauderon et al., 1998; Zumbusch et at., 1999; Campagnola et ai., 2002; Cheng et ai., 2002; Muller and Schins, 2002; Yelin et ai., 2002; Theer et ai., 2003; Zipfel et ai., 2003). The use of multi-photon phenomena provides several advantages over their single-photon counterparts. The most prominent is the confinement of signal generation to the focal region where the simultaneous occurrence of mUltiple photons is highest. Another important advantage is the capability to penetrate deeper into strongly scattering specimens (Denk and Svoboda, 1997; Centonze and White, 1998). Moreover, SHO, THO, and CARS (Zoumi et ai., 2002; Cox et ai., 2003) generate signals that are not accessible through single-photon interactions, thus complementing fluorescence imaging in a unique way. Unfortunately, multi-photon events have a low probability of occurrence, that is, they have a small cross-section. Small crosssections can be compensated by large excitation intensities. In microscopy, the strong focusing provided by the objective lens readily yields large intensities, in particular in conjunction with pulsed illumination. The only issue is that the applicable intensity is limited by photodamage, which also has a major nonlinear component (see Chapter 38, this volume). In some cases, multiphoton absorption may also reach (singlet-state) saturation. With the exception of the important application of imaging into strongly scattering tissue, the power of presently available lasers usually greatly exceeds the power required at a given point. Therefore, the use of several parallel foci may be regarded as an obvious solution to this problem. In multi-photon microscopy, this solution is particularly attractive because the optical sectioning is provided by the multi-photon interaction alone. No back-imaging onto an array of
pinholes is needed, which otherwise would require delicate alignment and the compensation of chromatic aberrations. In this chapter, we give an overview of parallelized multi-photon imaging methods, which are commonly referred to as multi-focal multiphoton microscopy (MMM).
Background Owing to their wavelength tunability, short pulse length, and high repetition rate, mode-locked titanium: sapphire (Ti : Sa) lasers have become the light sources of choice for multi-photon microscopes. Mode-locked Ti: Sa and similar laser systems typically provide I to 2 W of average power at a repetition rate of -80 MHz at pulse lengths of -200fs or 1 to 2 ps. This is ample light for a single scanning beam since nonlinear damaging effects normally limit the usable intensity to about 2000W/cm' at 200fs and 700W/cm 2 at 1 to 2ps in the focus (Hanninen et ai., 1995; Konig et al., 1996, 1999; Hopt and Neher, 2001). At typical repetition rates and focal spot sizes, this maximum focal intensity amounts to 3 to IOmW at 200fs and 10 to 30mW in the picosecond range average power. An important exception is the imaging of layers inside strongly scattering specimens, such as skin and brain at >250)lm depth, where most of the laser light is needed and femtosecond operation is preferable. Therefore, in regular, single-spot multi-photon microscopy, more than 90% of the laser power is discarded because applying more power would be detrimental. This holds both for the multi-photon absorption and the multi-photon scattering microscopy modes. By splitting up the beam of a mode-locked Ti : Sa laser into several beamlets and applying multiple, well-separated foci simultaneously, MMM exploits a much larger fraction of the available laser power, and at the same time it paral\elizes the imaging process without significant trade-offs in the resolution (Bewersdorf et ai., 1998; Buist et ai., 1998).
Determination of the Optimum Degree of Parallelization If photodamage, photobleaching, or saturation of an excited state of the chromophore can be neglected, the signal S from a single focus n-photon excitation microscope, per time unit, is proportional to crsPa",/'t"-'j"-', with P.'8 being the average laser power in the focus, 1: andfbeing the pulse length in the sample and the repetition rate, respectivel y. as is the multi-photon cross-section. In MMM, the laser beam is split up into N beam lets with an average power of P"" KMMM/N each. The signal of the N independent foci adds up to an overall signal S oc crSP~'N,MMM/('tfN)"- '. Within the framework of sheer signal generation, the parameters 1f and N are of equal
J6rg Bewersdorf, Alexander Egner, and Stefan W. Hell. Max Plan ck Institute for Biophysical Chemistry, 37070 G6ttingen, Germany 550
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC , New York, 2006.
Multifocal Multi-Photon Microscopy • Chapter 29
importance and therefore the change of one parameter can be compensated by adjusting one of the others. This can be illustrated by looking at the laser pulse train at a certain spot in the sample. The number of pulses arriving per second is proportional to j times N. Whether the repetition rate j is halved and N is doubled or vice versa is of no importance. A doubled pulse length 't can similarly be interpreted as two subsequent pulses. While 't and f are given by the laser system, the degree of parallelization N introduces a new degree of freedom to optimize the performance of a multiphoton microscope. It has to be noted though that N strongly influences the microscope design and thus can be changed only in a certain range without major technical modifications. The choice of the parameters for MMM depends on the limiting factors: saturation, photodamage, and the available laser power. Saturation obviously does not playa role for the scattering modes because in this case no long-lived state of the sample is involved. The overall damage per time unit can be written as a polynomial series Doc ·'2.J);P:VK.MMM/('tjN)i-l , with 8; expressing ;= 1
the relative weight (including the damaging cross-sections) of the damaging mechanisms of the different orders of non-linearity. For a certain range of laser and imaging parameters Pm'", 't, j and N , D can be approximated by D oc crDP;,gMMMI('tjN)"-', where d is the effective order of non-linearity that typically is not an integer. dis close to the order of non-linearity of the dominating damaging mechanism which can change, for example, with the applied laser power Pm'g. Similarly, cro is the effective damaging cross-section in this parameter range. As a result, the performance ratio ~ of the signal S to the damage D is proportional to crslcr,lPm 'K, M.1MI'tjN)"-d = crslcr/)P;',~~~. The goal obviously is to maximize the performance ~. For this purpose, one has to distinguish between two different situations:
• n > d (the excitation process is of higher order of nonlinearity than the dominating damaging process): Maximizing the peak power Ppm' yields the highest value for 13. Short pulses and low repetition rates are therefore favorable . Parallelization only decreases ~. However, an increase of ppm. is only reasonable up to a value where damaging processes of higher order become significant. • n < d (the excitation process is of lower order of nonlinearity than the dominating damaging process): p,,,", mu st be minimized to optimize ~. Apart from applying long pulses and high repetition rates, parallelization is the best alternative. Moreover, by increasing the overall average power Pm'K,M"'M and N simultaneously, ~ can be kept constant while at the same time the recorded signal per unit time S increases by a factor of N. This allows an acceleration of the imaging speed by this factor without increasing the damage. The maximum N is limited by the available laser power only as long as no low order damaging processes (such as heating) become dominant. In the case of n = d, ~ does not depend on the peak power. Therefore, parallelization or a change in 't or j has no real influence. We note that the distance between the focal spots and the size of the scanning field additionally influence the relative weights 8; of the damaging processes. Heating may be a problem if all of the average power is concentrated on a rather small scanning area of a few micrometers. With regard to the damage, parallelization is only reasonable in the case of a higher degree of non-linearity d of the dominant damage process as compared to that of the excitation process
551
(n > d). Another reason for the parallelization is enhanced scanning speed where parallelization is important even if n > d.
For the multi-photon excitation processes, the (rather rare case of) saturation is in the same way a highly nonlinear phenomenon. In this situation, a maximum acceptable saturation level can be defined with a corresponding focal average power Pm 30 f..lm, as can be seen in Figure 30.3(A-C) (Egner et ai., 2002a). • Slight differences in the refractive index of immersion and mounting medium lead to a z-dependent phase. For the refractive index mismatches typically encountered in the sample, it has been shown that this relationship is linear over a large distance (Egner et ai., 1995). If the magnitude of this distortion
is known, a linear change in the length of one of the 4Pi arms with changing z-position of the sample can correct for this phase shift during scanning. • Irregular changes of the refractive index within the specimen also influence the phase difference between the interfering wavefronts. To explore whether these conditions can be fulfilled, 3D stacks of (mammalian) Vera cells were recorded with water immersion lenses (Egner et ai., 2004). The cells were grown on a coverslip coated with Oregon Green and then covered with a similar coverslip so that they were essentially sandwiched between two ultrathin fluorescent layers. To correct for the difference in the refractive indices of the aqueous mounting medium 0.34) and water (1.33) used as immersion medium, the phase had to be adjusted linearly with the z-position. Figure 30.3(D) shows an xz-section of a typical 3D data stack containing a part of the nucleus. The xz-section through the layers on the coverslips and the resulting z-profile disclose the quality of the interference between the two wavefronts. Profiles are extracted from three different sites in the nuclear region, in which the local variations in refractive index are strongly pronounced: completely outside the nucleus, at the nuclear periphery, and straight through the nucleus [Fig. 30.3(E)]. Although slight changes become apparent, the phase does not notably change with the axial translation. This can be inferred from the comparison between the profiles in the same line, pertinent to the different coverslips. Comparison of the profiles from the same coverslip reveals that the presence of the nucleus introduces aberrations leading to elevated sidelobes, but again the changes in refractive index are not pronounced enough to distort the PSF of the 4Pi microscope. In the perinuclear region, the side lobe height is lower than the critical value of 50% of the central peak, but if the nucleus is directly involved, the lobes are slightly elevated. The upper profile of Figure 30.3(E) also shows that the refractive index challenges are relaxed when largely omitting the nucleus--cytosol interface.
Live Mammalian Cell 4Pi Imaging In order to further substantiate the applicability of 4Pi microscopy to living mammalian cells, the distribution of two Golgi marker enzymes labeled by EGPF inside living Vera cells have been imaged at 32°C with the MMM-4Pi microscope. The marker enzymes were UD P -galactosy I: gl ycoprotein, 1,4-13- galactosy 1 transferase (GalT), which is highly enriched in the middle and trans Golgi membranes and 3'-phosphoadenylyl-sulfate:uronyl-2O-sulfotransferase (2-0ST) which is a cis Golgi marker. The analysis of series of subsequent 3D recordings of the GaIT-EGFP distribution in a live Vera cell at 32°C with the MMM4Pi showed that the movement of the living Golgi apparatus does not pose particular challenges. The slight morphological changes observed between two subsequent recordings occur on a much larger time scale than the scanning ofthe 4Pi-PSF across each position in the cell. Thus, they did not compromise the space invariance of the PSF and the slight changes did not preclude the proper deconvolution of the data to achieve a 3D resolution of -100 nm. In addition, the repeated recording of the Golgi apparatus did not show a significant reduction in image brightness. Therefore, photobleaching of EGFP was not an issue in these samples. The cells remained viable even after prolonged mounting in the custom-built chamber for several days. The comparatively large space of 175 f..lm between the two coverslips and the addition of a suitable air/C0 2 mix is adequate to maintain division and propagation without apparent degradation.
4Pi Microscopy • Chapter 30
565
1
i\l'
~
0 1
C
4Pi
i\l'
I
~
\
/
\
/
' - cant.
0
0
-0.5
0 .0
0.5
z [11m]
32 11m
:51·· · ·~ · · · · · · · ·· · · . · · . ·t · · · · ·· ·Il
O. 0
::~l 0 .0
*(1,
,...A
,
,
ni. FP
~
,.,
i
n F,
'
,
..
,
,
,6 . . ,
•
. . k · · ··· · · · ················~ ···· ······· ,
a~
1.0~
I
4
~
~:~::=·:···:···:··~··:==~·:···i o
2
4
14
16
18
z [11m] FIGURE 30.3. Basics of MMM-4Pi imaging. (A) xy image of a fluorescent polydiacetylene monolayer taken from a 3D stack. The chessboard appearance is due to the different intensities in the neighboring 4Pi foci; the tilted stripes are inherent to the polydiacetylene monolayer. (B) Axial intensity profiles, I,(z), through the 3D stack exhibiting a sharp maximum and two lobes, recorded at three different coordinates (x,y). The spatial invariance of the three profiles indicates that the 4Pi-PSF is constant o ve r the whole field of view, irrespective of the micro lens used. Fast linear one-step deconvolution can be applied throughout the fi eld of view to remove the sidelobes (C). (D) xz-section recorded at the periphery of the nucleus of living Vero cell sandwiched between two coverslips each thinly coated with a fluorescence layer. The nucle us (N) is highlighted by Hoechst counterstaining and outlined; the region where the Oolgi resides (0 ) is briefly sketched. The cell is significantly larger than the whole xz-section. The three vertical lines on either side (orange) are due to the interference between the counte rpropagating focused fields, representing the main maximum of the 4Pi -PSF and the two ax iall y offset sidelobes. (E) Enlarged intensity profiles of the axial responses of the microscope to the thin fluo rescent layers, revealing the height of the lobes and the relative phase in greater detail. Outside and at the edge of the nucleus, the primary lobes are SOOnm). In recent years significant improvements of axial resolution have been achieved through the coherent use of two opposing lenses as is realized, for example, in 4Pi microscopy (see Chapter 30, this volume). However, these systems
are still limited by diffraction. In this chapter, we report on concepts for radically overcoming the diffraction barrier and attaining a resolution of few nanometers: a figure that has been considered impossible with focused light for more than a century.
BREAKING THE DIFFRACTION BARRIER: THE CONCEPT OF REVERSIBLE SATURABLE OPTICAL FLUORESCENCE TRANSITIONS Since the inception of nonlinear optics (Shen, 1984), it has been speculated that a nonlinear relationship between the applied intensity and the measured (fluorescence) signal could - at least in principle - expand the resolution capabilities of a focusing (farfield) optical system. However, these notions remained vague and without consequence because a concrete physical recipe could not be given. In fact, the multi-photon processes that had initially been considered for significantly improving the spatial resolution turned out to be unsuitable. Therefore, it was not until the early 1990s that concrete physical concepts appeared for breaking the diffraction resolution barrier with focused light (Hell and Wichmann, 1994; Hell and Kroug, 1995; Hell, 1997). In fact these concepts can be viewed as a family of concepts that utilize reversible saturable optical (fluorescence) transitions, which we now name reversible saturable optical fluorescence transitions (RESOLFT). They can be described in a common formalism (Hell, 2003). Let us assume a fluorescent molecule with two distinct states A and B, whereby the transition from A --7 B can be optically induced at a rate kAB = (51 [Fig. 31.1 (A)]. The variables (5 and 1 denote the transition cross-section and the light intensity, respectively. The rate for the reverse transition B --7 A is denoted with k BA . It may be driven by light, by a chemical reaction, by heat, or any other means, or simply be spontaneous. The kinetics of the molecular states is described by dNAldt = -dNBldt = kBANB- kABNA, with N A .B denoting the normalized population probability of each state. After a time period t » (kAB + kBAt t , the populations have reached a dynamical equilibrium at N; = kBA l(kAB + kBA)' The molecule's probability to be in A or B basically depends on kAB and hence on I. At the saturation intensity 1'"1 = k BA I(5, we have equal probability N; = 112. Increasing 1 » l,al renders kAB » kBA' so that the molecule is virtually shifted to B: N; --7 O. Figure 31.1 (B) illustrates how this behavior can be exploited for creating arbitrarily sharp regions of state A molecules. The scheme in Figure 31.I(B) is one-dimensional (x), but can be
Stefan W. Hell, Katrin I. Willig, Marcus Dyba, Stefan Jakobs, Lars Kastrup, and Volker Westphal. Max Planck Institute for Biophysical Chemistry, 37070 Gbttingen, Germany
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
571
572
Chapter 31 • S.W. Hell et al.
A
Fluorescent
k",,= 01
Non·fluorescent
(Non-fl uorescent)
k..
(Fluorescent)
0 ...................~ ®
/(x)
c
1=10 I
c:::J
Absorbing state
Molecules left in state A Light that drives A-+ B
®
Scanning
x
FIGURE 31.1. Breaking the diffraction barrier by reversible saturable optical transitions (RESOLFT) requires (A) two states A and B of a label that are distinct in their optical properties. The optical transition A ~ B takes place at a rate kAB = crI that is proportional to the light intensity I applied. The reverse transition B ~ A of rate kBA brings the label back to its initial state. (B) The profiles I to 4 show the spatial region in which the label is allowed to be in state A, if the region is subject to a standing wave of light with peak intensities 10 = 10, 50, 100, and 500 times I,", and with a zero at Xi' Increasing 10 ensures that the region in which the label may reside in A is squeezed down, in principle, indefinitely. If A is the fluorescent state of the label, this ultrasharp region functions as the effective fluorescent spot of the microscope and ax is its FWHM. The creation of a fluorescence image requires scanning that is moving the zero along the xaxis with subsequent storage of the recorded fluorescence. If B is the fluorescent state, then the ultrasharp regions of state A are dark. In this case, a sort of negative image is recorded. Nevertheless, with suitable mathematical postprocessing, similar optical resolution can be obtained. In any case, the resolution is no longer limited by diffraction, but only determined by the value of IJIwa. (C) The simplified energy diagram of a fluorophore depicts possible schemes for implementing saturable optical transitions.
readily extended to three dimensions. For this purpose we require the spatial intensity distribution 1= I(x) to be zero at a point Xi in space: I(x;) = O. Whereas in two dimensions or three dimensions, the light intensity distribution would be a two-dimensional (2D) or three-dimensional (3D) doughnut mode, in one dimension the zero is best produced by a standing wave I(x) = Iocos 2(21tnxIA). If we now apply I(x) to a spatial distribution of molecules (in x) that are first in state A, then for 10 » Isa[, virtually all the molecules will be transferred to B, except for those that are very close to Xi' The larger the ratio I1Isat » I, the sharper is the region where state A persists [note the increase in curve steepness with increasing saturation level l/Isat in Fig. 31.1 (B) J. The FWHM of the resulting spot of state A is readily calculated as:
(2) In microscopy, the spatial distribution I(x) may be produced by the objective lens itself. If it is produced through the finite aperture of the objective lens, the smallest spot that can be obtained is Ax",
A nnsina..j 111m,
(3)
which may be regarded as an extension of Abbe's equation (Hell, 2003, 2004). In fact, one can easily show that if the zero is produced by focusing the light through the lens, the equation becomes:
(4) For I1I,a, = 100, the theoretical resolution improvement over Abbe is by about 10. Despite the dependence of Ax on Alsina and in contrast to Eq. 1, the new Eqs. 2, 3 and 4 allow diffractionunlimited spatial resolution.
For a 3D doughnut, we obtain a confined spatial volume of molecules in state A whose dimensions scale inversely with ..,fTJT;;;, i = x,y,z, with Ii denoting the peak intensities along the respective axes. Hence, the reduction in volume scales with ..j I x I y I z /Psat . In RESOLFT microscopy while the resolution still scales with the wavelength A, its limit only depends on the applicable light intensity I at a given 1m" If we now assume that state A but not state B is a fluorescent state, the relevance to imaging becomes obvious: our scheme allows us to create arbitrarily small fluorescence spots (Hell and Wichmann, 1994; Hell, 2003, 2004; Hell et ai., 2003). Moreover, by scanning the zero Xi across (or through) the specimen, we can record the fluorophore distribution point by point, and thus assemble a fluorescence 3D image with arbitrary resolution. Identical fluorescent objects can be imaged as separate in space irrespective of their proximity and size because the fluorescence spot (state A) can be made so small that only one of the objects fluoresces. The concept of RESOLFT inevitably requires scanning (with a zero), but not necessarily with a single beam or a point-like zero. Multiple zeros or dark lines produced by the interference of counter-propagating waves (Cragg and So, 2000; Heintzmann et al., 2002) in conjunction with conventional charge-coupled device (CCD) camera detection can also be used, provided the zeros or the dark lines are farther apart than about the distance required by the diffraction resolution limit of conventional CCD camera imaging (Cragg and So, 2000; Hell, 2003). Dark lines increase the resolution in a single direction only, but stepwise rotation of the pattern plus interleaved scanning of the minima (e.g., by shifting the phase in the interference pattern) and subsequent computational reassignment (Heintz mann and Cremer, 1998; Heintzmann et al., 2002) may provide, under some conditions, similar transverse resolution as with points and do so at higher recording speed. The obligation for scanning remains. The need for scanning is also
Nanoscale Resolution with Focused Light: STED and Other RESOLFT Microscopy Concepts • Chapter 31
the reason why the saturable optical transition A ~ B has to be reversible. The molecule in state B must be able to return to the state A at the latest when the zero comes across its site.
DIFFERENT APPROACHES OF REVERSIBLE SATURABLE OPTICAL FLUORESCENCE TRANSITIONS MICROSCOPY The RESOLFf scheme and the subsequent breaking of the diffraction barrier is the actual idea behind stimulated emission depletion (STED) microscopy (Hell and Wichmann, 1994; Hell, 1997; Klar et aI. , 2000). In STED microscopy, many of the molecules that have just been excited to the fluorescent state S I (A) are immediately transferred by a further light intensity I to the molecular ground state So (B), so that fluorescence emission is prevented [Fig. 31.1(B,C) and Fig. 31.2(A»). The physical effect responsible for this transfer is stimulated emission, a basic single-photon phenomenon that has about the same cross-section as singlephoton absorption (0 '" 1O- 16_ 1O- 18 cm 2 ). Because STED competes with the spontaneous fluorescence decay of kfl '" (1 nsf l of the S" the saturation intensity I,m can be approximated as kio or about 1025 to 1027 photons per square centimeter and second, that is, several tens to a hundred megawatts per square centimeter [Fig. 31.2(B)]. Saturated depletion of the excited state with a focal spot containing a zero squeezes the extent of the fluorescent spot to a subdiffraction size that is not any longer limited by the wavelength.
C
Energy levels
A
573
but only by the applied intensity. The potential and details of STED microscopy will be discussed later. Other variants of RESOLFf microscopy employing different physical realizations of states A and B have also been suggested [Fig. 31.1(C») (Hell et al.. 2003). For example. in ground state depletion (GSD) microscopy (Hell and Kroug. 1995 ; Hell, 1997), the grou nd state So has the role of state A , while state B is a metastable triplet state; more precisely, A and B are the singlet and triplet systems of the dye, respectively. For a number of dyes, intersystem crossing (A ~B) occurs as a byproduct of the regular dye excitation becau se during each excitation cycle, the molecule crosses to the triplet state with a probability p '" 0.05 to 0.2. Because of its metastability, the triplet state is relatively easily filled up by repeated regular excitation. For saturation, the triplet buildup rate kAB == pol must be larger than the decay to the So, kBA '" (10-6 to 10-2 sfl , which is comparatively slow. Thus, typical l sul is several tens of kilowatts per square centimeter, which is by 2 to 4 orders of magnitude lower than with STED. Low I,al is invaluable with regard to the attainable resolution (see Eq. 2) , and with regard to sample compatibility. However, the experimental realization of this member of the RESOLFf family will be complicated by the fact that the triplet state is involved in the photobleaching pathway (Schafer, 1973). Perhaps the simplest way of realizing a saturated optical transition is through an intense excitation (Heintzmann et aI., 2002) . In this case, the ground state So (A) is depleted and expected to reside in the fluorescent state SI (B). The same RESOLFf for-
5TED microscope
51 c:
= s 'u 0
CI
x UJ
en
15
::> ~ u:::
50 E c:
Saturated depletion of state 51
B
~
1.0
2l
dye: MR-121
0.8
c:
~ 0.6 • _ _ I
sat
g 0.4
u::
""
244nm
0.2 1 ........
:
~ "-
0.00 100 200 300 400 500 600
Intensity I MW
cm·2
Excitation STED Saturated spot + spot + depletion
FIGURE 31.2. STED microscopy. (Al Molecules in the fluorescent state S, (state Al return to the ground state So (state B) by spontaneous flu orescence emission. Return to So may also be optically enforced through stimulated emission. To prevai l over the spontaneous emission, stimulated emission depletion of the S, requires relatively intense light pulses with durations of a fraction of the S, lifetime. (B) Saturated depletion of the S , with increasing peak intensity of STED pulses of - lOOps duration, as measured by the remaining fluorescence of a monomolecular layer of the dye MR-121. For the higher intensity levels. the S[ is optically depleted. The saturation intensity is defined as the intensity value at which the S, is depleted by half. (el Center: Sketch of a point-scanning STED microscope. Excitation and STED are accomplished with synchronized laser pulses focused by a lens into the sample, sketched as green and red beams, respectively. Fluorescence is registered by a detector. Bottom: Intensity distributions in the focus. The diffraction-limited excitation spot is overlapped with the doughnut-shaped STED spot featuring a central zero. Saturated depletion by the STED beam confines the region of excited molecules to the zero, leav ing a flu orescent spot of subdiffraction dimensions. Outer left and right insets: the confocal and the subdiffraction-sized spot left by STED, respectively. Note the doubled lateral and S-fold improved axial resolution over confocal microscopy. The reduction in dimensions (x, y , zl yields an ultrasma ll volume of subdiffraction dimensions, here 0.67 attoliter, corresponding to an 18-fold reduction compared to its confocal counterpart. In spite of using diffraction-limited beams, the concept of STED fluore scence microscopy may, under very favorable conditions, reach spatial resolution at the molecular scale.
574
Chapter 31 • S.W. Hell et al.
malism applies, except that it is state B that is now fluorescent. State A is depleted such that ultrasharp dark regions are created that are surrounded by bright fluorescent regions. In a sense, this approach records negative images, which are subsequently made positive by mathematical postprocessing. The dark regions can be lines produced by interference patterns, but also 3D doughnuts. In the latter case, one would produce dark 3D volumes that are confined by walls of intense fluorescence. The challenges with this otherwise very appealing approach are that the mandatory computations require an excellent signal-to-noise ratio. As with STED, excitation saturation competes with fluorescence emission, so that I sa' is also given by kfllcr which is of the order of 1025 photons per cm 2 - second, several tens of megawatts per square centimeter (Hell, 2003). Compared to STED, the intensity needed for saturated optical transitions should actually be up to 10 times lower because the dye can be excited at the maximum of the emission spectrum where cross-sections are largest. Still, intense excitation increases photobleaching. Relief could possibly be brought by using non-blinking semiconductor quantum dots as labels (Alivisatos, 1996; Bruchez et aI., 1998; Peng et aI., 2000). Hence, while the family of RESOLFT concepts is not subject to Abbe's diffraction barrier anymore, the dependence of the resentails another soft barrier which is the olution gain on maximum intensity I that the sample can tolerate. Fortunately, the remedy is the use of transitions with low values of 1m " that is, optical transitions that are easy to saturate. An example is the optical switching of bistable compounds from a fluorescent state (A) into a non-fluorescent state (B), or vice versa. Optical bistability can be realized by photo-induced cistrans isomerization (Dyba and Hell, 2002; Hell et ai., 2003). If both states A and Bare (meta)stable, the optical transition A -7 B or B -7 A can be completed at very long, if not arbitrary, time scales. Thus, the light energy needed for these transitions can be spread in time (Hell et aI., 2003), reducing I.w , to values that are lower by many orders of magnitude compared to those of STED or the other RESOLFT family members. These processes would also readily lend themselves for parallelization through large area widefield imaging. However, the principal advantage is the insight that nanoscale resolution does not necessarily require extreme intensities of light (Hell, 2003, 2004; Hell et ai., 2003). Suitable candidates for this concept are the optically switchable fluorescent proteins. For example, the protein asFP595 can be switched on by green light (B -7 A) and also switched off (A -7 B) by blue light recurrently (Lukyanov et aI., 2000; Chudakov et aI., 2003). Although known proteins such as asF595 may have major limitations, such as a low quantum efficiency and a strong tendency to form oligomers, these problems could possibly be solved by strategic mutagenesis. Alternatively, new switch able proteins could be found by targeted exploration. A RESOLFT concept based on genetically encoded optically switchable tags is extremely appealing, because it would allow highly specific imaging in live cells with unprecedented optical resolution. We expect that further variants of RESOLFT will emerge in the future.
..JTTT:
STIMULATED EMISSION DEPLETION MICROSCOPY So far, STED microscopy is the only member of the RESOLFT schemes that has been realized (Klar et aI., 2000, 2001). In its initial demonstration, STED microscopy has been realized as a point-scanning system, whereby excitation and STED is performed with two synchronized ultrashort pulses [Fig. 31.2(C)].
The first pulse excites the molecule into the fluorescent state SI at a suitable wavelength. The red-shifted second pulse that follows a few picoseconds later transfers the molecules away from the zero back to the ground state So. Although 1m , is several tens of megawatts per square centimeter, it scales inversely with the pulse duration of 10 to 300ps. Because the breaking of the diffraction barrier calls for I » 1m " focal intensities of 100 to 500MW/cm 2 are required (Hell and Wichmann, 1994). For comparison, live-cell multi-photon microscopy typically uses 103 to 104 times shorter pulses of 103 to 104 greater intensity: 200 GW/cm2 • As most sample damage mechanisms depend nonlinearly on the intensity, the typical values used for STED so far have been live-cell compatible (Klar et aI., 2000). Figure 31.2(C) shows a typical experimental focal intensity distribution of the excitation spot (green), overlapping with a STED spot (red) featuring a central hole. Saturated depletion inhibits fluorescence everywhere except for the very center of the focal region (Klar et ai., 2000). For the llIsa ,"" 100 applied for the measurement in Figure 31.2, the net 3D spot becomes almost spherical with a diameter of -100 nm, which amounts to an almost 6-fold and 2.3-fold increase in axial and lateral resolution, respectively. Although theory permits much higher resolution (in principle, molecular scale), in this experiment the production of a smaller spot was challenged by experimental imperfections (Klar et aI., 2000), such as a finite depth of the central zero, and increased photobleaching with increasing IIIsat • The fact that the spot is squeezed more in the z-direction than in the focal plane is due to the higher local intensity of this particular quenching spot along the optic axis. Using a STED beam of different shape, an improvement of more than 5-fold in the focal plane resolution, compared to Abbe's barrier, has recently been demonstrated with single molecules dispersed on a surface (see Fig. 31.3 and Westphal et aI., 2003). Furthermore, Figure 31.3(B) proves that objects separated by much less than the diffraction limit can clearly be distinguished. Recent experiments also indicate that, in accordance with Equations 2 and 3, even higher lateral resolution is possible with STED, provided that photobleaching can be avoided. In fact, it has been shown that a focal spot width of 16 nm, corresponding to little more than 2% of the wavelength used can be obtained [see Fig. 3 I .7(A)] (Westphal and Hell, 2005). Because the diffraction barrier is broken, STED microscopy does not have a firm resolution limit. The ultimate resolution solely depends on how well the operational conditions can be realized. In frequency space, the sample structure is described in terms of spatial frequencies. Therefore, microscope performance is defined by the OTF (optical transfer function), describing the strength with which these frequencies are transferred to the image. Thus, the resolution limit is given by the highest frequency that produces a signal above the noise level. Point spread function (PSF) and OTF are intertwined by Fourier mathematics: the sharper the PSF, the broader the OTE The OTF extension of STED is very smooth [see Fig. 31.3(C)], without any gaps, simplifying the deconvolution techniques needed to produce the best final results. At present, the fact that relatively few fast, pulsed, tunable, visible lasers are available places some practical limits on the dyes that can be used for STED microscopy. The STED laser system must be able to produce two short laser pulses that follow one another in the picosecond time domain. The first pulse must be at a wavelength capable of exciting the dye, and the second, more powerful pulse must be at a wavelength capable of quenching it. Although early STED studies were confined to red-emitting dyes by the availability of appropriate lasers, this is no longer the case.
Nanoscale Resolution with Focused Light: STED and Other RESOLFT Microscopy Concepts • Chapter 31
A
B
PSF
C
Separation of molecules
OTF -
STEO
62 nm
4---
Convent. 222 nm
\
0.1
....
- f------.
'-t' -
~
"-
STEO
-
Mol. no.2
' ~,col venl 0.01
-200
-1 0 0
0 X
100
200
[nm]
0
-100 X
100
575
o
.......
=~ ....
~ 40
10
[nm ]
FIGURE 31.3. Quantifying lateral resolutio n in STED microscopy through imag ing of point-like objects. (A) The effective point spread function (PSF) of a conventional microscope and a (laser-diode) STED mi croscope, determined on single dye molecules (JA 26). (B) Molecules spaced apart far below the diffraction limit could be clearly separated in STED microscopy (slightly augmented by deconvolution), (C) For STED microscopy the gain in transmitted bandwidth of the optical transfer function (OTF) is more than 5-fold, compared with con ve ntional microscopy. Objecti ve lens, N A = 1.4 (oil ); wavelengths A., 635 nm (excitatio n). 650 to 720 nm (fluorescence detecti on), 781 nm (STED),
A list of the dyes used so far, is found in Table 31, I. This list is likely to be expanded as the number of pulsed-laser diodes increases and lasers become available that are capable of doing STED on blue, green, and yellow fluorophores , as well as fluorescent proteins. Studies aimed at identifying suitable laser/dye pairs are ongoing. Recently, a very compact STED microscope was demonstrated using a laser diode for the blue excitation and a second diode laser for STED at around 780 nm (Westphal et aI., 2003). As the set of available wavelengths expands over the next decade, it should in the future become possible to realize STED microscopy at lower cost and on most dyes. Although shorter wavelengths will lead to higher spatial resolution, a furth er increase in intensity may be barred in aqueous media by intolerable photobleaching. Saturation factors of >200 might not be readily attainable. STED microscopy is still in its infancy. So far, most of the applications have been aimed at exploring its principles. Tackling cell biology questions will be a task for the years to come. Strong fluorescence suppression (reduction by 90%) is conceptually not mandatory, but practically important to attain subdiffraction resolution. Among the first biological stains described that allows thi s level of suppression to be reached were lipophilic dyes such as Styryl 6, 7, and 8, 9M, LDS 751 , Pyridine 1,2,4 and RH 414, or Oxazine 170 and Nile Red, Because STED was first reali zed with a titanium: sapphire (Ti: Sa) laser emitting in the far-red (750800nm), the emission maximum of these dyes is located around 650 to 700nm (Klar et aI. , 2000), Pyridine 4 was used to label the membranes of live Escherichia coli. A simultaneous doubling of both the axial and lateral resolution was observed using a 3D doughnut and STED pulses of
-30ps duration . This initial improvement is likely to be augmented by further optimization of the wavelengths, of the doughnut, and of the pulse duration. Indeed, recent studies revealed that STED-related photobleaching dramatically decreases with the duration of the STED pulse, which indicates a strongly nonlinear dependence of bleaching on the STED-pulse intensity (Dyba and Hell, 2003). Although bleaching is substantially reduced with pulses > 120 ps duration, more studies are required to address this critical issue. In budding yeast, the dye RH 414 is taken up by bulk membrane internalization and subsequently transported to the vacuolar membrane. The structural integrity of vacuolar membranes is sensitive to many stress factors. Therefore, the subdiffraction resolution imagi ng of vacuoles in living yeast by STED microscopy confirmed that STED microscopy is amenable to imaging living cells (Klar et aI. , 2000). Recent progress in laser techno logy has enabled STED microscopy employing dyes that fluoresce in the visible range. These dyes offer several significant advantages: The shorter wavelengths involved inherently improve the Abbe resolution, while a better quantum efficiency [compared with near-infrared (NIR) dyes] leads to brighter images . Furthermore, the fact that they are visible by eye simplifies sample inspection and image region selection, As a first example, a dramatic enhancement of the resolution in the focal plane is demonstrated on yellow-green fluore scent microspheres (FluoSpheres 505/515, Molecular Probes, OR) in Figure 3 1A. Although the principle of STED applies to any fluorophore, new flu orescent markers require prior investigation. Sometimes the narrow spectral window for efficient STED depends on the chemical environment and has to be established through meticu-
TABLE 31.1. Listing of Example Dyes That Have Been Used Successfully for STED Category Green dye* Yell o w dye* Red dye* Far red dye Infrared dye
Name
Emission nm
STED nm
Manufacturer
Notes
Att05 32 DY-510XL Atto647N Pyridine 2 Pyridine 4
540-570 560- 630 650-720 680- 750 710- 800
615 625 760 750-780 780-800
Atto-tec GmbH , Siegen, DE Dyomics GmbH , l ena, DE Atto-tec GmbH , Siegen, DE SigmaAldrich, Sl. Louis, MO SigmaAldrich, Sl. Louis, MO
Used for single-molecule studies Immunofluorescence label Used for single-molecule studies Me mbrane label Me mbrane label
* These dyes have functional groups and can be coupled
to proteins making simultaneous, 3-color, immuno-ft uorescence STED imaging possible.
576
Chapter 31 • S.W. Hell et al.
c
o ....,
~ 75nm
~nm
,
o
,
600nm
FIGURE 31.4. Beyond the diffraction barrier: STED versus confocal fluorescence microscopy on 40nm fluorescent micro spheres (emission maximum at 515 nm) spread out on a coverslip. The comparison of the confocal (A) and the STED (B) image demonstrates the superior resolution of STED microscopy; neighboring beads are clearly separated. (e) STED intensity distribution employed for depletion, featuring a prominent zero in the center. Drawn to scale with the images. (D) Vertical sum of marked region in (B) illustrates both the clear separation capabilities and subdiffraction resolution of STED microscopy. Objective lens, NA = 1.4 (oil); wavelengths A, 469 nm (excitation), 500 to 550nm (fluorescence detection), 585 nm (STED).
lous screening. Ongoing investigations on fluorescent proteins will establish the operational conditions for STED and its potential with these important labels. Recently, the resolving power of STED has been synergistically combined with that of 4Pi microscopy to achieve nanoscale axial resolution (Dyba and Hell, 2002). Destructive interference of the counterpropagating spherical wavefronts of the STED pulse at
the focal point produces a narrow focal minimum for STED with an axial FWHM of -AJ(4n) "" 100 to l20nm. Overlap with the regular excitation spot of a single lens has so far rendered focal spots down to Ilz = 40 to 50 nm. Linear deconvolution of the data removes the effect of the weak «30%) sidelobes that accompany the narrow focal spot. Moreover, it further increases the axial resolution up to 30 to 40 nm. This is exemplified in Figure 31.5(C,D),
A
Confocal
c
Confocal
B
STED-4Pi
o
STED-4Pi
FIGURE 31.5. Axial resolution increase provided by STED-4Pi (B, D) over confocal microscopy (A,e). xz-images from the immunolabeled microtubular network of a HEK cell as recorded with a (A) confocal and (B) STED-4Pi microscope. Both images have been recorded at the same site in the cell. The microtubules were labeled using a primary anti-/3-tubulin antibody and a secondary antibody coupled to MR-121. The STED-4Pi images were linearly deconvolved to remove the effect of the 4Pi sidelobes. Note the straight horizontal line which stems from an MR-121 layer on the coverslip. At this layer, the resolution of the STED-4Pi microscope is determined as 53nm after linear deconvolution. The HEK cell was mounted in aqueous buffer and recorded with water-immersion lenses. In another experiment the membranes of a live bacterium (Bacillus megaterium) were stained with the RH 414. Next it was simultaneously imaged in the confocal (e) and in the STED-4Pi microscopy mode (D). Note that the axial resolution of this focusing microscope is of the order of 30 to 40nm.
Nanoscale Resolution with Focused Light: STED and Other RESOLFT Microscopy Concepts • Chapter 31
which shows xz-images of the membrane-labeled Bacillus megaterium (Dyba and Hell, 2002). The STED-4Pi setup realized so far improves the resolution along the z-axis only, thus rendering a disk-shaped focal spot whose effect is also noticeable in Figure 31.S(B,D). The spot could be sculpted down to a spherical shape by applying a second, STED, pulse whose spatial form is designed to squeeze the spot laterally. STED-4Pi microscopy has also been applied to the imaging of the microtubular cytoskeleton of human embryonic kidney (HEK) cells (Dyba et aI. , 2003). The HEK cells were decorated with an anti-f3-tubulin antibody and a secondary antibody coupled to the red-emitting dye MR-121. The latter displays high STED efficiency (>90%) at a STED wavelength of -780 to 79S nm. Contrary to the confocal xz-sections, in the linearly deconvolved STED-4Pi counterpart, most of the microtubules appear as distinct objects [Fig. 31.S(A,B»). The axial resolution attained can be inferred from the FWHM of a fluorescent mono layer that has been deposited on the coverslip; it is -SOnm, corresponding to 1/16 of the irradiation wavelength of 793 nm. It is interesting to note that in the STED-4Pi image, the brightness of the monomolecular layer is of the same order as that of the microtubules. By contrast, in the confocal image, signal from the layer is overwhelmed by the total signal from the larger focal volume of the confocal microscope. An important task for the near future is to define the optimal parameters for specific imaging applications. The results obtained with STED and STED-4Pi microscopy demonstrate that the basic physical Abbe limit has been broken and we are now moving towards attaining a 3D resolution of the order of a few tens of nanometers.
CHALLENGES AND OUTLOOK Given the limits on the rate of excitation and fluorescence emission, high temporal and high spatial resolution may be mutually exclusive in many cases, simply because of the poor statistics of the collected photon signal. High spatial resolution also requires small pixels, which is not favorable for fast imaging of large areas. Downsizing the region of interest will be inevitable. A remedy is to parallelize the scanning system, either by applying many foci or by utilizing sophisticated structured illumination schemes. Likewise, the limited number of emission cycles that characterizes many fluorophores, caps the signal that is available from a sample, and, thus, the signal per sample volume. For a number of staining protocols and applications, the available signal might not match up with the number of photons that must be detected in order to benefit from the increase in spatial resolution. Therefore, potential improvements in fluorescent labels as well as strategies for avoiding photobleaching will play a vital role in firmly establishing nanoscale resolution in light microscopy. Still, even under poor signal conditions, a RESOLFT method creating ultrasmall focal volumes, such as STED, may be extremely helpful for techniques that exploit fluorescence statistics . For example, fluorescence correlation spectroscopy (Magde et aI. , 1972) depends on small focal volumes to detect rare molecular species or rare molecular interactions in concentrated solutions (Eigen and Rigler, 1994; Levene et ai., 2003). STED may be
577
the key to interrogating nanosized volumes in intact cells. In fact, it is so far the only method reported to squeeze a fluorescence volume to the zeptoliter scale without mechanical confinement. Published results imply the possibility of sampling spherical focal volumes of only 30nm diameter (Kastrup et aI., 200S). The past decade has witnessed the emergence of a family of physical concepts for attaining diffraction-unlimited spatial resolution in focusing fluorescence microscopy. Relying on reversible saturable optical (fluorescence) transitions (RESOLFT), the spatial resolution of these concepts is eventually determined by the saturation level that can be realized. Saturation brings about an essential nonlinear relationship between the signal and the applied intensity that allows one to overcome diffraction fundamentally. The nonlinearity brought about by saturation is radically different from that of the well-known multi-photon events. In the latter cases, the nonlinearity stems from the contemporaneous action of more than one photon, which inevitably demands high focal intensities. In contrast, the nonlinearity brought about by saturated depletion stems from the population kinetics of the states involved (Hell, 2003). This opened the door to attaining marked nonlinearities even with linear optical transitions such as singlephoton excitation and stimulated emission. Semistable molecular states enable saturable optical transitions at even lower light intensities. Bistable fluorophore constructs and switchable fluorescent proteins should allow very high levels of saturation at the low light intensities essential for live-cell imaging. This insight may be critical to opening up the cellular nanoscale with visible light and regular lenses (Hell, 2003; Hell et at., 2003). In fact, it is interesting that the demand for high intensities is the reason why the typical nonlinear optical processes of multi-photon fluorescence excitation and multi-photon scattering [second and third harmonic generation (SHG, THG), coherent antiStokes Raman scattering (CARS), etc.] could not substantially improve the spatial resolution. It is also clear that as n-photon excitation of the fluorescent state (i .e., the SI) eventually requires the subdivision of the excitation energy into n photons, it leads to an n times larger wavelength. As an n times larger wavelength leads to n times larger focal spots, multi-photon excitation processes are counterproductive when it comes to sharpening the focal spot size in the focal plane. Exceptions have been given (Hanninen et at., 1996; Schonle and Hell, 1999; Schonle et at., 1999) but they currently appear less promising than the concept of RESOLFT. The principles of the concept of RESOLFT have been validated through STED microscopy, whose ability to break the diffraction resolution barrier has been experimentally demonstrated. It is worth mentioning that the strategy of utilizing reversible saturable optical transitions also has the potential to break the diffraction barrier in nanoscale writing and data storage (Hell, 2004). The coming years will show whether STED and its RESOLFT cousins are going to establish themselves as part of the microscopic toolbox to elucidate dynamics and structure of cellular networks. The chances are better than ever. A taste of what is to come can be seen in Figure 31.6, which shows small patches of cell membrane, immuno-stained against the SNARE protein SNAP2S , and viewed both by confocal and STED microscopy using the exact same optics. The improvement is even more obvious in Figure 31.7.
578
Chapter 31 • S.W. Hell et al.
FIGURE 31.6. Resolution comparison of confocal versus STED-microscopy; plasma membrane patches immuno-stained against the SNARE protein SNAP2S; secondary antibody labeled with Atto S32-NHS ; Emission S40-S70 nm. STED at 6lS nm. The confocal image was recorded by simply turning off the STED beam with no other changes.
PSF
A
Resolution scaling
B
c
OTF
235nm
0 .45 A. E 200 oS ~
0.1
=::t: 100 LI.. LI..
en Il..
-200
-100
0
100
200 x [nm]
0 0
200 400 600 800 1000 STED Intensity [MW/cm2]
0.01 0
10
20
30
40 50 .6.x-1 [~m-l 1
FIGURE 31.7. STED-microscopy hits the nanoscale. (A) Comparison of the PSF (x-axis) of a conventional and a STED microscope probed with a single dye molecule whose orientation is parallel to the polarization of the STED-beam; for details see (Westphal and Hell, 200S). The up-to IS-fold reduction in lateral width underscores the potential of STED-microscopy to attain true nanoscale resolution. (B) The resolution of a STED microscope scales with the square root of the intensity used in the STED beam, with no firm limit, as predicted by equation (4). The gain in resolution also entails an increase of the OTF bandwidth over the diffraction barrier. (C) Shows an example where the usable bandwidth (magnitude> 1%) of the OTF is broader by about an order of magnitude than in a conventional microscope. All measurements were performed with the red emitting dye JA26 and with a STED-wavelength of -780nm. Being a far-field optical microscope, the resolution of STED-microscopy increases inversely with the wavelength. Therefore, in (A) reducing the wavelength for STED to SOOnm would decrease the FWHM down to -IOnm.
Nanoscale Resolution with Focused light: STED and Other RESOlFT Microscopy Concepts • Chapter 31
REFERENCES Abbe, E" 1873, Beitrage zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung, Arch. f Mikroskop. Anat. 9:413-420. Alivisatos, A.P., 1996, Semiconductor clusters, nanocrystals, and quantum dots, Science 271 :933-937. Bruchez, M. Jr., Moronne, M., Gin, P., Weiss, S., and Alivisatos, AP., 1998, Semiconductor nanocrystals as fluorescent biological labels, Science 281:2013-2015. Chudakov, D.M., Belousov, v.v., Zaraisky, A.G., Novoselov, v.v., Staroverov, D.B., Zorov, D.B., Lukyanov, S., and Lukyanov, K.A., 2003, Kindling fluorescent proteins for precise in vivo photolabeling, Nat. Biotechnol. 21: 191-194. Cragg, G.E., and So, P.T.c., 2000, Lateral resolution enhancement with standing wave evanescent waves, Opt. Lett. 25:46-48. Dyba, M., and Hell, S.W., 2002, Focal spots of size 1123 open up far-field fluorescence microscopy at 33 nm axial resolution, Phys. Rev. Lett. 88: 163901. Dyba, M., and Hell, S.W., 2003, Photostability of a fluorescent marker under pulsed excited-state depletion through stimulated emission, Appl. Opt. 42:5123-5129. Dyba, M., Jakobs, S., and Hell, S.W., 2003, Immunofluorescence stimulated emission depletion microscopy, Nat. Biotechnol. 21: 1303-1304. Eigen, M., and Rigler, R., 1994, Sorting single molecules: applications to diagnostics and evolutionary biotechnology, Proc. Natl. Acad. Sci. USA 91 :5740-5747. Hanninen, P.E., Lehtela, L., and Hell, S.W., 1996, Two- and multiphoton excitation of conjugate dyes with continuous wave lasers, Opt. Commun. 130:29-33. Hcintzmann, R., and Cremer, c., 1998, Laterally modulated excitation microscopy: Improvement of resolution by using a diffraction grating, SPIE Proc. 3568:185-195. Heintzmann, R., Jovin, T.M., and Cremer, c., 2002, Saturated patterned excitation microscopy - A concept for optical resolution improvement, 1. Opt. Soc. Am. A 19:1599-1609. Hell, S.W., 1997, Increasing the resolution of far-field fluorescence light microscopy by point-spread-function engineering, In: Topics in Fluorescence Spectroscopy (J.R. Lakowicz, ed.), Vol. 5, Plenum Press, New York, pp.361-422. Hell, S.W., 2003, Toward fluorescence nanoscopy, Nat. Biotechnol. 21: 1347-1355. Hell, S.W., 2004, Strategy for far-field optical imaging and writing without diffraction limit, Phys. Lett. A 326:140-145. Hell, S.W., and Kroug, M., 1995, Ground-state depletion fluorescence microscopy, a concept for breaking the diffraction resolution limit, Appl. Ph),s. B 60:495-497.
579
Hell, S.W., and Wichmann, J., 1994, Breaking the diffraction resolution limit by stimulated emission: Stimulated emission depletion microscopy, Opt. Lett. 19:780-782. Hell, S.W., Jakobs, S., and Kastrup, L., 2003, Imaging and writing at the nanoscale with focused visible light through saturable optical transitions, Appl. Phys. A 77:859-860. Kastrup, L., 810m, H., Eggeling, c., and Hell, S.W., 2005, Fluorescence fluctuation spectroscopy in subdiffraction focal volumes, Phys. Rev. Lett. 94: 178104. Klar, T.A., Engel, E., and Hell, S.W., 2001, Breaking Abbe's diffraction resolution limit in fluorescence microscopy with stimulated emission depletion beams of various shapes, Phys. Rev. E 64: 1-9. Klar, T.A, Jakobs, S., Dyba, M., Egner, A, and Hell, S.W., 2000, Fluorescence microscopy with diffraction resolution limit broken by stimulated emission, Proc. Natl. Acad. Sci. USA 97:8206-8210. Konig, K., Tadir, Y., Patrizio, P., Berns, M.W., and Tromberg, B.1., 1996, Effects of ultraviolet exposure and near-infrared laser tweezers on human spermatozoa, Hum. Reprod. II :2162-2164. Levene, MJ., Korlach, J., Turner, S.W., Foquet, M., Craighead, H.G., and Webb, W.W., 2003, Zero-mode waveguides for single-molecule analysis at high concentrations, Science 299:682-686. Lukyanov, K.A., Fradkov, A.F., Gurskaya, N.G., Matz, M.V., Labas, Y.A., Savitsky, AP., Markelov, M.L., Zaraisky, A.G., Zhao, X., Fang, Y., Tan, W., and Lukyanov, S.A., 2000, Natural animal coloration can be determined by a non fluorescent green fluorescent protein homolog, 1. BioI. Chern. 275:25879-25882. Magde, D., Elson, E.L., and Webb, W.W., 1972, Thermodynamic fluctuations in a reacting system - measurements by fluorescence correlation spectroscopy, Phys. Rev. Lett. 29:705-708. Peng, X., Manna, L., Yang, W., Wickham, J., Scher, E., Kadavanich, A, and Alivisatos, A.P., 2000, Shape control of CdSe nanocrystals, Nature 404:59-61. Schonle, A., Hanninen, P.E., and Hell, S.W., 1999, Nonlinear fluorescence through intermolecular energy transfer and resolution increase in fluorescence microscopy, Ann. Phys. (Leipzig) 8:115-133. Schonle, A, and Hell, S.W., 1999, Far-field fluorescence microscopy with repetitive excitation, Eur. Phys. 1. D 6:283-290. Westphal, v., and Hell, S.W., 2005, Nanoscale resolution in the focal plane of an optical microscope, Phys. Rev. Lett. 94: 143903. Westphal, v., Blanca, C.M., Dyba, M., Kastrup, L., and Hell, S.w., 2003, Laser-diode-stimulated emission depletion microscopy, Appl. Phys. Lett. 82:3125-3127. Westphal, v., Kastrup, L., and Hell, S.W., 2003, Lateral resolution of 28nm (lambdaJ25) in far-field fluorescence microscopy, Appl. Phys. B 77:377-380.
32
Mass Storage, Display, and Hard Copy Guy Cox
INTRODUCTION Confocal microscopes commonly generate their images not as real or virtual patterns of light, but as pixel values in the memory of a computer (Cox, 1993). This gives the image a measure of permanence - unlike a visual image, once acquired it will not fade - but it will be lost if the computer is turned off, or if that area of memory is overwritten. To store that image with all its information intact we must write it in digital form - a copy on paper or film, however good, cannot contain all the information of the original. However, a copy on disk or tape is not directly accessible to human senses. For publication or presentation of the image, or even just to access it, we must have a display or a hard copy, a picture which can be viewed by the human eye. This chapter reviews the range of possible solutions to these two problems. Because this is a rapidly moving area, new alternatives will doubtless become available almost as soon as this is printed. A measure of the rate at which this happens is that many of the technologies reviewed in the previous edition are now obsolete, leaving users with the task of copying images to new media if they are to retain access to their data. As well as assessing currently available technologies, therefore, I will try to provide enough background information to enable users to assess the latest high technology advances in a rational way. It is always worth considering the scale of the adoption of a technique as well as its technical efficiency because most of us will still want to be able to use our data in 10 or even 20 years' time, and only mass-market solutions are likely to survive on that timescale.
MASS STORAGE The major problem in storing confocal images is their sheer size. The smallest image we are likely to acquire would be 512 x 512 pixels, at one plane only and containing only one detector channel. Assuming that we store only 1 byte per pixel (that is, each point in the image can have one of 256 possible gray levels) this will require one quarter of a megabyte (MB) to store. We will require a little more space to store basic information about how the picture was acquired, either in a header at the start of the file, or at the end, or even in a separate file. Most confocal microscopes will capture larger images than this, and most will capture more than one channel. A three-channel, 2048 x 2048 pixel image (routine on any current system) will require 12 MB to store one plane. A three-dimensional (3D) image data set could easily contain 100 or more planes, thus requiring 1200MB (1.2 gigabytes, GB) or more to store. At the time of this writing, current personal computers typically have 80 to 200GB hard disks, a 200-fold increase
on the norm when the last edition of this chapter was written, but still not enough to be regarded as a permanent store. To provide archival storage, we must have some form of removable storage media.
Data Compression Before considering the contending bulk storage devices, is there any way we can reduce the size of the problem? Can we compress the image data to make it smaller? Lossless data compression systems, which preserve the integrity of our original image data, generally work on some variation or combination of three well-known algorithms. Run-length encoding (RLE) looks for sequences of identical values and replaces them with one copy of the value and a multiplier. It works very well with binary (black/white) images or simple graphics using a few bold colors, and is used, for example, for all the splash screens in Microsoft Windows. Lempel-Ziv-Welch (LZW) and Huffman encoding look for repeated sequences, assign each a token, then replace each occurrence by its token. Neither of these works well with real images (though they do an excellent job with computer-generated graphics). Thus, if you save a confocal image as a GIF (graphics interchange format) file , or as a compressed TIFF (tagged image file format) file, both of which use LZW compression, you will be lucky to get even a 10% to 20% decrease in size, and sometimes your file size will actually get larger. You will not do much better with the popular archiving systems PKzip, gzip, or WinZip, which (to avoid patent problems) use LZ77, an earlier Lempel-Ziv algorithm, and Huffman encoding (Deutsch, 1996). though these systems do at least recognize if compression is not working and insert the un compressed data instead, so your file should not get larger. There are a couple of exceptions to this generalization. First, some confocal microscopes store 12 bits of data at each pixel (4096 gray levels), but they store this as 16-bit numerical values. Clearly these images have redundant space - a quarter of the file contains no information - and they will therefore at least compress to 75% of the original size. The file will nevertheless become even smaller, often with little or no real loss, if it is converted to 8-bit data. Second, even though it may not immediately be obvious, a threechannel image of moderate size, saved as a 24-bit RGB file, must always have redundant information. Twenty-four bits of data can specify 16.7 million colors, but a 512 x 512 image with only a quarter of a million pixels can contain at most a quarter of a million colors. Efficient algorithms will automatically find this redundancy and yield effective compression (how this is done is explained in the description of the PNG format, below).
Guy Cox. University of Sydney, New South Wales 2006, Australia
580
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC , New York , 2006.
Mass Storage, Display, and Hard Copy • Chapter 32
PNG, which stands for portable network graphic, but is pronounced "ping," is a lossless compression system (Roelofs, 2003). It will usually offer the highest lossless compression currently attainable for confocal images. The formal compression system is identical to that of the "zip" systems; the deflate algorithm (Deutsch, 1996) a combination of Huffman and Lempel-Ziv algorithms that in essence looks for repeated patterns. The secret of PNG's improved performance lies in its prefiltering of the image to establish the best way to represent the data. In a real-world image of any kind, the difference between adjacent pixels will rarely be extreme so often the data can be reduced substantially by storing only the difference. The different filters vary essentially in the pixels used for comparison (no filtering, pixel before, or before and after, or before and above, etc). Any implementation contains all filters and so will decode any image, but the better implementations will offer improved compression by careful choice of which filter to apply. (The standard allows different filters to be used on each line of the image if required.) So if lossless compression is important it may be worth experimenting with different vendors' implementations of PNG (see below). It tends to be much slower than LZW to compress, partly because it is a two-pass process but mainly because, to get the best results, the program should test which algorithm will give best results. Decompression is fast (see Table 32.1). The demands of computer multi-media have led to the development of compression techniques specifically aimed at real-world images, both still and moving. Unlike the compression techniques mentioned above, which are completely reversible, these approaches discard information from the image. The picture created after compression and decompression will not be the same as the original. However, very large file compressions can often be achieved with losses which are barely detectable to the eye, though they may affect numerical properties of the image.
TABLE 32.1. TIme to Compress and Read Back an Image Using Different Techniques Compression Uncompressed TIFF LZWTIFF PNG Wavelet (Iossless) Wavelet (high-quality)" Wavelet (high-quality)b Wavelet (low-quality)" Wavelet (low-quality)b Lossless JPEG OCT JPEG (high-quality) OCT JPEG (low-quality)
Save Time (s)
5 7 40 9 16 8 15 6 8 4 4
Read Time (s)
3 5 5
12 10 10 6 6 7 4 4
File Size (KB) 9220 6014 2861 1934 841 821 9 10 3877 840 164
Specifying required quality. b Specifying required file size. The image used was that seen in Figure 32.1, but scaled up (using bicubic interpolation) 6-fold to 3072 x 3072 pixels in order to make the times measurable. All conversions were done using Paint Shop Pro version 8 (Jasc Software); the results should only be taken as relative and will vary greatly with processor speed. Scaling the image means that it contains substantial redundancy and therefore the compression levels achieved are unrealistic; the file sizes are given mainly to illustrate the trade-off between processing time and disk access. PNG was by far the slowest in compressing the image, but was rapid to read back. The processing requirements of OCT JPEG compression were more than compensated for by the reduction in disk access, so that it was very fast, but loss less JPEG was slower and its compression did not match lossless wavelet or PNG. Wavelet compression (JPEG 2000) showed the curious result that selecting a "compression quality" gave much longer save times than selecting the "desired output file size." At equivalent final sizes, the resulting images seemed similar. This is probably a quirk of the implementation of what is, at the time of this writing, a very new standard. Wavelet images were the slowest to read back, particularly at high image qualities. a
581
The most common still image format is the Joint Photographic Experts' Group (JPEG) compression protocol (Redfern, 1989; Anson, 1993; Pennebaker and Mitchell, 1993), which is supported by many paint and image manipUlation programs. This breaks the image into blocks of 8 x 8 pixels, each of which is then processed through a discrete cosine transform (OCT). This is similar to a Fourier transform, but much faster to implement, and gives an 8 x 8 array in frequency space. The frequency components with the lowest information content are then eliminated, after which highfrequency information (fine detail) will be selectively discarded to give the desired degree of compression. The remaining components are stored (using Huffman encoding) in the compressed image. The amount to be discarded in frequency space can be specified, which gives the user control over the trade-off between image quality and degree of compression. Typically, monochrome images can be compressed down to one fifth or less of their original size with no visible loss of quality (Avinash, 1993). Compression and decompression are similar operations, and require similar amounts of computer time. Ten years ago, when the standard was first published (Pennebaker and Mitchell, 1993), the time required was quite noticeable but with a modem processor the reduced amount of disk access will more than compensate for the processing time (Table 32.1). Color images can be compressed further than monochrome because luminance (brightness) and chrominance (color) are treated separately. The eye can tolerate a greater loss of information in the chrominance signal, so this is normally handled at half the resolution. (The standard allows many different options here but specific implementations usually do not make these evident to the user.) This has certain consequences in confocal microscopy because a three-channel confocal image is not a real-color, realworld image but three images which are largely independent of each other. A three-channel confocal image compressed as a color image will look quite adequate but should not be used reliably for numerical analysis; for example, the lower resolution of the color information would make many pixels show colocalization when in fact there is none. The JPEG standard itself specifies a compression technique, not a file format. As such it is used in many different situations (including one of the compression options in the TIFF standard and in programs such as Microsoft PowerPoint). However, it is most familiar to the end user in the form of files conforming to the JFIF (JPEG file interchange format) standard, which typically use the suffix .jpg. JPEG compression is designed for photographic images so that it only manipulates gray-scale or true color (RGB) images. Adding a false-color palette to a gray-scale image will make it less suitable for JPEG compression because the JPEG algorithm would convert it to a full color image, tripling its size, before compression. Lossless JPEG compression also exists; there have been two distinct lossless compression modes specified in the JPEG standard over the years, but these do not use OCT to compress the image and typically do not perform very well, so they have not become popular. The current version, JPEG-LS, uses a predictive algorithm formerly called LoCo, and is designed to be both fast and easy to implement. Other specific image compression techniques show considerable potential but have yet to achieve the popularity of JPEG (OCT). Fractal compression, a proprietary technique developed by Iterative Systems Inc. (Anson, 1993; Barnsley and Hurd, 1993), creates mathematical expressions which, when iterated, recreate the original image. It can give spectacular levels of compression. Unlike JPEG compression, creating the compressed image is a very time-consuming process but decompression is very quick.
582
Chapter 32 • C. Cox
This has made it most useful for such items as CD-ROM encyclopedias but its initial promise has not led to widespread adoption. Wavelet compression is currently the hot topic in image compression and will undoubtedly be in common use throughout the lifetime of the current edition of this book, though at the time of this writing it is only just appearing in the latest releases of mainstream implementations. It is, in a sense, mathematically comparable to JPEG in that it separates the frequency components in an image, but it works in real space rather than reciprocal space. The basic idea of separating an image into components of different resolution and discarding the lowest information content and highest frequencies first is similar, but it is achieved by passing a series of filters over the image at a range of different scales. The filters wavelet filters - are the key to this, and are designed to be reversible. The claim is that wavelets can offer useful compression without loss, and much greater compression with losses that are not obvious to the eye. Other advantages include the ability to rapidly generate a low-resolution image (using the coarsest wavelets) and fill in the detail afterwards. Wavelet compression can treat an image as a whole or break it down into blocks which are compressed individually. The JPEG has introduced wavelet compression into a new version of the JPEG standard (JPEG 2000), and it is in this format that most mainstream applications will offer wavelet compression. In the interests of speed and portability (wavelet compression is intrinsically slower than DCT), the JPEG 2000 implementation uses only two wavelet filters, one for loss less compression and one for lossy compression. Even so, the time required is quite noticeable even on a fast computer (Table 32.1). While a wider range could offer better performance by finding the best wavelet for each image, the practical difficulties involved were deemed to make it not worthwhile. Also, in the JPEG implementation the image is broken into blocks before compression. A major criticism of the DCT JPEG standard was that the 8 x 8 blocks could often become visible at high levels of compression and JPEG 2000 therefore offers variable sized blocks within a single image, so that one compression level can be applied to featureless regions (such as sky, or the background in a confocal image) and another to regions containing fine detail. In practice, however, wavelet compression does not seem to offer superior performance over DCT for confocal images, as Figure 32.1 shows. Figure 32.1(A) shows a cultured He-La cell labeled with fluoroscein isothiocyanate (FITC) tagged to an antibody against ~-tubulin. It is an average projection from 16 confocal optical sections - a 512 x 512 pixel 8-bit image. Using an average projection rather than a maximum brightness projection improves the signal-to-noise ratio, but it also reduces the total intensity (because so much of the image is dark) and this therefore reduces the number of gray values present (there are only 120 values in this image). Both factors make the image a better candidate for compression. To preserve the visual quality the contrast has been scaled and the gamma changed (see below); these operations simply change the values assigned to each of the 120 tones, they do not change the number of tones and should not affect how it will compress. Figure 32.1(B) is one of the original slice images with no modifications to gray values. It shows more noise than the projection, but contains 248 gray levels, showing that the gain and black level controls had been used optimally to make use of the full dynamic range without overflow or underflow. The raw image size in each case is 256 KB, and tif and bmp files are 257KB. An LZW-compressed tif file of Figure 32.1(A) offered a reasonably useful reduction to 170 KB, while a PNG file
created with the well-known program Paint Shop Pro (lASC Software) did rather better at 143KB. The PNG optimizing program Pngcrush (freeware; see Roelofs, 2003) made an insignificant improvement to 142 KB. This is 55% of the original file size and shows that with a restricted gray range and dark noise-free background reasonable compression can be achieved without loss. Lossless wavelet compression (JPEG 2000) was less effective, giving a file size of 168 KB, scarcely better than LZW-compressed tiff but taking very much longer to compress and decompress. Lossless JPEG was comparable, at 169 KB. As predicted, the original single-slice image [Fig. 32.1(B)] did not compress nearly so well; the LZW version, at 256 KB, was hardly changed from the original size. PNG did better, at 195 KB (204 KB before optimization). But at 76% of the original size it hardly seems worth the effort. It does, though, reinforce the point that PNG is the only format worth considering for lossless compression of confocal images. DCT (JPEG) compression of the projection [Fig. 32.1 (A)] to two different levels is seen in Figure 32.1 (C,D). Figure 32.1 (C) shows the image compressed to 26.4KB, around 10% of its original size. While some loss of quality is evident, the image remains perfectly usable and the compression is very substantial. In Figure 32.1 (D), compression has been increased to the point where the image is visibly degraded but still recognizable and even informative, though the file size is only 7.7KB, a mere 3% of the original! Figure 32.1(E,F) shows the same levels of compression but using wavelet compression with JPEG 2000. Both are substantially worse than equivalent DCT images. A specialist wavelet compression program (not using JPEG 2000) was also tried, and gave worse results at equivalent compression levels. It seems probable that the relative failure of wavelets to compete with DCT lies in the rather limited range of resolution levels which contain substantial information in these confocal images. The interest lies primarily in the microtubules, all of which are the same size. In reciprocal space, regions with no information will automatically compress to nothing, whereas the wavelet function may perhaps be chosen to treat all frequencies more or less equally because this may be the best strategy for conventional photographic images. There may therefore be scope for a wavelet implementation dedicated particularly to confocal images. Figure 32.2 shows the histograms of the images in Figure 32.1. In Figure 32.2(A) the missing gray values are obvious, whereas the single optical section [Fig. 32.2(B)] shows a continuous spectrum. At 10% compression the DCT image [Fig. 32.2(C)] shows a similar spectrum, but smoothed and with the gaps in the gray levels now filled. The wavelet version [Fig. 32.2(E)J also preserves the same shape, but is rather more smoothed at the same compression. At 3% of the original size the DCT histogram [Fig. 32.2(D)] is very much changed, while the wavelet one [Fig. 32.2(F)] shows little change from the 10% compression. In each case, the mean value remains unchanged. These figures show that photometric parameters are surprisingly well conserved even at levels of compression that would seldom be used in practice. While wavelet compression affects the histogram more than OCT at 10% compression, it is more accurate than DCT at 3% compression so that even though the image looks worse, its photometric parameters remain closer to the original. In practice these compression levels would only be used for such purposes as Internet transmission of images. Compression to between 25% and 50% of the original size would give images of more general usefulness, with little visible change from the original. Even essential photometric parameters are preserved. In spite of the current interest in wavelet compression, DCT still seems a
Mass Storage, Display, and Hard Copy • Chapter 32
583
FIGURE 32.1. Effects of image compression on a confocal fluorescence image of a cultured He·La cell immunostained with FITe against /3-tubulin. (A) Average projection of the original dataset of 16 optical sections, with contrast scaled and gamma subsequently corrected; original uncompressed image. (B) One optical section from the stack, with no subsequent processing. (e) JPEG compressed (DeT) to -10% of the original size. (D) JPEG compressed (DeT) to -3% of the original size. (E) Wavelet compressed (JPEG 2000) to -10% of the original size. (F) Wavelet compressed (JPEG 2000) to -3% of the original size. Insets in (A, e-F) are part of the image at 2x magnification to show the losses in compression more clearly.
584
Chapter 32 • G. Cox
B
A 101"" 61 «7 fn:t)
loin: 0
101"", 21221
Avg: 66
c
f~
loin: 0
Avg: ~
D 101"", 259117 f9.t)
loin: 0
Avg: 66
101"", 102814~)
loin: 1
Avg: 66
F
E tot"" 299012 f1 1:l:) loin: 0 Avg: 66
101 "'" 25940 f!IX)
loin: 0
Avg: 66
FIGURE 32.2. (A-F) Histograms of pixel intensities in Figure 32.I(A-F), respectively.
better choice for confocal images in cell biology. Not only is it more effective, it is much faster then wavelet compression (Table 32.1). Lossless compression only gives useful results on images with large amounts of uniform background and low noise but in these cases it can be effective. The most likely use would be for storing the output of 3D reconstructions, as in Figure 32.1(A).
Although generating a complex 3D movie sequence can take as long as acquiring the original confocal data, and the output files can be just as large, we typically do not have the same concerns about preserving data integrity. It is therefore sensible to use JPEG compression for storing the output. Some confocal datasets contain only very sparse information. Figure 32.3(A) provides an example, a frame (pre-calcium wave) from a time series of calcium transients induced by testosterone. There were 193 images in the series and without compression this dataset occupies close to 50MB. However, as only 12% of the pixels lie above the background noise level, the dataset even in its original form compresses without loss to below 100 kB per frame - 40% of the original- with LZW or PNG. If we remove background by setting pixels with a gray value of 14 or below to zero [Fig. 32.3(B)], we have a virtually unchanged image which is now highly compressible without further loss. PNG compression gave a file size of only 49.3 KB, less than 20% of the original. Our original 50MB dataset will now only be 10MB. Lossy JPEG compression makes no sense with such a dataset - using a typical setting for reasonable image quality the resultant file size was actually larger (59.3 KB) than the lossless one. What is more, the compression process brought background back into the dark areas. So the message is either use a lossy compression on the original data or compress it by background subtraction and then save it without further loss - do not do both. Other image manipulations will also affect the compressibility of images. Smoothing, to remove noise, will reduce the high-frequency content and therefore make images more compressible. Deconvolution, on the other hand, aims to restore high-frequency content. This will make images less compressible, or will mean that more is lost in lossy compression. Figure 32.4 illustrates this
FIGURE 32.3. Calcium imaging (non-ratiometric) of transients induced by testosterone in cultured cells. Pre-stimulation, time point 42 from 193 images taken at I-second intervals. (A) original image (B) background removed by setting all pixels below a value of 15 to zero, a process that permits no-loss compression to reduce file size of (B) by a factor of two compared to (A). A false-color palette has been added to show how the background has been set to black but none of the "data" pixels have changed. Taken using a 63xINA 1.2 water-immersion lens. Image width as shown (only part of the original image) is I54).1m. Image courtesy Dr. Alison Death.
Mass Storage, Display, and Hard Copy • Chapter 32
585
FIGURE 32.4. Three different angle projections (0°, 45°, and 90°) from a 3D dataset of the dinoflagellate alga Dinophysis. (A) Maximum intensity projection from the original dataset. (B) Maximum intensity projection after smoothing (3D median filter) and one-dimensional deconvolution.
point. This is a 3D dataset of the dinoflagellate alga Dinophysis, which was collected at 3 pixels per resel and therefore is slightly oversampled. This provides the opportunity to smooth the data down to the Nyquist limit and thereby reduce noise without adversely affecting resolution. Figure 32.4(A) shows maximum intensity projections, at three angles, from the original dataset. This type of projection leaves noise unchanged so the view shows an accurate impression of the noise content of the original set. Compressed with LZ77 the original 4.5 MB dataset reduces to 1.9 MB, a useful saving, reflecting the large proportion of background in the set. When the entire dataset is smoothed with a median filter, acting in three dimensions (Cox and Sheppard, 1999), it becomes more compressible, now reducing to 1.38MB. If we deconvolve this dataset we can restore some of the resolution lost in the z(depth) direction by the transfer function of the microscope (Cox and Sheppard, 1993, 1999). As expected, it is now rather less compressible, at 1.47 MB , but this is sti ll a useful saving on the original. The smoothed, deconvolved dataset is shown in the same projections in Figure 32.4(B). In any image compression strategy, it is important to bear in mind that confocal images can become virtually meaningless if the information about the acquisition is lost. Some confocal systems (e.g., Bio-Rad) store this data in a header within the same, single file as a series of optical sections. Even if the slice images are exported by the Bio-Rad software, the acquisition data is not exported and the images cannot be re-imported for subsequent processing. Other systems (e.g., Leica, Zeiss) store a database of information about the images - exported images generated within the acquisition software will still retain some of this information but typically 3D reconstructions can only be done from the original images. In either case it is important to ensure that the alI-important image acquisition data are preserved, and if possible that the images can be restored to their original file name and type. A final point: the most common waste of disk space consists of storing completely featureless areas! If your sample is rectangular, select a rectangular window to image it rather than collecting a strip of nothing on each optical section. And do not collect
three channels if you have only two labels! Modem systems make it all too easy to accept the default method, or configuration; it will save a lot of time in the long run if you spend a minute or two changing settings to collect only what you want.
Removable Storage Media Storage media can be divided into those which are sequential (records are written and read from one end only) and those which are random access (it is possible to move directly to any record, whenever it was written).
Sequential Devices Sequential devices are tapes of various formats and sizes storing up to 200GB on a single cassette. Tape is still the largestcapacity bulk storage medium available, but is no longer competitive in cost with optical storage. As an image storage system, it also suffers from the time taken to locate and recover anyone file . A single file cannot be erased and replaced by another; one must erase either the whole tape or a large group of files, depending upon the recording system. Also, although it is rewritable it will not stand an infinite number of uses. The tape surface has a much harder life than the surface of a disk - it comes into direct contact with the recording heads and capstans, and is coiled and uncoiled each time. Even reading files repeatedly wears the tape, and its long-term archival potential is dubious. Once tape drives were regularly used for data storage and transfer but now their use is almost exclusively for backup purposes - making a copy of a complete file system or subsystem which will typically be read only once, in the event of a hard disk failure. Modem tape systems are very specifically designed for this task; their purchase cost is high but cost per megabyte stored can be low compaced to other rewritable media. This gives them some attraction for long-term archival storage of images that will not need to be accessed regularly, and for very large collections of images. Dumping a 40 or 100GB hard disk full of images on to a single tape will be much quicker and simpler than writing to
586
Chapter 32 • G. Cox
dozens of compact disks (CDs) or digital video disks (DVDs). However, most tape systems now rely on specific software to handle them and both this software and suitable hardware will need to be available for the tape to be read in the future - past experience suggests that this will limit effective use to 5 years or so, and this is probably the realistic limit for tape life also. Transfer rates up to 24MB S-I are available on expensive highend systems, although systems designed for small computer use will offer no better than 3MBs- l • At 24MBs- 1 writing one CD worth of data will take only 30 s, but it will take a quarter of an hour to copy a 20 GB hard disk. At 3 MB S-I that same disk will take almost 2 h to copy. Manufacturers typically quote compressed capacities for their tape drives, based on a notional 2-fold compression ratio that they expect to achieve with their archiving software. This is unrealistic when dealing with image files, and when evaluating competing systems, it is important to compare actual, uncompressed, storage capacities; this is much closer to the figure achievable with microscope images.
Random-Access Devices Random-access devices comprise a range of disk media, either magnetic or optical, and solid-state devices.
Magnetic Disks The oldest and simplest of removable media, rewritable, randomaccess systems is the humble floppy diskette. These are now virtually obsolete, limited by their small capacity - 1.4MB in the only (marginally) surviving 3.5" version. As many will have found out, finding a drive to read the once ubiquitous 5.25" disks is already difficult. In any case, they are too small to be relevant for confocal images. Various types of super-floppy have had a vogue in the past, but the only current survivor seems to be the Iomega Zip disk, which originally held 100MB but now comes in capacities up to 750MB. These are robust and durable but seem unlikely to be current for very much longer, driven out by far cheaper optical technology. They are also too limited in space to meet most modem needs for confocal image storage. Cost per gigabyte is around US$20-100. Other removable platter magnetic devices have been current, and suffer from the same limitation that in the course of time there may no longer be hardware available to read them. One of the most successful at the time of writing is the Orb drive, available in capacities from 2 to 5 GB. Like many other portable devices they connect to the host computer by the USB (universal serial bus) port, or the parallel printer port. Parallel port connection is relatively slow and USB is by far the preferable option. Cost per gigabyte is of the order of US$IO to US$20, so it is a reasonably affordable option. There remains the option of just using conventional hard disks. Mounting kits are available to fit a conventional disk in a pull-out mount; disks are also available in cases for connecting to USB, Fire Wire, or SCSI (small computer systems interface) ports, and there are micro sized ones which fit the PC card (PCMCIA) slot in notebook computers. The recent fall in price and increase in capacity of hard disks has made this a surprisingly affordable option (below US$I.OO per gigabyte for IDE disks, more for SCSI). Data transfer is as fast as the disk - certainly faster than most other options - and rewriting capacity is effectively unlimited. The long-term potential is less certain because the durability of the system depends not only on the longevity of the magnetic medium, but also on the lifespan of the motor and heads.
Optical Disks In the previous edition, devices such as WORM (write once, read many) and MO (magneto-optical) disks were discussed. These, like so many technologies, are not only dead but virtually forgotten except by those laboratories which have a huge stock of the disks! However, optical technology is certainly the current preferred option because there is good reason to have faith in the archival durability of the media. Furthermore, mass-market devices now have sufficient capacity to meet many users' demands so that one can have some confidence in the longevity of the technology.
Compact Disks Compact disks (CDs) have already been with us for over 20 years, and writable CDs for 10. The cost, high when the previous edition was written, is now very low both in first cost and media (around US$0.70 per gigabyte). Speed, though it has increased about 12-fold since then, is still the major problem. The rate of data transfer for an audio CD is a rather pedestrian 150 KB S-I, and this is referred to as single speed. Read and write speeds up to 52x this base value are now available. A complete 700 MB CD can thus be written in 5 min or so, and modem software will adjust the writing speed on the fly so that the need to maintain a constant data stream is less of an issue. This means that CDs can now even be written across a network, though this will inevitably carry a speed penalty. It may still be preferable to carry the additional overhead of first copying files to the writing computer. Rewritable CDs are also widely available at a cost only a little higher than conventional single-use CDs. Erasing data for re-use is, however, a relatively slow process. They may be useful when images are to be stored for a short time only, but for long-term archival use it would seem wiser to use single-use disks. Many manufacturers have conducted accelerated-aging tests on their single-use CDs and their security as archival storage seems to be the best of all mass-market computer media. It seems inevitable that rewritable disks could not offer equal security, and the risk of accidental deletion is always present with any rewritable medium. In fact, the time spent trying to decide which files can be overwritten is usually worth much more than the disk space saved. Various formatting options now allow multiple use, either by writing multiple sessions (which does carry an overhead of about 15MB per session) or by using the packet CD format, which allows a CD to be treated almost exactly like a conventional mounted drive. Multi-session CDs can be read on most systems but the packet CD format cannot. Because it reduces compatibility with other systems and has little point when the content of a CD is relatively small compared to a modem hard disk, packet CD has not become widely popular. One of the limitations of the CD format is its handling of file names. The standard laid down by the International Standards Organization (ISO) requires file names to fit an 8 + 3 character format similar (but not identical) to that of MS-DOS. ISOcompatible CDs are readable on Apple, PC, and Unix computers, which is very convenient for data exchange. Unfortunately, most confocal microscopes give files and directories (folders) much longer names. Extensions to the standard allow for longer files names in both Macintosh and Windows computers, but these are unfortunately not cross-platform compatible. Because most confocal microscopes use Windows it is important to use the Joliet extension which caters for these file names, otherwise the disk will contain a useless collection of truncated names, particularly with microscopes such as current Leica models, which save each plane and channel as a separate file, and rely on a database program to
Mass Storage, Display, and Hard Copy • Chapter 32
identify these images. On an ISO disk, it will be impossible to identify which plane and channel are which, and the data becomes completely useless. Because CDs are likely to be a major archival medium, in the medium term at least, the question obviously arises as to how permanent they are. Pressed CDs have a polycarbonate blank into which the pits are pressed to carry the information. This surface is then coated with an evaporated metal layer, and then a coat of varnish and the printed label (Fig. 32.5, upper). The CD is read through the thickness of the blank. If the clear side of the blank gets scratched, it will hinder reading but it can often be repolished . The label side is more vulnerable because only a layer of varnish and the printed label lie between the data and the outside world. Recordable CDs have a dye layer between the polycarbonate blank and the metal film and it is this which is modified by the writing laser beam (Fig. 32.5, lower). In terms of pressed CDs, excluding physical damage, the key issues are the aluminum reflective coating (which can get oxidized, particularly if there are any flaws in the varnish) and the polycarbonate blank. So far, no plastic seems to last forever and I doubt if polycarbonate will stay clear and flexible indefinitely. However, as polycarbonate is vulnerable to almost all organic solvents, excluding light and solvent fumes will doubtless help. Archival quality recordable CDs usually use something better than aluminum. Several manufacturers offer silver, silver + gold, or pure gold. Obviously pure gold should be highly stable, but it is less reflective, which increases the risk of read errors. Whatever metal is used, archival life still depends on the dye layer in front of it remaining stable. The claims made by the manufacturer for their different dyes (typically cyanine, phthalocyanine, or azo) are difficult to evaluate. The cyanine dyes used in the earliest recordable CDs were rather vulnerable to bright ambient light. Some manufacturers have chosen to concentrate on extending the durability of these dyes, whereas others have turned to alternative dyes such as azo or phthalocyanine. All archival tests depend on accel-
Protective varnish
~
Label
Aluminum layer
Os.
_i-I~I:I:!I I
Reading side
Protective varnish
Dye layer
Metal layer
Reading side
FIGURE 32.5. Structure of a pressed (above) and recordable (below) CDs (not to scale).
587
erated aging (typically at higher temperature) and, while this is valid up to a point, it is unwise to trust it too implicitly (Nugent, 1989; Stinson et al., 1995) . The speed at which drives will write CDs has increased enormously over the years, with 52x now routine. This has placed pressure on manufacturers to increase the response time of the dye layer but it would seem logical that a dye which can be bleached at 50 times the original speed is unlikely to be as archivally stable as the older disks. Often the layer is made much thinner to enable the high-speed writing. Of course, dye technology is also evolving but, if an archival-quality disk will not support the latest writing speeds, there may be a good reason. Because one is likely to write or put a label on the back, the varnish is important. Most makers object to labels even though labeling kits are widely sold. The varnish is typically water based because the polycarbonate of the disks is very vulnerable to solvents. This leaves one in a cleft stick as to how to label it because water-based inks may loosen the varnish but solvent ones may attack the disk! Different manufacturers vary in their recommendations and the safest approach is to follow the recommendation for each particular brand. Quality disks will have an extra writable protective layer over the base varnish giving you a bit of extra security, and this is well worth having. Kodak recommends that CDs not be stacked adjacent to each other or to any other surface. They should therefore be stored in "jewel cases" or in a custom storage box which separates the disks, and not kept in envelopes or stacked on a spindle. Blank CDs are now so cheap that the cost of storage is below US$1 per gigabyte, depending on the brand and quality of the media. Common sense suggests that, however reasonable it may be to choose cheap disks when just sending data through the post or taking it from laboratory to laboratory, saving a few cents by choosing unknown brands is a false economy if the intention is archival storage. Writers are very cheap and quite fast (48 speed corresponds to 7 MB/s , comparable with modern tape systems). Best of all, every computer can read the disk without extra hardware. The huge range of commercial CD-ROMs ensures that readers will remain available for many years, so that archival material will be accessible as well as secure.
Digital Video Disk (DVD) DVD (digital video disks) represent the next stage of optical disk technology. Using similar technology, but shorter wavelength lasers so that resolution is better, 4.7 GB can be stored on one side of a disk with the same size as a CD. Because the optics that read the disk are confocal, a DVD can carry two separate layers of information, thus storing over 9 GB , but recordable two-layer disks are only just coming on to the market at the time of this writing. DVD-R write-once DVD disks - are now reasonable in price, at around US$I.OO per disk from the cheapest sources, and the writers are now reasonable at US$200 to US$300 (a mere I % of their price 4 years ago!). Two standards (DVD+R and DVD-R) exist for these disks. This has hindered their general acceptance, but newer players will handle both (Nathans, 2003). At 4.7 GB, it is clear that DVD is now both cheaper and more convenient than CD-R for storage of confocal images, though the question of the long-term stability and longevity of the format is still not so well known as that of the CD format. Nevertheless, DVD players are now common domestic appliances, so it seems likely that the standard will be durable. Rewritable DVD disks also suffer from incompatible standards and because they are of lower reflectivity are sometimes difficult to read in DVD players. The older standards (DVD-ROM and
588
Chapter 32 • C. Cox
DVD-RAM) were also lower capacity than 4.7 GB. As with CDRW, they are probably not the best alternative for archival storage, but could have their place for data transfer. Current drives mostly handle RW and R disks in both + and - options. As with CDs, rewritable disks are always slower to write. It is only in the past couple of years that the DVD market has really showed signs of maturity. Because most computer drives will read and write the CD format as well, it would seem to be the logical choice when purchasing a new system, and DVD writers are now routine on new confocal microscopes.
Solid State Devices A development which was not foreseen in the last edition of this chapter has been the proliferation of ultra-compact solid state memory devices which retain data even without a source of power. While small in capacity compared to a hard disk, these range up to more than the capacity of a CD in a tiny fraction of the space. Much of this development has been driven by the explosive growth of the digital camera market. Compact flash cards (Compact Flash Association, 2003) are used by many such cameras, making the computer accessories to read them an essential. Typically these use either the PC-card (PCMCIA) slots in notebook computers or else the USB or Fire Wire ports found on both desktop and notebook systems. A key feature is that both are designed to appear as hard disks to the computer without the need to install any drivers. As the card is the size of a postage stamp, and about 3 mm thick, it represents a highly portable data store, and many people use them for convenient portable storage or transfer between computers without reference to digital cameras. Flash drives currently can hold up to 2GB. Data transfer rates of the compact flash (confocal) chips are currently 5 to 7 MB/s, but the latest revision of the interface is designed to cope with rates of up to 16MB/s to allow for advances in chip technology in the future. Practical test speeds achieved in computers (Digital Photography Review, 2(03) are around 3 to 4MB/s with writing being slower than reading; performance in digital cameras will always be much slower. The cost is still around US$100 a gigabyte so it will not compete with CDs for archival storage, but as a fast, rewritable method of transporting relatively large files - whether images, documents, or digital presentations - compact flash has an important place. Memory Stick (Sony) and Smart Media (Samsung) are similar, more proprietary flash memory devices which fulfill similar functions, but so far offer a smaller range of useful options than the more open standard Compact Flash. They tend to be more popular in the portable music player market, showing again how several once-different technologies are converging. The Sony Micro Vault is a dedicated USB-only version that comes in capacities from 32MB to 256MB, and requires no further accessories; it even has a cover for the plug when removed from the computer. Similar "keychain" memory devices are available from other manufacturers. These flash memory devices have established a quite different market niche from other removable storage devices, but as photography becomes increasingly a digital process, this convergence seems likely to continue.
has an infinite gradation of tones within it, whereas the confocal image typically has just 256, if it is monochrome or false color. Merged two- or three-channel images may have up to 256 3 colors, but often have considerably fewer. Confocal images have a finite number of pixels, whereas photographic images have limited resolution, but a smooth transition from point to point. In more general terms, a confocal image is quantized in both spatial (x, y, z) and intensity dimensions (see Chapter 4, this volume). What the microscopi st actually sees is not the image itself, but a display on a monitor. Both the monitor and the way it is driven will have a major effect on the appearance of the image. This in tum is interpreted by the human eye when we see it directly, or by a camera if we record the image photographically. As a preliminary, we should therefore look at how monitors display confocal images.
Monitors Monochrome cathode ray tube (CRT) monitors simply have a layer of phosphor coated on the inside of the glass, so that an illuminated spot will be produced wherever the electron beam hits . The resolution of the monitor, therefore, depends solely on how small the electron beam hitting the screen can be. Color monitors, on the other hand, have red, green, and blue phosphors arranged either in dots (shadow-mask tube) or stripes (Trinitron tube). The image on a color monitor will always be made up of a mosaic of the three primary colors; the finer this mosaic, the better the image will be. This is specified by the dot pitch of the tube in millimeters - 0.28 mm would be a typical value for a good quality modern PC monitor, though pitches as small as 0.18mm are available, and cheaper or older monitors will have pitches up to OAmm. These are absolute values, so a larger monitor will have more dots in the total width of the image. The number of pixels that may be displayed on the monitor is a function of the speed at which the electron beam can respond to a changing signal, and is not related to the actual dot pitch of the CRT. Thus, a confocal image displayed on a color monitor will have each pixel subdivided into a pattern of red, green, and blue dots. If the pixel spacing of the data comes close to the dot pitch on the screen, aliasing (below) may occur, creating undesirable effects. Many color monitors can be set up to display more pixels than there are dot triplets available, so the full resolution of the image cannot actually be shown . Thus, a large monitor is essential on a confocal microscope if we are to be able to see the detail in a high-resolution image. A CRT-based monitor is intrinsically capable of displaying an almost infinite number of colors. I However, the video board inside the computer imposes its own restrictions. Display boards suitable for a confocal microscope will permit 256 gradations in each primary color, so that a 24-bit (three 8-bit channels) confocal image can be displayed without compromising the intensity range. (12-bit or l6-bit images will, however, need to be reduced to 8-bit for display.) The other factor determining the appearance of the image is the frequency with which the display is scanned. The more rapidly the screen is refreshed, the less the image will flicker, and a suitable monitor should redraw the entire image at least 70 times per second. Low cost boards will often compromise one or other of these attributes at their highest resolution and are therefore inher-
DISPLAY Before looking in detail at how the image is displayed and printed, we should consider the nature of the confocal image (see Chapter 4, this volume). The image in a conventional optical microscope
1
Limited only by Poisson noise as it affects the charge deposited by the electron beam during the pixel dwell time and the number of phosphor grains in each dot, etc.).
Mass Storage, Display, and Hard Copy • Chapter 32
ently unsuitable for confocal use, but because high-end display boards are now very cheap compared to confocal microscopes, this is not likely to be a problem so long as it is understood that just any computer will not do. Displaying large numbers of pixels on low-priced monitors can reduce the refresh rate to 60 per second, which is about the lowest tolerable value. Interlaced scanning is a strategy used to reduce flicker when it is not possible to scan the entire image at a sufficient rate. First, the odd lines of the image are drawn, then the next scan draws the even lines. This technique is primarily used for broadcast video signals to enable the signal to fit into the available bandwidth, but it has in the past also been used to obtain higher resolution on low-cost computer monitors. International television standards use 625 scan lines per frame, with each interlace drawn 50 times a second (and thus a full-frame rate of 25/s). The system used in the Americas and Japan uses only 525 lines, but a faster refresh rate of 60 interlaces per second. (Not all scan lines are available for display; standard video can display only 512 pixels vertically, US video only 483.) Video displays once played a significant part in confocal microscopy (they were standard, e.g., on the widely popular Bio-Rad MRC 500 and 600) but they are not used now. Apart from the low resolution and refresh rate, the problem of displaying multi-channel images adequately led to their demise. In broadcast television, the color signal is encoded as a chrominance signal at much lower resolution than the monochrome luminance signal, and this does not give adequate quality for a multi-channel confocal image. The alternative is to generate a three-channel video signal that will give a much higher quality display (on the same monitor so long as the appropriate inputs are provided). However, this signal cannot now be recorded on a standard video cassette recorder (VCR) or printed on a lowcost video printer. The utility of video in microscopy is primarily in recording fast-moving items and it may still have a part to play with direct-vision confocal systems (Nipkow disk or slit scanners), but it is no longer relevant to point-scan systems. In a confocal image, pixel intensity values are linearly related to the numbers of photons captured from the specimen. However, this linearity may not be preserved when the image is displayed. In the simplest case, if the value of the pixel is converted directly to the voltage at the control grid of the CRT, the actual brightness of the pixel on the screen will be proportional to the three-halves power of the pixel value. This may not be all bad because the human eye responds logarithmically, not linearly, to light (Mortimer and Sowerby, 1941). However, confocal images often look excessively contrasty on an uncorrected display, and fine detail in the mid-tones will be lost. This relationship may be modified by more sophisticated display electronics, and high-quality display cards typically come with software to allow the user to set up the display optimally. These are all too often ignored by researchers who "don't have time" (and then waste much more time struggling with pictures which fail to show details which "were there when they took the picture"). Failing such an option, image manipulation programs such as Photoshop, Paint Shop Pro, and Corel Photo Paint place the display gamma under user control; but there again the user must make the effort to use that control (for more discussion, see Chapter 4, this volume).
589
tion does not arise. Typically different resolutions are not available on LCD monitors; in the rare cases that they are, the pixels are remapped in software (see below and Fig. 32.6). Therefore, LCD monitors will often give a crisper display than CRTs, though on the other hand the pixels may be more visible simply because their edges are more clearly defined. Some older displays (typically used on lower cost notebook computers) used passive supertwisted nematic (STN) displays instead of active thin-film transistor (TFT) technology. These are both cheaper and far less demanding of power. However, they may not offer full 24-bit color, and the image may be less bright and have a smaller viewing angle. Large freestanding monitors are always TFT. Large LCD monitors have many advantages in the confocal laboratory. Because the display is not continually redrawn as in a
.. ....
"
...
..... .
~
.~c..........
.'"
'I . :.. .......,,#, .. .~ _-..:1_ .c.· ... _ ....._ .... 7·7.:.:~ ;! ~
'.: .JIt ,"" .. -
. !~.. :.... -:. *. :-.:: . .- : .. -.. ........."."
- ..... ,:
...-
~,
Liquid Crystal Displays Flat screen (liquid crystal) monitors are inherently different from CRTs. They use liquid crystal devices between crossed polarizers to display the image, and as each pixel is addressed independently, the question of display resolution not matching the pixel resolu-
FIGURE 32.6. Halftoning (A) versus dithering (B). Highly enlarged view of part of an image of an integrated circuit chip; above printed by halftoning using a 4 x 4 matrix of laser dots, below printed by dithering using a 3 x 3 matrix.
590
Chapter 32 • G. Cox
CRT, fiicker is not an issue. Consequently, neither is refresh rate (it can become an issue with video images). The screen is fiat and compact, and as confocal systems often include two monitors and three lasers in a small room, minimizing heat is worthwhile. However, LCDs are costly, and the cheaper models often sacrifice color quality, viewing angle, or both. These are sacrifices which are not worth making. Good saturation, wide control over contrast and gamma, and a wide viewing angle are all essential. If you can afford it, buy a pair of top-quality LCD monitors (and don't just take price as a criterion of excellence, check them out carefully yourself). If cost is a major issue, buy high-quality CRT monitors rather than low-quality LCDs.
Data Projectors Data projectors are now a very common display format for confocal images, and naturally they often represent the occasion when high-quality display is most important. However, they typically have lower resolution and often a poorer gray-scale rendition than computer monitors. In terms of resolution, 800 x 600 pixels (SVGA) is common on projectors intended for the home market and 1024 x 768 (XGA) on ones intended for academic and teaching use, though higher resolutions are available at correspondingly higher prices. Two different technologies are used in these projectors. A very good description is given by Powell (2004). LCD projectors use liquid crystal screens, as in fiat panel monitors, except these do not have a color mosaic. Instead, three panels are used, one for each of the primary colors. Digital light processor (DLP) projectors use a micromirror array, where the pixels are tiny mirrors and these are tilted to send more or less light to the image. Very expensive projectors use three DLP chips, one for each primary color, but these are rare. The projectors one is likely to encounter in a lecture room or conference have one DLP chip, with a rapidly spinning filter wheel in front of it. The colors are therefore generated sequentially and merged by persistence of vision. The two types have their own strong and weak points, and typically DLP projectors are favored for home theatre use and LCD for data projection. As a comparison, for this review the signal from a notebook computer was sent simultaneously to two moderately high-end projectors, projecting on to adjacent screens. One was a DLP projector, the other an LCD, and price and luminous output were comparable. Contrast, brightness, and other display parameters were set to midpoint values on both projectors. The native resolution of both projectors was 1024 x 768, and the computer was set to the same value. Both projectors were able to handle higher resolutions and scale them down, but the quality suffered very markedly when this was done. The first lesson, therefore, is to set your screen display to the resolution of the projector, if possible. Even at the native projector resolution there may still be some pixel re-mapping taking place because projectors correct for keystone distortion caused by a non-horizontal projection angle. This means that either the top or bottom of the image cannot use all the displayable pixels. The LCD projector gave a much sharper image, which was obviously preferable for fine text. However, its color rendition, particularly on real-world photographic images, was inferior, having a slight color cast and excessive saturation. The DLP projector gave images with a very accurate color rendition, free of any cast and natural in appearance. The two projectors differed markedly in gamma. The DLP projector had a gamma of I (measured with the gamma test function of an imaging package), while the LCD projector was around 1.6 (slightly higher in blue and red
than in the green). This means that the LCD projector was very comparable to both the screen of the laptop and to a CRT monitor, both of which checked out with similar values, but the DLP is more accurate for confocal images in which pixel value is typically linear with number of photons. To test the displayable gray scales, a test image with intensity scaling from 0 (black) on one side to 255 (white) on the other was used. All 256 values were present, and on both CRT and notebook monitors the change seemed totally smooth. On the LCD projector it also seemed smooth, though with some minor streaks, which may have been aliasing rather than posterizing (see Digital Printers, below). However, there were noticeable bands with the DLP projector. This implies it was incapable of reproducing a full 8 bits in each color, and in fact posterizing was noticeable in large pale areas of scanned pictures. Also relevant in this context is the contrast range of the projector: the difference between its whitest white and darkest black. This is an important figure of merit for a digital projector and is always quoted by manufacturers. The number of tones which can be reproduced has little relevance if they are squeezed into such a small range that the eye cannot distinguish them. In the past this has been a major concern when projecting confocal images, with detail disappearing in both highlights and shadows. This is an area in which digital projectors (of either technology) have made huge strides in recent years. DLP is normally regarded as leading in contrast ratio but in this test both projectors seemed comparable, with good rich blacks. Both projectors seemed evenly balanced in response time, with rapid mouse movements appearing equally (and acceptably) jerky in both (at 60Hz refresh rate). At very close quarters some misalignment of the different color images was visible with the LCD projector. This was invisible at normal viewing distance and may be inevitable in a projector with three different LCD arrays (especially one which is regularly transported). This would not be expected in a DLP projector because there is only one display element, but in fact some color fringing could still be seen at the edges of the screen, though not in the center. The verdict on this test was that the LCD projector was way ahead for text, diagrams, and other computer-generated graphics, but the DLP had the edge for micrographs and other real-world images. This is in line with the commonly accepted merits of the two technologies. The question of different gamma is likely to be significant when projecting confocal images, and in the rare case where one knows in advance which type of projector will be used, the images in a presentation could be adjusted to suit. But the most useful point to remember when giving a presentation at a conference is still to set your screen resolution to the native resolution of the projector.
HARD COpy When it comes to recording images, the confocal microscopist has to make a choice between two fundamentally different technologies. One option, photography, was in the past familiar ground to most microscopists. The other option, computer printers of one sort or another, are more likely to be relevant in the 21 st century.
Photographic Systems In the 10 years since the previous edition of this book, photography has almost completely disappeared from the cell biology laboratory. So far as confocal images are concerned, this is all to the good because there is a fundamental mismatch between film and the
Mass Storage, Display, and Hard Copy • Chapter 32
digital image. A photograph can reproduce far more tones than the 256 that an 8-bit image possesses, but the interposition of lenses and film means that pixel positions are not reproduced sharply and with complete accuracy. Blurring below the level of microscope resolution will not be noticeable in a conventional micrograph, but if pixels are not rendered clearly in a confocal image it will look soft - especially if there is any text superimposed on the image. At one time, screen-shooting devices were standard equipment with confocal microscopes but now they are no more than historical curiosities.
Digital Printers Printers typically work by putting dots on a page of paper. As printer technology has evolved, the size and resolution of these dots has become smaller and more precise, but in general the dots are still either present or absent, which limits the ability of a printer to represent images with a range of tones. However, these pixels are placed with extreme accuracy, so providing the data is handled properly (below), it is possible for each pixel in a confocal image to be printed sharply and in its correct place. There are two ways in which we can break up a gray-scale image into a pattern of black dots for printing: halftoning or dithering. Halftoning is the way images are reproduced in printed books and newspapers. The image is broken into a series of black dots of varying size, darker grays being represented by larger dots. Halftoning is unarguably the method of choice if the resolution of the output medium is sufficient. However, it will be clear that to produce halftone dots of varying sizes, each dot must be a multiple of the basic dot pattern of the printer. If the halftone dots are to be smail, the printer's basic dots must be very small. The halftone screen in a printed book is typically 133 or 150 dots per inch (Cox, 1987). A 1200dpi printer can give is 8 x 8 dots - 65 gray shades - within that resolution. To get 256 gray shades at 150dpi, we need a 2400dpi printer. Dithering uses a probabilistic method to decide whether a printer dot should be present or not. If the pixel is dark, there is a high probability that a black dot will be printed; if it is light that probability is low. The effect is of a grainy image without any reg-
591
ularly repeating pattern. When using a low-resolution output device, dithering is the only option; halftoning would result in an impossibly coarse screen. Figure 32.6 shows a magnified view of a confocal image of an integrated circuit, (IC) device, printed by halftoning and dithering. In the dithered image each pixel of the original micrograph is represented nine times, by a 3 x 3 matrix of laser dots, the probabilistic dithering calculation being applied independently each time. Thus, on average, a 50% gray would have either four or five dots black, the others being white, but which of the nine dots were black would vary each time. The eye can perceive about 64 shades of gray reflected from a solid surface. If fewer shades than this are used to reproduce an image, areas that should show a smooth transition in tone will reproduce as a series of bands. Because the human visual system is extremely sensitive to edges, this can be extremely distracting. This effect is termed posterizing. In printing it occurs when the printer is unable to reproduce at least 64 shades of gray. Figure 32.7 shows an example of this. In Figure 32.7(A), using 16 gray tones (roughly what an old 300dpi laser printer can give) the smoothly-graded gray appears as a series of discrete bands. To some extent, the problem can be reduced by combining dithering with halftoning - values which bridge the boundaries between the levels the printer can produce are randomized to decide which value they should have [Fig. 32.7(B)]. Ultimately, though, good reproduction of a confocal image will require at least 64 gray levels to be reproduced on paper [Fig. 32.7(C)]. Proper reproduction of a confocal image by a printer will generally require the image to be reproduced either pixel for pixel, or with an integer multiple (or fraction) of printer pixels or half-tone dots to each dot of the confocal image. If this is not the case, aliasing (Chapter 4, this volume) will generate artifactual patterning in the image. This is shown in Figure 32.8; scaling the original 465pixel wide image [Fig. 32.8(A)] to fit the common 512-pixel size has given the diagonal lines and the circle a very jagged appearance [Fig. 32.8(B)]. This introduces a considerable constraint on printing confocal images; photography can reproduce them with equal accuracy at any size, but printing works best with integer multiples or fractions of the image pixels. The only ways out of this are either to use such a high-resolution print device that it
FIGURE 32.7. Posterizing. The smooth ramp of shade in the original shows banding when reproduced with only 16 gray levels (A). This can be partly disguised by dithering, still only using 16 levels (8), but using 80 gray levels is enough to give a smooth-looking result (C).
592
Chapter 32 • G. Cox
FIGURE 32.S. Aliasing. The circle and the diagonal lines appear as smooth as the horizontal and vertical ones in the original image (A), but when it is scaled up slightly they become jagged (B). In the original image the curved and slanting lines are actually made up of black and various shades of gray, shown much enlarged in (C).
exceeds the Nyquist criterion at the output resolution or else to remap the image with sophisticated software (several algorithms are in general use, and bilinear or bicubic res amp ling are probably the commonest) to the output resolution. The first option is becoming more common as printer resolution improves but cannot yet be guaranteed. Figure 32.9 shows the effect of different remapping algorithms. The image (shown in the inset) is a tiny part of Figure 32.1(A), enlarged by the odd amount of 467%. Direct pixel scaling [Fig. 32.9(A)] gives, as expected, very poor results, and bilinear resampling [Fig. 32.9(B)] is very little better. Bicubic resampling [Fig. 32.9(C)] is hugely better and obviously the only useful choice in this case. Aliasing can also appear if the printer's gray levels do not match the intensity levels of the image. When a black diagonal, or curved, line is reproduced in a pixelated image it will appear as a mixture of gray and black pixels [Fig. 32.8(C)]. Pixels which lie wholly within the line are black; those which were partially intersected by the line are gray. So long as they are reproduced at their original intensity the line will retain the illusion of smoothness, but as soon as they are made lighter or darker the line will appear jagged. Color images present all the above problems, as well as some of their own. Most laser-scanning confocal microscopes (CLSM) do not produce color images in the sense that a conventional optical microscope does. Color images produced on a CLSM are either pseudo-color images in which a false-color palette is applied to a gray-scale image, or multi-channel images in which two or three different signals are each assigned to a different primary color. An image with a fluorescein signal in the green channel and a rhodamine image in the red channel might look very similar to a real color photomicrograph of the same slide, but the way the image is made up is very different. A further problem arises because multi-channel images, and some false-color palettes, tend to use fully saturated colors. These almost never occur in nature. On a monitor, which emits light, these can look very effective, not least because confocal microscopes commonly operate in dimly lit rooms. They can also make good slides. However, when printed on paper, where the image is created by light reflected from the paper through the ink, the image will look very dark. This problem is exacerbated by the different color models used to form the image. A computer image is usually stored, and always displayed, using an RGB (red/greenlblue) or additive color model. Adjacent points on the monitor screen emit light of the three primary colors,
FIGURE 32.9. Scaling techniques. A small part of the image shown in Figure 32.1 (A) enlarged by the odd amount of 467%, which cannot be achieved by simple scaling of pixel size. (A) Pixel enlargement, as well as looking blocky. severe aliasing is obvious. (Inset) The original image. (B) Bilinear resampling: aliasing is much reduced but the image is still blocky. Little use is made of the extra pixels now available. (C) Bicubic resampling gives a hugely better result; it cannot produce more resolution than the original data contains, but it does the optimum job of mapping it to the output resolution.
Mass Storage, Display, and Hard Copy • Chapter 32
and the different colors are created by adding these together. A printed page uses the CMY (cyan/magenta/yellow) or subtractive model. A red point is created by putting on the paper both magenta (which subtracts green from the reflected light) and yellow (which subtracts blue) so that only red will be reflected. [Commercial printing, and many computer printers, also add a black ink to compensate for the fact that the three color inks may not add up to a perfect black. This is then a CMYK (cyan/magenta/yellowlblack) color model.] Thus, reproducing each of the primary colors that look so brilliant when formed by a single phosphor on a monitor requires the light to pass through two separate color dyes before being reflected by the paper. To put back brilIiance into such an image, it is necessary to unsaturate the colors - to add some "white" into them. Many computer image-manipulation programs provide this facility. Often, some experimentation is required before a good screen image can be turned into a good print and it can be very useful to have a program that permits one to print an array of small test images, each made using slightly different settings.
593
with an inkjet of printing at lower quality and hugely lower cost for proofing and layout purposes.
Dye Sublimation Printers These printers have changed little since they the previous edition of this book. They are still an excellent output medium for routine production of photographic quality output, and although still not cheap, the price has not increased in line with inflation so they are not much more expensive per page than the inkjet for photo-quality output, though the purchase price is much greater. They use a fullpage sheet of ink for each color they print, but in this case the ink sublimes when heated and is absorbed by a specially coated sheet of paper. The vaporized ink will tend to diffuse laterally to a limited extent, making pixelation less obvious. More or less ink can be transferred, depending upon the amount of heat applied, so that true gray scales can be produced. The output from the best dye sublimation printers can be comparable to photography, but the cost is also similar - up to US$S .OO for an A4 size color print. The cost per page is fixed, unlike a laser or inkjet printer, where the cost per page depends upon the degree of coverage.
Laser Printers Laser printers are the workhorse of the modem office. Their crisp black type and graphics are unrivaled for most computer output. For reproduction of confocal (or other) images their abilities are more limited, though they can produce quite reasonable proof prints. Laser printers use a low power laser to write dots directly onto the charged drum of a photocopier. The limitation of a laser printer is that it is difficult to get the pattern of dots fine enough to reproduce a full range of tones by halftoning (above). However, laser printers with 1200dpi resolution are now common, and many also have the capability of modulating the size of the spot to some extent. Output from such a printer is adequate for many purposes, and the cost is far lower than either photography or higher-class printing. Their weakest point is that large expanses of black are still not rendered as uniformly as in a photograph. Color laser printers have onl y recently started to make a significant impact on the marketplace. This is partly because of a huge decrease in price, and partly because of improvements in resolution (600dpi is common) and in the tricks used by the built-in firmware, which at last make near-photographic quality routinely attainable. The quality does not yet match that offered by inkjet and dye sublimation printers, but on a cost-versus-quality basis it hugely exceeds it, so that a color laser printer is ideal for such purposes as printing preprints of journal articles in quantity.
Ink Jet Printers These have long ago taken over from dot matrix as the everyday printer for home and single-user office use. They operate by squirting small jets of ink on the paper (for best results, a slightly absorbent paper). They typically offer three- or four-color printing at much finer resolution than any other printers. Some use seven inks (high and low intensities of the three subtractive primaries) to give more realistic results. Though the tendency of the ink to spread limits their ultimate performance, it also helps improve the perceived realism of the image by making individual pixels less visible. Printing to photographic quality requires special paper and also uses large amounts of the expensive ink, so it is not cheap, but the results bear comparison with those from expensive dye sublimation printers. Because (unlike other printer technologies) there is no inherent limitation on the size of the paper, A3 and larger printers are readily available for such tasks as printing conference posters. Unlike dye sublimation printers there is always the option
CONCLUSION The big change since the previous edition of this book 10 years ago is that now mass-market media are effectively equal to the demands of the confocal user. It is a truism that for several years now developments in the personal computer market have been driven not by business or scientific usage, but by the domestic market. The requirements of games, music, digital photography, and video have been the driving factors for processor speed, interfaces, display quality, print output, and data storage. We no longer need specialist image manipulation hardware, custom video cards, non-standard monitors, dedicated data buses, or expensive storage devices. Computers are fast enough for the necessary image manipulation, everyday video cards handle 24-bit images at high resolutions and fast refresh rates, as well as providing hard-wired image manipulation functions. Monitors offer megapixel displays at high bit depths and refresh rates. Fire Wire and US8-2 will carry data faster than any point-scanning confocal can scan, and domestic video disks are big enough to handle huge data sets. Just about every home has a printer giving photographic quality output and these are printing higher resolution images than most microscope users generate. The major computer magazines generally run annual surveys of color printers and it is always worth seeking out the latest of these before making a purchase. A final word of warning: If you are evaluating a hard copy system, of whatever sort, insist on testing it on real confocal images from your own work. Every manufacturer has a gallery of images which reproduce superbly on his own hardware, and if you try to judge a system on the basis of such pictures you will be disappointed once you start using it yourself.
SUMMARY Bulk Storage Image compression has made huge strides but still , as always, needs to be used with care. The new wavelet compression system seems to offer little to the confocal user but the PNG format has at last gi yen us a lossless technique that works. Recordable CDs
594
Chapter 32 • G. Cox
are currently the most common and most reliable choice for mass storage, though recordable DVDs seem likely to take over during the lifetime of this edition. For archival purposes, there seems little merit in selecting the rewritable version of either of these media. Tiny solid-state FLASH memory devices have become a very effective way to carry presentations from place to place.
Display Monitors are no longer a problem; 24-bit displays of adequate size and refresh rate, as well as the display boards to drive them, are now the norm rather than expensive exceptions. Flat-screen LCD displays are still expensive, but worth the cost for their extra sharpness and for the complete elimination of flicker.
Hard Copy Inkjet printers have now essentially replaced dye sublimation printers for optimal photographic output. They have much lower purchase cost but the cost per glossy print is still high. Color laser printers are now more than adequate for proofing and preprints. Finally, if you are evaluating a hard copy system, insist on testing it on real confocal images from your own work.
ACKNOWLEDGMENTS I am very grateful to Mike Speak and Dennis Dwarte, who have been generous with their time and effort in helping me prepare this chapter. I also owe a debt of gratitude to Todd Brelje, who wrote the original outline but who was unable to proceed with the chapter due to other commitments. In spite of this, he gave much assistance in producing the original edition, for which Martin Wessendorf, Nick White, Andrew Dixon, and Jim Pawley also provided a tremendous amount of information and help. The revision for the 3rd edition was carried out while I was at the Instituto Gulbenkian de Ciencia, Oeiras, Portugal, and I thank them for their hospitality; Nuno Moreno and Jose Fefjo provided much help and assistance, and critical reading of the manuscript.
REFERENCES Anson, L.P., 1993, Fractal image compression, Byte 19(11): 195-202. Avinash, G.B., 1993, Image compression and data integrity in confocal microscopy, Proc. Microsc. Soc. Am., Jones and Begell, New York, 51: 206-207. Barnsley, M.P., and Hurd, L.P., 1993, Fractallmage Compression, A.K. Peters, Wellsley, Massachusetts. Compact Flash Association, 2003, The Compact Flash Standard, available from: http://www.compactftash.org/. Cox, G.e., 1987, What went wrong with my micrograph? Aust. EM Newslett. 16:12-13. Cox, G.e. , 1993, Trends in confocal microscopy, Micron 24:237-247. Cox, G.e., and Sheppard, e., 1993, Effects of image deconvolution on optical sectioning in conventional and confocal microscopes, Bioimaging 1:82-95. Cox, e.G., and Sheppard, C., 1999, Appropriate image processing for confocal microscopy, In: Focus on Multidimensional Microscopy (p'C. Cheng, P.P. Hwang, J.L. Wu, G. Wang, and H, Kim, eds.), World Scientific Publishing, Singapore, pp. 42-54. Deutsch, P.L., 1996, DEFLATE Compressed Data Format Specification version 1.3, available from: ftp://ftp.uu.netlgraphics/png/documentslzliblzdocindex.htm!. Digital Photography Review, 2003, Digital Film Comparison, available from: http://www.dpreview.com/articles/mediacompare/. Mortimer, P.J., and Sowerby, A.L.M., 1941, Wall's Dictionary of Photography, Sixteenth Edition, Iliffe & Sons, Ltd., London. Nathans, S.P., ed., 2003, Writable DVD Drives, available from: http://www.emedialive.com/r 18/. Nugent, W.R., 1989, Estimating the Permanence of Optical Disks by Accelerated Aging and Arrhenius Extrapolation, ANSI X3B 11/89-101. Pennebaker, W.B., and Mitchell, J., 1993, JPEG Stilllmage Compression Standard, Van Nostrand Rheinhold, New York. Powell, E., 2004, The Great Technology War: LCD vs. DLP, available from: http://www.projectorcentral.comllcd_dlp_update.htm. Redfern, A., 1989, A discrete transformation, Australian Personal Computer 10(12):144-154. Roelofs, G. , 2003, Portable Network Graphics, available from: http://www.libpng.org/pub/png/. Stinson, D, Ameli, P., and Zaino, N., 1995, Lifetime of KODAK Writable CD and Photo CD Media, available from: http://www.cd-info.com/CDIC/ Technology/CD-RlMedia/Kodak.html.
33
Coherent Anti-Stokes Raman Scattering Microscopy X. Sunney Xie, ji-Xin Cheng, and Eric Potma
INTRODUCTION Advances in biological sciences are often facilitated by new tools in microscopy. Confocal and nonlinear or multi-photon fluorescence microscopy (Chapters 21 and 28) have become powerful techniques for three-dimensional (3D) imaging of living cells. This coincides with developments of various natural and artificial fluorescent probes for cellular constituents (Chapters 16 and 17). For biochemical species or cellular components that neither fluoresce nor tolerate labeling, other contrast mechanisms with molecular specificity are needed. Vibrational microscopy based on infrared absorption and spontaneous Raman scattering has been used for chemically selective imaging. Although infrared microscopy is a powerful tool, it is limited to low spatial resolution because of the long wavelength of light used (Humecki, 1995; J amin et at., 1998). Furthermore, the absorption of water in the infrared region makes it difficult to image through living cells. In contrast, Raman microscopy can overcome these limitations, as the wavelength of the excitation light is much shorter and there is essentially no absorption of water at these wavelengths. Raman spectroscopy has been applied extensively to biological molecules and cellular constituents (Turrell and Corset, 1996; Puppels, 1999; Shafer-Peltier et al., 2002). Confocal Raman microscopy of biological samples has resulted in high-resolution images of living cells. However, the intrinsically weak Raman signal necessitates high laser power (typically >IOOmW) and long integration times, and the signal is often overwhelmed by the fluorescence background of the sample, limiting its application in biology. In this chapter, we present coherent anti-Stokes Raman scattering (CARS) microscopy, a nonlinear vibrational imaging technique that overcomes the limitations of the linear techniques mentioned above. Coherent anti-Stokes Raman scattering as a nonlinear optical process was first reported in 1965 by Maker and Terhune at Ford Motor Company (Maker and Terhune, 1965), and later named CARS by Begley and colleagues (1974). Since then, CARS spectroscopy has been used widely as a spectroscopic tool for chemical analyses in the condensed and gas phases and has become the most extensively used nonlinear Raman technique (Clark and Hester, 1988; Tolles et at., 1977). In CARS spectroscopy, a pump laser and a Stokes laser, with center frequencies of 1U
0
2
4 6 D lAp
p'
8
s
FIGURE 33.3. (Al Sketches of the Hertzian dipole radiation pattern; (B) Hertzian dipoles are coherently added up in the sample plane. (C) Addition of multiple Hertzian dipoles leads to constructive interference in the forward direction and destructive interference in the backward direction. (D) Forwardand backward-detected CARS signals as a function of the diameter D of a spherical sample centered at the focus.
o
10
20
30
Distance ( 11m)
40
50
thin disk, the signal goes in both forward and backward directions [Fig. 33.3(B)] (Cheng et aZ., 2002d). The forward and backward CARS signals provide complementary information about a sample. Forward-detected CARS (F-CARS) microscopy is suitable for imaging objects of a size comparable to or larger than the excitation wavelength. For smaller objects, the F-CARS contrast can be overwhelmed by the large non-resonant background from the solvent. The main purpose of E-CARS microscopy is to provide a sensitive means of imaging objects with an axial length smaller than the excitation wavelength, circumventing the large background from the solvent (Cheng et aZ., 200lb, 2002d; Volkmer et aZ., 2001). As a demonstration, Figure 33.4 shows the simultaneous FCARS and E-CARS images of an epithelial cell with the Raman shift tuned to the CH-stretching vibration frequency at 2847 cm- I • The resonant CARS signal arises from lipids that are rich in C-H vibration. Although the F-CARS image has large signal amplitudes, it has a constant non-resonant background from the surroundings. A better contrast for small objects is seen in the E-CARS image with minimal non-resonant background from the surroundings. Figure 33.5 shows the experimental configuration of the CARS microscope. The collineariy-overlapped pump and Stokes laser beams are tightly focused into a sample by a high-NA objective in an inverted microscope. A condenser lens (or objective lens) is used to collect the forward CARS signal. The forward propagating signal has a relatively small cone angle and an air condenser generally suffices for efficient signal collection (Cheng et aZ., 2002a). The large working distance of the air condenser increases the accessibility to the sample and thus facilitates imaging ofliving cells. The CARS signal is conveniently separated from the excitation beams using bandpass filters. The backward CARS signal is
iJ o~ o
10
20
30
40
50
Distance (11m)
FIGURE 33.4. Forward-detected (F-CARS) and epi-detected (E-CARS) images of an epithelial cell with the Raman shift tuned to the C-H stretching vibration at 2850cm-'. The intensity profiles of the two white lines are shown in the lower panels. Note the offset in the F-CARS image due to the non-resonant background of the solvent. The background-free E-CARS image shows better contrast for small features.
Coherent Anti-Stokes Raman Scattering Microscopy • Chapter 33
Non-descanned F-CARS detector Bandpass filter Condenser ~:!,,!!!!~- Sample
Objective lens
x-v
o
Dichroic mirror Bandpass filter
scanner Non-descanned E-CARS detector
FIGURE 33.5. Schematic of a CARS microscope with both forward and backward detection. The synchronized pump and Stokes picosecond pulse trains are collinearly combined on a dichroic mirror and are directed to a beam-scanning microscope. A high-NA objective focuses the beams to a diffraction-limited spot in the sample. CARS signal is detected simultaneously in the forward and backward direction. Non-descanned photomultipliers (PMT) are typically used as detectors.
collected with the same objective used for focusing the laser beams. passing through the dichroic beam-splitter that is used to reflect the incoming excitation beams. F- and E-CARS images are simultaneously recorded by raster scanning the two laser beams with a pair of galvanometer mirrors. With a laser-scanning CARS microscope using two near-infrared (IR) laser beams of high repetition rate (Cheng et al., 2002a), forward- and epi-detected images (512 x 512 pixels) from the same sample can be taken simultaneously in less than I s. The imaging properties of F-CARS and E-CARS have been characterized by using laser-scanning CARS images of polymer beads embedded in agarose gel (Cheng et ai., 2002a). The typical FWHMs of the lateral and axial intensity profiles for a 0.2 flm bead are 0.28 flm and 0.75 flm, respectively. Using a NA = 1.2 waterimmersion objective lens, the lateral FWHM for 0.1 flm beads is 0.23 flm. These FWHM should be used with caution as simple deconvolution methods cannot be applied to CARS images. It should be noted that, besides the mechanism associated with small scatterers, there are two additional mechanisms responsible for backward CARS. One occurs at the interface of two homogeneous media with different i 3 ) (Cheng et ai., 2002d), while the other results from the back reflection or backscattering of the forward CARS in a heterogeneous sample. Such animal tissue (Evans et aI., 2005).
OPTIMAL LASER SOURCES FOR COHERENT ANTI -STOKES RAMAN SCATTERING MICROSCOPY It is known that the non-resonant electronic contribution to the
signal can be enhanced by two-photon electronic resonance (Maeda et al., 1988). This is evidenced by the fact that CARS images of liposomes taken with visible dye lasers are dominated by the non-resonant background (Duncan, 1984). The use of nearIR laser beams reduced the two-photon enhancement of X~~ and
599
accordingly improved the image contrast (Zumbusch et al., 1999; Cheng et al., 200lb). On the other hand, CARS spectroscopy with visible laser beams can take advantage of one-photon electronic resonance (Hudson et al., 1976; Carreira et at., 1977; Dutta and Spiro, 1978; Dutta et ai., 1980; Andrews et at., 1981; Igarashi et al., 1981; Schneider et al., 1988; Ujj et al., 1994; Voroshilov et at., 1995) for chemical species with absorption in the visible wavelength range, such as cytochrome c (Dutta et al., 1980; Andrews et al., 1981), l3-carotene (Carreira et al., 1977; Dutta et al., 1980), bateriorhodopsin (Ujj et ai., 1994), and hemoglobin (Voroshilov et at., 1995). Electronically-resonant CARS provides a way to suppress the non-resonant solvent background by enhancing the signal from the scatterer. The signal enhancement and the photo damage associated with one-photon electronic resonance has yet to be explored in CARS microscopy. The 1999 work used femtosecond lasers that have a high peak power and are widely used in multi-photon fluorescence microscopy. However, it turns out that femtosecond pulses are not optimal for vibrational imaging. Whereas a few molecules such as water have broad Raman spectra, the typical Raman line width at room temperature is around IOcm- 1 in the condensed phase. The spectral width of a 100 femtosecond pulse is 333 cm- I , much wider than the Raman line width. Therefore, most of the pulse energy ends up being used to generate the non-resonant background. On the other hand, the spectral width of a picosecond pulse is comparable to the Raman line width, so that the excitation energy is fully utilized to maximize the vibrationally resonant CARS signal. It has been shown theoretically that the optimal signal-tobackground ratio occurs with spectral pulse widths of 1-2ps for a typical Raman band at room temperature (Cheng and Xie, 2003). A pulse width of a few picosecond provides a good compromise between the spectral resolution and the peak power, and improved the signal-to-background ratio. In 2001, Cheng and colleagues constructed a new CARS microscope with two synchronized picosecond titanium: sapphire (Ti: Sa) lasers, and experimentally demonstrated improved spectral resolution and high signal-tobackground ratio (Cheng et at., 2001 b). In general, the timing jitter between the pump and the Stokes pulses introduces a fluctuation to the CARS intensity and this limits the image acquisition rate. This is particularly important when two independent, passively mode-locked lasers are used. Recently, tight synchronization between the pulse trains with jitters of 20 to 100fs has been realized (Jones et al., 2002; Potma et al., 2002). This has allowed acquisition of high-quality CARS images using 2 ps pulse trains (Potma and Xie, 2003). Picosecond pulses from a Ti: Sa oscillator typically have an energy -6 nJ at an average power level of about 500 m Wand a repetition rate of around 80 MHz. Such laser pulses have been successfully applied for high-speed CARS imaging (Cheng et al., 2002a). The integrated CARS signal intensity can be enhanced by a factor of m 2 (= m 3/m) if the pump and Stokes pulse energies are increased by a factor of m and the repetition rate is lowered by the same factor. On the other hand, high peak power can be hazardous to the sample and a high repetition rate is beneficial for fast image acquisition. A repetition rate of 100 kHz to I MHz is a good compromise. Recently, Ye and co-workers have shown that enhancement of pulse energy can be realized through coherent storage of radiation in a high finesse cavity without a gain medium (Jones and Ye, 2002). Using a high finesse cavity equipped with a cavity dumper, Potma and colleagues demonstrated amplification of mode-locked picosecond pulses that is greater than a factor of 30 at a repetition
600
Chapter 33 • X.S. Xie et al.
rate of 253 kHz (Potma et at., 2003). These novel sources provide new possibilities for improving the speed and sensitivity of CARS imaging. The timing jitter between the pump and probe pulses can be avoided if one uses synchronously pumped optical parametric oscillator (OPO) systems. When a commercial, near-IR picosecond OPO based on periodically-poled nonlinear crystals, pumped by a mode-locked laser with semiconductor saturable absorber mirrors (SESAMs) becomes available, it can serve as an ideal laser source for CARS imaging.
SUPPRESSION OF THE NON - RESONANT BACKGROUND The suppression of both the non-resonant background from the solvent and that from the scatterers is key to improving the detection sensitivity and spectral selectivity of CARS microscopy. Several schemes for background suppression have been developed and are summarized as follows.
Use of Picosecond Instead of Femtosecond Pulses A straightforward way to minimize the non-resonant background is not to generate it in the first place. As outlined above, compared to the broadband pulses of fs lasers, the spectral bandwidth of 1 to 2 ps pulses matches the linewidth of the Raman bands and minimizes generation of non-resonant spectral components. A picosecond laser system thus represents a better light source for CARS imaging in terms of the signal-to-non-resonant background ratio.
Epi-Detection As already discussed, this method introduces a large wave vector mismatch, which acts as a size-selective filter that rejects the background signal (resonant or non-resonant) from the bulk solvent and allows high-sensitivity imaging of small objects (Volkmer et al., 2001; Cheng et al., 20mb). Because E-CARS rejects signal contributions based on size and not on vibrational properties, the nonresonant signal from the small objects themselves still contributes to the image. Hence, for weak resonant signals (e.g., from the protein amide I band), the vibrational sensitivity of E-CARS can be limited due to the non-resonant background from the small objects themselves.
detector blocks the non-resonant signal, whereas a portion of the differently polarized resonant signal leaks through the analyzer. The detected resonant signal is optimized when the polarization difference between p(3)R and p(3)NR is maximized. This is achieved by introducing a polarization difference of71.6° between the pump and the Stokes beams (Brakel and Schneider, 1988). The efficient background rejection of the P-CARS microscope permits vibrational imaging of intracellular proteins (Cheng et at., 2001 a). Figure 33.6(A) shows the P-CARS spectrum of Nmethylacetamide, a model compound containing the characteristic amide I vibration at 1652cm- 1, which is a signature band for peptides and proteins. As the P-CARS band positions coincide with the corresponding Raman band positions, one can assign the P-CARS bands based on the Raman literature. The difference in the relative intensity of the bands in the two spectra arises from the quadratic dependence on the number of vibrational oscillators and the Raman depolarization ratios of the bands. Figure 33.6(B,C) shows the background-free P-CARS images of an unstained epithelial cell with (Op - (Os tuned to the amide I band [Fig. 33.7(B)]. Tuning (Op - (Os away from the amide I band to 1745 cm- 1 resulted in a faint contrast [Fig. 33.7(C)], proving that the image contrast was due to proteins, the distribution of which is heterogeneous in the cell (Cheng et at., 2001 a).
Time-Resolved Coherent Anti-Stokes Raman Scattering Detection The vibrationally-resonant signal can be separated from the nonresonant electronic contribution by use of pulse-sequenced detection with femtosecond pulse excitation (Laubereau and Kaiser, 1978; Kamga and Sceats, 1980). In time-resolved CARS detection, a signal-generating probe pulse is time delayed with respect to a temporally-overlapped pump/Stokes pulse-pair. Because of the instantaneous dephasing time of the non-resonant signal, the non-resonant CARS signal only exists when the pump/Stokes pulse-pair overlaps with the probe pulse. On the other hand, the vibrationally-resonant CARS signal decays with the finite dephasing time of the vibrational mode. The dephasing time is related to the spectral width of the corresponding Raman band and is typically several hundred femtoseconds for a mode in the condensed phase (Fickenscher et at., 1992). The non-resonant background in the CARS images can thus be eliminated by introducing a suitable delay between the femtosecond pump/Stokes and the probe pulses. Time-resolved CARS imaging has been demonstrated by Volkmer and colleagues with a three-color excitation scheme (Volkmer et al., 2002). The disadvantages of this approach are the photodamage induced by the femtosecond pulses and the necessity of adding a third laser beam of a different color.
Polarization-Sensitive Detection This method is based on the different polarization properties of the electronic (p. C,,)
c:
Q)
::I
0"
... C,,)
t.I..
~ FIGURE 34.20. An amplitude modulation waveform for the RF carrier at the Larmor frequ ency (A) produces a band-limited excitation of the nuclear spins (8). If such an RF waveform is used in combination with a linear magnetic field gradient, then only a limited slice or slab will be excited perpendicular to the gradient direction, Three such slice-selective pulses could be combined to limit excitation to a cubic sub-volume within a sample,
Related Methods for Three-Dimensional Imaging • Chapter 34
621
c Primary Solenoid for Main Magnetic Reid Shim Coils Sample/RF Coil Assembly Gradient Coils Sample/RF Coil Assembly Base
D
r"i=~~~5i~~~!:: 9i1~~~:~~"'~:'--
SampleTube Sample RF Coli (Solenoid)
Sample/RF Coil Assembly Base
FIGURE 34.21. A typical MRM hardware configuration: (A) an 11.7 Tesla vertical bore superconducting magnet. This hardware is identical to that used for high-resolution NMR spectroscopy. (8) Three millimeter transverse RF solenoid (copper winding at upper center) and tuning circuitry for excitation and detection of the MR signal. (C) Simplified schematic cross-section of a vertical bore MRM setup. The gradient and shim coils are secured within the magnet bore and the RF coil/sample assembly introduced for each experiment. (D) Enlargement of a transverse configuration for the solenoidal RF coil and sample tube. The sample is often surrounded by a suitable fluid, such as buffered saline or culture medium. For fixed samples where background signal needs to be eliminated, the sample can be surrounded by 'H-free, fluorinated fluids. Environmental control can be provided by thermostatically controlled airflow through the bore of the coil/sample assembly base.
the carrier frequency of the excitation pulse limits the frequency content. so that the excitation becomes spatially limited in the presence of a field gradient. For example, alms shaped pulse might have a bandwidth of 4kHz that, in the presence of a lOmT/m field gradient would limit NMR excitation to a region approximately 1 cm wide in the gradient direction. but would be unlimited perpendicular to the gradient direction. This process is termed slice or slab selection and can be extended to line and volume selections by using multiple, selective RF pulse/gradient combinations. Slice selection is often used for in vivo imaging where interleaved acquisition of multiple 2D kxy-spaces separated in z is more time-efficient than scanning the complete 3D kxv:-space. In a typical MRM experiment, a sample, such as an anesthetized mouse or a fixed embryo, is secured within an appropriate holder and placed within the RF coil. Additional physiological monitoring, anesthesia lines, and environmental control are often integrated into the RF coils/sample assembly. The gradient and
shim coils are normally secured more permanently within the main magnet bore (Fig. 34.21) requiring only the RF coil and sample to be introduced for each experiment. The RF coil/sample assembly is then placed at the center of the main field and gradient coils for imaging. The shim coil currents are adjusted to minimize magnetic inhomogeneities in the sample, the synthesizer frequency set to the Larmor frequency determined from the received signal and the strength of the RF excitation pulses calibrated. The duration of an image acquisition may vary from less than a second to more than a day and is largely determined by the required spatial resolution and signal-to-noise ratio. The RF power deposition in the sample varies widely with pulse sequence design and has the potential to cause significant heating. Changes to the pulse sequence timing and RF pulse waveform specifications can minimize peak and average power deposition. Heating effects increase rapidly with the Larmor frequency (increasing field strength) and power deposition can be estimated, monitored and controlled for in vivo imaging studies.
622
Chapter 34 • J.M. Tyszka et al.
Magnetic Resonance Microscopy Hardware The hardware for MRM consists of three major components and their associated electronics: (I) a permanent or electromagnet that generates the main polarizing field, (2) one or more radiofrequency coils for excitation and signal detection, and (3) a three-axis set of magnetic-field-gradient coils. The sample is placed within the RF coil which in tum is placed within the gradient and the main solenoid. The main, polarizing field is typically generated by a cryogenically-cooled, superconducting electromagnet, although permanent magnets and even resistive electromagnets are in use. The uniformity or homogeneity of the polarizing field plays a critical role in successful high-quality MRM. Additional roomtemperature field correction or shim coils are routinely used to improve the fundamental homogeneity of the magnetic field within the imaging volume. RF coils for MRM are most commonly simple loops, solenoidal resonators, or a volume resonator design called a birdcage. A single coil may be used for both transmission/excitation and reception, or a larger transmitting volume coil with good RF field homogeneity can be used with a high-sensitivity receiver loop coil (surface coil). The RF power absorbed by a sample during imaging varies widely with the type of pulse sequence employed. Some sequences involve rapid RF pUlsing at relatively high power which ultimately will lead to tissue heating. Power deposition also increases rapidly with field strength for a given pulse sequence. The signal received from the RF coil is amplified by a low noise preamplifier prior to demodulation from the Larmor frequency (see sidebar) and digitized at rates from about 100kS/s to 2 MS/s. Raw k-space data is reconstructed and displayed by the operating-console computer. The coils generating magnetic field gradients (gradient coils) are usually wound on a cylindrical former within the main magnet bore. The coil windings are designed to produce independent, linear field gradients over the imaging volume in each of the three cardinal directions (x, y, z). The maximum gradient strengths required for MRM typically exceed SOOmT/m and may rise to more than S Tim in high-performance hardware. Even at these strengths, the maximum field perturbation from the gradient coils rarely exceeds 1% of the main magnetic field. A functional MRM system requires a high-stability, highaccuracy frequency synthesizer that provides the reference oscillation for both RF transmission and reception. Amplifiers for the RF and gradient coils are controlled by a pulse sequencer that in tum is coordinated by the operating-console computer.
Strengths and Limitations of Magnetic Resonance Microscopy Whereas magnetic resonance imaging of humans is very well established as a medical diagnostic tool, microscopy has developed at a slower rate. MRM cannot compete with optical microscopy in terms of spatial resolution and sensitivity, but has several unique strengths that benefit certain applications. Specifically, MRM is not limited by the opacity of an object to light, because it employs radiofrequency radiation with wavelengths much larger than the typical sample size. Optimizing MRM is a complex trade-off between acceptable signal-to-noise ratio, temporal resolution, and spatial resolution. The NMR experiment has an intrinsically low sensitivity, so requirements for minimum SNR tend to place lower bounds on spatial and temporal resolution. Consequently, acquiring high spatial resolution MRM with voxel sizes in the 10 to 100 micron
range typically requires tens of minutes to hours to achieve an acceptable SNR. Various estimates place the ultimate limit for MRM spatial resolution in the neighborhood of l!-lm for liquid water at room or physiological temperatures.! These estimates consider the effects of molecular diffusion, T2 relaxation, microscopic field inhomogeneity, and sensitivity (Callaghan, 1991). Other factors, including hardware design and sample size, lead to practical MRM isotropic resolution limits greater than 10 !-lm. The achievable spatial resolution is limited by sensitivity factors, specifically the signal-to-noise ratio achievable in a given acquisition time. As a general rule, high-resolution MRM with spatial resolutions less than 100!-lm benefits from the use of polarizing magnetic fields greater than 3 Tesla, sometimes as high as 17.S Tesla. Although this leads to a greater nuclear paramagnetic polarization, other factors such as increasing T! relaxation times, decreasing T2 relaxation times, and increased susceptibility-based field inhomogeneities tend to diminish the expected gains in sensitivity and signal-to-noise ratio efficiency. High-performance gradient hardware, with higher maximum amplitudes (Tim) and faster slew rates (Tlmls) are almost always an advantage in MRM. Microscopy gradients capable of 10 Tim allow high spatial and temporal resolution imaging of small samples. Another approach to increasing the sensitivity of MRM is through RF coil design. Various groups have developed microcoils with dimensions smaller than Imm to boost sensitivity in very small samples. Some of the highest resolution MRM images obtained have employed such designs (Grant et aI., 2001; Lee et aI., 2001; Ciobanu and Pennington, 2004).
Image Contrast in Magnetic Resonance Microscopy Signal differences between tissues in !H MRM of biological samples largely arise from differences in the microscopic environment of intracellular and interstitial water. Chemical differences between molecules containing !H nuclei also contribute to image contrast, for example, between water and lipid-rich tissues. Parameters influencing MRM image contrast include T! and T2 relaxation times, NMR nucleus concentration, temperature, diffusion coefficient, fluid velocity, magnetic susceptibility, and magnetization transfer coefficients of a material. The influence that each of these parameters has on image contrast can be accentuated or suppressed by careful design of the sequence of radiofrequency and field gradient pulses played out prior to signal acquisition. For example, specific pulse sequence designs allow the time delay between excitation and signal acquisition to be increased, accentuating differences in the T2 relaxation time of materials. Reducing the time delay between excitation pulses accentuates short T! materials that appear brighter in such images. Motion of water and other fluids can be encoded in the phase or amplitude of MRM images using gradient pulse combinations. Both coherent motion (flow) and incoherent motion (diffusion) can be quantified by MRM over a 2D plane or 3D volume of an optically opaque material. MRM measurements of water diffusion within organized tissues have been particularly well developed and are capable of quantifying subresolution ensemble molecular interactions with restrictive or hindering boundaries (Tuch, 2003). For example, diffusion tensor imaging (DTI) maps diffusion
I
About 3 to 4!lm has been obtained on very small specimens (9 Tesla; Ciobanu and Pennington, 2004).
Related Methods for Three-Dimensional Imaging • Chapter 34
anisotropy within tissues using a simplified model of moleculeboundary interactions (Pierpaoli et al., 1996). Diffusion tensor MRM has been applied to brain development in mice (Zhang et al., 2003), dysmyelination models (Song et al., 2002), and myocardial fiber structure (Jiang et al., 2004). In addition to endogenous image contrasts, such as relaxation time and molecular diffusion, it is possible to introduce exogenous MRM contrast agents that target specific tissues or physiology within a living system. Gadolinium chelates are widely used, both individually and bound to larger molecules, to reduce the T, relaxation time of surrounding water (Weinmann et aI. , 1984, 2003). Ionic manganese is a calcium analog and TI contrast agent that, at low concentrations, has been used as an in vivo trans-synaptic axonal transport tracer (Pautler and Koretsky, 2002; Pautler et al., 2003; Aoki et al., 2004). Tracking of progenitor, stem, or immune cell s using MRM is a recent growth area (Hinds et al., 2003; Shapiro et al., 2004) typically employing variants of superparamagnetic iron oxide (SPIO) particles as intracellular T2* contrast agents (Foster-Gareau et al., 2003). The development of molecular imaging for MRM, targeting both gene expression and metabolism within a living organism is likely to be the next significant field of investigation (Louie, 2000; Allen, 2004) and holds great promise for future non-invasive biomedical research.
623
Magnetic Resonance Microscopy Applications Phenotyping MRM is widely used for phenotyping genetically manipulated organisms such as transgenic mice (Johnson et al., 2002a, 2002b; Lo et al., 2003) leading to the development of high throughput techniques for imaging more than one animal simultaneously (Bock et al. , 2003). MRM brain atlases of adult inbred mouse strains are being constructed by several groups (Kovacevic et al., 2005; Segars et al., 2004).
Histology Ex vivo MR histology exploits the lack of physiological motion and extended data acquisition times to generate high-resolution structural images of various organisms (Figs. 34.22 and 34.23) (Johnson et al., 1993 ; Johnson et al., 2002a). MRM lacks the spatial resolution and stai ning flexibility of optical histology, but preserves the 3D structure of tissues and eliminates the need for dehydration, embedding, and physical sectioning and their associated artifac ts. As a non-invasive technique, MRM can even precede conventional histology, providing a valuable structural reference for subsequent histological sections.
FIGURE 34.22. Examples of MRM histology of biological samples. (A) High-resolution 3D M RM of a fixed late gastrula stage Xenopus laevis embryo with a nominal isotropic spatial resolution of 16 microns. The vegetal cell mass (vern), blastopore (bp), and archenteron (arch) are clearly visualized in this section. (B) Volume texture rendering of a 3D MRM image of a fixed 8.5dpc mouse embryo within its yolk sac. (C) Central section through a composite MRM image of a fixed mouse eye. The red channel encodes the isotropic-diffusion-weighted image, highlighting regions of restricted diffusion within the chorioretina (orange). The green channel encodes the To-weighted image in which free-fluid appears bright. The low intensity circular feature within the globe of the eye is the crystall in e lens. (D) Montage of sections from a 3D MRM image of a fixed 7 dpc mouse embryo with nominal 20 micron isotropic spatial resolution.
624
Chapter 34 • J.M. Tyszka et al.
FIGURE 34.23. Volumetric rendering of an ll-day postcoital mouse embryo. Because the sample has not been dehydrated. the global geometric preservation of the embryo is excellent.
Developmental Biology MRM has been applied to structural imaging of mouse (Jacobs, 1999; Chapon, 2002) and quail embryos and to serial in vivo imaging of the Xenopus iaevis embryo (Pap an et ai., 2001 ; Jacobs et ai., 2003). MRM leverages its ability to image optically opaque specimens to generate 3D data which would be difficult or impossible to obtain by other methods . MRM is particularly suited to studies of dynamic developmental processes such as morphogenesis. As for phenotyping, MRM is proving valuable in the construction of reference developmental atlases for mouse, quail, and other embryos that are too large or opaque at later stages for optical imaging (Dhenain et ai., 2001; Matsuda et at., 2003).
Other Applications MRM is in many ways underexploited outside the realm of vertebrate imaging. MRM is an excellent resource for botanical imaging, exploiting minimal physiological motion to generate high spatial resolution images (Edzes, 1998; Kockenberger, 2001, 2004) and is especially suited to imaging flow and water distribution within plant samples (Kuchenbrod et ai., 1998). MRM has also seen applications in opaque bioreactors and biofilms, studying local flow and structure beyond the capabilities of optical microscopy (As and Lens, 2001; Paterson-Beedle et al., 2001; Manz et al., 2003; Seymour et al., 2004).
Future Development of Magnetic Resonance Microscopy The future of MRM is likely to include incremental improvements in spatial resolution and sensitivity as hardware design and fabrication methods are refined. Ironically, clinical MRI has led the way in technical advances, driven by the needs and resources of medical diagnosis and therapy. Some of these technologies, including
phased-array RF coils and advanced pulse sequence designs, will be reapplied to MRM applications. However, it is in the development of new molecular imaging agents in model organisms that we will most likely see the greatest advances in MRM in the near future.
CONCLUSION The imaging modalities discussed in this chapter are emerging tools directed at meeting the growing need for volumetric microscopy of challenging specimens. They join confocal laserscanning and two-photon laser-scanning microscopy in the suite of microscopic imaging techniques that enable entire, wholemounted specimens to be studied. Each of these volumetric imaging tools plays an important role in the imaging toolkit, as none answers all needs. SIMIEFIC is a high-resolution method that is capable of capturing the entire volume of large specimens; however, it is inherently destructive as biological samples must be fixed, dehydrated, and embedded, and then must be sectioned. Although imaging the block face avoids the distortions resulting from the sectioning process, the specimen is still subject to any fixation and dehydration artifacts. As presently employed, OPT requires extensive clearing of a specimen to render it translucent, and so is subject to the same fixation and shrinkage artifacts as SIMIEFIC. A major strength of OPT on fixed specimens is its ability to capture the distribution of molecular markers such as antibody staining and gene expression in situ using common molecular biological and histological techniques. Of course, on living specimens that are small and transparent, OPT can be performed without processing, and at the time of writing, live OPT imaging experiments are being performed. A subset of the volumetric imaging tools are designed to image living specimens: OCT, SPIM, and MRM. OCT, a reflected light
Related Methods for Three-Dimensional Imaging • Chapter 34
modality, requires no extrinsic labels and is ideal for optical imaging deep within specimens, and for following motions such as blood flow. SPIM offers an important tool at the confluence of imaging and molecular biology technologies, as it is a fluorescent optical sectioning technique, ideal for capturing the expression of GFP reporter genes in living specimens. The use of transgenic GFP reporter constructs allows for the direct in vivo visualization of gene transcription during development within the context of the living embryo: the SPIM sample chamber is designed for maintaining physiological conditions and the use of the light sheet minimizes sample bleaching and reduces blur from structures not in the focus plane. Using SPIM, dynamic events ranging from fractions of a second to days can be recorded at voxel sizes of -0.5~m.
MRM is not an optical modality and therefore is not limited by the optical properties of the subject. MRM is non-invasive and allows long-term observation of living biological specimens as no ionizing radiation is employed. Although MRM is slower, more expensive, and offers lower resolution than most optical methods, it offers direct observation of morphogenic movements in opaque embryos. Previously, these movements could only be inferred from static sections. Clinically, MRI has a proven diagnostic history but is only in its infancy as a microscopic technique. Tools proven in the clinic are being modified for microscopic use, promising a growing set of techniques for following tissue fine structure and blood flow. As with any review of finite length, there are a great many modalities and modifications to existing techniques that have not been treated here. Adapting non-conventional elements to conventional devices pushes microscopy development forward. The field of view of confocal microscopy has been greatly increased (20 mm 2 at 2 ~m resolution) with the use of a high-NA f-theta telecentric objective (Dixon, 1995). Microscopic versions of clinical imaging tools are emerging and improving at an alarming rate. These include positron emission tomography (micro-PET), and ultrasonography (micro-US). Typing in any of these key terms in a Web- or library-based search will find a rapidly growing set of offerings. The resolution of these techniques has improved dramatically. The omission of these powerful approaches is in no way intended to diminish their potential role for volumetric imaging, but instead a statement of how rapidly these fields are advancing. The future will see the toolkit for volumetric imaging grow at an increasing rate.
ACKNOWLEDGMENTS The authors thank Jeff Fingler for additional expert OCT content.
REFERENCES Allen, M.1., and Meade, T.1., 2004, Magnetic resonance contrast agents for medical and molecular imaging. Met. Ions. BioI. Syst. 42: 1-38. Aoki, I., Wu, Y.J., Silva, A.e., Lynch, R.M., and Koretsky, A.P., 2004, In vivo detection of neuroarchitecture in the rodent brain using manganeseenhanced MRI, Neuroimage 22:1046-1059. As, H.Y., and Lens, P., 2001, Use of IH NMR to study transport processes in porous biosystems, f. Ind. Microbiol. Biotechnol. 26:43-52. Bock, N.A., Konyer, N.B., and Henkelman, RM., 2003, Multiple-mouse MRI, Magn. Reson. Med. 49:158-167. Boppart, S.A., Bouma, B.E., Pitris, e., Southern, J.F., Brezinski, M.E., and Fujimoto, J.G., 1998, In vivo cellular optical coherence tomography imaging, Nat. Med. 4:861-865.
625
Boppart, S.A., Brezinski, M.E., Bouma, B.E., Tearney, G.J., and Fujimoto, J.G., 1996, Investigation of developing embryonic morphology using optical coherence tomography, Dev. Bioi. 177:54-63. Boppart, S.A., Tearney, G.J., Bouma, B.E., Southern, J.F., Brezinski, M.E., and Fujimoto, J.G., 1997, Noninvasive assessment of the developing Xenopus cardiovascular system using optical coherence tomography, Proc. Natl. Acad. Sci. USA 94:4256-4261. Callaghan, P.T., 1991, Principles of Nuclear Magnetic Resonance Microscopy, Oxford University Press, Oxford, United Kingdom. Chapon, e., Franconi, F., Roux, J., Marescaux, L., Le Jeune, J.1., and Lemaire, L., 2002, In utero time-course assessment of mouse embryo development using high resolution magnetic resonance imaging, Anat. Embryol. 206(1-2): 131-137. Ciobanu, L., and Pennington, e.H., 2004, 3D micron-scale MRI of single biological cells, Solid State Nuc!. Magn. Reson. 25:138-141. Dhenain, M., Ruffins, S.W., and Jacobs, R.E., 2001, Three-dimensional digital mouse atlas using high-resolution MRI, Dev. BioI. 232:458-470. Dixon, A.E., Damaskinos, S., Ribes, A., and Beesley, K.M., 1995, A new confocal scanning beam laser macroscope using a telecentric, f-theta laser scan lens, f. Microsc. 178:261-266. Edzes, H.T., van Dusschoten, D., and Van As, H., 1998, Quantitative T2 imaging of plant tissues by means of multi-echo MRI microscopy, Magn. Reson. Imaging 16(2):185-96. Ewald, A.J., McBride, H., Reddington, M., Fraser, S.E., and Kerschmann, R., 2002, Surface imaging microscopy, an automated method for visualizing whole embryo samples in three dimensions at high resolution, Dev. Dyn. 225:369-375. Ewald, A.J., Peyrot, S.M., Tyszka, J.M., Fraser, S.E., and Wallingford, J.B., 2004, Regional requirements for Dishevelled signaling during Xenopus gastrulation: Separable effects on blastopore closure, mesendoderm internalization and archenteron formation, Development 131:6195-6209. Fercher, A.F., 1996, Optical coherence tomography, f. Biomed. Opt. I: 157-173. Foster-Gareau, P., Heyn, C., Alejski, A., and Rutt, B.K., 2003, Imaging single mammalian cells with a 1.5T clinical MRI scanner, Magn. Reson. Med. 49:968-971. Grant, S.e., Buckley, D.L., Gibbs, S., Webb, A.G., and Blackband, S.1., 2001, MR microscopy of multicomponent diffusion in single neurons, Magn. Reson. Med. 46:1107-1112. Haskell, Re., Williams, M.E. , Petersen, D.e., et al., 2004, Visualizing early frog development with motion-sensitive 3-D optical coherence microscopy, In: Proceedings of the 26th Annual International Conference of the IEEE EMBS, IEEE, San Francisco, CA, USA. Herman, G.T., 1980, Image Reconstruction from Projections: The Fundamentals of Computerized Tomography, Academic Press, San Francisco. Hinds, K.A., Hill, J.M., Shapiro, E.M., et al., 2003, Highly efficient endosomal labeling of progenitor and stem cells with large magnetic particles allows magnetic resonance imaging of single cells, Blood 102:867-872. Hoeling, B.M., Fernandez, A.D., Haskell, Re., and Petersen, D.C., 2001, Phase modulation at 125 kHz in a Michelson interferometer using an inexpensive piezoelectric stack driven at resonance, Rev. Sci. Instrum. 72: 1630-1633. Hoeling, B.M., Fernandez, A.D., Haskell, Re., Huang, E., Meyers, W.R., Petersen, D.e., Ungersma, S.E., Wang, R, and Williams, M.E., 2000, An optical coherence microscope for 3-dimensional imaging in developmental biology, Opt. Express 6:136-146. Holdsworth, D.W., and Thornton, M.M. , 2002, MicroCT in small animal and specimen imaging, Trends Biotechnol. 20(Suppl.):S34-S39. Huang, D. , Swanson, E.A., Lin, e.P., Schuman, J.S., Stinson, W.G., Chang, W., Hee, M.R, Flotte, T., Gregory, K., Puliafito, e.A., Fujimoto, J.G., 1991, Opt. Coherence Tomogr. Sci. 254: 1178-1181. Huisken, J., Swoger, l., Del Bene, F., Wittbrodt, l., and Stelzer, E.H.K., 2004, Optical sectioning deep inside live embryos by selective plane illumination microscopy, Science 305:1007-1009. Izatt, l.A., Kulkarni, M.D., Wang, H.-W., Kobayashi, K., and Sivak, M.V. Jr., 1996, Optical coherence tomography and microscopy in gastrointestinal tissues, IEEE f. Sel. Topics Quant. Electron. 2: 1017-1028. Jacobs, RE., Ahrens, E.T., Dickinson, M.E., and Laidlaw, D., 1999, Towards a microMRI atlas of mouse development, Comput. Med. Imaging Graph. 23(1):15-24.
626
Chapter 34 • J.M. Tyszka et al.
Jacobs, R.E., Papan, c., Ruffins, S., Tyszka, J.M., and Fraser, S.E., 2003, MRI: Volumetric imaging for vital imaging and atlas construction, Nat. Cell Bioi. SslO-SsI6. Jiang, Y., Pandya, K., Smithies, 0., and Hsu, E.W., 2004, Three-dimensional diffusion tensor microscopy of fixed mouse hearts, Magn. Reson. Med. 52:453-460. Johnson, G.A., Benveniste, H., Black, R.D., Hedlund, L.W., Maronpot, R.R., and Smith, B.R., 1993, Histology by magnetic resonance microscopy, Magn. Reson. Q 9:1-30. Johnson, G.A., Cofer, G.P., Fubara, B., Gewalt, S.L., Hedlund, L.W., and Maronpot, R.R., 2002a, Magnetic resonance histology for morphologic phenotyping, I. Magn. Reson. Imaging 16:423-429. Johnson, G.A., Cofer, G.P., Gewalt, S.L., and Hedlund, L.W., 2002b, Morphologic phenotyping with MR microscopy: The visible mouse, Radiology 222:789-793. Keng, W.T., Sharpe, J., et al., 2002, Optical projection tomographic examination of miscarried human embryos, I. Med. Genet. 39:S23-S23. Kerwin, J., Scott, M., Sharpe, J., et al., 2004, 3 dimensional modelling of early human brain development using optical projection tomography, BMC Neurosci.5:27-27. Kockenberger, W., 2001, Nuclear magnetic resonance micro-imaging in the investigation of plant cell metabolism, I. Exp. Bot. 52(356):641-652. Kockenberger, W., De Panfil is, c., Santoro, D., Dahiya, P., and Rawsthorne, S., 2004, High resolution NMR microscopy of plants and fungi, I. Micros. 214(Pt2): 182-189. Kovacevic, N., Henderson, J.T., Chan, E., et aI., 2005, A three-dimensional MRI atlas of the mouse brain with estimates of the average and variability, Cereb. Cortex. 15(5):639-645. Kuchenbrod, E., Kahler, E., Thurrner, F, Deichmann, R., Zimmermann, U., and Haase, A., 1998, Functional magnetic resonance imaging in intact plants-quantitative observation of flow in plant vessels, Magn. Reson. Imaging 16:331-338. Lee, S.c., Kim, K., and Kim, J., 2001, One micrometer resolution NMR microscopy, I. Magn. Reson. 150:207-213. Lo, c., Nabel, E., and Balahan, R., 2003, Meeting report: NHLBI symposium on phenotyping: Mouse cardiovascular function and development, Physiol. Genomics 13(3):185-186. Louie, A.Y., Huber, M.M., Ahrens, E.T., et aI., 2000, In vivo visualization of gene expression using magnetic resonance imaging. Nat. Biotechnol. 18(3):321-325. Mansfield, P., 1977, Multi-planar image formation using NMR spin echoes, I. Phys. C 10:L55-L58. Manz, B., Volke, F, GolI, D., and Horn, H., 2003, Measuring local flow velocities and biofilm structure in biofilm systems with magnetic resonance imaging (MRI), Biotechnol. Bioeng. 84:424-432. Masters, B.R., 1999, Early development of optical low-coherence reflectometry and some recent biomedical applications, I. Biomed. Opt. 4:236247. Matsuda, Y., Utsuzawa, S., Kurimoto, T., 2003, Super-parallel MR microscope, Magn. Reson. Med. 50:183-189. Papan, C., Velan, S.S., Fraser, S.E., and Jacobs, R.E., 2001, 3D time-lapse analysis of Xenopus gastrulation movements using mu MRI, Dev. Bioi. 235:189. Paterson-Beedle, M., Nott, K.P., Macaskie, L.E., and Hall, L.D., 2001, Study of biofilm within a packed-bed reactor by three-dimensional magnetic resonance imaging, Methods Enzymol. 337:285-305. Paulus M.1., Gleason, S.S., Easterly, M.E., and Foltz, C.1., 2001, A review of high resolution X-ray computed tomography and other imaging modalities for small animal research, Lab. Animal 30(3):36-45. Paulus, M.J., Gleason, S.S., Kennel, S.1., Hunsicker, P.R., and Johnson, D.K., 2000, High-resolution x-ray computed tomography: An emerging tool for small animal cancer research, Neoplasia 2:62-70. Pautler, R.G., and Koretsky, A.P., 2002, Tracing odor-induced activation in the olfactory bulbs of mice using manganese-enhanced magnetic resonance imaging, Neuroimage 16:441-448.
Pautler, R.G., Mongeau, R., and Jacobs, R.E., 2003, In vivo trans-synaptic tract tracing from the murine striatum and amygdala utilizing manganese enhanced MRI (MEMRI), Magn. Reson. Med. 50:33-39. Pickhardt, P.1., Halberg, R.B., Taylor, A.J., Durkee, B.Y., Fine, J., Lee, Jr., FT., and Weichert, J.P', 2005, Microcomputed tomography colonography for polyp detection in an in vivo mouse tumor model. PNAS 102:3419-3422. Pierpaoli, C., Jezzard, P., Basser, P.1., Barnett, A., and DiChiro, G., 1996, Diffusion tensor MR imaging of the human brain, Radiology 201 :637-648. Robson, S.c., Woods, H.M., et al., 2004, 3-dimensional modelling of human development using optical projection tomography, I. Soc. Gynecol. Invest. 11:296A. Rollins, A.M., and Izatt, J.A., 1999, Optimal interferometer designs for optical coherence tomography, Opt. Lett. 24:1484-1486. Rollins, A.M., Kulkarni, M.D., Yazdanfar, S., Ung-arunyawee, R., and Izatt, J.A., 1998, In vivo video rate optical coherence tomography, Opt. Express 3:219-229. Segars, WP., Tsui, B.M., Frey, E.C., Johnson, G.A., and Berr, S.S., 2004, Development of a 4-D digital mouse phantom for molecular imaging research, Mol. Imaging Bioi. 6:149-159. Seymour, J.D., Codd, S.L., Gjersing, E.L., and Stewart, P.S., 2004, Magnetic resonance microscopy of biofilm structure and impact on transport in a capillary bioreactor, I. Magn. Reson. 167:322-327. Shapiro, E.M., Skrtic, S., Sharer, K., Hill, J.M., Dunbar, C.E., and Koretsky, A.P., 2004, MRI detection of single particles for cellular imaging, Proc. Natl. Acad. Sci. USA 101:10901-10906. Sharpe, J., 2003, Optical projection tomography as a new tool for studying embryo anatomy, I. Anat. 202:175-181. Sharpe, J., 2004, Optical projection tomography, Annu. Rev. Biomed. Eng. 6:209-228. Sharpe, J., Ahlgren, U., Perry, P., et aI., 2002, Optical projection tomography as a tool for 3D microscopy and gene expression studies, Science 296:541-545. Song, S.K., Sun, S.W., Ramsbottom, M.J., Chang, C. Russell, J., and Cross, A.H., 2002, Dysmyelination revealed through MRI as increased radial (but unchanged axial) diffusion of water, Neuroimage 17: 1429-1436. Tearney, G.J., Bouma, B.E., Boppart, S.A., Golubovic, B., Swanson, E.A., and Fujimoto, J.G., 1996, Rapid acquisition of in vivo biological images by use of optical coherence tomography, Opt. Lett. 21: 1408-141 O. Tearney, G.1., Brezinski, M.E., Bouma, B.E., Boppart, S.A., Pitris, c., Southern, J.F., and Fujimoto, J.G., 1997, In vivo endoscopic optical biopsy with optical coherence tomography, Science 276:2037-2039. Tuch, D.S., Reese, T.G., Wiegell, M.R., and Wedeen, V.J., 2003, Diffusion MRI of complex neural architecture, Neuron 40:885-895. Weichert, J.P., 2004, Micro-computed tomography of mouse cancer models, In: Mouse Models of Human Cancer (E.C. Holland, ed.), John Wiley and Sons, New York, pp. 339-348. Weinmann, H.J., Brasch, R.c., Press, W.R., and Wesbey, G.E., 1984, Characteristics of gadolinium-DTPA complex: A potential NMR contrast agent, AIR Am I Roentgenol 142:619-624. Weinmann, H.J., Ebert, W., Misselwitz, B., and Schmitt-Willich H., 2003, Tissue-specific MR contrast agents, Eur. I. Radiol. 46:33-44. Weninger, W., and Mohun, T., 2002, Cardiac phenotyping of transgenic embryos: A rapid 3D-screening method based on episcopic fluorescence imaging capturing (EFIC), Clin. Exp. Pharmacol. Physiol. 29:A 70--A 71. Wojtkowski, M., Srinivasan, V.J., Ko, T.H., Fujimoto, J.G., Kowalczyk, A., and Duker, J.S., 2004, Ultrahigh-resolution, high-speed, Fourier domain optical coherence tomography and methods for dispersion compensation, Optics. Express 12(11):2404-2422. Yazdanfar, S., Kulkarni, M.D., and Izatt, I.A., High resolution imaging of in vivo cardiac dynamics using color Doppler optical coherence tomography, Opt. Express 1:424-431. Zhang, J., Richards, L.I., Yarowsky, P., Huang, H., van Zijl, P.c., and Mori, S., 2003, Three-dimensional anatomical characterization of the developing mouse brain by diffusion tensor microimaging, Neuroimage 20: 1639-1648,
35
Tutorial on Practical Confocal Microscopy and Use of the Confocal Test Specimen Victoria Centonze and James B. Pawley
INTRODUCTION The other chapters in this book give the reader an in-depth description of every important aspect of biological confocal microscopy that we could think of. This chapter is to provide the novice user of this instrument with a basic understanding of the practical information needed to use it effectively. Because the computer interfaces of the various commercial instruments vary greatly, this chapter will stress the important features of microscopical optics and the basics of sampling that are common to all instruments. The underlying agenda of these suggestions is based on two principles: • Don't waste photons . • Get all of the optical performance that you have paid for.
few seconds, the image will begin to fade and become indistinct. Reducing the zoom magnification will reveal the terrible truth: the observations so far have severely bleached the dye in the rectangular area that was being scanned at the higher zoom magnification. This is the moment at which would-be users either decide right then and there that the confocal microscope clearly produces far more bleaching than they are used to and leave the room, never to return, or they decide that there may be a bit more to operating this instrument than they had bargained for and set out to do better. This chapter is written for those who fall into the latter category.
Bleaching - The Only Thing That Really Matters
Getting Started
There are three reasons that the bleaching rate of a confocal microscope often seems to be much higher than that of a conventional epi-fluorescent microscope:
In order to be able to operate a confocal microscope properly, one must first be thoroughly familiar with the basic principles of nonconfocal microscopy. The user must understand the concept of conjugate image planes, Kohler illumination, and how to set up a microscope to produce it [i.e. , the field diaphragm must be in focus in the image plane and the condenser aperture set so that the illumination almost fills the back-focal plane (BFP) of the objective lens. See the Appendix to this chapter or Chapter 6, this volume, for a refresher]. The user must also understand how the phenomenon of diffraction acts to place the fundamental limitation on the resolution that any optical system can attain (see Chapters I, 8, and II , this volume, or an introductory text such as Bradbury, 1984, Inoue 1986, Pawley and Centonze, 1997). The next step is to get an image of some specimen that you understand. Without going into the details of the many available commercial confocal systems, we note that most confocal microscopes permit the user to set up the microscope for normal, nonconfocal use, and then switch over to confocal operation by changing the position of a single control. The user should follow the in structions from the manufacturer to produce such a confocal image using a live mode (i.e., producing an image that is continuous and not frame-averaged or Kalman filtered). In most cases, the fluorescent specimen chosen for this first attempt at using the confocal microscope should be similar or identical to one that has just been viewed with success on a conventional instrument. Almost invariably, new users will close the detection pinhole in order to get the highest resolution and increase the zoom magnification to increase the size of the image on the display. They will then try to adjust the focus , but after a
I. The instrument can be used in such a way that the supposition is correct! A laser is a far more intense light source than a Hg arc. Although the total power striking the entire specimen may be less with the laser, the area of the specimen over which it is absorbed can be reduced arbitrarily by increasing the setting of the zoom magnification control, thus increasing the power/unit area. The crucial difference between the two microscopical methods is that in normal epi-fluorescence, the power/area on the specimen is fixed by the type of arc and the illumination optics, while in confocal microscopy, the power/area increases with the square of the zoom magnification. In normal epi-fluorescence, one does not expect to sec an image that is both large enough to see the finest details and bright enough to view by eye because it is not possible. In the confocal microscope , the extreme intensity of the laser does make it possible, though only for the moment until bleaching occurs! 2. If one is accustomed to viewing a widefield (WF) image in which fluorescence from a large number of focal planes is added together into a single image, a confocal image may look somewhat anemic because, ideally, it records only the fluorescence features present in a single optical section. When presented with an image of the few features that happen to be in the focus plane, one may be tempted to use a much longer exposure than is necessary in a vain attempt to record features that are not really there. 3. A final factor contributing to the perception that confocal bleaches faster is that, in normal epi-fluorescence, one expects to make one exposure that may require 30 to 60 s, while in the confocal microscope one often wants an image that is bright enough to look at after collecting for only I to 2 s and then sets out to
Vi cto ri a Centonze • University of Texas Health Science Center, San Antonio, Texas 78229 James B. Pawley. University of Wisco nsin, M adison, Wisconsin 53706 Handbook of' BioloRical Confocal Microscopy. Third Edition, edited
bv James B. Pawley. Springer Science+Business Media. LLC. New York. 2006.
627
628
Chapter 35 • V. Centonze and J.B. Pawley
record 20 to 30 similar images of the adjacent planes to form a three-dimensional (3D) stack. Each of these 20 to 30 images will cause additional bleaching. In actual fact, what one should be interested in is not so much the bleaching rate but some sort of efficiency ratio of information recorded to illuminating dose. This ratio is set by the numerical aperture (NA) and transmission (T) of the optics and the quantum efficiency (QE) of the photodetector. Therefore, assuming equal optics, the confocal microscope is lOx to 100x more efficient than photography, because at these low light levels, the photomultiplier tube (PMT) photodetector that it uses is lOx to 100x more efficient than film (Chapter 12, this volume). The only photodetector more efficient than the PMT is the cooled charge-coupled device (CCD), and this detector can be used in disk-scanning confocal microscopes (Chapters lO, 12,21,25, and Appendix 3, this volume). While on the subject of bleaching, there are a few additional items that should be mentioned. It is generally assumed (1) that bleaching is proportional to the total dose of light where dose is the illumination level (in photons//lm2S ) x time, and (2) that, as absorption is usually low, the total number of photons passing through any horizontal plane is a constant. From these facts, some jump to the conclusion that bleaching occurs at the same rate in the focus plane as it does outside this plane. While this may be true for short distances above and below the focus plane and when using lenses of low NA, it is not true once the height of a stack of images becomes comparable with the width of the scanned area. Figure 28.3(B) in Chapter 28 shows an xz-image through such a bleach pattern made with an NA 1.4 in a specimen of fluorescent plastic. It confirms that there is a pyramid of bleached dye expanding above and below the plane of focus. As the illumination intensity at any level is proportional to the cross-sectional area of the light cone at that level, one can see that, with a large NA lens, the illumination level and, therefore, the bleach rate drops off quite quickly with distance away from the focus plane. The illumination level at these planes is not constant within the cone but is less strong near the edges. Figure 35.1 is a diagram to help visualize this process. We assume that the BFP of the objective is uniformly illuminated and, therefore, light is focused into the focused spot from all angles equally. As a result, the total number of photons passing through any particular level within the cone of the beam is a constant. The intensity itself decreases roughly with the square of the distance away from the focus plane but, as the beam scans over the raster, this effect is counterbalanced by the fact that most points in nonfocus planes will be illuminated for a longer time. Although points that are only illuminated by the outer rays of the cone when the beam is scanning near the edge of the raster will only be illuminated once per scan, those near the center of the raster will be illuminated for a time that is longer in direct proportion to the reduction in the illumination intensity at any instant. All of the points for which this is true for a particular focus position will lie inside a bleaching octahedron having the scanned area as its base and a defining angle equal to the acceptance half-angle of the objective (ex). Inside this octahedron the bleach rate is constant, while outside it the bleach rate decreases slowly until it becomes zero in the region where even the outer rays of a beam scanning the edge of the raster never reach. As the focus plane moves up or down in the specimen during a z-series, the total bleaching will be proportional to the superposition of the octahedra of the individual planes. The result is that points inside the octahedron associated with the central focus plane will receive a maximum dose while those farther from the center
c
z axis FIGU RE 35.1. Pattern of bleaching in the confocal microscope. A stationary, focused light beam will bleach a conical volume in a uniform fluorescent specimen. The severity of the bleaching at anyone level in the cone is inversely proportional to the area of the cone at that level and is most severe at the focus plane. However, as the beam is scanned over the focus plane, points in this plane are only illuminated a few times while points in adjacent planes will be illuminated for a longer time. As the beam is scanned over a line in the focus plane, the convergence angle of the beam (a) defines a triangle inside which reside points at which the lower intensity of the flux associated with not being in the focus plane is exactly compensated for by the increased amount of time they are illuminated [black diamond (A)]. As this line scans to create a 20 image, the triangle of equal total illumination becomes an octahedron (B). As the 20 focus plane is scanned in z to create a 30 image, the octahedrons add up to produce more damage near the center of the raster, where the planar constant bleach pattern is thickest (C). It should be noted that, because of the rather slow response of the mechanical scanning mirrors. a considerably higher level (lOx-20x) of bleaching will occur to either side ofthe imaged area of the focus plane where the horizontal motion of the mirrors slows down and reverses. In some confocal instruments, the beam is blanked during this retrace period.
will receive less. The reason for mentioning these simple consequences of geometrical optics here is to emphasize the difficulty of devising a computer program to correct for bleaching artifacts even if the bleach rate is assumed to be directly proportional to the energy deposited and the dye stays in one place. The best plan is to be careful not to waste photons and thereby minimize the amount of bleaching that occurs. Although the bleaching problem may be no worse in confocal microscopy than in normal fluorescence microscopy, it is still a very serious problem. In fact, it is not unlikely that, particularly when viewing living cells, one's success at using either method will depend on how successful one is at collecting and recording as great a fraction as possible of the fluorescent photons produced and at recording only the information that one really needs while using no more than the minimum necessary exposure. We explain how to do this in the sections that follow.
Tutorial on Practical Confocal Microscopy and Use of the Confocal Test Specimen • Chapter 35
629
FIGURE 35.2. Alignment of the laser in the back-focal plane (BFP) of the objective as seen using a 45° prism that allows one to view a circular target situated in the BFP. (A) Properly centered, (C) misaligned. The two lower images show the bundle of light rays projected into the specimen under each alignment condition. These images were made by photographing the fluorescent light emanating from a cube of uranium-doped glass placed in the specimen plane of a IOx/0.45 objective. The angles involved are somewhat smaller than expected because the glass has a very high index of refraction.
GETTING A GOOD CONFOCAL IMAGE To obtain an optimal confocal image, a fluorescent specimen must first be in focus under the right type of objective: that is, one that is designed to focus into a medium having the refractive index of your specimen, thereby avoiding problems with spherical and chromatic aberration (see Chapters 7 and 20, this volume). Because of the optical sectioning property on the confocal microscope, specimens that are far out of focus will produce no signal! If you are having trouble finding the specimen, open the pinhole all the way, focus on the specimen, then close it down again. Next, the scanhead must be properly aligned. When the beam scans over the center of the raster pattern, the light from the laser should be both on axis and fill the BFP of the objective, and the light passing up from the focused spot to the PMTs must be aligned to pass through the pinhole. Once alignment has been achieved, the pinhole size should be set for the best compromise between signal strength and xy- and z-resolution, (for details see Chapter 22, this volume) the PMTs should be adjusted to produce an appropriate signal level and the monitors should be adjusted so that the full range of signal can be viewed.
Proper alignment l of the scanhead first requires adjustment of the mirror(s) that direct(s) the laser light down the optical path of the microscope. A 45° angle prism (or even just a small piece of lens-cleaning tissue) can be positioned in place of an objective so that the laser light path can be viewed with respect to the BFP (Fig. 35.2). Although the angle at which it passes the BFP is proportional to the instantaneous position of the beam in the imaged plane, the laser beam should always pass through the center of the BFP and then be focused by the objective to form a symmetrical spot at the specimen plane [Fig. 35.2(A,B)]. Misalignment at the BFP may mean that the light does not fill the pupil of the objective, reducing its effective resolution and causing it to pass through the specimen at an angle [Fig. 35.2(C,D)]. Next, a standard fluorescent or reflective test specimen must be used to align the scanhead so that the central maximum of the
I
Many newer confocal microscopes come pre-aligned and there are no user controls for adjusting the alignment as described in this section. However, some of the tests used to check performance can still be used to see if you need a service visit.
630
Chapter 35 • V. Centonze and J.B. Pawley
diffraction-limited spot (Airy disk) in the specimen is focused through the pinhole where it can be detected by the PMT. When aligning for peak signal, it may be helpful to change the display to have a colorized lookup table (LUT). A colorized LUT emphasizes small changes in signal intensity because the eye is more sensitive to absolute hue than to absolute gray level. While scanning, adjust the scanhead mirrors or pinhole location to maximize the signal received by the PMT. For the time being, you need only be concerned with obtaining the brightest image, not one of maximum quality. It is advisable to make the mirror adjustments using the objective you plan to use for your final high-quality image. Because wedge errors in the dichroic beam-splitter can displace the apparent position of the pinhole (Chapter 9, this volume), older instruments will require re-alignment of the scanhead whenever the dichroic is changed. Once one has a reasonable image, it is probably worthwhile to make a note of the settings of all the user controls. These can later be used to monitor instrument (and operator!) performance and to serve as starting conditions after one has implemented major service or development changes. With basic alignment complete, you may find it interesting to view the Airy disk in the pinhole plane. On older scopes this can be accomplished by placing a small CCD camera in the intermediate image plane, but if your instrument uses a larger, iris-diaphragm aperture rather than a real pinhole, you may be able to see the light directly by viewing it scattered by a piece of paper placed in front of the pinhole. Figure 3S.3 shows the sort of image you can expect to see if you do this by placing a piece of paper in the second filter well of a Bio-Rad MRC series scope that has been set up to measure BSL from a plane mirror surface that is either in focus [Fig. 3S.3(A)] or just out of focus [Fig. 3S.3(B)]. The in-focus image shows some astigmatism, probably caused by imperfections in the beam-splitter. It is informative to see how the image of the Airy disk degrades if one reduces the zoom magnification, a process that reveals the deleterious effect of off-axis aberrations such as astigmatism, curvature of field, and coma (see also Chapter II, this volume). One can also use this setup to view the asymmetrical increase in Airy disk size (and decrease in resolution) caused by misaligning the illumination in the objective BFP (as in Fig. 3S.2). Actually, viewing the Airy disk helps to make clear the idea that each objective has an optimum pinhole size because it becomes evident that the light at the pinhole plane is simply a magnified view of the light in the image plane and that, therefore, at a fixed NA, the size of the image of the spot will be proportional to the magnification of the objective. A pinhole aperture equal to the diameter of the first minimum in the Airy disk (i.e., 1 Airy unit) will pass about 80% of the in-focus signal and still have a bit better xy-resolution than a widefield microscope. Once this benchmark detector aperture is known for a given optical setup (wavelength, NA, magnification), it is possible to make informed choices about the most appropriate aperture for any other objective. Specimen preparation, objective lens, pinhole size and alignment, and focus plane all affect the amount of light collected from the specimen. Adjustment of the display monitor and the PMT affect the image viewed. The black level of the PMT amplifier must be set so the full range of signal can be detected. This can be done by scanning while light is blocked from entering the PMT and adjusting the black level until the rastered area becomes just visible compared to the unscanned part of the monitor screen. 2
2
When the black-level control is set correctly, black areas will register as 2 to 3 ADU levels out of 255.
A
8 FIGURE 35.3. The Airy disk reflected back from a mirror test specimen can be viewed directly at the pinhole plane of a Bio-Rad MRC confocal microscope by removing the second filter block and placing a piece of paper into the bottom of the well. (Turn the PMT all the way down before doing this!) (A) In focus, but showing slight astigmatism, probably because of distortion by the dichroic; (B) slightly out of focus. Random points of light well away from the optical axis defined by the Airy disk represent stray light. This is less evident in (A) because of the higher relative brightness of the central spot and because the beam dump in first filter block had been improved by the addition of a piece of black velvet.
Note that if the black level is set too high, some intensity values may fall below the voltage corresponding to the zero value of the analog-to-digital converter (ADC) and so not be recorded. If, after setting the black level in this way, the image on the display has insufficient brightness, this can usually be best adjusted by modifying the display LUT in the computer. The gain of the detector circuitry (i.e., how much it amplifies) is changed by adjusting the accelerating voltage of the PMT. In round figures, a SO V increase in the accelerating voltage doubles the gain. Practically speaking, the gain controls both the brightness and the contrast of the image. When it is properly adjusted, the pixels in the brightest areas of the image should be bright white in order to utilize the full gray-scale range of the data handling system. However, one must also be careful not to saturate the ADC and this may be more difficult to avoid than expected because, in fluorescence confocal microscopy, the brightest signal may represent only 5 pixels across. These noise features can and should be eliminated by deconvolving the data, preferably in 3D. As an aside, because the confocal PSF can be approximated as a 3D Gaussian blob, 3D Gaussian filtering has almost the same
636
Chapter 35 • V. Centonze and J.B. Pawley
• The amount of light passing through the specimen is not affected by the zoom setting. • At high zoom settings, the confocal microscope can easily illuminate the specimen with an intensity (and a bleaching rate!) that is 100x greater than in a normal fluorescence microscope using a Hg arc source. • Fortunately, because the PMT detector is more sensitive than film, less illumination is actually required to record a usable confocal image. • Set the zoom to produce a pixel size satisfying the sampling criterion for the lens in use. • Don 't forget to deconvolve (or 3D Gaussian filter) your data before viewing it.
USING A TEST SPECIMEN
Why Use a Test Specimen?
FIGURE 35.6. The relationship between resolution and the number of voxels needed to Nyquyist-sample the point-spread function. According to standard light microscopy theory, the Abbe resolution , 0, is defined as the radius of the first dark ring of the Airy disk. [n confocal microscopy, this defines a blob with a diameter of 20. If we assume that a pixel is 0/2, then the 20 image of a point object will put signal into at least 12 to 16 bright pixels in the focus plane. The 3D image of a point requires collecting signal from at least four planes, separated by a distance equal to one half of the z-resolution. As a consequence of these factors, the 3D image of a point obj ect requires measuring signal in 50 to 100 voxels. Averaging signal over these voxels (e.g., by deconvolution) does not reduce spatial resolution but does greatly improve the SIN of the resulting data . (See Chapter 25, this volume.)
effect as a full-scale 3D deconvolution, as least in tenns of suppressing noise features. Filtering the data in this way has another benefit: it effectively averages the voxel intensity data over the number of voxels containing significant counts in the Gaussian blob that represents the PSF. If we estimate that the blob is -5 pixels wide in x, y, and z (see Fig. 35.6), then it will have significant counts in 5 x 5 x 5 = 125 voxels. Therefore, 3D deconvolution will effectively average out Poisson noise over -125 voxels : a significant factor (Chapters 19 and 25, this volume)!
The tendency to use tried-and-true specimens when testing or learning how to operate the confocal microscope is so marked that it would be fruitless to recommend against it. However, it does have some disadvantages, and the chief of these is that one probably really does not know what a good confocal image of such a specimen would look like. This is because any stained biological specimen is probably too complex to understand so well that one can predict what a 3D image of it should look like and then adjust the instrument until this result is attained. Although we may have prepared hundreds of similar specimens and may recognize some as successes and others as failures, we seldom really know in a quantitative way either the exact size of any of the specific features that they contain, how much they have been stained, or how much this stain may have faded or bleached (Chapters 16 and 39, this volume). We may believe that such a specimen possesses a particular 3D structure when, if fact, this structure has been lost when it was inadvertently flattened during specimen preparation (Chapter 18, this volume). In short, such a specimen is not a test specimen, and one cannot really use it to measure either your own skill or the performance of the microscope.
Description of the Test Specimen Microscope manufacturers have long produced a variety of test specimens for measuring the performance of their instruments. The most common of these is the stage micrometer, which usually consists of a graticJe etched into a metal film on the lower side of a coverslip. Such patterns can be very useful for checking magnification but, because they are made using light optics, the finest spacings that can be produced on them are -1 to 211m, and this is not fine enough to really test the ultimate resolution of a good optical microscope. To fill this void, we followed the lead of Oldenbourg and Inoue (Chapter I, this volume) and designed a test specimen that we then fabricated at the National Nanofabrication Facility at Cornell University using electron-beam lithography (Pawley et al., 1993).6 The patterns were etched into a 50nm Al film on the lower side of a 1 x I cm piece of #1.5 coverslip and
Pixel Size Summary To preserve all of the information in your data, you must use a pixel (or, in three dimensions, a voxel) size at least 2.3x smaller than the Abbe resolution limit of your optical system in (x, y, and z! i). This means that, for each objective, there is an optimal setting for both the zoom control and the interplane spacing of 3D data sets.
6
Similar test specimens may soon be available again from Louie Kerr, at the Marine Biological Laboratory, Woods Hole, MA. Other test specimens are discussed in Chapter 36.
Tutorial on Practical Confocal Microscopy and Use of the Confocal Test Specimen • Chapter 35
637
FIGURE 35.7. Scanning electron micrographs of a confocal test specimen fabricated at the National Nanofabrication Facility at Cornell University using electronbeam lithography. The pattern is etched into a 50nm Al film on the lower side of a I x 1 cm piece of # 1.5 coverslip and laid out as a square about 500/lm on a side (left). There is an L-shaped, resolution test pattern, 40/lm along each arm, at each corner with one lOx 10/lm square of l/lm period at the corner and six 5 x 10/lm rectangles of 0.7, 0.5, 0.35, 0.25, 0.20 and 0.17/lm period along each arm (right).
were laid out as a square about 500!lm on a side [Fig. 35.7(A)] . There is an L-shaped resolution test pattern, 40!lm along each arm, at each corner with one lOx lO!lm square of l!lm period at the corner and six 5 x lO!lm rectangles of 0.7,0.5,0.35,0.25,0.20, and 0.17!lm period along each arm [Fig. 35.7(B)]. The periods of the spacings are listed next to the pattern in the top right corner when the coverslip is mounted, AI-side down, so that the writing is legible. The procedure for measuring the optical transfer function of your microscope from images of such patterns is explained at the end of Chapter 1. Such a test specimen can be used to record images using transmitted light, reflected light, or fluorescence. For the latter, the coverslip is mounted over a well into which dye-laced immersion oil has been introduced. As the test pattern is predominantly opaque, light can only reach the dye through the clear areas of the pattern [Fig. 35.8(A,B)]. Although the dye itself is essentially infinite in thickness, as long as the NA is reasonably high, most of the excitation is confined to a thin triangular region just under a clear space in the pattern and next to the coverslip [Fig. 35.8(C)]. Although some excitation light does reach the dye through adjacent clear lines in the pattern, this is only a minor effect as can be seen in the xy- and xz-images of such a specimen, shown in the upper half of Figure 3S.8(A,B). Because the dye is a liquid, any bleaching that may occur is masked by the diffusion of new dye into the imaged area.
by the Nyquist criterion (i.e., 4 to 6 samples/resolution element). This is because the Nyquist analysis is based on information theory and, according to information theory, a periodic object contains only two items of information: the frequency and the phase. Although these can be determined from data sampled at the Nyquist rate, the image can look very poor. Figure 35.9 shows how,
... .
··1"111
'1
tIll",
r'
'I"
I
l'Ill! 1
t
I
. A
Using the Test Specimen To image the test specimen in reflected light, treat it as you would any other. However, you may find it difficult to adjust the plane of focus to exactly coincide with the metal film. This is because, as is mentioned in Chapters 14 and 15, the position of maximum signal can be determined to a small fraction of the z-resolution, and in the case of reflected light, the stage-motion control on your instrument may not have sufficient precision to allow you to focus exactly on the metal surface. Because you are attempting to view a periodic object, the zoom setting should be set about 2x higher than that normally required
FIGURE 35.8. (A) Diagram of the fluorescent test object. Because it is difficult to fabricate a planar fluorescent test object that does not bleach, this device works by placing a pool of fluorescent immersion oil just below the etched metal pattern on the bottom of the coverslip. As long as the BFP of the objective is filled uniformly, much of the illumination will approach the specimen at relatively large angles, selectively exciting a triangular prism of dye beneath each etched line. (B) As the focus plane moves into the dye, some of the dye deeper in the specimen is excited by rays that pass through adjacent etched lines. However, this is only a minor effect that makes the z-response somewhat asymmetrical.
638
Chapter 35 • V. Centonze and J.B. Pawley
---
--~
~
----
---------
-
-
--~
-
FIGURE 35.9. Images of the te~t specimen made using fluorescence (upper pair) or backscatteredlrellected light (lower pair). The images were collected simultaneously and show mirror symmetry because the fluorescent light emerges from those areas not covered with the metal film that provides the reflected signal. The upper member of each pair shows an xy-image and the lower member shows an xz-image that makes more evident the asymmetry of the fluorescent signal. Nikon 60x/1.4 on a Bio-Rad 60010ptiphot. The vertical ~pacings have the following spacings left to right: 0.35, 0.5, and O.7/lm (vertical) and 1.0/lm (hori zontal).
when sampling at twice the frequency of the analog data but doing so just as the signal crosses zero, it is possible to miss the variations in a periodic signal entirely. As a result, one should choose 0 .05 J.1.m pixels or about 35 J.1.m field width for a 768 pixel line for use with an NA 1.3 to 1.4 objective. If you have done things correctly, the picture that you see should look like that in Figure 35.10(A). In these images, it can be seen that the edges of the vertical lines have a high-frequency wiggle (period about 4-5 raster lines = 16.6-20ms), indicating the presence of mains-frequency electronic, magnetic, or mechanical interference. In Figure 35.1 O(B), the sets of horizontal lines above and below the axis of the computer-calculated Fourier transform (FT) below each image clearly demonstrate the presence of some other periodic instability present in these single-scan images. As the vertical spacings of the two finer periods are near the cut-off frequency of the contrast-transfer function, only a single spot is shown along the horizontal axis on either side of the vertical axis, but the spacing of all the first-order spots can be seen to be inversely related to the actual spacing of the features in each segment of the image. The horizontal bars in the FTs to the left of Figure 35.l0(C,D) show the effect of Kalman averaging. Not only do the spots corresponding to the fine serrations of the edge get averaged
out, but the SIN of the FT is markedly higher. The FT in the lower right of Figure 35.1 O(E) was made from a larger patch of the spec~ imen having a 0.8 J.1.m spacing. Three harmonics can be seen easily, but the fourth , representing 0.8/4 = 0.2 f.!m data, is barely visible. If you have the equipment for simultaneous fluorescentlBSL imaging, the specimen can be imaged in fluorescence light simply by switching to the other channel. Otherwise the microscope must first be reconfigured for fluorescent light imaging. The image that you see should look like that in Figure 35.11.
The Diatom: A Natural 3D Test Specimen The test specimens discussed have the disadvantage of being planar. Making a test specimen having precise structural features in the third dimension is more difficult. The best solution so far is to immerse diatom frustules (North Carolina Biological) in fluorescent oil and view them as negative objects, as suggested by Roger Tsien (University of California, San Diego). Figure 35.12 shows a stereo view of such a preparation, imaged with BSL using the system mentioned above. Although careful attention to the exact species of diatom can result in highly reproducible spacings, this specimen is not without its problems.
"Lucky" Nyquist
-B
2x oversampled
c o FIGURE 35.10. Data sampling of periodic objects. Although the Nyquist criterion states that only -2 samples are required for each period of the highest spati al frequency in the data, this does not always work for periodic objects. Although a fOltuitous phase relationship between the sampling instant and the positive and negative peaks of the highest frequency can lead to an accurate digitization (A), a phase relationship that sampled the signal only at the instant that the signal crossed the axis would not record this contrast at all (8). For this reason, it is safer to digitize images of periodic objects with at least four samples/period (C).
E FIGURE 35.11. (A) High resolution refl ected light image of test specimen recorded using a Nikon 60xll.4 lens o n a 8io-Rad 600/0ptiphot. Spacing, left to right: 0.5,0.35, and 0.25 ~m. In (8 ), the sets of horizontal lines above and below the axis of the computer-calculated Ff below each image clearly demo nstrate the presence of a periodic instability in these si ngle-scan images . As the two finer spac in gs are near the cut-ofl frequ ency of the contrast- transfer fun ction , on ly a single spot is shown along the horizontal axis on either side of the vertical axis, but the spacing of all the first-order spots can be seen to be inversely related to the actual spacing of the features in each segment of the image. Panels (C, DJ are reflected light images of 0.7 ~m hori zontal spacings made either live (Cl or after averaging 20 scans (D). The horizontal bars in the Ff plots to the left of each image show the eflect of averaging. Not on ly do the spots corresponding to the fine serrations of the edge get averaged out, but the SIN of the lower Ff is marked ly higher. The Ff in the lower ri ght (El was made from an image of a larger patch of the specimen having an 0.8~m spacing. Three harmonics can be seen easily, but the fourth, representing 0.8/4 = 0.2~m data, is barely visible.
640
Chapter 35 • V. Centonze and J.B. Pawley
FIGURE 35.12. Stereo image of part of a diatom immersed in immersion oil made with BSL using a Nikon 60xl1.4 lens on a BioRad 60010ptiphot equipped with the optimized BSL system described in the text and diagrammed in Figure 2.3. The diatom is resting on a microscope slide and is viewed from the top through the coverslip and two layers of immersion oil. The signal reflected from the slide surface can be seen around the edge of the image but the presence of the diatom distorts the focus plane away from the surface in the center of the field of view. (Specimen kindly provided by Nelson Navaro, CU.P.R., Puerto Rico.)
Because the amorphous silica out of which the frustule is made has an index of refraction different from that of either water or immersion oil, it refracts and also scatters considerable light. While this is convenient for BSL imaging, it produces aberrations whenever the focus plane penetrates too far. This effect can be seen in Figure 35.12 where, around the edge of the image, one can see the signal reflected back from the flat surface of the glass slide on which the diatom is resting. However, in the center of the field the beam must pass the through the frustule to reach the surface and its presence distorts the focus plane away from the surface of the glass making it appear dark.
REASONS FOR POOR PERFORMANCE
Sampling Problems As mentioned above, when imaging periodic objects, it is necessary to sample the image data at higher than the Nyquist rate. Failure to do so will result in the aliasing artifacts shown in Figure 35.13.
Optical Problems Aberrations As is mentioned above and covered in Chapters 1, 7, 8, and 20, one can only obtain diffraction-limited resolution if all optical aberrations are absent. The major aberrations are spherical aberration and chromatic aberration. In a high NA objective, spherical aberration is usually corrected for a number of wavelengths but only for one immersion medium. To get the recommended performance, you must use the immersion liquid for which the lens was designed or use a lens having a correction collar to adjust for different media. As the focal length of any objective changes slightly with wavelength, its magnification also changes with wavelength, only in the opposite sense. As a result, when operating in fluorescence, an off-axis ray at the excitation wavelength will cross the intermediate image plane at a different point from the longer wavelength emission ray coming from the same point on the specimen. As a result, the emission ray may partially or totally miss the pinhole, resulting in loss of signal (Fig. 35.14)! Two other points are worth mentioning in regard to this second point:
FIGURE 35.13. The effect of under-sampling on a periodic image. The upper image was recorded using reflected light with a Nikon 60x/1.4 lens on a Bio-Rad 60010ptiphot with 0.041.un/pixel (twice the Nyquist rate for 0.2J..lm structures) and then printed on a dye sublimation printer having pixels the equivalent of 0.018 J..lm in size. In it, both the 0.25 J..lm (left) and 0.2 J..lm (right) spacings are clearly seen. The lower image was made with identical optical conditions but the pixel size was 0.08J..lm (exactly the Nyquist limit) and the image was recorded photographically from a video monitor having a screen pitch equivalent to 0.05 J..lmlelements. To some extent, the reduced contrast in the 0.25 J..lm spacing may be due to the bandwidth limitations of the monitor or the vagaries of the photographic process, but the virtual absence of the 0.2 J..lm spacings is probably due to sampling problems. as is evidenced by the presence of a faint aliasing pattern at about twice the period of the actual features in the image.
Tutorial on Practical Confocal Microscopy and Use of the Confocal Test Specimen • Chapter 35
Return ing ray at longer wavele~~~~ ....
:::.1
641
A
//1 .'
Illuminating ray from laser
.' .' Pinhole in intermediate ~ image plane, with beam scanned off axis
B Focus plane
FIGURE 35.14. Lateral chromatic aberration. The effect of chromatic magnification error on signal level in the confocal microscope. The focal length and, hence, the magnification of any optical system varies with wavelength. Consequently, an off-axis ray of short wavelength light will not follow the same path as that of a ray of longer wavelength light originating from the same point. In the laser-scanning confocal microscope, the scanning mirrors are supposed to deflect both the source and the detector pinhole off-axis by the same amount. However, this will not happen if the optical system has high chromatic aberration and the system is used for fluorescence, because the exciting and emitting wavelengths are magnified by different amounts. As a result, the signal light will miss the pinhole and the image will become darker away from the axis. The problem can be reduced only by using optical systems that are highly corrected for chromatic aberration and by placing the field-of-view as near to the optical axis as possible.
250r-----:jllljiiC::&:"'-----,
c
200 • In many older microscopes, it was common to correct chromatic aberration in the eyepiece. Objectives from such microscopes will demonstrate totally unacceptable levels of chromatic aberration if they are not used in conjunction with the appropriate correcting eyepieces. Generally one should use only modern, highly-corrected objectives for fluorescence confocal microscopy and use them only with other components that are properly matched. • The magnitude of the displacement caused by the chromatic magnification error is generally proportional to the distance that the point in the image is away from the optical axis. Therefore, the effect will be less severe if imaging is restricted to an area near the optical axis.
Curvature of Field Unless concrete steps are taken to prevent it, simple lenses will focus a plane surface onto a segment of a sphere. Objectives designated "plan" are supposed to have flatter fields of view than most, but this feature is usually incorporated at the cost of additional elements and often lower light transmission. Figure 35.15 shows two images of a plane mirror, one made with a non-plan lens (upper) and one with a planapo-correction (lower). The images have been posterized to emphasize the variation in signal caused by field curvature. The variation in signal strength across the field is shown with greater precision in the x-profile plots at the bottom. We should emphasize that these plots show variation in the measured intensity of reflected light passing the confocal pinhole, not the actual geometrical curvature of the field! Although the upper image displays a variation in focus that could degrade the image of a flat test object such as that mentioned above, one
.e- 150 0;; c: .!! c: 100
-
50 o~--------
________________
~
FIGURE 35.15. Curvature of field. In general, lenses focus a plane surface onto a spherical one. In objectives not specifically corrected for plan operation, the curvature of field has the effect that an image of a plane surface loses intensity off-axis. This figure shows two images of a plane mirror, one made with an uncorrected lens (A) and one with a planapo correction (B). The variation in signal strength across the field is shown with greater precision in the two xprofile plot~ (C). (We should emphasize that these plots show variation in the measured intensity of reflected light passing the confocal pinhole, and are not a measure of the actual geometrical curvature of the field!) Although field curvature can be a serious problem when viewing planar specimens, it is often less serious when making 3D fluorescent images of biological specimens because in the latter case there is no signal loss but simply a slight dish-shaped distortion in the final 3D data. However, if this distortion is to be kept symmetrical, it is important not to pan the scanned area away from the center of the field of view when using such a lens.
should remember that, with a z-resolution of -0.6 ~m, even an intensity drop of 90% between the center and the edge of the field would only indicate a field curvature of 0 and S2i.coloc = 0 if Sl; = O. Each numerator represents colocalized pixels - the sum of intensities from pixels within one channel that also contain some intensity component from the other channel. The denominators constitute the summed intensities of all pixels, above threshold, in the channel. Some software applications use upper- and lowercase to distinguish colocalization coefficients calculated from the global image from those calculated from a thresholded image or a region selected from the fluorogram.
Setting Thresholds Quantitation in colocalization studies is usually dependent upon the subjective application of thresholds. Establishing the threshold to eliminate noise or background is sometimes carried out by subtracting the threshold from the image, or by setting limits in the 2D histogram, which many find to be more intuitive. One may have to test multiple thresholds before finding the levels that are supported by correlative methods and the controls. When software recommends threshold settings (or setting background levels), it is important to know how those settings are being calculated. A background derived from the darkest pixels within the image more correctly represents the noise floor of the image rather than the amount of background from bleed-through, autofluorescence, or non-specific labeling. A new approach has been developed that estimates thresholds for both channels simultaneously over a range of decreasing thresholds until the probability of correlation equals zero (Costes et aI., 2004). This approach promises to be less subjective, but can be still skewed by an imbalance in average channel intensities.
Spatial Deconvolution in Colocalization Studies Noise and background in colocalization can be reduced by employing restorative deconvolution (Van Steensel et at., 1996; Landemann, 2002; Landmann and Marbet, 2004). Even analysis intended for 2D images can benefit from deconvolution if the data is collected as a shallow z-series and the single 2D image selected following deconvolution of the z-series. Alternatively, 2D confocal images can be Gaussian filtered if sampled adequately. In the example shown in Figure 36.19, a field of peripheral nerves in rat tooth pulp is shown as a single optical slice from a z-stack collected at low resolution [Fig. 36.19(A)], with a small region of the sample [small box in panel (A)] shown at higher optical resolution [Fig. 36.19(B)l The arrowhead points to the same fiber in both panels. Panel (B) indicates that there are actually two populations of closely intertwining nerve fibers, larger fibers predominantly labeled red (arrowhead), and the smaller, somewhat beaded fibers labeled green (arrows). Closer examina-
Practical Confocal Microscopy • Chapter 36
FIGURE 36.19. Colocalization. Nerve fibers labeled for peripherin (red) and CGRP (green) in tooth pulp from rats 3 weeks of age. Initial observations with a 20 x 0.75 NA Fluor lens (Nikon) (A) suggest a population of nerves coimmunostained for both proteins, appearing yellow. A 60 x 1.4 NA PlanApo lens (Nikon) with zoom = 2 (B) reveals two distinct populations of fibers: a thicker population labeled with peripherin (arrowhead), and smaller, beaded fibers, labeled with CGRP (small arrows). Both images are single planes from z-stacks. High magnification images were undersampled 60% laterally (150 nm vs. 83 nm) for this sample. Image (B) has been deconvolved by a maximum likelihood estimate algorithm (MLE). Scale bars represent 251lm (A) and IOllm (B).
tion shows that the two labels may be co-existing (as noted by yellow) only where the fibers cross, or in some of the thickened regions along the small green fibers. Given that the z-resolution of the confocal microscope is significantly less than the xy-resolution, such overlaps may simply be due to the diminished z-resolution. The 2D histograms in Figure 36.20 are derived from the highresolution volume shown in Figure 36.l9(B), and graphically demonstrate the impact on co localization of background removal through deconvolution. The 2D histogram of the original (nondeconvolved) volume [Fig. 36.20(A») displays two putative correlations, suggesting that each fiber is predominantly labeled for one protein, but containing the other protein in reduced amount. Pixels plotting close to the x-axis represent low intensity green label, such as the line of pixels very close to the x-axis (below the threshold line) indicating low-intensity background fluorescence in the green channel that associates with all levels of red fluorescence. Pixels plotting near the y-axis represent low intensity red label. The 2D histogram changes dramatically after restorative deconvolution by maximum likelihood estimation (MLE, using Huygens, Scientific Volume Imaging) of the z-stack
669
FIGURE 36.20. Fluorogram analysis of colocalization. 2D histograms of the volumes shown in Figure 36.19. Green intensities are plotted on the y-axis (0-255), and red intensities are plotted on the x-axis (0-255). The fluorogram before deconvolution (A) shows low intensity background in the green channel indicated by a layer of points lying above the x-axis and two correlations of unequal intensity possibly existing between the two channels. After restorative deconvolution the 2D histogram (B) has less background and shows the effect of background removal on putative colocalization. While CA) suggests a possible weak correlation in a bimodal distribution, (B) suggests a lack of colocalization. The white lines in both plots represent threshold levels applied to both channels for subsequent analysis on the volume (Table 36.3), and shown by white masks in panel (C) and (D). Panel (C) is the same field as that shown in Figure 36.19(B), but taken from the volume before deconvolution, and panel (D) is the same field after deconvolution. The scale bar represents 10 11m.
[Fig. 36.20(B»). The low-intensity background from the green channel has been removed and the two regions of correlation observed in Figure 36.20(A) have been sharply reduced, suggesting that they were due to blur, as does the fact that most pixels are now shifted towards the x- or y-axis. The impact of MLE on the correlation coefficients is shown in Table 36.3. The red and green labels show very high correlations in the original, undeconvolved low resolution volume [Fig. 36.19(A»), with values of 0.817 and 0.896 for Rand T, respectively, in Table 36.3, Line 1. The colocalization coefficients MI and M2 indicate that all voxels are colocalized because 100% of voxels in each channel are contributing to the colocalization. The higher
670
Chapter 36 • A.R. Hibbs et al.
TABLE 36.3. Effect of MLE and Threshold on Colocalization Sample 1 2 3 4 5
Fig. Fig. Fig. Fig. Fig.
36.19(A) 36.19(B) 36.19(8) 36.19(8) 36.19(8)
MLE
Threshold
R
r
MI
M2
No No No Yes Yes
No No 37/37 No
0.817 0.624 0.457 0.345 0.298
0.896 0.803 0.461 0.475 0.303
1.00 1.00 0.496 0.807 0.254
1.00 1.00 0.789 0.833 0.556
20120
resolution volume [Fig. 36.19(B)] displays some reduction in both Pearson's coefficient and overlap coefficient (Table 36.3, Line 2) as the two populations of fibers are now resolved into separate objects. However, the colocalization coefficients for both channels are still at unity suggesting either perfect correlation or, more likely, significant levels of background. MLE processing also changes the colocalization values, as shown in Table 36.3, Line 4. The amount of correlation from Pearson's coefficient drops to 34.5% and the overlap coefficient is reduced to 47.5%. The proportion of contribution to colocalization by the two channels (Ml and M2) are reduced to 80.7% for the red channel (Ml) and 83.3% for the green channel (M2). The threshold is important for determining the level of fluorescence that is to be considered significant. Thresholds were set for both channels by measurements of presumed background from the darker regions between the labeled fibers. These thresholds are represented by the white lines in Figure 36.20(A), at a threshold of 37 for the original image stack and at a threshold of 20 in Figure 36.20(B), the fluorogram of the volume after MLE. Table 36.3 provides the corresponding coefficients for the volumes following application of thresholds. The field in Figure 36.20(C) is the same field as in Figure 36.19(B), but taken from the high resolution volume before deconvolution. The white mask represents the pixels above threshold, as applied to this image plane. Figure 36.19(0) is the same field as Figure 36.19(B), taken from the volume after deconvolution. Pixels above the threshold applied to this image after deconvolution arc represented by the white mask. The differences between Figure 36.20(C) and Figure 36.20(0) not only include the difference in number of colocalized pixels, but extend to the apparent thicknesses of the nerve fibers and background around the fibers. Clearly, the apparent colocalization in the low magnification image and the high-resolution image was due to blending of the two channels by blur, noise and bleed-through. Restorative deconvolution reduced these contributions in a far more meaningful manner than was possible with simple thresholding. The remaining colocalization in this example is likely due to bleed-through, blur at the intersections of fibers, as well as blur from fibers in adjacent volume planes (not shown). Using deconvolution, direct inspection of the image, and the 20 histogram in conjunction with the quantitative analysis tools available for colocalization can provide far greater insight into the relationships between the two labels in a volume than can any single tool alone. While we may be left with questions regarding the exact nature and extent of colocalization in this example, these tools can be used together to indicate directions that might be taken to determine the relationship between these two proteins more closely. Careful use of adequate controls to establish acquisition parameters that avoid bleed-through, noise, blur and autofluorescence, as described in detail by Costes et at. 2004, is essential for obtaining sound datasets for colocalization.
DISCUSSION Recent advances spanning only the past 20 to 30 years have fueled extremely rapid development and popular adoption of laserscanning microscopy as a practical and uniquely powerful tool for scientific discovery. The capabilities of these instruments have two faces - on one hand, they can gather light with unprecedented sensitivity, lack of noise, and resolution so that we can work at the limits of physics on biological questions that were unapproachable in an earlier era. On the other hand, they can also generate exquisitely stunning images of artifact or elevate the most mundane source of background into signals that overwhelm all meaningful information from the sample. Adequate controls are an absolute requirement to understand the interplay of the instrument controls and your samples. The regular application of standard confidence tests applied to both the instrument and the biological samples serve to demonstrate the outstanding capabilities of a well-maintained imaging system as well as providing meaningful information on the limitations of such equipment. To ensure maximum productivity, it is advantageous to monitor the progress of declining performance so that the system may be restored to peak performance before the potential for grievous artifact or frank system failure becomes a reality. Murphy's Law dictates that there is a disproportionate chance such problems will be discovered at the onset of an important experiment or when results are needed in the face of looming deadlines. Quantitative measures of performance may serve to aid remote diagnosis by a field engineer and expedite subsequent ordering of the necessary parts. It is hoped that the tests outlined in this chapter will aid you in realizing the goals of data integrity, peak performance, and high instrument reliability when using laser-scanning confocal microscopes in a variety of environments.
ACKNOWLEDGMENTS The images used to illustrate colocalization (Fig. 36.19 and Fig. 36.20) were generously provided by Dr. Orapin Veerayutthwilai (DDS, MS), Department of Oral Biology, and Dr. Margie Byers (PhO), Department of Anesthesiology, University of Washington, Seattle, WA. The image of a subresolution fluorescent bead in Figure 36.6 was kindly provided by Michael Weiss, Pacific AgriFood Research Centre, Summerland, BC, Canada.
REFERENCES Agnati, L.F., Fuxe, K., Torvinen, M., Genedani, S., Franco, R. , Watson, S., Nussdorfer, G.G., Leo, G., Guidolin, D., 2005, New methods to evaluate colocalization of fluorophores in immunocytochemical preparations as exemplified by a study on A'A and D, receptors in Chinese hamster ovary cells, 1. Histochem. Cytochem. 53:941-953. Akinyemi, 0., Boyde, A., 8rowne, M.A., Hadravsky, M., and Petran, M., 1992, Chromatism and confocality in confocal microscopes, Scanning 14:136-143. 800th, M.J., and Wilson, T., 2001, Refractive-index-mismatch induced aberrations in single-photon and two-photon microscopy and the use of aberration correction, 1. Biomed. Opt. 6:266-272. Brakenhoff. G.J. Wurpel, G.W.H., Jalink, K., Oomen, L., 8rocks, L., and Zwier, 1.M. , 2005, Characterization of sectioning fluorescence microscopy with thin uniform fluorescent layers: Sectioned imaging property or SIPcharts. 1. Microsc. 219: 122-132. Browne. M.A., Akinyemi, 0., and Boyde, A., 1992, Confocal surface profiling utilizing chromatic aberration, Scanning 14: 145-153.
Practical Confocal Microscopy • Chapter 36
Carlsson, K., 1991, The influence of specimen refractive index, detector signal integration, and non-uniform scan speed on the imaging properties in confocal microscopy, J. Microsc. 163:167-178. Costes, S.v., Daelemans, D., Cho, E.H., Dobbin, Z., and Pavlakis, G., 2004, Automatic and quantitative measurement of protein-protein colocalization in live cells, Biophys. J. 86:3993-4003. Cox, G., 1999, Measurement in the confocal microscope, Methods Mol. Bioi. 122:357-371. Cox, G., and Sheppard, C.J.R., 2004, Practical limits of resolution in confocal and nonlinear microscopy, Microsc. Res. Tech. 63:18-22. Demandolx, D., and Davoust, J., 1995, Multicolor analysis in confocal immunofluorescence microscopy, J. Trace Microprobe Tech. 13:217-225. Demandolx, D., and Davoust, J., 1997, Multicolour analysis and local image correlation in confocal microscopy, J. Microsc. 185:21-36. Dickinson, M.E., Bearman, G., Tille, S., Lansford, R., and Fraser, S.E., 2001, Multi-spectral imaging and linear unmixing add a whole new dimension to laser scanning fluorescence microscopy, Biotechniques 3 I: 1272-1278. Garini, Y, Gil, A., Bar-Am, I.. Cabib, D., and Katzir, N., 1999, Signal to noise analysis of multiple color fluorescence imaging microscopy, Cytometry 35:214-226. Hibbs. A.R., 2004, Confocal Microscopy for Biologists, Kluwer Academic/Plenum Publishers, New York. Hiraoka, Y., Sedat, J.W., and Agard, D.A., 1990, Determination of threedimensional imaging properties of a light microscope system, Biophys. J. 57:325-333. Ho, Roo and Shao, Z., 1991, Axial resolution of confocal microscopes revisited. Optik 88:147-154. Huth, U., Wieschollck., A, Garini, Y, Schubert, R., and Peschka-Siiss, R., 2004, Fourier transformed spectral bio-imaging for studying the intracellular fate of liposomes, Cytometry 57 A: 10-21. Hutter, H., 2004, Five-colour in vivo imaging of neurons in Caenorhabditis eiegans, J. Microsc. 215:213-218. LaMorte. v.J., Zoumi, A., and Tromberg, B.J., 2003, Spectroscopic approach for monitoring two-photon excited fluorescence resonance energy transfer from homodimers at the subcellular level, J. Biomed. Opl. 8:357-361. Landmann, L., Deconvolution improves co localization analysis of multiple fluorochromes in 3D data sets more than filtering techniques, 2002, J. Microsc. 208: 134-147. Landmann, L., and Marbet, P., 2004, Colocalization analysis yields superior results after image restoration, Microsc. Res. Tech. 64:103-112. Lansford, R., Bearman, G., and Fraser, S.E., 200 I, Resolution of multiple green fluorescent protein color variants and dyes using two-photon microscopy and imaging spectroscopy, J. Biomed. Opt. 6:311-318. Maly, M., and Boyde, A., 1994, Real-time stereoscopic confocal reflection confocal microscopy using objective lenses with linear longitudinal chromatic dispersion. Scanning 16: 187-193. Manders, E.M.M .. Stap, J., Brakenhoff, G.J., van Driel, R., and Aten, J.A., 1992, Dynamics of three-dimensional replication patterns during the S-phase, analysed by double labeling of DNA and confocal microscopy, J. Cell Sci. 103:857-862. Manders, E.M.M., Verbeek, FJ., and Aten, 1.A., 1993, Measurement of colocalization of objects in dual-colour confocal images, J. Microsc. 169:375-382. Model, M., and Burkhardt, 1., 2001, A standard for calibration and shading correction of a fluorescence microscope, Cytometry 44:309-316.
671
Neher, R., and Neher, E., 2004, Optimizing imaging parameters for the separation of multiple labels in a fluorescence image, J. Microsc. 213: 46-62. Pawley, 1.B., 1994, Sources of noise in three-dimensional microscopical data sets, In: Three-Dimensional Conj;,cal Microscopy: Volume Investigation of Biological Specimens (J.K. Stevens, L.R. Mills, and J.E. Trogadis, eds.), Academic Press, London, pp. 47-93. Pawley, J.B., 2000, The 39 steps: A cautionary tale of quantitative 3-D fluorescence microscopy, Biotechniques 28:884-886, 888. Pawley, J.B., 2002, Limitations on optical sectioning in live-cell confocal microscopy, Scanning 24:241-246. Scalettar, B.A., Swedlow, J.R., Sedat, J.W., and Agard, D.A., 1996, Dispersion, aberration and deconvolution in multi-wavelength fluorescence images, J. Microsc. 182:50-60. Shaw, P.J., and Rawlins, D.J., 1991, The point-spread function of a confocal microscope: Its measurement and use in deconvolution of 3-D data, J. Microsc. 163:151-165. Smallcombe, A., 2001, MuIticolor imaging: The important question of colocalization, Biotechniques 30: 1240-1246. Stark, P.R.H .. Rinko, L.J., and Larson, D.N., 2003, Fluorescent resolution target for super-resolution microscopy, 1. Microsc. 212:307-310. Tran, P., 2005, CCD cameras for fluorescence imaging of living cells, In: Live Cell Imaging: A Laboratory Manual (R.D. Goldman and D.L. Spector, eds.), Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, pp. 87-100. van den Doel, L.R., Klein, A.D., Ellenberger, S.L., Netten, H., Boddeke, F.R., van Vlict, L.J., and Young, I.T, 1998, Quantititative evaluation of light microscopes based on image processing techniques, Bioimaging 6:138-149. van der Voort, H.TM, Brakenhoff, G.J., and Janssen, G.C.A.M., 1988, Determination of the 3-dimensional optical properties of a confocal scanning laser microscope, Optik 78:48-53. Van Steensel, B., van Binnendijk, E.P., Hornsby, C.D., van der Voort, H.TM., Krozowski, Z.S., de Kloet, E.R., and van Driel, R., 1996, Partial colocalization of glucocorticoid and mineralocorticoid receptors in discrete compartments in nuclei of rat hippocampus neurons, J. Cell Sci. 109: 787-792. Visser, TD., Brakenhoff, G.J., and Groen, FC.A., 1991, The one-point fluorescence response in confocal microscopy, Optik 87:39-40. Wallace, W., Scbaefer, L.H., and Swedlow, J.R., 2001, A workingperson's guide to deconvolution in light microscopy, Biotechniques 31:1076-1097. Wilson, T, and luskaitis, R., 1995, The axial response of confocal microscopes with high numerical aperture objective lenses, Bioimaging 3:35-38. Young, M., 2000, Impulse response and transfer function, In: Optics and Lasers (M. Young, ed.), Springer-Verlag, New York, pp. 181-192. Zucker, R.M., and Lerner, J.M., 2004, Calibration and validation of confocal spectral imaging systems, Cytometry 62A:8-34. Zucker, R.M., and Price, OT, 1999, Practical confocal microscopy and the evaluation of system performance, Methods 18:447-458. Zucker, R.M., and Price, O.T, 2001a, Evaluation of confocal microscopy system performance, Cytometry 44:273-294. Zucker, R.M., and Price, O.T., 2001b, Statistical evaluation of confocal microscopy images, Cytometry 44:295-308. Zwier, 1.M., Van Rooij, G.J., Hofstraat, J.W., and Brakenhoff, GJ., 2004, Image calibration in fluorescence microscopy, J. Mierosc. 216:15-24.
37
Selective Plane Illumination Microscopy Jan Huisken, Jim Swager, Steffen Lindek, and Ernst H.K. Stelzer
INTRODUCTION In this chapter we present an alternative way of optically sectioning the sample in fluorescence microscopy. By illuminating the sample from the side with a sheet of light, and viewing the light emitted or scattered by this layer with a widefield microscope oriented on an axis perpendicular to the sheet, a sectioning effect that is similar to that in confocal microscopy can be produced. Such an approach has several advantages over confocal scanning microscopy and these make the technology especially well suited for relatively large samples such as embryos. However, the principle is universal and can also be applied to micrometer-sized samples. This chapter introduces light-sheet microcopy by providing an overview of microscopy techniques for large samples and the history of techniques employing plane illumination. The lightsheet microscope that we have developed at the EMBL is illustrated and explained, its characteristics are described and evaluated in the context of confocal microscopy and its performance demonstrated by a few applications. The stereo microscope is the most commonly used microscope in developmental biology. It provides a stereoscopic image of the sample while keeping it under ideal conditions for sorting and selecting embryos (e.g., in a Petri dish filled with an appropriate medium). A camera can be used to record images or even movies of the developing embryo for later reference. However, no volumetric quantification can be made because the stereo microscope offers little depth discrimination and no optical sectioning. The same holds true for conventional widefield microscopes, which in addition often suffer from short working distances that are inadequate for imaging large samples. Quantitative analysis of three-dimensional (3D) structures requires sectioning (optical or otherwise). Out-of-focus light is rejected in confocal microscopy by coupling point excitation with a laser beam to point detection through a pinhole, so that in general only in-focus light is detected. This principle works well in relatively thin samples (up to ca. 100/-lm); however, in large and scattering objects, the signal is lost since most of the fluorescent light is scattered and therefore rejected by the pinhole. The confocal microscope, therefore, suffers from a limited penetration depth, and cannot be applied for in toto studies of most embryos. In such samples, the best resolution is obtained by physically sectioning the sample (Weninger, 2002), which is slow, laborious, destroys the sample irretrievably, and is not applicable for studies of living preparations. The many methods that have recently been developed to enhance the resolution, e.g., 4Pi-confocal (Hell, 1992), confocal theta microscopy (Stelzer, 1994), 15M microscopy (Gustafsson, 1999), and Stimulated Emission Depletion (STED) microscopy (Klar, 2000) do not address the challenges encountered in large and
scattering samples such as embryos. Their improved resolution is easily degraded by sample-induced wavefront distortions, and in addition, they generally require quite complex instrumentation. Chapter 34 presents an overview of imaging techniques for large samples, such as Optical Projection Tomography (OPT; Sharpe, 2002) and Micro Magnetic Resonance Imaging (/-lMRI; Louie, 2000), which to some extent attempt to overcome these challenges.
COMBINING LIGHT SHEET ILLUMINATION AND ORTHOGONAL DETECTION "Light-sheet" microscopy now provides an alternative approach for imaging large samples at high resolution. Like confocal microscopy, techniques based on sheet illumination have been "invented" a number of times and all owe much to the "Ultramikroskop," an orthogonal, darkfield illuminator invented by Siedentopf and Zsigmondy in 1903, to visualize gold particles much smaller than the resolution limit by detecting light scattered perpendicular to the illumination. The light-sheet concept has been used in several other microscopes since then. For decades the slit-lamp was used for viewing the human eye (Campbell et al., 1974) and similar techniques were reviewed by Webb in Chapter 34 of the second edition of this Handbook. More recently, Voie used the high coherence of laser light to develop orthogonal-plane fluorescence sectioning (OPFOS) and applied it to the cochlea (Voie, 1993, Voie et al., 1995). I Although initially operating at a fixed magnification and a resolution set by the close-focus camera lens used to convey the image to the CCD camera, more recent implementations (Voie, 2002) used a 4x optical system and a rotating specimen immersed in index-matching medium to record details about lO/-lm in size. In 2002, Fuchs described the Thin Laser Light Sheet Microscope (TLLSM) (Fuchs, 2002) to detect, discriminate and document microbes passing through the light sheet. Although it has the same optical arrangement as Siedentopf and Zsigmondy, TLLSM uses fluorescence, which was not practical 100 years ago. Because in this system, the stage is stationary relative to the light sheet, only the slice of the sample that happens to intercept the sheet is imaged, making the system unsuitable for generating threedimensional images. Another technique with orthogonal illumination and detection is 3D Light Scanning Macrography (Huber, 2001), which is used for generating macroscopic reconstructions of samples using scat-
I
More information on spencertechnologies.com
this
system
can
be
obtained
at
www.
Jan Huisken, Jim Swoger, Steffen Lindek, and Ernst H.K. Stelzer· European Molecular Biology Laboratory (EMBL), 69012 Heidelberg, Germany
672
Handbook of Biological Confocal Microscopy, Third Edition. edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
Selective Plane Illumination Microscopy • Chapter 37
tered light. The orthogonal arrangement of illumination and detection was also used in theta microscopy (Stelzer, 1994), although no light sheet is formed in this variant of the single-point confocal scanning microscope. Our current implementation of the technique is the latest of three versions of the selective plane illumination microscope (SPIM). It is embodied in a compact, stable, and economical system based on three operating principles: illumination of the sample with a light sheet, observation of light coming from the sample in a direction perpendicular to the illumination axis, and, optionally, rotation of the sample about an axis parallel to gravity. A 3D image is recorded by scanning the sample through the stationary light sheet and recording the fluorescence or scattered light with a CCD camera. The sample, which can be as small as a few micrometers or as large as several millimeters depending on the working distance of the objective lenses, can be embedded in a gel, immersed in a liquid or held in air. As it is mounted on a stage that can be rotated as well as translated, one can record 3D images from multiple directions, and these data sets can be combined into a single 3D data set with a spatial resolution dominated by the lateral resolution of the detection system.
673
SELECTIVE PLANE ILLUMINATION MICROSCOPY SETUP Figure 37.1 shows the main components of SPIM. A series of lasers [several helium-neon (He-Ne) and a multi-line argon (Ar)ion laser] provide a range of excitation wavelengths, which can be focused by a cylindrical lens to form a thin light sheet. The sample is mounted in a transparent cylinder, typically made of agarose prepared using an aqueous solution suitable for the sample (e.g., phosphate-buffered saline, PBS). The agarose cylinder is then immersed in a bath of this solution, virtually eliminating refractive index artifacts at the surface. The cylinder is held from above by a micropositioning device with four degrees of freedom (three translational, one rotational), which allows positioning the sample such that the excitation light illuminates any plane of interest. An objective lens placed perpendicular to the illumination plane, a detection filter, and a tube lens are used to image fluorescent light excited by the light sheet onto a CCD camera. A variety of objective lenses can be used (preferably those designed for imaging in water without a coverslip), with magnifications ranging from 2.5x
chamber
FIGURE 37.1 . Schematic setup of a selective plane illumination microscope. The light sources (back), illumination path (front, left), and detection path (front, right) are shown. The beams of two lasers are superimposed and the acousto-optical tunable filter (AOTF) is used to select and to attenuate the desired laser line. (The use of the optical fiber to deliver the laser light to the microscope helps to reduce mechanical vibrations, but is actually optional.) The laser light is focu sed by the cylindrical lens to form a light sheet that traverses the chamber. The sample is placed into the chamber in the path of the light sheet. Widefield detection is performed with the objective lens, filter, tube lens, and CCD camera. (Inset) Detailed view of the chamber with sample. illumination beam, and detection lens. The light sheet enters the aqueous medium-filled chamber (shown cut open) through a thin glass window. The sample is embedded in a cylinder of agarose, which is supported from above and can be moved by a translating and rotating stage. The objective lens is sealed into the chamber with a rubber 0ring, and can be moved axially to focus on the plane of fluorescence excited by the light sheet. The sample is then moved through the light sheet to acquire a 3D data set.
674
Chapter 37 •
J.
Huisken et al.
illumination
system
detection
2
2
2
1.5
1.5
1.5
E 0.5
E 0.5 ::1. -..
.....
0
N
n;
E 0.5
:::I. ..... N
N
n;
n;
'x
.~
.~ 0.5
'"
1.5 2 2
0.5 0 0.5 lateral r I lJm
1.5
1.5
2 lateral r him
lateral r him
FIGURE 37.2. Theoretical intensity point spread functions (PSF) in the SPIM. The system PSF (right) is obtained by multiplying the illumination (left) and detection PSFs (center). "'ill = 488nm, "'del = 520nm, NAill = 0.15, NAdot = 0.8, n = 1.33. Contour lines are plotted for intensities of 0.9,0.7,0.5,0.3,0.2,0.1, 0.05,0.03,0.02,0.015,0.01,0.005,0.002, and 0.001. The intensities are shown with a lookup table (LUT) of Y= 0.25, in blue for excitation, in green for detection, and in red for the system.
to 100x, numerical apertures (NAs) from 0.16 to 1.0, and working distances of 1 to l2mm.
LATERAL RESOLUTION The lateral resolution of SPIM is determined by the NA of the detection lens, as is generally the case in widefield microscopy. However, if the images are under-sampled by the CCD, the lateral resolution of the detection lens is not fully exploited, and the pixel size of the camera determines the lateral resolution. This is a general problem in digital microscopy using currently available objective lenses and cameras, and is not unique to SPIM. Figure 37.2 shows the point spread functions (PSF) of illumination and detection in SPIM, as well as the product of the two, that is, the system PSF, for typical NAs. The light sheet illumination does not affect the lateral resolution.
LIGHT SHEET THICKNESS AND AXIAL RESOLUTION There are two distinct yet related effects of the light sheet dimensions that are relevant to SPIM. First, selective illumination by the light sheet provides optical sectioning in SPIM, and the extent of this capability depends on the thickness of the light sheet. Second, the light sheet improves the axial resolution of the SPIM if it is thinner than the axial extent of the detection PSF (Fig. 37.2). For both of these reasons the optimal light sheet for SPIM is made as
OL
CL
c:
thin as possible, while maintaining its properties approximately uniform across the field of view. The relations between the light sheet dimensions and the detection field of view are schematically illustrated in Figure 37.3. The minimum thickness of the light sheet is inversely proportional to the NA of the cylindrical illumination lens, while the width of this thin waist (i.e., its extent along the illumination axis) is inversely proportional to its NA2. While the importance of the thickness of the light sheet is obvious (for the reasons given above), that of its width is perhaps less so. However, when one considers that it is desirable to have the extent of the optical sectioning uniform across the field of view, it becomes clear that it is necessary to have the light sheet width at least as large as the detection field of view (see Fig. 37.3). Thus, the optimal SPIM design involves a compromise between strong optical sectioning (requiring a large illumination NA) and a large field of view (requiring a small illumination NA). The simplest means of balancing these two parameters is to change the focal length of the cylindrical lens or the size of its aperture. As an example of a configuration suitable for relatively large samples, with a lOx/0.30 detection objective lens the SPIM has a field of view of 660 11m. A Gaussian light sheet can be formed with a thickness that varies between 6 11m and 9 11m across this distance by using a cylindrical lens of focal length 100mm and aperture 6mm. In addition to providing optical sectioning, this considerably improves the axial resolution of the system, which would be -141lm with conventional illumination. If SPIM is to be used for higher-resolution imaging of smaller samples, a I OOx/l.O with an axial PSF width of 1.08 11m and a field
:=:> fov
- ---------1 -
wd
- --==-=--*-----w------1.....:
-----j
FIGURE 37.3. Light sheet illumination (blue) and widefield detection (green) in the SPIM. Illumination: CL, cylindrical lens; NAill = n sin ai; t, light sheet thickness; w, light sheet width. Detection: OL, objective lens; NA"eI = n sin a,,; fov, field of view; wd, working distance.
Selective Plane Illumination Microscopy • Chapter 37
of view of 66 ~m is more appropriate. For this detection lens, a light sheet thickness of 2 ~m yields a reasonably uniform illumination across the field of view (thickness variation C 0
50 0
I::::
'-'
Ql
a:
'-' CI)
~
Ql
10
> 0
'-' Ql
a:
0.1
Ql
as
°[2]
10
1
~
Ql
'-'
30
100
Power (mW)
Ql
100 '-' c Ql
B
'-' CI)
~
~
0
0
:::J
50
u::: 10
:::J
u:::
:!2 0
'if!-
10
740 Energy [Il-J]
0
Wavelength (nm)
FIGURE 39.10. (A) Percentage of single-protein fluorescence recovery versus the illumination time at nOnm for different excitation powers: 3.34mW (solid circles), 4.34mW (solid squares), 5mW (open squares), and 5.7mW (open circles). The solid lines are best-fit to a sigmoidal function. (B) Recovery efficiency with laser excitation at no nm versus the product of the illumination time with excitation power: 3.34mW (solid circles), 4.34mW (solid squares), 5mW (open squares), and 5.7mW (open circles). The solid line is the best-fit power law with an exponent of 3.8 ± 0.2. (C) Recovery efficiency at a fixed illumination time of 16.5 ms versus the excitation wavelength at 5.7 mW (open circles), 4.34mW (filled squares), and 3.34mW (filled circles). The error bars are the 95% statistical errors due to the number of events observed. The inset shows the photobleaching rate 1fTh versus excitation power for the two states (A, open squares; B, filled squares). (After Chirico et aI., 2004.)
range. A subtle interplay between different electronic coupling mechanisms is responsible for an efficient transfer of the absorbed energy to the reaction center. Recently, a study of photobleaching and energy transfer in single phycoerythrocyanin (PEC) monomer has been presented by Zehetmayer and colleagues (2002). The PEC monomer contains two different chromophores, phycoviolobilin (PVB) and phycocyanobilin (PCB) (Zhao and Scheer, 1995). Zehetmayer and co-workers (2002) recorded single-molecule images of phycoerythrocyanin monomers. Their photobleaching behavior was studied by simultaneously exciting at two wavelengths according to a method earlier applied on the E222Q GFP mutant by the same group (lung et ai., 2001). The PVB chromophore was found to be responsible for the photobleaching of PEe. On the other hand, it was possible to ascertain that the l5E form of PVB corresponds to one of the short-lived dark states of PEe. This form does not induce real photobleaching but simply reduces the average fluorescence emission, as is the case in most of the GFP mutants. The difficulty in discriminating between nonreversible photobleaching and long-term blinking therefore seems to be ubiquitous in single-protein photophysics.
CONCLUSION It is not easy to indicate conclusions that define an optimal strategy
for either reducing or exploiting the photobleaching process. We hope that the facts reported in this chapter could be useful to design new experiments or to revise old ones with the aim of a better understanding of the delicate, intricate, and complex structure-function
700
Chapter 39 • A. Diaspro et al.
relationship that is the basis of our job as microscopists or, more generally, as biophysicists. We decided to attack photobleaching by considering both its imaging and its single-molecule aspects because we think that feedback between these related aspects of the problem can greatly improve our knowledge of photobleaching.
ACKNOWLEDGMENTS We are greatly indebted to Michael W. Davidson of Florida State University (Tallahassee, FL) for providing Figure 39.7. We also acknowledge the work of collaborators in our laboratories by mentioning, in random order, Davide Mazza, Maddalena Collini, Valentina Caorsi, Michele Caccia, Paolo Bianchini, Fabio Cannone, Laura D' Alfonso, Giuseppe Vicidomini, Tytus Bernas, Marc Schneider, Silke Krol, Raffaella Magrassi, Miroslaw Zarebski, Ilaria Testa, Federica Morotti, Mattia Pesce, and Francesca Cella. Alberto Diaspro thanks Teresa and Claudia for accepting long weekends and nights writing and revising. Last, but not least, the authors are very grateful to Jim Pawley for his patience and constructive criticism. This work was supported by MIUR, INFM, IFOM, compagnia di San Paolo (AD), and The Wellcome Trust (JD) grants.
REFERENCES Agard, D.A., Hiraoka, Y., Shaw, P.J., and Sedat, J.W., 1989, Fluorescence microscopy in three-dimensions, Methods Cell Bioi. 30:353-378. Amos, W.B., and White, J.G., 2003, How the confocal laser scanning microscope entered biological research, Bioi. Cell 95:335-342. Arndt-lovin, D.J., Nicoud, R.M., Kaufmann, J., and Jovin, T.M., 1985, Fluorescence digital-imaging microscopy in cell biology, Science 230: 13330-13335. Axelrod, D., Koppel, D.E., Schlessinger, J., Elson, E., and Webb, W.W., 1976, Mobility measurement by analysis of fluorescence photobleaching recovery kinetics, Biophys. J. 16:1055-1069. Basche, T, 1998, Fluorescence intensity fluctuations of single atoms, molecules and nanoparticles, J. Lurnin. 76-7:263-269. Bastiaens, P.I., and Hell, S.W., 2004, Recent advances in light microscopy, J. Struct. Bioi. 147: 1-89. Benson, D.M., Bryan, J., Plant, A.L., Gotto, A.M. Jr., and Smith, L.c., 1985, Digital image fluorescence microscopy: Spatial heterogeneity of photobleaching rate constants in individual cells, J. Cell BioI. 100:1309-1323. Berglund, A.J., 2004, Nonexponential statistics of fluorescence photobleaching, J. Chern. Phys. 121:2899-2903. Bernas, T, Cook, P.R., and Dobrucki, J.W., 2005, Confocal fluorescence imaging of photosensitised DNA denaturation in cell nuclei, Photochern. Photobiol. 81(4):960-969. Bernas, T, Zantbski, M., Cook, P.R., and Dobrucki, J.W., 2004, Minimizing photobleaching during confocal microscopy of fluorescent probes bound to chromatin: role of anoxia and photon flux, J. Microsc. 215:281-296. Bianco, B., and Diaspro, A., 1989, Analysis of the three dimensional cell imaging obtained with optical microscopy techniques based on defocusing, Cell Biophys. 15:189-200. Birks, J.B., 1970, Photophysics of Aromatic Molecules, Wiley Interscience, London. Bloom, J.A., and Webb, W.W., 1984, Photodamage to intact erythrocyte membranes at high laser intensities: Methods of assay and suppression, J. Histochern. Cytochern. 32:608-616. Bock, G., Hilchenbach, M., Schauenstein, K., and Wick, G., 1985, Photometric analysis of antifading reagents for immunofluorescence with laser and conventional illumination sources, J. Histochern. Cytochern. 33:699-705. Bonetto, P., Boccacci, P., Scarito, M., Davolio, M., Epifani, M., Vicidomini, G., Tacchetti, C., Ramoino, P., Usai, c., and Diaspro, A., 2004, Threedimensional microscopy migrates to the Web with PowerUp Your Microscope, Microsc. Res. Tech. 64: 196-203.
Braeckmans, K., De Smedt, S.c., Roelant, c., Leblans, M., Pauwels, R., and Demeester, J., 2003, Encoding microcarriers by spatial selective photobleaching, Nat. Mater. 2:169-173. Braga, J., Desterro, 1.M.P., and Carmo-Fonseca, M., 2004, Intracellular macromolecular mobility measured by fluorescence recovery after photobleaching with confocal laser scanning microscopes, Mol. BioI. Cell 15:4749-4760. Brakenhoff, GJ., Muller, M., and Ghauharali, R.I., 1996, Analysis of efficiency of two-photon versus single-photon absorption of fluorescence generation in biological objects, J. Microsc. 183:140--144. Cannone, F., Chirico, G., and Diaspro, A., 2003, Two-photon interactions at single fluorescent molecule level, J. Biorned. Opt. 8:391-395. Chen, R.F., and Scott, C.H., 1985, Atlas of fluorescence spectra and lifetimes of dyes attached to protein, Anal. Lett. 18:393-421. Chirico, G., Cannone, F., Baldini, G., and Diaspro, A., 2003, Two-photon thermal bleaching of single fluorescent molecules, Biophys. 1. 84:588-598. Chirico, G., Cannone, F., Beretta, S., and Diaspro, A., 2001, Single molecule studies by means of the two-photon fluorescence distribution, Microsc. Res. Tech. 55:359-364. Chirico, G., Cannone, F., Beretta, S., Diaspro, A., Campanini, B., Bettati, S., Ruotolo, R., and Mozzarelli, A., 2002, Dynamics of green fluorescent protein mutant2 in solution, on spin-coated glasses, and encapsulated in wet silica gels, Protein Sci. II: 1152-1161. Chirico, G., Cannone, F., Diaspro, A., Bologna, S., Pellegrini, v., Nifosi, R., and Beltram, F., 2004, Multiphoton switching dynamics of single green fluorescent proteins, Phys. Rev. E Stat. Nonlin. Soft. Matter Phys. 70:03090 I (R). Cinelli, R.A.G., Ferrari, A., Pellegrini, v., Tyagi, M., Giacca, M., and Beltram, F., 2000, The enhanced green fluorescent protein as a tool for the analysis of protein dynamics and localization: Local fluorescence study at the single-molecule level. Photochern. Photobiol. 71 :771-776. Cinelli, R.A.G., Pellegrini, V., Ferrari, A., Faraci, P. , Nifosi , R., Tyagi, M., Giacca, M., and Beltram, F., 2001, Green fluorescent proteins as optically Controllable elements in bioelectronics, Appl. Phys. Lett. 79:33533355. Cole, N.B., Smith, c.L., Sciaky, N., Terasaki, M. , Edidin, M., and LippincottSchwartz, J., 1996, Diffusional mobility of Golgi proteins in membranes of living cells, Science 273:797-801. Cubitt, A.B., Woollenweber, L.A., and Heim, R., 1997, Understanding structure-function relationship in the Aequoria Victoria Green Fluorescent Protein, In: Green Fluorescent Proteins (K.F. Sullivan and S.A. Kay, eds.), Academic Press, San Diego, California, p. 19. Davis, S.K., and Bardeen, CJ., 2002, Using two-photon standing waves and patterned photobleaching to measure diffusion from nanometers to microns in biological systems, Rev. Sci. Instrurn. 73:2128-2135. Delon, A., Us son, Y., Derouard, J., Biben, T., and Souchier, c., 2004, Photobleaching, mobility and compartmentalisation: inferences in fluorescence correlation spectroscopy, J. Fluoresc. 14:255-267. Denk, W., Strickler, J.H., and Webb, W.W., 1990, Two-photon laser scanning fluorescence microscopy, Science 248:73-76. Deschenes, L.A., and van den Bout, D.A., 2002, Single molecule photobleaching: increasing photon yield and survival time through suppression of two step photolysis, Chern. Phys. Lett. 365:387-395. Diaspro, A., 2002, Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances, Wiley-Liss, New York. Diaspro, A., 2004, Confocal and multiphoton microscopy, In: Lasers and Current Optical Techniques in Biology (G. Palumbo and R. Pratesi, eds.), RSC-Royal Society of Chemistry, Cambridge, United Kingdom, pp. 429-478. Diaspro, A., and Robello, M., 2000, Two-photon excitation of fluorescence for three-dimensional optical imaging of biological structures, J. Photochern. Photobiol. B 55:1-8. Diaspro, A., Chirico, G., and Collini, M., 2005, Two-photon fluorescence excitation in biological microscopy and related techniques, Quart. Rev. Biophys. In press. Diaspro, A., Chirico, G., Federici, F., Cannone, F., Beretta, S., and Robello, M., 2001, Two-photon microscopy and spectroscopy based on a compact confocal scanning head, J. Biorned. Opt. 6:300--310.
Photobleaching • Chapter 39
Diaspro, A., Federici, F, Viappiani, C., Krol, S., Pisciotta, M., Chirico, G., Cannone, F, and Gliozzi, A., 2003, Two-photon photolysis of 2-nitrobenzaldehyde monitored by fluorescent-labeled nanocapsules, J. Phys. Chern. B 107:11008-11012. Dickson, RM., Cubitt, A.B., Tsien, R.Y, and Moerner, W.E., 1997, On/off blinking and switching behaviour of single molecules of green fluorescent protein, Nature 388:355-358. Dittrich, P.S., and Schwille, P., 2001, Photobleaching and stabilization of fluorophores used for single molecule analysis with one- and two-photon excitation, Appl. Phys. B 73:829-837. Dixit, R., and Cyr, R, 2003, Cell damage and reactive oxygen species production induced by fluorescence microscopy: Effect on mitosis and guidelines for non-invasive fluorescence microscopy, Plant J. 36: 280-290. Dobrucki, 1.W., 2001, Interaction of oxygen-sensitive luminescent probes Ru(phen)j+ and Ru(bipy)j+ with animal and plant cells in vitro. Mechanisms of phototoxicity and conditions for non-invasive oxygen measurements. J. Photochem. Photobiol. B. 65:136-144. Dobrucki, 1.W., and Darzynkiewicz, Z., 2001, Chromatin condensation and sensitivity of DNA in situ to denaturation during cell cycle and apoptosis - a confocal microscopy study, Micron. 32:645-652. Drummond, D.R., Carter, N., and Cross, R.A., 2002, Multiphoton versus confocal high-resolution z-sectioning of enhanced green fluorescent microtubules: Increased multi photon photobleaching within the focal plane can be compensated using a Pockels cell and dual widefield detectors, J. Microsc. 206:161-169. Edidin, M., Zagyansky, Y, and Lardner, Tl., 1976, Measurement of membrane protein lateral diffusion in single cells, Science. 191:466-468. Eggeling, c., Brand, L., and Seidel, C.A.M., 1997, Laser-induced fluorescence of coumarin derivatives in aqueous solution: Photochemical aspects for single molecule detection, Bioimaging. 5:105-115. Eggeling, c., Widengren, 1., Rigler, R, and Seidel, C.A.M., 1998, Photobleaching of fluorescent dyes under conditions used for single molecule detection: Evidence of two step photolysis, Anal. Chern. 70:2651-2659. Ellenberg, 1., Lippincott-Schwartz, 1., and Presley, 1.F, 1998, Two-color green fluorescent protein time-lapse imaging, Biotechniques. 25:838-842, 844-846. Emptage, N.l., 2001, Fluorescent imaging in living systems, Curr. Opin. Pharmacol. 1:521-525. Garcia-Parajo, M.F, Segers-Nolten, G.M.l., Veerman, l.-A., Greve, 1., and van Hulst, N.F, 2000, Real-time light-driven dynamics of the fluorescence emission in single green fluorescent protein molecules, Proc. Nat. Acad. Sci. USA 97:7237-7242. Greenbaum, L., Rothmann, C., Lavie, R, and Malik, Z., 2000, Green fluorescent protein photobleaching: a model for protein damage by endogenous and exogenous singlet oxygen, Bio!. Chern. 381: 1251-1258. Harper, I.S., 2001, Fluorophores and their labeling procedures for monitoring various biological signals, In: Methods in Cellular Imaging (A. Periasamy, ed.), Oxford University Press, New York, pp. 20-39. He, 1.A., Hu, Y.Z., and liang, L.1., 1997, Photodynamic action of phycobiliproteins: In situ generation of reactive oxygen species, Biochim. Biophys. Acta 1320:165-174. Heim, R, Prasher, D.C., and Tsien, RY, 1994, Wavelength mutations and posttranslational autoxidation of green fluorescent protein, Proc. Natl. Acad. Sci. USA 91:12501-12504. Hines, M.A., and Guyot-Sionnest, P., 1996, Synthesis and characterization of strongly luminescing ZnS-Capped CdSe nanocrystals, J. Phys. Chern. 100:468-471. Hirschfeld, T, 1976, Quantum efficiency independence of the time integrated emission from a fluorescent molecule, Appl. Opt. 15:3135-3139. Hockberger, P.E., Skimina, TA., Centonze, V.E., Lavin, c., Chu, S., Dadras, S., Reddy, 1.K., and White, 1.G., 1999, Activation of flavin-containing oxidases underlies light-induced production of H 20 2 in mammalian cells, Proc. Nat. Acad. Sci. USA 96:6255-6260. laiswal, 1.K., and Simon, S., 2004, Potentials and pitfalls of fluorescent quantum dots for biological imaging, Trends Cell Bioi. 14:497-504. lung, G., Wiehler, 1., Gilhde, W., Tittel, 1., Basche, Th., Steipe, B., and Brauchle, c., 1998, Confocal microscopy of single molecules of the green fluorescent protein, Bioimaging 6:54-61.
701
Kao, Fl., Wang, YM., Chen, 1.C., Cheng, P.C., Chen, RW., and Lin, B.L., 2002, Photobleaching under single photon and multi-photon excitation: Chloroplasts in protoplasts from Arabidopsis thaliana, Opt. Commun. 201:85-91. Kawano, H., Nabekawa, Y, Suda, A., Oishi, Y, Mizuno, H., Mijawaki, A., and Midorikawa, K., 2003, Attenuation of photobleaching in two-photon excitation fluorescence from green fluorescent protein with shaped excitation pulses, Biochem. Biophys. Res. Commun. 311:592-596. Ko, D.S., 2004, Photobleaching time distribution of a single tetramethylrhodamine molecule in agarose gel, J. Chern. Phys. 120:2530-2531. Kohen, E., Reyftmann, 1.P., Morliere, P., Santus, R, Kohen, C., Mangel, W.F, Dubertret, L., and Hirschberg, 1.G., 1986, A microspectrofluorometric study of porphyrin-photosensitized single living cells. II. Metabolic alterations, Photochem. Photobiol. 44:471-475. Konig, K., and Tirlapur, U.K., 2002, Cellular and subcellular perturbations during multiphoton microscopy, In: Confocal and Two-Photon Microscopy: Foundations, Applications and Advances (A. Diaspro, ed.), Wiley-Liss, New York, pp. 191-205. Krenik, K.D., Kephart, G.M., Offord, K.P., Dunnette, S.L., and Gleich, G.l., 1989, Comparison of antifading agents used in immunofluorescence, J. Immunol. Methods 117:91-97. Kriete, A., 1992, Visualization in Biomedical Microscopies, VCH, Weinheim, Germany. Kubitschek, U., Ktickmann, 0., Kues, T, and Peters, R, 2000, Imaging and tracking of single GFP molecules in solution, Biophys. 1. 78: 2170-2179. Lakowicz,l.R, 1999, Principles of Fluorescence Microscopy, Plenum Press, New York. Landon, TM., 1997, Alexa dyes, tracers and conjugates, Bioprobes 26: 1-6. Larson, D.R., Zipfel, W.R, Williams, R.M., Clark, S.W., Bruchez, M.P., Wise, FW., and Webb, W.W., 2003, Water-soluble quantum dots for multiphoton fluorescence imaging in vivo, Science 300:1434-1436. Lippincott-Schwartz, 1., and Patterson, G.H., 2003, Development and use of fluorescent protein markers in living cells, Science 300:87-91. Maher, RC., Cohen, L.F, and Etchegoin, P., 2002, Single molecule photobleaching observed by surface enhanced resonant Raman scattering (SERRS), Chern. Phys. Lett. 352:378-384. Manitto, P., Speranza, G., Monti, D., and Gramatica, P., 1987, Singlet oxygen reactions in aqueous solution. Physical and chemical quenching rate constants of crocin and related carotenoids, Tetrahedron Lett. 28: 4221-4224. Mathies, RA., and Stryer, L., 1986, Single-molecule fluorescence detection: A feasibility study using phycoerythrin, In: Applications of Fluorescence in the Biomedical Sciences (D.L. Taylor, A.S. Waggoner, RF Murphy, F Lanni, and RR. Birge, eds.), Alan R Liss, New York, pp. 129-140. Matsumoto, B., 2002, Cell Biological Applications of Confocal Microscopy, 2nd ed., Academic Press, San Diego, California. Mattoussi, H., Mauro, 1.M., Goldman, E.R., Anderson, G.P., Sundar, V.c., Mikulec, FV, and Bawendi, M.G., 2000, Self-assembly of CdSe-ZnS quantum dot bioconjugates using an engineered recombinant protein, J. Am. Chern. Soc. 122:12142-12150. McNally, 1.G., and Smith, c.L., 2002, Photobleaching by confocal microscopy. In: Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances (A. Diaspro, ed.), Wiley-Liss, Inc., New York, pp. 525-538. Mertz, 1., 1998, Molecular photodynamics involved in multi-photon excitation fluorescence, Eur. Phys. 1. D3:53-66. Michalet, X., Kapanidis, A.N., Laurence, T., Pinaud, F, Doose, S., Pflughoefft, M., and Weiss, S., 2003, The power and prospects of fluorescence microscopies and spectroscopies, Annu. Rev. Biophys. Biomo!. Struct. 32:161-182. Moerner, W.E., Peterman, E.J.G., Brasselet, S., Kummer, S., and Dickson, RM., 1999, Optical methods for exploring dynamics of single copies of green fluorescent protein, Cytometry 36:232-238. Molski, A., 2001, Statistics of the bleaching number and the bleaching time in single-molecule fluorescence spectroscopy, J. Chern. Phys. 114: 1142-1147. Ness, 1.M., Akhtar, RS., Latham, C.B., and Roth, K.A., 2003, Combined tyramide signal amplification and quantum dots for sensitive and photostable immunofluorescence detection, J. Histochem. Cytochem. 5:981-987.
702
Chapter 39 • A. Diaspro et al.
Nifosi, R., Ferrari, A., Arcangeli, C., Tozzini, V, Pellegrini, V, and Beltram, E, 2003, Photoreversible dark state in a tristable green fluorescent protein variant, J. Chem. Phys. B 107:1679-1684. Ono, M., Murakami, T, Kudo, A., Isshiki, M., Sawada, H., and Segawa, A., 2001, Quantitative comparison of anti-fading mounting media for confocal laser scanning microscopy, J. Histochem. Cytochem. 49:305-311. Ormo, M., Cubitt, A.B., Kallio, K., Gross, L.A., Tsien, R.Y., and Remington, S.J., 1996, Crystal structure of the Aequorea victoria green fluorescent protein, Science 273: 1392-1395. Patterson, G.H., and Piston, D.W., 2000, Photobleaching in two-photon excitation microscopy, Biophys. J. 78:2159-2162. Pawley, 1.B., 1995a, Handbook of Biological Confocal Microscopy, Plenum Press, New York. Pawley, 1.B., 1995b, Fundamental limits in confocal microscopy, In: Handbook of Biological Confocal Microscopy (l.B. Pawley, ed.), Plenum Press, New York, pp. 19-37. Periasamy, A., 2001, Methods in Cellular Imaging, Oxford University Press, New York. Peterman, E.l.G., Brasselet, S., and Moerner, W.E., 1999, Optical methods for exploring dynamics of single copies of green fluorescent protein, J. Phys. Chem. A 103: 10553-10560. Peters, R, Peters, 1., Tews, K.H., and Bahr, W., 1974, A microfluorimetric study of translational diffusion in erythrocyte membranes, Biochem. Biophys. Acta 367:282-294. Poo, M., and Cone, RA., 1974, Lateral diffusion of rhodopsin in the photoreceptor membrane, Nature 247:438-441. Reits, E.A., and Neefjes, 1.1., 2001, From fixed to FRAP: measuring protein mobility and activity in living cells, Nat. Cell Bioi. 3:EI45-EI47. Reyftmann, 1.P., Kohen, E., Morliere, P., Santus, R, Kohen, C, Mangel, W.E, Dubertret, L., and Hirschberg, 1.G., 1986, A microspectrofluorometric study of porphyrin-photosensitized single living cells. 1. Membrane alterations, Photochem. Photobiol. 44:461-469. Rizzuto, R, Brini, M., DeGiorgi, E, Rossi, R, Heim, R, Tsien, R.Y., and Pozzan, T, 1996, Double labelling of subcellular structures with organelletargeted GFP mutants in vivo, Curr. Bioi. 6: 183-188. Schneider, M.B., and Webb, W.W., 1981, Measurements of submicron laser beam radii, Appl. Opt. 20: 1382-1388. Sokolov, K., Aaron, 1., Hsu, B., Nida, D., Gillenwater, A., Follen, M., MacAulay, C, Adler-Storthz, K., Korgel, B., Descour, M., Pasqualini, R, Arap, W., Lam, W., and Richards-Kortum, R, 2003, Optical systems for in vivo molecular imaging of cancer, Technol. Cancer Res. Treat. 2:491-504. Song, L., Hennink, E.J., Young, LT, and Tanke, H.l., 1995, Photobleaching kinetics of fluorescein in quantitative fluorescence microscopy, Biophys. J. 68:2588-2600. Song, L., Varma, CA., Verhoeven, 1.W., and Tanke, H.l., 1996, Influence of the triplet excited state on the photobleaching kinetics of fluorescein in microscopy, Biophys. 1. 70:2959-2968. Souchier, L.A., Ffrench, M., and Bryon, P.A., 1993, Comparison of antifading agents used in fluorescence microscopy: image analysis and laser confocal microscopy study, J. Histochem. Cytochem. 41: 18331840. Stavreva, D.A., and McNally, J.G., 2004, Fluorescence recovery after photobleaching (FRAP) methods for visualizing protein dynamics in living mammalian cell nuclei, Methods Enzymol. 375:443-455. Stelzer, E.H.K., 1998, Contrast, resolution, pixelation, dynamic range and signal to noise ratio: fundamental to its resolution in fluorescence light microscopy, J. Microsc. 189: 15-24.
Stephens, D.l., and Allan, Vl., 2003, Light microscopy techniques for live cell imaging, Science 300:82-86. Straub, T, 2003, Heterochromatin dynamics, PLoS Bioi. I:EI4. Taroni, P., and Valentini, G., 2004, Fluorescence spectroscopy and imaging (non microscopic). Part I: Basic principles and techniques, In: Lasers and Current Optical Techniques in Biology (G. Palumbo and R. Pratesi, eds.), RSC-Royal Society of Chemistry, Cambridge, United Kingdom, pp. 259-290. Tsien, R.Y., 1998, The green fluorescent protein, Annu. Rev. Biochem. 67:509-544. Tsien, R.Y., 2003, Imagining imaging's future, Nat. Rev. Mol. Cell Bioi. 4:SS16-SS21. Tsien, R.Y., and Waggoner, A.S., 1995, Fluorophores for confocal microscopy. Photophysics and photochemistry, In: Handbook of Biological Confocal Microscopy (1.B. Pawley, ed.), Plenum Press, New York, pp. 267-279. Voura, E.B., laiswal, 1.K., Mattoussi, H., and Simon, S.M., 2004, Tracking metastatic tumor cell extravasation with quantum dot nanocrystals and fluorescence emission-scanning microscopy, Nat. Med. 10:993-998. Watson, A., Wu, X., and Bruchez, M., 2003, Lighting up cells with quantum dots, Biotechnology 34:296-303. Weber, G., 1960, Enumeration of components in complex systems by fluorescence spectrophotometry, Nature 190:27-29. Weber, G., and Teale, EW.l., 1958, Fluorescence excitation spectrum of organic compounds in solution, Trans. Faraday Soc. 54:640-648. Wennmalm, S., and Rigler, R., 1999, On death numbers and survival times of single dye molecules, J. Phys. Chem. B 103:2516-2519. White, 1., and Stelzer, E., 1999, Photobleaching GFP reveals protein dynamics inside live cells, Trends Cell Bioi. 9:61-65. Widengren, 1., and Rigler, R, 1996, Mechanisms of photobleaching investigated by fluorescence correlation spectroscopy, Bioimaging 4: 149-157. Wilson, T, and Sheppard, Cl.R, 1984, Theory and Practice of Scanning Optical Microscopy, Academic Press, London. Wu, X., Liu, H., Liu, 1., Haley, K.N., Treadway, 1.A., Larson, 1.P., Ge, N., Peale, E, and Bruchez, M.P., 2003, Immunofluorescent labeling of cancer marker Her2 and other cellular targets with semiconductor quantum dots, Nat. Biotechnol. 21 :41-46. Xie, X.S., and Trautman, 1.K., 1998. Optical studies of single molecules at room temperature, Annu. Rev. Phys. Chem. 49:441. Xu, C, 2002, Cross-section of fluorescent molecules in muitiphoton microscopy, In: Confocal and Two-Photon Microscopy: Foundations. Applications, and Advances (A. Diaspro, ed.), Wiley-Liss, New York, pp. 75-99. Zehetmayer, P., Hellerer, Th., Parbel, A., Scheer, H., and Zumbusch, A., 2002, Spectroscopy of single phycoerythrocyanin monomers: dark state identification and observation of energy transfer heterogeneities, Biophys. J. 83:407-415. Zhang, 1., Campbell, RE., Ting, A.Y., and Tsien, RY., 2002, Creating new fluorescent probes for cell biology, Nat. Rev. Mol. Cell Bioi. 3:906-918. Zhao, K.H., and Scheer, H., 1995, Type I and type II reversible photochemistry of phycoerythrocyanin a-subunit from Mastigocladus laminosus both involve Z, E isomerization of phycoviolobilin chromophore and are controlled by sulfhydryls in apoprotein, Biochem. Biophys. Acta 1228:244-253. Zimmer, M., 2002, Green fluorescent protein (GFP): Applications, structure, and related photophysical behavior, Chem. Rev. 102:759-781. Zondervan, R, Kulzer, E, Kol'chenko, M.A., and Orrit, M., 2004, Photobleaching of rhodamine 6G in poly(vinyl alcohol) at the ensemble and single-molecule levels, J. Phys. Chem. A 108:1657-1665.
40
Nonlinear (Harmonic Generation) Optical Microscopy Ping-Chin Cheng and C.K. Sun
INTRODUCTION In recent years, light microscopy, particularly fluorescence microscopy, has been extensively useful in the study ofliving cells and tissues. Although it has become an important tool in biological research, both single- (Sheppard and Shotton, 1997) or twophoton (Denk et al., 1990; Cheng et al., 1998, 2001) excitation schemes require that the specimen contain either intrinsic or extrinsic fluorescent probes. These probes include fluorescent dyes, fluorescent proteins, and quantum dots and common problems include probe penetration, probe toxicity, and photobleachingl damage (Konig, 1995; Cheng et ai., 2001a). To be useful, the fluorescent probes must usually be bound to specific biostructures or molecules, either by partition of the dye into various compartments, antigen-antibody reactions, affinity or site-specific binding of dye(s), or the transgenic expression of fluorescent and luminescent proteins. In addition, the probe may interact with the ionic environment to give a specific fluorescence signature. In all these cases, the fluorescence signals are related to the chemistry of the dye and the chemistry of the interaction between the dye and the cell or tissue or the genetic expression of the probe. Therefore, the term "chemical" and "biochemical" contrast is frequently used to describe the fluorescence imaging modality. On the other hand, nonlinear imaging modalities exist that create contrast based on the organization and orientation of nanostructures in the specimen, and these can often provide an alternative tool for studying the dynamics of cellular structures and functions. These nonlinear imaging modalities include second and third harmonic generation (SHG and THG), in which specific structural features of the optical configuration of the native specimen generate prompt signals. In this case, signal is generated by nonlinearity in the physical properties of the specimen (such as short-period modulations or discontinuities in its refractive index, RI) particularly those that occur at very high excitation intensity, such as that present at the focus point of an objective lens illuminated with high-intensity, femtosecond pulsed laser. Unlike fluorescence, these harmonic signals are generated with no time delay at all and emerge at wavelenghts that are exact integer submultiples of the excitation wavelength and traveling in the same direction (forward). Recent studies on man-made nanoperiodic structures (e.g., super-lattices) indicate a strong enhancement in SHG when high-intensity light is incident on non-centri-symmetric structures (Zhao et ai., 1999; Kao et ai., 2000; Sun et al., 2000). SHG occurs because the structure violates the condition of optical centro-symmetry. This sort of breakdown occurs in the inorganic crystals used for doubling the frequency of laser light (see Chapter
5, this volume). A typical application is the frequency doubling of 1064 nm emission from a semiconductor laser to produce the 532 nm light in the green-laser pointer. More generally, SHG occurs in many other structured samples, some of which are biological. A super-lattice structure is basically a one-dimensional nonlinear X(2) photonic crystal (where X(2) is the second-order nonlinear susceptibility), which is defined as a material with a periodicity in its second-order nonlinear dielectric properties (Berger, 1998). Strong SHG-enhancement has also been reported in one-dimensional and two-dimensional nonlinear photonic crystals (Broderick et ai., 2000; Dumeige et al., 2001). The mechanism is similar to the bandgap-resonant enhancement that occurs in common SHG-conversion crystals and, although SHG conversion efficiency is always highest near the nonlinear photonic bandgap, it does not vanish even when the illumination (pump) wavelength is far away from the spatial modulation period (the bandgap). Quasi-phasematching is an example of this, as the operating wavelength of the excitation is much shorter than the spatial modulation period of the nonlinear coefficient X(2). Recently, strong SHG has been reported in several types of biological material, mainly in orderly packed biomolecules or macromolecular structures (Chu et ai., 200 I; Cheng et ai., 2002; Sun et al., 2003). In contrast, no SHG occurs in amorphous materials, such as the almost randomly arranged macromolecules and other nanostructures that make up a cell. The enhancement of SHG by the nanophotonic crystals of the membrane protein bacteriorhodopsin (bR), has recently been demonstrated using hyper-Rayleigh scattering (Clays et ai., 2001). The bR protein forms a two-dimensional (2D) crystal in the purple membrane of Halobacterium salinarium (Birge et ai., 1990). This structure has an alternating change in second-order nonlinearity with a -5 nm period, causing it to act as a nonlinear photonic crystal (Berger, 1998). A number of other highly organized biological nanostructures have been reported that appear to break optical ccntro-symmetry and behave as SHG-active nonlinear photonic crystals. Such structures include stacked membranes such as those found in the myelin sheath, the endoplasmic reticulum (ER), the Golgi apparatus, and the grana in the chloroplast, microtubules, cellulosic microfibrils, collagen fibers, enamel prisms, bone matrix, starch granules, and mineral deposits in plants (Chu et al., 2001; Cheng et ai., 2002; Sun et ai., 2003, 2004). Similar to the backward SHG and THG detected in a waveguide, Gu and colleagues (1999) and Sun and co-workers (2005) reported strong backward SHG from microtubules, a fact that allows imaging the cytoskeleton and the mitotic spindle in living tissue. Collagen, the major protein of the extracellular matrix, is one of the best-known SHG structures in biology. The collagen mole-
Ping-Chin Cheng. State University of New York, Buffalo, New York 14260, and National University of Singapore, Singapore C.K. Sun· National Taiwan University, Taipei, Taiwan, Republic of China
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
703
704
Chapter 40 • P.-c. Cheng and C.K. Sun
cule is a long, stiff, triple-stranded helical structure, in which three collagen polypeptide chains, called a-chains, are wound around one another in a rope like superhelix. After being secreted into the extracellular space, these collagen molecules assemble into highorder polymers called collagen fibrils that are 10 to 300nm in diameter and many hundreds of micrometers long. Collagen fibrils often aggregate into larger, cable-like bundles, several micrometers in diameter, which can be seen in the light microscope as collagen fibers. Under electron microscopy, collagen fibrils have characteristic cross-striations every 67 nm, reflecting the regularly staggered packing of the individual collagen molecules in the fibril. This regularly staggered packing order provides the needed structural condition for efficient SHG (Williams et aI., 2005; Sun et aI., 2004). The SHG signal is created in proportion to the square of the instantaneous local intensity (Gannaway and Sheppard, 1978). As a result, like two-photon fluorescence (2PF) microscopy (Denk et aI., 1990; Cheng et aI., 1998), SHG provides intrinsic axial resolution. SHG microscopy was first demonstrated in studies of SHG photonic crystals (Gannaway and Sheppard, 1978), surfaces/interfaces (Shen, 1989), electric field distribution in semiconductors (Sun et aI., 2001), and was also shown to be present in studies of tissue polarity (Freund et aI., 1986; Guo et aI., 1997) and of membrane potentials in cells (Campagnola et al., 1999; Peleg et aI., 1999; Moreaux et aI., 2000b). Because most biological structures are not highly ordered, they are optically isotropic and do not produce any SHG signal. Only those few biological structures that are ordered or that involve some spatial organization that breaks the optical centro-symmetry can produce harmonic signals. In contrast to fluorescence processes that involve the excitation of the fluorescent molecule to an excited state having a finite lifetime, harmonically generated signals involve only virtual states that involve no time delay. The marked advantage of this virtual transition is the lack of energy deposition in the specimen. As a result, SHG produces no photodamage or bleaching, and like backscattered light, can be regarded as a truly "non-invasive" imaging modality. However, it is important to remember that the very high instantaneous power levels required to produce the effect may cause other novel and damaging nonlinear damage mechanisms to occur (see Chapter 38, this volume). As both harmonic generation and nonlinear absorption processes can occur simultaneously, it is also possible to produce photodamage due to absorption processes if absorbers are present. In fact, as some small amount of nonlinear absorption always occurs in biological specimens, the main advantages of SHG are not just that the signal generation process involves no energy deposition, but that it provides specific structural information. In contrast to the chemical specificity that characterizes fluorescence imaging, harmonic generation (SHG and THG) provides an imaging modality specific for structural configuration. Discontinuities in RI or the optical dispersion properties of biological tissues can generate third harmonic generation (THG) signals (Barad et aI., 1997; Muller et aI., 1998; Moreaux et aI., 2000a,b; Sun et aI., 2001) and the intensity of these signals is proportional to the third power of the illumination intensity. Using the same illumination wavelength, THG provides even better opticalsectioning resolution than SHG or 2PF, but is also more sensitive to changes in the intensity of the light in the focused spot, such as those caused by laser instability or by scattering or defocusing of the illumination. The THG can be used to study optical interfaces in the cell such as those at cell membranes or organelle surfaces. For example, the surface of the erythrocyte can generate significant THG (Sun et aI., 2004).
HARMONIC GENERATION Shortly after the first demonstration of the laser by Maiman in 1960, the next year Franken and co-workers discovered the process of SHG in man-made materials. This work is often taken as the beginning of the field of nonlinear optics. Nonlinear optical phenomena are "nonlinear" in the sense that they occur when the response of a specimen depends upon the strength of the optical electric field in a nonlinear manner. For example, SHG occurs as a result of that part of the atomic response to the oscillating field of the light that depends quadratically on the strength of this optical field. Consequently, the intensity of the signal generated by the SHG mechanism increases as the square of the intensity of the incident light. By the same token, THG signals vary with the cube of the intensity of the incident light. An image is a two-dimensional recording of the interaction between light and a specimen, and therefore is a representation of the optical properties of the specimen. In conventional optical imaging, contrast mechanisms include interactions such as absorption, reflection, scattering, and fluorescence, and the response recorded is linearly dependent on the intensity of the incident light. In the nonlinear optical domain, high-intensity light causes a variety of "anomalous" optical responses, and therefore the image contrast mechanism represents not only the differences in optical properties of the specimen, but also the result of the modification of those optical properties that depend on the intensity of the light in a nonlinear manner. In the case of conventional (linear) optics, a linear relationship exists between the electric field strength of the light and the induced polarization of the object. At relatively low incident intensity, the optical response can be approximated to be the first-order response as: Pet) = £oX(l)E(t)
(1)
where X(l) is the linear susceptibility, Pet) is the dipole moment per unit volume, or polarization of a material system, E(t) is the strength of an applied optical field, and (0 is the frequency.
Second Harmonic Generation In contrast, at high incident intensity, the nonlinear optical response can be described by: Pet) = EoX(1lE(t) + £OX(2)E(t)2 + £oX(3)E(t)3 + ...
== pOl(t) + p(2l(t) + p(3l(t) + . . .
(2)
where X(2 l is the second-order nonlinear susceptibility and X(3) is the third-order nonlinear susceptibility. In the above equation, because the fields are vectors, the nonlinear susceptibilities are tensors. As each atom acts as an oscillating dipole that radiates in a dipole radiation pattern, the radiation phase among the enormous number of atoms must be matched to induce constructive interference and thus nonlinear generation is allowed under phase-matching conditions (i.e., when the scattered light is in phase). This leads to the generation of radiation at the second harmonic frequency (half the wavelength of the illumination). However, this situation does not lead to the generation of electromagnetic radiation because its second time derivative vanishes and a static electric field is created within the nonlinear crystal. Second harmonic generation can also be considered as an interaction involving the exchange of photons between the various frequency components of the field. In SHG, two photons of frequency (0 are destroyed and one photon of frequency 2(0 is simultaneously created in a single quantum-mechanical process. The
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
•
705
•
Virtual Transitions
Real Transitions
FIGURE 40.1. Energy state diagram. Both SHG and THG involve virtual transitions in which no energy is absorbed by the specimen. In contrast, two-photon ftuorescence (2PF) involves the absorption of energy (real transitions) and excitation of molecules. The wavelength of SHG is half and THG is one third that of the incident wavelength, while 2PF has an emission wavelength more than half of the incident wavelength.
solid lines in Figure 40.1 represent the atomic ground states, and the dashed lines represent virtual levels. These levels are not energy eigenlevels of the free atom, but rather represent the combined energy of one of the energy eigenstates of the atom and of one or more photons of the radiation field. The fact that SHG vanishes in any material system that possesses centro-symmetry (i.e., one that has a center of inversion) can be explained by changing the sign of the applied electric field in Eq. 2. In a medium possessing inversion symmetry, the sign of the induced polarization must also change. Hence, relation 2 is replaced by (3)
By comparing Eq. 3 with Eq. 2, pet) must equal-Pet), and this can occur only if pet) vanishes identically. So X(2) is equal to zero for centro-symmetric media and no SHG signal is expected. Because the molecules making up most biological materials are oriented more or less "at random" (compared to a crystal, for example), they generate no SHG signals. The phase-matching condition in SHG is that the wave vector difference between input excitation light and output harmonic light is zero. Therefore, phase-matching is seriously affected by polarization, and SHG is sensiti ve to the angle between the polarization of the incident light and the symmetry condition of the material. Pol-dependent SHG can provide information about crystal orientation and imperfection, macromolecular structure and orientation, and regions where the centro-symmetry either breaks down, such as at surfaces and optical interfaces or where it is induced by organized, submicron structures. It can also be created by large localized residual electric fields, such as those across electrically polarized cell membranes.
Third Harmonic Generation Third harmonic generation involves a process whereby three photons of frequency (0 are destroyed and one photon of frequency 3405 nm). However, the THG signal generated by a 810 nm primary beam will fall at 270 nm, in the deep ultraviolet (UV) region, while threephoton fluorescence (3PF) will be found at >270nm. Consequently, the THG and 3PF signals will suffer from the high UV absorption of most biological specimens (Fig. 40.2; Cheng et aI., 2001a; Lin et aI., 2001), making signal detection difficult. In addition, one would need special (and expensive!) UV objectives, mirrors and photomultiplier tubes (PMT) in the detecting path. In contrast, the Cr:forsterite laser operates in the range of 1230 to 1270nm in the IR spectrum, making it an excellent light source for multi-modality microscopy (Bouma et aI., 1996; Chu et aI., 2002). For example, operating a Cr:forsterite laser at 1230nm allows SHG (615nm), THG (41Onm), 2PF (>615nm), and 3PF (>41Onm), all to fall within the visible spectrum. In addition, the lowest light attenuation in biological material is generally found
4
Ytterbium (1030nm)
Nd;glass (1064nm)
Laser System
Wavelength Range (nm)
Pulse Width (fs)
Repetition Rate (Hz)
Ti:Sa
700-980
100
1053-1064 1230-1270 1030
150 65 200
76-100 MHz 2 GHz, Chu et al., 2003b 70-150MHz 76-120MHz 50 MHz
Nd:glass Cr:forsterite Diode-pumped ytterbium
in the 1000 to 1300nm range (Fig. 40.2). By moving the excitation wavelength to 1230nm, not only the visible but also the NIR spectrum is open for signal recording. This could be important in imaging botanical specimens that have a high autofluorescence output over nearly the entire visible spectrum (see Fig. 20.2, this volume). The fact that autofluorescence diminishes at wavelengths longer than the fluorescence emission of chlorophylls makes NIR fluorescent probes, such as AlexaFluor 750, Cy5.5 and Cy7 (Molecular Probes Inc., Eugene, OR), very attractive. Development of endogenous fluorophore mutants (i.e., long-wavelength fluorescent proteins) having emission wavelengths longer than DsRed will be helpful for the study of botanical specimens. Liu and colleagues (2001) and Cheng and co-workers (200l) have demonstrated that a number of commonly used fluorescent probes can be efficiently excited at 1230nm by 2PF and 3PF. On the other hand, mode-locked femtosecond Nd:glass and diode-pumped ytterbium lasers provide an alternative choice. A typical femtosecond Nd:glass system offers a tuning range of 1053 to 1064nm, with a repetition rate of 70 to 150MHz and a 150fs pulse width (Time-Bandwidth Products, Zurich). This laser will produce SHG signal at 526 to 532 nm, which is at the second attenuation minimum of green botanical specimens (Fig. 40.2). The
- - Rice leaf ( Oryza sativa) - - Chicken dermis
11:5
CrF
(780nm)
(1230nml
3
1.0
~
~ .s:
2
:;:::
'iii
c
(])
0.5
(j)
0
(5
..c a..
500
1000
1500
Wavelength nm
2000
2500
FIGURE 40.2. Light attenuation spectra (absorption and scattering) in rice leaf (Oryza sativa) and chicken dermis. Note the operating ranges ofTi:Sa, ytterbium, Nd:glass, and Cr:Forsterite lasers. The light green lines indicate a typical emission wavelength of Cr:Forsterite (1230 nm) and its corresponding SHG (6l5nm) and THG (4IOnm). The blue lines represent a typical Ti:Sa emission wavelength (780nm) and the corresponding SHG (390nm) and THG (260nm) wavelengths. The purple lines indicate the emission line of Nd:glass and corresponding SHG (532 nm) and THG (352nm) lines. The dark green line indicates the emission line of ytterbium laser (1030nm). The curves labeled Sl336 and G8l98 represent the sensitivity of typical visible and NIR photodiode detectors.
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
THG is situated at 351 to 354 nm. Because both the excitation and emission wavelength are optimized at an attenuation minimum, using Nd:glass greatly increases the imaging depth in tissue, particularly green plant tissue. Using this laser system, one can obtain useful 3PF, 2PF, and SHG simultaneously and it has the added advantage that many commercially available IR objectives have reasonable transmittance at the 1064nm wavelength. On the other hand, the diode-pumped ytterbium laser operates at 1030nm, with SHG and THG at 515nm and 343nm, respectively. Figure 40.2 summarizes the operating wavelengths and the SHG and THG signals of the four mode-locked ultra-fast lasers in reference to the attenuation spectra of biological specimens and detector sensitivity. In the near UV wavelength region, attenuation is dominated by the strong absorption of common tissue constituents such as proteins and carbohydrates. For IR wavelengths longer than 1400 nm, however, the molecular resonance absorption of water starts to dominate the attenuation. As a result, biological specimens have a relatively transparent window between 350 to 1300nm. Although the Ti:Sa laser, operating at around nO-950nm, has been widely adopted as the preferred light source for multi-photon microscopy, one can easily see that the attenuation minimum is around 1000 to 1300 nm, not in the Ti:Sa range. In fact, this should not be surprising because scattering cross-sections become smaller as the wavelength increases. This plot is also in good agreement with previous measurements of the attenuation coefficients of biological materials such as human skin (Anderson and Parrish, 1981), maize stem (Cheng et al., 1998), and leaves (Lin et at., 2001). The advantage of the Ti:Sa laser is its large tuning range compared to the Cr:forsterite and Nd:glass lasers. In addition, while most objective lenses work well in the NIR spectrum (up to 900nm) and some IR objectives have reasonable transmission up to -1100 nm, obtaining reasonable transmission and optical correction in the l230nm range requires special objective lenses and other optical components (see Chapter 7, this volume).
707
How do the signals produced by these four lasers match the performance of the available detectors? Because the Ti:Sa wavelengths are well within in the sensitivity range of both high quantum efficiency (QE) silicon-based detectors and most PMT photocathodes, any scattered excitation illumination will be detected by these detectors and recorded as background. Although this reduces the signal-to-noise (SIN) ratio of fluorescence signals, it also affords the benefit of being able to detect the backscattered light (BSL) signal as an additional, non-damaging imaging modality. On the other hand, because the emission wavelengths of Nd:glass and Cr:forsterite lasers are beyond the sensitivity of most of the commonly used PMT and silicon detectors (Table 40.2), detector insensitivity guarantees a low background level from scattered light but sacrifices availability of BSL modality unless a special NIR detector (e.g., InGaAs photodiode, Edmund Scientific) is installed. Table 40.2 lists the sensitivity, the sensitivity range, and the wavelength of maximum sensitivity for each of the photodetectors commonly used in microscopy (see also Chapter 12, this volume). Because the peak intensity of the ultra-fast laser is so high, a dichroic beam-splitter and a barrier filter may not be sufficient to exclude the background signal caused by light leakage. This is a particular problem when collecting SHG and THG signals in the transmission mode where the excitation striking the emission filter may be 107x brighter than the SHG/THG signal. In this case, using detectors insensitive to the illumination wavelength can further limit the background signal level, improving the SIN of the image. Using a Cr:forsterite or Nd:glass laser at 1230nml1064nm for nonlinear microscopy allows one to fully utilize the transparent window in most biological specimens. At l230nm, Chu and colleagues found that the SHG and 2PF signal dropped by only I order of magnitude, when generated at a depth of 360 f..tm in a maize stem fixed in 70% ethanol (Chu et at., 2001). This is in good agreement with previous light attenuation measurements of maize stems (Cheng et at., 1998). Comparing the attenuation spectra of
TABLE 40.2. Characteristics of Photodetectors Used for SHG and THG Microscopy
Type SI S4 Sll S13 S20 S20 (extended-red multi-alkal) Bialkali Bialkali (green extended) Solar blind
Silicon photodiode InGaAs photo-diode
Photocathode Composition
Photoemission Threshold (nm)
Wavelength at Maximum Sensitivity (nm)
AgOCs SbCs, SbCs,
1100 680 700
800 400 440
2.3 50 80
0.4 16 22
SbNa2KCs SbNa,KCs
850 900
420 550
70 35
20 8
SbKCs SbKCs
630 700
400 440
90 100
28 28
CsTe
340 Spectral range
235
20 Photo sensitivity
10
320 (l90)-1100nm 900-1700nm
960nm l550nm
Radiant Sensitivity at Am" (mNW)
Quantum Efficiency at Am" (%)
(AfW) 0.33 (Ie = 633 nm)
0.9 (Ie = l300nm) 0.95 (Ie = 1550nm)
The short wavelength limit of the PMT is determined by the window material (lime glass, 300nm; borosilicate glass, 250nm; fused silica. 180nm. The short wavelength sensitivity cut-off of photodiode is also limited by the package window material; number in parentheses represents fused silica window. (Data adopted from Photonics, Brive, France and Hamamatsu Inc., Japan. QE data refer to the photocathode response only.)
708
Chapter 40 • P.-c. Cheng and C.K. Sun
Confocal aperture' IR cut-off
FIGURE 40.3. (A) Diagrammatic representation of a nonlinear laser-scanning microscope. The xy-beam scanner can be removed (by parking the xy-mirror at the center position) from the system to allow stage-scanning operation. An autocorrelator can be introduced into the beam to monitor the pulse width. A second harmonic generating crystal can also be introduced into the beam to allow the illumination to operate at 1.12. The spectrometer can also be connected to the epiillumination path to detect the backward SHG signal. CCD, charge-coupled device; *, optional.
_-=:'1":=_ t===F=:==
Backward detection Forward detecti:-o-n-tC:===OC::::==::J x;t;ge scanner
plant tissue with that of a chicken skin (Fig. 40.1), over the emission range of the Ti:Sa laser, one can expect significantly better depth penetration performance from Cr:forsteritelNd:glass/ytterbium laser sources in brain-slice research. This superior depth performance agrees well with previous studies comparing the penetration depth of optical coherence tomography (OCT) between 800nm and l300nm light (see Chapter 34, this volume; Bouma et ai., 1996). Moreover, a significant reduction in photo-induced damage was observed in plant materials when 2PF microscopy was conducted using a femtosecond Cr:forsterite laser rather than a femtosecond Ti:Sa laser, under similar illumination intensity and exposure conditions (Chen el ai. , 2002).
NONLINEAR OPTICAL MICROSCOPY SETUP Figure 40.3 shows a diagrammatic representation of a typical imaging system for multi-modal, nonlinear microscope using either Cr:forsterite, Nd:glass, ytterbium, or Ti:Sa lasers (Cheng et ai., 2001b). In our setup, the laser beam profile is first shaped with a spatial filter and collimated by a beam expander to fill the back aperture of the objective lens. The collimated beam was coupled into an Olympus IX7l microscope through the confocal scanning unit with a 45° dichroic beam-splitter (Chroma 970dcspxr) to reflect the IR and transmit visible light. The original aluminum coating on the xy-scanning mirrors was replaced with a silver coating to enhance IR reflectivity. A second beam path was constructed to bypass the scan mirrors for fixed-beam, stagescanning operation. The excitation IR light is focused onto the biological sample with a spot size close to diffraction limit using high-NA objectives [such as Olympus ULWD MIR 80x/0 .75; or UPlanApo water-immersion 60x/1.20j and the excited photoemission spectrum was collected using an opposing high-NA objective
(of the same type as the illuminating objective). The fundamental IR beam was removed with an infrared-blocking filter in conjunction with a dichroic beam-splitter (Chroma 970dcspxr). The collected light (visible and NIR) was then directed into a spectrometer (SpectraPro-2300i, Roper Scientific) and recorded by a Peltiercooled charge-coupled device (CCD) detector (DV42-0E, Andor Technology). Transmission detection is used because most SHG and THG is emitted in the forward direction (i.e., in the same direction as the pump beam; Cheng and Lin, 1990). I In addition, a mechanical scanning stage is included to permit stable raster scanning so that xyzlc-images can be obtained. Is these images, a detailed nonlinear spectrum is recorded at each pixel (xy-plane is defined as the plane perpendicular to the laser propagation direction). The signal spectrum is obtained using a scanning spectrometer equipped with an InGaAs photodiode (Edmund Scientific). When beam scanning, a flat mirror is placed in the back-focal plane of the collector objective to send the SHGITHG signal back up the optical path so that it can be de-scanned and detected by the PMTs (the same setup as the transmission confocal microscope method discussed in Chapter 8, this volume). To obtain an image with a set signal wavelength, we used a simple scanning-monochrometer method, adopted from astronomy. The actual system is shown in Figure 40.4. Although it is possible to select harmonic signals using narrow bandpass interference filters, it is important to take special precautions if the bandpass filter is used to isolate the SHG signal from multi-photon-excited fluorescence. In certain types of specimen, where the fluorescence emission wavelength generated by
1
Recent results have shown that, in certain biological structures, backward SHG signals can be even stronger than forward SHG signals (Sun et aI., 2005).
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
709
FIGURE 40.4. The nonlinear microscope setup in the laboratory of one of the authors (pee) at the Department of Biological Sciences, National University of Singapore (A, B). The system is based on a mode-locked, ultra-fast er:forsterite laser (65fs; 120MHz; tuning range, 1230-1270nm) (e) and an Olympus Fluroview FV300 confocal scanning unit (D). I, transmission detector fiber connection; 2, er:forsterite laser; 3, SH generator; 4, beam scanner; 5, autocorrelator; 6, transmission detector; 7, pump input for Cr:forsterite laser.
three-photon excitation coincides with the SHG wavelength, spectral contamination can be serious even when a narrow-bandpass SHG isolating filter is used . Figure 40.5(A) shows a fluorescence spectrum when LysoTracker Red (Molecular Probes, Inc. , L7528) is excited by Cr:forsterite laser at 1240 nm. If a biological specimen is stained with the dye and imaged simultaneously in SHG and fluorescence mode, the expected SHG signal is at 620nm and THG at 413 nm, but the multi-photon-excited fluorescence (both 2PF and 3PF) of LysoTracker Red extends from -580nm to 720nm. Therefore, even a very narrow bandpass filter centered at 620 nm will allow significant fluorescence contamination in the SHG signals, particularly if the SHG is relatively weak. The best way to obtain a "pure" SHG signal is to use a monochrometer. Figure 40.5(B) shows the emission spectrum of the red leaf of Euphorbia pulcherrima (poinsettia) excited by 1234 nm IR,
and the fluorescence emission consisting of 2PF and 3PF. Note the position of the SHG at 617 nm. No trace of SHG can be recognized in this spectrum because, in this case, the SHG is in the forward direction with respect to the illumination beam, while this spectrum was recorded from the same side as the incident illumination. Figure 40.5(C) shows the emission spectrum of the leaf of Zea mays (corn). The fluorescence emission is in the red region and a small peak at 617 nm is the SHG signal scattered back from the leaf tissue. Because the SHG signal propagates in the same direction as the illumination beam, placing detector on the far side of the specimen will ensure better signal strength. Figure 40.6 shows nonlinear absorption by a methanol extract from the yellow petals of Canna. Readers are also referred to the nonlinear absorption spectrum of the highly efficient APSS fluorophore [(Fig. 8.5(A»). Sun and colleagues (2003) have reported that using a longer
71 0
Chapter 40 • P.-c. Cheng and C.K. Sun
0.7
LysoTracker Red (L-7528) 594nm
12
• 0.6
~
'00 c
10
Q)
C Q)
'-' c '-'
0.5
3PF
~ Iii
8
Q)
'"~ 0
:!:
:;
::J
c.
4
0
500
600
700
800
900
Wavelength (nm)
Euphorbia pulcherrima Poinsettia red leaf Ex: 1234nm, Integration time: 9.6 sec
644nm
Q)
C CQ) '-'
'"~
12 10
u:::
400
500
600
700
40
800
900
686nm 35
Zea mays leaf Ex: 1234nm, Integration time: 1 sec
'0
30
...l
~ ~
'00 c
0.4
0.5
0.6
0.7
FIGURE 40.6. Nonlinear absorption of the methanol extract of the yellow petal of Canna. (Measurement was performed by Ms. Mei-Hsin Chen, Graduate Institute of Electro-Optical Engineering and Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan.)
Optically Active Structures in Plants
Wavelength (nm)
~
0.3
4
a
0'
0.2
OPTICAllY ACTIVE BIOLOGICAL STRUCTURES
6
2
:c (f)
0.1
8
0
::J
0.1
Input power (W )
14 ~
0.2
0.0 0.0
16
'00 c
0.3
:;
413nm (THG)
2 400
0.4
0
c.
6
u:::
Canna yellow
- - y=-O.005+0.96x - - y = -O.005+0.96x-l.37' 10--1·x-2 R-2 = 99.9% . SD = 0.004
617nm (SHG)
25
Q)
C 20 c o '00
.~ 15 w
________________
10+-~~~-.~~~~~~~~~~~~~~~~,
400
500
600
700
800
900
Wavelength (nm)
FIGURE 40.5. Two-photon fluorescence spectra of three different samples. (Top) LysoTracker Red (L-752, Molecular Probes), excitation at l240nm; (middle) red leaf of Euphorbia pulcherrima (excitation, l234nm); and (Bottom) leaf of Zea mays (excitation, l234nm), note the small SHG peak at 617 nm. The top two spectra show overlap between the 2PF emission and that from SHG.
wavelength excitation (l230nm vs. SOOnm) can minimize the twophoton autofluorescence signal, hence improve the SIN in harmonic generation images. However, it is not clear if this reduction in autofluorescence intensity corresponds to a lower nonlinear absorbance of the specimen or a reduction in the quantum yield of the fluorophore.
The cell wall of higher plants consists of macrofibrils, which are orderly bundles of cellulose microfibrils each with a diameter about 10 nm. Within the microfibrils are micelles, representing another degree of highly ordered crystalline structure (Peter et at., 1992). These photonic-crysta1-like structures produce the optical anisotropy that gives rise to SHG. Figure 40.7(A) shows the nonlinear emission spectra measured from the cell wall of a parenchyma cell in maize stem (Zea mays L.). Symmetric THG and SHG emission peaks are visible, centered at 410nm and 615 nm, with an intensity similar to or stronger than the twophoton autofluorescence centered at 6S0 nm. The strong THG signal is induced by the optical inhomogeneity within, and surrounding, the cell wall, while the SHG signal reflects the highly organized crystalline structures in the wall that break threedimensional (3D) optical centro-symmetry. Figure 40.7(B-D) shows the THG, SHG, and 2PF images made by detecting wavelengths corresponding to the peaks in the spectrum shown in Figure 40.7(A). The source of SHG is further confirmed by the strong SHG signal obtained from the stone cell of pear (Pyrus seratina R) (Fig. 40.S). The extensive secondary wall development of the sclerenchyma cell generates significant SHG signals. The starch granule, a highly birefringent structure, consists of crystalline amylopectin lamellae organized into effectively spherical blocklets and large concentric growth rings (Gallant et al., 1997) (Fig. 40.9). The crystalline lamellae in starch granules are believed to consist of the ordered, double-helical amylopectin side chains and are interleaved with more amorphous lamellae consisting of the amylopectin branching regions. The amylopectin sidechain clusters within the crystalline lamellae have varying sizes but, on average, are around 10nm wide by 9 to 10nm long (the length represents the thickness of the lamellae). These orderly nanolayers form the biophotonic structure, breaking the centrosymmetry and producing strong SHG. On the other hand, starch granules contain other, larger structures made up of crystalline hard shells and semi-crystalline soft shells having dimensions of hundreds of nanometers. These alternating crystalline and semi-
,.,,,
16
14 12
10
I~
8 6
:01-0--==4~00~5:::00=6::-0--=O=-7~O:0:8::;O=0=9:"0'---'0 wavelength(nm)
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
-
----
711
~~ .,
\,
-
~
T HG c
B
~
\ HG
, 0
.. 2PF
FIGURE 40.7. Non linear emission spectra measured from the cell wall of a parenchyma cell in maize stem (Zea mays). (A) The spectrum present a point in the image of (6); (6) THG image; (C) SHG; and (D) 2PF images.
crystalline rings have spatially-modulated nonlinear properties that could behave as 3D nonlinear photonic bandgap crystals (Berger, 1998) and may be responsible for the strong SHG observed. For example, the SHG signal from a potato (Solanum tuberosum L.) starch granule is so strong that is visible to the naked eye, even under ambient room light (Fig. 40.10). This unexpectedly strong SHG activity may be the result not only of the superhelical amylopectin nanostructure but also suggests that the specimen may be acting as a collection of 3D photonic bandgap crystals, the reciprocallattice basis vectors of which would be capable of producing SHO by meeting the non-collinear phase-matching condition. Depending on the illumination wavelength and the materials of which the specimen is composed, the spatial frequencies present in high-order structures from 200nm to 10 11m in size, can provide the non-collinear phase-matching base vector needed to produce SHG. The potato specimen is acting as an array of nonlinear, biophotonic bandgap crystals. Figure 40. l1(A-C) shows THO, SHG, and 2PF images, respectively. of a mesophyll cell of Commelina communis L. ; Figure 40.11 (D) is the corresponding three-channel, false-color
image. The THG image shows the interface signal between the chloroplast and the surrounding cytoplasm while the SHG reveals starch granules and possibly grana and thylakoid membranes in the chloroplasts and the 2PF results from chlorophyll autofluorescence. Enlarged images of individual chloroplasts are shown in Figure 40. 11 (E-H). The distribution of chlorophyll inside the chloroplasts is shown by the 2PF signals while THO signal provides information on various sub-organelle interfaces. SHG signal, on the other hand, indicates the presence of nano-organi zed biophotonic structures in the chloroplast. By matching the SHG image with transmission electron microscope (TEM) images of similar specimens, it has been possible to conclude that the SHG signals are produced by stacked thylakoid membranes present in the grana (crescent shaped) and by highly birefringent starch granules (oval or round shaped) in the chloroplasts (Gunning and Steer, 1996). The stacked thylakoid membranes of the grana and the orderly deposits of amylopectin in the starch granules provide the structural requirement for efficient SHG, again resembling the behavior of photonic crystals.
7000r-------------------------------. Stone cell of pear THG Ex: 1230nm 6000 ~
5000
SHG
.~ 4000 Ql E 3000 2000 1000
Oc=~~==========~==========~
400
500
600
700
800
Wavelength(nm)
FIGURE 40.8. Spectrum of the stone cell of a pear frui t (Pyrus sero/ina R.) recorded with O.I -s integration time.
FIGURE 40.9. Conventional polarization microscopy of maize starch granules. The image was taken with an oblique-illuminated microscope equipped with cross-polarizer. Note the strong birefringence of the starch granules.
712
Chapter 40 • P.-c. Cheng and C.K. Sun
16000
Potato
SHG
14000 >.12000
~ 10000 ~ 8000 6000 4000 Cr:forsterite laser
_
B
2000 0
SHG Potato slice
c
400
500 600 700 Wavelength(nm)
800
FIGURE 40.10. Photograph showing the red SHG signal generated from a thin slice of potato tuber (blue arrow) by a focused laser beam at a wavelength of l230nm and projected onto a white paper (orange arrow) (A). As this photograph was taken with a conventional camera. which is insensitive to 1230nm light, the small red spot on the potato slice is the scattered SHG signal. The corresponding setup is schematically shown in (B). The emission spectrum is shown in (C).
Plant cell
FIGURE 40.11. Harmonic and fluorescence images of a mesophyll cell of Commelina communis L. (A) THG image; (B) SHG image; (C) 2PF image; (D) falsecolor image of the combination of THG-SHG and 2PF. (E-H) High magnification images of chloroplasts showing strong SHG generation from starch granules (s) and possible grana (G). Ex, 1230nm.
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
5000 4000 ~
·05 c
Rice leaf epidermal call
THG
3000
~
(1230nm)
SHG
1PF .j~V)
2000
~ ___ --··-(780~'mr/
0 400
A
Maize stem parenchyma
500
·05 c
3000
E
2000
Q)
2PF
"
"
1000
THG
4000
Q)
E
5000
713
600
,. Ll,
1000 0 400
700
Wavelength (nm)
B
500
600
700
Wavelength (nm)
FIGURE 40.12. Nonlinear emission spectra from the cell wall of (A) an epidermal cell in rice leaf and (B) parenchyma cells in maize stem with three differ· ent sources: UV light (dotted line), 780nm pulsed signal (dashed line), and 1230nm pulsed laser (solid line). The suppression of background 2PF and the sym· metric peaks of THG and SHG in 1230nm excitata spectra are evident.
The tumbling movement of chloroplasts in cytoplasm results in changes in the apparent periodicity of the thylakoid membrane stacks along the direction of the illuminating beam. This tumbling therefore changes the SHG efficiency, a fact that can be used to study chloroplast tumbling, a phenomenon that is difficult to study by other means. Compared with the emission spectra excited by a femtosecond Ti:Sa laser in these plant specimens, the intrinsic 2PF signals are generally reduced by using a longer wavelength as the pumping source (Chu et aI., 2001). As an example, Figure 40.12 shows a quantitative comparison between the emission spectra excited by a 150mW 1230nm Cr:forsterite femtosecond laser (l20fs pulse width, solid line), a 50mW 780nm Ti:Sa femtosecond laser (l20fs pulse width, dashed line), and a 0.45mW 365nm UV light source from a frequency-doubled Ti:Sa laser (dotted line). The emission spectra from the cell wall of rice leaf (Oryza sativa) epidermis and of the parenchyma cells in a maize stem are shown in Figure 40.12(A,B), respectively. Broad autofluorescence covering the whole visible and NIR region is evident with both UV (singlephoton fluorescence) and 780nm excitation (2PF); whereas with 1230 nm excitation, only weak residual 2PF and efficient harmonic generation were observed. With background autofluorescence suppressed by using a longer excitation wavelength, the whole visible and NIR region is available for detecting signals from specially designed, multi-photon excited fluorescence dyes that can be used to label different functional molecules (Cheng et at., 1998). Furthermore, as there is no energy deposition at all during harmonic generation processes, no photodamage effect is expected with SHG and THG. Thus, with efficient SHG and THG signals, along with the reduced but still-measurable intrinsic 2PF signals, the longer wavelength light source appears to be a better choice for intrinsic harmonic generation imaging because it leaves a wider spectrum available for extrinsic multi-photon dye labeling. Figure 40.13 compares the emission spectra of maize leaf and stem (epidermal cells) when excited by 1230nm, 780nm, and 400nm. Figure 40.14 shows the specific XYA-images corresponding to THG, SHG, and 2PF in the parenchyma cells of a maize stem, taken at a depth of 110 J.!m from the sample surface. The total sample thickness was around 500 J.!m. With an average power of
100 m W before entering the sample and a focused spot diameter of -1.3 J.!m, the intensity at focus ranges from 9 to 50 X 1010 Wlcm,2 depending on the focal depth inside the sample. As expected, strong SHG and THG can be observed in the cell wall. THG shows the longitudinal cell walls in the center of the image as well as the transverse cell walls of several adjacent cells [Fig. 40. 14(A)], This demonstrates the ability of THG to pick up the outline of the whole cell. SHG may show the relative position of the secondary walls (arrows). In regions with extensive secondary wall growth, the separation of the secondary wall can be clearly observed in Figure 40. 14(B ) and the 2PF signal indicates the distribution of fluorescent molecules [Fig. 40.14(C)]. By comparing different images made using different modalities, one can image the relation-
FIGURE 40.13. Normalized emission spectra of maize tissue under different types of excitation. IPF at 400nm (blue). Two-photon excited spectra generated at 1234 nm from a leaf (green) lacks the hump due to fluorescence of the cell wall found in stems (magenta). It also lacks the SHG peak at 617 nm, probably generated by the secondary cell walls in the stem sample. Two-photon excitation at 780 nm produces fluorescence (maroon) that peaks at a shorter wavelength than the fluorescence excited by the other modes.
714
Chapter 40 • P.-c. Cheng and C.K. Sun
FIGURE 40.14. Optical section obtained near the longitudinal wall (LW) of a parenchyma cell in maize stem. Note the strong THO from the radial wall (RW) (A), the SHO (B), 2PF (C), and combination ofTHO-SHO-2PF (D). Arrows: possible secondary wall material. The image was obtained IIO!lm below the surface of the specimen.
ships between structural (harmonic generations) and chemical information (two- and multi-photon fluorescence) in biological tissues. A penetration depth of more than SOO 11m has been achieved in the maize stem sample with 1230nm excitation (Chu et aI., 2001). Figure 40.1S shows paradermal optical sections of the adaxial surface of rice (Oryza sativa L.) leaf. THG provides structural interfaces, such as the papillae from the cuticular layer and the cell wall of bulliform cells [Fig. 40.1S(B)]. As expected, SHG reflects biophotonic structures including the cuticular papillae and longitudinal cell walls [Fig. 40.1S(B)], due to the orderly arrangement of cutin, waxes, and cellulose micro fibrils , respectively. 2PF, on the other hand, picks up the fluorescent chromophores [Fig. 40.1S(C)]. Figure 40.16 shows a through-focus series (Sl1m step) of an area on the surface of rice leaf where silica cells can be found (Hodson and Sangster, 1989; Cheng 1987, Cheng et aI., 1990). Note the outline of the silica cell in the THG images and the high SHG signals from the silica deposits. A strong SHG signal can also be obtained from the silica wall of a diatom.
Optically Active Structures in Animal Tissues There are many structures in animal tissue that are also good candidates to produce a strong biophotonic effect (Table 40.3). For
example, the sarcomeres in skeletal muscles are composed of crystalline myosin and actin nanofilaments, with periods of 40 and 20nm, respectively, that fall into the spatial range required for strong SHG activity. Figure 40.17 shows longitudinally sectioned xy-images obtained from the somites of a zebrafish embryo. The cardiac muscle fiber produces intense SHG signal [green, FigAO.17(A-D)] and the surfaces of the red blood cells (RBC) produce intense THG signal [blue, Fig. 40.17(C,D)]. Figure 40.18 shows a two-channel harmonic image of somites in a zebrafish embryo. At low magnification, the somites are seen to be separated by clefts [Cl, Fig. 40.18(A)]. At higher magnification [Fig. 40.18(B,C)]. the green SHG image of the sarcomeres (s: between the two arrows) indicate that the strong blue THG signal in Figure 40.18(A) is probably generated by the optical discontinuity occurring between the somites. The SHG intensity difference in individual sarcomeres [Fig 40.18(C), between arrows] reflects differences in the spatial packing of Factin and myosin in dark bands and light bands. The strong SHG activity from the actin/myosin complex can make harmonic imaging a useful tool for the study of muscle cell dynamics as the arrangement changes during muscle contraction and relaxation. Microtubule bundles are birefringent structures that can be studied in living cells using polarization microscopy. Recent
FIGURE 40.15. Through-focus, three-channel harmonic fluorescence images of the adaxial surface of rice leaf. (A) Near surface, (B) epidermal cells, (el epidermal-mesophyll interface, (D) mesophyll cells. Note the strong 2PF emission from the chlorophyll of the me sophy II cells. White arrows: cuticular papillae.
Nonlinear (Harmonic Ceneration) Optical Microscopy • Chapter 40
715
FICURE 40.16. Through-focus, two·channel images of the surface of rice leaf, showing a silica cell. As it can be seen in the focal series (A-D), the SHG signal is generated by the bulk of the phytolith (Si), not by the surface of the silica deposit.
advances in dynamic polarization microscopy utilizing tunable liquid-crystal polarizers now allow one to study cytoskeleton dynamics and spindle behavior. The technique not only provides information on the retardance of the birefringent structure, but can also provide data of the molecul ar orientation of the structures (s-axis) (Oldenbourg and Mei, 1995). Figure 40.19 shows a typical example in which polarization microscopy has been used to determine the orientation of the collagen fiber scaffold in an engineered tissue. Polarization microscopy can also allow one to directly visualize the spindle during cell division under low light conditions. Figure 40.20 shows the spindle of a fertilized oocyte during a cloning operation. Microtubules have also been shown to produce strong SHG signal in both the forward and the backward configurations (Sun et aI., 2005). Figure 40.21 demonstrates the specificity of the SHG signal from the spindle apparatus of a developing zebrafish embryo. The interface between the cell membrane and the surrounding aqueous medium produces a strong THG signal (blue) that provides the
TABLE 40.3. Biological Structures that Produce Harmonic Signals Structures Microtubules Microfilaments Spindle Collagen fi ber Elastic fiber Cuticle Cuticular wax Sarcomere Grana and thylakoid Si02 deposit Starch granules Cell wall Bone matrix Dentine Enamel
Birefringence
SHG
+ + + + + + + +
+ + + + +
?
+ + + + + +
? ?
+ + + + + + + +
716
Chapter 40 • P.-C Cheng and CK. Sun
FIG URE 40.17. Two-channel, harmonic images of developing heart in a zebrafish embryo. The cardiac muscle fiber produces intense SHG signal (green, A-D). The red blood cells (RBC) produce intense THG signal (blue, e and D).
FIGURE 40.18. Two-channel harmonic image of somites in a zebrafish embryo. (A) Low magnification view of the somites separated by clefts (el); SHG, green; THG, blue. (B) SHG image showing developing myofibrils. Note: individual sarcomeres are clearly visible. (e) Higher magnification, SHG image of the sarcomere (s: between the two arrows) indicating that the strong SHG signal is likely generated from the dark band of sarcomere.
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
717
FIGURE 40.19. Polarization microscopy of man-made collagen fiber. The image demonstrates the orientation of the collagen fiber imaged by using a Polscope. The color wheel indicates the orientation of the s-axis of the collagen fiber. (Image courtesy of Prof. Hanry Yu, Department of Physiology, National University of Singapore.)
general outline of the embryo. Figure 40.20 is a set of high magnification, polarization microscope images showing a cell in various stages of cell division. As multi-harmonic imaging (SHG and THG) allows deeptissue imaging of biological specimens, it is suitable in developmental biology where larger specimens are commonly encountered. The use of a high-repetition-rate Ti:Sa laser offers the possibility of real-time, harmonic imaging of biological specimens (Chu et aI., 2003a; Sun et aI., 2002). Collagen fibers are abundant in animal tissue, and are both birefringent and capable of producing SHG signals. SHG is now often used to visualize the orientation-dependent properties of connective tissue and the extracellular matrix (Roth and Freund, 1979, 1981; Freund et al., 1986; Guo et al., 1997; Stoller et aI., 2002a,b; Zipfel et aI., 2003; Williams et aI., 2005).
Polarization Dependence of Second Harmonic Generation SHG signal strength from biophotonic structures varies according to the relative orientation between the beam and the organized
structure. This allows one to study the orientation of a structure using SHG with controlled illumination polarization. For example, the dumbbell-shaped silica deposits on rice and maize leaves produce intense SHG signals. By varying the incident light polarization (Fig. 40.23), the concentrically deposited silica layers in the lumen of silica cells (Hodson and Sangster, 1989) produces SHG images that depend on the orientation of the illumination polarization. The polarization direction of the excitation is marked in the LR comer of the left-hand image in each row. The white outlines in Figure 40.23(A) demarcate the locations of two dumbbell-shaped silica cells. In contrast, no polarization dependency is evident in THG images. It is possible to isolate the epidermal cuticle with the attached phytolith (silica deposits) by ZnCJz-HCl treatment. The fact that images of these isolated, silica deposits reveal little polarization dependency in the SHG signal (Fig. 40.24) suggests that the organic matrix on which the silica is deposited is the source of the photonic activity in the phytolith. Polarization-dependent contrast can also be used to study the orientation of collagen fibers in tissue engineering research (Freund et aI., 1986; Stroller et aI., 2002a,b; Chu et aI., 2004).
FIGURE 40.20. The spindle apparatus in the first division of a mouse zygote imaged by polarization microscopy. [The image was obtained by Ms. Gina Chen of Ming-mei Technology, Taipei, using a Spindleview dynamic polarization microscope (Cambridge Research Instrument, Cambridge, MA).]
718
Chapter 40 • P.-c. Cheng and C.K. Sun
FIGURE 40.21. Dual-channel. harmonic image of a zebrafish embryo showing the cells and the embryo outlined by THG (blue) and a mitotic spindle (green, arrows) in SHG. (A) Overall view of the embryo; (8) part of the embryo showing cells in telophase (yellow arrow) and anaphase (green arrow).
FIGURE 40.22. High magnification views of various stages in the formation of the spindle in a zebrafish embryo. (A) Telophase, (8) anaphase, (C) late anaphase, and (D) after cytokinesis, viewed in SHG (green) and THG (blue).
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
18000
J
16000
THG 5000
~12000
>.
1i5 4000
'00
10000
E
8000
K
6000
SHG
14000
a5
719
c
Q)
E 3000
6000 2000 4000 2000
300
THG 400
SHG 2PF
1000 500
600
700
800
900
Wavelength (nm)
300
400
500
600
700
800
900
Wavelength (nm)
FIGURE 40.23. Polarization dependence of the SHG signal. (A-C) Ninety·degree polarization; (D-F) 45° polarization; and (G-I) 60° polarization. Note that the change of illumination polarization has little effect to the THG image, but the SHG signal does show polarization dependence. The spectra at the bottom of this figure (J, K) were obtained at the poin ts labeled in (G).
SUMMARY Highly organized nanoperiodic structures in biological samples exhibit strong SHG activity resembling that of nonlinear photonic crystals and thus they can be treated as nonlinear biophotonic crystals . Many biologicaL structures, such as microfibrils in cell walls, alternating crystalline Lamellae in starch granules, cuticular papillae on the leaf surface, crystalline myosin and actin nanofilaments in the myofibrils of skeletal and cardiac muscle, thylakoid membranes and grana in chloroplasts, and microtubules in both the cytoskeleton and the mitotic apparatus, all exhibit strong biopho-
tonic effects. Many of the birefringent structures found in biological specimens also exhibit SHG properties (Table 40.3). The relative locations and orientation of these biophotonic structures can be examined using SHG microscopy, while optical interfaces and the functional molecules can be separately located by THG and 2PF contrast. In contrast to single- and multi-photon absorption, harmonic generation involves only virtual states and does not involve energy deposition. The harmonic signals allow 3D structural visualization with minimal or no additional preparation of the sample. Meanwhile, 2PF imaging modes can be added to monitor multiple molecular probes in living cells and tissues, such as those
720
Chapter 40 • P.-c. Cheng and C.K. Sun
FIGURE 40.24. Isolated phytoliths, illuminated using different polarization directions. Note that the SHG (green) and THG (blue) signals vary little with different illumination conditions. The THG patch (arrow) in the 45° image may be the result of focal drift bringing the underlying cuticular surface into focus.
composed of transformed cells or taken from transgenic organisms. Multi-modal microscopy can provide a powerful tool for investigating the dynamics of structure-function relationships at the molecular and subcellular levels.
ACKNOWLEDGMENTS Most of the images presented in this chapter come from research performed at the Graduate Institute of Electro-Optical Engineering and Department of Electrical Engineering. National Taiwan University, Taipei, Taiwan, by a number of researchers: I. H. Chen, T.-M. Liu, Mei-Hsin Chen, Szu-Yu Chen, and Shi-Wei Chu. Most of the biological work was done by Huai-Jen Tsai and Chung-Yung Lin of the Graduate Institute of Molecular and Cell Biology, National Taiwan University, and B.-L. Lin, S.-P. Lee of Molecular and Cell Biology Division, Development Center for Biotechnology, Taipei. The transmission spectroscopy was done by M.-X. Kuo, D.-J. Lin, and H.-L. Liu. Thanks are also due to Nick White, Sir William Dunn School of Pathology, University of Oxford for his help editing the proofs.
REFERENCES Anderson, R.R., and Parrish, J.A., 1981, The optics of human skin, J. Invest. Dermato!' 77: 13-19. Barad, Y., Eisenberg, E., Horowitz, M., and Silberberg, Y., 1997, Nonlinear scanning laser microscopy by third harmonic generation, App!. Phys. Lett. 70:922-924. Berger, Y., 1998, Nonlinear photonic crystals, Phys. Rev. Lett. 81:4136-4139. Birge, R.R., Fleitz, P.A., Lawrence, A.F, Masthay, M.A., and Zhang, C.F, 1990, Nonlinear optical properties of bacteriorhodopsin: Assignment of second order hyperpolarizabilities of randomly oriented systems using two-photon spectroscopy, Mol. Cryst. Liq. Cryst. 189:107-122. Bouma, B.E., Tearney. G.l., Bilinsky, I.P., Golubovic, B., and Fujimoto, 1.G., 1996, Self-phase-modulated Kerr-lens mode-locked Cr:Forsterite laser source for optical coherence tomography, Opt. Lett. 21:1839. Broderick, N.G.R., Ross, G.W., Offerhaus, H.L., Richardson, D.J., and Hanna, D.C., 2000, Hexagonally poled lithium niobate: A two-dimensional nonlinear photonic crystal, Phys. Rev. Lett. 84:4345-4348. Campagnola, P.J., Wei, M.-D., Lewis, A., and Loew, L.M., 1999, High resolution nonlinear optical imaging of live cells by second harmonc generation, Biophys. J. 77:3341-3349.
Chen, I.-H., Chu, S.-w., Sun, C.-K., Cheng, P.c., and Lin, B.L., 2002, Wavelength dependent cell damages in multi-photon confocal microscopy, Opt. Quantum Electron. 34:1251-1266. Cheng, P.c., 1987, Sample preparation for X-ray imaging and examples ofbiological X-ray images, In: X-Ray Microscopy-Instrumentation and Biological Applications (P.C. Cheng and G.J. Jan, eds.), Springer-Verlag, Berlin, pp. 289-310. Cheng, P.c., and Lin, T.H., 1990, The use of computer-controlled substage folding optics to enhance signal strength in fluorescent confocal microscopy, Trans. Royal Microscop. Soc. 1:459-462. Cheng, P.C., Lin, B.L., Kao, FJ., Gu, M., Xu, M.-G., Gan, X., Huang, M.-K., and Wang, Y.-S., 2001a, Multi-photon fluorescence microscopy - The response of plant cells to high intensity illumination, Micron 32:661-669. Cheng, P.C., Newberry, S.P., Kim, H.G., Wittman, M.D., and Hwang, I.-S., 1990, X-ray contact microradiography and shadow projection microscopy, In: Modern Microscopy (P. Duke and A. Michette, eds.), Plenum Press, New York, pp. 87-117. Cheng, P.c., Pan, S.J., Shih, A., Kim, K.S., Liou, W.S., and Park, M.S., 1998, Highly efficient upconverters for multi photon fluorescence microscopy, J. Microsc. 189:199-212. Cheng, P.c., Sun, C.-K., Kao, F.-J., Lin, B.-L., and Chu, S.-W., 2001b, Nonlinear multi-modality spectro-microscopy: Multiphoton fluorescence, SHG and THG of biological specimen, SPIE Proc. 4262:98-103. Cheng, P.c., Sun, c.K., Lin, B.L., and Chu, S.W., 2002, Bio-photonic crystal: SHG imaging. Maize Genetics Cooperation Newsletter 76:8-9. Chu, S.w., Chen, I.S., Li, T.M., Lin, B.L., Cheng, P.C., and Sun, C.K., 2001, Multi-modality nonlinear spectral microscopy based on a femtosecond Cr:Forsterite laser, Opt. Lett. 26: 1909-1911. Chu, S.-w., Chen, I.-S., Liu, T.-M., Sun, C.-K., Lee, S.-P., Lin, B.-L., Cheng, P'-C., Kuo, M.-X., Lin, D.-J., and Liu, H.-L., 2002, Nonlinear bio-photonic crystal effects revealed with multi-modal nonlinear microscopy, J. Microsc. 208: 190-200. Chu, S.-w., Chen, S.-Y., Tsai, T.-H., Liu, T.-M., Lin, c.-Y., Tsai, H.J., and Sun, C.-K., 2003a, In vivo developmental biology study using noninvasive multi-harmonic generation microscopy, Opt. Express 11:3093-3099. Chu, S.w., Liu, T.M., Sun, c.K., Lin, c.Y., and Tsai, H.J., 2003b, Real-time second-harmonic-generation microscopy on a 2-GHz repetition rate Ti:Sapphire laser, Opt. Express 8:933-938. Chu, S.-W., Chen, S.-Y., Chern, G.-W., Tsai, T.-H., Chen, Y.-C., Lin, B.-L., and Sun, c.-K., 2004, Studies of (2)/(3) tensors in submicron-scaled biotissues by polarization harmonics optical microscopy, Biophys. J. 86:3914-3922. Clays, K., Elshocht, S.Y., Chi, M., Lepoudre, E., and Persoons, A., 2001, Bacteriorhodopsin: A natural, efficient (nonlinear) photonic crystal, J. Opt. Soc. Am. B 18:1474-1482. Denk, W., Strickler, J.H., and Webb, W.W., 1990, Two-photon laser scanning fluorescence microscopy, Science 248:73-76.
Nonlinear (Harmonic Generation) Optical Microscopy • Chapter 40
Dumeige, Y., Vidakovic, P., Sauvage, S., Levenson, lA., Sibilla, c., Centini, M., D' Aguanno, G., and Scalora, M., 2001, Enhancement of secondharmonic generation in a one-dimensional semiconductor photonic band gap, Appl. Phys. Lett. 78:3021-3023. Freund, I., Deutsch, M., and Sprecher, A., 1986, Connective tissue polarity: Optical second-harmonic microscopy, crossed-beam summation, and small-angle scattering in rat-tail tendon, Biophys. 1.50:693-712. Gallant, D.J., Bouchet, B., and Baldwin, P.M., 1997, Microscopy of starch: Evidence of a new level of granule organization, Carbohydrate Polym. 32:177-191. Gannaway, J.N., and Sheppard, CJ.R, 1978, Second-harmonic imaging in the scanning optical microscope, Opt. Quantum Electron. 10:435439. Gunning, B., and Steer, M., 1996, Plant Cell Biology: Structure and Function, Jones and Bartlett Publishers, Sudbury, Massachusetts, p. 21. Gu, X., Makarov, M., Ding, Y.J., Khurgin, J.B., and Risk, WP., 1999, Backward second-harmonic and third-harmonic generation in a periodically poled potassium titanyl phosphate waveguide, Opt. Lett. 24: 127. Guo, Y., Ho, P.P., Tirksliunas, A., Liu, E, and Alfano, RR, 1997, Optical harmonic generation from animal tissues by the use of picosecond and femtosecond laser pulses, Opt. Lett. 22:1323-1325. Hodson, MJ., and Sangster, A.G., 1989, Silica deposition in the inflorescence bracts of wheat (Triticum aestivum). II X-ray microanalysis and backscattered electron imaging, Can. 1. Botany 67:281-287. Kao, EJ., Huang, M.K, Wang, Y.S., Hung, S.L., Lee, M.K, Sun, C.K, and Cheng, P.c., 2000, Two-photon optical-beam-induced current microscopy of indium gallium nitride light emitting diodes, SPIE Proc. 4082:92-98. Konig, K.H., Liang, M., Berns, W., and Tromberg, BJ., 1995, Cell damage by near-IR microbeams, Nature 377:20-21. Lin, B.-L., Cheng, P.C., and Sun, C.-K, 2001, Optical density of leaf, Maize Genetics Cooperation Newsletter, 75:61-62. Liu, T.M., Chu, S.W, Sun, C.K, Lin, B.L., Cheng, P.c., and Johnson, I., 2001, Multiphoton confocal microscopy using a femtosecond Cr:Forsterite laser, Scanning 23:249-254. Moreaux, E, Sandre, 0., and Mertz, J., 2000a, Membrane imaging by second-harmonic generation microscopy, 1. Opt. Soc. Am. B. 17:16851694. Moreaux, L., Sandre, 0., Blanchard-Desce, M., and Mertz, J., 2000b, Membrane imaging by simultaneous second-harmonic generation and twophoton microscopy, Opt. Lett. 25:320-322. Muller, M., Squier, J., Wilson, KR., and Brakenhoff, G.J., 1998, 3D microscopy of transparent objects using third-harmonic generation, 1. Microsc. 191:266-274. Oldenbourg, R., and Mei, G., 1995, A new polarized light microscope with precision universal compensator, 1. Microsc. 180: 140-147.
721
Peleg, G., Lewis, A., Linial, M., and Loew, L.M., 1999, Nonlinear optical measurement of membrane potential around single molecules at selected cellular sites, Proc. Natl. Acad. Sci. USA 96:6700-6704. Peter, H.R, Ray, EE., and Susan, E.E., 1992, Biology of Plants, Worth Publishers, New York, p. 36. Roth, S., and Freund, I., 1979, Second harmonic-generation in collagen, 1. Chern. Phys. 70:1637-1643. Roth, S., and Freund, I., 1981, Optical second-harmonic scattering in rat-tail tendon, Biopolymers 20:1271-1290. Shen, Y.R, 1989, Surface properties probed by 2nd harmonic and sum frequency generation, Nature 337:519-525. Sheppard, C.J.R, and Shotton, D.M., 1997, Confocal Laser Scanning Microscopy, BIOS Scientific Publisher, Oxford, United Kingdom. Stoller, P., Kim, B.M., Rubenchik, A.M., Reiser, K.M., and Da Silva, L.B., 2002a, Polarization-dependent optical second-harmonic imaging of a rattail tendon, 1. Biomed. Opt. 7:205-214. Stoller, P., Reiser, KM., Celliers, P.M., and Rubenchik, A.M., 2002b, Polarization-modulated second harmonic generation in collagen, Biophys. 1. 82:3330-3342. Sun, C.K, 2005, Higher Harmonic generation biopsy, Focus on Microscopy 2005, Jena, Germany. Sun, C.-K, Chen, C.-C., Chu, S.-W, Tsai, T.-H., Chen, Y.-c., and Lin, B.-L., 2003, Multiharmonic generation biopsy of skin, Opt. Lett. 28:2488-2490. Sun, C.-K, Chu, S.-W., and Liu, T.-M., 2002, High frame-rate secondharmonic generation micoscopy based on a 2 GHz Ti: sapphire laser, Proc. OSA paper TuH4. Sun, C.-K, Chu, S.-W, Chen, S.-y', Tsai, T.-H., Liu, T.-M., Lin, C.-Y., and Tsai, H.-J., 2004, Higher harmonic generation microscopy for developmental biology, 1. Struct. Bioi. 147:19-30. Sun C.-K., Chu, S.-W, Tai, S.P., Keller, S., Abare, A., Mishira, U.K., and DenBaars, S.P., 2001, Mapping piezoelectric-field distribution in gallium nitride with scanning second-harmonic generation microscopy, Scanning 23: 182-192. Sun, C.-K., Chu, S.-W., Tai, S.-P., Keller, S., Mishra, U.K, and DenBaars, S.P., 2000, Scanning second-harmonic-generation and third-harmonic-generation microscopy of GaN, Appl. Phys. Lett. 77:2331-2333. Williams, RM., Zipfel, WR, and Webb, WW, 2005, Interpreting secondharmonic generation images of collagen I fibrils, Biophys. 1. 88:13771386. Zhao, T., Chen, Z.-H., Chen, E, Shi, W-S., Lu, H.-B., and Yang, G.-Z., 1999, Enhancement of second-harmonic generation in BaTi0 3/SrTi0 3 superlattices, Phys. Rev. B 60:1697-1700. Zipfel, WR., Williams, R.M., Christie, R, Nikitin, A.Y., Hyman, B.T., and Webb, W.W., 2003, Live tissue intrinsic emission microscopy using multiphoton excited intrinsic fluorescence and second harmonic generation, Proc. Nat/. Acad. Sci. USA 100:7075-7080.
41
Imaging Brain Slices Ayumu Tashiro, Gloster Aaron, Dmitriy Aronov, Rosa Cossart, Daniella Dumitriu, Vivian Fenstermaker, Jesse Goldberg, Farid Hamzei-Sichani, Yuji Ikegaya, Sila Konur, Jason Maclean, Boaz Nemet, Volodymyr Nikolenko, Carlos Portera-Cailliau, and Rafael Yuste
INTRODUCTION Brain slices are convenient preparations to study synapses, neurons, and neural circuits because, while they are easily accessed by experimental manipulations such as drug applications, intracellular recordings, and optical imaging, they preserve many of the essential functional properties of these circuits. In this chapter, we describe techniques of live brain-slice imaging used in our laboratory. We cover in detail experimental protocols and know-how acquired over the years about preparing neocortical and hippocampal slices and slice cultures, loading neurons with dyes or using biolistic transfection techniques, two-photon and second harmonic imaging, morphological reconstructions, and image processing and analysis. These techniques are used to study the functional or morphological dynamics of synaptic structures, including dendritic spines and axon terminals, and to characterize circuit connectivity and dynamics. The importance of developing methods is underestimated in modem biology. The education of biomedical researchers and the federal granting agencies are dominated by the ideology that good research is question-driven, whereas technique-driven research is of lesser quality. We disagree with this exclusive view because it seems to us that the specific technique used is as important as the question addressed. As Sydney Brenner put it: "Progress in science depends on new techniques, new discoveries, and new ideas , probably in that order." (Brenner, 2002). As an example, we would argue that the invention of high-affinity, selective calcium indicators have revolutionized many fields of biology (Grynkiewicz et ai., 1985 ; Tsien, 1989). We feel that methods are essential, not only for performing and validating experiments, but as exploratory tools that generate new ideas, leading into new fields . Moreover, in our experience, the difference between a difficult experiment working or not often depends on minute technical details. These details are normally acquired with great effort by the investigator, yet generally they must be left out of publications. To help compensate for this, we present in this chapter a detailed account of current methods used in our .laboratory to image living brain slices. The general goal of our work is to use brain tissue specimens thin enough so that they can be successfully imaged optically. As explained in detail below, we use different types of brain slices and keep them in submerged chambers, where we seek to preserve ideal conditions of
temperature, ionic composition , and nutrients to enable the slices to survive as long as possible. Slices are imaged normally in upright microscopes, in order to enable electrical, as well as optical, access to the surface of the slice. In this respect, fixed stage microscopes are ideal because they enable the stable positioning of micromanipulators and mechanical independence of the focusing of the objective. Although inverted microscopes enable better optics, they are very difficult to use for electrophysiological experiments with slices because the electrical approach to the preparation comes from the opposite side of the slice than the optical approach. In this chapter we will discuss a combination of methods to image brain slices that are used in our laboratory. We will first cover in detail the preparation of different types of brain slices, discuss how to label cells in slices with optical probes, and then specifically discuss different types of imaging approaches to slices. We finish with an additional section of useful methods to morphologically reconstruct ne urons from slices for histological or ultrastructural work and a brief discussion of different image processing strategies that we use. We hope that other investigators will profit and learn from our experience and that this will enable more research teams to enter the exciting territory of imaging slices.
MAKING BRAIN SLICES Acute Slices Acute live slices prepared from the brain have become a standard preparation commonly used to study electrophysiological properties of neurons in circuits (Alger et ai., 1984) and, more recently, imaging (Yuste, 2000b). Most of our work is carried out with slices from mouse primary visual cortex (Fig. 41.1). The relatively high degree of preservation of neuronal organization after slicing and the availability of a variety of easy experimental manipUlations make acute slices an attractive experimental preparation. Generally, acute slices can be maintained in good condition for up to 12 h. At the same time, we find a lot of variability in the quality of slices from day to day and even from slice to slice. The large number of variables that are likely to be important in the preservation of the health of the slice make obtaining good slices somewhat of an art form. Unfortunately, systematic studies to determine
Ayumu Tashiro, Gloster Aaron, Dmitriy Aronov, Rosa Cossart, Daniella Dumitriu, Vivian Fenstermaker, Jesse Goldberg, Farid Hamzei-Si chani, Yuji Ikegaya, Sda Konur, Jason Maclea n, Boaz Nemet, Volodymyr Niko lenko, Carlos Portera-Cailliau, and Rafael Yuste· HHMI, Columbi a University, N ew York, New York 10027 722
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC. New York, 2006.
Imaging Brain Slices • Chapter 41
723
A
B
c
FIGURE 41.1. Brain slices of mouse visual cortex. Representative coronal slices from mouse primary visual cortex. Sections cut through the posterior pole of the cerebral hemisphere. (A) Photomicrograph of an acetylcholinesterase-labeled corona l section taken from the Franklin and Paxinos (I997) atlas. Note the more intense staining in layer 4 due to the high den sity of small granule cell s. Violet and blue shaded areas indicate the monocular and binocular regions of the mouse primary visual cortex, respectively. (B) Nissl stain of a mouse visual cortex slice from a PIS animal. Note the prominent layer 4. (C) Cytochrome oxidase staining of a similar section. Note the intense staining ncar layer 4, which is an indication of the primary visual cortex. Scale bar = 500 ~m. (Courtesy of Z. Peterlin and A. Tsiola.)
which variables are important to make healthy brain slices have not yet been done.
Protocol for Acute Neocortical Slices Mice are anesthetized with l20mg/kg ketamine and 10mg/kg xylazine (intraperitoneally) and decapitated with scissors. Some investigators in our laboratory prefer to perfuse the mouse with a gravity-fed cold saline solution prior to decapitation. The skin covering the skull is severed with a fresh razor blade above the midsaggital line of the skull. The skull is then eut along this line and forceps are slid under the skull posteriorly, such that an air bubble forms anterior to the forceps tips. This air bubble provides a perfect pocket for the forceps to slide anteriorly, with care taken not to
touch cortical tissue with the forceps. Each half of the skull may then be retracted laterally. The brain is then exposed and should be immediately placed into ice-cold sucrose artificial cerebrospinal fluid (sucrose-ACSF; 222mM sucrose, 2.6mM KCl, 27mM NaHC0 3 , 1.5mM NaH 2P0 4 , 2mM CaCI 2 , 2mM MgS04' bubbled with 95 % 0" 5% CO 2 ) , After approximately 3 min in ice-cold sucrose ACSF, the brain is removed and situated on the cutting block such that the cortex faces the approaching blade. Slices 300 to 400 11m thick are cut with a vibratome (Leica VT100OS; Leica, Nussloch, Germany; high vibration and slow speed setting) and incubated at 37°C for 30 min in a submerged slice chamber. Slices are then incubated at room temperature for up to 12 h, until used for experiments.
724
Chapter 41 • A. Tashiro et al.
Identification of Primary Visual Cortex The primary visual cortex, or area 17, of the mouse is located in the occipital region of the brain. In the adult animals it extends 1.3 mm anteriorly from the posterior end of the cortex (interaural line). In coronal sections it extends between 1 to 2mm laterally from the medial line at its anterior border and between 2 to 3 mm in the most posterior (Franklin and Paxinos, 1997). The primary visual cortex of the mouse is divided into two regions: the monocular region, which receives input from the contralateral retina and is located in the medial part, and the binocular region, which receives input from both retinas and is placed laterally (Zilles and Wree, 1985). The primary visual cortex can be identified in coronal sections by the densely arranged granule cells of layer 4 (Fig. 41.1) .
Thalamocortical Slice Protocol Thalamocortical slices are an ideal preparation to investigate the effect of thalamic inputs onto cortical neurons or circuits because it preserves both structures and connections between the ventrobasal nucleus of the thalamus and the somatosensory cortex. Preparation of the thalamocortical slice is slightly modified from Agmon and Connors (1991), as previously described (Beierlein et al., 2002). Briefly, C57BLl6 mice postnatal (P) 10 to 18 are anesthetized with 120mg/kg ketamine and lOmg/kg xylazine and decapitated. The brain is quickly removed and placed into cold artificial CSF (ASCF) containing the following (in millimolars): 126 NaCl, 3 KC1 , 1.25 NaH 2P04 , 26 NaHC0 3 , 10 dextrose, 1.3 MgS0 4 , and 2.5 CaCh (saturated with 95% O 2 and 5% CO 2). The brain is midsagittally sectioned into the left and right hemispheres. Each hemisphere is glued (standard cyanacrylate "super" glue) to a plastic block, midline down, anterior end pointing toward the floor and ventral surfaces facing in, toward one another. The hemispheres are rotated 10° from center line of the block. The plastic block is actually a right triangle with the hypotenuse 55° from the level (floor). Slices, 400llm thick, are cut with a vibratome (VT1000S) and then incubated at 32°C for 45 min. Usually two or three viable thalamocortical slices can be made from each hemisphere.
totemporal axis. The tissue is transferred onto a tissue chopper (TC-2 Tissue Sectioner; Smith & Farquhar) with two spatulas. The rectangular block of tissue is positioned such that the chopping orientation is perpendicular to the septotemporal axis of the hippocampus. Slices 300llm thick are obtained. With two flatended spatulas, the slices are transferred to a fresh culture dish containing cold sucrose-ACSF. The slices are separated from each other with a surgical blade and a flat-ended spatula as soon as possible. Slices are cultured in culture medium, 100mL of which contains 50mL Basal Medium Eagle (catalog #21010-046, Invitrogen), 25 mL Hank's Balanced Salt Solution (catalog #24020-117, Invitrogen), 25 mL heat-inactivated horse serum (Hyclone), 0.65 g dextrose, 0.5 mL L-glutamine (catalog #25030-149, Invitrogen) , 1.0mL HEPES (catalog #15630-106, Invitrogen), and 1.0mL 100X Pen-strep (catalog #15140-148, Invitrogen). Approximately 1 mL of medium is poured onto and under the culture inserts (catalog #PICM 030 50 or PICM ORG 50, Millipore) in the sterile hood, so that the membrane of the inserts is completely submerged in culture medium. We find that the use of serum from Hyclone (Logan, UT) is a particularly important variable because cultures made with serum from other sources were not successful. Individual slices are then transferred onto the membrane with a flatended spatula. Three to six slices are cultivated on single inserts. Most of the medium (but not all) is removed from the inserts with a pipette, and the slices are positioned with the spatula at the center of the insert, but separated from each other by at least 2 to 3 mm. Then, all remaining medium in the inserts is aspirated. The inserts are transferred into 6-well culture plates, in which each well contains 1 mL of culture medium . The culture plates are kept in the incubator (5% CO 2 , 37°C). Every other day, 0.6mL of culture medium is changed with fresh medium. During the first few days in culture, slices spread slightly and become flattened to a 150 to 250llm thickness.
LABEliNG CELLS
Cultured Slices
Biolistic Transfection
Because acute slices cannot be maintained in good condition for more than 12h, long-term culture is required for manipulations involving long-term experiments, such as those requiring the expression of genes. Below, we describe the protocol for mouse hippocampal slice cultures, which we have used extensively to image the morphology of single neurons transfected with the green fluorescent protein (GFP) (Dunaevsky et al., 1999; Tashiro et ai., 2000).
To image neuronal morphology, we transfect GFP using biolistics ("gene gun") (Arnold et al., 1994; Lo et al., 1994). The principle of this method is that metal particles coated with DNA are transferred physically into nucleus by pressured gas. Transferred GFP genes are expressed and the whole neuronal cytoplasm, including their axons and dendrites, can be visualized (Fig. 4l.2).
Protocol for Hippocampal Cultured Slices Neonatal mice (PO-P3) are cryoanesthetized on ice for 1 min and decapitated with scissors. In a tissue culture hood, skin and skull are cut with scissors and separated with forceps. The brain is then gently removed and placed into a 35 mm tissue culture dish filled with cold sucrose-ACSF (see above). Under a dissecting microscope, the two hemispheres are separated with a surgical blade and oriented such that the medial surface faces down. The cerebellum and the mesencephalon are carefully dissected away and discarded. Then, after the hemisphere is rotated such that its medial side faces up, the diencephalon is removed with a surgical blade and a flatended spatula. The remaining piece of tissue, representing the cortex and the hippocampus, is trimmed into a rectangular block along the anterior edge of the hippocampus parallel to the sep-
Biolistic Protocol We use the Helios Gene Gun System (Bio-Rad), and the following protocol is modified from the procedure described in the manual for the system. Plasmids are purified on a Qiagen Maxiprep kit. Weigh 12.5 mg gold (111m diameter) in a 1.5 mL tube and add 100 ilL of 0.05 M Spermidine. Sonicate the tube for 5 s to di ssociate aggregated gold particles (FS30, Fisher Scientific; 40kHz, l30W). Add solution containing desired amount of plasmid [for eGFP-Cl plasmid (Clontech), lOOJ.1g] and precipitate plasmid onto gold particles with 100llL of 1 rnM CaCI 2 . In lOmin, gold particles precipitate to the bottom of the tube. Then, remove supernatant solution without disturbing the particles and wash particles three times with 100% EtOH. The particles are then suspended in 100% EtOH and transferred into a new 15 mL tube to a final volume of 3 mL. Transfer particles in EtOH into Tefzel tube (Bio-Rad). In approximately 90 s, gold particles precipitate to
Imaging Brain Slices • Chapter 41
C
725
0 min
FIGURE 41.2. Two·photon imaging of living neurons in slices. Two-photon micrographs of GFP-transfected neurons in hippocampal brain slice cultures. (A) A eA3 pyramidal neuron at 14 days in vitro. (8) Apical dendrite of a eA3 pyramidal neuron at II days in vitro. (e) Time-lapse sequence of a dendritic spine in (8). Note the morphological rearrangements occurring in a few minutes. (D) Dentate granule cells at 14 days in vitro. Note that the entire dendritic and axonal processes are visualized. Scale bar: (A) 50Jlm, (8) 2.5/lm. (C) l/lm. (D) 150/lm.
the bottom of the tube. Then, EtOH is removed from the tube slowly without disturbing the particles. Tube is dried with nitrogen gas until the inside surface of the tube becomes completely dry. With tubing cutter (Bio-Rad), the tube is cut into small pieces, which are used for single shots. Tubing sets can be stored with desiccant at 4°C for up to a month and at -80°C for longer storage. To drive gold particles into neuronal nuclei, we use highpressure helium flow. We adjust the helium pressure to 100 to 150psi for transfection of slice cultures and acute slices. Two to three "preshots" are fired with an empty cartridge holder to clean the helium pathway and make sure that pressure is stable after each shot. In order to reduce the damage to slices caused by highpressure flow, the tips of barrel liners are covered by a nylon mesh (90 11m, Small Parts, Inc). For cultured slices. the cover of the culture plate is removed, and the gun is fired perpendicular to the plate with a distance of 10 mm between the tip of the barrel
liner and the insert. The culture plates are immediately put back into the incubator. Slices are incubated for 2 to 5 days before imaging.
Genetic Manipulation with Dominant-Negative and Constitutively Active Mutants One of the advantages of biolistics over other transfection methods is that cotransfection of multiple genes is quite easy. If the two genes are in separate mammalian vectors, they can be cotransfected with high cotransfection efficiency (>90% in our hands) by simultaneously coating gold particles with these two vectors. With cotransfection of GFP and dominant-negative or constitutively active mutant genes, the roles of specific molecular signaling cascades in the regulation of neuronal morphology can be examined. For example, we have been studying the roles of Rho GTPases in
726
Chapter 41 • A. Tashiro et al.
regulating morphology and motility of dendritic spines (Tashiro et ai., 2000; Tashiro and Yuste, 2004).
Diolistics and Calistics Conventionally, lipophilic dyes, such as DiO and DiI, are used to label axonal projections between different regions of the brain by placing a crystal of these dyes in a defined region (Honig and Hume, 1989). However, this method is not suitable for imaging the morphology of single neurons because the region near the crystal is stained too densely to visualize single neurons. To label single neurons, Gan and colleagues developed diolistics, a variant of particle-mediated gene transfer, which transfers metal particles coated by fluorescent dyes onto cells (Gan et ai., 2000). Our group has extensively used this method to visualize neurons in fixed tissue and classified cortical pyramidal neurons in mouse VI into different morphological categories (Tsiola and Yuste, 2003). Calistic, another variant of biolistics, serves to inject Ca2+ indicators such as calcium green-1 and fura-2 into neuronal cytoplasm (Kettunen et al., 2002). With this method, an apparently higher concentration of calcium indicators can be injected into neurons than using acetoxymethyl ester (AM) loading, and a large number of neurons can be visualized. Calistic also allows the simultaneous measurement of morphological and calcium dynamics of single neurons (Lohmann et ai., 2002).
Slice Loading and "Painting" with Acetoxymethyl Ester Indicators In our past work we have pioneered the use of calcium imaging to characterize the activity of neuronal populations (Yuste and Katz, 1989, 1991; Smetters et ai., 1999; Peterlin et aI., 2000). The bulk loading method for double incubation of cortical slices with fura2 AM or indo-1 AM (Molecular Probes) calcium indicators has been previously described (Yuste and Katz, 1989; Yuste, 2000a). Briefly, cortical slices are initially incubated with 2 to 5/lL of a 1 mM fura-2 AM or indo-l AM in 100% DMSO solution for 2 min, followed by a second incubation in 3 mL of 10 11M fura-2 AM in ACSF for 60min. However, the use of the thalamocortical slice preparation (Beierlein et al., 2002) necessitated the development of a modified bulk-loading procedure because the long-projecting thalamocortical (TC) axons are particularly sensitive to the double incubation methodology, even though local connections within the cortex remain intact (Kozloski et ai., 2001). We have been able to circumvent this problem by applying the fura-2 AM or indo-1 AM solution directly to the region of interest in the cortical slice with a pipette, tip diameter approximately 30/lm, filled with fura-2 AM. The region of interest of the slice can be "painted" with fura-2 AM or indo-1 AM. In this way one is able to achieve good loading while preserving TC connections up to postnatal day 18 in mouse barrel cortex (Fig. 41.3). The maintenance of intact TC projections was confirmed using thalamic stimulation that elicits a calcium response in barrel cortex indicative of intact thalamocortical axons (Beierlein et ai., 2002).
Dye Injection with Whole-Cell Patch Clamp Whole-cell patch clamp is commonly used to study electrophysiological properties of neurons in brain slices (Edwards et ai., 1989). Using electrodes filled with fluorescent dyes, the whole-cell configuration of patch clamp injects the dyes into neurons by diffusion through the pipette tip into the neuron. This technique has the advantages that the labeling procedure is rapid and that any neuron in the slice can be targeted and therefore visualized. When electrophysiological measurements are combined with imaging, a lower concentration of dyes is used and the whole-cell patch clamp is maintained during an experiment (Yuste and Denk, 1995). However, an extended period of patch clamp may interfere with cellular functions such as spine motility (Majewska et al., 2000a), possibly because the biochemistry in the neuron is perturbed by the perfusion of intracellular solution or the diffusion of cytoplasm into the patch electrode. Because of this problem, we routinely fill the electrode with higher concentration of fluorescent dyes and pull out the electrode a few minutes after whole cell recording is established (bolus technique, see below; Helmchen et ai., 1996; Majewska et ai., 2000a).
Protocol for Slice AM Painting 1. Deposit TC slice carefully onto the bottom of a small Petri dish (35 x 10 mm) filled with 2 mL of ACSF aspirated with 95% O 2 and 5% CO 2 and place onto microscope stage. 2. Fill a fire-polished pipette (tip diameter -30/lm) with fura-2 AM from a previously prepared aliquot of 50 /lg of fura-2 AM
Bolus Injection Protocol Neurons of interest are identified using differential interference contrast (DIC) optics, and then patched and recorded using the whole-cell patch clamp technique in current-clamp configuration to ensure the neurons are healthy. Electrodes are filled with a solution containing (in millimolars): 5 NaCl, 10 KC1, 10 HEPES, 135 KMeS04, 2.5 to 4 Mg-ATP, 0.3 Na-GTP, and 100 to 500/lM fluorescent dye such as Calcium Green-lor Alexa-488 (Molecular Probes, Inc., Eugene, OR). Electrodes are then pulled out 1 to 3 min after patching to prevent dialysis of cytoplasm. The resistance of patch electrodes is typically 7 to 14 MO. Diffusion of dyes is so rapid that the whole dendritic tree is visualized in a few minutes.
FIG U RE 41.3. Two-photon imaging of neuronal ensembles. Two-photon micrograph of an acute cortical brain slice, loaded with indo-l AM. A number of neurons are loaded with the calcium-sensitive indicator, indo-I. Note that dendritic processes are also visualized in many neurons. 60 pixels correspond to 20 11m. Scale bar: 50llm.
Imaging Brain Slices • Chapter 41
dissolved in IO~L of DMSO and 2~L of pluronic acid (F127, Molecular Probes). 3. Insert the filled pipette into a standard patch clamp electrode holder, with tubing attached, and using a micromanipulator, place pipette tip 100 to 200 ~m above the surface of the slice. Apply 5 to 10 psi positive pressure to the pipette. Slowly move the pipette across the surface of the slice using the manipulator, covering the area of interest with the dissolved fura-2 AM . 4. Incubate the slices at 32°C for 24 to 28 min depending on the age of the animal from which the slices were taken (younger animals require shorter incubation times), aspirated with 95% O 2 and 5% CO 2 throughout. 5. Finally transfer the slices to oxygenated ACSF at room temperature at least 15 min before use for the experiment.
Green Fluorescent Protein Transgenic Mice Recently, a number of different types of GFP transgenic mice have become available commercially or from independent investigators. If these mice express GFP in neurons of interest at the right age, the labeling procedures described above are circumvented. For example, we have used the GFP-M line of GFP transgenic mice developed by Feng and colleagues (Feng et aI. , 2(00). At the second postnatal week, this line of mice expresses GFP weakly in V I, but strongly in pyramidal neurons in pyriform cortex. In addition, too many pyramidal neurons in the hippocampal CA 1 region are labeled by GFP, so background fluorescence makes it difficult to visualize single neurons.
IMAGING SLICES
Two-Photon Imaging of Slices For imaging brain slices we almost exclusively use upright microscopes (Olympus BX50WI) because they can provide easy combination of electrophysiological techniques (whole-cell patch clamp and extracellular stimulation) with simultaneous imaging of patched and/or stimulated cells. As explained, with inverted microscopes it is difficult to position the patch/stimulating electrode from one side of the slice and image from the same side. Thi s requirement is satisfied in the case of upright microscope and dipping-type water-immersion objectives with a working distance large enough to enable bringing the e lectrode in the field of view from the same side. Two-photon imaging is carried out with a custom-built twophoton laser-scanning microscope (Majewska et ai. , 2000b). A more recent description of our system can be found at www.twophoton.com or at http://www.columbia.edulcu/biology/ faculty/yustelindex.html. The microscope consists of a modified Fluoview (Olympus, Melville, NY) confocal microscope with a titanium: sapphire (Ti: Sa) laser providing -130 fs pulses at 76 MHz at wavelengths of 700 to 900 nm (Mira, Coherent, Santa Clara, CA) pumped by a solid-state source (Verdi, Coherent). We detect the fluorescence with a non-descanned detector (see below). Imaging is done at low excitation intensities (3-10mW at sample). Under these conditions no significant photobleaching or photodamage is observed, allowing us to image for long periods of time. For fast time resolution we can record continuous mov ies (1000 frames per movie, 0.2-1.6 s/frame), acquiring individual calcium fluorescence signals from hundreds of neurons simulta-
727
neously. Alternatively, in a time-lapse mode (I framellS s), we can image the same region for up to 6h without appreciable photodamage. The major limitation of the use of two-photon imaging to monitor the activity of large neuronal populations is the slow time resolution associated with laser-scanning methods. It is therefore best suited for the study of slow events rather than to detect single spike correlations. By performing online analysis, we can identify prominent features of spontaneous activity, target key elements of the network, perform whole-cell recordings while continuing to image the slice, and characterize electrophysiologically the ne urons that participate in these events.
Slice Chamber Protocol Acute and cultured slices are continuously perfused with standard ACSF containing (in ruM): 126 NaCI, 3 KCI, 2 CaCh, I MgS0 4 , 1.1 NaH 3 P0 4 , 26 NaHC0 3 , and 10 dextrose and saturated with 95 %0 2 and 5%C02 • To hold the slices on the microscope, we use a temperature-controlled chamber (Series 20 imaging chamber, Warner Instrument). Flow is gravity-driven by raising a container (60ml syringe, for example) above the chamber and controlled by a flow regulator. ACSF is sucked from the chamber by a vacuum pump. Sometimes it is necessity to use an additional flow regulator in the vacuum line to stabilize the level of liquid in the recording chamber, especially when a powerful vacuum pump ("Air Admiral"; ColeParmer) is used. Medium flow in the chamber can cause movement of the whole slice, which is a serious problem in time-lapse imaging, especially with small structure such as presynaptic and postsyn aptic structures. To minimize movement artifacts, medium flow is reduced to I mUmin and the sli ces are stabilized with a slice anchor (Warner Instrument). As an alternati ve to using a weight that can damage the slice, we also use the direct adherence of the slice to the chamber. To do so, we position the slice on the bottom glass of the chamber and drain all the ACSF. After a few seconds, we reperfuse the chamber carefully so as not to lift the slice. In most cases, the slice has adhered to the chamber and will not move for the rest of the experiment. ACSF is heated before fl owing into the chamber by an in-line heater (SH-27B, Warner Instru ment), and the base of the chamber is also heated by a pl atform heater (Series 20 platform, Warner Instrument). These heaters are controlled by a dual channel heater Controller (TC-344B, Warner In strument). The temperature of the in-line heater is set at - 39°C and the platform heater is kept at - 39°C , in order to keep the liquid in the chamber at -36-37C. As an independent control of the liquid temperature, we use an additional thermosensor (Warner) or a thermocouple-based handheld digital thermometer (TES 1300). If ACSF is saturated with O 2 and CO 2 at room temperature, these gases come out of solution in the heated imaging chamber and produce a number of small bubbles. These bubbles degrade image quality and can damage the slices. To prevent this, we keep the ACSF container in a hot bath and saturate the ACSF with the gases at 37°C.
Choice of Objectives We use 40x (0.8NA) or 60x (0.9NA) dipping-type water immersion objectives (Olympus), although we have recently started to use the new 20x 0.95 NA objective (Olympus) to image a larger number of cortical neurons. With this low-magnification, highnumerical aperture (NA) objective, we can simultaneously monitor the activity of large neuronal populations (average 650 neuron s, range 184-1396) in a thin optical section of the slice. The area
728
Chapter 41 • A. Tashiro et al.
viewed covers up to 5 different layers of the primary visual cortex (-400 x 700 11m). The improved depth penetration and the high sensitivity of two-photon imaging allows us to image at depths of > I 00 11m from the slice surface, where connectivity is less affected by the slicing procedure. High-NA objectives further increase the fluorescence collection and thus allow deeper fluorescence measurements with good resolution than conventional objectives. Even at low magnification, we can resolve individual neurons and some of their dendritic processes. However, a major difficulty associated with the use of the 20x objective for patching, stems from its large dimensions, which restricts access for electrodes. In general, it is also necessary to minimize the working depth in the tissue. Deeper imaging leads to loss of excitation light and fluorescence signal because of light scattering by the tissue. Also deeper imaging leads to lower contrast in bright-field and makes patching extremely difficult. That is why one of the major requirements for objectives is a long working distance. The Olympus 20x, 40x and 60x water-immersion objectives satisfy this requirement, with working distances of at least 2 mm. On the other hand it is possible to use objectives with shorter working distances for studies which do not require simultaneous imaging and electrophysiological recording (morphological studies of cells labeled genetically or by bolus injection). The efficiency of signal collection in the case of two-photon laser scanning microscopy using the whole-area detection mode directly depends on the NA of the objective lens used. Also, a higher NA decreases the diffraction-limited size of the excitation spot, which gives better spatial resolution and provides higher local intensity of the excitation light, thus increasing the efficiency of nonlinear optical effects (multi-photon absorption or secondharmonic generation). In this respect, the Olympus 20x 0.95 NA water-immersion objective was crucial and allows 2P Ca2+ imaging of large neuronal populations with excellent signal to noise ratio (Fig.41.3A). Choosing the right magnification for the objective lens is another practical issue. Lower magnification is good for imaging larger structures; although low-magnification objectives are not convenient for simultaneous electrophysiological recordings, if cells have to be patched while visualizing them through the eyepieces. In some cases, increasing magnification with a separate system of lenses solves this problem. As an example, we have a "U-CA" adaptor from Olympus for the BX 50WI microscope, which works as a magnifying telescope and is inserted into the optical pathway of the microscope in the region where the light is collimated. On the other hand it is not ideal to use magnifying adaptors for imaging, because they lead to a loss of fluorescence signal in collection pathway. Hence this adaptor should be used only for convenience when observing with low-magnification objectives (20x) through the eyepieces. High magnification lenses (40x, 60x, lOOx) with higher numerical apertures provide images with higher spatial resolution in 3D. Using the "digital zoom" option available in majority systems for laser scanning microscopy, allows one to set the pixel size according to Nyquist. This kind of spatial sampling is not always necessary. For example, when imaging neuronal populations, we are interested mostly in the integrated signal from individual cells. Nyquist sampling, or even oversampling, is important if we ask questions about sub-resolution movements of small structures (i.e., quantitative analysis of spine movement). In this case, one should not consider a laser scanning microscope as an imaging device with resolution limit defined by diffraction, but more as a position-measuring instrument that measures the centroid of the
distribution of fluorescent molecules and the dynamics of the position of this centroid. For two-photon laser scanning microscopy, it is important to make sure that all the components of the excitation pathway have good transmission in the near IR. Users should use objective lenses corrected for optical distortions and made to be transparent in the NIR. In many cases, additional changes are needed in the installed optics to make them IR-transparent (pupil-transfer lens in our case; (Majewska et al., 2000b). Objective lenses used for multi-photon imaging should be free of geometrical (spherical) aberrations. The requirement for the absence of chromatic aberrations is not so important - excitation light is practically monochromatic (the spectrum widening caused by the finite length of pulse from mode-locked lasers is negligible). Also, the absence of a confocal aperture in front of the detector and the general architecture of the collection system, emphasizes collecting the maximum fraction of the emitted light, and allows one the freedom of using collecting optics (objectives in case of 2P-fluorescence and condenser lens in the case of SHG) not well-corrected for chromatic aberration.
Beam Collimation and Pulse Broadening The majority of modem microscopes are designed for infinitycorrected objective lenses, so it is important to provide collimated laser light to the back-aperture of the objective (Tsai and Kleinfield, 2002). Even if initially the microscope system and scanning head are designed to provide collimated light at the backaperture of objectives, custom modifications of the optical pathway and the switch to NIR excitation can distort this collimation. This indeed happened in our custom-made 2P-microscope and we solved the problem by introducing additional optics into the excitation pathway (Nikolenko et aI., 2003). Specifically, we use a simple system of 2 lenses in order to collimate the light to the objective lens. Our system also works as a "beam expander" - it modifies the laser beam in such a way that the excitation beam at the back aperture of objective is not only collimated, but also is large enough to slightly overfill the objective pupil. One of the major requirements for laser scanning microscopes is that the back aperture should be overfilled by the excitation light (Tsai and Kleinfield, 2002). This minimizes variation of excitation power across the field of view and guarantees that the full numerical aperture of the objective lens is used. The level of overfilling should be minimal to maximize the amount of excitation power deliveried to the sample. Another important factor in 2P microscopy is the problem of pulse broadening. Nonlinear microscopy requires using pulsed laser light. Mode-locked lasers generate a train of pulses of finite length at certain repetition rate. Even though each pulse represents monochromatic light, the finite length of each pulse leads to a spectrum with certain width in Fourier-space. Linear dispersion of this light in the intermediate optical elements disturbs the phase relations between the different spectral components of the pulse, which in practice leads to the pulse being broadened in time. This decreases the peak excitation intensity, and hence decreases the efficiency of nonlinear excitation. In order to avoid this, the optical system should use the minimum number of lenses between the laser and the specimen. Alternatively, one can add additional optics with negative dispersion in order to compensate for the positive dispersion of the rest of the optics (Lechleiter et al., 2002).
Imaging Brain Slices • Chapter 41
Image Production, Resolution, and z-Sectioning In our two-photon microscope, fluorescence is detected with photomultiplier tubes (PMTs; HCI2S-02, Hamamatsu, Japan) used in an external, whole-area detection mode, and images are acquired using Fluoview software (Olympus). Images are sometimes taken at the highest digital zoom, resulting in a nominal spatial resolution of 20-30 pixels per 11m with the 40x. This spatial resolution is suitable for imaging very small structures including dendritic spines, the size of which is typically on the order of 11m. Since brain slices are three-dimensional, we collect a series of images (z-stack) from different focal planes to cover the whole neuronal structure of interest. In principle, three-dimensional structures can be reconstructed from the z-stack. However, this is not practical when the same structure is imaged repeatedly, particularly in time-lapse imaging, for the following reasons: (I) perfusion causes small movements of the slice so the reconstructed structures are not accurate and (2) to achieve pixelation in the z-direction at a similar level to the x- and y-direction, many focal planes have to be scanned. This is impractical because it compromises fast timelapse imaging and increases the possibility of photodamage. To circumvent these problems, we scan the images with a 111m difference between focal planes (up to 9 planes), and then project the z-stack into a single, two-dimensional image. Since the point spread function of the 60x objective lens in our microscope measures approximately 0.4 x 0.4 x 1.311m (Majewska et ai., 2000b), the images with 111m focal distance have enough overlap to produce a good projection. A major problem associated with time-lapse imaging of brain slices is slice movement in the x, y and z-directions. To minimize movement in z, we routinely scan extra focal planes at the top and the bottom of the z-stack. If movement in the structure of interest is evident in these extra focal planes, we move the whole z-stack 111m up or down. Thus, structures of interest are not lost from the z-stack. Structures of interest can also move out of the images by moving in the x or y directions. To minimize this, we try to make the slice adhere to the bottom glass of the chamber. In addition, we avoid placing the structures of interest near the edge of the image, and if the structures move near the edge, we reposition the specimen so that the structures move toward the center of the imaged area. Although this prevents the loss of structures of interest from the z-stack, xy movement results in the drift of the whole image in the time-lapse movies. In these movies, it is extremely difficult to observe and analyze changes in morphology and fluorescence intensity. Therefore, in the analysis we compensate the xy movement as described below.
Choice of Indicators for Two-Photon Imaging of Calcium Two-photon excitation of calcium indicators loaded via whole-cell recording is ideal for imaging calcium during action potential generation or during synaptic excitation (Yuste and Denk, 1995). We have used both calcium green-l and fluo-4 successfully and find that both indicators are excited well by a mode-locked laser at 800nm. However, we find that each indicator is suited for different conditions. Because calcium green-l is brighter at low calcium concentrations, it is ideally suited for visualizing fine structures such as dendritic spines. However, in part due to its high fluorescence at rest, its increase in intensity on binding calcium is compromised (FjFfree = -14), such that it is imperfect for detecting small or heavily buffered signals. On the other hand, although fluo-4 is dim at rest and therefore demands higher excitation laser power, it
729
undergoes a large change in fluorescence on binding calcium (FmaJFmin= -I 00). Thus, fluo-4 is ideal for imaging small or heavily
buffered signals, such as in cortical interneurons, which have high endogenous buffering capacities. We have used calcium green-l over the concentration range of SO to 200I1M, and fluo-4 from 100 to 400 11M. Because increased indicator concentrations cause larger distortions of the true calcium signals, the smaller the concentrations yield more physiologically-relevant data; however, if higher laser intensities are necessary to observe signals in environments with low concentration, photodamage may be accelerated. In our hands, indo-l is the best choice when shorter wavelengths «7S0nm for two-photon excitation) are needed to combine calcium imaging with uncaging techniques. The two-photon sensitivity of calcium indicators is available from http:// cell science .bio-rad.comlproducts/mul tiphotonIRadiance21 OOMP/ mpspectra.htm.
Photodamage The Achilles' heel of all live imaging is phototoxicity. Using our two-photon microscope, we experience two types of photodamage, when imaging neurons with too-high laser power (>20mW at the sample). First, unstained cells that show strong autofluorescence suddenly collapse, emitting high-intensity light, like an explosion. This often masks the structure of interest. We recommend not including any structures with high autofluorescence in the imaging area. Second, stable structures in labeled neurons can show abnormal morphological rearrangements, particularly beading. This type of photodamage is nonlinear and, in some cases, can start to occur even minutes after the illumination has stopped. Moreover, even though previous scans can be normal, the same intensity of excitation can suddenly cause photodamage, particularly if the imaged structures are near the surface of the slice and if the concentration of dye is high. As is the norm in all microscopy, we recommend imaging with as Iowan excitation intensity as possible and adjusting this intensity for each sample depending on the brightness of the image. In some experiments where photo damage becomes a persistent problem, we use the antioxidant Trolox (Sigma, 1010011M), added to the ACSF (Sheenen et ai., 1996). We have not noticed any effect of Trolox on the physiology of the neurons, although it has been suggested that high concentrations of it can block NMDAR (A. Konnerth, personal communication).
Second Harmonic Imaging Second harmonic generation (SHG) is a nonlinear optical effect in which the incident light is coherently scattered by the specimen at twice the optical frequency and at certain angles (Lewis et al., 1999; see Chapter 40, this volume). The signal can be produced by endogenous structures or from inserted chromophores. Unlike fluorescence, in which emitted photons are best detected in the epi configuration, SHG photons are best detected in the transmission path of the microscope. The SHG photons, generated at the focal spot of the laser in the sample, are collected by a condenser lens which has to have the same NA as the objective lens in order to collect the whole cone of light. This is important because the SHG radiation in the forward direction is restricted to certain off-axis angles. It is best to have a spatially filtered laser beam for SHG because it is a coherent process - the spatial filter acts as a point source and restores the Gaussian wavefront and phase. Spatial filtering can be achieved by a telescopic system of two positive lenses and a pinhole placed between them such that the pinhole-
730
Chapter 41 • A. Tashiro et al.
to-lens distance equals the focal length, respectively. In front of the photomultiplier tube (PMT), which is placed in the transmission path, is a narrow band (-20 nm) filter centered at half the wavelength of the laser. The amplification of the signal is done by standard methods as in fluorescence detection. Specific instructions to adapt a two-photon microscope for SHG imaging can be found in Nikolenko and colleagues (2003).
Silicon-Intensified Target Camera Imaging While two-photon imaging results in a high spatial resolution imaging with the least photobleaching and most depth penetration, there is a cost, as with any laser-scanning microscopy, in terms of temporal resolution when many neurons are simultaneously imaged because the laser beam must be systematically moved over a large area. While future scanning modifications may alleviate this problem, at present single-photon fluorescence imaging has proved a useful technique for measuring the fluorescence changes in several neurons simultaneously (Peterlin et al., 2000; Kozloski et al., 200 I). For example, using calcium imaging with fura-2, the typical detectable change in fluorescence that can be measured in a neuron from a single action potential has a fast onset « 100 ms rise time) and slow decay (>2 s; Smetters et al., 1999). This allows one to detect the time of occurrence of action potentials in identifiable neurons with 100 ms temporal resolution, provided that frames are acquired at a rate of 50ms or less. We have achieved this using a SIT camera (Dage), BX50WI microscope (Olympus), 40x 0.8 NA water-immersion lens, an LG-3 frame grabber (Scion Corp.) in a Power Macintosh 7600, and NIH image software. With this equipment, we can view an area of cortex of 320 x 240 j..lm with a spatial resolution of 640 x 480 pixels, and we can capture frames at a rate of 30 frames/so Using fura-2 AM loaded slices (see above for loading technique), we can image dozens of neuronal somata using a mercury source (Olympus), a 380nm excitation filter and 510nm emission filter. The gating of the light source is accomplished via a triggered shutter (Uniblitz), which fully opens in less than 30ms after the triggering TTL pulse. The acquisition of a frame or a set of frames (a "movie") can also be initiated from an external trigger.
A
The delay from triggering the movie to acqUISItIon of fully illuminated frames is from 100 to 150 ms to compensate for the 100 to 150ms lag of the camera. Each captured frame uses 307.3 kB of memory. Homemade macros have been written using NIH image, controlling the shutter through the modem port of the computer. These macros enable the acquisition of movies timelocked to either depolarizing current pulses in current-clamped neurons (as in Kozloski et al., 2001) or to a large PSC recorded in a voltage-clamped neuron via a window discriminator (WPI).
MORPHOLOGICAL PROCESSING AND ANALYSIS As our most reliable method for morphological reconstructions, we use biocytin fills and processing to recover the morphologies of the neurons imaged (Fig. 41.4).
Biocytin Protocol Following electrophysiological recordings, the slices are immediately placed in 4% paraforrnaldehyde in 0.12M phosphate buffer (PB) and kept at 4°C overnight. Slices are then cryoprotected in 20% sucrose in 0.12 MPB for 2 to 8 h and frozen on dry ice in tissue freezing medium (catalog #H-TFM, Triangle Biomedical Sciences). Upon defrosting, slices are rinsed in 0.12 MPB three times and pretreated with I % hydrogen peroxide in 0.12 MPB for 30 min under agitation at room temperature. The tissue is then rinsed in 0.02M potassium phosphate saline (KPBS) and incubated in AvidinBiotin-Peroxidase Complex (catalog #PK-6100, Vector Laboratories, Inc.) overnight under agitation at room temperature (lOj..lL solution A and 10j..lLsolution B per I mLofO.02MKPBS and 0.3% Triton-X). Slices are rinsed in 0.02M KPBS three times and incubated in 0.7 mg/mL 3,3'-diaminobenzidine, 0.2 mg/mL urea hydrogen peroxide, 0.06MTris buffer (catalog #D-4293, Sigma-Aldrich) in 0.02M KPBS for 5 to 15 min. Upon completed DAB reaction, the slices are rinsed in 0.02M KPBS and mounted in Vectashield mounting medium (catalog #H-1000, Vector Laboratories, Inc.).
B
FIGURE 41.4. Histological reconstruction of neurons using biocytin and Neurolucida. (A) Biocytin staining. A pyramidal ccll in a coronal section from mouse visual cortex was filled intracellularly with biocytin and then processed for visualization. Intracellular biocytin staining enables a large signal/noise and allows a fairly accurate reconstruction of the dendritic arbor of the neuron. Structures as small as spines (-1 j.!m in diameter) can be visualized. Arrowhead indicates the axon. Scale bar = 50j.!m. (Courtesy of Z. Peterlin and A. Tsiola.) (B) Neurolucida reconstruction. (Left panel) Confocal image of a pyramidal neuron from a P7 mouse slice, cultured for 6 days. The cells were transfected with EGFP using a gene gun. (Right panel) Reconstruction of the neuron using Neurolucida.
Imaging Brain Slices· Chapter 41
Anatomy with a Two-Photon/Neurolucida System We have experimented with direct two-photon reconstructions of the cells in the brain slices. This procedure enables the investigator to quickly reconstruct the morphology of the imaged cell. Images of neurons from both live and fixed tissue can be taken by the two-photon microscope and the stacks of images can be imported into a computerized reconstruction and measuring program, such as Neurolucida (Microbrightfield, Brattleboro, VT).
Two-Photon/Neurolucida Protocol When examining dendritic morphology alone, z-stacks of the neuron of interest can be captured using a 20x or 40x objective. When spines and filopodia are also of interest, a 60x objective with a 2.Sx digital zoom yields good images. For detailed reconstructions of the protrusions from the entire neuron, small overlapping sections of the neuron are imaged using the 60x objective and 2.Sx digital zoom. After capturing the z-stack of images, they should be saved as a Fluoview Multi-Tiff (* .tiff) and transferred to a computer running the Neurolucida software. The images can either be burned onto a recordable CD, or can be transferred to the computer over a local area network. The neurons can then be reconstructed using Neurolucida software. The stack can be opened using the Image Stack Open command under the File icon on the menu bar. Once the image is opened, the brightness and contrast of the image can be adjusted by selecting Image Effects under the Video icon. Once the picture is in clear focus, the neuron is ready to be reconstructed. The image must be calibrated and a reference point chosen. The PgUp and PgDn keys on the keyboard will allow for scrolling through the stack of images. The mouse is used to trace the neuron and the type of tracing can be set by right clicking on the image. We find that rubber line tracing is very effective. The thickness of the line can be determined by the scroll feature on the mouse (the wheel or center button). In the tool bar we are able to select the section of the neuron we are drawing, for example, apical dendrite or cell body. Nodes and branches are added by right clicking on the image during tracing. After the contour is completed, it can be saved and opened in the Neuroexplorer program. This program will allow for easy analysis of the reconstruction.
731
rabbit IgG antibody (Roche Diagnostics) for 1 h at room temperature. After developing with DAB, the sections are postfixed in 1% osmium tetroxide in PB, dehydrated, and then infiltrated with Epox 812 resin (Fullam), placed flat in resin between two plastic slides, and polymerized in the oven at 60°C. After polymerization, the plastic slides are separated and the imaged areas of interest were cut out, mounted on a blank block, and sectioned to 10 11m for examination under phase contrast. The 10 11m sections with the imaged areas of interest were then remounted on a blank block, thin sectioned, and examined in the electron microscope (JEM 1200EX). Imaged mossy terminals with filopodia are reconstructed from serial sections (see also Chapter 49, this volume).
Protocol for Two-Photon/EM Imaging of Biocytin Labeled Cells Neurons are bolus injected with electrode containing 0.4% biocytin and I mM Alexa-488 as described before. Immediately after imaging, brain slices are immersion fixed with 4% paraformaldehyde and O.S% glutaraldehyde in 0.1 MPB overnight at 4°C. They are washed three times 10 min on shaker at room temperature with 0.1 MPB. Then, slices are incubated in 1% hydrogen peroxide, SO% ethanol, and O.OSMPB for 30min at room temperature on shaker for eliminating internal peroxidase activity. After washing again, they are incubated with ABC (Vectastain) overnight at 4°C and 1 h at room temperature. After washing again, they are DAB reacted using fast DAB%o from Sigma for about 3 min. Then, pictures are taken from the imaged region to facilitate its EM reconstruction. At this point, slices can be postfixed with glutaraldehyde. Finally, slices are osmicated by 1% osmium tetroxide together with 7% glucose and O.OOS% CaCIz. If there is a problem of revealing the imaged region with biocytin, slices can be resectioned after fixation. For this, slices are embedded in 3% agarose and resectioned using a vibratome. Freeze-thawing is another method to increase the penetration of reactants like ABC. For this, slices are kept in 30% sucrose until the slices sink to the bottom, and then dipped (in the plastic container with sucrose) into liquid nitrogen. As a variant on this protocol, for fixing slices for EM, prepare a fix solution of 4% paraformaldehyde, O.OS% glutaraldehyde, and IS mL of saturated picric acid in 100 mL of 0.1 MPB. Slice should be kept in this solution for approximately 3 h. Thereafter, slices may be stored in a solution of O.OS M sodium azide in 0.1 MPB. Then follow the protocol above.
Correlated Electron Microscopy Although we can image neuronal structures in live brain slices at quite high resolution with two-photon microscopy, for some questions, such as confinning the existence of a synapse, we find it necessary to use electron microscopy. We have pioneered the combination of two-photon live imaging with serial thin-section electron microscopy to enable us to examine the ultrastructure of dendritic spines and axonal filopodia (Dunaevsky et al., 2001; Tashiro et ai., 2003).
Protocol for Two-Photon/Electron Microscopy Imaging of GFP-Labeled Cells Neurons are transfected with eGFP using biolistics. Slices imaged with two-photon microscopy are fixed with S% glutaraldehyde in PB for 1 h. The slices are then embedded in 3% agar and resectioned at 7S 11m with a vibratome (Technical Products International, St. Louis). After locating the imaged neuronal structure in the sections using a fluorescence microscope, the sections are immunostained with anti-GFP antibody (Roche Diagnostics Corp.) overnight at 4°C, and then with peroxidase-conjugated goal anti-
Morphological Classification of Neurons Using Cluster Analysis One of the problems in classifying cortical neurons is their heterogeneity and the vast number of parameters that can be used for this purpose. These parameters usually encompass a massive array of physiological, morphological, and, most recently, gene expression data. A rigorous approach to classifying cortical neurons must involve a thorough analysis of the structure of the data before one attempts to assign neurons to certain clusters. Principal component analysis (PCA) and cluster analysis (CA) are valuable multivariate data analysis methods that can be used jointly with CA to address these issues (Kozloski et al., 2001).
Protocol for PCNCA As a first step, we perfonn a PCA analysis using Statistica on the variables automatically measured by the Neurolucida program. In a second step, we perfonn cluster analysis (Wards's methods), also using Statistica.
732
Chapter 41 • A. Tashiro et al.
Let X denote an m x n matrix of m cells (cortical neurons) with n measured or computed parameters. The elements of this matrix denoted by xi} represent the value of the jth parameter for the ith neuron. For values of n ~ 3, visualization of data is impossible and thus the overall distribution of cells in their n-dimensional parameter space is not accessible. In the case of cortical neurons the number of dimensions can easily exceed tens or even hundreds with the inclusion of gene expression data. The major advantage of PCA methods is to help reconstruct a new parameter space with (optimally) three dimensions in such a way that it still faithfully represents the data with the minimum loss of information during the process of space transformation. This transformation involves computating eigenvalues (L) and eigenvectors (u) of the n x n correlation matrix (R) of the original data matrix X. The goal is to map the n-dimensional vector describing the parameters of each neuron to a vector with, for example, three dimensions, so that each neuron can be plotted in a three-dimensional (3D) graph. The desired new dimensions, also called principal factors or components, are extracted from the original space through the eigenvalue decomposition of the correlation matrix. This decomposition will provide the eigenvectors (principal components) and its related eigenvalue (total variance accounted for by each component). Principal component scores computed for each principal component give the coordinates of neurons in the reduced space. PCA methods also provide a matrix of factor or principal component loadings that show the correlation value of each original parameter with the newly computed components. These loading values show which parameters contribute significantly (highly positive or negative correlations) to the derived dimensions. It can be seen that, while some original parameters contribute significantly to some components, they contribute very little to others. Meanwhile, some original variables only contribute to principal components that carry a relatively low proportion of the total variance (low relative eigenvalue). These latter parameters are thus found to be less important in the characterization of the data structure. The final stage of classification involves application of CA to the new coordinate values (principal component scores) of the neurons in the reduced space. Appropriate linkage rules for CA can be chosen based on the apparent shape of the clusters as seen in the scatter plot of neurons in the principal component space. Application of PCA before CA allows a rational choice of linkage rules which would lead to a better segregation of clusters.
IMAGE PROCESSING Compensation for the Drift and the Vibration of the Slices As described above, one of our problems is the movement of slices that produces the drift of images during long (> lOmin) time-lapse movies. This drift of structures of interest makes it particularly difficult to observe and analyze changes in morphology and fluorescence intensity. Indeed, spine motility was only discovered in our laboratory after the alignment of time-lapse movies was performed (Dunaevsky et ai., 1999). Therefore, we always compensate for the movement of slices. Although manual alignment of time-lapse movies using the structures which are always stable as references works well, this is time consuming. Instead, we have been using three automatic methods of alignment. Two of them are custom-made programs based on the overlap between the images and the center of mass,
respectively. These programs are written in NIH image and ImageJ software, respectively. The other is commercially available software from AutoQuant Imaging, Inc.
Alignment Based on the Overlap Between Images This alignment program is written as an NIH image macro. The principle of this macro is quite simple. Take two images, project these two images, and compare the average pixel values of the projected image and the original image. The more similar the two original images are and the more overlap they have, the closer the two average pixel values will be.
Protocol for Overlap Alignment Before performing automatic alignment, images are thresholded to highlight neuronal structures. Ideally, the original images should work as well. However, because optical noise changes during timelapse sequences, in practice, thresholded images work better. The macro selects two images and performs multiple iterations of an alignment procedure using the above principle. Each iteration comprises shifting the second original image by 1 to a certain number of pixels in four directions (up, down, left, right), comparing the average pixel values of the projected images and the first image, determining the optimal shift, where the average pixel values of the projected image and the original image are closest, and moving the second image by this optimal shift. From the second iteration on, images are only shifted in three directions because one of four directions is toward a starting point to the previous iteration. To avoid including the blank peripheral area which arises from the shift of the second image, only the area where the first original image and the shifted second image overlap are used to calculate the average pixel values of the projected image and the first original image. These iterations are repeated until an iteration finds that the original position of the second image in the iteration is optimal, or, in other words, the average pixel values of the projected image and the original image are closest in the ending position of the iteration. This alignment macro works very well for most time-lapse sequences of projected images when the images have more area with stable structures than with unstable ones. For example, in the case of the time-lapse imaging of dendritic spine motility, the morphological changes in spines are quite small compared to the stability of the much larger dendritic shafts. Thus, in most cases, the alignment macro works reasonably well. When, in rare cases, alignment does not work completely, we align the movies manually using NIH image software. We use this alignment for z-stacks where the structure of interest shifts between focal planes because of slice movement. Images in z-stacks are not aligned as well as time-lapse movies consisting of projected images since each image in a z-stack is different from the next image (with small overlapping). Although not ideal, the same macro helps to align z-stacks. We check all aligned stacks and correct them manually if the alignment is not good.
Alignment Based on the Center of Mass We have created a different alignment program as a plug-in for ImageJ software using Java programming language. This program is based on the calculating relative positions of the center of mass of the drifting objects.
Imaging Brain Slices • Chapter 41
Protocol for Center of Mass Alignment The coordinates of center of mass are calculated by using the pixel value as a mass. For meaningful calculations, a cut-off value is used in order to prevent including background pixels in the calculation. Our program calculates the center of the first frame in the image stack and uses these coordinates as the reference point for aligning the rest of the image stack (i.e., It considers the first frame as not drifted). The program then calculates coordinates of the center of mass of the each image in the stack. The program then shifts each image in order to align images in such a way that centers of mass of all frames have the same coordinates (they overlap each other). If a region of interest (ROl) is chosen, the program aligns to the center of mass of the ROI. The ROI can be any shape and allows aligning by using center of mass of a selected object, not whole image. This algorithm is quite simple, fast, and works well if the center of mass is always calculated from the same structure. However, drift of the slices can make a new structure appear or a part of a structure disappear from the edge of the image. Because this edge effect can make the center of mass move to completely different positions in different images, purely automatic alignment is sometimes unsuccessful. On the other hand, choosing a stable structure as the ROI for alignment can prevent this artifact.
733
The next logical step in the development of this type of algorithm could be using spatial moments of higher order (center mass coordinates are spatial moments of the first order; and the total mass is spatial moment of zero order). For example, including in the alignment algorithm, spatial moments of the second order and third order will allow compensating not only for drift, but also for image rotation and squeezing.
Online Cell Detection of Neurons Using AM loading, we can simultaneously image over 3000 neurons (Fig. 41.3). To analyze fluorescence changes, such as those indicating Ca2+ concentration, of a large number of individual cells, we need to identify and select all the neurons. As manual selection of this many cells is quite tedious and cannot be performed online, automatic cell detection algorithms were developed in ImageJ (NIH, Bethesda, MD) and Matlab (MathWorks, Natick, MA) [Fig. 41.5(A)].
Protocol for Center of Mass Alignment First, time-lapse movies are collapsed in time, creating a single projected image by averaging the fluorescence of each pixel throughout the recording . This effectively reduces the amount of spatial noise in the image and reveals smaller elements, such as dendrites. Due to the slightly unequal loading of different regions
A
.
\
'. . .. .. ,.... .....I. ,,'. ".. ...''''!. .. ..... " "'. '.... ...• ~ t , •• ~ ••
'. \ .
.,
'
J
• ,
t
•
• ' "
'"
.,
.. .. '
•• f .
.\~.. \."
." " . "
,'
.. .•.. ", •
""
,.
\f. ".
" •
"i t. .~ •• ::.
,.
. .' .. .. ., .
..
...
"
ori gi nal
..:. -
' " , ' _'.,
•
' : ., " ....
, ,
I'
.:..... " , . B
,A.'_,,'.
"
'(1 •
'
..
. "' t \ ••
',
",". "
"
'
.
~
...
,
,
I ... ...
.'
• '
'
"
! ....
gaus ian
wavelet
FIGURE 41.5. Image processing algorithms. (A) Cell detection algorithm. (Left) An original image of a fura-2l oaded acute slice. Note how the staining is not even throughout the image. (Right) An image with outlines of all detected cells. The contours of the cells are drawn. (B) De-noising algorithms. (Upper panels) Images of a dendritic growth cone from a cortical pyramidal neuron. All three images are shown with the same brightness and contrast. (Lower panels) Binarized images. All three images are thresholded with the same pixel value. (Left) Original images. (Center) Gaussian-filtered images. (Right) Images de-noised with wavelet transformation.
734
Chapter 41 • A. Tashiro et al.
of the slice, the average fluorescence level of cells in the same slice often vary, making automatic detection problematic. To account for these spatial fluorescence variations, the value of each pixel is normalized, dividing it by the average fluorescence of a 25 x 25 )Jm square centered at that pixel. In the resulting image, the fluorescence level is almost constant for all cells and is typically between 1.2 and 1.5. Contours can be plotted at a manually chosen value in this range using the Matlab contour drawing algorithm. Every closed contour corresponds to the accurately detected boundary of a fluorescent entity in the imaged field. This method identifies practically all cells, as well as a large number of smaller non-cellular elements in the slice [Fig. 41.3(A)J. By measuring the average fluorescence value of the pixels inside each contour as a function of time, we can quickly reconstruct fluorescence changes of large populations with single-cell resolution . In total , this allows us to measure calcium fluorescence signals from roughly 400 to 1400 entities per slice.
Image De-Noising Using Wavelets The external PMT in our two-photon system produces uncorrelated dark noise, primarily of thermal origin. This type of noise has a strong dependence on applied bias voltage (Majewska et ai., 2000b; www.twophoton.com) . Itis therefore important to correctly choose the bias voltage in order to balance the resulting gain of the PMT versus noise. This type of noise is intrinsically random and does not have specific spectral components. Therefore, it cannot be distinguished from the real signal by classic methods of linear filtration . For example, the widely used mean and Gaussian windows filters are not efficient in terms of removing this type of noise (Fig. 41.5). Indeed, background noise removal is very important for quantitative analysis of thresholded images, and linear filtration of noisy images usually gives artifacts [Fig. 41.5(B)]. As an alternative approach, we use wavelet transformation for the purpose of image de-noising (Lio, 2003). Wavelet transformation is widely used for signal compression and de-noising and represents further development of classic methods of analysi s such as Fourier, Gabor, and short-time Fourier transformation. The wavelet transformation gives full representation of the signal [for one-dimensional (lD) time signals, a correct representation in time and frequency domains; for two-dimensional (2D) images, a representation of spatial frequencies and coordinates]. Whereas Fourier transformation presumes the infinite dimensions of the image space, wavelet transformation is inherently local and gives a better representation of naturally occurring finite-size objects in image.
Protocol for Wavelet De-Noising The general de-noising procedure consists of the following steps: In case of image de-noising, an individual image is represented as 2D array of numbers (pixel values). The wavelet transformation decomposes this 2D signal into wavelet space by using a specified wavelet family. In case of discrete wavelet transformations, it computes the detail coefficient of the signal up to the certain predefined level. For the purpose of signal de-noising, the detail coefficients at all levels of decomposition have to be thresholded. The numerical value of the threshold can be chosen based on the noise model used. There are several methods of thresholding. The practical choice of the method depends on the nature of the signal and the chosen model. Reverse wavelet reconstruction is then performed using the modified detail coefficients, and the filtered image is
regenerated from thresholded detail coefficients by using the inverse wavelet transformation. The latest release of the Wavelet Toolbox (version 2.2) for Matlab (The MathWorks Inc., Natick, MA) provides a variety of ready-to-use tools for wavelet transformation and signal denoising. There is an interactive graphical user interface in Wavelet Toolbox, which simplifies the task of choosing the parameters for signal de-noising and compression. In the simplest case, practical de-noising can be done based on the visual perception of the denoising quality, but also using different recovery criteria (e.g., based on entropy estimation). For large-scale image processing, we created a custom code in Matlab, which performs simple image de-noising based on the chosen model. As a model, we use wavelet decomposition to level 4 with symlet-6 wavelets, and soft thresholding (see more theory of de-noising procedure in Wavelet Toolbox documentation). The script processes raw images in a multi-TIFF format, gives acceptable de-noising, and does not change the quantitative values of intensity in the principal image details [see Fig. 41 .3(B)]. Our algorithm processes each frame individually, therefore processing time depends linearly on the size of the image stack. The algorithm is not memory demanding and turned out to be relatively fast - it takes approximately 5 s to process a 800 x 600 pixel , 16-bit image on a 1.9GHz Pentium 4 PC.
SUMMARY • Brain slices are convenient preparations because they permit the easy manipulation of their environment, access for imaging or electrophysiological equipment, and preservation of threedimensional organization of the brain region studied. • We describe the techniques of live-slice imaging we use in our laboratory, including slice preparation (acute and cultured slices), cell labeling (biolistics, diolistic, calistics, injection with patch electrodes, and AM loading), morphological processing and analysis, imaging procedures (two-photon, second harmonic, and camera imaging), and image processing. • Although in vivo imaging techniques have recently developed in many species, brain-slice imaging has advantages for studying many questions and will be increasingly important for cortical research.
ACKNOWLEDGMENTS We thank the National Eye Institute (EY1l787 and EY 13237), the NINDS (NS40726), the New York STAR Center for High Resolution Imaging of Functional Neural Circuits, the HFSP, and the John Merck Fund for their support.
REFERENCES Agmon , A. , and Connors, B.W., 1991 , Thalamocortical responses of mouse somatosensory (barrel) cortex in vitro, Neuroscience 41 :365- 379. Alger, B.E., Dhanjal , S.S. , Dingledine, R. , Garthwaite, J., Henderson, G., King, G.L., Lipton, P., North, A., Schwartzkroin, P.A., Sears, T.A., Segal, M., Whittingham, T.S ., and Williams, J. , 1984, Brain slice methods, In: Brain Slices (R. Dingledine, ed.), Plenum Press, New York, pp. 381-437. Arnold , D. , Feng, L., Kim , J. , and Heintz, N. , 1994, A strategy for the analysis of gene expression during neural development, Proc. Natl. A cad. Sci. USA 91:9970-9974. Beierlein, M., Fall, c.P., Rinzel , J., and Yuste, R. , 2002, Thalamocortical bursts trigger recurrent activity in neocortical networks: Layer 4 as a frequencydependent gate, 1. Neurosci. 22:9885- 9894.
Imaging Brain Slices • Chapter 41
Brenner, S., 2002, Life sciences: Detective rummages, investigates, The Scientist, 16(6): 15-16. Dunaevsky, A., Blazeski, R., Yuste, R., and Mason, e., 2001, Spine motility with synaptic contact, Nat. Neurosci. 4:685-686. Dunaevsky, A., Tashiro, A, Majewska, A., Mason, e.A., and Yuste, R., 1999, Developmental regulation of spine motility in mammalian CNS, Proc. Natl. Acad. Sci. USA 96:13438-13443. Edwards, FA., Konnerth, A., Sakmann, B., and Takahashi, T., 1989, A thin slice preparation for patch recordings from neurones of mamalian central nervous system, Pflugers Arch. 414:600-612. Feng, G., Mellor, R.H., Bernstein, M., Keller-Peck, e., Nguyen, Q.T., Wallace, M., Nerbonne, J.M., Lichtman, J.W., and Sanes, J.R., 2000, Imaging neuronal subsets in transgenic mice expressing multiple spectral variants ofGFP, Neuron 28:41-51. Franklin, K.B.J., and Paxinos, G., 1997, The mouse brain in stereotaxic coordinates, Academic Press, San Diego, California. Gan, W.B., Grutzendler, J., Wong, W.T., Wong, R.O., and Lichtman, J.W., 2000, Multicolor "DiOlistic" labeling of the nervous system using lipophilic dye combinations, Neuron 27:219-225. Grynkiewicz, G., Poenie, M., and Tsien, R.Y., 1985, A new generation of Ca 2+ indicators with greatly improved fluorescence properties, J. BioI. Chem. 260:3440-3450. Helmchen, F, Imoto, K, and Sakmann, B., 1996, Ca2+ butfering and action potential-evoked Ca2+ signalling in dendrites of pyramidal neurons, Biophys. J. 70: 1069-1081. Honig, M.G., and Hume, R.I., 1989, Dil and DiO: Versatile fluorescent dyes for neuronal labelling and pathway tracing, Trends Neurosci. 334:333-340. Kettunen, P., Demas, J., Lohmann, e., Kasthuri, N., Gong, Y., Wong, R.O., and Gan, W.B., 2002, Imaging calcium dynamics in the nervous system by means of ballistic delivery of indicators, J. Neurosci. Methods 119:37-43. Kozloski, J., Hamzei-Sichani, E, and Yuste, R., 2001, Stereotyped position of local synaptic targets in neocortex, Science 293:868-872. Lechleiter, J.D" Lin, D.T., and Sieneart, I., 2002, Multi-photon laser scanning microscopy using an acoustic optical deflector, Biophys. J. 83:2292-2299. Lewis, A., Khatchatouriants, A., Treinin, M., Chen, Z., Peleg, G., Friedman, N., Boucvitch, 0., Rothman, Z., Loew, L., and Sheves, M., 1999, Second harmonic generation of biological interfaces: Probing membrane proteins and imaging membrane potential around GFP molecules at specific sites in neuronal cells of C. elegans, Chem. Phys. 245: 133-144. Lio, P., 2003, Wavelets in bioinformatics and computational biology: state of art and perspectives, Bioin/ormatics 19:2-9. Lo, D.e., McAllister, AK, and Katz, L.e., 1994, Neuronal transfection in brain slices using particle-mediated gene transfer, Neuron 13: 1263-1268. Lohmann, e., Myhr, K.L., and Wong, R.O., 2002, Transmitter-evoked local calcium release stabilizes developing dendrites, Nature 418: 177-181. Majewska, A., Tashiro, A., and Yuste, R., 2000a, Regulation of spine calcium compartmentalization by rapid spine motility, J. Neurosci. 20:8262-8268.
735
Majewska, A, Yiu, G., and Yuste, R., 2000b, A custom-made two-photon microscope and deconvolution system, Eur. J. Physiol. 441 :398-409. Nikolenko, v., Nemet, B., and Yuste, R., 2003, A custom two-photon and second harmonic microscope, Methods. 30(1):3-15. Peterlin, Z.A., Kozloski, J., Mao, B., Tsiola, A., and Yuste. R., 2000, Optical probing of neuronal circuits with calcium indicators, Proc. Natl. Acad. Sci. USA 97:3619-3624. Sheenen, WJJ.M., Makings, L.R., Gross, L.R., Pozzan, T., and Tsien, R.Y., 1996, Photodegration of indo-I and its effect on apparent Ca concentrations, Chem. Bioi. 3:765-774. Smetters, D.K, Majewska, A, and Yuste, R., 1999, Detecting action potentials in neuronal populations with calcium imaging, Methods 18:215-221. Tashiro, A., and Yuste, R., 2004, Regulation of dendritic spine motility and stability by Rac1 and Rho kinase: evidence for two forms of spine motility, Mol. Cell Neurosci. 26(3):429-440. Tashiro, A., Dunaersky, A., Blazeski, R., Mason, e.A., and Yuste, R., 2003, Bidirectional regulation of hippocampal mossy fiber filopodial motility in kainate receptors: a two-step model of synaptogenesis, Neuron, 38(5):773-784. Tashiro, A., Minden, A., and Yuste, R., 2000, Regulation of dendritic spine morphology by the Rho family of small GTPases: Antagonistic roles of Rac and Rho, Cerebral Cortex 10:927-938. Tsai, P.S., Nishimura, N., Yoder, EJ., White, A, Doluick, E., and Kleinfeld, D., 2002, Principles, design and construction of a two-photon scanning microscope for in vitro and in vivo studies, in Methods for in vivo Optical Imaging (R. Frostig, ed.), CRC Press, pp. 113-171. Tsien, R.Y., 1989, Fluorescent probes of cell signaling, Ann. Rev. Neurosci. 12:227-253. Tsiola, A., and Yuste, R., 2003, Classification of neurons in the mouse primary visual cortex, J. Compo Neurol. 461 (4 ):415-428. Yuste, R., 2000a, Loading populations neurons in slices with AM calcium indicators, In: Imaging Neurons: A Laboratory Manual (R. Yuste, F Lanni, A. Konnerth, eds.), Cold Spring Harbor Press, Cold Spring Harbor, New York, pp. 34.31-34.39. Yuste, R., 2000b, Imaging Neurons: A Laboratory Manual, Cold Spring Harbor Press, Cold Spring Harbor, New York. Yuste, R., and Denk, W., 1995, Dendritic spines as basic units of synaptic integration, Nature 375:682-684. Yuste, R., and Katz, L.e., 1989, Transmitter-induced changes in intracellular free calcium in brain slice of developing neocortex, Soc. Neurosci. Abstr. 4.5:2. Yuste, R., and Katz, L.e., 1991, Control of postsynaptic Ca2+ influx in developing neocortex by excitatory and inhibitory neurotransmitters, Neuron 6:333-344. Zilles, K., and Wree, A., 1985, Cortex: A real and laminar structure, In: The Rat Nervous System (G. Paxinos, ed.), Academic Press, Sydney, pp. 375-415.
42
Fluorescent Ion Measurement Mark B. Cannell and Stephen H. Cody
INTRODUCTION It is likely that cell behavior is regulated by the summation of subcellular transduction events and these are generally linked, nonlinearly, to subsequent stages of signal transduction. Hence, a full understanding of signal transduction can only arise from the analysis of subcellular behavior and thi s is one reason why microscopic imaging has become so important. To carry this idea further, since the cell is a three-dimensional (3D) structure, methods that provide 3D imaging (as in confocal and multi-photon methods) are likely to provide better insight than even widefield/deconvolution imaging which is, itself, an improvement on either widefield or simple photometric methods . To analyze cell behavior, we need to measure something, and then after a period of time, we need to measure it again. This often implies the use of probes that can report the behavior of interest. In the case of intracellular ions, because calcium has been found to be a ubiquitous second messenger system, it is not surprising that many methods for measuring it have been developed, each with unique advantages and disadvantages. Here, we will tend to focu s on calcium and pH measurements in living cells, but the ideas and problems that will be discussed are generally applicable to all methods that employ optical probes of function; I the differences between approaches and techniques reside in the degree of applicability rather than the presence or absence of any single problem. The continuing effort to improve signal-to-noise ratio, specificity, and resolution (spatial and temporal) of these techniques has brought wonderful new insight into our understanding of cell function . Each major new advance seems to have taken approximately a decade to achieve and each decade has improved spatiotemporal resolution by something like 2 orders of magnitude (for constant signal-to-noise). Thus, from the development and application of aequorin to giant nerve and muscle cells in the 1960s (which supplanted tissue-level chemical methods), we are now measuring subcellular events on the scale of microns with fluorescent probes. Resolution in both space and intensity is currently limited by the quantal nature of light itself. Further major advances in resolution may be confounded by the problems of excessive buffering (or perturbation) of responses by the probe and by cell damage caused by the excitation light. This pessimism is tempered by the fact that a little more than two decades ago, it was thought unlikely that it would ever be possible to resolve the microscopic gradients of calcium that exist inside the muscle sarcomere during activation [which at that time were only predicted on the basis of computer modeling (Cannell and Allen, 1984»). However, the early
1990s saw the potential of the laser-scanning confocal microscope (LSCM) applied to Roger Tsien 's dyes (e.g., for reviews see special edition of Cell Calcium 11 (2-3), Williams (1993). The discovery of calcium sparks (Cheng et al., 1993; Cannell et aI., 1994; see Fig. 42.1) while one of the authors (MBC) was developing methods needed to look for some spatial non-unforrnity in the cardiac Ca++ transient (during a sabbatical in Dr. Lederer's laboratory) was completely unexpected and would probably not have occurred without the LSCM. Imagine the excitement at seeing these events rise out of the background noise displayed on a dimmed computer screen in a blacked-out laboratory. There is something compelling about imaging responses under carefully-optimized experimental conditions, perhaps related to the immediacy and tangibility of the results . Seeing is believing and imaging, in its many modalities, allows us to see otherwise invisible and unexpected cell signaling events as they occur. Provided artifacts are recognized and rejected, new insights develop as we try to understand the cellular basis of what has been observed. Therein lies a caveat: Poor science inevitably results from not appreciating the limitations of the method. Having a new instrument or method does not remove the possibility that artifactual results may be obtained from excessive perturbation of the experimental system and measuring something about a system inevitably involves some sort of perturbation. Both light and probes can easily damage cells or modify their responses. As we seek to resolve ever smaller and faster events, the perturbation of natural responses becomes larger. In general, experimental problems arise from: 1. The quantal nature of light (it is not possible to directly measure a response that is associated with a change of signal of less than a single photOn/pixel). 2. Buffering the cell response (any probe must interfere with the cell processes by its interaction with the chemical process of interest). 3. Limited time resolution (all chemical reactions take time before the products become available for analysis). 4. Damage to the cell either by the toxicity of the probe or damage by the light irradiating the system. 5. Limited spatial resolution associated with the wavelength of light itself. In this chapter, we will discuss how these problems relate to the measurement of calcium levels inside living cells, although the discussion applies to all other optical measurements as well.
The limiting Case I
Functional probes also exist for cAMP, Na, K, Mg, and many other ions (see Chapters 16 and 17, this volume) .
If we confine our discussion to optical resolution, the smallest cell volume that can be observed in practical light microscopy is about
M ark B. Cannell • University of Au ckland, Auckland, New Zealand Stephen H. Cody. Ludwig Institute for Cancer Research, Royal Melbourne Hospital, Parkville Victori a, Au strali a
736
Handbook of Biological Confocal Mic roscopy , Third Edition, edited by James B. Pawley, Springer Science+Business Media. LLC, New York, 2006.
Fluorescent Ion Measurement • Chapter 42
FIGURE 42.1. Two consecutive confocal images of a fluo-3 loaded cardiac myocyte. The areas of increased fluorescence are spontaneous Ca++ sparks. Note the scale of this image. The lower panel shows sequential confocal images of spontaneous Ca++ waves propagating and colliding within a cardiac cell. Scale bar 20flm. Frame scan time, O.7s.
0.06 fl (i.e., 0.25 x 0.25 x 0.5/lm). If a single ion resides in this volume (high) the concentration is about IIlMIL. In the case of calcium, this is a high physiological level. However, we cannot directly measure an ion in solution inside a cell: it must be bound to a probe whose physical properties are altered by that binding. If the ion is always bound to the probe then we must detect a single probe molecule in that volume (note that we have also reduced the level of free calcium to zero at the same time - a serious problem
737
that will be discussed later). It is unlikely that we could detect the bound probe by a change in absorbance because this would be akin to directly imaging a single molecule in transmission light microscopy and Beers Law shows the futility of such an approach. Instead, we could use a change in probe fluorescence and current fluorochromes have a fluorescence lifetime of 1 to lOns. This implies that, with the maximum possible excitation rate, we might expect to generate 108 photons/s/molecule. This large number rapidly declines when we consider the detection efficiency of the instrument. A good water-immersion dipping objective lens may collect about 10% of the emitted light (as limited by the numerical aperture of the lens and bathing solution refractive index) and after accessory optics to separate the excitation light from the emitted light, focus the light on a confocal pinhole and detector, typically I 00 ~ allows determination of F m",. Adding Mn++ quenches the dye fluorescence, allowing a determination of cell autofluorescence. It is assumed that the latter was not affected by Mn++ and separate experiments should be carried out to show whether this is the case.
and such effects have been observed for several indicators (see above). The possible errors caused by not using a ratiometric approach are illustrated in Figure 42.9 where a proximal tubule has been labeled with pH indicator BCECF and imaged at a single wavelength. The low intensity fluorescence evident in the brush border, could be caused by one of three factors: a lower pH in the brush border, a lower concentration of dye in this region, or the fact that there is more unstained, extracellular space in the brush border (path-length error). However, when a similar preparation is labeled with the ratiometric SNARF-l (Fig. 42.10), the ratio image shows the true pH distribution map and it becomes evident that apparent pH differences are simply dye-concentration artifacts. Similar artifacts were also observed in a single wavelength study that used Fura-2 in heart cell nuclei (Wier et ai., 1987). Application of the calibration equations also assumes that the indicator is sufficiently selective not to form complexes with other ions to any significant extent. Although this may be a questionable assumption, the fluorescent indicators generally perform quite well in this respect. This is particularly true for Ca++ indicators based on the BAPTA molecule, for example, fura-2 and indo-l (Grynkiewicz et at., 1985), which are more selective against Mg++ and protons than are (say) any of the metallochromic Ca++ indicators, where competing effects from such ions pose a major calibration problem (Thomas, 1991). In the case of pH indicators, selectivity is quite good so intracellular pH may be simply manipulated by adding nigericin and a CO 2 - bicarbonate-based buffer system. For example, in the experiments shown in Figure 42.4 a high potassium solution was used (to zero the membrane potential) and the pH of the superfusate measured with a pH electrode. By using three, premixed gasses (100% O 2 , 5%COzl95% O 2 and 100% CO 2 ) to bubble the superperfusate (which included 18.5 mM NaHC0 3 ), the intracellular pH changes could be calibrated without needing multiple solution changes. This increased the reproducibility of the intracellular calibrations as the specimen did not need to be manipulated unnecessarily (Fig. 42.11). A disadvantage with this method is that Ca++ must be removed if alkaline pH calibrations are needed, otherwise the Ca++ in the perfusate will precipitate with the bicarbonate. When the interference effects from other ions are significant, the simplest approach is to calibrate the indicator in a solution that contains appropriate concentrations of those ions. Of course, such calibrations only apply so long as the concentrations of the inter-
FIGURE 42.8. Cultured monolayer of rabbit proximal tubule cells stained with the acetoxymethyl ester of SNARF-l. (Al Wavelengths greater than 595 nm. (B) Wavelengths lower than 595 nm . C: A ratio image of A divided by B (>595 nml595nm) provides a pH distribution map that is independent to photobleaching, dye distribution and path length artifacts. Bar = 5 11m. (Cody and Williams 1999.)
Fluorescent Ion Measurement • Chapter 42
Intracellular SNARF-l Standard Curve 1.6 Excitation = 488nm
E a5
I
1.4
Ol LO
745
mapping projects were conducted by SHC in the Department of Physiology, University of Melbourne under the mentorship of Prof. David A. Williams and with a major contribution from Mr. Philip N. Dubbin. Major contributors to the muscle fatigue project were made by Dr. Gordon S. Lynch and Dr. Noel D. Duncan. Kidney tubule preparations kindly supplied by Prof. Peter Harris, University of Melbourne.
.t:.
E
~ 1.2 Ol
REFERENCES
LO
:::!...2
til a: 1.0
0.8 6.8
6.6 i
25
7.0 i
i
20 15
7.2 i
10
7.4 pH i
i
5 4 % CO 2
7.6 i
3
7.8 i
2
8.0
8.2
o
FIGURE 42.11. A typical intracellular calibration curve of SNARF-1. Muscles stained with SNARF-l and treated with the hydrogen ion ionophore nigericin (50 1lM). The pH of the modified Kreb's solution was altered by bubbling with different concentrations of CO,. The extracellular pH was monitored with a pH electrode, and dual channel ratiometric images of intracellular SNARF-l fluorescence were recorded. (Cody et al., previously unpublished.)
fering ions in the cell do not change significantly during the experimental procedures and autofluorescence remains constant. As accurate calibration remains problematic and time consuming to do properly with appropriate controls, it may be better to design experiments where one need only determine the direction of change rather than measure the absolute ion levels.
CONCLUSION The new fluorescent probes coupled with the LSCM allow unprecedented sensitivity and resolution for the detection of ion levels in four dimensions within living cells. The current range of dyes and the equipment available has enabled resolution of microscopic signaling on the scale of 100 photons/pixel. Although such photon-counting methods place severe limits on acquisition speed, because dense bodies are stable structures and, hence, essentially motionless, sampling of discrete points is an adequate substitute for full-field sampling, and this reduces the likelihood of phototoxicity (Breusegem, 2002). Although the use of FRET in embryos is in its infancy, these results suggest that in the near future this imaging tool, the use of which was once exclusive to cell biologists working on flat cells in tissue culture, will soon be available to developmental biologists.
CONCLUSIONS: A BRIGHT FUTURE FOR 3D IMAGING OF LIVING EMBRYOS Few disciplines within biology have benefited more from improvements in confocal, multi-photon, and related technologies than developmental biology. As more exotic probes are developed for use in living cells, particularly those that can be genetically encoded, the imaging of embryos will continue to become more sophisticated. As each new probe is developed, there may be new challenges for developmental biologists, but based on current imaging modalities, several generalizations will probably continue to hold true. First, although there are times when more expensive equipment, such as multi-photon microscopes, 4D deconvolution microscopes, or high-speed imaging approaches are necessary, empirical tests of viability using off-the-shelf confocal equipment should be performed first before such equipment is assumed to be necessary.
While similar experiments are now being performed using multiphoton microscopy, numerous published experiments indicate that both standard single-beam confocal microscopes and disk-scanning microscopes can often be used to image living embryos in four dimensions without the need for additional equipment. Second, as computer processing speed continues to increase, deconvolution and 3D projection of 4D datasets will become more routine. This can only improve the ability of the developmental biologist to visualize complex processes in four dimensions. The more routine use of multi-wavelength probes will likewise improve the ability of the developmental biologist to perceive and comprehend the complex beauty of embryogenesis. Finally, although there has typically been a lag between the application of new imaging modalities in cultured cells and their subsequent use in embryos, the history of the field suggests that eventually many of the approaches first worked out in cultured cells will be adapted for use in embryos. Extension of the successful combination of microsurgical methods and confocal and multi-photon methods presented in this chapter, which effectively make the embryo flatter, will likely allow techniques such as FRET to be used routinely in embryos in the near future. Ultimately, 4D live imaging of fluorescent probes in embryos will cease to be the preserve of the specialist, and will become a part of the standard repertoire of the developmental biologist.
ACKNOWLEDGMENTS I thank the editor of this volume and my wife, Susie, for their remarkable patience as this work was completed. I am grateful to Brian Burkel and Bill Bement for sharing unpublished data regarding PA-GFP, and to Sophie Breusegem and Bob Clegg for sharing unpublished results regarding FRET in C. elegans. Tim Walston, Mark Sheffield, and Chris Lockwood provided movies and images used in this work, and other members of the Hardin laboratory provided moral support. This work was supported by NIH grant GM058038 and NSF grant IBN-0l12803, awarded to the author. CLSM footage was generated using a Bio-Rad 1024 confocal microscope originally purchased by NSF grant 9724515 awarded to J. Pawley. Movies associated with this chapter will be available on the Web site associated with this publication. http://www.springer.comlO-387-25921-X.
REFERENCES Beck, J.e., Murray, J.A., Willows, A.O., and Cooper, M.S., 2000, Computerassisted visualizations of neural networks: Expanding the field of view using seamless confocal montaging. 1. Neurosci. Methods 98:155-163. Bement, W.M., Sokac, A.M., and Mandato, e.A., 2003, Four-dimensional imaging of cytoskcletal dynamics in Xenopus oocytes and eggs, Differentiation 71:518-527. Benink, H.A., and Bement, W.M., 2005, Concentric zones of active RhoA and Cdc42 around single cell wounds, 1. Cell BioI. 168:429-439. Benink, H.A., Mandato, e.A., and Bement, W.M., 2000. Analysis of cortical flow models in vivo, Mol. Bioi.} Cell 11 :2553-2563. Bloor, l.W., and Kiehart, D.P., 2002, Drosophila RhoA regulates the cytoskeleton and cell-cell adhesion in the developing epidermis, Development 129:3173-3183. Breusegem, S.Y., 2002, In vivo investigation of protein interactions in C. elegans by Forster resonance energy transfer microscopy, Ph. D. dissertation, University of Illinois, Urbana-Champaign. Campbell, R.E., Tour, 0., Palmer, A.E., Steinbach, P.A., Baird, G.S., Zacharias, D.A., and Tsien, R.Y., 2002, A monomeric red fluorescent protein, Proc. Natl. Acad. Sci. USA 99:7877-7882.
Confocal and Multi-Photon Imaging of Living Embryos • Chapter 43 Centonze, Y.E., and White, J.G., 1998, Multiphoton excitation provides optical sections from deeper within scattering specimens than confocal imaging, Biophys. 1. 75:2015-2024. Chalfie, M., Tu, Y., Euskirchen, G., Ward, W.W., and Prasher, D.e., 1994, Green fluorescent protein as a marker for gene expression, Science 263:802-805. Chapman, S.C., Collignon, J., Schoenwolf, G.e., and Lumsden, A., 2001, Improved method for chick whole-embryo culture using a filter paper carrier, Dev. Dyn. 220:284-289. Chong, FK., Coates, e.G., Denvir, DJ., McHale, N.G., Thornbury, K.D., and Hollywood, M.A., 2004, Optimization of spinning disk confocal microscopy: Synchronization with the ultra-sensitive EMCCD, Proc. SPIE 5324:65-76. Chudakov, D.M., Belousov, Y.Y., Zaraisky, A.G., Novoselov, Y.Y., Staroverov, D.B., Zorov, D.8., Lukyanov, S., and Lukyanov, K.A., 2003, Kindling fluorescent proteins for precise in vivo photo labeling, Nat. Biotechnol. 21:191-194. Clegg, R.M., Holub, 0., and Gohlke, e., 2003, Fluorescence lifetime-resolved imaging: Measuring lifetimes in an image, Methods Enzymol. 360:509542. Clendenon, J.L., Phillips, C.L., Sandoval, RM., Fang, S., and Dunn, K.W., 2002, Voxx: A PC-based, near real-time volume rendering system for biological microscopy, Am. 1. Physiol. Cell Physiol. 282:C213-C218. Cockell, M.M., Baumer, K., and Gonczy, P., 2004, lis-1 is required for dyneindependent cell division processes in C. elegans embryos, 1. Cell Sci. 117:4571-4582. Concha, M.L., and Adams, R.J., 1998, Oriented cell divisions and cellular morphogenesis in the zebrafish gastrula and neurula: A time-lapse analysis, Development 125:983-994. Cooper, M.S., D' Amico, L.A., and Henry, e.A., 1999, Confocal microscopic analysis of morphogenetic movements, Methods Cell Bioi. 59: 179-204. Cooper, M.S., Szeto, D.P., Sommers-Herivel, G., Topczewski, J., SolnicaKrezel, L., Kang, H. e., Johnson, I., and Kimelman, D., 2005, Visualizing morphogenesis in transgenic zebrafish embryos using BODIPY TR methyl ester dye as a vital counterstain for GFP, Dev. Dyn. 232:359-368. Czirok, A., Rongish, B.J., and Little, e.D., 2004, Extracellular matrix dynamics during vertebrate axis formation, Dev. BioI. 268:111-122. Czirok, A., Rupp, P.A., Rongish, B.J., and Little, e.D., 2002, Multi-field 3D scanning light microscopy of early embryogenesis, 1. Microsc. 206:209-217. DeGrauw, e.J., Frederix, P.L.T.M., and Gerritsen, H.e., 2002, Aberrations and penetration in in-depth confocal and two-photon-excitation microscopy, In: Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances (A. Diaspro, ed.), Wiley-Liss, New York, pp. 153-169. Dubertret, B., Skourides, P., Norris, D.J., Noireaux, Y., Brivanlou, A.H., and Libchaber, A., 2002, In vivo imaging of quantum dots encapsulated in phospholipid micelles, Science 298: 1759-1762. Edwards, K.A., Demsky, M., Montague, R.A., Weymouth, N., and Kiehart, D.P., 1997, GFP-moesin illuminates actin cytoskeleton dynamics in living tissue and demonstrates cell shape changes during morphogenesis in Drosophila, Dev. Bioi. 191: 103-117. Galbraith, J.A., and Terasaki, M., 2003, Controlled damage in thick specimens by multiphoton excitation, Mol. Bioi. Cell 14: 1808-1817. Gao, X., Yang, L., Petros, J.A., Marshall, FF, Simons, J.W., and Nie, S., 2005, In vivo molecular and cellular imaging with quantum dots, Curr. Opin. Biotechnol. 16:63-72. Gimlich, R.L., 1991, Fluorescent dextran clonal markers, Methods Cell BioI. 36:285-297. Girdham, e.H., and O'Farrell, P.H., 1994, The use of photoactivatab1e reagents for the study of cell lineage in Drosophila embryogenesis, Methods Cell BioI. 44:533-543. Gomez, TM., and Spitzer, N.C., 1999, In vivo regulation of axon extension and pathfinding by growth-cone calcium transients, Nature 397:350-355. Grevengoed, E.E., Loureiro, JJ., Jesse, T.L., and Peifer, M., 2001, Abelson kinase regulates epithelial morphogenesis in Drosophila, .l. Cell BioI. 155:1185-1198. Hadjantonakis, A.K., Dickinson, M.E., Fraser, S.E., and Papaioannou, Y.E., 2003, Technico1our transgenics: Imaging tools for functional genomics in the mouse, Nat. Rev. Genet. 4:613-625.
767
Hammond, A.T, and Glick, B.S., 2000, Raising the speed limits for 4D fluorescence microscopy, Traffic 1:935-940. Heid, PJ., Raich, W.B., Smith, R, Mohler, W.A., Simokat, K., Gendreau, S.B., Rothman, J.H., and Hardin, J., 2001, The zinc finger protein DIE-I is required for late events during epithelial cell rearrangement in C. elegans, Dev. BioI. 236: 165-180. Heid, PJ., Voss, E., and Soli, D.R, 2002, 3D-DIASemb: A computer-assisted system for reconstructing and motion analyzing in 4D every cell and nucleus in a developing embryo, Dev. BioI. 245:329-347. Hell, S., Reiner, G., Cremer, e., and Stelzer, E.H.K., 1993, Aberrations in confocal fluorescence microscopy induced by mismatches in refractive index, 1. Microsc. 169:391-405. Henry, e.A., Hall, L.A., Burr Hille, M., Solnica-Krezel, L., and Cooper, M.S., 2000, Somites in zebrafish doubly mutant for knypek and trilobite form without internal mesenchymal cells or compaction, Curr. BioI. 10: 10631066. Huisken, J., Swoger, J., Del Bene, F, Wittbrodt, J., and Stelzer, E.H., 2004, Optical sectioning deep inside live embryos by selective plane illumination microscopy, Science 305:1007-1009. Jacinto, A., Wood, W., Balayo, T, Turmaine, M., Martinez-Arias, A., and Martin, P., 2000, Dynamic actin-based epithelial adhesion and cell matching during Drosophila dorsal closure, Curro BioI. 10:1420-1426. Jaffe, L.A., and Terasaki, M., 2004, Quantitative microinjection of oocytes, eggs, and embryos, Methods Cell BioI. 74:219-242. Jonkman, J.E.N., and Stelzer, E.H.K., 2002, Resolution and contrast in confocal and two-photon microscopy, In: Confocal and Two-Photon Microscopy: Foundations, Applications, and Advances (A. Diaspro, ed.), Wiley-Liss, New York, pp. 101-125. Keller, R., 2002, Shaping the vertebrate body plan by polarized embryonic cell movements, Science 298:1950-1954. Kerr, R., Lev-Ram, V., Baird, G., Vincent, P., Tsien, R.Y., and Schafer, W.R., 2000, Optical imaging of calcium transients in neurons and pharyngeal muscle of C. elegans, Neuron 26:583-594. Kilian, B., Mansukoski, H., Barbosa, Fe., Ulrich, F, Tada, M., and Heisenberg, e.P., 2003, The role of PptlWnt5 in regulating cell shape and movement during zebrafish gastrulation, Mech. Dev. 120:467-476. Koppen, M., Simske, J.S., Sims, P.A., Firestein, B.L., Hall, D.H., Radice, A.D., Rongo, e., and Hardin, J.D., 2001, Cooperative regulation of AJM-I controls junctional integrity in Caenorhabditis elegans epithelia, Nat. Cell BioI. 3:983-991. Kozlowski, D.l., Murakami, T, Ho, R.K., and Weinberg, E.S., 1997, Regional cell movement and tissue patterning in the zebrafish embryo revealed by fate mapping with caged fluorescein, Biochem. Cell Bioi. 75:551562. Labbe, J.e., McCarthy, E.K., and Goldstein, B., 2004, The forces that position a mitotic spindle asymmetrically are tethered until after the time of spindle assembly, 1. Cell Bio/. 167:245-256. Langenberg, T, Brand, M., and Cooper, M.S., 2003, Imaging brain development and organogenesis in zebrafish using immobilized embryonic explants, Dev. Dyn. 228:464-474. Lippincott-Schwartz, J., and Patterson, G.H., 2003, Development and use of fluorescent protein markers in living cells, Science 300:87-91. Lippincott-Schwartz, J., Allan-Bonnet, N., and Patterson, G.H., 2003, Photobleaching and photoactivation: Following protein dynamics in living cells, Nat. Cell Bioi. (Suppl.):S7-S 14. Megason, S.G., and Fraser, S.E., 2003, Digitizing life at the level of the cell: High-performance laser-scanning microscopy and image analysis for in toto imaging of development, Mech. Dev. 120:1407-1420. Michalet, X., Pinaud, FF, Bentolila, L.A., Tsay, J.M., Doose, S., Li, J.J., Sundaresan, G., Wu, A.M., Gambhir, S.S., and Weiss, S., 2005, Quantum dots for live cells, in vivo imaging, and diagnostics, Science 307:538-544. Miyawaki, A., Griesbeck, 0., Heim, R., and Tsien, R.Y., 1999, Dynamic and quantitative Ca2+ measurements using improved came1eons, Proc. Natl. A cad. Sci. USA 96:2135-2140. Miyawaki, A., Sawano, A., and Kogure, T, 2003, Lighting up cells: Labelling proteins with fluorophores, Nat. Cell BioI. (Supp\.):SI-S7. Mohler, W.A., 1999, Visual reality: using computer reconstruction and animation to magnify the microscopist's perception, Mol. BioI. Cell 10:3061-3065.
768
Chapter 43 • J. Hardin
Mohler, W.A., Simske, J.S., Williams-Masson, E.M., Hardin, J.D., and White, J.G., 1998, Dynamics and ultrastructure of developmental cell fusions in the Caenorhabditis elegans hypodermis, Curro BioI. 8:1087-1090. Molecular Probes, 2004, The Handbook - A Guide to Fluorescent Probes and Labeling Technologies. Carlsbad, CA: Invitrogen Corp. Molyneaux, K.A., Zinszner, H., Kunwar, P.S., Schaible, K., Stebler, J., Sunshine, M.J., O'Brien, W., Raz, E., Littman, D., Wylie, C., and Lehmann, R, 2003, The chemokine SDFlICXCL12 and its receptor CXCR4 regulate mouse germ cell migration and survival, Development 130:4279-4786. Oegema, K., Desai, A., Rybina, S., Kirkham, M., and Hyman, A.A., 2001, Functional analysis of kinetochore assembly in Caenorhabditis elegans, J. Cell BioI. 153:1209-1226. Paddy, M.R, Saumweber, H., Agard, D.A., and Sedat, J.w., 1996, Timeresolved, in vivo studies of mitotic spindle formation and nuclear lamina breakdown in Drosophila early embryos, J. Cell Sci. 109:591-607. Patterson, G.H., and Lippincott-Schwartz, J., 2002, A photoactivatable GFP for selective photolabeling of proteins and cells, Science 297: 1873-1877. Periasamy, A., Skoglund, P., Noakes, c., and Keller, R, 1999, An evaluation of two-photon excitation versus confocal and digital deconvolution fluorescence microscopy imaging in Xenopus morphogenesis, Microsc. Res. Tech. 47:/72-181. Pertz, 0., and Hahn, K.M., 2004, Designing biosensors for Rho family proteins - Deciphering the dynamics of Rho family GTPase activation in living cells, J. Cell Sci. 117:1313-1318. Poteryaev, D., Squirrell, J.M., Campbell, J.M., White, J.G., and Spang, A., 2005, Involvement of the actin cytoskeleton and homotypic membrane fusion in ER dynamics in C. elegans, Mol. BioI. Cell. 16:2139-2153. Rueden, c., Eliceiri, K.W., and White, J.G., 2004, VisBio: A computational tool for visualization of multidimensional biological image data, Traffic 5:411-417. Sekar, R.B., and Periasamy, A., 2003, Fluorescence resonance energy transfer (FRET) microscopy imaging of live cell protein localizations, J. Cell Bioi. 160:629-{)33. Shaner, N.C., Campbell, RE., Steinbach, P.A., Giepmans, B.N., Palmer, A.E., and Tsien, R.Y., 2004, Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein, Nat. Biotechnol. 22:1567-1572.
Skop, A.R, Bergmann, D., Mohler, W.A., and White, J. G., 200], Completion of cytokinesis in C. elegans requires a brefeldin A-sensitive membrane accumulation at the cleavage furrow apex, Curro BioI. 11:735-746. Squirrell, J.M., Wokosin, D.L., White, J.G., and Bavister, B.D., 1999, Longterm two-photon fluorescence imaging of mammalian embryos without compromising viability, Nat. Biotechnol. 17:763-767. Strickland, L., von Dassow, G., Ellenberg, J., Foe, v., Lenart, P., and Burgess, D., 2004, Light microscopy of echinoderm embryos, Methods Cell BioI. 74:371-409. Strome, S., Powers, J., Dunn, M., Reese, K., Malone, C. J., White, J., Seydoux, G., and Saxton, W., 2001, Spindle dynamics and the role of gamma-tubulin in early Caenorhabditis elegans embryos, Mol. BioI. Cell 12:17511764. Terasaki, M., 1998, Imaging of echinoderm fertilization, Mol. BioI. Cell 9:1609-1612. Terasaki, M., and Jaffe, L.A., 2004, Labeling of cell membranes and compartments for live cell fluorescence microscopy, Methods Cell BioI. 74:469489. Thomas, c., DeVries, P., Hardin, J., and White, J., 1996, Four-dimensional imaging: Computer visualization of 3D movements in living specimens, Science 273:603-607. Thomas, C.F, and White, J.G., 2000, Acquisition, display, and analysis of digital three-dimensional time-lapse (four-dimensional) data sets using free software applications, Methods Mol. BioI. 135:263-276. Ulrich, F, Concha, M.L., Heid, P.J., Voss, E., Witzel, S., Roehl, H., Tada, M., Wilson, S.W., Adams, R.J., Soil, D.R, and Heisenberg, c.P., 2003, SlbfWntll controls hypoblast cell migration and morphogenesis at the onset of zebrafish gastrulation. Development 130:5375-5384. Voie, A.H., Burns, D.H., and Spelman, FA., 1993, Orthogonal-plane fluorescence optical sectioning: three-dimensional imaging of macroscopic biological specimens. J. Microsc. 170:229-236. Wallrabe, H., and Periasamy, A., 2005, Imaging protein molecules using FRET and FUM microscopy, Curro Opin. Biotechnol. 16:19-27. Wilson, P., and Keller, R, 1991, Cell rearrangement during gastrulation of Xenopus: Direct observation of cultured explants, Development 112:289300. Wood, W., and Jacinto, A., 2004, Imaging cell movement during dorsal closure in Drosophila embryos, Methods Mol. Bioi. 294:203-210.
44
Imaging Plant Cells Nuno Moreno, Susan Bougourd, Jim Haseloff, and Jose A. Feij6
INTRODUCTION The history of imaging plant cells is intimately related to the very development of microscopes and microscopical techniques. Some of the early microscopists made extensive use of plant specimens, and Hooke's description of cork microstructure (Fig. 44.1) will stand in the imagination of many as the structural foundation for the cell theory. There are several reasons for this to have happened: in many respects plant tissues are easier to deal with, easier to slice and peel to the necessary thickness for observation, they have more water, and consequently are less optically dense than many other tissues, often they are naturally pigmented and the cells are usually larger. Above all, the existence of a skeletal cell wall composed of cellulose and other molecules makes plant cells extraordinarily geometric and highly regulated in their structural features. In many instances, these structural components of the plant cell form the basis of its function, making microscopical analysis a recurrent method for cell and developmental biology research. Since the 17th century, many gifted microscopists have given cumulative accounts of the different levels of plant cell structure. Perhaps the first truly specialized plant microscopist was Grew (1673), who initiated the first systematic study of plant microanatomy. Robert Brown, Amici, Schleiden, and Nawaschin are among the many notables who contributed to the description of some of the fundamental biological features of plants, such as their reproductive cycles, mostly using microscopes as tools. More recently, the seminal textbooks of Esau (1977a,b) and Fahn (1990) systematized definitively the histological and cellular features of higher plants. Plant cell ultrastructure has subsequently been described and systematized in detail in many textbooks and atlases, such as the ones produced by Gunning and Steer (1996). This effort was recently complemented by an extensive review on CD-ROM that stands, and probably will stand for a long time, as a central reference for anyone seeking information on microscopical data concerning plant cells (Gunning, 2003; http:// www.plantcellbiologyoncd.com). The developments in technology and reagents that brought fluorescence-based methods to microscopy have also become prevalent in plant cell biology (Lloyd, 1987). However, plant cells do present a challenge to fluorescence microscopy because they often contain pigments and complex excitable molecules in subcellular structures that generate copious autofluorescence. In many circumstances this is a nuisance in terms of signal-to-noise ratio: the autofluorescence can swamp signal from other fluorescent probes
being studied. Furthermore, without the optical sectioning of the confocal microscope, autofluorescence glaring from all planes may obscure the signal from any in-focus structural information. Naturally, confocal imaging has made a strong impact in the area of plant-cell imaging. All these problems have been addressed in extensive reviews about applications of confocal microscopy to plant cell biology (Hepler and Gunning, 1998), imaging ions and other advanced methods (Blancaflor and Gilroy, 2000), and in hand- and textbooks on methods and applications (Galbraith et aI., 1995; Hawes and Satiat-Jeunemaitre, 2001). Green fluorescent protein (GFP) and other genetically encoded fluorescent probes have made a substantial impact on the field (Haseloff and Amos, 1995; Haseloff et al. 1997), and extensive lists of references of different applications, spectral conditions, and transient expression systems are available (Brandizzi et al. 2002). Detailed comparisons of the relative merits between conventional and widefield (Shaw, 2001) and between two-photon excitation and confocal (Feij6 and Moreno, 2004) have shown that there are specific niches for all methods, and probably none should be considered universal, irrespective of their price and sophistication. Previous reviews also included detailed protocols for image acquisition using these different methods. A very useful collection of practical criteria for probe choice, empirical methods, and many tricks for immobilization, perfusion, and loading protocols are described by Fricker and colleagues (2001). Specific methods for some of the most-used cell and tissue types have also been described (e.g., Kodama and Komamine, 1995; Raghavan, 1995; Sheen, 1995). Various fixation and other histological methods specific for plant cells are extensively described in many references (e.g., Spence, 2001), including complex and sophisticated freeze-fixation and freeze-substitution methods (Galway et aI., 1995; Parthasarathy, 1995). However, recent years have been marked by the introduction of less invasive methods and vital probes, with a strong emphasis on those that are genetically encoded. The focus now is on keeping cells and tissues intact and alive, and this offers the added value of enabling one to study the true dynamics of vital processes. Vesicle trafficking (Bolte et at., 2004), individual gene expression (Shav-tal et at., 2004), and cytoplasmic streaming dynamics (Shimen and Yokota, 2004) are just a few examples of the living processes now reachable using state-of-the-art microscopy in combination with genetic and molecular techniques. We will thus focus this chapter on recent developments that might affect plant biological research beyond the topics covered in the earlier reviews.
Nuno Moreno· Instituto Gulbenkian de Ciencia, PT-2780-1S6 Oeiras, Portugal Susan Bougourd • University of York, York YOlO 500, United Kingdom Jim Haseloff • University of Cambridge, Cambridge CB2 3EA, United Kingdom Jose A. Feij6 • Instituto Gulbenkian de Ciencia, PT-2780-1S6 Oeiras, Portugal, and Universidade de Lisboa, PT-1749-0l6 Lisboa, Portugal
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC , New York, 2006.
769
770
Chapter 44 • N. Moreno et al.
FIGURE 44.1. Cork now and then! Comparison of Robert Hooke's picture (A) and state-of-the-art laser-scanning microscopy (B) of a cork slice, having a time difference of almost 350 years. It was in the book Micrographia published in 1665 that Hooke used the term "cells" for the first time to describe the element of this regularity. Despite the obvious differences (one drawn by hand, the other based on digital acquisition of optical slices and software enhancement by using maximum-intensity projection), the evidence of the highly regular structure of plant cells is clearly evident in Hooke's image, and underlies the close relationship between form and function in plant cells. However, the confocal image also shows one of the most prevalent features of plant cells: the almost ubiquitous existence of autofluorescence in various subcellular structures, a factor that can be both informative and a problem to deal with when imaging probes with overlapping fluorescence spectra.
THE EVER PRESENT PROBLEM OF AUTOFLUORESCENCE Light is potentially damaging to cells, yet plants live in the fast lane. In order for photosynthesis to proceed, leaves and other aerial parts are often exposed to high levels of light radiation, and evolution has resulted in a number of mechanisms for filtering photons before they reach sensitive parts inside the cell. In addition, many plant cell walls accumulate complex hydrophobic molecules that regulate apoplastic water loss and movement (e.g., suberin in cork; Fig. 44.1), and a plethora of secondary metabolites, many working as pigments, have evolved for ecological or allelopathic reasons. Many of these molecules use varied pathways to dissipate photoexcitation, namely, non-radiative decay or relaxation, energy transfer, photosynthesis, and fluorescence. Unfortunately, the latter is common and autofluorescence is a natural feature of almost every plant cell (see Chapter 21, this volume, for detailed spectra). This autofluorescence is a nightmare for many fluorescence applications. On fixed, sectioned, and stained specimens, protocols using strong oxidizing agents such as chlorine bleach, chloral hydrate, or sodium-borohydride have long been used to reduce autofluorescence (Shaw, 2001; Chapter 18, this volume). Efficient as they are, these techniques cannot be used when imaging living cells, and the molecular degradation they produce can even destroy the specificity of immunostaining. Many fixation reagents, such as glutaraldehyde, also add to the problem by generating autoftuo-
rescent Schiff bases and delocalized electron resonance transfer when they react with cellular components. The problem is illustrated in Figures 44.1 and 44.2. In Figure 44.1 (A), the same cork that Hooke could describe because of its opacity to photon transmission is also shown by state-of-the-art confocal microscopy because the walls of dead cells are impregnated with suberin, a complex lipid with strong autofluorescence. As the cells are empty, after extended focus stacking, the cell walls can be seen with great detail and the confocal capacity of rejecting out-of-focus light renders images with great visual depth. Figure 44.2 shows what is probably the most common source of imaging problems: the green tissues. Chlorophyll accumulates inside plastids that occur in great numbers in green tissue (the round red organelles in this image). Although in the shoot meristem region, depicted in Figure 44.2, the cell wall is still relatively immature and thus has less autofluorescence (shown in blue), these two signals render the observation of other fluorophores (in this case GFP in the endoplasmic reticulum (ER), shown in green) almost impossible unless the out-of-focus light can be rejected and strong spectral separation is available. Cell wall autofluorescence swamps the signal from added probes and the absorption and scattering of both the excitation and the signal severely limits the distance that one can image into the tissue. The common way of dealing with spectral mixing is to use selective dichroic/emission filters, with the excitation peak as narrow as sensitivity allows. In a relatively young organ such as that shown in Figure 44.2, good results are obtainable with standard confocal settings (in this case, using a Zeiss Pascal). The importance of being able to produce stringent optical sections is illustrated in Figure 44.2(B), which shows a two-photon image of the GFP in the central section of the same plantlet. The image is almost clear of the other sources of signal. On a more physical basis, absorption in green tissues usually involves both excitation of a first excited singlet state (responsible for red absorption) and excitation of a second excited singlet state (responsible for blue absorption). In both cases, emission is mainly in the near-infrared because the blue-excited state relaxes to the first singlet [Malkin and Niyogi 2001; Figure 44.2(C)j. This complex response can be either a drawback or a bonus. Chloroplasts emit farther into the red [Fig. 44.2(D) in false blue], but other emissions in the visible spectrum can superimpose important information. Overall, they usually have a broad emission spectrum, and it is not a trivial matter to discriminate it from labeling, even in a spectral microscope. Many recent confocal microscopes come equipped with the capacity for spectral analysis, and this facility is becoming a useful tool for discriminating against autofluorescence. There are two ways of using the new spectral tools. In the first, one is only concerned with the emission from a single dye, and chooses the narrowest, most selective emission window for the dye involved. This is a one-step method, and only a single reference spectrum is needed for each series of observations. In the second method, one uses linear spectral unmixing to separate the emission of different dyes. Although speed is usually not an issue with the first method, it may be with the second, especially if the signal in the different channels must be acquired sequentially. However, parallel acquisition often implies channels only lOnm wide and, when an already weak signal is distributed among several narrow spectral windows, there is even less signal in each one. As a result, low signal level can be a problem and no commercial system is devoid of limitations in that respect. Although low signal can be overcome to some extent by more laser power, this can lead to other sorts of physical limitations, particularly singlet-state saturation of the dye and photodamage to the specimen.
Imaging Plant Cells • Chapter 44
C
771
in situ spectra
o +----.-----.----~--_.----~
500
550
600
650
700
750
Wavelength / nm
FIGURE 44.2. Three-dimensional reconstruction of an Arabidopsis hypocotyl/cotyledon imaged using a Zeiss Pascal with three emission channels and three laser lines (405, 488, and 633 nm) using a 63x Plan Apochromat NA 1.4 oil-immersion objective. This is an enhancer trap line with GFP linked to an ER domain protein, which in this case was activated on the meristematic primordial cells. Using the normal dichroic/emission filters, GFP (green) is well resolved from the chloroplasts (red) because their emission shows up mainly in the near infrared. The immature (thus dimmer in terms of autofluorescence) cell wall appears as shown in false blue. Insert (B) shows a different way of resolving the GFP signal, by two-photon excitation (TPE) at 870nm. Because the excitation wavelength is not as optimal for the chlorophyll as it is for GFP, the red plastids seen in (A) are much dimmer or invisible. (e) Arabidopsis seedling spectra traced in situ with a Leica SP2 AOBS. After making a lambda scan from 500 to 700nm with a spectral gate width of 30nm, using a 488 laser line and a lOx Plan Apochromat NA 0.4, a spectrum from each part of the tissue was traced for eGFP, cell wall and chloroplast autofluorescence. With these emission spectra, it becomes possible to unmix the lambda stack (D). This software tool comes with all the spectral confocal systems (Leica SP2 AOBS and Zeiss 510 Meta) and works in a similar way to spatial deconvolution but in this case the bleed-through is not caused by out-of-focus signal, but from overlapping fluorescence emission. Bar = 200 11m.
Figure 44.2 shows an Arabidopsis seedling with GFP expression in the ER. Reference spectra were generated for the different pigments [Figure 44.2(C)] and the respective signals extracted from the raw xyzA-image data (a so-called lambda stack), and then merged again with false colors [Fig. 44.2(D)]. It is clear that the channels have been sharply separated. While the principle works, there are limitations, and when several fluorescent proteins are
present, it can be more difficult to separate colors, especially if more than one of them occurs in the same voxel. In any case, one needs a specimen capable of producing lots of signal. As usual there are several engineering solutions to the problem of spectral detection. Leica uses a prism and a set of moveable mirrors to break the spectrum of emitted light into three or four segments, each going to a separate photomultiplier tube (PMT).
772
Chapter 44 • N. Moreno et al.
Zeiss uses a grating to disperse the emission spectrum over an array of 32 mini-PMTs, and has facilities to digitize as many as 8 channels from any combination of the 32 outputs (Nikon and Olympus are also launching their own solutions). There are pros and cons to each method: generally speaking one can say that the latter gains in terms of temporal resolution during live imaging, but perhaps at the cost of some sensitivity, especially in the red extreme of the spectrum. Once the spectral data have been obtained, they must be deconvolved (or unmixed) using programs similar to those used to deconvolve widefield structural data (see Chapters 23, 24, and 25, this volume), except that, instead of starting with a point-spread function, one starts with stored spectra of each of the dyes present. Least-squares algorithms are used to fit the spectral data measured in each voxel or region to a linear sum of the spectra of the dyes expected to be present. As all the light used by the spectral detector comes through a single pinhole, one cannot use pinhole size to balance the signal intensities from different dyes. Consequently, the procedure works best when the signals from all the various dyes are approximately equal in strength.
SINGLE-PHOTON CONFOCAL MICROSCOPY Because plant tissue generally consists of deep layers of highly refractile cell walls and aqueous cytosol and contains various autofluorescent and light-scattering components, intact tissue proves a difficult subject for fluorescence microscopy. However, direct imaging of living tissue is possible using suitably corrected microscope optics. Plant seedlings or excised tissues can simply be mounted in water for microscopy and examined using a longworking-distance water-immersion objective to minimize the effects of spherical aberration when focusing deep into an aqueous sample. Even with the use of such specialized objectives, using single-photon excitation, image quality degrades rapidly for optical sections deeper than 60 to 80l1M within the tissue. However, the small size of seedlings, such as those of the model plant Arabidopsis thaliana, allows very useful imaging despite this limitation. For example, median longitudinal optical sections can be obtained from intact roots. This direct approach to imaging plant materials has been reviewed elsewhere (Haseloff, 2003).
FIGURE 44.3. (A) Optical section of an Arabidopsis root with a random insertion (Cuttler et aI., 2000) of Clontech-YFP targeted to the plasma membrane (confocal image excited at 525 nm using a Leica SP2 A08S). (8) Seedling shoot epidermis from a line similar to that shown in (A), but with the Clontech-GFP imaged with TPE at 930nm. Despite the non-optimal optical response to this wavelength, the beam has enough power to produce the desired exc itation and yields outsta nding results in terms of tissue penetration and signal-to-noise ratio. (C) Dual-e mission confocal imaging from a stoma in which the ER is stained with mGFP5 (enhancer trap line) and the chloroplasts are emitting auto-fluorescence. Mixed widefield and fluorescence (D) and confocal fluorescence (E) of mGFP5 enhancer-trap transformed Arabidopsis root hair. GFP is fu sed to an ER motif, and the dynamics of the big ER inclu sions are clearly visible. (F) Doubleemission of ER-targeted mGFP5 (Siemering et al.. 1996) fusion product with the SCARECROW gene (Wysocka-Diller et aI., 2000). Despite the low expression level. the ER and nuclear envelope are clearly resolved in the endodennis. four cell layers deep into the root. The cell wall is stained with vital PI. The optical-sectioning properties of TPE are welI ilIustrated in the sharp fading pattern of the cell wall tangential section.
Imaging Plant Cells • Chapter 44
FIGURE 44.4. Leaf epidermal cell of a transgenic Arabidopsis with a microtubule-associated protein (MAP4) GFP fusion, imaged with a Zeiss Pascal confocal. z-stacks were acquired (B) and 3D volume-rendered. Red organelles are plastids. The subcortical distribution of the microtubules is clearly visible due to the depth created by the rendering algorithm. This construct, while producing excellent results for epidermal microtubules, seems useless for labeling microtubules in other cells. (C, D) Leaf trichome of a transgenic Arabidopsis with the actin-bundling protein talin fused with GFP, imaged with a Zeiss Pascal confocal and processed as in (A). Amplification in (D) is particularly noticeable by the details of the actin bundles inside these cells. However, great care must be used in interpreting this result, because under a constitutive strong promoter, talin is prone to creating "artificial" cables of actin, which are not supported by other means of visualization (Ketelaar et ai., 2004).
Direct visualization of GFP fluorescence in living tissues is not prone to fixation or staining artifacts, and can provide images of exceptional clarity. Moreover, the activities of living cells, such as cytoplasmic streaming, are clearly evident during microscopy. Ordinarily, movement within a sample is a nuisance, placing constraints on the use of sometimes protracted techniques for noise reduction during confocal microscopy, such as frame averaging. However, it is also possible to monitor dynamic events by timelapse confocal microscopy, and this combination of a vital fluorescent reporter with high-resolution optical techniques has proven valuable in cell biological and physiological experiments. We have also found that autofluorescent chloroplasts, normally present in the upper parts of the plant, and certain red fluorescent dyes can provide useful counterfluors for GFP. For example, propidium iodide can be applied to live seedlings in water, to specifically label root cell walls, and allow accurate identification of GFP expressing cells [Fig. 44.3(F)]. It is now possible to genetically mark cells or subcellular compartments within a living organism using GFP and to visualize these directly during development. A number of collections of transgenic lines have been developed where GFP gene expression has been targeted to particular cell types or where GFP protein fusions have been used to decorate cell compartments in Arabidopsis. For example, Cutler and colleagues (2000) have produced a library of transgenic Arabidopsis lines that express random cDNA-GFP fusions. The fluorescent protein is targeted to various subcellular compartments in these lines, and they provide a useful source of dynamic markers for nuclei, plastids, different membranes, and other compartments [Fig. 44.3(A,B)]. In addition, enhancer trap strategies have been used to direct the expression of a foreign transcription activator, GAL4, in different cell types in Arabidopsis (Haseloff, 1999a,b). The GAL4 gene was inserted into the Arabidopsis genome, using Agrobacterium tumefaciensmediated transformation. Expression of the GAL4 gene is dependent upon the presence of adjacent genomic enhancer sequences, and so different patterns of expression were generated. The inserted DNA also contains a GAL4-responsive GFP gene, and patterns of GAL4 gene expression are immediately detectable, with each GAL4-expressing cell marked by green fluorescence. These lines provide a valuable set of markers, where particular cell types are tagged and can be visualized with unprecedented ease and clarity in living plants [Figs. 44.2 and 44.3(C-E)]. The collections of cDNA fusions and GAL4 enhancer trap lines are available through the Arabidopsis stock centers. A particularly exciting field has also emerged with the successful development of GFP-fusion products with cytoskeleton associated proteins, which enable high resolution imaging of both microtubules (Fig. 44.4) and actin microfilaments
773
774
Chapter 44 • N. Moreno et al.
[Figs. 44.4(C,D)]; see also Chen et aI., (2003); and Shaw et ai., (2003). Use of over-expressing constructs of cytoskeleton binding proteins can, however, disrupt the delicate balance of the dynamic instability that usually takes place during the polymerization of the cytoskeleton under physiological conditions, as recently demonstrated for talin (Ketelaar et ai., 2004). Despite their obvious beauty and apparent information, these images and their spatial and temporal kinetics should be interpreted with great care.
Staining Plant Tissues As an alternative to direct imaging of live specimens, it is possible to fix and stain intact plant tissues and then to clear the material using a high-refractive-index mounting medium. It is then possible to obtain high-resolution optical sections from deep within intact tissues using objectives corrected for oil immersion. A wide variety of staining techniques have been adopted for plant specimens over the last 150 years. Perhaps the most widely used general tissue stains are Safranin 0 and Haematoxylin. These are often accompanied by the use of counterstains such as Fast Green, Orange G, or Alcian Blue. In addition, there are a large variety of more specific staining techniques that have been developed for particular plant materials and organelles. For example, Feulgen staining has been used for specifically labeling DNA, the periodic acid-Schiff reaction can be used to label carbohydrates, Aniline Blue for labeling callose, Nile Red for oil bodies, and Phloroglucinol for lignin . A multitude of published protocols are available. An excellent, recently published source of procedures can be found in Ruzin (1999). Interestingly, many of the synthetic dyes used for plant microtechnique are highly fluorescent. This is particularly so for the red, orange, and yellow dyes in the azine (e.g., Safranin 0), acridine (e.g., Acridine Orange), and xanthene (e.g., Rhodamine) families. Thus, many classic histological techniques unintentionally produce specimens that are intensely fluorescent. In addition, aldehyde fixation, certain mountants, and long-term storage of stained preparations can result in tissue fluorescence, and the high concentration of stains deposited in the sections can lead to metachromasia (Mason, 2000). In our hands, it is rare to find stained and sectioned botanical material that is not highly fluorescent. Current confocal microscopes can sometimes allow the clean separation of different fluorescent emission signals and the balancing of signal levels in different channels. Thus, fluorescent images of exceptional clarity and vivid color can be easily obtained (Fig. 44.5). In addition, the optical sectioning properties of the confocal microscope can be used to collect clear images from within thick sections and whole mounts.
Clearing Intact Plant Material The three-dimensional (3D) anatomical arrangements of plant cells have conventionally been observed using microtomy techniques. However, the laborious nature of thin sectioning, the problem of obtaining the desired plane of section, and difficulty of obtaining a complete series of sections has limited its use to the skilled and patient. Optical sectioning has many advantages from the point of view of speed and simplicity, and it allows software reconstruction of whole mount specimens, assembled from a series of z-axis images. However, it is difficult to observe cellular details deep in living plant tissue. Any light penetrating the tissue must pass through many layers of cytoplasm, watery vacuoles, and highly refractile cell walls. The different refractive indices of the material contribute to spherical aberration, and particulate subcellular
FIGURE 44.5. Confocal microscopy of Equisetum arvense spores. Equisetum sporangiophore tissue was fixed, cleared in xylene, embedded in paraffin, sectioned using a microtome, and stained with Safranin 0 and Fast Green FCE A z-series of three-channel color images were collected using a Leica SP confocal microscope with laser excitation at 488 nm, 568 nm, and 633 nm. These were visualized using a maximum intensity projection algorithm. Spores can be seen within the sporangiophore. The spores are surrounded by exosporial elaters, and lignified cells involved in sporangial dehiscence are seen (top right).
matter also causes light scattering. Various techniques have been developed in order to produce samples with glass-like optical properties and to maximize image quality. Clearing agents, such as xylene, clove oil, cedar oil, and chloral hydrate have been adopted and combined with compatible mountants such as Canada balsam and Hoyer's solution. All of these reagents have a high refractive index similar to that of glass (-1 .55). As a result of these clearing treatments, tissue sections become more transparent, greatly reducing problems with light scattering and spherical aberration as long as oil-immersion lenses are used. This allows high-resolution imaging of thin sections. When the same techniques are applied to thick sections or whole mounts, the results of clearing are even more startling. However, the stain or fluorochrome is generally distributed throughout the cleared tissue and details deep in the sample are still obscured by overlying signal. Here the confocal laser-scanning microscope proves its worth. High-resolution optical sections can be collected to distances of greater than 200 IJ.m in such cleared samples. The depth of image collection is limited mainly by the working distance of available high numerical-aperture objectives. The application of classic, highly fluorescent stains and clearing techniques creates a new field of opportunities for modem confocal microscopy and computerized display methods. Recently, classic botanical methods have been modified to allow intense and specific staining of plant cells and clearing for 3D microscopy. For example, Aniline Blue has been used as a stain for the cell contents of Arabidopsis embryos subjected to 3D imaging (Bougourd et aI., 2000). More recently, plant carbohydrates have been labeled by treatment with periodic acid to
Imaging Plant Cells • Chapter 44
FIGURE 44.6. (A) Deep optical sectioning of cleared Arabidopsis thaliana root tissues. Mature Arabidopsis embryos were treated with periodic acid to produce aldehyde groups within carbohydrates, and stained using a pseudo-Schiff reaction to specifically label the cell walls. The tissues were mounted in a chloral-hydrate-based clearing agent for microscopy. The combination of clearing and intense staining allows deep optical sectioning of an entire embryo cotyledon. A series of 736 optical sections were obtained to span the cotyledon, producing a dataset with a depth of 147 11M. (8) 3D segmentation of plant cells. A series of confocal optical sections, corresponding to a segment of cotyledon from an Arabidopsis embryo, is visualized using the AMIRA orthogonal slicing routines. The AMIRA 3D segmentation editor was used to seed and label particul ar voxels that correspond to chosen plant cells within the confocal dataset. The use of a specific cell wall stain allows easy selection of the intern al volumes of individual cells. A closed, tri angulated surface could then be formed over the selected group of voxe ls, using a marchingcubes algorithm. Rendered cells are displayed at the correct position and scale within the dataset, to build an accurate representation of the shapes, arrangement, and connectivity of cells within the tissue.
---~-----.--.--.-.::
produce aldehyde groups that are reacted with fluorescent pseudoSchiff reagents. If fixed plant tissue is treated in this way, cell walls (and starch-containing plastids, if present) become intensely and covalently labeled with the fluorophore . The labeling of wall material produces a complete outline of each cell. The tissue can then be directly cleared in a high-refractive-index agent containing chloral hydrate, and mounted in Hoyer's solution for microscopy [Haseloff and Bougourd, unpublished results; Fig. 44.6(A)]. The combination of high level s of fluorescence and high refractive index mountant allows the collection of extended z-series images at very fine resolution (0.1-0.5 f..LM steps), using a confocal aperture close to I Airy unit, and without fear of photobleaching or
..
-- -
- --
- --
-
- - --
775
"---
signal and resolution loss due to spherical aberration. Imaging to a depth of around 200 f..Lm allows simple optical sectioning throughout an entire Arabidopsis root at high resolution. In fact, every cell within a mature Arabidopsis embryo can be clearly visualized (Fig. 44.6).
3D Reconstruction The basic features of a plant's body plan are established during embryogenesis, but its final form results from the continued growth of meristems and the formation of organs throughout its life, often in a modular and indeterminate fashion . Because plant
776
Chapter 44 • N. Moreno et al.
cells are constrained by rigid cell walls and are generally nonmotile, there is the clear possibility that cell fates within a meristem are determined by lineage. However, evidence from plant chimera and wounding studies have demonstrated a more important role for cell---{;ell interactions during fate determination. It is likely that, during plant development, positional information is exchanged between cells, and that the fate of cells within a developing tissue is determined by a network of local cellular interactions. In order to dissect such a network, it is crucial that we can clearly map individual cells and their neighbors inside intact meristems - and have means to manipulate them. Thus, the cellular anatomy of plants is of particular relevance to the understanding of development and morphogenesis. Three-dimensional visualization techniques similar to those used in medical imaging can be applied to confocal datasets. This involves the selection and labeling of particular voxels that correspond to a 3D object of interest. Various techniques are available for selecting volumetric objects, which range from the fully manual to automatic tools that detect volume boundaries through differences in local intensity or texture (see Chapters 14 and 15, this volume). The use of specific staining techniques can aid the labeling process. For example, cell wall staining produces an outline of every cell in the tissue. This is very helpful as it allows the use of automatic segmentation tools to determine the interior volume of a chosen cell. We routinely use AMIRA, a generalpurpose physical modeling and data visualization program (Mercury Computer Systems, www.tgs.com). for our 3D visualization and segmentation [Fig. 44.6(B»). The software provides an interface for visualizing large multi-dimensional confocal
microscopy datasets. AMIRA provides a very useful set of input/output, data handling and visualization modules, and allows software routines to be combined in a modular fashion. This allows specialized 3D reconstruction and visualization techniques to be applied in a flexible way to different types of data, including confocal datasets. In addition, a developer's version, AMIRADev, allows the incorporation of custom visualization techniques. The program provides a simple interface, sophisticated, fast visualization routines, is affordable and robust, and is suitable for both highend PC and UNIX hardware.
3D Segmentation High-resolution confocal datasets can be rapidly segmented to allow direct visualization of cell arrangements within intact plant embryos and meristems. The large data files, up to 250 MB in size, can be directly rendered to allow excavation of the data, production of sections in arbitrary planes, and rendering of surface features. In practice, cells are generally chosen by placing a "seed point" manually within the center of a cell in the confocal dataset. This seed is then inflated in three dimensions to find the cell boundaries that are marked by a high staining intensity (Fig. 44.7). The program provides a segmentation editor for this purpose [Fig. 44.6(B»). The exterior geometry of a segmented cell can then be described using a "marching cubes" algorithm, which if needed can be converted to a solid geometry for the computer-assisted milling of 3D models or for finite element analysis [Fig 44.8]. We can now routinely reconstruct the cellular structure of entire meristems for various experiments.
FIGURE 44.7. Balloon model-based segmentation of plant cells. A deformable-mesh algorithm for segmentation was implemented in C++ as a software module inside AMIRA. The module allowed the interactive placement of a seed mesh within the dataset (top left panel). The surface was then inflated using a discretetime physical model. Expansion was accompanied by adaptive subdivision of the surface during inflation, and ultimately was constrained by a "force" based on the image intensity gradient vector. Vertices on the surface are attracted to intensity maxima in the data, which correspond to the stained boundaries of the cell.
Imaging Plant Cells • Chapter 44
777
FIGURE 44.8. (A) Watershed algorithm for segmentation. Code from the National Library of Medicine Insight Toolkit (www.itk.org) was adapted to provide a watershed algorithm for the AM IRA visualization platform. The algorithm was applied to a 3D dataset derived from the root primordium of an Arabidopsis thaliana embryo. A single, root-cap protodermal initial cell from the segmented dataset is displayed, and its different shared walls are shown color-coded. Each colored segment corresponds to a connection with a different neighboring cell within the meristem. (B) Graph of 3D connections between cells. Using the dataset shown in (A), an adjacency network was computed and cell-cell contact measured. This data was represented using graph-generation software (AT&T Graphvis). The 16 cells making contact with the root cap-protodermal initial (RCP_initial) are shown on the diagram, with measurements of the area of each cell-cell contact.
There are many opporturutles for improved analysis of this kind of 3D dataset. For example, plant cells are generally convex and simple in shape, and this allows the use of a more robust model-based segmentation approach: a deformable mesh can be placed within a 3D dataset at chosen seed points and inflated (McInerney and Terzopoulous, 1996). The mesh simulates an elastic surface expanding from the interior of a cell. The surface evolves through a discrete-time physical model and adaptively subdivides to fit the object boundary (Fig. 44.7; Rudge and Haseloff, unpublished results). This type of model-based segmentation is much less sensitive to noise in the experimental data and produces a compact description of plant cell shapes directly. Other techniques produce an intermediate segmented volume, from which a surface must be generated and smoothed. In addition, it is possible to automatically obtain a measurement of the number and area of shared walls between cells and their neighbors. These values are highly relevant biologically as they correspond to shared walls that contain plasmodesmata,
provide conduits for informational molecules that regulate cell behavior, and are an important parameter for modeling approaches . The watershed algorithm is based on the metaphor of water catchment basins in a landscape. First, an initial classification of all points into catchment-basin regions is done by tracing each point down its path of steepest descent to a local minima. The confocal image intensity is used as the landscape height. This process gives ridges between cells and shows slightly uneven terrain in the cell interiors. Next, neighboring regions and the boundaries between them are analyzed according to minimum-boundary height to produce a tree of merges among adjacent regions. By changing the "flood level," we can interactively traverse the merge tree and thus finely tune the segmentation. The technique produces a single boundary between cells with no empty, unclassified intervening space. We can use this property to compute an adjacency network for the cells and measure cell-cell contact areas. This can then be represented diagrammatically using graph-generation software (Fig. 44.8; Rudge and Haseloff, unpublished results).
778
Chapter 44 • N. Moreno et al.
FIGURE 44.9. Three-dimensional reconstruction of cell arrangements within an Arabidopsis thaliana embryo. (A) An 8-cell Arabidopsis embryo was treated with periodic acid and stained, using propidium iodide as a pseudo-Schiff reagent. The sample was cleared with chloral hydrate and mounted in Hoyer's Solution. A z-series of confocal optical sections was collected, and a single optical section from near the middle of the 3D dataset is shown. (B) The 3D image was then processed using a deconvolution algorithm (Huygens Essential, Scientific Volume Imaging). A single optical section is shown from the processed volume. Deconvolution results in improved image detail and SIN. (C) Visualization of the deconvolved 3D dataset, using the Voltex module in AMIRA, provides direct volume rendering. The stalk-like embryo and some surrounding maternal tissues can be seen. (D) Cells in the lower tier of the embryo (yellow) and upper cell of the suspensor (green) have been segmented, converted to new geometrical surfaces, and visualized within the dataset. The cells are shown within a semitransparent outline of the embryo. Extracted cell geometries can be used to define regions of interest, or co-visualized with the original microscopy data.
These computer visualization methods. which are adapted from the medical imaging field, reduce large datasets to a much more compact and simple description of the 3D shapes and relative arrangement of cells in a meristem or embryo (Fig. 44.9). Because cell-cell signaling plays such an important role in plant development, these techniques show much promise for the analysis of genetically perturbed plants, and as a basis for modeling the cellular interactions themselves.
TWO-PHOTON EXCITATION: ARE TWO BETTER THAN ONE? Recently, two-photon excitation (TPE) fluorescence microscopy has become a common tool in many advanced cell biology laboratories. A growing body of literature points to several advantages of TPE over other fluorescence imaging methods, with improved
signal-to-noise ratios, deep penetration, and benefits for living-cell imaging (Feij6 and Moreno, 2004). TPE microscopy is still in an early stage of development and reproducible protocols, probes, and applications remain relatively scarce. However, botanical techniques already in use provide clear advantages. In particular, TPE is relatively immune to the presence of out-of-focus absorbing structures. This section covers some of the aspects of imaging plant cells using TPE.
Improved Signal-to-Noise Ratio and Dynamic Range Of all the possible advantages of TPE microscopy, the one we feel most confident to stress is the high signal-to-noise ratio (SIN) and dynamic range of the final signal. Generally speaking, once conditions have been sufficiently optimized to produce a good image, TPE provides better contrast, crispness, and image quality. Numer-
Imaging Plant Cells. Chapter 44
ous examples of this rule have been shown elsewhere (Feij6 and Moreno, 2004, and references therein; see also Chapter 21, this volume). Comparison of images collected using single- and twophoton excitation showed much better detail and signal-to-noise ratio for the TPE data, especially for tangential sections at the extremes of a deep z-stack. On such sections, TPE resolves without loss of dynamic range, whereas confocal microscopy gives blurry and ill-defined images (Fig. 44.10). Consequences of this increase in SIN ratio and dynamic range become even more evident on deep/thick or whole-organ imaging. In a quantitative comparison with confocal and widefield epi-fluorescence, TPE was recently shown to outperform the other methods in terms of effective resolution (Cox and Sheppard, 2004).
Imaging Thick/Opaque Specimens The second generally accepted TPE advantage is its capacity to image deep into tissues that usually allow poor or no penetration in confocal microscopy. This trend is well reflected in the literature and in our own studies on whole-organ imaging of Arabidopsis (Feij6 and Moreno, 2004). However, it should be pointed out that the ability to image whole organs or thick tissues varies tremendously, especially in plants. Although people have been able to image >500 J..lm into living brain (M. Vaz Afonso and T. Bonhoeffer, personal communication), brain tissue is far less opaque than most plant tissues. As a result, penetration values this large are very uncommon for botanical specimens unless they have first been fixed and cleared. On some tissues, we could not penetrate through more than the epidermis (e.g., Arabidopsis living styles), apparently because it is covered with an outer cuticle that is so opaque that deeper penetration requires laser power levels that damage the tissue (they can literally boil the epidermal cells). Other tissues (e.g., immature anthers of Agapanthus) are reasonably transparent and allow live imaging up to the full working-distance of a high-resolution oilimmersion lens (ca. 200J..lm). Roots usually allow much deeper imaging than leaves or other green tissues [Fig. 44.1O(A,B)]. Add a low level of stain to these absorption problems, with the frame averaging and the extra excitation/fading this implies, and one can see that any estimate on how deep one can go must be assessed on a case-by-case basis. Several direct comparisons of confocal and TPE microscopy are offered in the literature. Vroom and colleagues (1999), working on microbial biofilms, made a quantitative comparison of signal intensity, and concluded that, compared to confocal, they were able to record images four times deeper and that these deep images did not lose contrast. In fixed material embedded in Nanoplast resin, TPE showed improvements both in penetration, contrast, and fade resistance (Decho and Kawaguchi, 1999). A practical example in which this improvement led to novel information was documented by Meyer and Fricker (2000). While studying glutathione distribution in different tissues, TPE provided more detail from optical sections deep in the tissue with less signal attenuation, and this ability was pivotal in being able to distinguish vacuoles from cytosol and to get a better separation of the sequestered signal. On a more technical level, Sun and colleagues (2001) showed that the attenuation of the excitation signal in plant tissues is reduced with TPE. More significantly, Cheng and colleagues (200lb) made probably the most thorough analysis of signal attenuation as a function of the excitation wavelength (see Chapter 21, this volume). They determined that in mesophyll cells and whole leaf, while major attenuation of the signal occurs only below 700nm (the major peak of attenuation at ca. 690nm, attributable
779
to chlorophyll absorption), the attenuation decreases continuously up until about 1000nm, which makes it clear that the longer the wavelength, the less appreciable will be the attenuation of the excitation signal. Better dynamic resolution and improved penetration are easy to demonstrate in a number of Arabidopsis organs. Figure 44.10 (A,B) shows an application to whole, living roots, in which the cell walls have been stained with propidium iodide (PI) at a vital concentration. Comparison of Figure 44.1 O(A) (confocal) with Figure 44.10(B) (TPE) makes it clear: TPE provides high contrast and resolution to the mid-sagittal level of the root, in both yz- and xzprojections, and up to the 10th cell layer (the red lines indicate the relative position of the zz' -projections). This comparison is especially noteworthy as exactly the same specimen, optics, and acquisition protocol were used to obtain both images, and the only changes were the excitation source and the pinhole aperture (1 Airy unit for confocal and wide open for TPE). Because roots are cylindrical, this level of performance makes it reasonable to speak of the high-resolution characterization of an entire living root. In contrast, under the very same imaging conditions, confocal resolves no better than five cell layers (note that the red line indicating the projection plane is not even over the central stellar tissues), and the contrast and definition shown on the negative-contrast projection shows a clear deterioration of the signal as the image plane goes deeper. In other organs, such as living leaves of Arabidopsis, penetration is much more limited (Feij6 and Moreno, 2004). While one can appreciate improved dynamic range and better structural accuracy (specially with external, non-descanned PMTs), compared to confocal, TPE penetration was still barely more than 50 J..lm (see also Chapter 21, this volume). The conspicuous presence of abundant chloroplasts explains this result. Not only is penetration impaired, but specimen damage also becomes an issue. Complex interactions among all the pigments present can generate a number of potentially harmful products that may make TPE even more damaging than confocal in green tissues (see next section). We have imaged living anthers at power levels consistent with structural recording of dynamic processes (Feij6 and Cox, 2001). Both DNA stained with 4'-6-diarnidino-2-phenylindole (DAPI) and autofluorescence from the cell walls contribute to the observed fluorescence at 780nm [Fig. 44.1O(C,D)]. Under TPE, DAPI fades very little and produces a strong signal over an excitation bandwidth of more than 100nm (720-850nm). In confocal microscopy, the autofluorescence of the inner cell layers makes it difficult to image deeper than two to three cells, leaving the important tapetum and the sporogenic/meiotic cells occluded. However, under TPE, anthers show more transparency, allowing sections close to 200 J..lm [Fig. 44.10(B); 6-7 cell layers deep], which is the working distance limit of a high-NA immersion lens (in this case, a Nikon PlanApo, 60x, NA 1.4, oil). Remarkably, the same level of laser excitation was used to acquire the entire z-stack from the surface epidermis to the central sporogenic tissue [Fig. 44.1O(AB)]. With these improvements, an exquisite degree of contrast and detail is obtained in images of the pit fields and wall stress fibers in tangential/oblique optical sections of intermediary cell walls [Fig. 44.10(A), arrow]. The ability to image the mid-sagittal section of an anther allows direct viewing of the activities of meiotic cells (Feij6 and Cox, 2001).
Fading, Vital Imaging, and Cell Viability Two-photon excitation is effectively confined to a sub-femtoliter volume at the point of focus within the sample, and although it is
780
Chapter 44 • N. Moreno et al.
Imaging Plant Cells • Chapter 44
generally agreed that TPE can minimize phototoxicity and fading during microscopy (Potter, 1996), a word of caution is due in many situations. As both damage and signal is proportional to (at least) the square of the power, power control is often a critical issue. Fragile tissues or those that dissipate heat poorly may pose a problem. For example, thin, fixed, sectioned materials (e.g., specimens used for immunolabeling) are sometimes much more difficult to image with two-photon than using widefield or confocal microscopy, because, in the focus plane, fading occurs more strongly with TPE. Clearly, TPE is not a cure-all. It takes some effort to determine the best experimental conditions for vital TPE imaging, and it is particularly important to find the best balance between power, excitation wavelength, and probe concentration (when controllable). It usually implies a careful search for the best wavelength (one that maximizes SIN while minimizing photodamage) and then setting the power level so that there is just enough signal to form an image. The tuneability of titanium: sapphire (Ti: Sa) laser sources allows one to search for the optimum excitation wavelength, and this often leads to the choice of a wavelength away from the peak in the excitation spectrum. Continuous wave laser sources used for confocal microscopy are generally restricted to a small number of discrete excitation wavelengths, which makes this sort of optimization impossible. As a rule of thumb, the longer the wavelength the less the damage. Even at low power, using wavelengths less than 800nm often causes much more damage in the form of fading, arrest of streaming or even slowing growth. A second rule of thumb is that wavelengths longer than 870nm are almost always more suitable for imaging plant cells. Finally, dye molecules that are more asymmetrical (e.g., DAPI) tend to show better results than symmetrical ones [e.g., fluoroscein isothiocyanate (FITC)]. In symmetrical molecules, usually more than one excitation peak is obtained, and using the longer wavelength tends to produce much better results in terms of less fading, even though it may also show less emission. An exhaustive analysis and comparison of metabolic imaging led Fricker and Meyer (2001) to conclude that TPE might be the best means to study primary metabolism in vivo, if and when the relevant probes are produced and optimized. Because pollen tubes
781
show easily observable signs of vitality in the form of streaming and growth, we have used them to assay comparative phototoxicity. Being relatively resistant to radiation, they can also be used to studying fading (Feij6 and Moreno, 2004). When viewing sensitive features such as actin dynamics, streaming, growth rate, pump activity, etc., TPE always seems to produce superior results when operated above 870 nm. For example, expression of an ADF: GFP construct that labels actin micro filaments in pollen tubes, combined with TPE allowed the collection of up to 600 sequential frames without interval, without producing noticeable fading, and with minimal effect on cell growth rates. Figure 44.1O(E-N) illustrates another example of a result difficult to achieve using confocal microscopy. Phil Benfey's laboratory produced lines of Arabidopsis with GFP (von Arnim et ai., 1998) fused to the putative transcription factor Shortroot (SHR) under the control of its own promoter and these were then imaged using TPE in our laboratory. Again, better contrast, dynamic range, penetration, and sensitivity were obtained, and the whole root was imaged down to the distal side of the central stele using TPE at 920nm. Despite the weak signal, a diffuse cytosol labeling was discernible in the central stele cells and labeling in the nuclei of the endodermis shows a perfectly resolved and exclusive perinuclear location of SHR. Many of these features are not easily resolved in the equivalent confocal images. The TPE images helped to support the exciting finding that transcription factors may move between cell layers (Hake, 2001; Nakajima et ai., 2001), and SHR was assigned a role in defining cell fate in the root (Nakajima et at., 2001). The sequence in Figure 44.1O(E-N) shows frames of a division of a root meristematic cell (arrow), in which the putative transcription factor diffuses out during nuclear envelope breakdown and later re-aggregates in the two daughter nuclei. Using standard confocal, the signal faded before all the data could be collected because the extremely low fluorescence levels required a number of frames to be averaged to produce each highdefinition optical section (G. Senna, 2002, personal communication). In the TPE results shown here, not only was fading reduced to a manageable level, but the radiation seems not to have affected either the division cycle or root growth to any measurable extent.
FIGURE 44.10. Imaging of a whole, living root of Arabidopsis, vitally stained with PI (lOf.!g/mL; wild-type root at ca. 8~12 days after germination) with confocal (A) and TPE (B) microscopy. In healthy cells, the membrane is impermeable to PI, rendering the cell wall fluorescent and clearly defining the cell boundaries. (A) The central confocal image shows an optical section 30f.!m deep. The right and bottom panels show negative images of xz and yz reconstructions of the whole zz-stack acquired (confocal excitation using the three visible-light lines of the Kr: Ar-ion laser, emission filter HQ 598/40; the stack was acquired with 0.5 f.!m steps). The lines mark the relative position of the reconstructions on the central image, or the depth of the specific plane shown on the side panels. Wall boundaries, although visible, are diffuse, despite generous use of Kalman averaging. Three-dimensional projections accentuate that feature, and the xz-image shows little contrast and resolution beyond 4 to 5 cell layers. Thus, it is only possible to image the half-diameter of the whole root up to about 100f.!m from the tip and the important context of the central stele is lost. Bar = 50 f.!m. Plan Fluo 40x, oil-immersion objective (NA 1.3). However, under TPE (B), anthers show more transparency, allowing sections close to 200f.!m [Fig. 44.5(B); 6---7 cell layers deep], the working distance limit of a high-NA immersion lens. (Reproduced from Moreno and Feij6, 2004, with permission from Springer-Wien.) (C) Image of a living Agapanthus umbelatus anther, stained with DAPI. This organ is very translucent, permitting imaging up to the working-distance limit of a high-NA objective (ca. 200f.!m, montage from surface to ca. 200f.!m is shown from a to f). Nuclei (C, double arrow), cell walls (arrow), and plastids (arrowhead) all show some autofluorescence. (D) A tangential optical section through the cell wall of the third to the fourth cell layer reveals pit fields and cellulose stress directions with exquisite detail. Detailed nuclear structure is clearly visible up to the seventh cell layer, eventually reaching the central meiocyte tissue (B, arrow). Imaged with a Plan Apo 60x, oil-immersion objective (NA 1.4); excitation, 780nm; no barrier emission filter, internal PMT. Bars: (C) 20f.!m; (D) 50f.!m. (Reproduced from Feijo and Cox, 2001, with permission from Elsevier.) (E~N) Time-lapse sequence of the division of a meristematic cell in the apex of a lateral root of A. thaliana that is expressing a fusion protein constructed of SHR and GFP. The montage was extracted from a sequence acquired with a rate of I frame/5 min over almost 3 h. During that period the root grew unaffected, and several meristematic and endodermal cells divided. In this sequence, a meristematic cell is shown throughout the cell cycle: the SHR: GFP fusion protein is first located at the periphery of the nucleus (E, arrow), but diffuses to the cytosol as soon as the nuclear membrane disappears (G). At this stage, the fluorescence seems continuous with the central stele cells (G~I). Following division and by the onset of cytokinesis (J~L) the nuclei are again discrete and SHR: GFP is again visible in the two daughter nuclei (K~N). (Excitation, 920nm; emission, 530nm; external PMT.) Bar = 10 11m. (Reproduced from Feij6 and Moreno, 2004, with permission from Springer-Wien.)
782
Chapter 44 • N. Moreno et al.
Again, a word of caution should be issued. Potential pitfalls will remain until the right conditions to image specific probes in particular tissues are better defined. Until then, any study should be undertaken systematically, with time allowed for long sessions in which the procedures can be optimized for a particular experimental condition. There are also cases where confocal provides the faster and more reliable path. In a recent case reported by Reddy and colleagues (2004), the shoot apical meristem (SAM) was imaged in an attempt to resolve a number of cell and organ-fate issues (Running, et al. 1995). A number of transgenic lines with markers of cell division and cell fate were created, and methods were developed for imaging the shoot growth on the same confocal plane for over a week. The workers were able to image deep stacks encompassing the whole meristem down to the first primordia every 4 h (Reddy et aI., 2004). Using Zeiss optics and confocal scanner, they made extensive comparisons with TPE and concluded that confocal significantly outperformed TPE in terms of keeping the SAM growing and dividing (Reddy, personal communication). Being a green tissue, these results come as no great surprise because, as we discussed in the autofluorescence section, the complex interaction between pigments and the high levels of near-infrared radiation makes it predictable that green tissues will always pose a problem for TPE imaging. This example should definitely be kept in mind when deciding on the amount of time and effort one might need to develop a viable protocol with a new method.
Two-Photon Imaging of Plant Cells and Organelles To date there is a growing body of literature showing the application to TPE of practically all of the common dyes and labels used for visible, one-photon fluorescence. Here we restrict ourselves to plant applications, plus our own experience managing a multi-user facility (see also Table II in Feij6 and Moreno, 2004; Chapter 17, this volume). Again, most information regarding the use of different dyes with TPE should be considered preliminary and dependent on the specific experimental context (dye, vitaUfixed samples, microscope, etc.). The available information should be used as a starting point for the process of fine-tuning. Implicit in this statement is the recognition that it is difficult to obtain clear reproducible spectra for TPE excitation. In contrast to confocal microscopes and spectrofluorimeters, in which power can be kept within close tolerances and the average level has little variation, TPE depends critically on the peak power, which changes dramatically (i) along the tuning range of most ultrafast lasers, (ii) with position inside the focal volume, and even more so (iii) with the pulse width. Generally speaking, longer wavelengths provide shorter pulses, hence higher peak powers. To complicate matters, the average output power of a Ti : Sa laser peaks at around SOO nm, but falls off rapidly at the extremes of the tuning range. As most biology laboratories do not possess (or want to!) the very complex analytical instruments necessary to measure TPE cross-sections properly, wavelength-optimization routines are more likely to be based on somewhat crude empirical experiments. Data are often hard to reproduce directly on another system with different optics and design, but the experimental parameters are often close enough to indicate a range for optimization. One important reference that should be kept in mind is the classic work of Xu and Webb (1996) listing spectra for most of the commonly used dyes. [For updates, check the largest library of TPE spectra at www.drbio.comell.edu; a small list of commonly used dyes is also given in Diaspro and Robello (2000).]
The ultraviolet (UV)-excited DNA dyes DAPI and Hoechst 33342 are TPE dyes par excellence and it is of no surprise that one or the other has been used in every paper published so far to illustrate the TPE principle (e.g., Denk et aI., 1990; Konig, 2000; Tirlapur and Konig, 2002). In plant tissues other than the classic onion root system, these dyes have been used to image live meiotic nuclei in anthers (Feij6 and Cox, 2001), nuclear distribution during arbuscular mycorhization of Aspergillus nidans and tomato (Bago et aI., 1999) and deep imaging in nanoplast-embedded tissues (Decho and Kawaguchi, 1999). The popularity of these dyes is due to the fact that they are extremely bright and fade resistant under TPE, and that they can be excited over a wide range of wavelengths (720-SS0nm for DAPI in our system), a feature that facilitates double- and triple-labeling as well as searches for the best vital imaging conditions. TPE of Nile Red has been used to follow the mobilization of lipid globules in fungal hyphae (Bago et aI., 2002). Autofluorescence was used for high-resolution dynamic analysis of chloroplast division (Tirlapur and Konig, 2001) and to study the cell wall (Konig, 2000). Perhaps one of the most interesting and informative trends relates to TPE visualization of glutathione (GSH). The method was introduced to plants by Fricker and colleagues (2000), and it uses the dyes monochlorobimane (MCB) and monobromobinane (MBB), which complex with GSH to produce glutathione-S-bimane (GSB). Under normal confocal, these workers determined the concentration of GSH to be 2 to 3 mM in most cell types. However, using TPE, they were able to obtain more detail with less signal attenuation deep in the tissue and this was pivotal in distinguishing vacuoles from cytosol to get a better separation of the sequestered signal (Meyer and Fricker, 2000). Only under TPE was it possible to measure that GSH concentration in trichoblasts (2.7 ± O.S mM) was significantly different from that in atrichoblasts (S.S ± O.S mM). More recently, trichomes were shown to have 300x more GSH than cells from the basement layer and the epidermis (Gutierrez-Alcahi et aI., 2000). These, and other results, support the conclusion that TPE is currently the best approach to studying primary metabolism in vivo (Fricker and Meyer, 2001; Meyer et aI., 2001).
Two-Photon Excitation Imaging of Green Fluorescent Protein TPE is well suited for viewing GFP in plants (Xia et aI., 1999, Volkmer et aI., 2000), and indeed GFP is easily excitable at a wide range of wavelengths and in combination with other dyes, such as propidium iodide [Fig. 44.3(F)]. Special manipulation of the GFP levels have been used to raise signals above the wild-type chlorophyll autofluorescence by using "super" promoters, such as 3SS3SSAMV, driving especially bright GFPs, such as S6ST, a combination successfully used to visualize mitochondria using TPE with good SIN even in green tissues (Kohler et al., 1997). This example points to the need to consider the different kinds of GFP currently available. The first one introduced was the wild-type GFP which, while excitable at about SOOnm, faded rapidly, perhaps even more so than with confocal. This and other problems led to the development of several enhanced forms of GFP. Here we offer data about some of the commonly used forms: Clontech's EGFP, Jim Haseloff's mGFP forms (Siemering et al. 1996; Haseloff et aI., 1997) and also the forms engineered by Chiu and colleagues (1996) and by von Arnim and co-workers (199S). Clontech's version is perhaps currently the most popular. It shows a relatively broad TPE excitation peak from 920 to 940nm, but can hardly be visualized below 900, unless under very strong
Imaging Plant Cells • Chapter 44
783
FIGURE 44.11. Fast dynamics in plant cells is illustrated by the movement of the protein inclusion on the ER of the hypocotyl epidermis on enhancer trap lines with GFP fused to an ER motif. In this specific sequence, TPE at 870nm was used to stretch the limits of viability of this kind of imaging. The images shown were extracted from a time series of a consecutive (no interval) sequence of acquisitions, which was extended up to 650 frames. Despite the amount of exposure, we observed no effect on fading or streaming rates.
promoters or when fused to abundant and stable proteins. Although these wavelengths are near the edge of the tuning range of the Ti : Sa laser and produce some dye fading, they still produce better images than with confocal microscopy [Fig. 44.3(B)], and even very faint markers such as membrane transporters can be imaged in living cells. The same result was obtained for the S65T mutation described by von Arnim and colleagues (1998) [Fig. 44.1O(E-N)]. Nevertheless, we have found that the two other enhanced-GFP versions (Chiu et al., 1996; Siemering et al., 1996) are even better for our purposes. First, they seem to behave best at 870 to 890nm, a much more convenient wavelength range for the Ti : Sa laser, and one that any user can mode-lock. Second, if reasonably expressed, they seem to experience no quantifiable fading, even when sequentially imaged (Fig. 44.11). They provide high-resolution signals, and the mGFP5-ER, expressed in enhancer trap lines, provides superb material for structural characterization, either alone LFig. 44.3(D,E)] or in combination with PI [Fig. 44.3(F)]. Fine details of the nuclear envelope and of the ER-derived system are resolved, and the exclusive location of the tag in the endodermis is evident from the cell-edge fading that can be seen in tangential optical sections of the root. In streaming movements of the cortical ER, large particles have been followed in mGFP5-ER enhancer trap lines during many hundred consecutive frames, without any visible fading or quantifiable effect on the streaming rates and patterns (Hawes et aI., 200 I; Fig. 44.11). These two engineered forms of GFP have proven to be exceptional tools for TPE in plants, and can be expected to become the source of many important advances in our understanding of dynamic cell and developmental processes.
DYNAMIC IMAGING While cell mobility in plants is strongly limited by the presence of a semi-rigid extracellular matrix, plant cells can display impressive spatial dynamics at the subcellular level. Generally speaking,
cytoplasmic streaming is much faster than mechanical movements in animal cells, with organelles moving up to 21lmls [e.g., ER inclusions on Arabidopsis hypocotyl epidermis (Fig. 44.11) or mitochondria in pollen tubes [(Fig. 44.12)]. The easiest way to
FIGURE 44.12. Comparison of spinning-disk and TPE imaging effects on pollen tube growth, Rhodamine 123-labeled mitochondria in Lilium longiflorum pollen tubes after 600 consecutive acquisitions. Frames were taken consecutively without any interval (1 frame/2s) in the case of the TPE (820nm excitation with an emission at 598/40nm) at a speed of SIs in a scanning disk (488 nm excitation with a triple dichroic). Unlike laser-scanning confocal (Feij6 and Moreno, 2004), both systems show a very low phototoxicity. However, in the case of the TPE we should point out the low noise picture, and, in the case of the spinning-disk, the fast image acquisition. With TPE, the growth rate decreases by 30% compared to 10% with the NSDC.
784
Chapter 44 • N. Moreno et al.
image such a fast process is to use a widefield microscope with a scientific-grade charge-coupled device (CCD) camera, and apply a cost-effective algorithm to the two-dimensional (2D) data, such as no-neighbors 2D filtering (Monck et aI., 1992; McNally et al., 1999). This method is a similar to one developed by Castelman (1979), in which three planes are acquired and the information of the outer planes subtracted to produce an improved image of structures present in the central plane. However, more stringent applications do need optical sectioning. In many cases, deconvolution is not fast enough to provide enough temporal resolution on a moving sample, and other methods, such as single-beam laser-scanning microscope (LSM), TPE, or Nipkow spinning-disk confocal (NSDC), or even videorate confocal (YRC) must be used . Frame size is the next critical choice. In a LSMffPE, one can go up to a few frames per second by choosing a box of only -10 kilopixels, usually at the cost of proper sampling and, therefore, resolution. For many applications, as long as one does not need to go faster than 2 to 3 frames per second (fps), it is possible to find a compromise between speed and resolution. On the other hand, if speed and resolution are needed simultaneously, a different approach is needed. NSDC and YRC use detectors that need time to integrate: thus, the dimmer the sample, the more time needed to record a decent picture. As most spinning-disk confocals use cooled CCD cameras [or the new intensified CCDs or even an electron multiplier (EM)CCDs], one can adjust the exposure time and the binning independently. The first parameter affects acquisition speed and the second, resolution. Sensitivity is an issue, and exposures as long as several seconds are not unusual for lightly stained specimens. Binning 2 x 2 in a megapixel, cooled CCD improves this parameter dramatically. Although it will bring the effective pixel size close to the Nyquist sampling criteria, either the exposure time can be reduced to roughly one quarter or the SIN increased by a factor of 2. Though common sense suggests that plant cells should be able to handle lots of light, this is not always true. Because their growth rate is quite sensitive to light, pollen tubes can be used to quantify photo toxicity. As previously mentioned, as long as a reasonable amount of signal is present, TPE above 870nm produces almost no effects on the growth and streaming rates of pollen tubes [Fig. 44.12(A)] . However, even when imaging continuously, without interval, good resolution still implies a 2s frame-scan time. At this speed, much of the dynamics of fast-moving organelles, such as ER (Fig. 44.11) or especially streaming mitochondria (Fig. 44.12) will be lost. Provided that there is enough signal to allow use of a fast CCD (the case in Fig. 44.12, stained with the very bright Rhodamine 123), Nipkow-disk-based systems, such as the PerkinElmerIYokogawa, definitely show a much more informative view [Fig. 44.l2(B)] . In this case, it was easy to achieve an increase of one order of magnitude in the time resolution (2 s/frame on TPE and 200 ms/frame with the NSDC) and this allows one to clearly visualize different cytoplasmic domains in which mitochondria move faster or slower. We were unable to record similar data with any single-beam scanning method. As can be appreciated by comparing Figure 44. 12(A,B), although the optical thickness is lower, the more accurate dynamic picture clearly outweighs the information lost by lower z-resolution. The ability to record a z-stack of a large, intact specimen in a fraction of the time also makes the technique suitable for time-lapse studies of fast-moving or growing organs. We expect that NSDC microscopes will be crucial to being able to resolve fine temporal dynamics with minimal fading. The sensitivity issue does exist, and lightly stained samples such as
many GFP-fusion lines could either not be imaged at all, or required an integration time so long that the system performance was less than that of TPE. Intensified CCDs or EM-CCDs may soon change this picture quite dramatically, and bring sensitivity to the level of TPE and LSM. Although the new generation of VRCs use high-speed computer interfaces with a bandwidth of 100 MB/s that permit high frame rates, problems with singlet-state saturation prevents single-beam confocal scanners from providing useful information about biological specimens except at low spatial resolution. On the other hand, high-speed line-scanners such as the Zeiss 5-Live avoid this trap by scanning many points at one time (see Chapter 10, this volume).
DECONVOLUTION In the last two decades there has been a boom in the use of 3D microscopy in biology, and botanical samples are no exception (Hepler and Gunning, 1998). Although this trend has been driven largely by the emergence of the confocal microscope, widefield microscopes can also generate raw 3D data, and software that can run on any modem personal computer is now available to extract sharp, 3D reconstructions from these data. If one considers a microscope to be a "convolution machine," its transfer function can be determined by measuring its response to a subresolution fluorescent bead: namely, its point spread function (PSF). To the extent that one knows the PSF and records the widefield 3D data stack accurately, lone should be able to invert the convolution process (deconvolution) to obtain a 3D image of the object imaged. The details of both convolution and deconvolution are covered in detail in Chapters 23, 24, and 25. Because of the difficulties encountered when trying to measure the PSF, we usually make an average PSF from the 3D images of several beads (Holmes et aI., 1995). Although the biggest drawback of this approach is the requirement that the sample must not move or change during data acquisition (restricting the technique to fixed cells or structures moving very slowly), even minimal problems with the optics, such as misalignment, spherical aberration, and dirtiness, will affect the final image to a much greater extent than in confocal and TPE. To some degree, it may be possible to compensate for the degradation of the PSF brought about by spherical aberration (Boutet de Monvel et aI., 2001) or by the heterogeneous refraction index of a living cell (Kam et aI., 1998). As these techniques improve, they may be important when deconvolution techniques are applied to images of plant cells in which, for example, spheroidal central vacuoles can act as miniature lenses. Despite these limitations, deconvolution retains some substantial advantages for living-cell microscopy: phototoxicity, photobleaching, and price are all often lower. It is also important to note that deconvolution protocols not only can, but should, be applied to all 3D confocal and TPE datasets. Doing so not only substantially reduces the "single-pixel noise features" produced by the statistical uncertainty attendant on measuring small numbers of photons, but it also substantially improves the statistical accuracy of the image data by effectively averaging the signal over many voxels (see Chapter 25, chis volume).
1
The main practical limitation to doing this is the presence of both Poisson and read noise in the final data .
Imaging Plant Cells • Chapter 44
785
FIGURE 44.13. The effect of deconvolution on a widefield stack. (A) A plane from a stack of a unicellular green algae. (B) A deconvolved version of the same plane using a measured PSF and a glycerol immersion objective (PlanApo 63x NA \.3). The whole projection was about 30llm deep (autofluorescence). The improvement in contrast is obvious, and is highlighted in the line scan. The PSF was generated using orange beads from Molecular probes (PS-Speck) with 170nm diameter.
Figure 44.13(A) shows an autofluorescence image of a unicellular green alga. In this specific sample, several attempts with both confocal and multi-photon microscopes failed to produce a sharp image over extended focus stacks of the entire depth of the cell. The main problem was that the high intensity illumination caused the signal to get dimmer and the autofluorescence to shift from red to green (Cheng et aI., 2001). Deconvolution of data stacks collected using a much lower intensity of excitation light made a considerable improvement [Fig. 44.13(B)].
CONCLUSION Ever since the time of Hooke, plant cells have been the foundation of many of the fundamental discoveries that have shaped cell biology. Underlying all these findings were significant advances in microscopy that have helped to push forward our conceptual thinking, for example, supporting the acceptance of the cell theory. In this chapter we have highlighted what is now the leading edge of this technological effort. Imaging is now more than lenses and microscopes. Computers are also essential, both to accumulate and display research images, and to extract and analyze enormous amounts of quantitative information from them. Recent advances range from major hardware (e.g., two-photon equipment) to the development of computer software that now enables us to derive fresh insights from old histological techniques. We have tried to emphasize that all these techniques are only of value when they enable us to describe previously-unknown biological features, and indeed many of the most important technical developments were developed only as part of specific research projects. The ever-growing field of genetically encoded probes, such as GFP, seems likely to trigger more new and important technical adaptations that will enable us to obtain more and better dynamic information from living systems. Plant cell biology continues to bloom, and much of this growth is supported by modem imaging methods.
ACKNOWLEDGMENTS The JAF laboratory is supported by FCT grants POCTVBCV and POCTV 4645312002, POCTVBIA-BCM/60046/2004, BIA-BCM/6 I 270/2004. Some GFP lines from the Haseloff laboratory were imaged during the 2nd and 3rd Gulbenkian Biology Course on Plant Development (Oeiras, Portugal, 2003) by groups involving the students Niko Geldner, Vaiva Kazanaviciute, Jirina Prochazkova, Kasia Olczak, Martha Alvarado (a very special thanks for the MAP4 and talin lines), Lisa Gao, Martin Potocky, and Chris Nichols. The SB laboratory is supported by grants from BBSRC.
REFERENCES Bago, B., Zipfel, W., Williams, R.M., Cahmberland, J.G., Lafontaine, J.G., Webb, W.W., and Piche, Y., 1998, In vivo studies on the nuclear behavior of the arbuscular mycorrhizal fungus Gigaspora rosea grown under axenic conditions, Protoplasma 203:1-15 Bago, B., Zipfel, W., Williams, RM., Jun, J., Arreola, R, Lammers, P.J., Pfeffer, P.E., and Shachar-HiII, Y. 2002, Translocation and utilization of fungal storage lipids in the arbuscular mycorrhizal symbiosis, Plant Physiol. 128: 108-124 Bago, B., Zipfel, W., Williams, RM., and Piche, Y., 1999, Nuclei of symbiotic arbuscular mycorrhizal fungi as revealed by in vivo two-photon microscopy, Protoplasma 209:77-89. Blancaflor, E.B., and Gilroy, S., 2000, Plant cell biology in the new millennium: New tools and new insights, Am. 1. Botany 87: 1547-1560. Bolte, S., Talbot, c., Boutte, Y., Catrice 0., Read, N., and Satiat-Jeunemaitre, B., 2004, FM-dyes as experimental probes for dissecting vesicle trafficking in living plant cells, 1. Microsc. 214: 159-173. Bougourd, S., Marrison, J., and Haseloff, J., 2000, Technical advance: an aniline blue staining procedure for confocal microscopy and 3D imaging of normal and perturbed cellular phenotypes in mature Arabidopsis embryos, Plant 1. 24:543-550. Boutet de Monvel, J., Le Calvez, S., and Ulfendahl, M., 2001. Image restoration for confocal microscopy: Improving the limits of deconvolution, with application to the visualization of the mammalian hearing organ, Biophys. 1. 80:2455-2470.
786
Chapter 44 • N. Moreno et al.
Brandizzi, E, Fricker, M., and Hawes, C., 2002, A greener world: The revolution in plant bio-imaging, Nat. Rev. Mol. Cell Bioi. 3:520-530. Castleman, K.R., 1979, Digital Imaging Processing, Prentice Hall, Englewood Cliffs, New Jersey. Chen, e.Y., Cheung, A.Y., and Wu, H.M., 2003, Actin-depolymerizing factor mediates Rac/Rop GTPase-regulated pollen tube growth, Plant Cell 15:237-249. Cheng, P.-e., Lin, B.-L., Kao, E-J., Gu, M., Xu, M.-G., Gan, X., Huang, M.-K., and Wang, y'-S., 2001a, Multi-photon fluorescence microscopy - The response of plant cells to high intensity illumination, Micron 32:661670. Cheng, P.C, Sun, C.K., Lin, B.L., Kao, EJ., and Chu, S.W., 2001b, Nonlinear multimodality spectromicroscopy: Multiphoton fluorescence, SHG and THG of biological specimen, Proc. SPIE 4262:98-103. Chiu, w.-L., Niwa, Y., Zeng, w., Hirano, T., Kobayashi, H., and Sheen, J., 1996, Engineered GFP as a vital reporter in plants, Curr. BioI. 6:325-330. Chu, S.w., Chen, I.-H., Liu, T.-M., Sun, e.-K., Lee, S.-P, Lin, B.-L., Cheng, P-C., Kuo, M.-X., Lin, D.-J., and Liu, H.-L., 2002, Nonlinear bio-photonics crystal effects revealed with multi modal nonlinear microscopy, f. Microsc. 208: 190-200. Cox, G., and Sheppard, C., 2004, Practical limits of resolution in confocal and nonlinear microscopy, Microsc. Res. Tech. 62: 18-22. Cutler, S.R., Ehrhardt, D.W., Griffitts, J.S., and Somerville, e.R., 2000, Random GFP: :cDNA fusions enable visualization of subcellular structures in cells of Arabidopsis at a high frequency, Proc. Natl. Acad. Sci. USA 97:3718-3723. Decho, A.W., and Kawaguchi, T., 1999, Confocal imaging of in situ natural microfibril communities and extracellular polymeric secretions using nanoplast resin, Biotechniques 27:1246-1252. Denk, W., Strickler, J.H., and Webb, W.W., 1990, Two-photon laser scanning fluorescence microscopy, Science 248:73-76. Diaspro, A., and Robello, M., 2000, Two-photon excitation of fluorescence for three-dimensional optical imaging of biological structures, f. Photochem. Photobiol. B 55: 1-8. Esau, K., I 977a, Anatomy of Seed Plants, John Wiley and Sons, New York. Esau, K., I 977b, Plant Anatomy, 2nd ed., John Wiley and Sons, New York. Fahn, A., 1990, Plant Anatomy, 4th ed., Pergamon, London. Feij6, J.A., and Cox, G., 2001, Visualization of meiotic events in intact living anthers by means of two-photon microscopy, Micron 32:679-684. Feij6, J.A., and Moreno, N., 2004, Imaging plant cells by two-photon excitation, Protoplasma 223: 1-23. Fricker, M., Parsons, A., Tlalka, M., Blancaflor, E., Gilroy, S., Meyer, A., and Plieth, C., 2001, Fluorescent probes for living plant cells, In: Plant Cell Biology, A Practical Approach (e. Hawes and B. Satiat-Jeunemaitre, eds.), Oxford University Press, Oxford, United Kingdom, pp. 35-83. Fricker, M.D., and Meyer, A.J., 2001, Confocal imaging of metabolism in vivo: Pitfalls and possibilities, f. Exp. Botany 52:631-640. Fricker, M.D., May, M., Meyer, AJ., Sheard, J., and White, N.S., 2000, Measurement of glutathione levels in intact roots of Arabidopsis, f. Microsc. 198:162-173. Galbraith, D.W., Bohnert, HJ., and Bourque, D.P, eds., 1995, Methods in Cell Biology, Academic Press, New York. Galway, M.E., Heckman, J.w. Jr., Hyde, G.J., and Fowke, L.e., 1995, Advances in high-pressure and plunge-freeze fixation, In: Methods in Plant Cell Biology (D.W. Galbraith, HJ. Bohnert, and D.P. Bourque, eds.), Academic Press, New York, pp. 4-20. Grew, N., 1673, An idea of a phytological history propounded. Together with a continuation of the anatomy of vegetables, particularly prosecuted upon roots. With an account of the vegetation of roots grounded chiefly thereupon. Published by Nehemiah Grew. Gunning, B.. 2002, Plant cell biology on CD, Author edition, www. plantcellbiologyoncd.com. Gunning, B.S., and Steer, M.W., 1996, Plant Cell Biology: Structure and Function, Jones and Bartlettt Publishers, Boston. Gutierrez-Alcala, G., Gotor, e., Meyer, A.J., Fricker, M., Vega, J.M., and Romero, L.e., 2000, Glutathione biosynthesis in Arabidopsis trichome cells, Proc. Natl. Acad. Sci. USA 97:11108-11113. Hake, S., 2001, Mobile proteins signals cell fate, Nature 413:261-263. Haseloff, J., 1999a, GFP variants for multispectral imaging of living cells, Methods Cell BioI., 58: 139-151.
Haseloff, J., 1999b, Imaging green fluorescent protein in transgenic plants, In: Imaging Living Cell (R. Rizzuto and C. Fasolato, eds.), Springer-Verlag, Berlin, pp. 40-94. Haseloff, J., 2003, Old botanical techniques for new microscopes, Biotechniques 34: 1174-1178. Haseloff, J., and Amos, B., 1995, GFP in plants, Trends Genet. 11 :328-329. Haseloff, J., Siemering, K.R., Prasher, D.e., and Hodge, S., 1997, Removal of a cryptic intron and subcellular localization of green fluorescent protein are required to mark transgenic Arabidopsis plants brightly, Pmc. Natl. Acad. Sci. USA 94:2122-2127. Hawes, e., and Satiat-Jeunemaitre, B., eds., 2001, Plunt Cell Biology, A Practical Approach, Oxford University Press, Oxford, United Kingdom. Hepler, P.K., and Gunning, B.E.S., 1998, Confocal fluorescence microscopy of living cells, Protoplasma 201: 121-157. Holmes, TJ., Bhattacharyya, S., Cooper, J.A., Hanzel, D., Krishnamurthi, v., Lin, W., Roysam, B., Szarowski, D., and Turner, J., 1995, Light microscopic images reconstructed by maximum likelihood deconvolution, In: Handbook of Biological Confocal Microscopy (J.B. Pawley, ed.), Plenum Press, New York, pp. 389-402. Kam, Z., Hanser, B., Gustafsson, M.G.L., Agard, D.A, and Sedat, J.W., 2001, Computational adaptive optics for live three-dimensional biological imaging, Proc. Natl. Acad. Sci. USA 98:3790-3795. Ketelaar, T., Anthony, R.G., and Hussey, P.J., 2004, Green fluorescent proteinmTalin causes defects in actin organization and cell expansion in Arabidopsis and inhibits actin de-polymerizing factor's actin depolymerizing activity in vitro, Plant Physiol. 136:3990-3998. Kodama, H., and Komamine, A., 1995, Synchronization of cell cultures of higher plants, In: Methods in Plant Cell Biology (D.W. Galbraith, H.J. Bohnert, and D.P Bourque, eds.), Academic Press, New York, pp. 315-330. Kohler, R.H., Zipfel, W.R., Webb, W.W., and Hanson, M.R., 1997, The green fluorescent protein as a marker to visualize plant mitochondria in vivo, PlantJ. 11:613-621. Konig, K., 2000, Multiphoton microscopy in life sciences, f. Microsc. 200:83-104. Lloyd, e.W., 1987, The plant cytoskeleton: The impact of fluorescence microscopy, Annu. Rev. Plant Physiol. 38: 119-139. Malkin, R., and Niyogi, K., 2001, Photosynthesis, In: Biochemistry & Molecular Biology of Plants (B.B. Buchanan, W. Gruissem, and R.L. Jones, eds.), American Society of Plant Biologists, New York, pp. 568-628. Mason, W.T., 1993, Fluorescent and luminescent probes for biological activity, Academic Press, London. McInerney, T., and Terzopoulos, D., 1996, Deformable models in medical image analysis: a survey, Med. Image Anal. 2:91-108. McNally, J.G., Karpova, T., Cooper, J., and Conchello, J.A, 1999, Threedimensional imaging by deconvolution, Microsc. Methods 19:373-385. Meyer, A.J., and Fricker, M.D., 2000, Direct measurement of glutathione in epidermal cells of intact Arabidopsis roots by two-photon laser scanning microscopy, J. M ic rosc. 198: 174-1 81. Meyer, A.J., May, M.J., and Fricker, M., 2001, Quantitative in vivo measurement of glutathione in Arabidopsis cells, Plant f. 27:67-78. Monck, J.R., Oberhauser, A.E, Keating, T.J., and Fernandez, J.M., 1992, Thinsection ratiometric Ca2+ images obtained by optical sectioning of fura-2 loaded mast cells, f. Cell BioI. 116:745-759. Nakajima, K., Sena, G., Nawy, T., and Benfey, P.N., 2001, Intracellular movement of the putative transcription factor SHR in root patterning, Nature 413:307-311. Parthasarathy, M.V., 1995, Freeze-substitution, In: Methods in Plant Cell Biology (D.W. Galbraith, HJ. Bohnert, and D.P. Bourque, eds.), Academic Press, New York, pp. 57-70. Potter, S., 1996, Vital imaging: Two photons are better than one, Curr. BioI. 6: 1595-1598. Raghavan, V., 1995, Manipulation of pollen grains for gametophytic amd sporophytic types of growth, In: Methods in Plant Cell Biology (D.W. Galbraith, H.J. Bohnert, and D.P. Bourque, eds.), Academic Press, New York, pp. 367-377. Reddy, G.V., Heisler, M.G., Ehrhardt, D.W., and Meyerowitz, E.M., 2004, Real-time lineage analysis reveals oriented cell divisions associated with morphogenesis at the shoot apex of Arabidopsis thaliana, Development 131:4225-4237.
Imaging Plant Cells • Chapter 44
Ruzin, S.E., 1999, Plant microtechnique and microscopy, Oxford University Press, Oxford. Running, M.P, Clark, S.E., and Meyerowitz, E.M., 1995, Confocal microscopy of the shoot apex, In: Methods in Plant Cell Biology (D.W. Galbraith, H.J. Bohnert, and D.P. Bourque, eds.), Academic Press, New York, pp. 217-230. Shav-Tal, Y., Singer, R.H., and Darzacq, X., 2004, Imaging gene expression in single living cells, Nat. Rev. Mol. Cell BioI. 5:856-862. Shaw, P.J., 2001, Introduction to optical microscopy for plant cell biology, In: Plant Cell Biology, A Practical Approach (c. Hawes and B. SatiatJeunemaitre, eds.), Oxford University Press, Oxford, United Kingdom, pp. 1-33. Shaw, S.L., Kamyar, R, and Ehrhardt, D.W., 2003, Sustained microtubule treadmilling in Arabidopsis cortical arrays, Science 300:1715-1718. Sheen, 1., 1995, Methods for mesophyll and bundle sheath cell separation, In: Methods in Plant Cell Biology (D.W. Galbraith, HJ. Bohnert, and D.P. Bourque, eds.), Academic Press, New York, pp. 305-314. Shimmen, T., and Yokota, E., 2004, Cytoplasmic streaming in plants, Curro BioI. 16:68-72. Siemering, K.R, Golbik, R, Sever, R., and Haseloff, J., 1996, Mutations that suppress the thermosensitivity of green fluorescent protein, Curro Bioi. 6:1653-1663. Spence, 1., 2001, Plant histology, In: Plant Cell Biology, A Practical Approach (c. Hawes and B. Satait-leunemaitre, eds.), Oxford University Press, Oxford, United Kingdom, pp.189-206. Sun, c.-K., Chu, S.-W., Chen, I.-S., Liu, T.-M., Cheng, P.-c., and Lin, B.-L., 2001, Multi-modality nonlinear microscopy, Con! Lasers Electro Opt. Eur. Tech. Dig. 2001 :222-227.
787
Tirlapur, U., and Konig, K., 2001, Femtosecond near-infrared lasers as a novel tool for non-invasive real-time high-resolution time-lapse imaging of chloroplast division in living bundle sheat cells of Arabidopsis, Planta 214:1-10. Tirlapur, U., and Konig, K., 2002, Two-photon near-infrared femtosecond laser scanning microscopy in plant biology, In: Confocal and Two-Photon Microscopy: Foundations, Applications and Advances (A. Diaspro, ed.), Wiley-Liss, New York, pp. 449-468. Volkmer, A., Subramaniam, v., Birch, D.l.S., and lovin, T.M., 2000, One and two-photon excited fluorescence lifetimes and anisotropy decays of GFPs, Biophys. J. 78: 1859-1898. von Arnim, A.G., Deng, X.M., and Stacey, M.G., 1998, Cloning vectors for the expression of green fluorescent protein fusion proteins in transgenic plants, Gene 221:35-43. Vroom, 1.M., de Grauw, KJ., Gerritsen, H.C., Bradshaw, DJ., Marsh, P.D., Watson, J.J., Birmingham, c., and Allison, C., 1999, Depth penetration and detection of pH gradients in biofilms by two-photon excitation microscopy, Appl. Environ. Microbiol. 65:3502-3511. Wysocka-Diller, 1.W., Helariutta, Y., Fukaki, H., Malamy, 1.E., and Benfey, P.N., 2000, Molecular analysis of SCARECROW function reveals a radial patterning mechanism common to root and shoot, Development, 127:595-603. Xia, A.D., Wada, S., Tashiro, H., and Wuang, W.H., 1999, One and two-photon fluorescence from recombinant GFP, Arch. Biochem. Biophys. 372:280-284. Xu, c., and Webb, W., 1996, Measurement of two-photon excitation cross sections of molecular fluorophores with data from 690 to 1050 nm, J. Opt. Soc. Am. B 13:481-491.
45
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells Irina Majoul, Yiwei Jia, and Rainer Duden
INTRODUCTION
How to make a good science La luna densa The moon is dense, Ogni densa e grave Everything dense is substantial. Come sta la luna? How, moon, do you exist? Leonardo da Vinci (1452-1519}.1
For the purpose of this chapter, a free paraphrasing of Leonardo's point might be: Living cells are moving. Movement reflects their functional activity. How, cells, do you function? After formulating this philosophic question in a poetic form, Leonardo the Scientist, provides us with a real experimental (optical) setup. "As I propose to treat the nature of the moon, it is necessary that I first describe the perspective of mirrors, whether plane, concave, or convex," (B .M.94r - Arundel MS in British Museum). Next, in the pages of Codex Atlanticus (C.A.190r), Leonardo invites us to "Construct the glasses to see the moon magnified" and half a millennium later we are still following him for, as Bulgakov famously said, "Manuscripts do not bum! " As a master of light and shadow, he knew how the uneven surface of the ocean reflects light in all directions, breaking it into myriad beams. Analyzing the brightness of the light reflected from the moon, Leonardo suggests that its outer surface might be covered with a rippled liquid. As a scientist he was able to understand how the images of the moon's phases would appear to an observer on the Earth [Fig. 4S .2(A), arrow]. In its monthly cycle the moon always exposes the same side towards the Earth. Only when it is opposite to the Sun (with respect to an Earth observer) is the moon fully illuminated by sunlight. Leonardo follows his interest further to conclude that when the moon is on the same side of the Earth as the Sun, it will be illuminated by brilliant daylight reflected from the watery surface of the Earth. This situation, known in astronomy as lumen cinerium or "new moon in the old moon's arms," was described and drawn by Leonardo (see Fig. 4S .l) . Perhaps it was this under-
1
Leonardo - Art and Science, 2000, Crispino, ed., Giunti Editore S.p.A. a, Florence Milan.
standing that later led him to question whether the Sun really does circle around the Earth. Much later, Johannes Kepler reached similar conclusions in "Astronomiae pars optica." The scientific prediction of Leonardo (based on optics) may have met "real life" when Amerigo Vespucci described the New World to him, possibly during the painting of a portrait by Leonardo that is mentioned in Lives of Artists by Giorgio Vasari: "sketch of Amerigo Vespucci shows head of a very handsome old man drawn in charcoal." For us here, what Leonardo discovered is less important than the approach he used to experience nature and to obtain new knowledge. His further research in optics led him to conclude that perspective images would be closer to reality if they are projected not onto a flat paper but onto a concave spherical surface, as happens in the human eye. In Figure 4S.2(B) we see a painter, that might be Leonardo himself, working with perspectrograph (from Codex Atlanticus, folie Sr). Next, Leonardo concludes, that "the gradation of light but not of the colors is what defines the depth in space" (Manuscript G, 8r). Today, in modem microscopy we still use these ideas: DIC images code shape and brightness, color information is used to draw attention to the locations of transfected fluorescent proteins and our best 3D rendering algorithms are
FIGURE 45.1. Leonardo da Vinci, image of a moon phase from the Codex Leicester: 35v, 2r (2A).
Irina Majoul and Rainer Duden· Roya l Holloway University of London, Egh a m TW20 OEX, United Kingdom Yiwe i Jia • Olympus America Inc., Me lville, New York 11747 788
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of living Cells • Chapter 45
789
FIGURE 45.2. Leonardo da Vinci. (A) Schematic drawing of the light reflected from the moon that can be analyzed by an observer on the Earth: Codex Leicester [drawn between IS08-ISIO, 3Sv, 2r (2A)]. (B) Painter working with perspectrograph, Codex Atlanticus, folie Sr (fragment). (C) Schematic drawing of the laser light path used to illuminate the object in an inverted microscope. (Right) Fluorescence image of a living cell obtained by this setup, inserted DIC image (for details, see text). Both arrows indicate positions of the observer's eye.
designed to code objects that are farther away so that they appear less bright, less distinct and sometimes even bluish [Fig. 45.2(C)]. Although Leonardo had never seen a microscope, his mind worked with the precision of this useful device: starting from a general overview, dissecting and moving to higher magnifications where he captured the smallest details necessary to finally understand the essence of the entire object under investigation. In a sense, Leonardo's habit of portraying biological structures in 3D foreshadows our modern interest in 3D microscopy. Not yet trapped in the 2D world of the standard light microscope, his view was inherently three-dimensional. He knew instinctively that 3D insight was necessary to understand how biological surfaces must
interdigitate to maximize the surface area of the adjoining regions. Figure 45.3(A), shows the interface between contacting surfaces in a womb wall (RL manuscript, 19102r, detail). His 3D image of these surfaces is impressively correct and in agreement with our modern understanding of cellular contacts on a much smaller scale: that of synapses [Fig.45.3(B)] and cell-cell contacts [Fig. 45.3(C)], resolyed by modern microscopes almost 500 years later. His way of following "functionality" and performing "preliminary studies" (that might take him many years) was so correct that, even in the 21 st century, his approach remains as important for us as it was for him. Leonardo was a great philosopher and the first one to integrate the four most productive aspects of scientific
FIGURE 45.3. (A) Taken from an anatomical study of Leonardo da Vinci IS08-ISI0 (Manuscript RL, 19102r, Windsor Library, UK). The left panel shows the interface between dissected tissues (wall of a womb) where Leonardo used the tools available to him to analyze how two biological surfaces create an adhesive junctional interface. (B) An insert representing junctional folds of a synaptic terminal that creates integrity of the structure (reproduced with permission from Cohen-Cory, 2002). (C) Junctional cell-cell inerface between two cultured primary astrocytes revealed with antibodies against a gap junction protein, Connexin43 (Butkevich et ai., 2004).
790
Chapter 45 • I. Majoul et al.
thinking: (1) Follow your naive interest; (2) Open your mind, get new ideas and try to understand how things work; (3) Have the patience to develop new and versatile experimental approaches to get new information; (4) Test the new knowledge against your wider experience. We too are at the beginning of a new era in biology and biotechnology, with neurobiology and stem cell research leading to prominence. Although our tools are now molecular imaging, 3D fluorescence microscopy, proteomics, genetic screens, spectroscopy, nanobiophysics, animal models, and cellular physiology, the overall process remains much the same.
Beauty, Functionality, Cell Cycle, and living-Cell Imaging Rule 1: Get To Know Your Cells Before moving on to the methods of fluorescence resonance energy transfer (FRET) analyses, it is important to include a reminder about the main object of our investigations: living cells. Living cells imaged in the microscope allow one to obtain insight into the complexity of dynamic processes in real time, whereas fixed cells maintain only the last steady state distribution (at the moment of death) for the molecules of interest. For live-cell FRET analyses, it is crucial to keep cells in good condition during the time course of an experiment. This includes temperature and CO 2 control of the medium, proper transfection of the cells with fluorescent proteins tested for functional activity, minimal level of laser illumination throughout the experiment, etc. The beauty of transfected cells may often serve as an indicator oftheir biological functionality, and thus can be used as an (almost) scientific criterion. In a sense, beauty is a biological functionality. For example, in cells transfected with yellow fluorescent protein (YFP)-tubulin to highlight the microtubule network, microtubules are impressively beautiful when the cells are taken straight from the incubator and recorded at 37°C. Cells forgotten on the table without temperature control or adequate CO 2 will show partially depolymerized, ugly microtubules. In such cells, transport of proteins in the secretory and endocytic pathways is impaired, and the cells are dying. Investing hard work into badly transfected cells or attempting functional studies on dying cells is a wasted effort (except, of course, when studying apoptosis). In contrast, beautifully transfected cells are a source of a great happiness to the researcher, stimulating himlher to acquire the best possible images. Figure 4S.4 shows two examples of cells after transfection: Figure 4S.4(A) shows a cell expressing a physiological level of the transmembrane KDEL receptor (ERD2) that is recognized by cholera toxin to enable toxin transport through the Golgi complex (Majoul et at., 1996, 2001). Figure 4S.4(B) shows a cell imaged at longer times after transfection and that is now over-expressing KDEL receptor, which aggregates around the Golgi. Cholera toxin recognizes and binds to functional KDEL receptor molecules [Fig. 4S.4(A)], but does not enter the over-expressing structures [Fig. 4S.4(B)]. Instead, cholera toxin recognizes native, still functional receptors between the aggregates and thus enters the Golgi [Fig. 4S.4(B), arrow]. Clearly, the cells in Figure 4S.4(B) should not be used for experiments. The transfection setup, transfection reagents, cells, and possibly even the chimeric proteins must be further optimized to prepare a better experiment. Because many biological processes are cell cycle-dependent, the cell cycle stage should be considered for FRET experiments. Figure 4S.4(C) shows different degrees of surface expression for a glycosylphosphatidylinositol (GPI). GPI-anchored protein is
detected with antibodies that make FRET under the plasma membrane. Acceptor bleach revealed different degrees of FRET, showing that the level of GPI-anchored protein expression varies with the stage of the cell cycle. We have also shown that the appearance of lipid receptors (glycosphingolipids) on the cell surface, and thus the lipid composition of the plasma membrane as well as the distribution of certain lipid-binding proteins under the plasma membrane, is strongly cell cycle-dependent (Majoul et at., 2002a). Surface expression of lipid receptors can be mapped to a specific cell cycle stage using incorporation of BrDU to mark S phase [Fig. 4S.4(D)] or by monitoring the expression of cyclins, for example, cyclin B (mitotic marker). To map different lipid microdomains on the plasma membrane, we used fluorescently labeled ABS toxins (cholera toxin and shiga toxin) and found a strong cell cycle-dependent heterogeneity of binding to cell surfaces [Fig. 4S.4(E)]. The molecular basis of this mechanism is an intracellular lipid modification that involves modification of a common lipid precursor (e.g., lactosyl ceramide) with sugar moieties by galactosyl- and glucosyltransferases and the expression of these enzymes turned out to be cell cycle-dependent [Fig. 4S.4(F); Majoul et at., 2002a]. Thus, the behavior of many analyzed proteins will be different in the G I, S, G2, or M phases of the cell cycle. The take-home message here is: if a strong FRET heterogeneity is observed in a cell population, it may be useful to synchronize cells and to test directly whether this heterogeneity is cell cycle-dependent.
FLUORESCENCE RESONANCE ENERGY TRANSFER THEORY More than SO years ago, the German scientist Forster discovered that close proximity of two chromophores changes their spectral properties in predictable ways (Forster, 1948a). If two biological molecules in a living cell interact functionally, the interaction will bring into close proximity any chromophores attached to these molecules. After collecting spectral data and deciphering this information based on Forster's theory, we can understand changes that occurred in the near-vicinity microenvironment of two chromophores and thus draw biologically meaningful conclusions about the specificity of protein-protein interactions. The idea of non-radiative dipole-dipole interactions was first formulated by Forster in 1948 (Forster, 1948a,b). Further development of this idea, biological application, and introduction of spectrally distinct fluorescent molecules came later (Stryer and Haugland, 1967; Clegg, 1992; Tsien, 1998; Lacowicz, 1999). Forster's theory explains FRET as a dipole-dipole interaction between neighboring molecules (Fig. 4S.S). Forster showed that the presence of an appropriate acceptor in FRET proximity will decrease the time a donor spends in the excited state due to nonradiative energy transfer to the acceptor in close proximity:
kt D+A+hvE::::} D*+A::::} D+A * (1)
Here, kt is the rate of energy transfer from the donor to the acceptor, ko = kOf + kOi and kA = kA + kAi , ko and kA correspond, respectively, to the rate of the radiative decay of the donor and the acceptor (for us here, fluorescence emission), and kOi and kAi to the rates of the non-radiative decay constant of the donor (D) and the acceptor (A). The ratio between the number of photons emitted
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
791
A
B
c
o
E
F
PM
1
Goigi
t
ER
FIGURE 45.4. (A) Cells transfected with a transmembrane Golgi receptor (ERD2-CFP/ERD2-YFP). At near-physiological expres sion levels, it is locali zed in the Golgi where it binds to cholera toxin (red). (B) When over-expressed, the receptor aggregates and is excluded from the Golgi. In this case. cholera toxin sti ll recognizes a portion of the native receptor (shown by arrows) and can thus enter the Golgi compartment. (C) FRET demonstrated by acceptor bleach on the surface of a cell monolayer labeled with antibodies against GPI-anchored proteins and 5' NT. The cell-to-cell variation in FRET (Ecn) distribution of 5' NT molecules on the cell surface may have deep biological roots. (Taken from Kenworthy and Edidin, 1998.) (D) Incorporation of BrDU can be used to determine the stage of the cell cycle in indi vidual cells in a non-synchronized cell monolayer. Only cells carrying out DNA synthesis (S phase) will be labeled. (E) Overlay of the same monolayer with labeled CTX (cholera toxin, used as a lipid raft marker) and ST (Shiga toxin) revealed that cell s in S phase have weak binding of both toxins. Cells in G I phase express more GMI on the cell surface and bind CTX more strongly (shown in green). Gb3-bound ST is shown in red . (F) Cartoon representing cell cycle-specific modifications of a common lipid precursor lactosyl ceramide that can be modified by glycosyl and galactosyltransferases to produce more branched GMI (G I phase) or Gb3 (around M phase). Thus, differential expression of lipid receptors on the plasma membrane can be a marker for the phase of the cell cycle.
792
Chapter 45 • I. Majoul et al.
(2) Donor - Acceptor orientation (kappa squared)
(~)
Thus, the quantum yield of the donor in the presence of the acceptor will be:
1(2= (Cos 9T - 3 Co 9D' Co 9A)2
Arbitrary FRET
y
(3)
The fraction of the photon energy absorbed by the donor that can be transferred without radiation to an acceptor represents the efficiency of energy transfer. E ell . The relationship between Eell and the quantum yield of the donor in the presence and absence of acceptor can be described as:
(4) D~---
9D = 9A = 9T = 0; COS 9T = Co 90 = COS 9A = 1
Maximal FRET
where QDAIQD is the efficiency of the radiative process that, added to the efficiency of the non-radiative energy transfer (FRET) equals 1. Eell can be also described in terms of k" k D, kDi by substituting of Eqs. 2 and 3 into Eq. 4:
y
Y
1(2= 4
Eell
--
kt
MD
D Minimal FRET
A
1M"
QDAIQD
=1-
(kD + kDi)/(k, + kD + kDJ
k, I k, + kD + kDi
z X
X
=1-
DA
90 = 9T = 90; 9A = 0; CO 9T = Cos 90 = 0; COS 9A = 1
(5) (5a)
To perform practical FRET measurements, one needs to keep in mind that the efficiency of energy transfer (Eell) to the acceptor will be strongly decreased by an increase in both radiative (k D ) and non-radiative (kD ;) decay of the donor. FRET efficiency is a direct measure of the photon energy absorbed by the donor and transferred to an acceptor (Figs. 45.5 and 45.6). Eell depends strongly on the actual proximity R between the donor (D) and the acceptor (A) molecules and on their mutual orientation (Fig. 45.5). For efficient FRET to occur, the distance (R) between the excited donor and the fluorescent acceptor needs to be typically 2 to 7 nm (Forster, 1948a,b; Stryer, 1978; Lakowicz, 1999; Patterson et al., 2000). The dependence of Eell on R was described by Forster as:
(6) The distance between the donor and the acceptor for which Eell = 0.5 is called Ro and reflects the properties of a particular DIA
l1li BP 4eII17
8IIlLPI5IOnm
FIGURE 45.5. FRET theory. (Upper panel) Proximity and spatial orientation between the donor (D) and the acceptor (A) pair. An increase in the Forster radius between donor and acceptor (R DA ) will lead to a decrease in efficiency of energy transfer. Mutual orientation (angles en, eA, eT ) between the donor and acceptor molecules influence the transfer efficiency E off' (Lower panel) Jablonski diagram for single- and two-photon excitations of the donor may result either in the emission of the donor (CFP Aem = 474nm) or (in case of a close proximity) to the non-radiative dipole-dipole transfer of energy (kt ) to the acceptor molecule followed by the Stokes shift and the emission of fluorescence by the acceptor molecule at a longer wavelength (YFP, Aem = 526nm).
to the number of photons absorbed by the donor or the acceptor is known as quantum yield of the donor (QD) or of the acceptor (QA)' The lower the quantum yield of the donor (QD). the more photon energy will undergo either non-radiative decay kDi compared to the radiative kD fluorescence emission of the donor or possible FRET (with k" the rate of FRET transfer). The quantum yield of the donor in the absence of an acceptor is represented as:
pair. It is a function of the quantum yield of the donor (QD), integral spectral overlap between the donor emission and the acceptor excitation, iDA and a factor reflecting their relative spatial orientation (~) (Figs. 45.5 and 45.6). If 1( is known and fixed, then the FRET efficiency between two chromophores can be taken as a direct measure of R, using Forster's conversion formulae (Eq. 6). Quantitative FRET analyses of different subcellular compartments in a living cell requires that we know the real value of the spatial orientation factor (1(2). Figure 45.5 describes the common arbitrary case and two extreme cases: maximal FRET to no FRET based on orientation factor (1(2). A Jablonsky diagram for single- and two-photon excitation in FRET [cyan fluorescent protein (CFP) donor, YFP acceptor] is shown below (Fig. 45.5). Our preliminary data revealed that the same chromophores, CFP and YFP, measured with the identical microscope will display different levels of FRET when attached to proteins from the different subcellular compartments. Two transmembrane Golgi donor and acceptor proteins (i.e., donor and acceptor that can oligomerize only by diffusing laterally in the plane of the Golgi membrane) display less FRET than the combination of the transmembrane acceptor-YFP and a free cytosolic donor-CFP that interact with each other (as shown in Fig. 45.8). It is clear that the rotational mobility, and thus relative spatial orientation factor (1(2), can be higher for the cytosolic proteins than for the laterally fixed transmembrane fluorescent proteins. Because in most experiments the actual orientation is difficult to determine, it is common to assume a random orientation of the
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45 FICURE 45.6. Spectra illustrating the FRET effect. Spectrofluorimetry allows one to follow the time-dependent appearance of FRET (left). The decrease in donor fluorescence (long arrow) correlates with the increase in sensitized emission of the acceptor (short arrows). Spectral overlap (lOA) between the emission spectrum of the donor (CFP) and the excitation spectrum of the acceptor (YFP) ensures that FRET can potentially occur between a selected pair of biological molecules (e.g., proteins X and Y of interest), if they functionally interact in living cells (upper right comer). Formulae used for Ro calculations (R" corresponds to E,,, = 50% of a maximal). For a detailed description see the FRET Theory section in the text.
793
FRET
o . decrease In donor fluorescence
kl
12000000
O· A
10000000
CA·
Eeff
8000000
0 ,5
0
6000000
0
Ro
4000000 2000000 , O~
emission of acceptor __ __________p -______ ~
486
green fluorescent proteins (GFPs) even when they are fused to macromolecules. This uncertainty, plus the uncertainty in chromophore location within the entire chimeric protein, means that FRET with GFP variants is more suited to detecting changes in conformation or percentages of association than to quantifying the absolute value of R. The dominant uncertainty causing Ro or EFRET fluctuation s is the relative orientation of dipoles, as is described for three particular cases in Figure 45.5. When the labeled proteins are not known to be attached to a larger structure, it is common to assume that the two dyes are freely rotating on a time scale comparable to, or faster, than the fluorescence lifetime and that 12 can be dynamically averaged to a fixed value of 12 = 2/3. To test whether this assumption holds true for new constructs under investigation, ensemble polarization anisotropy measurements can be performed on selected constructs (Lacowicz, 1999; Mattheyses et aI., 2004). For free cytosolic proteins, or a combination of a cytosolic and a membrane-attached protein, it has been found that both free and attached GFP have large anisotropy values (around r = 0.20), indicating obstructed rotational diffusion. In this chapter, calculated values of R o are based on the ass umption that ~ = 2/3, even though we recognize that it is possible that in some subcellular compartments K2 can be as high as 2 or even 4 (Fig. 45.5). If the wavelength is measured in centimeters and the molar absorption in M - I cm- I , and the IDA value is given in M- 1 cm- 3 , then the following formula results: R~ = 8.79
X
10- 25 (QD
K2 ·n-4 ID A }
cm 6
10-5 {QD
K2·n-4
JDA} A6 (in Angstroms)
nm
between selected GFP variants can be used for FRET calculations (Patterson et aI., 2000). As the energy absorbed by the single dipole will induce this dipole (e.g., donor) to oscillate along X-, Y-, and z-oriented directions ·(Fig. 45.5), this energy will be in the reverse proportion to the radius R3. With the appearance of the second oscillating dipole in resonance proximity (acceptor), their common energy will be calculated in reverse proportion to the ~. Eeff is highly and inversely dependent on the distance (R) between the donor and the acceptor chromophores, and can be obtained in the experiment. The Forster equation can in principle be used to calculate the intermolecular distances on a nanometer scale (i.e., FRET = nanobioscopy):
(9) The actual distance between a particular pair of donor and acceptor molecules can be calculated if their Forster radius (Ro) is known from previous experiments. Calculations of R", based on the spectral properties of some donor- acceptor pairs of fluorescent proteins are given in Table 45 .1. From Eq. 9, it follows that an increase in the distance between acceptor and donor from R = Ro to R = 2R" will decrease Eerr from 50% to 1.5%. Ro is determined experimentally as the distance in Angstroms (1 nm = lOA). Eeff can be obtained directly from the experiment: (10)
(7)
However, if the wavelength is depicted in nanometers and the molar absorption in M - ' em- I, and the IDA value is given in M- I cm- I nm-4, the Eq. is the following (see also Fig. 45.6):
R! = 8.79 x
560
~~~~~~~--~
TABLE 45.1. Calculated Donor-Acceptor Distance (Ro) in Nanometers (Patterson et a/., 2000) Acceptor (GFP Variant or DsRed)
(8)
A number of experimental data have confirmed the linear dependence of Eeff on the value of I in the range of distances used to measure FRET. It is important to mention here that, as the GFP variants exhibit relatively small Stokes shifts, Ro values calculated
Donor Cyan Green Yellow
Green 4.82 4.65 3.25
Yellow
Red
4.92 5.64 5.11
4.17 4.73 4.94
794
Chapter 45 • I. Majoul et al.
where F is the fluorescence intensity measured in the experiment. FOA is the fluorescence intensity of the donor in the presence of acceptor and Fo is the fluorescence intensity of the donor alone. FRET efficiency Eeff can also be measured as the relative fluorescence of the donor in the presence (Dbb) and absence (Dab) of the acceptor (i.e., after photobleaching of the acceptor). As it is usually difficult to achieve complete photoinactivation of the acceptor in living cells, an image of the acceptor is acquired both before bleach (Abb) and after bleach (Aab). The ratio of Aab/Abb then reflects the percentage of non-bleached acceptor that needs to be included in calculations of FRET efficiencies. Acceptor photobleaching allows one to obtain the experimental parameters Dbb and Dab, used for FRET calculations. Ectf
=1-
DbblDab
(11)
Anisotropy analysis of chimeric fusion proteins can also provide a solid basis for structural analysis (Velez and Axelrod, 1988). Different experimental variations of FRET measurements have been developed for biological applications within the last two decades and continue to be improved in a growing number of laboratories around the world. Nevertheless, it is still useful to perform spectral analyses for any new chromophores selected for FRET to control for their exact properties once they are inside the cell line selected for experiments. Alternatively, one can calculate FRET efficiency from fluorescence lifetime data [1 from a fluorescence lifetime imaging microscope (FUM) (Lakowicz, 1999; and Chapter 27, this volume]. Using the previous equations, we can describe lifetime as: '1:
= Qlk
where Q is the quantum yield of the f1uorophore and kr the number of photons emitted. The lifetime of the donor in the presence of the acceptor is described as: '1: 0A =
QOAlko = I/(k, + ko + k Oi ) or kT + ko + k Oi = l!'!oA
(12)
The lifetime of the donor in the absence of the acceptor will be '1:0
= QoIko = l/(ko + k Oi ) ko + k Oi = 1/'1:0
(13)
Substituting Eq. 12 and Eq. 13 into Eq. 5, FRET efficiency can be calculated directly from the experimental measurements of the lifetime of the donor in the presence ('1: 0A ) and in the absence of the acceptor ('1: 0 ): (14)
FLUORESCENT PROTEINS AND FLUORESCENCE RESONANCE ENERGY TRANSFER Considering all the non-fluorescent proteins in living organisms, the discovery and, ultimately, cloning of a naturally fluorescent one, GFP, from the marine organism Aequoria victoria came as a lucky surprise (Ward et at., 1982; Prasher et al., 1992; Chalfie et al., 1994; Chalfie, 1995; Tsien, 1998). What drove evolution to create a fluorescent protein? One reasonable theory suggests that GFP molecules appeared as a byproduct in the development of antioxidant systems that evolved into the pocket of an excitable chromophore (Heim et aI., 1994; Labas et aI., 2002). The story is made more interesting by the fact that the native GFP chromophore is excitable with blue light (at a peak of 395 nm) that is internally
transferred by a FRET mechanism, resulting in emission at 508 nm (Heim and Tsien, 1996). Thus, FRET already exists as an intrinsic property of native GFP (Ward et aI., 1982; Heim et aI., 1995). It took about 20 years after the discovery of GFP before it was put to practical use in the field of cell biology, but then it produced a scientific revolution (Ward et aI., 1982; Prasher et aI., 1992; Chalfie et aI., 1994; Tsien, 1998, 2004; Lippincott-Schwartz and Patterson, 2003). GFP-based methodologies and live-cell imaging, combined with FRET and other biophysical approaches, yielded a vast amount of information, allowing one to study and dissect biochemical pathways at the single cell level (Hirose et aI., 1999; Miyawaki et at., 1999; Janetopoulos et at., 2001; LippincottSchwartz and Patterson, 2003; Jones et aI., 2004; Kusumi et aI., 2005). GFP, and its spectrally shifted variants, CFP and YFP (Heim and Tsien, 1996; Tsien, 1998; Patterson et aI., 2001) are particularly useful, either for colocalization studies or as partners for FRET-proximity imaging experiments (Mochizuki et al., 2001). CFP was created by introduction of the S65A, Y66W, S72A, N146I, M153T, and Vl63A mutations into GFP, resulting in an excitation peak of 434nm and an emission maximum at 474nm. YFP differs from GFP by the S65G, V68L, S72A, and T203Y mutations and has an excitation peak of 514 nm and an emission maximum at 527 nm (Heim and Tsien, 1996; Tsien, 1998). Good spectral overlap between CFP and YFP currently favors these two chromophores as the best pair for analyzing protein-protein interactions. GFP provides the great advantage of allowing us to follow in a living cell any cellular protein for which the coding sequence is known. GFP-based chimeric proteins have been expressed in an almost unlimited range of different cell types, including neurons, in tissue slices, and even in whole living organisms (transgenic mice, rats, rabbits, fish, etc.). Often GFP chimeras are placed under tissue-specific promoters. The spectral variants of GFP that represent donor-acceptor pairs allow us not only to map intracellular protein localization, but also to analyze their interactions in different cells and subcellular compartments. GFP chimeras can be targeted to different cellular localizations. As GFP itself is a relatively small protein (-27kDa) with a chromophore inside a cylinder of 2.4nm in diameter, the centers of the two chromophores cannot get closer to each other than their diameter. Maximal FRET between CFP and YFP was measured at a distance of 4.9 to 5.2nm, demonstrating FRET (proximity) in a size range of protein-protein complexes (Tsien, 1998). Stryer's dream of FRET as a "Spectroscopic Ruler" started to become a reality. FRET between f1uoroscein and rhodamine molecules was first measured a long time ago and Ro values of 4.9 to 5.4 nm have been known and used for the FITC (fluorescein isothiocyanate) and TRITC (tetramethyl rhodamine) pair as chromophores with high energy transfer efficiency (Johnson et aI., 1984). DsRed, derived from the Discosoma coral, is genetically different from GFP (Matz et aI., 1999). An intrinsic tendency of DsRed to tetramerize and its slow maturation compared to GFP prevented its broad application for colocalization or FRET studies. However, DsRed has recently been "re-evolved" by extensive mutagenesis, resulting in bright monomeric versions, called mRFPl or mCherry, which show improved fluorescence and maturation times (Campbell et ai., 2002; Shaner et aI., 2004). Using GFP as a donor, these mRFPs can now be used for FRET analyses. Small fluorescent molecules (e.g., Cy2, Cy3, Cy5, Alexa-488, Alexa-543) can be easily attached to bacterial toxin molecules without changing their activity and can then be used for monitor-
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
ing toxin actions in living cells. These dyes can either form FRET pairs by themselves (e.g., Cy3 and Cy5) or can be combined with GFPs in a FRET pair (e.g., CFP and Cy3 , or potentially GFP and Cy5).
NANOBIOSCOPY OF PROTEIN- PROTEIN INTERACTIONS WITH FLUORESCENCE RESONANCE ENERGY TRANSFER
Qualitative Analysis
Methods of Fluorescence Resonance Energy Transfer Measurement
Because in many cases it is difficult either to estimate the level of cellular expression to determine the donor-acceptor ratio or to correct for non-transfer energy loss pathways! in a system as complicated as a living cell, many researchers settle for qualitative data. Additionally, lack of precise filters, suboptimal laser excitation wavelengths, and the limited speed of data readout are common sources of errors in FRET analysis. The spectroscopic properties required for successful FRET measurements in live cells (both for GFP-like chromophores and for organic dyes) include: • A suitable spectral integral overlap between donor emission and acceptor excitation. • High donor molar extinction coefficient. • High fluorescence quantum yield of the donor. • Comparable rates for donor and acceptor photobleaching. • Comparable maturation times and intensity of chromophores during FRET records. • Known ratio of donor:acceptor molecules. To avoid common mistakes when choosing either a single probe or a FRET donor/acceptor pair, the most important first step is a literature search. For example, the strong pH dependence of YFP makes it unsuitable for fusions on the luminal side of transmembrane proteins of acidic membranous organelles, such as Iysosomes. However, when YFP is attached to such proteins on the cytosolic side it will work well and give bright fluorescence (Elsliger et aI., 1999; Griesbeck et ai., 200 I). For each chromophore, it is useful to know whether the fluorescent signal obtained derives from the intrinsic concentration of the chromophore, or if it also reflects properties of the microenvironment (e.g., pH dependence).
Preparation Preparations for live cell experiments thus should include: 1. Selection of appropriate fluorescent probes and cell lines. 2. Correct design of experiments (strategy) to answer the question. 3. Proper labeling or correct cloning of chimeric fluorescent proteins. 4. Finding a suitable method for introducing fluorescent molecules into cells: transfectionlelectroporation/microinjection. 5. Proper microscope setup for FRET analyses [lasers, filters, charge-coupled devices (CCDs), photomultiplier tubes (PMTs)l. 6. Best conditions for data acquisition from live cells (ensure good cell condition during the experiment, e.g. , temperature, CO 2 , isotonicity, illumination levels). 7. Appropriate image analysis (e.g., using programs such as MetaMorph 4-6, Imaris, NIH ImageJ, Photoshop, or custommade software programs) based on controls in which only the donor or only the acceptor is expressed.
J
For example. a normal cellular protein may act as a non-fluorescent acceptor, partially depleting energy of the donor.
795
FRET is never observed directly but can only be monitored immediately after the actual act of energy transfer from donor to acceptor. Therefore, all methods discussed here determine FRET indirectly. Changes in the degree of donor-acceptor interactions result in changes in FRET efficiency which is measured by comparing two states: (1) donor signal in the presence of acceptor FDA, compared to (2) donor signal without the acceptor FD (i .e., after bleach). Several experimental FRET setups are described below.
Sensitized Emission of Acceptor FRET by sensitized (i.e., increased) emission of the acceptor relies on measurements of acceptor emission resulting from excitation of a donor (Bastiaens and Jovin, 1998; Lakowicz, 1999; Periasamy et al., 200 I). Sensitized emission of the acceptor occurs when it is in close enough proximity to an excited donor that it can accept energy non-radiatively from the donor and then emit its own fluorescence without direct excitation. This method is simple but always requires a subsequent correction for cross-talk of the donor fluorescence into the acceptor channel , something that is especially difficult to measure for chromophores with closely related spectra. The concentration ratio of donor/acceptor molecules present in living cells will also influence FRET results. FRET data on molecular interactions are easier to obtain if the donor/acceptor ratio is in the region of I : I to I : IO (Gordon et aI., 1998; Herman et al., 2001; Hoppe et aI., 2002). The optimal D/A pair must be chosen for each particular experiment. Donor-acceptor pairs that are far away in wavelength allow one to avoid significant spectral overlap but usually have poor FRET efficiencies. Alternatively, although D/A pairs that share a large spectral overlap are often difficult to separate and to use reliably in experiments, they almost always show high FRET efficiencies. The sensitized emission (FRET signal) can also be easily detected by spectrofluorimetry (Fig. 45.6). However, in this case, the FRET signal will be averaged over the entire cell population. Only microscopes equipped with the appropriate filters allow one to resolve both FRET signals together with spatial information at cellular or subcellular level. A FRET signal from a corresponding cellular image obtained with a microscope will thus uniquely provide an additional magnification of the process, surpassing the optical resolution needed to visualize the Golgi , endoplasmic reticulum (ER), lysosomes, mitochondria, and other cellular subcompartments. Often, FRET efficiency (Ecft) can be directly visualized under the microscope throughout the entire cell and even resolved at the level of subcellular compartments, for example, the differential increase in FRET at the GolgiJcytosol interface shown in Figure 45.8. Currently, a typical living-cell FRET experiment involves usually capturing the emitted fluorescence from both the donor CFP and the acceptor YFP after excitation of only the donor (CFP) . The increase in sensitized emission corresponds to the degree of physical association between the two fusion proteins. In addition, spectral measurement or ratiometric imaging of the emission profile can be performed. This should include recording (a) emission of a sample expressing donor alone as a control, and (b) recording the interacting FRET pair. Comparison between the
796
Chapter 45 • I. Majoul et al.
emission spectra of these two samples allows one to assess the degree of FRET (see Fig. 45.6). Sensitized emission of the acceptor is a quick method for monitoring FRET in living cells, and perhaps the most promising approach for analyzing protein-protein interactions that have fast kinetics. It is also useful for recording neuronal activities and for studying the effects of drugs on specific cellular targets. These experiments require that the detection of the acceptor emission is not significantly contaminated with a non-FRET contribution from the donor. The only way to observe FRET in a dynamic situation is to have filters that effectively eliminate cross-talk between the emitted light from the two fluorophores. The underlying assumption is that the ratiometric analysis of donor and acceptor signals will allow one to monitor and to calibrate the kinetics of the biological responses. Precalibration of the expected effect in the correct time frame and the correct range of fluorescence intensities, laser sources, and filters is required to create an intensitydependent lookup table, which can then be used during an experiment. Using sensitized emission, we were able to monitor the timedependent effect of a bacterial toxin, cholera toxin, on the oligomerization of a transmembrane Golgi receptor over a period of 2h [Fig. 45.7(A)].
Donor Fluorescence Acquisition of a clean donor image, that is, non-contaminated with acceptor fluorescence, is the most critical step in this type of FRET analyses. This is especially true for CFP because the CFPNFP emission spectra are closely overlapping and CFP chimeras are usually less bright than YFP chimeras. The main requirements for acquiring a correct donor image are proper donor excitation wavelength, laser power adjusted to be appropriate for living-cell experiments, and a correct detector gain in the acquisition channel. The microscope must be precalibrated to ensure that the detector gains are linear within the working range of the experiment, and, therefore, that each channel is tuned separately for a maximal signal. For each set of experiments, we used a donor-only sample (CFP) to set the gain in the donor channel and to determine the minimum
FIGURE 45.7. (A) FRET monitored by sensitized emission of acceptor revealed the time-dependent state of oligomerization at the single cell level. Vero cells were double transfected with CFP and YFP versions of ERD2. During its intracellular transport, cholera toxin binds to the KDEL receptor and induces receptor oligomerization, revealed here by FRET as an increase in sensitized emission of acceptor (large arrow). (For details, see Majoul et aI., 2001.) (B) The distribution of a CFP signal in II nm lambda windows using the LSM-SI 0 META, single transfection with a CFP chimera. (C) The distribution of a YFP signal in 11 nm lambda windows using the LSM-SIO META, single transfection with a YFP chimera. Note that the upper four spectral windows contain the CFP signal non-contaminated by YFP, whereas a YFP signal is best detected in the middle windows. The spectral windows are very important for monitoring the FRET-induced increase in the CFP signal after acceptor photobleaching. (D) Spectral characterization of FRET induced between CFP and YFP chimeras; FRET proximity leads to the disappearance of the donor emission peak (I"em = 474nm) and appearance of the sensitized emission of acceptor O"em = S26nm) (gray curve). (E) Time-dependent FRET induced by cholera toxin transport between cytosolic ARFGAP-CFP and transmembrane ERD2-YFP revealed by spectrofluorimetry. Disappearance of CFP (474nm) peak and the appearance of YFP (S26nm) sensitized emission peak can be followed during these records. Red curve shows typical CFP emission at 474nm at the beginning of toxin transport, indicating no or low FRET, whereas a later time point (black curve) shows strong FRET. For details, see Majoul et al. (2001).
YFP: only acceptor
CFP: only donor
E
498 ~~~--~~----------~ 450
450
500
550
500
550
wavelength, nm
600
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
laser power necessary to acquire a donor signal at maximum detector gain. The same process was repeated for the acceptor channel using an acceptor-only (YFP) sample. To acquire a FRET image the same laser power and wavelength was used as for CFP alone. If FRET occurs, it should result in a decrease in donor fluorescence, and this can be measured by different methods, including fluorescence lifetime. Usually it is done by acquiring sequential sets of donor images and comparing donor intensities, I D, with IDA (donor in the presence of acceptor) after different (t + 't) and after different amounts of donor photobleaching. The advantage of this approach is the ability to follow FRET-induced changes in the donor fluorescence. The disadvantage is the absence of proper controls for acceptor and other non-FRET ways of donor deactivation.
797
Olympus FV500
Acceptor Bleach The best way to prove that an observed FRET signal reflects a true interaction between two molecules is to remove (photo-inactivate) the acceptor (Bastiaens et ai., 1996; Bastiaens and lovin, 1998). In the case of true FRET, an increase in donor fluorescence will result; from the portion of the donor fluorescence that was recovered after accepter inactivation. The ratio of the donor fluorescence intensities before and after acceptor photobleaching (Dbb/Dab) reports the FRET efficiency. FRET efficiency (Eeff) is calculated as the ratio of two intensities generated in the same detection channel but from the same sample before and after acceptor bleach. Depending on how much FRET occurred in a particular region of interest, the ratio of the two intensities measured on the same cell in two different regions of interest (e.g., a bleached and an unbleached region of the same cell treated with bacterial toxin) often varies (Bastiaens et ai., 1996). The application of the acceptor bleach method is often delicate in living biological samples because of the long bleaching times involved and the associated phototoxicity. We used a special builtin laser emitting at 532 nm for acceptor photo-inactivation, and calculated FRET from the percentage of acceptor bleach instead of performing a total inactivation of the acceptor. The degree of acceptor bleach can be calculated from the ratio of acceptor after and before bleach (Aab/Abb), images. In this type of experiment, one needs to ensure also that the process of acceptor photobleaching does not affect the donor intensity, otherwise the experimentally measured FRET value will be lowered. An example of a strong increase in donor fluorescence after acceptor bleach (Dab), detectable even by eye, at the Golgi/cytosol interface is shown in Figure 45.8. This figure represents data from two different setups used for measuring FRET: a Leica-based, twophoton multi-focal, multi-photon microscopy (MMM) FRET setup (described in detail in Majoul et aI., 2001, 2002b) and a singlephoton Olympus FV500 setup. Here we compare FRET results measured on the same type of biological samples consisting of interacting CFP and YFP fusions (Fig. 45.8). In both cases, the fluorescent donor was a CFP chimera of the cellular protein ARFGAP. This normally predominantly cytosolic protein becomes associated with membranes of the Golgi complex in a biologically meaningful way upon its interaction with the acceptor, the YFP-tagged transmembrane Golgi KDEL-receptor (Majoul et ai., 2001). We analyzed fluorescence of the donor ARFGAP-CFP before (Dbb) and after acceptor photo-inactivation (Dab) (Fig. 45.8). The FRET-MMM setup included an lOOx oil objective and a cooled back-illuminated CCD camera. For the Olympus setup, a PL APO 60x oil NA 1.4 objective lens was used at Aex = 458 nm
FRET-MMM (2-photon Leica)
FIGURE 45.8. An example of FRET induced by cholera toxin transport between cytosolic ARFGAP-CFP and transmembrane ERD2-YFP recorded under the microscope. It is obvious that a cytosolic donor can only efficiently interact with the transmembrane Goigi acceptor at the Goigi/cytosol interface. (Upper panel) The Olympus FV 500 setup allowed us to resolve the increase in CFP (donor) fluorescence at the Goigi/cytosol interface even after partial photoinactivation of acceptor (compare Dbb and Dab). (Lower panel) The same combination of cytosolic donor ARFGAP-CFP and the transmembrane acceptor ERD2-YFP was used for FRET-MMM analyses. Although we were able to see an increase in donor fluorescence after acceptor bleach. the fine resolution of FRET at Goigi/cytosol interface was not achieved in this experimental setup. (For more details see text.)
798
Chapter 45 • I. Majoul et a!.
(argon laser, 10% power) and the PMT voltages set to 729Y for the CFP channel and 653 Y for the YFP channel. Acceptor emission was detected within a narrow 15 nm wavelength window between 480 and 495 nm. The YFP acceptor was photobleached at Aex = 532 nm and emission was sensed between 535 and 565 nm. The FRET-MMM two-photon excitation wavelength for the donor (CFP) was Aex = 820nm. Acceptor bleaching was done using a Nd:YAG laser at 532nm and CFP donor fluorescence was detected with the a BP485117 nm emission filter (Zeiss) in front of a backilluminated CCO camera. Up to >90% of acceptor (YFP) was photobleached within 72 s. After the irradiation of the acceptor was completed, the de-quenched donor fluorescence was analyzed (Dab) and compared to the initial state of the donor CObb) to estimate the degree of donor-acceptor interactions in FRET. Using FRET-MMM, we were able to see the increase in donor fluorescence (Fig. 45.8, lower panel, big arrows) but were not able to distinguish the Golgi/cytosol interface. We expected to see more FRET at the Golgi membrane/cytosol interface because the cytosolic donor (ARFGAP-CFP) can interact with the transmembrane Golgi acceptor (ER02-YFP) only in this location, that is, the multiple cytosolic donor molecules accumulating on the Golgi membrane interact with the transmembrane acceptor. Remarkably, using the Olympus setup we could discriminate between outer (looking into the cytosol) and inner surfaces of the Golgi complex (Fig. 45.8, upper panel). The Olympus setup revealed a much stronger de-quenching of the donor at the Golgilcytosol interface after the acceptor photoinactivation (Fig. 45.8, left panel, see small arrows).
FLUORESCENT PROTEINS AS FLUORESCENCE RESONANCE ENERGY TRANSFER PAIRS Cyan Fluorescent Protein and Yellow Fluorescent Protein - The Commonly Used Fluorescence Resonance Energy Transfer Pair As native CFP and YFP molecules normally do not interact, ideally only a biologically meaningful interaction between the non-GFP part of chimeras will bring these chromophores into FRET proximity. However, in over-expressing cells, unspecific stochastic interactions induced by a very high local density of CFP-YFP chimeras can also generate an artificial FRET signal. Such cells should not be used. Normally the interaction time of an unspecific pair of CFP and YFP chimeras should be significantly shorter than that displayed by a pair of CFP and YFP chimeras undergoing a biologically meaningful interaction.2 To measure sensitized emission complementary to the FRET microscope data, we used Fluoromax-2 and Fluoromax-3 spectrofluorimeters (ISA, Jobin-Yvon Instruments, Edison, NJ) as a control in the same transfected cell population in bulk cell experiments. CFP was excited at Aex = 425 nm, and fluorescence emission was detected between 450 to 600 nm. For comparison, YFP was excited separately at A.x = 498 nm, and its fluorescence was measured between 510 and 600 nm. For a time-dependent analysis of the interactions between CFP and YFP fusion proteins, the culture plate was divided into segments and the cells from different segments were examined by spectrofluorimetry before and after toxin application (Majoul et at., 2001, 2002b).
2 3
Except when these fusions may co-aggregate in over-expressing cells. The 458 om laser line can be used for CFP excitation but is not optimal. A line at - 420 to 430nm would be better, but these are relatively rare.
The ability to detect meaningful signals clearly depends on the signal/noise ratio. For example, although trying to detect GFP fluorescence in the presence of the intrinsically strong autofluorescent background signal found in transfected primary hepatocytes often results in a signal/noise ratio of -1, the same GFP expressed in primary fibroblast-like cell cultures will produce a signal/noise ratio of -10 : 1. Compared to fibroblasts, hepatocyte cultures will always present much bigger problems with photobleaching and phototoxicity. An understanding of cell physiology and the photophysics of the dye used in the microenvironment of each cell type is essential for evaluating FRET data obtained with the fluorescence microscope.
Cyan Fluorescent Protein or Green Fluorescent Protein Forms a Fluorescence Resonance Energy Transfer Pair with mRFP1 The development of monomeric forms of OsRed (Campbell et at. , 2002; Shaner et at., 2004) provided a new partner for FRET analyses. The current leading FRET pair CFPIYFP usually requires donor excitation at a wavelength (A.x peak = 430 nm) not found on many confocals. Although confocal microscopes equipped with argon-ion laser lines at 458 nm, 488 nm, and 514 nm, a green helium:neon (He-Ne) laser Aex = 543nm or a red He-Ne lasers A.x = 633 nm can be used / those employing dual-gas krypton/argonion lasers with lines at Aex = 488 nm, 568 nm, and 657 nm are not suitable for exciting either CFP or YFP (excitation peak Aex = 514nm). Although CFP can be efficient excited by a krypton-ion laser at Aex = 413 nm or violet laser diodes at Aex = 405 nm, both of these are close to the ultraviolet (UY) and potentially cytotoxic, making them unsuitable for long, time-dependent experiments. As the 488 nm line is common on most confocals and is optimal for exciting GFP, the introduction of mRFP provides us with a new FRET pair that will be highly appreciated by many cell biologists. Whereas the large spectral overlap between CFP and YFP emission spectra necessitates the use of expensive, narrow bandpass filters and sensitive CCOs, mRFP, and other members of a growing family of red-shifted sea coral fluorophores, emits in the red. This favors their use as FRET acceptors with either GFP or CFP (Galperin et at., 2004). Integral spectral overlaps (lOA) between CFP/mRFP I and GFP/mRFPl are critical for FRET measurements (see also Fig. 45.6). The Ro values typical of GFP and DsRed are -5 nm (Patterson et at. , 2000). Emission maxima of the donor (CFP or GFP) and the acceptor (mRFP1) measured on a Fluoromax-3 are shown in Figure 45.9, together with integral spectral overlaps. One remaining problem is that mRFP matures slowly in the cytoplasm; mCherry seems to be an improvement in this respect (Shaner et at., 2004). This is a disadvantage for both comparative and quantitative FRET measurements with GFP or CFP, but it may be partially overcome by using a FRET detection algorithm, recently developed to compensate for the slow acceptor maturation in CFPIDsRed or GFPIDsRed pairs (Erickson et aI., 2003). Combinations of laser excitation and specific filter sets suitable for different FRET pairs of chimeric fluorescent proteins as well as fluorescent dyes are depicted in Table 45 .2.
Fluorescence Resonance Energy Transfer-Based Sensors Many biological processes are successfully analyzed with fluorescent biosensors where FRET between CFP and YFP reflects a wide variety of events such as phosphorylation of specific sequences, activity of cellular kinases or small GTPases, oscillations of intra-
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
cellular Ca++, pH, or binding of specific molecules that increase or abolish FRET (Miyawaki et al., 1997; Miesenbock et aI., 1998; Kraynov et al., 2000; Ting et ai., 2001; Del Pozo et al., 2002; Wouters et ai., 2001; Bunt and Wouters, 2004). Quite often, a fluorescent biosensor is a fusion protein comprising three main components, a functional or targeting domain fused to the C- or Nterminus of enhanced cyan and yellow fluorescent proteins. FRET measurable with the biosensor will change in a predictable way in response to changes in the intracellular environment [pH, Ca++, cyclic adenosine monophosphate (cAMP) levels, redox potential, adenosine triphosphate (ATP) levels, glucose levels, etc.] or the binding of a protein or peptide to the biosensor (reviewed in Hahn, 2003; Meyer and Teruel, 2003; Bunt and Wouters, 2004). GFPbased biosensors have been successfully applied to analyze compartmentalization of many cellular dynamic processes, notably the spatio-temporal analysis of cellular transport, signaling, and development. Phosphoinositide signaling can be studied using GFP-tagged pleckstrin homology (PH) domain constructs. Different reporters for serine, threonine, and tyrosine protein kinase activities have been reported. Such GFP-based phosphorylation substrates are designed for precise subcellular targeting and to not impair phos-
aoo
.00
700
800
799
TABLE 45.2. Lasers and Filter Configurations for Selected Fluorophores Used CFP YFP Alexa488 or GFP Cy3 or Rhod-2 Green DsRED/mRFPl Cy5
Laser, Ah
AEm Filters (nm)
Argon, 457 Argon, 514 Argon, 488 nm He-Ne, 543 nm He-Ne, 543 He-Ne 633 or He-Ne 594
485/30 or 485117 (Zeiss) 530/50 515/30 or 535/50 590170 590170 660LP 660LP
FRET: Acceptor (YFP) bleach: external laser of 530 nm Filters for the Widefield FRET: Donor/Acceptor Ae,IBP(nm); AEm(nm); Dichroic (nm) Donor/Acceptor BFP/GFP
BFPYFP BFP/mRFPl CFP/mRFPl CFP/YFP GFPIRhod-2 Green FITClRhod-2 Green FITC/Cy3 Cy3/Cy5 Alexa488/Cy3
365115 460/50 365/15 460/50 365/15 460/50 436/20 485117 436/20485/17 488/20 535/50 488120 535/45 488120 535/45 525/45 595/60 488/20 535/45
535/50 535/26 610/60 610/60 535/20 595/60 595/60 595/60 695/55 595/60
390 390 390 455 455 505 505 505 560 505
A 1 photoD, DnI ----_.... CFP/mRFP1 configuration for FRET:
.•....
. ..
650
GFP/mRFP1 configuration for FRET: 400000
••,
•, •, ••, 1
•.
. ..
••
OL---------------~
250
300
350
400
450
550
. .....
650
FIGURE 45.9. Monomeric red fluorescent protein (mRFP) is potentially a good FRET acceptor for both CFP and GFP chromophores. Excitation and emission spectra of CFP, GFP, and mRFP together with their integral spectral overlaps (lDA) for CFP/DsRed and GFPlDsRed are given for comparison. Different filter sets used to separate CFP, GFP, and DsRed signals are indicated.
phory lation of endogenous substrates. As a result, they allow spatiotemporal resolution of phosphorylation processes in live cells. For example, imaging of FRET-based reporters for protein-kinase C (PKC) translocation (with CFP and YFP fused to the N- and C-terminus of PKC), phosphoinositide bisphosphate conversion to IP3, and diacylglycerol has shown that in HeLa cells PKCmediated oscillatory phosphorylations correlate with Ca2+_ controlled translocation of conventional PKC to the membrane without oscillations of phospholypase-C (PLC) activity or diacylglycerol (Violin et ai., 2003). Time-resolved visualization of growth factor-induced activation of Ras and Rapl in living cells is yet another example of dissecting cellular signaling events using microscopic techniques and FRET sensors (Mochizuki et al., 2001). A fluorescent biosensor that is useful for temporal measurements of the mitotic clock has been recently described (Jones et al., 2004). High throughput methods of FRET screening using biosensors will likely be useful in pharmaceutical and clinical screens for modulators of tyrosine kinases and phosphatases, and many other cellular activities. The list of useful biosensors is growing by the day. Previously based mostly on biochemical data, real-time observations with FRET in living cells are now becoming major sources for our knowledge of how cells move, divide, and are activated for coordinated cellular actions in signaling and intracellular transport.
FLUORESCENCE RESONANCE ENERGY TRANSFER AND OTHER COMPLEMENTARY METHODS
Fluorescence Resonance Energy Transfer and Fluorescence Lifetime Imaging Microscope FUM is a neat tool for detecting multiple fluorophores in living cells, particularly spectrally overlapping molecules such as GFP
800
Chapter 45 • I. Majoul et al.
variants, which have different fluorescence lifetimes despite their overlapping spectra (Bastiaens and Squire, 1999). The detection of fluorescence is typically achieved by counting the number of photons emitted by the excited state of a fluorophore. Alternatively, the lifetime of the excited state can be measured by FUM (Lakowicz, 1999; Chapter 27, this volume). FUM is an excellent tool to measure FRET because the lifetime of the excited state decreases strongly when FRET provides an additional means for decay from the excited state. FUM can also be used to monitor FRET-based protein-protein interactions. The lifetime of the donor is the time spent in the excited state ('t) before returning to the ground state (Lakowicz, 1999). Analyses of donor-acceptor interactions thus relies on measurements of the length of time (typically less than 4ns for GFP molecules) that the donor remains fluorescent after being excited with a fast pulsed laser. FRET can be calculated in such an experiment according to the following Eq. (Lakowicz, 1999): (14)
Because the change in the fluorescence lifetime of the donor can be analyzed independently of acceptor emission, this technique
can be used to detect FRET between GFPs with nearly identical emission spectra but different 't (Bastiaens and Squire, 1999). FLIM measurements are also of advantage for those donor molecules that are prone to quick photobleaching. Changes in 't can easily be detected by advanced modem technology using the frequency or time domain, or direct photon counting (see Chapter 27, this volume). A FUM setup, suitable for measuring FRET, as well as some results from living cells are shown in Figure 45.10. Comparison between the life time of the donor alone ('to) and the donor life time in the presence of the acceptor ('to A ) will provide information on R. The biggest advantage of FLIM analyses is the independence of donor lifetime 'to on the concentration of the dye (except in some cases of homo-FRET). Nevertheless, the complex calibration procedure for each particular FUM setup will require professional support. The FUM approach has a number of limitations: • Each microscope must be calibrated. • Spatial resolution may be limited if multiple images must be acquired. • More photons must be detected to determine 't for a pixel than are required to merely detect presence or absence .
. Frequency domain
1.5 intensity Modulation
1.8
D
= ~~
°0~----1~0~0----~2~0~0----~3~00~--~4~00~--~50~0~ FRET
degrees Time domain
total
FRET: donor channel
counts 0
==-___
(3 700
FIGURE 45.10. FRET resolution by FUM. Upper panel, left: Schematic presentation of FRET in the frequency domain: the pulsed excitation frequency (dark blue); the delayed signal of the CFP-dono r (red curve); and the less-delayed signal of the donor when FRET is occurring (light blue) because FRET decreases the do nor lifetime. Right: The intensity of the signal having the CFP, (donor) time constant collected under the same conditions, and showing that FRET causes the donor (CFP) intensity to decrease when both CFP- and YFP-tagged connexin molecules are oligomerized in a perinuclear Golgi region of this cell. Middle right: Alternative representation of the data in image above. Histograms represent the shorter donor lifetime when FRET to an acceptor is occurring (T FRET ) , compared to when only the donor is present (To). Lower panel, left: Schematic presentation of the individual parameters in the time domain: Log (intensities) versus lifetime: (TD ) is longer than (TFR ET) ' Left: Images of FRET in the Golgi region of a cell: comparison between total counts and counts sampled only from the donor channel reveals the portion of the donor molecules that are not transferring energy to the acceptor, and thus still emitting in the donor channel.
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
Despite the existence of commercial hardware and software programs for the task, FLIM data will generally require more complex (global) mathematical analysis of acquired results (Bastiaens and Jovin, 1998; Bastiaens and Squire, 1999; Bastiaens and Pepperkok, 2000; Gerritsen et al., 2001 ; Herman et al., 2001; and the overview of FLIM given in Chapter 27, this volume).
Fluorescence Recovery After Photobleaching and Fluorescence loss in Photobleaching Fluorescence recovery after photobleaching (FRAP) is a technique in which a region of interest is selectively photobleached with a high-intensity laser and the recovery that occurs as molecules move back into the bleached region is monitored over time with lower-intensity laser light. Depending on the protein studied, fluorescence recovery can result from protein diffusion, binding/dissociation, or transport processes. FRAP experiments can thus determine the kinetic parameters of a protein, including its diffusion constant, mobile fraction , transport rate, or binding/dissociation rate from other proteins in living cells (Lippincott-Schwartz et aI., 2001, 2003). Fluorescence recovery is usually quantitatively monitored by a microscope-mounted CCD camera that is calibrated to the fluorescence signal before photobleaching. Fluorescence loss in photobleaching (FLIP) is a photobleaching technique complementary to FRAP. In a FLIP experiment, a region of interest in a cell is repeatedly photobleached while fluorescence in the whole cell is continuously monitored. Any cellular regions with connections to the area being bleached will lose fluorescence due to lateral movement of mobile proteins into the bleached area. On the other hand, the fluorescence in unconnected cellular regions will be unaffected (Cole et at., 1996). FRAP and FLIP techniques have been extensively used for studying the trafficking within the secretory pathway in animal cells, especially ER-to-Golgi trafficking (Cole et aI., 1996; Lippincott-Schwartz et al., 2oo1; Presley et at., 2002; Lippincott-Schwartz and Patterson, 2003). In principle, FRET techniques can be nicely combined with both FRAP and FLIP to obtain supplementary information on protein-protein interactions together with kinetic parameters of protein localization, for example, about mobility of plasma membrane receptors and lipids .
Fluorescence Resonance Energy Transfer and Fluorescence Correlation Spectroscopy Confocal fluorescence correlation spectroscopy (FCS) is a unique tool to analyze processes that can be recorded from a very small area of illumination in a selected region of less than 1 ~m (Magde et al., 1972). FCS is thus suitable for quantitative measurements of the local concentrations and diffusional mobility of fluorophores through a small volume of a living cell, for example, in endocytic membrane-bound carriers or other compartments (Bacia et aI., 2oo2; Kim and Schwille, 2003). Diffusion coefficients measured by FCS can be used to calculate the approximate sizes of protein complexes in living cells, and based on quantification of the fluorescence intensity of the diffusing complex the number of molecules can be estimated, as has, for example, been done for the Gag complex during retrovirus assembly (Larson et aI., 2004). FRET and FCS can be considered as complementary techniques, where FRET is used to estimate the distance between the donor and acceptor molecules, and FCS to provide information on dynamic properties and the sizes of the protein complexes analyzed.
801
Fluorescence Resonance Energy Transfer and Total Internal Reflection Fluorescence In total internal reflection fluorescence (TIRF) microscopy, excitation light approaches an aqueous specimen as a coherent beam of light that strikes the coverslip/water interface at an angle greater than the critical angle. As a result, only an evanescent excitation field penetrates into the water to excite dye molecules, such as transfected GFP chimeras, that are within a few tens of nanometers of the interface. The thickness of the excitation band depends on the wavelength and the approach angle. Images are recorded in widefield using intensified- or electron muliplier (EM)-CCDs. As this technique excites and visualizes only fluorophores in close proximity to the plasma membrane, it has the obvious advantage of strongly reducing photobleaching and phototoxicity compared to other microscopy techniques. It also reduces background light from out-of-focus planes essentially to zero. Therefore, the parts of living cells near the coverslip can be imaged over longer time periods and with high contrast and improved temporal resolution. TIRF is the method of choice for those who want to resolve processes occurring close to the plasma membrane. Such processes include endocytosis and exocytosis, cell-cell signaling, formation of cell-cell contacts, and release of neuromediators. If the biological process under study allows photobleached GFP chimeras to be quickly exchanged from the non-bleached pool of chimeras in deeper regions of the cell, this results in a very stable fluorescent signal for long-term observations. TIRF can be combined with conventional widefield fluorescence or DIC microscopy for localized analyses of cellular processes and/or FRET measurements (Bezzi et aI. , 2004; Jones et aI., 2004).
Quantum Dots and Fluorescence Resonance Energy Transfer Semiconductor quantum dots were introduced recently into microscopy for applications in cell biology (MichaIet et al. , 2005). Although they seem to be very promising (Lidke et at., 2004), their application for FRET will be limited. Quantum dots exhibit broad absorption profiles: as both quantum dots at 565 nm and 655 nm can be excited with a single excitation filter 435170, long-pass emission filters must be used to separate the signals. These parameters will prevent simultaneous use of quantum dots with CFP- or GFP-like live chromophores that could otherwise be complementary donors for live-cell FRET analyses. Nevertheless, doublelabeling with quantum dot conjugates is a powerful tool in immunohistochemistry and immunocytochemistry applications (Wu,2003).
CLONING AND EXPRESSION OF FLUORESCENT CONSTRUCTS FOR FLUORESCENCE RESONANCE ENERGY TRANSFER Cloning of Fluorescent Chimeras After deciding to use a FRET approach to study a protein of interest, fluorescent donor-acceptor pairs need to be generated. The most common approach is to obtain the coding regions for the chosen proteins and use commercially available vectors with multiple cloning sites (e.g., BD Biosciences Clontech, Stratagene, Qbiogene, and other companies) to generate in-frame fusions with the CFP, YFP, GFP, or mRFPl genes. GFP fusions can be gener-
802
Chapter 45 • I. Majoul et al.
ated at the N- or C-tenninus of proteins of interest, or even internally, with GFP being interspaced between domains of a protein coding sequence. It is generally a good idea to simultaneously prepare both N- and C-tenninal fusions, hoping that at least one of these chimeras retains functional activity. One fairly common problem is that the linker between the protein of interest and GFP may be recognized as a nuclear import signal (NLS) by the cellular machinery and the chimera may thus unexpectedly be detected in the nucleus instead of its proper cellular location. This chance is particularly high if the linker contains NLS-like motifs such as repeats of lysines (K) or arginines (R). Such common biological mistakes in linker generation should be avoided. Combinations of other amino acids, especially secondary structure breakers such as glycine (G) or proline (P) or small amino acids such as serine (S), alanine (A), or threonine (T) are less likely to create problems. However, there are no dead-sure recipes for generating functional GFP chimeras, beyond following standard protocols for the cloning itself (Sambrook et al., 1989). Therefore, performing functional tests is critical. After DNA sequencing to verify the constructs, DNA for mammalian cell transfections needs to be prepared with great care. In general, commercial kits (e.g., Qiagen Maxiprep) work well. For some cell lines (e.g., our favorite cell line, Vero cells) removal of endotoxins (bacteriallipopolysaccharides) is important and easily accomplished with commercial kits. Many transfection reagents routinely used to introduce plasmid DNA into living cells are lipophilic agents that penetrate biological membranes but may also induce unwanted side effects, such as aggregation or fusion of intracellular membranes (see below). This makes cells transfected with these reagents often look clumpy and non-beautiful, and the appearance of intracellular membranous organelles may also be abnormal. As a word of caution, any strong overproduction of a transfected protein (10- to lOO-fold above nonnal cellular levels) may impair cell function, an effect that needs to be controlled by appropriate functional assays. (For useful www links see Table 45.4).
Functional Activity of Expressed Constructs Successfully cloned chimeras will be introduced into the cells selected for initial test experiments. First, the expression levels and protein stability of the donor and acceptor needs to be tested in a time course, and the spectral properties of the chimera established. Often, a newly generated chimera may be brightly fluorescent but has lost functional activity, displays the wrong localization, and/or has lost the ability to interact with its normal partners in a protein complex. Therefore, it is most important to check the functional activity of chimeras before attempting to draw any conclusions from their use. Such checks require extreme delicacy because one can only attribute aberrant behavior to the construct if one is sure that the behavior is not being caused by your method of observation. If an antibody against the protein under study is available, the GFP chimera should display the same or a very similar intracellular distribution as the endogenous protein. If normal interactions between the chimera with known cellular partner proteins can be established by co-immunoprecipitation, this is a good sign!
Expression and Over-Expression Proper expression levels of chimeras are crucial. In the case of many of our CFP/YFP chimeras that have functions in the secretory pathway, we use the intracellular transport of cholera toxin (CTX) as a good functional test. We learned that in cells overexpressing fluorescent chimeras, transport of the toxin from the
TABLE 45.3. Transfection Buffer for Electroporation, Internal Medium 1M. Transfection Buffer 120mM final MW: 74.56g/mol Stock solution 1.2M ~ for 50 mL 4.48 g
KH 2PO. 10 mM final MW: 136.09 g/mol Stock solution 100 mM ~ for 50 mL 680.45 mg EGTA 2 mM final MW: 380.4 g/mol Stock solution 20mM ~ for 50mL 380mg
MgCI2 5 mM final MW: 203.31 g/mol Stock solution 50 mM ~ for 50 mL 508.3 mg HEPES 25 mM final MW: 238.3 g/mol Stock solution 250mM ~ for 50mL 2.98 g
CaCI 2 0.5mM final MW: 147.2g/mol Stock solution 15mM ~ for 50mL 1l0mg To prepare 50mL of cytomix: Mix 5 mL of each stock solution, except for CaCh, only 500 ilL Add H 20 until 50mL, adjust pH = 7.5-7.6 For complete cytomix: Add GSSG 5 mM final MW: 656.6 g/mol Stock solution 100mM ~ for lOmL 660mg Add 50 IlLlmL ATP 2mM final MW: 550g/mol Stock solution 100 mM ~ for ]0 mL 550 rng Add 20 IlLlmL Modified from Majoul et at. (2001). ]0'_108 cells washed in PBS are transferred into 400/lL of total Internal Medium Mix for electroporation in Bio-Rad (green cuvette) Cat. #165-20086. Add 20-40/lL of endo-free plasmid DNA (from stock of l/lg//lL). Bio-Rad electroporator: 0.7 kV, main position SO/capacitance extension 200/lF (see on the panel of electroporator). Resulting time constant for this combination should be 1.44-1.64 to get 50%-80% of finally transfected cells.
plasma membrane to the ER, or even to Golgi compartments, was blocked. Extending the time after transfection (in our case using electroporation) to 16 to 20h for the transmembrane protein ERD2-YFP, which is the receptor for cholera toxin in the Golgi complex, led to severe disturbances not only in the intracellular localization of the chimeras but also resulted in impaired toxin transport. Although cholera toxin was able to recognize the CFP or YFP-fused ERD2 protein 6 to 8h after transfection [Fig. 45.4(A)], after 16h of expression, it was no longer able to enter the structures labeled by ERD2 fusions [compare Figs. 45.4(A,B)]. However, even in these damaged cells, CTX still bound to the areas of Golgi containing functional endogenous receptor and exhibited only low labeling with the chimeric receptor [Fig. 45.4(B), long arrows]. Conventional fluorescence microscopy can be used to establish the time course of expression and to select the time when the expression is sufficient to measure FRET, when the chimeras are still localized in the same areas as in their endogenous wild-type counterparts, where their functional activity can be established. In our hands, using CMV promoter constructs, this was usually -6 to 12h after electroporation, depending on the construct. In general, the physiological validity of FRET measurements in living cells expressing a given pair of CFPIYFP fusion proteins is strengthened if the observed FRET signals are obtained in response to a known, meaningful physiological trigger, especially when compared with other pairs of CFPIYFP fusion proteins that remain unchanged.
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
803
TABLE 45.4 Links Useful for the Design of FRET Experiments Proteins Collection of useful links related to microscopy around the world: Molecular Expressions: The Microscopy Primer. A comprehensive Tutorial Site that includes fascinating "Virtual Microscopy" BIND: Protein-protein interaction database Fluorophores & other probes: Proteins MoBiTec: Tools for Cell and Molecular Biology Invitrogen: Cloning & expression BD Biosciences Clontech: GFP news Protein sequence and classification database Society for Molecular Imaging
http://bind.ca! http://www.probes.com/ http://www.mobitec.de/ http://www.invitrogen.com! http://www.bdbiosciences.com!clontechlgfp/ http://pir.georgetown.edu/ http://www.molecularimaging.org
Optics Fluorescent Spectra Database Chroma Technology: Optical filters for biological applications Omega: Optical filters for biological applications Semrock: New generation of filters AHF analysentechnik AG Optical filters including GFPs Sutter: Filter changer and light sources for microscopes Bioimage: Biological images for scientific research Semrock: FRET filters
http://www.mcb.arizona.eduIlPC/spectra_page.htm http://www.chroma.com! http://www.omegafilters.com! http://semrock.com http://www.ahf.de http://www.sutter.com! http://www.bioimage.org/pub/collaborators.jsp http://www.laser2000.co.uk/semrock/brightline.htm
METHODS FOR INTRODUCING CHROMOPHORES INTO LIVING CEllS Electroporation We have described an efficient electroporation technique (Majoul et aI., 20Gl) that has been routinely used during the Vancouver 3D microscopy courses (2002-2005) for multiple transfections with different plasmids. With slight modifications (see Table 45.3), this protocol is now successfully used in many laboratories for transfection of a large variety of cell lines, including primary cultured neurons, to study GFP-highlighted organelles in living cells and for FRET analyses. Instead of using PBS (phosphate-buffered saline), which is not very physiological, we use a high potassium "internal media" buffer condition for the electroporation step (K-glutamate; up to 140mM, pH 7.6). Other essential components are fresh ATP and GSSG (see Table 45.3 for details). We prefer to use a Bio-Rad electroporator, green Gene Pulser cuvette (BioRad Cat. No. 165-2086; gap 0.2 cm) and use plasmid DNA in a total volume of up to 300 ilL "internal media" per transfection reaction. In Vero cells (a Green monkey kidney fibroblast cell line), we routinely obtain 90% or more of successfully transfected cells with this method. The advantage of electroporation for transfer of DNA for FRET analyses is the total absence of the unspecific fluorescence background produced by other transfection reagents.
Transfection Reagents For transfecting cells with foreign DNA, scientists commonly use different commercially available agents with lipid-like properties, able to form micellae around the DN A: for example, Lipofectamine, Lipofectin, Metafecten, FUGENE 6, etc. Transient transfections are usually performed on routine cell lines, such as HEK 297, COS-5, COS-7, NRK (normal rat kidney), 3T3 fibroblast-like cultures, BHK-2l (hamster kidney cell line), or the Vero cells used for the experiments described in this chapter. Cells prepared for trans-
fection are typically grown in media such as Dulbecco's modified Eagle medium (DMEM), supplemented with 10% fetal calf serum (FCS), glutamine, and antibiotics such as penicillin and streptomycin. Exact transfection protocols are often supplied by the man-
http://www.ou.edu/researchlelectron/www-vl/long.shtml http://micro.magnet.fsu.edu/primer/index.htmi
ufacturer of the transfection reagents. Many of them work well, although certain cell types, especially neurons, may require different transfection procedures. Ballistic methods to introduce neuronal tracers, plasmid DNA or intracellular calcium indicators into neurons have also been reported, are now in common use (Grutzendler et aI., 2003). Although usually the commercial reagents do not create problems for biochemical experiments, for microscopic, single-cell analyses, remnants of precipitated DNA or aggregates of transfection reagents may cause a strong fluorescent background, especially when two-photon excitation is used. Tetracyclin-regulated promoters cannot be used in combination with two-photon excitation of CFP, as even traces of tetracyclin create a strong fluorescent background in the cell. Expression systems induced by agents such as ecdysone or hygromycine may be more suitable in such cases. We routinely express our fusion proteins of interest under the control of the widely used, strong CMV promotor. Again, selection of the appropriate time window after transfection, with regard to obtaining good fluorescence signals for imaging without having vast over-expression, is crucial for best results.
Microin jection Microinjection is an elegant way to deliver biological material into living cells, for example, DNA, RNA, and proteins (e.g., fluorophore-labeled or unlabeled recombinant proteins or inhibitory antibodies). Microinjection was used as early as the 1960s (Kohen et al., 1966). One of the first attempts to measure intracellular Ca++ with aequorin also was based on microinjection techniques (Blinks et al., 1978). In the 1970s and 1980s rnicroinjection was applied for the transfer of dye between neighboring cells and used to study the permeability of gap junctions (Conn, 1991) (see Fig. 45 .11). With the development of antibodies against specific proteins or protein domains, microinjection of antibodies to inhibit the function of the protein under study became popular. Microinjection of antibodies against COPI vesicular transport machinery (anti-[3COP antibodies raised against the EAGE sequence of this subunit) prevented the transport of cholera toxin through the Golgi apparatus (Majoul et al., 1998). The construction of plasmids encoding specific proteins and the development of anti-sense RNA and RNAi provided another
804
Chapter 45 • I. Majoul et a!. FIGURE 45.11. (A, B) Two examples of microinjection: (A) a dying cell after being impaled by the needle, with collapsed cellular membranes, and (B) successful microinjection into the cytosol of a cell. (C) Dye transfer between astrocytes after microinjection of caIcein into a single cell (indicated by a rrow) ; (D) The same field viewed in transmission light. (E, F) A decrease in dye transfer from a microinjected cell upon lowering of the extracellular free Ca++ concentration. Living-cell nuclei stained with DAPI; (F) the same field viewed in transmission light.
area for rapid development of microinjection techniques. The discovery of GFP and its color variants has led to the use of microinjection for inducing expression of fluorescent chimeras in many different types of cells, including cultured neurons and even live organisms by injecting DNA into Xenopus oocytes, embryos, developing zebrafish, Drosophila, and other species. In microinjection experiments, Alexa-350 or Cascade blue conjugated to BSA can be co-injected to easily identify the injected cells. In experienced hands and with availability of good equipment, microinjection into cells and micromanipulation of cells and nuclei with glass pipettes, sometimes combined with electrophysiology, can address many complicated biological questions (for an example, see Wakayama et ai., 2003 ).
In short, microinjection can be suitable for: 1. Introduction of plasmid DNA: Mostly for cell imaging, while for a macroscale biochemical analysis electroporation is more advisable. 2. Introduction of RNA: si-RNA, labeled oligonucleotides, antisense RNA.
3. Introduction of proteins: Dominant-negative functionally active molecules (e.g., dynamin) or microinjection of Fab fragments of antibodies because whole divalent antibody molecules may produce unspecific effects by cross-linking the proteins of interest. 4. Introduction of dyes and other fragments: Fluorescent markers (e.g., labeled phalloidin to visualize actin), membrane-permeable dyes (e.g., calcein) to study cell-cell contacts, or impermeable agents (e.g., GTP-y-S).
FUTURE PERSPECTIVES: 3D MICROSCOPY, BIOLOGICAL COMPLEXITY, AND IN VIVO MOLECULAR IMAGING The 21 st century should see a huge wave of novel microanalytical techniques. It is now becoming obvious that the major goals of the "post-genomics" and proteomics era will be to analyze protein function and protein-protein interactions at the single-cell level. We now realize that structural anatomy provides only limited
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of living Cells • Chapter 45
information about pathophysiological consequences, and that the future will involve a struggle to visualize complexes of single molecules (Ha et ai., 1999; Kusumi et al., 2005), biochemical events (Hirose et ai., 1999; Weijer, 2003) and activities of signaling networks (Hurtley and Helmuth, 2003). Many of these scientific questions are well-suited to molecular proximity methods such as FRET, FCS, FRAP. Eventually, these types of analysis will lead us to a new branch of cell biology - functional cell physiology. Imaging applied not only to living cells, but also to living organisms, should keep us occupied for the next few decades! (Fig. 45 . 12). Today, in addition to having a wide variety of fluorescent markers for subcellular compartments (Lippincott-Schwartz and Patte rson , 2003), we also have a broad palette of established fluorescent biosensors (Ting et ai., 200 I) and reporter molecules that can be used to measure cellular activities (Miyawaki et al., 1997; Miyawaki et al., 1999; Zhang et ai., 200 I; Sato et al., 2002; Sato et 01 .• 2003; Violin et al., 2003; Umezawa, 2005, see also Chapters 16 and 17 , this volume).
805
The importance of using FRET to probe the functional assembly of multi-subunit protein complexes, such as the neuronal receptors AMPA, NMDR, the hexameric connexins in the plasma membrane or the immuno-recognition complexes, will drive the development of new FRET partners with custom-designed spectral properties (Tsien, 2004; Galperin et ai., 2004). New high-speed/ high-resolution microscopes with improved sensitivity are under constant development (Stephens and Allan, 2003; Goldman and Spector, 2005), and will provide new tools transforming the singlemolecule analyses of "cellular imaging" into "molecular imaging." Cellular processes in the plasma membrane, the nucleus and other cellular compartments such as Golgi . ER, etc., will be resolved in rcal-time and at high spatial resolution. The resulting improvements in the understanding of the cellular physiology of both healthy and diseased organisms (including tumors), will support new frontiers in drug discovery. These are fields in which FRET can be applied to elucidate the core of many of the biological processes that must be understood before we movc on to clinical diagnostics and drug screening.
The 2004 palette of nonoligomerizing fluorescent proteins GFP-derived
mRFP1-denved
Evolved by SHM
~ ~~-------------~-------~7~~ ~~'~~
D
Elcc. 3804:W452488 516487/504 540 548 554 568 574 587 595 596 605 590 nm Em. 440 47!11505 509 529 537~ 553 562 581 585 596 610 620 625 636 648 nm
- I I" ~ " ~ FIGURE 45.12. Experiment performed during the UBC 3D microscopy cou rse. 2004. A simple experimental se tup for whole-organism imaging of a transgenic GFP mouse is shown in (A, B). (C) Fluorescence of GFP-expressing neuronal cells in cerebellum can be seen under the microscope even with 5x magnifi cation. and this allows preparation of tissue slices for high-resolution microscopy: (D) the insert shows the new "2004 palette of non-oligomerizing fluorescent proteins" (taken from Tsien, 2004), illustrating the future of live-cell imaging.
806
Chapter 45 • I. Majoul et al.
IN VIVO MOLECULAR IMAGING The field of molecular imaging is developing rapidly. Two relatively mature disciplines - molecular biology and radiology are coming together to create an interdisciplinary approach to the biology of tissues and cells and the interactions of molecules. Molecular imaging will add a new dimension to anatomical imaging and the evaluation of histological slides, providing us with the detailed information on the distribution and function of reporters and marker molecules needed to deepen our insight into physiological differences between normal and diseased tissues. The equipment required to provide this information is now becoming available. A new generation of "stick lens" objectives (Fig. 45.13) was recently introduced by Olympus to permit one to image planes inside the living animal without resorting to the fiberoptical confocal endoscopes described in Chapter 26 (this volume). Exciting developments in high-content screening are described in the next chapter. Many other initiatives are currently focused on the development of high-specificity/high-sensitivity probes designed to improve the detection sensitivity of molecular imaging by 10- to lOO-fold. A feature of this new generation of probes will be the development of "Molecular Libraries" that will help to overcome current limitations in our ability to detect and image specific mo-
lecular events in animal models and humans. Methods, such as microSpect imaging, as recently described by Choi et (II. (2005), demonstrate the potential for the pre-clinical application of in vivo molecular imaging, a topic also discussed in Chapters 26 and 38. The outlook looks both bright and multi-colored!
ACKNOWLEDGEMENTS We are grateful to Prof. Natalia B utkevich for extensi ve discussions of aspects of quantum mechanics and spectroscopy, to Dr. Alexander Goroshkov for his useful suggestions in modern microscopy and image acquisition, Dr. Clemens Kaminski and L.K. Bert van Geest for imaging and discussions about FUM, to Professor Alberto Luini for help in reading Leonardo da Vinci's original Italian "cryptic" text. We also want to thank Tom lovin, Bob Clegg Philippe Bastiaens and many others for guiding our initial steps into FRET-science. Finally, we are deeply grateful to Professor lames Pawley for his patience, generous interest and continuous support of our often "unconventional" performance. Some of the experimental data presented in this chapter (as indicated in the text) were acquired during the practical course "International 3-D Microscopy of Living Cells," held annually at the University of British Columbia, Vancouver, Canada during the years 2002, 2003, 2004 and 2005.
FIGURE 45.13. (A) Three, new "stick lens" objectives have been developed by Olympus especially for in vivo imaging of fluorescent signals in whole animals. These objectives were designed to be used with a new, miniaturized , tiltable confocal scan-head, the Olympus IVIOO Intravital LSM (B). The system uses an AOTF to select between 488, 561 , 633 and 748nm lasers. Three different stick objectives: 27xfNA 0.7 (3.5mm diameter, FOV: 200~m) , 20xfNA 0.5 and (1.3mm diameter, FOV: 200~m), 6xfNA 0.14 (1.3mm diameter, FOV: 670~m) and a comprehensive double-galvanometer scanner provided flexible raster scanning over a wide magnification range. A fixed , 4-channel dichroic beamsplitter and 3 turrets , each with 6 emission filters, provides wavele ngth selection for 3 PMT detector channels. To minimize the effect of invasive microscopy on living animals, the angle at which the objective and the scanning head approached the animal is adjustable (B) and the whole apparatus is enclosed in a light shield that can also serve as means of administering anesthetic gases.
Practical Fluorescence Resonance Energy Transfer or Molecular Nanobioscopy of Living Cells • Chapter 45
REFERENCES Bacia, K., Majoul, LV., and Schwille, P., 2002, Probing the endocytic pathway in live cells using dual-color fluorescence cross-correlation analysis, Biophys.l.83:1184-1193. Bastiaens, P.I.H., and Jovin, T.M., 1998, Fluorescence resonance energy transfer (FRET) microscopy, In: Cell Biology: A Laboratory Handbook (J.E. Celis, ed.), Academic Press, New York, pp. 136-146. Bastiaens, P.I.H., and Pepperkok, R, 2000, Observing proteins in their natural habitat: The living cell, Trends Biochem. Sci. 25:631-637. Bastiaens, P.I., and Squire, A., 1999, Fluorescence lifetime imaging microscopy: Spatial resolution of biochemical processes in the cell, Trends Cell BioI. 9:48-52. Bastiaens, P.LH., Majoul, LV., Verveer, P.J., Soling. H.D., and Jovin, T.M., 1996, Imaging the intracellular trafficking and state of the AB5 quaternary structure of cholera toxin, EMBO 1. 15:4246-4253. Bezzi, P., Gundersen, v., Galbete, J.L., Seifert, G., Steinhauser, c., Pilati, E., and Volterra, A., 2004, Astrocytes contain a vesicular compartment that is competent for regulated exocytosis of glutamate, Nat. Neurosci. 7:613-620. Blackman, S.M., Piston, D.W., and Beth, A.H., 1998, Oligomeric state of human erythrocyte band 3 measured by fluorescence resonance energy homotransfer, Biophys. 1. 75: 1117-1130. Blinks, J.R, Mattingly, P.H., Jewell, B.R., van Leeuwen, M., Harrer, G.c., and Allen, D.G., 1978, Practical aspects of the use of aequorin as a calcium indicator: Assay, preparation, microinjection, and interpretation of signals, Methods Enzymol. 57:292-328. Bunt, G., and Wouters, ES., 2004, Visualization of molecular activities inside living cells with fluorescent labels, Int. Rev. Cytol. 237:205-277. Butkevich, E., Hulsmann, S., Wenzel, D., Shirao, T., Duden, R., and Majoul, I., 2004, Drebrin stabilizes connexin-43 and links gap junctions to the submembrane cytoskeleton, Curr. BioI. 14:650-658. Campbell, R.E., Tour, R., Palmer, A.E., Steinbach, P.A., Baird, G.S., Zacharias, D.A., and Tsien, RY, 2002, A monomeric red fluorescent protein, Proc. Nat!. A cad. Sci. USA 99:7877-7882. Chalfie, M., 1995, Green fluorescent protein, Photochem. Photobiol. 62:651-656. Chalfie, M., Tu, Y, Euskirchen, G., Ward, W.W., and Prasher, D.C., 1994, Green fluorescent protein as a marker for gene expression, Science 263:802-805. Choi, S.R., Zhuang, Z.P., Chacko, A.M, Acton, P.D., Tjuvajer-Gelovani, J., Doubrovin, M., Chu, D.C., and Kung, H.E, 2005, SPECT imaging of herpes simplex virus Type I thymidine kinase gene expression by l(123)I]FIAU(l).Acad. Radiol. 12:798-805. Clegg, R.M., 1992, Fluorescence resonance energy transfer and nucleic acids, Methods Enzymol. 211:353-388. Cohen-Cory, S., 2002, The developing synapse: Construction and modulation of synaptic structures and circuits, Science 298:770-776. Cole, N.B., Smith, c.L., Sciaky, N., Terasaki, M., Edidin, M., and LippincottSchwartz, J., 1996, Diffusional mobility of Golgi proteins in membranes of living cells, Science 273:797-801. Conn, P.M., ed., 1991, Electrophysiology and Microinjection, Academic Press, London. Del Pozo, M.A., Kiosses, W.B., Alderson, N.B., Meller, N., Hahn, K.M., and Schwartz, M.A., 2002, lntegrins regulate GTP-Rac localized effector interactions through dissociation of Rho-GDI, Nat. Cell BioI. 4:232239. Eisliger, M.A., Wachter, RM., Hanson, G.T., Kallio, K., and Remington, S.J., 1999, Structural and spectral response of green fluorescent protein variants to changes in pH, Biochemistry 38:5296-5301. Erickson, M.G., Moon, D.L., and Yue, D.T., 2003, DsRed as a potential FRET partner with CFP and GFP, Biophys. 1. 85:599-611. Forster, v.T., 1948a, Zwischenmolekulare Energiewanderung und Fluoreszenz, Ann. Phys. 6:54-75. Forster, T.H., 1948b, Versuche zum zwischenmolekularen Ubergang von Elektronenanregungsenergie, Naturwissenschaften 33:93-100. Galperin, E., Verkhusha, v.v., and Sorkin, A., 2004, Three-chromophore FRET microscopy to analyze multi protein interactions in living cells, Nat. Methods 1:209-217.
807
Gerritsen, H.C., and de Grauw, K., 2001, One- and two-photon confocal fluorescence lifetime imaging and its applications, In: Methods in Cellular Imaging (A. Periasamy, ed.), Oxford University Press, New York, pp. 309-323. Gordon, G.W., Berry, G., Liang, X.H., Levine, B., and Herman, B., 1998, Quantitative fluorescence resonance energy transfer measurements using fluorescence microscopy, Biophys. 1. 74:2702-2713. Griesbeck, 0., Baird, G.S., Campbell, RE., Zacharias, D.A., and Tsien, RY, 2001, Reducing the environmental sensitivity of yellow fluorescent protein. Mechanism and applications, 1. Bio!. Chem. 276:29188-29194. Grutzendler, J., Tsai, J., and Gan, W.B., 2003, Rapid labeling of neuronal populations by ballistic delivery of fluorescent dyes, Methods 30:79-85. Ha, T., Ting, A.Y, Liang, J., Caldwell, B., Deniz, A.A., Chemla, D.S., Schultz, P.G., and Weiss, S., 1999, Single molecule fluorescence spectroscopy of enzyme conformational dynamics and cleavage mechanism, Proc. Nat!. Acad. Sci. USA 96:893-898. Hahn, K., 2003, Monitoring signaling processes in living cells using biosensors, Sci. STKE 205:tr5. [DOl: 10.1126/stke.2003.205.tr5] Heim, R, Cubitt, A.B., and Tsien, RY, 1995, Improved green fluorescence, Nature, 373(6516):663-664. Heim, R, and Tsien, RY, 1996, Engineering green fluorescent protein for improved brightness, longer wavelengths and fluorescence resonance energy transfer, Curro Bio!. 6: 178-182. Heim, R, Prasher, D.C., and Tsien, RY, 1994, Wavelength mutations and posttranslational autooxidation of green fluorescent protein, Proc. Natl. Acad. Sci. USA 91:12501-12504. Hell, S.W., 2003, Toward fluorescence nanoscopy, Nat. Biotechnol. 21:1347-1355. Herman, B., Gordon, G., Mahajan, N., and Centonze, V.E., 2001, Measurement of fluorescence resonance energy transfer in the optical microscope, In: Methods in Cellular Imaging, (A. Periasamy, ed.), Oxford University Press, New York, pp. 257-272. Hirose, S.K., Kadowaki, M., Tanabe, H., Takeshima, and lino, M., 1999, Spatiotemporal dynamics of inositol 1,4,5-triphosphate that underlines complex Ca2+ mobilization patterns, Science 248: 1527-1530. Hoppe, A., Christensen, K., and Swanson, J.A. , 2002, Fluorescence resonance energy transfer-based stoichiometry in living cells, Biophys. 1. 83:36523664. Hurtley, S.M., and Helmuth, L., 2003, The future looks bright, Science 300:75. Janetopoulos, C., Jin, T., and Devreotes, P., 2001 , Receptor-mediated activation of heterotrimeric G-proteins in living cells, Science 291 :24082411. Jones, J.T., Myers, J.W., Ferrell, J.E., and Meyer, T., 2004, Probing the precision of the mitotic clock with a live-cell fluorescent biosensor, Nat. Biotech. 22:306-312. Johnson, D.A., Voet, J.G., and Taylor, P., 1984, Fluorescence energy transfer between cobra a-toxin molecules bound to the acetylcholine receptor, 1. BioI. Chem. 259:5717-5725. Kohen, E., Legallais, V., and Kohen, c., 1966, An introduction to microelectrophoresis and microinjection techniques in microfluorimetry, Exp. Cell Res. 41:223-226. Kenworthy, A.K., and Ec!;
~
0
5" ~
0 ~ c-
o OJ
5·
(")
U>
~
~
CD
~
0 ~ c-
o OJ
U>
!!!. 0
a.
5·
CD
CD
OJ
5· --i
0:
3
~
~
0
3
OJ
0 0 ~ c-
5· CD
U>
0, then the maximum timing error is I tR/tM l·tM (Eq. 2). From Table 48.1, one would expect that the theoretical maximum error would be near 2 . tR and that this would be worse at the start and the end of a movie clip. However, when we captured a series of out-of-register frames it became clear that the movie player had assigned each frame to its closest refresh period rather than to the next one in sequence, that is, they use a centric update that reduces the worst frame timing error by half. Despite this smart feature, Equation 1 has two noteworthy consequences. If there is any timing error whatsoever, the maximum error will always be larger than half a refresh cycle. Even more unexpectedly, any apparently rational ratio between frame rate and refresh rate will result in a largest possible timing error of one entire refresh cycle! So it is not wise to somewhat align the frame rate; they need to be identical. Practically, this means that if you cannot align your movie frame rate to the refresh rate of the display (something that you might not know) then a timing error of up to one refresh cycle is nearly unavoidable. This timing error is less dramatic if the refresh rate is significantly higher than the movie frame rate. A 7 Hz movie will look OK on all types of displays, whereas the 50Hz animation noticeably varies in playback speed when viewed on a 60 Hz device (often referred to as breathing). We keep hammering this issue because, very often, one has a choice as to how fast your animation plays back, and slowing it down might actually improve the visual impression. Also, some advanced movie formats (such as MPEG-2 and MPEG-4) do not encode individual movie frames but evolve one frame from the previous one. These formats can produce frames at any desired point in time, not just at the original frame rate. This feature is called pull down and is used for standards conversion between television formats. In the case of MPEG-2, it actually can be performed in real time on the very fastest single-CPU machines available. It does not yet offer the quality of offline pull down but will definitely become a method of choice in the future. The MPEG-2 player from Philips already offers this feature. So far no MPEG-4 player can do online pull down. As an intermediate solution, one could create the same movie for each of the refresh rates one is likely to face, one for your native display speed and one for 60 Hz projectors.
Motion Picture Artifacts Besides the patch and pixelation artifacts inherited from still images, movies face timing subsampling artifacts (temporal artifacts). If a scene evolves too much from one frame to another, the movie will be plagued with temporal aliasing. As in the case of an image with insufficient resolution, this has absolutely nothing to do with the encoding of the movie or technical restrictions and depends solely on the mismatch between movie acquisition and human perception. Fortunately, because human vision is outstand-
839
ingly tolerant with temporal artifacts and can compensate for it, temporal aliasing catches much less attention than subsampling images spatially. In addition, automated software that can be applied to time sequences is still much less widespread than imageprocessing algorithms for still images. Finally, because movie formats are seldom used for image-processing applications, we have limited the issue of temporal artifacts to coding errors. A movie with a fixed data rate cannot always perfectly encode the amount of change between two frames. The different formats cope with this problem in very different ways. The MPEG formats have switches to handle this scenario. MPEG-4 can be set to "scene change detection" that will cue restarting encoding at steep scene changes. MPEG-2 can be allowed to arbitrarily increase its data rate (called variable bitrate encoding) when a scene change requires this. Both solutions come at the expense of higher storage requirements. The other movie compressors also invest more in coding data to handle changes but they do so much less consciously, to the extent that image noise can drive their coding effort to the same level as real scene changes. Conversely, limiting their noise encoding restricts their scene update accuracy. Naturally, none of this holds if one uses a coding format that does not include any compression. TIFF series are lossless as they offer no data reduction either within the frame or over a sequence of frames. For today's large micrographs of one or more channels and megapixel resolution per frame, these formats have little relevance except for programs that allow us to pan and zoom small windows in these datasets. The second type of artifact comes not from the information encoded in the file but from insufficient machine performance for decoding the animation or from the limited capabilities of the decoder. This can be a very serious restriction when playing back large frames at high rates, especially if one must use an unknown computer. All MPEG-4 formats, independent of resolution, are computationally expensive. Do not expect an aged computer to perform seamless MPEG-4 playback. At rigid timing, the player will drop frames, causing motion under-sampling to become even worse. In addition, when forced to play all frames, the player may not decode all the image information and may play coarse images with ripples in them. The results look very similar to badly encoded movie formats ill general. As outlined above, decoders usually do an excellent job of squeezing the best possible result from the movie format and computer resources available. The same cannot be said about encoders (also referred to as compressors) because the task of the compressor is infinitely complex and as long as the compressors do not understand what they are encoding, they have to rely on heuristics to decide what image information they can throwaway without visible harm to the movie quality. After all, this is the only way to reduce the data. The criteria for useless information are unique to each compressor but all are optimized to handle cinema movies or TV shows, a condition that sets clear limits on the color space they must handle and how well they replay motion. Unfortunately, cinema-optimized encoders are notoriously bad at displaying small moving objects against a large constant background, that is, images that look like a fluorescent object on a black background. This is why some encoders have a special setting to encode sports shows, a setting that can be useful to process time sequences of micrographs. Without this optimization, small moving objects drag strong ripple artifacts around with them. On the other hand, fast motion compensation comes at the price of a generally higher noise level. MPEG-2 encoders can achieve both low noise and
840
Chapter 48 • F. Margadant
small object motion at once but only by requiring higher data throughput. It is usually worthwhile pay the price of a large data file as it usually costs less than more computing power.
The MPEG Formats A notable feature of movie formats is that the more advanced formats can arbitrate between performance and quality, and are able to sacrifice resolution and sharpness, in order to maintain the frame rate. Depending on the data to be displayed, this can be either useful or noxious. Two major approaches exist to implement this so-called rigid timing. MPEG-l (as it is implemented in the Video For Windows format, also known as H.261) and MPEG-2 (also named H.262) limit the complexity of the movie format (also referred to as bitrate capping) upon coding the movie. If content changes too much between frames, capping can only be achieved by jettisoning some of the resolution. MPEG-2 can encode two different bitrates at once, called the basic and the helper stream. If performance allows for it, the helper stream is decoded as well and this helps to improve the quality. If the resources are insufficient, the basic stream supplies the output. The concept is not very flexible and it can only either boost resolution or lower noise. The unique feature of MPEG-2 is that because the helper stream neither has to be decoded nor read, playback performance is assured. Because of this, frame-rate tests for MPEG-l and MPEG-2 yield trivial results - independent of the scene complexity, the frame rate stays constant and frame dropping will only occur when CPU power is very limited and any sort of movie playback is not really possible anyway. On current low-end 2 GHz systems, the CPU load is below 25% of the CPU resources even for very complex movies (see description of entropy) and hence not worth investigating. MPEG-4 incorporates MPEG-2 modes but adds the capability of limiting the format complexity during playback. Therefore, an MPEG-4 file can be generated at a high bitrate and then be played back at a much lower one (Ebrahimi and Pereira, 2002). In contrast to the MPEG-2 helper stream, however, all data must be read to decode an MPEG-4 stream and this can impede playback performance. Consequently, the player software has a substantial influence on MPEG-4 performance (Walsh and Bourges-Sevenier, 2002). Many custom players such as the QuickTime or DivX players use the settings of the MPEG-4 codec and the movie format and do not curtail the bitrate. If playback content is too complex to be mastered at full frame rate, the players drop frames. The DivX player incurs a significant jitter because it only notices the timing loss after the frame has been rendered, and CPU cycles that could have been used to render the next frame have just been wasted. However, both players can be performance-tuned by changing the codec settings while leaving the movie file untouched. The Windows Media Player reduces the content complexity within a few frames, so that jitter and frame loss occur only briefly (adaptive bitrate adjustment sometimes called elastic bitrate). As useful as these procedures are, they make it hard for the user to tune the trade-off between the available performance and the image. The long startup time and thc huge computing requirements require one to justify when to use MPEG-4. A small 6 MB MPEG-4 file included into PowerPoint did consume an initial startup delay of 2.5 s, allocate 28 MB of system memory, and cause 78% CPU load while playing. Due to the serial coding of the format, however, memory allocation will not exceed some 40MB, even for long movies. Also startup times will not vary with file size. However, the decoding effort prior to anything being
displayed can be cumbersome during a presentation, as can the high CPU load. The use of MPEG-4 in microscopy should therefore be restricted to what it does best: playing long, highly compressed movies at guaranteed video rates. Like MPEG-2, MPEG-4 can play back an unlimited number of frames from the disk without interruption. Unlike MPEG-2 however, it eats many more CPU cycles and the number depends strongly on the movie content. The complexity of the MPEG-4 resource requirement is touched in the benchmarks listed below. Of more concern, one cannot be sure that an MPEG-4 file that runs smoothly on one system will perform well or even acceptably on a slightly less-powerful system. The only recommendation we can give here is to play the movie on your test system at maximum quality and run the CPU meter (in the TaskManager in Windows systems). Check that the CPU load does not exceed that of the presentation system. Be sure to leave plenty of margins when planning to play an MPEG-4 file this way. In contrast to MPEG-2 hardware, MPEG-4 accelerators are still scarce and - to make things worse - depend on the player software. MPEG-4 exists in two completely different levels of operation: the more widespread H.263 level, referred to as MPEG-4 part 2, and the widely hailed H.264 (MPEG-4 part 10) (Richardson, 2003). H.263 achieves its very high compression rate by using a plethora of different compression techniques that all rely on certain assumptions about the movie content. For example, H.263 can encode 2D scene rotation, panning, dimming, and even object rotation without storing any image information whatsoever. Sadly, few of these abilities matter for micrographs. On the other hand, H.264 deploys a new psycho-visual coding technique that is more effective than MPEG-2 but abandons the scene understanding of H.263. Therefore, it compresses less than H.263 but is also less prone for artifact creation. As a consequence, H.263 only works well in microscopy in sparse scenes and when the moving objects are not too small. When maximum compression is not essential there will be no serious use for H.263 in microscopy. Because modern Apple systems support the H.264 decoders as standard/ one cannot use such a system to assure your movie will play well on an inferior machine. For PCs, the situation is less encouraging but easier to control. Although later Radeon or X-series cards from ATI support the decoding of MPEG-4, they do so to a lesser degree than the Apple H.264 decoders but still boost performance. To get a better estimate, the hardware support for the MPEG-4 codec can be switched off and hence one can get a more reproducible performance estimate.
MPEG Display Formats The most critical drawback of the MPEG formats is that they come in very few display raster sizes. MPEG-l is available only in 352 x 288 resolution and hence limited to coarse VHS quality movies only. The upside is that it can be universally played and is not very computing power hungry (decoding effort about one third of main format MPEG-2, that is also well within modest computer limits). The so-called main profile of MPEG-2 is defined for the resolutions (called levels) 352 x 288 (low), 720 x 576 (main), 1440 x
7
H264 offers the same seamless playback and guaranteed frame rates as MPEG-2 does but at roughly half the bandwidth.
Display and Presentation Software • Chapter 48
1152 (high 1440), and 1920 x1152 (high) and for two frame rates 30Hz (NTSC) or 25 Hz (PAL). There are three more MPEG-2 profiles but they offer no additional resolutions or frame rates. The hardware installed on most new graphics boards decodes the main profile but does not go beyond the 720 x 576 resolution. If your movie can be played back at one of these two frequencies or an integer fraction of it, then MPEG-2 offers low load for the computer, universal playback, and color calibration for many systems, and can be written to a DVD. In a presentation, it still may not be wise to playa 2 framels (fps) movie as an NTSC MPEG-2 because doing so may inflate a single frame to 15 frames. Also, as the jitter measurement in Table 48.1 shows, it can be cumbersome to encode a 27 fps movie to either PAL or NTSC. For the same reasons that most projectors offer a 30 Hz refresh rate, the NTSC format in a presentation will often play more fluently than the PAL format. If the movie is played full screen, then this limitation does not apply as the refresh will be switched to the 25 Hz needed for PAL. The MPEG-4 format allows for arbitrary formats but playback software will not support deviations from the MPEG-2Ievels. The codecs we tested would work on the main and the low MPEG-2 level for both MPEG-2 and MPEG-4. None would support the two high levels, and one MPEG-4 codec even crashed when using alternate resolutions. The MPEG standard (Watkinson, 2000, pp. 9-16, 47) assumes that the analog image data is filtered in such a way that it cannot change more rapidly than the resolution that the digital movie contains. In micrographs, this can be assured by not under-sampling the images. Only those digital micrographs that are properly sampled and not too noisy constitute suitable targets for either JPEG or movie compression algorithms. Unlike JPEG series, high-quality MPEG compression also requires proper temporal sampling to avoid compression artifacts. Consequently, MPEG compressors incorporate digital smoothing filters to handle noisy source images and, when an MPEG-compressed animation appears unnaturally sharp or hard edged with weak ripples and creases, this is a limitation of the standard and not of the player software.
Very High Resolutions Except for the new H.264 standard, there is no widespread standard for playing movies at 1000 pixel square or higher resolution as is needed by microscopists. The other MPEG formats are practically limited to the MPEG-2 main level, the DVD's resolution. Although QuickTime can assemble single frames of any resolution in use today into movies (there is a filesize limit and a 214 total pixels resolution limit) apart from the dated QuickTime compressor, it does not interface to any advanced movie standard to compress and store the result. Microsoft's AVI format offers some of the same capabilities but with the even worse restriction that they provide no compressor at all and merely accommodate playback. Movie authoring software such as Adobe Premiere, Final Cut Pro, or Jasc AnimationShop can write megapixel movies, and so do the tools of most microscopy software. When it comes to compression ratios, QuickTime performs better than the antique Sorensen or Cinepak coders used to support the high resolutions for AVI. Apple Final-cut Pro can now create HDTV MPEG-4 (H.264) and H.262 movies.
Movie Compression and Entropy To obtain some generally valid results and establish a concept of movie performance and quality, it is necessary to focus on the
841
complexity of the movie. The more information a movie conveys (not necessarily the same as useful information!) the harder it is to compress, to decompress, and to play back in real time. Complex movies will both suffer from lower frame rates and more compression artifacts. In addition, they will not compress as well. Therefore, the compression factor can be used as a scale for the movie complexity. The compression depends on the entropy of the movie information - the more random the images and animation are, the less it can be compressed without apparent degradation. To get "clean" measurements, we have used an image generator that yields image sequences in which the randomness is controlled. As a result, the image or movie content will be perfectly random to test the movie format but the local changes in the image and the fluctuations in time will be more rapid with higher entropy. In information theory, one can either use this "channel entropy" when talking about compression or when assessing it by taking the autocorrelation of the image sequence. This measure can be scaled to 1 for a completely constant image sequence and 0 for one where neighboring pixels have no relationship to each other in space and time. To benchmark the players and the formats, we devise three movies, one with autocorrelation 0, that is, completely random signal (Format A). One with the highest useful complexity, that is, a movie that is Nyquist sampled in space and time but is as random as possible within these restraints (Format B). And finally, a sparse movie as they are popular in fluorescence labeling, with a real two-channel sequence forced to proper sampling in space and time (Format D). Very often, temporal and spatial sampling is not performed properly. Because, despite all efforts to preach the ultimate importance of proper sampling, under-sampled material still enjoys popularity, we also include an under-sampled version of Format D as Format C. For the compressor, A is harder to handle than B, B harder than C, and C harder than D. Instead of some meaningless autocorrelation factor, we give the compression ratio (q) as a percentage, that is, the size of the compressed movie file divided by the size of the uncompressed image series. Unfortunately, decompression mostly behaves inversely and a well-compressed movie (low q) will take more computer cycles to play than a more complex one that is compressed less. The exceptions are the MPEG-2 and H.264 formats with a variable bitrate, as here compression and playback effort is symmetric. Because they all guarantee 25 to 30 frames per second, we do not include playback speed in the benchmarks.
Performance Benchmark Speed benchmarks give you a very coarse idea of what is doable in movie animation. We intentionally do not use the latest Nvidia GeForce6 series or the AT! X600 and X800 engines that are common in G5 Apple machines because we need to get the presentation running on an average system available today. Hence, we test a notebook with a 2-year-old GeForce4MX mobile engine and a desktop with a GeForce 5600. Both cards do not support hardware movie features for non-MPEG playback and hence we should obtain movie performance that is somewhat related to system power.
Compression Ratios for TV-Sized MPEG Movies VBR is variable bitrate, that is, a compression that adapts to the movie content. CBR stands for constant bitrate, meaning that the compression ratio is held constant, independent of the content.
842
Chapter 48 • F. Margadant
TABLE 48.2. Compression of TV-Sized Movies (720 x 576) Fonnat/fype MPEG-2 MPEG-4 MPEG-2 MPEG-4
CBR CBR VBR VBR
A
B
C
0
2.3% 0.8% 2.5% 2.0%
2.3% 0.8% 2.5% 2.0%
2.3% 0.8% 2.1% 1.7%
2.3% 0.8% 2.1% 1.7%
Please note that the two modes cannot be compared to each other as they serve completely different quality needs. CBR is used wh~n there is only a limited bandwidth that can be transferred. VBR IS used when one needs at least a certain quality level. The settings used here are default settings for 2-h DVDs (M2 settings), resulting in a 5000 to 6000 kbitls bitrate. The CBR ratios are trivial as the movie size is given via the bitrate. A high-quality DVD movie would use around 6000kbitls. The limit of the MPEG-2 standard (even for VBR movies) is 15,000 kbitls, including all soundtracks. The MPEG-2 compressors one can buy generally cap at 9200 kbitls. Commercial DVD movies will not go below 4000kbitls. The raw movie material assumed in developing Table 48.2 would consume a datarate of 30 Hz x 720 width x 576 height x 8 bits per sample, that is, 298,500kbits/s. MPEG-4 goes far below this mark. Movies are often encoded at 600 to 700kbitls and 1500 kbitls is considered high quality. These numbers (Tables 48.3-48.6) list file size and frames per second of different formats. Determining quality involves examining the artifacts introduced and for most rich micrographs, the uncompressed and motion JPG formats may very well be the methods of choice, despite their lack of speed and high memory requirements. Cinepac may be a good and universal ch~ice f~r a Web page but hardly ever lives up to the needs of dlsplaymg micrographs. Important note: the numbers for QuickTime are QT 6.5 benchmarks, QT 5 actually has higher frame rates for the uncompressed case (17.5 fps) and identical rates for motion JPEG. Because QT 6 (and higher) is more efficient while compressing and supports MPEG-4, we use the latest version. As in the case of displaying static images, one must realize that the presentation system may have a different screen resolution from the system you tuned the movie on. In this case, the content must either be zoomed or - much more likely - downsized to match the projection system. Respect the restrictions on downsampling given early in this chapter and also keep in mind that down-sampling will cost CPU power, though less than 5% on a 2 GHz system. If you must run your animation on an unknown computer, make sure that you have plenty of leeway.
TABLE 48.3. Compression Ratios for PAL TV-Sized Movies (720 x 576) Format/fype
q of A
q ofB
q ofC
q of 0
QuickTime NC QuickTime JPG QuickTime QT AVI Cinepac
100% 27% 17% 7%
100% 18% 18% 7%
100% 3.5% 3.5% 5%
100% 3.5% 3.5% 4.5%
TABLE 48.4. Playback Frames per Second for PAL TV-Sized Movies (30-s Window) Format/fype QuickTime NC QuickTime JPG QuickTime QT AVI Cinepac
720 x 576 A
720 x 576 B
720 x 576 C
720 x 576 0
27fps 22fps 32fps 60fps
27fps 22fps 34fps 60fps
27fps 24fps 34fps 60fps
27fps 24fps 34fps 60fps
Storing Your Presentation for Remote Use Besides texts and drawings, Keynote and PowerPoint presentations can include image and animation data as well as the complex scripts needed to present them. Images and animations can either be stored within the presentation or they can be linked to the latter by reference. In the latter case, PowerPoint or Keynote will retrieve the original image or movie by a path and file name (the link), and then display it by rules stored within the presentatio~. The display rules at a minimum consist of dimensions and POSItions but can also contain timing and movement, triggers that react to certain other events or scripts. Due to the native viewer concept, PowerPoint hosts images in their original format and hence does not compress or reformat the image data for storage. Any resizing and resampling is done only when the image or movie is displayed. This means that any image included in a PowerPoint presentation will increase the PPT file by the size of the original image (plus a base overhead of about 4kB for the displaying rules) and will have to be decompressed and resized each time it is displayed in a slide. The data of each image is then cached by PowerPoint so that, the display is faster, the second time it is shown, even though the image must still be decompressed each time. As a result, it is better to only use images that have no more resolution than is actually needed. 8 There are several reasons for only storing references within a presentation. The premier one is that only a single cop~ of an image or a movie has to be stored no matter how many dIfferent presentations may use it. This also assures update consis.tency, as changes to a file will be available in all the presentatlOns that include it. The main drawback is that keeping the references consistent when transporting a presentation to a different computer is not trivial because the path referenced in the document must also function properly on the presentation computer. There is no perfect solution to this situation and all approaches are plagued with obvious shortcomings: 1. PowerPoint offers apack-and-go mode. The file made using this command includes the images and animations (obtained from their references). This is the fastest (and fairly safe) way to complete the task but it leaves one with a single-use present~tio~. S~ould you perform changes and add material to the presentatlOn, It WIll .be cumbersome to revert it back into a presentation with links agam. The inclusions have to be deleted manually and replaced by new, valid links. So if there is the chance that you will want to update
8
An exception might be if one is using Keynote or PowerPoint .not primarily for a presentation but instead as a way of storing a number of figures for a printed publication. This can be a convenient means of accumulatlOg and annotating the figures and in this case, you can store the Images at the resolution appropriate for their final use in the article.
Display and Presentation Software • Chapter 48
TABLE 48.5. Compression Ratios for Large Movies (1 024 Square) PlayerlFormat
q of A
q ofB
q ofC
q ofD
QuickTime NC QuickTime JPG QuickTime QT AVI Cincepac
1000/c 24.2% 16.6 % 5.2%
100% 15.1% 16.7% 5.2%
100% 2.75% 2.9% 4.1%
100% 2.75% 2.9% 4.0%
your presentation after packing, this approach will not work well. Also, failures of this procedure have been reported. It works flawlessly for images and PowerPoint animation software but large movies that may be processed by codecs not originally shipped with Windows, sometimes do not play. This is particularly common with movies formats that contain references in themselves such as do MPEG-4 movie streams. Even under OSX, non-standard codecs can lead to incomplete wrapping of the presentation. The danger here lies in the fact that this situation cannot be avoided by packing then testing the presentation, as the references on the machine the packing is performed on are still valid and failure will only occur once the presentation is loaded onto the target computer. A safe way to force a valid test is to unlink the references - if all your included data is distributed in a path (that would be a directory name in Windows or a folder in OSX) say "all," rename "all" as "former_all." Then running the packed presentation will actually invalidate improper links and the error will show when the presentation is run. 2. Moving the entire filesystem (or file tree) is a very flexible option for people who try to avoid version conflicts between presentations. If you keep all the data you work with and intend to present in a certain path (a drive and directory tree under Windows or a mounted folder under OSX), then you will be able to move this file system between different computers and keep all references in your presentations valid. For Mac users, this goal can be achieved almost trivially by having a portable disk with a name that does not occur on any of the target computers. So when the disk with the cryptic name is plugged to an alien Mac, it will appear as a folder accessible from the desktop. All presentations can then be moved without any changes. pes have an unpleasant restriction that the drive name assigned to such a disk must be free on the target computer. So it is not wise to use any disk drive letter either between A and E, as they tend to be occupied by disks, optical drives, and memory sticks, or using the last characters V through Z as they may also be used for images of network drives. Other than that, the more remote letters work reliably and using a drive letter far from these zones of confusion usually works well. Newer operating systems, such as Windows 2000 and XP, allow the use of "soft links" similar to the folder concept of the Apple operating system. You can link your portable drive with a soft link from the desktop and when transporting presentations simply re-create that soft link on the target machine. This approach
will not work reliably with dated PowerPoint versions as soft links are often resolved to their absolute paths. Given that you use a drive X: with all your presentations, hence using the basic path X:\all_my_presentations, you could soft link it from the desktop, say as "all_my_presentations." PowerPoint 97 may still store the references with the drive name though and hence limit portability. However, as long as this link name is not in use on the target system, this provides a very reliable transportation mechanism for newer PowerPoint versions. 3. Approach 2 won't work if either you cannot use your own disk or want to hand out a presentation. Sharing a presentation is made much easier if the sharing is planned in from the start. If all images and animations for a presentations are made available in a single directory - or folder - from the beginning, then simply copying that folder will make the presentation mobile. This approach is referred to as aflatfile structure, in contrast to the directory tree which is hierarchical. For older PowerPoint versions (especially the still-popular 95 and 97 releases), this can lead to a soft error the first time the presentation is run and PowerPoint will feel obliged to ask each reference for an updated location. The newer versions, however, will deal with the flat file structure smoothly. Note, however, that with this approach, the person who receives the presentation has copies of the images and movies available in native format. Whether or not you want to give them away, PowerPoint itself does not prevent your data from being extracted. PowerPoint files are documented and even a pack-andgo presentation can be disassembled into its components again. To obtain more sophisticated protection of your images and animations, make sure to incorporate a copyright watermark into the scans and the movies. Adobe PhotoS hop and Premiere, as well as some plug-ins for QuickTime, will perform this task. Watermarks are very difficult to remove with current tools. Another strong protection measure is to lower the resolution of the image material. Scans reduced to half their size, say 512 square will look fine on a projector but will look inferior when printed. Animations in MPEG-l format can even look great despite the fact that they are of far lower resolution (352 x 288 pixels at native resolution) than the time series from which they were obtained (usually of megapixel resolution). Both measures, watermarks and lowered resolution, provide the best available protection of your data at this stage. More advanced protection algorithms, such as steganography, which encodes additional information into file formats, is sometimes removed when "saved again" on some tools. Field marks or any other watermarks that cover the entire image are very safe but do impair the image quality. They imprint a noise pattern on the image that is unique to a password you choose and the pattern is chosen in a way that standard filters, such as smoothing or noise reduction, will not remove. In contrast to normal watermarks, however, it is not visible to the eye and hence has no meaning for daily copyright issues. It also does not survive reliably in print and so will not stop others from using it. In Apple Keynote 2, the Save As command yields a window that asks if you want to copy theme images, audio, or movies into
TABLE 48.6. Playback Frames per Second for Large Movies (30-s Window) PlayerlFormat QuickTime NC QuickTime JPG QuickTime QT AVI Cinepac
843
1024 x 1024 A
1024 x 1024 B
11.5fps 7.0fps 16.0fps 26.5fps
Same 8.0fps 15.5fps 27.0fps
1024 x 1024 D Same 8.0fps 17fps 25fps
1024 x 1024 D Same 8.0fps 17fps 25fps
844
Chapter 48 • F. Margadant
the document and this seems likely to produce a file similar to the PowerPoint pack-and-go file.
Taking Your Presentation on the Road: Digital Rights Management and Overlaying When playing animation formats, some advanced features of the native player software can cause substantial inconsistencies. All players are optimized to provide high performance and will exploit hardware features if the drawing libraries (DirectX and QuickDraw) allow this. One of the very popular features shared by the two main players in the field - literally, for once - is dubbed overlaying. For this, the content of the movie frames is not drawn directly into the video memory but into a secondary buffer named the overlay. When the graphics engine refreshes the display, it reads the contents of the overlay, if one is present, rather than the contents of the main image buffer. Overlays must be of rectangular shape and the hardware places strict limits on how many overlays there can be in a display frame . The limit can actually be as low as I or O. The advantage of doing this is that the display does not have to be redrawn when the movie plays. This permits a much more naturally integrated movie window that will respond to scaling and moving much like any other application, and that hides its thirst for computing power. However, on some hardware, overlaying becomes a problem when a projector is added in parallel to the primary display (i.e., on a laptop, the number of overlays for desktop machines is often more than I and hence the problem may not arise - repeat "may"). Under these conditions, sometimes the overlay only works for the primary display and the movie output on the projector is blank (actually, black). Overlaying can be turned off in MediaPlayer: choose Tools -7 Options, switch to the Performance tab and choose Advanced, and there, under Video Acceleration, you can disable Use Overlays. There is a second overlay switch for DVD pl ayback below the first one that you might have to deactivate if you play your movie from a DVD. And in QuickTime: go to Edit -7 Preferences -7 QuickTime Preferences , toggle to Video Settings, and deactivate Enable DirectDraw. As thi s will disable all acceleration via the graphics library and hence impair performance, it is not the generic setting to be used all the time. It is just a possible solution to the blank-movie window problem described above. A much more general problem that seems to be dramatically increasing over time is the Digital Rights Management mechanism built into the standard players. This is a protection mechanism intended to uphold copyrights and inhibit duplication of copyrighted material and is capable of preventing you from grabbing screen shots and sequences (clips) from DVDs or other restricted movies. The problem is that, as there is no open publication of what DRM does and how it affects your system, the normal user can have problems when the player interprets your movie as being copyrighted even when it isn' t. This type of error can happen when using advanced formats 9 in QuickTime and MPEG2 and the results
can vary from not being able to play the movie in PowerPoint to just not being able to display it on a projector. A lesser nuisance is that movie startup may be delayed while the player tries to check to see if it is authorized to play the format involved. The safest way to prevent problems of this kind is to keep your movies in the original format when playing them. Even playing a DVD as part of a presentation may lock the player to a fixed frame rate and format. If the movie files are present in the original, simple format (MPEG files or AVIs that you wrote yourself), it is very unlikely that you will encounter these problems. Finally, resist the temptation to update the version of your native players just before a presentation. Continuou sly escalating DRM requirements may make unplayable a movie that had been playable on an earlier version. Although the chances that this will happen are small, they are fatal for a presentation. In general, it is inherently risky to update your player as some of your movie formats may no longer be supported or the new player may insist that your formats must comply with certain rules and will not play them unless they do. This applies particularly to QuickTime files and MPEG-2 or MPEG-4 formats. At present, most player software still behaves benignly. They will play all formats to the projector if the overlaying is switched off, and when they cannot retrieve a license, they search for a movie file, and then play this format. One notable exception involves movies on DVDs, and these come with severe restrictions and may not even be able to be played at the frame rate of your projector. However, the topic of DRM continues to evolve quickly and, at present, the only way to retain control over your animations is to keep them in the original formats that you can edit (i.e., the production format) . In other words, you can use the QuickTime, MPEG-4, and MPEG-2 formats, but you must keep them original (i.e., after writing them, do not package them or write them onto a DVD or VCD), unless you are sure that you have not included or activated an any copyrights when producing the movie . If you do have to resort to DVDs, make sure that you do not create a Region Code that will prevent the DVD from playing on another continent than the one on which the DVD was created. This problem can be a particular nuisance when assembling presentations for use at international conferences. The DVD feature to avoid at all costs is User Prohibitions, that is, locking the way the DVD has to be played and accessed. These constraints will prevent the player from playing even an excerpt from the movie and they will also lock down the frame rate of the playback. Unless you can make sure that these features are switched off when creating your DVD, you run the ri sk that others may not be able to play your presentation and you , yourself, may not be able to play it on any other system or, at least not in the sequence you intend to. To be prepared, it is always wise to carry a copy of your production files with you. Good Luck!
HELPFUL URlS
The TIFF Standard Archives http ://www.digitalpreservation.gov/fo rmats/fdd/fdd000022.shtml " If you use an animation editing tool such as Pre miere, you can copyright the output material which gives you a different file format from the orig inal data. Editing lingo is production format for the origi nal fil e and consumer format for the distributed, that is, the copyrighted file . These latter files are susceptible of being handled more strictly by DRM mech anisms.
Nvidia Graphics Cards Technical Details http://developer.nvidia.com/page/home
Display and Presentation Software • Chapter 48
ATI Graphics Card Technical Details http://www.ati.cmnldeveloper/index.html
Video and Image Compression lingo Site http://streaming. wisconsin.edu/terms.html
Wikipedia Encyclopedia http://en.wikipedia.org
The OpenGl Committee and Graphics language Reference http://www.opengl.org
General http://developer.apple.com/ http://www.vitecmm.coml http://www.jpeg.org/
ACKNOWLEDGMENTS Thanks are owed to Valentin Guggiana, ETH Zurich, for his help with setting up benchmarks on systems I do not comprehend; to Dr. Chris Pudney of Focal Technologies for his recommendation
845
and advice on software tools and standards; to Dr. Urs Moser and Felix Gattiker of Sulzer Innotec for providing image material, image-processing software, and numerous hardware platforms to test our assumptions; to Dr. Guy Cox, University of Sydney, for advice on the chapter and providing material from his own chapter; and, finally, to Dr. James Pawley, University of Wisconsin, as the concept and the structure of the chapter and most of its final outcome were conceived by him.
REFERENCES Castleman, K.R., 1995, Digital Image Processing, Prentice Hall, Englewood Cliffs, New Jersey, pp, 247-248. Bracewell, R., 1995, Two Dimensional Imaging, Prentice Hall, Englewood Cliffs, New Jersey. Ebrahimi, T., and Pereira, F., 2002, The MPEG-4 Book, Prentice Hall, Englewood Cliffs, New Jersey. Nikolaidis, N., and Pitas, I., 2000, 3-D Image Processing Algorithms, John Wiley and Sons, Inc., New York, pp. 20-24. Richardson, I.E.G., 2003, H.264 and MPEG-4 Video Compression: Video Coding for Next Generation Multimedia, John Wiley and Sons, Hoboken, New Jersey. Watkinson, J., 200l, MPEG Handbook, Focal Press; Linacre House, Oxford, UK.
Walsh, A.E., and Bourges-Sevenier, M., 2002, MPEG-4 Jump-Start, Prentice Hall, Englewood Cliffs, New Jersey. Woo, M., Neider, J., and Davis, T., 2000, Texture mapping, in OpenGL(R) Programming Guide, Addison Wesley, pp. 351-466.
49
When Light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy Paul Sims, Ralph Albrecht, James B. Pawley, Victoria Centonze, Thomas Deerinck, and Jeff Hardin
INTRODUCTION Early Correlative Microscopy Notwithstanding the many ways, amply documented elsewhere in this book, in which fluorescent light microscopy can elucidate biological structure and function, there are times when the available spatial resolution is just not sufficient to answer the biological question. Consequently, there is a need to follow up the initial light microscope (LM) findings by subsequently viewing the LM specimens in the transmission or scanning electron microscope (TEM or SEM). Correlative microscopy of this type has a long history. In 1973, shortly after LM stains had been developed that identified T- and B-lymphocytes, Wetzel viewed the same exact cells, first in the LM and then in the SEM to show that being a T or a B cell bears no relation to whether the cells appeared to be "rough" or "smooth" when viewed in the SEM (Wetzel et al. , 1973). Somewhat later, Sepsenwol used time-lapse studies in the LM followed by high-voltage electron microscopy (HVEM) and SEM to study the unusual protein on which Ascaris sperm motility is based (Pawley et al., 1986; Sepsenwol et al., 1989; Sepsenwol and Taft, 1990).
Early 4D Microscopy Albrecht and his group tracked the motion of colloidal gold-labeled proteins on the surface of activated platelets, using time-lapse, rectified, differential interference contrast (DIC), LM (Fig. 49.1), before determining the final position of the gold particles using both low-voltage SEM (LVSEM; Figs. 49.2, 49.3, and 49.4; Pawley, 1990, 1992) and HVEM (Figs. 49.5 and 49.6; Albrecht et at., 1989, 1992; Loftus et al., 1984). This pioneering work did much to elucidate the mechanism of clot formation. It was possible in part because the platelet is small enough to be viewed in the HVEM as a critical-poi nt-dried (CPD) whole mount and in part because the colloidal gold used to label the surface receptors of interest could be seen in the LM, the SEM and the HVEM. The LM allowed one to watch the movement of the Au-labeled, fibrinogen receptors; the LVSEM at 1 and 5 kV allowed one to see how these markers were bound to the surface (Pawley and Albrecht, 1988); and stereo views from the HVEM allowed one to correlate these surface changes with changes in the location and orientation of the cytoskeleton. The details of these early studies are found in the captions of Figures 49.1 through 49.9.
More recently, this group extended the technique by adding an additional, sensitive charge-coupled device (CCD) camera and an ingenious system of dichroic filters to permit them to image living cells using ultraviolet (UV) fluorescence at the same time that they were being viewed using rectified DIe. This has allowed them to follow the motion of the 20 nm gold particles in the DIC image while monitoring the intracellular Ca++ concentration using fura-2. As shown in the previous images, the binding of fibrinogen to the receptor (and associated receptor cross-linking) triggers a centripetal movement (edge to center) of the receptor-ligand complexes over the platelet surface. The new instrumentation has allowed them to determine that the movement of the receptors across the platelet surface coincides with a Ca++ transient (Fig. 49.10) and to do this on a platelet that they can view subsequently by LVSEM.
CORRELATIVE LIGHT MICROSCOPE/ ELECTRON MICROSCOPE TODAY Light Microscope and Electron Microscope Have Different Requirements Because the structural details of electron microscope (EM) samples must be preserved in much greater detail and because almost all EM specimens must be viewed in vacuo, procedures for preparing them differ markedly from those used in the LM. Correlative studies can be segregated in several ways. The first way is based on the order in which observations are made: 1. The LM observations are made first, on fixed or living cells, and these are then prepared for EM. 1 2. The preparation is fixed and stained before LM observation. 3. Cells incorporating fluorescent markers are prepared for thinsection TEM, but these are viewed in the LM just before being viewed in the EM.
I
Because the act of observing a biological spec imen in an EM subjects it to a flux of radiation so high that virtually all the organic molecules present (such as dyes or stains) are irre trievably damaged, there is seldom much point in viewing a specimen in an LM after it has been viewed in an EM. The exception to thi s rule is the quantum dot, a fluorescent label that, as di scussed below, is not destroyed by being observed in the TEM.
Paul Sims, Ralph Albrecht, james B. Pawley, and j eff Hardin. University of Wisco nsin, Madison, Wisco nsin 53706 Victoria Centonze • University of Texas Health Science Center, San Antonio, Texas 78229 Thomas D eerinck • University of California, La jo lla, California 92093
846
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley, Springer Science+Business Media, LLC, New York, 2006.
When light Microscope Resolution Is Not Enough: Correlational light Microscopy and Electron Microscopy • Chapter 49
847
FIGURE 49.1. Time-series of ole images showing initial binding of Au-conjugated fibrinogen (black) to the platelet surface membrane over the subjacent peripheral web and outer-filamentous zone of the cytoskeleton of a fully spread, substrate-adherent platelet. This label binds to the integrin receptor for fibrinogen on platelets. Over several minutes, bound labels are transported over the surface, towards the platelet center coming to rest, still on the membrane surface, but now overlying the inner filamentous zone. The platelet is then fixed, stained with osmium and uranyl-actetate, and dried by the critical-point procedure for subsequent LVSEM and HVEM. Bar = 11..un.
FIGURE 49.2. Surface image of the same uncoated platelet as Figure 49.1 viewed in LM at I kV accelerating voltage in a modified Hitachi S-900 SEM. It shows the platelet surface and labels in detail. What is actually seen is the platelet surface and the fibrinogen-covered individual gold particles. Bar = I J,lm.
FIGURE 49.3. SEM at SkY, still in the SE mode. The increased beam penetration clearly demonstrates the location of the gold particles, bright spots, relative to stained internal cytoskeletal structures. However, surface structure is less apparent. Bar = I J,lm.
848
Chapter 49 • P. Sims et al.
FIGURE 49.4. SEM at 20kV and somewhat higher magnification. This demonstrates the relationship of the labels on the membrane surface to the underlyi ng cytoskeletal organization. Bar = 0.25 11m.
FIGURE 49.6. A higher magnification HVEM of the same area of the platelet as seen in Figure 49.4. P, peripheral web; OF, outer filamentous zone; IF, inner filamentous zone; G, granulomere; mt, microtubules; dark arrowheads, margin of inner filamentous zone; white arrowheads in Figure 49.3 point to labels trapped under the platelet. Bar = 0.25 11m.
FIGURE 49.5. HVEM stereo-pair whole mount of the same platelet, which offers a clear view of the platelet cytoskeleton. The gold labels, black spots, are clearly seen in relationship to the subjacent platelet cytoskeleton. Bar = 111m.
FIGURE 49.7. DIe imaging tracking of movement of individual label particles across the surface of a fully spread platelet.
FIGURE 49.8. The same platelet, following fixation, staining, and dehydration, is seen via LVSEM at 1.5kV accelerating voltage. The position of the individual tracked particles at the time of fixation relative to the platelet surface can be seen.
FIGURE 49.9. The same platelet, HVEM stereo-pair, demonstrating the position of the particle labels relative to internal structure. P, peripheral web; OF, outer filamentous zone; IF, inner filamentous zone; G, granulomere zone; m, micro filament bundles. Arrows point to individual labels also seen in Figure 49.8, asterisk indicates particle tracked in Figure 49.7 and seen in Figures 49.8 and 49.9. Labels fixed in transit generally still appear over the outer filamentous zone while labels that have completed their movement are generally seen over the inner filamentous zone. Bar = 1.0 flm.
850
Chapter 49 • P. Sims et al.
Another way of dividing correlative methods is between those using the TEM and those using the SEM. A final criterion that might be used to characterize correlative LMIEM studies is whether it is essential to view exactly the same cell using all methods or if it is sufficient to use the EM image merely to find out the general type of structures being labeled in the LM preparation. The following sections will include a brief review of recent studies illustrating all of these approaches.
Finding the Same Cell Structure in Two Different Types of Microscope: light Microscope/Scanning Electron Microscope
FIGURE 49.10. (Al and (8) show a plate let imaged simultaneously via DIC and UV fluorescence. In (A), the diffraction images of colloidal goldfibrinogen particles can be followed over time as they move across the platelet surface. (8) Fura-2 340nrnl380nm ratio images that provide a measure of the free internal [Ca++] as the process proceeds (blue represents resting [Ca++] levels while red and white show increasin g [Ca++]). Although the initial fibrinogenbinding produces no increase in free [Ca++] (left panels), once the movement of the fibrinogen-receptor complexes is initiated, [Ca H ] is seen to increase (middle panels) until the movement is complete (right panels). (C) An LVSEM image of the same platelet shows the final position of the gold-fibrinogenreceptor complexes (bright spots) relati ve to platelet surface structure.
Control
10ms
Because both the LM and the SEM permit one to view quite large specimens, it is often easier to find the exact cell previously viewed in an LM using an SEM rather than a TEM. Centonze and coworkers used this combination to investigate the effect of fluorescence recovery after photobleaching (FRAP) on the molecular structure of the labeled structures. In FRAP, a structure or volume that has been labeled with fluorescent dye in a living cell is intentionally bleached and one then measures the rate at which fluorescence returns to the bleached area (see Chapters 5, 8, 9, 17, and 45 , this volume). Alternatively, one may wish to derive inferences from the motion of the bleached area. One well-studied example is the motion of a band bleached across a mitotic apparatus made of fluorescent tubulin. In this case, the band did not move even though mitosis proceeded (Gorbsky et al., 1987). Centonze used the SEM to investigate this further. To do this, gold was evaporated through an EM finder grid onto the surface of a glass coverslip to create a fiduciary pattern that could be seen in both the LM and the SEM. These covers lips were then mounted into a hole in the bottom of a plastic Petri dish using silicone grease and isolated, fluorescent microtubules were allowed to adhere to the glass. After obtaining reference transmission and fluorescence LM images, some of the microtubules not located over the gold were bleached, using 546 nm light from an argon-ion laser, for a measured period of time at a known power level. The preparation was then fixed, critical-point dried, coated with ion-beam-sputtered Pt, and viewed in a high-resolution, lowvoltage SEM at 1.5 kY. Figure 49.11 shows the results. While a 10 ms pulse caused little visible damage, 30ms caused total destruction. Fortunately, as is shown in Figure 49.12(A), it is possible to produce significant bleaching using a pulse only I ms in duration , a period thought unlikely to cause severe structural damage. It should also be noted that dissolution or disruption of microtubules could also be induced by repeated illumination under conditions used for normal fluorescence observations. Figure 49.13 shows this system applied to bleaching microtubules in a living cell grown on a marked coverslip. In this case,
30ms
50ms
FIGURE 49.11. Low-voltage scanning e lectron micrographs of single fluorescent microtubules that had earlier been subjected to irradiation by a 546 nm bleaching beam for the times listed. Any exposure above lOms destroyed the microtubule.
When light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy • Chapter 49
851
fiGURE 49.12. Fluorescence micrographs showing single microtubules, composed of in vitro polymerized microtubules made from rhodamine-conjugated tubulin , across which a bar has been bleached by exposure to a 546nm laser beam for the noted times. (A) shows that substantial bleaching is produced by a I ms exposure, a level lOx lower (8) than that shown to produce structural damage in Figure 49.11. The plots below each image show the intensity along a line down the center of the microtubule (upper trace) and along the straight white line, showing the background signal (lower trace).
FIGURE 49.13. (A) Phase-contrast image of a spread cell containing microtubles made out of rhodamine-conjugate tubulin. (8) Fluorescence image of the same field. (e) Bleached bar and (D) fluorescence image of the same areas of the specimen after it has been prepared for SEM. (E) Tiled montage of low-voltage SEM images in which one can see that all the microtubules visible in thc LM images, even those in the bleached zone (brackets), remain physically intact after the bleaching event.
852
Chapter 49 • P. Sims et al.
fibroblast cells were grown on gold-sputtered, locator coverslips and injected with fluorescent derivatized tubulin . After the injected tubulin was allowed to incorporate into the microtubule cytoskeleton for at least 1 h, the injected cells were relocated using phasecontrast microscopy. Fluorescent microtubules were imaged both before and after photobleaching a discrete region [Fig. 49.13(D)). Because the fixation protocol involved lysing the cells in Triton X-lOO detergent before fixation in 2% glutaraldehyde in a cyto skeletal-stabilizing buffer containing, Pipes, Hepes, EGTA and Mg++ (PHEM) buffer, the montage of SEM images shows the cytoskeleton, rather than the cell membrane. In this case, bleach levels were low enough to create a photobleach mark without disrupting the microtubules so that all of the microtubules visible in the fluorescence image are visible in the SEM montage. LVSEM images of specimens bleached with higher power, or for a longer time, showed that the affected microtubules could be completely ablated. The ablation was documented in three ways: (1) the fluorescence recovery rates were observed to be different in the bleached region; (2) treatment with fluorescent anti-tubulin produced no staining in the photobleached region because the structures were destroyed; and (3) SEM images clearly show no microtubule structures in the photobleached region (data not shown) . This correlational study should give pause to anyone convinced that the ONLY possible effect of photobleaching is that the dye in the irradiated volume becomes non-fluorescent. More insight into photodamage mechanisms is provided in Chapters 38 and 39.
successful methods for doing this, called fluorescence photo-oxidation, was developed by Maranto (1982) and later extended to immunolabeling and in situ hydridization by Ellisman (Deerinck et al., 1994). It relies on the fact that, when certain fluorescent dyes, such as eosin, are excited in the presence of diaminobenzidine (DAB), the reactive oxygen produced by the triplet-excited fluorescent compound causes the DAB to form a deposit very close to the reaction site. This deposit can then be stained with considerable specificity with osmium tetroxide (Fig. 49. 14), rendering it visible by both transmitted light and electron microscopy. Because the reaction is limited to only the region near the excited fluorescent dye, the precise area can be located in the epoxyembedded specimen by light microscopy and then prepared for thin section electron microscopy. More recently, this group has adapted the tetracysteine genetic marking technique to produce the reactive oxygen needed to deposit the DAB in labeled cells (Griffin et al., 1998; see Chapter 16, this volume). In this technique, a tetracysteine tag sequence is
Finding the Same Cell Structure in Two Different Types of Microscope: Light Microscope/Transmission Electron Microscope Although correlative microscopy uses both light and electron microscopy to examine the same sample, each microscopic method produces contrast in a very different way. When viewing living biological specimens in the LM, one commonly uses phase, DIC, or more recently, fluorescence imaging. However, in the TEM, none of these factors produces significant contrast, and, in addition, the TEM specimen must be considerably thinner than most cells. To produce enough mass-thickness contrast to make biological structures visible in the TEM, one must somehow decorate them with heavy metal stains, such as uranyl-acetate and lead compounds. As a result, LMffEM correlative methods generally depend on the use of some technique that will deposit heavy metals at or near the site of the fluorescent dye. Possibilities include double labeling with both fluorescent and Au-conjugated antibodies and using light captured by the fluorescent dye to initiate a chemical reaction that later results in the deposition of a heavy metal. Finally, quantum dots are particularly useful because they are both fluorescent and directly visible in the TEM (Niesman et al., 2004). Correlative LMffEM techniques are also complicated by the fact that the visible area of a TEM grid is usually less than 2mm in diameter and quite a lot of this area is obscured by grid bars. Moreover, it is in general hard to keep track of changes in the orientation of specific structures in the LM specimen as the sample passes the many steps needed to prepare it for thin-section TEM. This can make it very difficult to find exactly the same feature using both methods, especially if the handedness of the image has been changed by the section having been mounted upside-down .
Making LM Labels Visible in the Transmission Electron Microscope There are a few techniques that allow the same stained sample to be used for both light and electron microscopy. One of the most
FIGURE 49.14. (Al Using immunofluorescently stained bovine aortic epithelial cells for both light and electron microscopy, cultured cells were labeled with an antibody to beta-tubulin foll owed by a secondary antibody-eosin conjugate. Following confocal imaging of the eosin fluorescence, the specimen is intensely illuminated in the presence of diaminobenzidine. The resulting reactive oxygen creates a reaction product that can be subsequently vi sualized by electron microscopy (B). (Image kindly provided by the laboratory of Mark Ellisman at the National Center for Microscopy and Imaging Research, University of California, San Diego.)
When Light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy • Chapter 49
FIGURE 49.15. Using a genetically encoded tetracysteine tag for labeling proteins for light and electron microscopy, cultured cells expressing a recombinant version of the major gap junction protein Cx43 that contains a small tetracysteine tag. Living cells are stained with the biarsenical red fluorescent compound ReAsH (A). Following imaging, the cells are fixed and the specimen is intensely illuminated in the presence of diaminobenzidine. The resulting reactive oxygen creates a reaction product that can be subsequently visualized by electron microscopy (B). (C) is a higher magnification view of (B). (Images kindly provided by the laboratory of Mark Ellisman at the National Center for Microscopy and Imaging Research, University of California, San Diego.)
introduced into the gene of the protein of interest in such a way that the four cysteines can bind to a bi-arsenical derivative of fluorescein (green fluorescence) or resorufin (red fluorescence). Living cells can be bathed in a solution of these membranepermeant dyes without damage as long as the arsenicals are neutralized with small vicinal dithiols such as 2,3-dimercaptopropanol or 1,2-ethanedithiol. The dye becomes fluorescent only when it bonds specifically to the tetracysteine moiety. Like eosin, the redfluorescent version of this dye (called ReAsH), is capable of producing reactive oxygen and depositing DAB. The result is a genetically encoded marker that can be seen in both LM (by fluorescence) and TEM (by specific staining) (Gaietta et al., 2002; Fig. 49.15). LMIEM correlation can be obtained by embedding the cells still attached to the coverslip and making high- and low-magnification images of them with a confocal microscope before the coverslip is removed. These fluorescent images can them be correlated with low-magnification TEM images of serial-section ribbons made starting from what had been the plastic/coverglass interface. This same group has also had success labeling structures in such a way that the label can be seen in both LM and EM using quantum dots. These nanocrystals are both fluorescent (Nisman et al., 2004) and directly visible in thin-section TEM (Giepmans, 2005; Fig. 49.16). One of the major advantages of using quantumdot fluorophores is that different colors of quantum dots can be
FIGURE 49.16. Using quantum dots as labels for light and electron microscopy. cultured cells were labeled with an antibody to beta-tubulin followed by a secondary antibody-quantum dot 655 conjugate (cell nuclei were counterstained with Hoechst 33342). Following confocal imaging (A), the specimen is prepared for electron microscopy (B). The crystalline core of the quantum dots are readily visible by electron microscopy. (Image kindly provided by the laboratory of Mark Ellisman at the National Center for Microscopy and Imaging Research. University of California. San Diego.)
853
854
Chapter 49 • P. Sims et al.
discriminated by both light and electron microscopy (via size and shape differences), making double or even triple labeling possible. Furthermore, they are not so readily destroyed by electron irradiation as are organic dyes. As a result, it is possible to obtain fluorescent images from the sections after they have been viewed in the TEM. On the other hand, the fluorescence of quantum dots is destroyed if heavy metal stains are used to produce contrast. Both of these techniques allow one to test and optimize antibody-labeling parameters first by LM before continuing the more laborious processing for electron microscopy.
Marrying Fluorescence with TEM Replicas to Analyze the Cytoskeleton Phase-contrast and fluorescence microscopy have been used to visualize actin dynamics at the leading edge of living cells in which various cytoskeletal proteins have been labeled with GFP. Svitkina and colleagues followed up such studies by removing the membranes with an extracting buffer, and then shadowing the preparation with Pt after it had been critical-point dried to visualize actin filament structure in the TEM. This led them to propose a filopodial initiation model (Svitkina et at., 2003). Such studies provide additional insights into the detailed topology of cytoskeletal assemblies in cells. Because processing for EM is performed on cells with a known history based on LM and potential shrinkage artifacts are carefully monitored during fixation, this approach provides greater confidence that the more highly resolved EM images directly reflect events imaged using LM.
FluoroNanoGold for Cryosections to be Viewed by LM, then TEM In addition to its utility in analyzing living, migrating cells, viewing specimens first using LM followed by TEM also has other uses. In some cases (e.g., immunostaining), it is far easier to label the structures of interest in fixed and sectioned specimens using LM techniques. Knowing that a given specimen contains the features of interest provides more confidence that subsequent processing for TEM will yield a specimen worth analyzing in detail. Although it has long been clear that, compared to chemical fixation, cryopreparative techniques both arrest cellular processes faster (milliseconds vs. seconds) and preserve biological structure down to the molecular level better, the complexity of the equipment and procedures needed to freeze even modest-sized specimens without creating ice-crystal artifacts has delayed its widespread use in light microcopy (Biel et at., 2003). However, because cryotechniqes are unsurpassed when it comes to preserving antigenicity, in 1998 Takizawa and colleagues stained thin, cryosections with FluoroNanoGold (FNG) antibodies for correlative LM and TEM on the same thin section (Takizawa et at., 1998). Ultrathin cryosections were cut from a frozen suspension of fixed cells embedded in gelatin/sucrose. The sections were picked up on finder grids, stained with FNG, and viewed first by fluorescence LM and then by TEM. FluoroNanoGold labels consist of ondeca gold conjugated with a fluorescent dye. As the ondeca gold contains only 11 gold atoms, it often penetrates better than the large colloidal-Au labels. On the other hand, it is difficult to see even in TEM unless it has been decorated by enhancement using silver or gold salts. Several examples of this technique have been reported. To determine the diameter of transcription sites in the nuclei of HeLa cell, Pombo and colleagues made cryosections 100 to 200nm thick, stained them with fluorescent antibodies for imaging in the
LSCM, and then re-embedded them in Epon to image the same structures by TEM (Pombo et at., 1999). In 2001, Robinson reviewed the techniques then used for correlative LM and TEM in cryosections and the advantages of viewing thin sections by LM (Robinson et at., 2001). Any selective staining protocol must face the problem of how to see structure not stained with the selective agent. One solution is to use an energy-filtering TEM (EFTEM) to maximize the contrast between unstained protein and embedding resin. Ren and coworkers used this technique to image Quetol-embedded cells in which promyelocytic leukemia (PML)-bodies had been antibodylabeled with Cy-3. They found cyanine and Alexa dyes to be stable in this resin (Ren et al., 2003) and also used quantum dots for correlative fluorescence and EFTEM (Nisman et al., 2004).
Green Fluorescent Protein Methods The introduction of the genetically based marker, green fluorescent protein (GFP), has revolutionized the study of cell biology (see Chapter 16, this volume). In 2003, Luby-Phelps showed that GFP fluorescence, which was destroyed by most TEM embedding techniques, survived embedding in LR White (Luby-Phelps et al., 2003). Adjacent thin and 111m sections were cut from preparations of zebrafish eyes and viewed using the TEM and LM, respectively. Recently, we have extended this approach to thin sections. Specifically, we were interested in the relationship between two junction proteins, HMP-l [Caenorhabditis etegans (Ce) alphacatenin] and DLG-l (Cediscs large). Confocal fluorescence LM had shown that DLG localizes near HMP-l. To find out if these molecules colocalized at the EM level, we imaged HMP-l::GFP in a thin section by LSCM and immuno-gold-labeled DLG-l in the same section. EM localization of DLG-I gold particles with HMP1::GFP expression is consistent with LM data and confirms the gold labeling parameters needed to label DLG. Thin sections, adhered to EM finder grids and floating on coverslips, were first imaged in an LSCM using both fluorescence and backscattered light (BSL), to reveal the section surface [Fig. 49.17(A)], and fluorescence, to show HMP-l::GFP [Fig. 49.17(B)]. The combination of BSL (displayed in red) and GFP, in green [Fig. 49. 17(C)], shows the location of GFP within each embryo and is useful for locating a specific embryo in the TEM [Fig. 49.17(D)]. At higher magnification, the boxed area shows the 20 nm gold labeling due to DLG-I [Fig. 49.17(F)]. We have also imaged AJM-l::GFP (a novel Ce junction molecule) in the pharynx with 20nm-gold anti-GFP in LR-Gold thin sections after chemical fixation [Fig. 49.18(A)]. Although colloidal gold particles have been shown to quench the fluorescence of molecules in close proximity to them (Kandela et at., 2004), we have found that rhodamine-linked secondary antibodies, not actually conjugated to the gold-label [Fig. 49.18(B)] still fluoresce [Fig. 49.18(C)] (Sims and Hardin, 2004). Like the ReAsH system mentioned above, this procedure also allows one to visualize a genetic marker in a thin section, using both fluorescence LM and a specific stain visible in TEM.
Using Phalloidin as a Correlative Marker As GFP fluorescence had been shown to survive dehydration and embedding in methacrylate resin, we were curious if rhodamine-phalloidin might also survive this process. C. elegans embryos that had been bleached to remove the eggshell, were fixed and incubated in buffer containing saponin and rhodaminephalloidin. After low-temperature embedding in LR-Gold, 100nm sections were cut and imaged by confocal laser scanning
When Light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy • Chapter 49
855
FIGURE 49.17. (A) BSL image of a thin section on an EM finder grid. (B) HMP-l::GFP from the same thin section. (C) A red-green merge of BSL and GFP images which is useful to orient fluorescence to specific embryos. (D) TEM of the same area imaged in (A-C) with the same embryo highlighted by the boxed area. (E) Higher magnification TEM montage of the comma-staged embryo also observed in all previous figures. (F) Boxed area in (E) with 20nm gold labeling OLG-I, a junctional protein that colocalizcs with AJM-I.
FIGURE 49.18. (A) Postembedding AJM-I::GFP observed at epithelial junctions in thin sections (100 nm) of fixed C. elegam' larvae. A brief fixation followed by freezing permits infiltration and embedding of larva in LRGold (methacrylate) embedding resin. Both GFP fluorescence and antigenicity endure the rigors of partial dehydration and low temp embedding. GFP fluorescence can still be detected in thin sections cut from this original block over I year after embedment. (B) Rhodamine goat anti-rabbit (secondary antibody) bound to a rabbit anti-GFP antibody. If one couples both Au and a fluorophore to the same antibody, the Au quenches the fluorescence. (C) 20nm gold particles conjugated to rhodamine goat anti-rabbit antibody label AJM-I ::GFP at junctions. Mixed conjugates can be useful to confirm specific labeling by LM, but results in reduced gold particle labeling seen by electron microscopy. Bar = I ~m.
856
Chapter 49 • P. Sims et al.
FIGURE 49.19. TEM cross-section of a C. elegans larva, en bloc-stained with rhodamine-phalloidin and viewed simultaneously in backscattered (A) and fluorescent (8) light on a confocal microscope. (C) Overlay of (A) and (8). (D) A low-magnification TEM (bar = 2~m). The two boxed areas are seen at higher magnification in (E), a close-up of the intestine (bar = 1 ~m) and (F) a hypodermal adherens junction (bar = 0.5 ~m). Phalloidin staining is seen decorating the intestinal microvilli and the four muscle quadrants on the sides of the worm.
microscopy (CLSM). A BSL image [Fig. 49.19(A)] and a fluorescent image [Fig. 49.19(B)] were merged together [Fig. 49.19(C)] showing actin-rich, phalloidin-stained muscle quadrants and intestinal microvilli in red. The contrast from the rhodamine signal was increased and overlayed on a low-magnification TEM image [Fig. 49.19(D)]. The boxed regions in Figure 49.19(D) are shown at higher magnification in Figure 49.19(E), covering the intestinal microvilli, and Figure 49.19(F), showing an adherens junction. Overall, this study demonstrates that, at least for proteins as concentrated as actin in muscle and microvilli, one can assess which structures are labeled by analyzing only the fluorescent LM images.
Cryo-Immobilization Followed by Post-Embedding Confocal laser Scanning Microscopy on Thin Sections Chemical fixation is known to be slow and to introduce a wide variety of artifacts. Unfortunately, the only alternative is cryopreservation, a technique that depends for its success on freezing the important part of the specimen without forming detectable icecrystal artifacts. At atmospheric pressure, ice crystals can only be avoided by using cryoprotectants, such as sucrose and polyethyl-
ene glycol, or by freezing the tissue very fast indeed (> 105 Kls; Studer et al., 1989; McDonald et at., 1993). As using a cryoprotectant requires at least some prefixation, it is not a great improvement on other types of chemical fixation . However, fast freezing is also complex; largely because of the high heat -of-fusion of water and the low thermal conductivity of ice, even cryogens as efficient as liquid-He-cooled copper are unable to extract the heat from the specimen fast enough to prevent ice-crystal artifacts from more than the outer 10 to 15).tm of the specimen. Objects the size of C. elegans (60).tm diameter) can only be successfully frozen using a high-pressure freezer (Studer et al. 1989; McDonald et at., 1993). In this device, pressurized LN2 is used to cool a specimen about 2 mm in diameter and 200).tm thick from ambient temperature to 77 K in about 20ms, under a transient pressure spike of about 2000 bar. Under these conditions, water freezes at a lower temperature (-253 K rather than 273 K), and ice crystals propagate more slowly for various reasons, including the fact that the viscosity of the water increases with ambient pressure. The result is that a relatively high fraction of such specimens are frozen without noticeable artifact. For TEM observation, freezing is followed by either freeze fracture/metal shadowing/carbon replication, or freeze substitution with ethanol or acetone followed by "normal" epoxy embedment in Lowicryl HM-20. With the latter approach, the plastic-
When Light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy • Chapter 49
857
FIGURE 49.20. (A) TEM longitudin al section of an adult C. eiegans. (B) The same section viewed by fluorescent confocal microscopy showing that the flu orescence of AJM-I::GFP has not been eliminated by embedding in LR-Gold. Boxed areas in (A) and (B) are shown at higher magnification in (C) and (D ). where the fluorescence image has been overlaid in green over the TEM image. As the section was also labeled with anti-GFP-gold, these insets show good agreement between the distribution of these two markers in pharynx (e) and around the gut (D). Bar = (A, B) 10 J..lm; (C) 0.5 J..lm; (D) I J..lm.
embedded specimen can be viewed directly using CLSM, and thi s image can be used to choose the best area of the block to be sectioned for TEM (Biel et ai., 2003) .
Cryopreparation of C. e/egans Because the eggshell and the vitelline membrane reduce penetration by chemical fixatives, C. eiegans embryos are difficult to fix chemically. As a result, we have adapted the freeze-substitution protocol to prevent it from damaging frozen GFP specimens. Although Ward observed that GFP fluorescence was completely extinguished if specimens were placed in absolute ethanol (Ward, 1998), Walther and Ziegler (2002) found that adding 1% to 5% water to the freeze-substitution medium increased the visibility of membranes, and we reasoned that a similar approach might preserve GFP structure. We found that freeze substitution in 95 % ethanol/5 % water, followed by low-temperature embedding in LR Gold, preserves AJM-l::GFP in thin sections (Fig. 49.20). Figure 49.20(B) shows a thin section of a specimen containing AJM-l: :GFP and also stained with Au-conjugated anti-GFP,
imaged with a disk-scanning confocal microscope. In Figure 49.20(A), an image of GFP fluorescence is overlaid on a lowmagnification TEM image. Higher magnification TEMs of the two areas highlighted in Figure 49.20(B) are shown in Figure 49.20(C,D). The green overlays roughly colocalize with the 20nm gold label. This technique allows one to combine the ability of fluorescence LM to search large areas of the section rapidly to identify rare stained structures and then find and view these rare structures in the TEM. The result confirms that this HPF protocol preserves GFP fluorescence and specimen immunoreactivity and also allows one to correlate AJM-J::GFP with specific cellular structures visible only at the ultrastructural level. Another application of GFP and cryopreservation involves identification of wild-type and mutant embryos prior to cryopreservation and processing with immunogold . Although C. elegans embryos containing mutant copies of the ajm-l gene arrest soon after the 2-fold stage, an average 60% of them can be rescued to viability if the hermaphrodite (parent) is microinjected with an extrachromosomal array that includes a functional copy of the
858
Chapter 49 • P. Sims et al.
FIGURE 49.21. (A) A transmitted light image of living embryos embedded in agar. (B) A brightest-pixel projection (Image]) of AIM-I ::GFP expression in the same embryos in (A). Transmitted light and fluorescence images were obtained simultaneously on a Bio-Rad 1024 confocal using 488nm excitation. Embryos older than comma stage which do not express GFP, lack the ajm-I gene. (C) Overlay of the fluorescent projection over the transmitted light image. Red-shaded area is a 2-fold embryo which lacks ajm-l. (D) A TEM image of the same embryos after HPF and embedding in Epon. Same area red-highlighted in (C) and (D) is shown in (El, where arrows point to epithelial cell membrane separations associated with the loss of ajm-l.
ajm-I gene, in this case ajm-I fused to the GFP-coding region. The GFP label allows the observer to determine which of the living embryos carry the ajm- I gene and which do not. Although one could make a similar discrimination by applying Au-conjugated, anti-GFP antibodies to the thin sections, optimal fixation and staining could be compromised by the need to preserve the antigenicity of the GFP. Correlating the LM fluorescence image of a living embryo with the TEM image of a section from this embryo avoids this problem, particularly if rescued and non-rescued embyos are next to one another on the same TEM grid. Any observable differences are more likely to be real when one knows that the high-pressure freezing, fixation, and staining are identical. Figure 49.21(A) shows a transmitted light image of a group of 10 embryos, 6 of which have been rescued (Koppen et al., 200 I; Simske et al., 2003). Although the non-rescued embryos look normal in transmitted light, they are clearly marked by the absence of GFP fluorescence [Fig. 49.21(B,C)]. Following high-pressure freezing, freeze substitution, embedding in Epon, and sectioning,
the same group of embryos can be visualized at low magnification by TEM [Fig. 49.21(D)]. The small red-shaded boxes in Figure 49.21(C,D) indicate the area shown at higher magnification in Figure 49.21 (E). Arrows point to separations at epithelial cell junctions believed to be associated with the loss of ajm- I .
Tiled Montage Transmission Electron Microscope Images Aid Correlation Many modem TEMs incorporate both electronic image recording and motor-driven stage motion. This combination of features greatly facilitates LMffEM correlational studies by making it much easier to obtain high-resolution images over a wide area of the specimen. Figure 49.22 shows such a tiled montage of a specimen prepared in the same way as that shown in Figure 49.21. This image covers an area 30 x 50 flm in size and consists of 36 images recorded at an original magnification of 3.5x. The green overlay represents AJM-I ::GFP fluorescence recorded on a Bio-Rad 1024 confocal microscope.
When Light Microscope Resolution Is Not Enough: Correlational Light Microscopy and Electron Microscopy • Chapter 49
859
FIGURE 49.22. A tiled montage TEM image of a specimen prepared in the same way as that shown in Figure 49. 11 . The original montage covered an area of the specimen 50 x 32/.lm in size and consisted of 36 images recorded at 80kY using an original magnification of 3400x on a Phillips 120 using a Soft Imaging Systems. Keenview, CCD camera coupled to the phosphor by a 3.4x fiber-optic taper. The green overlay represents AJM-l ::GFP fl uorescence recorded on a BioRad 1024 confocal microscope.
860
Chapter 49 • P. Sims et al.
CONCLUSION Although during the 1950s and 1960s light microscopy languished beneath the high-resolution shadow of the electron microscope, starting in the early 1970s, it flowered as a variety of new technical improvements were introduced: video-enhanced DIC, videointensified fluorescence, improved CCD cameras, and myriad new fluorescent probes. Suddenly, the fact that one could observe function in wholly new ways seemed to more than offset the ability to image the internal features of intracellular organelles. These LM developments were soon followed in the 1980s and 1990s by the widespread use of a variety of confocal microscopes, and more recently by the advent of GFP and its many relations. This entire book is a testament to the fact that live-cell fluorescent light microscopy is blossoming as never before. On the other hand, we lose nothing by admitting that much of the business of cells occurs at a size scale that is much smaller than that which can be imaged with the light microscope. When this occurs, it is useful to remember that help is at hand; given the effort needed to boost LM spatial resolution by a factor of 2, it is salutary to acknowledge how much more clearly one can see when it is increased by a factor of 40.
ACKNOWLEDGMENTS This work was supported by grants from the National Institutes of Health: GM-058038 (Hardin), RR-04050 (Ellisman), and NIGMS63001 (Albrecht); and from the National Science Foundation IBN-0712803 (Hardin).
REFERENCES Albrecht, R.M .. Goodman, S.L. , and Simmons, S.R., 1989, Distribution and movement of membrane-associated platelet glycoproteins: Use of colloidal gold with correlative video-enhanced light microscopy, low-voltage high-resolution scanning electron microscopy, and high-voltage transmission electron microscopy, Am. J. Anat. 185:149- 164. Albrecht, R.M., Olorundare, O.E .. Simmons, S.R., Loftus, J.e.. and Mosher, D.F., 1992, Use of correlative microscopy with colloidal gold labeling to demonstrate platelet receptor distribution and movement, Methods En zymol. 2 J 5:456-479. Biel, S.S. , Kawaschinski, K., Wittem, K.P., Hintze, U. , and Wepf, R. , 2003, From tissue to cellular ultrastructure: Closing the gap between micro- and nanostructural imaging, J. Microsc. 212:9 1-99. Deerinck, TJ. , Martone, M.E. , Lev-Ram, v., Green, D.P., Tsien, R.Y. , Spector, D.L.. Huang. S.• and Ellisman, M.H., 1994, Fluorescence photooxidation with eosin: A method for high resolution immunolocalization and in situ hybridization detection for light and electron microscopy, J. Cell Bioi. 126:901-910. Gaietta, G., Deerinck, TJ., Adams, S.R., Bouwer, J., Tour, 0., Laird. D.W. , Sosinsky, G.E.. Tsien, R.Y., and Ellisman, M.H., 2002, Multicolor and electron microscopic imaging of connexin trafficking, Science 296: 503- 507. Giepmans, B.N., Deerinck, TJ., Smarr, B.L., Jones , Y.Z., and Ellisman, M.H., 2005, Correlated light and electron microscopic imaging of multiple endogenous proteins using Quantum dots, Nature Methods 2(10):743749. Gorbsky G.J., Sammak, P.J., and Borisy. G.G .. 1987, Chromosomes move poleward in anaphase along stationary microtubules that coordinately disassemble from their kinetochore ends, J. Cell Bioi. 104:9-18. Griffin, B.A., Adams. S.R., and Tsien, R. Y., 1998. Specific covalent labeling of recombinant protein molecules inside live cells, Science 28 J :269- 272.
Kandela, I.K., Bleher, R., and Albrecht, R.M., 2004, Correlative labeling studies in light and electron microscopy, Microsc. Microanal. 10(Suppl, 2): 1212-1213. Koppen, M., Simske, J.S. , Sims, P.A., Firestein, B.L., Hall, D.H., Radice, A.D., Rongo. e., and Hardin, JD., 2001, Cooperative regulation of AIM-I controls junctional integrity in Caenorhabditis elegans epithelia, Nat. Cell Bioi. 3:983-991. Loftus, l.e., Choate, 1. , and Albrecht, R.M., 1984, Platelet activation and cytoskeletal reorganization: High voltage electron microscopic examination of intact and Triton-extracted whole mounts. J. Cell Bioi. 98:2019-2025. Luby-Phelps, K., Ning, G., Fogerty, 1., and Besharse, l.e. , 2003, Visualization of identified GFP-expressing cells by light and electron microscopy, J. Histochem. Cytochem. 51:271 - 274. Maranto, A.R.. 1982. Neuronal mapping: A photooxidation reaction makes Lucifer yellow useful for electron microscopy, Science 217:953-955. McDonald, K., and Morphew, M.K., 1993, Improved preservation of ultrastructure in difficult-to-fix organisms by high-pressure freezing and freeze-substitution: 1. Drosophila melanogaster and Strongylocentrotus purpuratus embryos, Microsc. Res. Tech. 24:465-473. Nisman, R., Dellaire, G., Ren, Y. , Li , R. , and Bazett-lones, D.P., 2004, Application of quantum dots as probes for correlative fluorescence, conventional, and energy-filtered transmi ssion electron microscopy, J. Histochem. Cytochem.52:13-18. Pawley, J.B., 1990, Practical aspects of high resolution LVSEM, Scanning 12:247-252. Pawley, J.B., 1992, LVSEM for high resolution topographic and density contrast imaging, Adv. Electron. Electron Phys. 83:203-274. Pawley, 1.B., and Albrecht, R., 1988, Imaging colloidal gold labels in LVSEM, Scanning 10:184-189. Pawley, 1.B., Sepsenwol, S., and Ris, H., 1986, Four-dimensional microscopy of Ascaris sperm motility, Ann. NY A cad. Sci. 483: 171-180. Pombo, A., Hollinshead, M., and Cook, P.R. , 1999, Bridging the resolution gap: Imaging the same transcription factories in cryosections by light and electron microscopy, J. Histochem. Cytochem. 47:471-480. Ren, Y. , Kruhlak, M.J. , and Bazett-Jones, D.P., 2003, Same serial section correlative light and energy-filtered transmission electron microscopy, J. Histochem. Cytochem. 51 :605--612. Robinson, 1.M., Takizawa, T. , Pombo, A. , and Cook, P.R. , 2001, Correlative fluorescence and electron microscopy on ultrathin cryosections: Bridging the resolution gap, J. Histo chem. Cytochem. 49:803-808. Sepsenwol, S., Ris, H. , and Roberts, T.M., 1989, A unique cytoskeleton associated with crawling in the amoeboid sperm of the nematode, Asca ris suum, J. Cell Bioi. 108:55-66. Sepsenwol, S., and Taft, S.J. , 1990, In vitro induction of crawling in the amoeboid sperm of the nematode parasite, Ascaris suum, Cell Moti!. Cytoskeleton 15:99-110. Sims, P.A., and Hardin, J.D. , 2004, Visualizing green fluorescent protein and fluorescence associated with a gold conjugate in thin sections with correlative confocal and electron microscopy, Microsc. Microanal. 10(Suppl. 2): 156-157. Simske, 1.S. , Koppen , M., Sims, P., Hodgkin, 1., Yonkof, A. , and Hardin, 1., 2003, The cell junction protein VAB-9 regulates adhesion and epidermal morphology in C. elegans [comment], Nat. Cell Biol. 5:619-625. Studer, D., Michel, M., and Muller, M., 1989, High-pressure freezing comes of age, Scanning I I (Suppl. 3):253-269. Svitkina, T.M., Bulanova, E.A., Chaga, O.Y., Vignjevic, D.M., Kojima, S., Vasiliev, 1.M., and Borisy, G.G., 2003, Mechanism of filopodia initiation by reorganization of a dendritic network, J. Cell Bioi. 160:409-421. Takizawa, T., Arai , S., Osumi, M., and Saito, T., 1998, Ultrastructure of human scalp hair shafts as revealed by freeze-substitution fixation , Anat. Rec. 251:406-413. Walther, P., and Ziegler, A., 2002, Freeze substitution of high-pressure frozen samples: The visibility of biological membranes is improved when the substitution medium contains water. J. Microsc. 208:3-10. Ward, W.W., 1998, Green Fluorescent Proteins: Proteins, Applications and Protocols, John Wiley and Sons, New York. Wetzel, B., Erickson, B.W., and Levis, W.R., 1973, The need for positive identification of leukocytes examined by SEM, Part III:535-541.
50
Databases for Two- and Three-Dimensional Microscopical Images in Biology Steffen Lindek, Nicholas
J.
Salmon, and Ernst H.K. Stelzer
INTRODUCTION Modem computerized and automated microscope technology enables researchers, technicians, and quality-assurance personnel to collect large numbers of high-quality images. However, despite the addition of digital storage (see Chapter 32, this volume) and networking, microscopes are still generally used in a stand-alone manner insofar as the output is simply a huge number of image files. There is little attempt at proper cataloguing, and little metadata beyond the filename to identify and describe the contents of each image (Gonzalez-Couto et al., 2001). This is not only a problem for the individual scientist who wants to document his or her work and master the flood of data that accumulates over the years, or for the manager of a microscopy facility, who has to maintain the huge amounts of data generated by different users on different instruments. It also concerns the scientific community on a global scale: in contrast to textual documents such as journal articles, microscopy data is difficult to discover (usually based only on keyword searches in literature databases, rather than on image-specific information). In addition, getting copies of the data usually relies on a personal correspondence with the author, in which he/she is asked to send the data and any relevant information regarding the experiment (Anonymous, 2004). The first steps towards a better archive infrastructure are the networking of instruments so that data can be easily transferred from one point to the other, and the introduction of databases both at the individual microscope level and at the large-scale level to hold the metadata that allows others to manage and use the data. The latter two developments go hand in hand: Having local solutions does not solve the global problem of data access, and having a universal database for microscopy data requires local solutions to develop an efficient data submission procedure. This chapter outlines the developments made in both areas. It discusses the benefits of integrating a database into a microscope to store information about both the instrument and the image data generated by it, and it presents preliminary efforts made to develop a database infrastructure at the global level.
DATA AND METADATA MANAGEMENT IN MICROSCOPES In early confocal microscope systems, information about the machine configuration was contained within a single text file that could be read and written at any time during data acquisition or
processing. Such a fiat file format avoided the need for recompilation of the entire microscope program, but as its length increased, it rapidly became unsuitable for human viewing. Online configuration of a number of system parameters was possible, often by running a setup file containing a subset of the information required to obtain the image. However, the information was intended as an aid to simply getting an image, rather than adjusting the large number of machine variables. Indeed, storage of some component settings was impossible or useless because these were machine settings that did not provide information about what was actually happening to the specimen. Moreover, if a hardware component were replaced, then apparently identical machine settings could result in different image data, for example, because filter selection was implemented by reference to filter wheel positions, rather than to actual filters. Image intensity data was saved as a stream of bytes, and information about a particular image was mostly limited to saving pixel spacing information in an adjacent text file. Thus , reading such data into image processing programs often required some programming effort and, in most cases, manual intervention. Currently, most light microscopes generate a small metadata set that helps retrieve the image data and refers mainly to storage related issues (i.e. , format, number of lines/pixels, date, etc.). Traditionally, developers have not offered the possibility of storing much extra information because they could see only limited opportunities for its later use. Standard file systems permitted restricted searches of the text files , often using laboriously constructed string searches where the results were returned in a similarly user-unfriendly manner.
Recent Developments More recently, it has become standard practice to store image data in more sophisticated, generic file formats that may also hold a number of parameters, for example, in an expanded TIFF format. This offers the advantage of being automatically readable by a large number of image-processing programs, where simple measurements can be performed. TIFF is actually an excellent example because one can add as many tags as one wants and could add a lot of metadata. Whatever is not recognized by the program reading the file is ignored. The increasing number of proprietary raw image formats created by camera manufacturers has recently prompted Adobe (http://www.adobe.com) to release a unified TIFF-based file format (DNG, Digital Negative) that supports a wide range of metadata declared by manufacturers.
Steffen Lindek and Ernst H.K. Stelzer· European Molecular Biology Laboratory (EMBL), Heidelberg, Germany Nicholas ). Salmon· SLS Software Technologies GmbH, Heidelberg, Germany
Handbook of Biological Confocal Microscopy, Third Edition, edited by James B. Pawley. Springer Science+Business Media. LLC. New York. 2006.
861
862
Chapter 50 • S. Lindek et al.
Recent initiatives to remedy these problems led to the development of MPEG-7 (Pereira and Koenen, 2001) as the new standard for content description of multi-media files. MPEG-7 provides descriptive elements ranging from basic features, such as colors, textures, and shapes, to complex information about motion. However, its complexity, and its lack of clarity for those illiterate in metadata management, are an impediment to its implementation. Although standard database technology has been available for many years, it is only recently that commercial relational database management systems (RDBMS) have ventured from mainframe and workstation systems to systems suitable for use on the desktop. Improvements in the quality of such systems and in the range of interfaces now available for development environments now offer microscopists the chance to store data in personalized databases. Moreover, there are now more data processing programs, such AYS Express (http://www.avs.com). available to access such database stores. Even the ubiquitous MS-Office package can access databases from within Word, Access, and Excel. Nevertheless, currently available systems are far from providing seamless integration of image data acquisition and analysis (Andrews et at. , 2002). Most recently, microscopes with integrated RDBMS have been developed (Salmon et at., 1999) and some of these are incorporated into systems such as the LSM51 0 (Carl Zeiss Jena), and the compact confocal camera (CCC), the microscope used for research at the European Molecular Biology Laboratory (EMBL). These machines use an RDBMS to describe and store their hardware configurations to an unprecedented level of accuracy and completeness . Parts can be changed without affecting performance, and the machine is automatically configured, which requires tagging the components such that they can be recognized by the software. Image data and associated metadata are all managed via databases, and common software such as Word, Access, and Excel can be used to access the databases.
IMAGE INFORMATION MANAGEMENT Image information has to be managed both at the local level, at the scientist's desk, and at the global level, for example, at scientific journals. With the advent of digital microscopy, many scientists have replaced their folders of glossy prints or slides by shelves with hundreds of CDslDYDs and drawers full of tapes . Some may have adopted desktop-based image management systems, either using software coming with their instrument (e.g., Zeiss' LSM Image Browser or Leica's IMlOOO) or commercial packages which are becoming more widespread due to the increased usage of digital photography (see Appendix). Alternatively, they may use an image archiving system tailored to microscopy such as DDL's Aequitas or RYe's Research Assi stant (see Appendix). At the other end, journals have started to make the image data submitted by authors available electronically (Anonymous, 2004). Initially this was done as an annual CD containing supplementary data, now as files for download from the publishers' Web sites (e.g., in Science' s Supporting Online Material or Nature's Supplementary Information). This also enabled publishers to provide data that could not be published in the printed journals due to its dimensionality or its huge volume. However, almost none of this image data is managed in a database, and - at least for the external user of the Web site - there is no way to bypass the individual articles and any relevant keywords appearing in the text body and directly access all images on one publisher's site relating to a particular
topic, such as fluorescence microscopy of GFP-labeled proteins in the Golgi apparatus. It is also not possible to search for a particular topic across the Web sites of different publishers (in case of the two leading journals mentioned above, a link to the image section is shown only together with the individual article). Increased usage of laboratory information management systems (LIMS) (Nakagawa, 1994) that typically use a database format to store information demonstrates how the continual improvement of data collection and processing operations is also vital for running many types of laboratory procedures, such as screening assays. As legal demands increase, LIMS may become mandatory for demonstrating conformance to stringent quality requirements. Potential benefits for LIMS users include smoother work flow, better access to information, improved quality of results, better control of costs, and increased processing speed. And, last but not least, these systems replace the traditional laboratory books in which scientists keep their handwritten notes.
THE AIMS OF MODERN MICROSCOPE SYSTEM DESIGN We have pointed out two important aspects to microscope design: On one side of the system there should be a networked microscope that functions within database structures. On the other side there should be an image database that provides easy querying and access to both data and metadata. That the database system plays a key role as the mediating component between data acquisition, image processing, and data analysis software has been recognized by the Open Microscopy Environment (OME, http://www.openmicroscopy.org), a project that seeks to improve networking of biological microscopy data and to transfer the data from the instrument to image processing and analysis software for quantitative analysis (Andrews et at., 2002; Swedlow et at. , 2003). There is actually an additional element in this data flow : the software must maintain contact between the acquisition parameter and the processed result so that the chain from data acquisition to the global database remains unbroken . The complete infrastructure envisaged for the management of information is shown in Figure 50. I. With computer-controlled microscopes, information about the operational state of the instrument can be stored automatically. A single program should control a family of instruments and organize the interaction between them and the user. The details of this interaction are retrieved from a database that describes the hardware as completely as possible, storing at least all variable parameters and periodically accessing results to see if benchmark s of resolution and stability are met. Changes in the microscope hardware are reflected in changes to database entries and affect the user interface. To this end, the components of the microscope should be labeled such that the software can recognize them automatically, as it is now done in Zeiss Axiolmager. Further information concerning the specimen and the experiment being performed can be gathered from the user in a non-intrusive manner by fully integrati ng the data collection process with the microscope control program. This information is saved in a database, together with a pointer to the image data such that the user can always be offered seamless navigation from the acquisition parameters to the image data. Figure 50.1 doesn 't make any statement regarding the nature of the archive. It may be a central file server or a network of distributed servers. As the major concern with modern digital microscopy is the size of the data and the required bandwidth, a
Databases for Two- and Three-Dimensional Microscopical Images in Biology • Chapter 50
metadata
863
BioAgency
~OOIS ~0
data
BioArchive
FIGURE 50.1. The life cycle of microscopy data and metadata. Data from one or more microscopes is stored on a file server (1). Technical information is automatically written into a database (2). The user adds details on the biology of the experiment (3). Metadata and data are reused by image processing software, resulting data is saved on the file server, and the metadata database is updated (4). Metadata is transferred to a global database, such as Biolmage (5), while the image data stays on the author's server or is archived somewhere else (6). Third parties access the metadata in the Biolmage database, which directs them to the data. The Biolmage user can use tools that access the metadata to get and process the data (7). Note that the microscope can be considered as a tool in the sense of the diagram: using the metadata saved with the original data, it can reconfigure itself and repeat the experiment. The black arrows symbolize pointers to the data. The left-hand side is a local computer system, while the right-hand side is a global system. The three-components concept for databases of biologically relevant data (BioAgency, BioArchive, and BioTools) was introduced in Carazo and Stelzer (1999).
distributed storage system will be preferable for a large project with global focus. Use of an integrated microscope and database system can provide the following benefits (Salmon et at .• 1999): • Fast, simple, and automatic machine configuration can be achieved by associating a complete set of configuration parameters with each dataset. The microscope can then be adjusted and optimized by repeatedly changing the parameters, rescanning the same sample area into a new scan window, and analyzing the effect of these changes. The best image can then be quickly selected by the user and the associated parameters reused as a start point for all subsequent scans. Moreover, functions to reuse subsets of parameters can also be created. Thus, preselected areas can be quickly scanned with the same illumination and detector sensitivity parameters, or the same area can be scanned with different detector sensitivity parameters. Loops can also be inserted to check that the data flow has not changed so much as to imply a hardware malfunction requiring operator intervention. • Configuration of different machines to the same state is possible if the microscope configuration interface uses realworld physical units rather than arbitrary units, for example, the actual laser power at some location close to the microscope objective rather than the voltage used to set the laser power output. Also, the microscope magnification should be calibrated rather than relying on the magnification value engraved into the side of the objective lens. Finally, storing the detected signal data in units of photons or photoelectrons would permit one to make a close estimate of its statistical accuracy. With parameters properly handled as physical values, users are no longer tied to finishing an experiment within a single session
or using a particular machine, and collaborating scientists can compare images acquired under conditions that are strictly comparable.
• Improved analysis of results can be achieved by storing more information about the system and using more sophisticated models. Although this is often assumed, many properties of a light microscope are not described by a constant relationship (e.g" resolution is not constant over the entire field of view and often changes with focus plane). Saving such information allows for a better calibration, and makes room for the subsequent development of algorithms to compensate for this variance. For example, because of aberrations induced by refractive index mismatch (Hell et at., 1993), image signal intensity is reduced the deeper the focus plane moves into the sample. Image processing programs such as AVS can improve image data using information about the immersion medium, the optical system, and the position of data samples relative to optical boundaries, such as the coverslip. Although the corrected image data has a more uniform contrast, making images easier to view and giving a considerable improvement in the results of threedimensional (3D) rendering techniques such as ray tracing, it is not reasonable to imagine that such programs can correct for all the vagaries that beset fluorescence microscopy (Pawley, 2000). • Improved access to data is possible when more information is stored in a modern RDBMS. Such programs are not only optimized to perform fast queries on the data they contain, but typically also offer an array of tools and interfaces that serve both novice and expert programmers in the task of data processing. RDBMS also offer presentation interfaces to many commonly used office programs, as well as dynamically generated HTML files. Users can perform sort, select, retrieve, and compare operations on image archives, querying not only the filenames, but also the entire spectrum of information associated with the image.
• Semi-automatic submissions to other databases becomes possible when more information is stored in a database. In the past, users were satisfied if they could import the image files
864
Chapter 50 • S. lindek et al.
into image-processing programs. However, now that more details can be stored with the source microscope images, users can make more useful, semi-automatic submissions to data warehouses. For example, for biological applications, it is highly valuable to export data to databases that also store biology-related information, such as in the BioImage Database described below. • Remote monitoring and maintenance is possible if the actual state of the instrument is stored and made available via a computer network. A system administrator can examine the information online, and users can benefit from fast fault diagnosis, rectification, or advice on the microscope configuration. • Repetition of experiments is a key requirement in science that is much easier to satisfy if the experimental process can be automated. By saving data and constructing systems that use this information to repeat the experiment automatically, the potential for error is reduced. • Performance and development: A well-defined database structure enables not only faster runtime performance, but also simplifies programming and program structure, speeding development time of the overall machine control program.
INSTRUMENT DATABASE MODEL The database must contain sufficient information to enable the microscope control program to facilitate the communication between the user and the instrument. This technical information can be divided into four types (Fig. 50.2). Additional information describing the biological specimen is required to meet the fundamentals of solid scientific work. • The first group of tables describing the microscope comprises essentially datasheet-like descriptions of items that could potentially be used within the instrument. This information is general, and therefore valid for any machine within a family of instruments that is served by the database. The collection of tables describing all possible components is like a library for
machine construction. Typically, entries provide information on how to control a part, how to incorporate it into the system, and indicate its expected performance. • Each particular machine contains either 0, 1, or many instances of a specific component. Each of these available parts requires a unique database entry through which it can be referenced and controlled. Because many parts require calibration, separate calibration details must also be included. Available parts and their calibration form the second and third groups. • Every time a sample is scanned and data is collected, parameters of some parts such as the photomultiplier tube (PMT) voltage or the laser power are set to values that may be different from those used for other scans. To record these temporally changing values, a fourth group of tables is required. These can exist in a separate database because they do not relate to a particular machine. • Any specimen examined under a microscope has been subjected to processing steps prior to observation. It may also be subject to processing steps, such as the addition of chemicals, between observations. In addition, different features or areas of interest may be imaged. Finally, information obtained by the microscope may be subject to one of a number of postprocessing steps such as deconvolution. Therefore, backward references to other instruments that looked at or manipulated the sample before, and forward references to what will or might happen next, complete the information network. The design of the model to hold this metadata is a timeconsuming task, but it has to be done carefully if the model is to remain optimal over the long term. Recently, ontologies have been developed that describe the knowledge in many scientific areas in a systematic computer-processable form (for a review, see Bard and Rhee, 2004). Such a framework enables consistent indexing of experimental data and allows systems in different domains to communicate with each other: Without it, one cannot exchange information without loss. The development of ontologies for the different expertise areas of microscopy is the major milestone that must be passed before the integration of databases for different individual instruments can work seamlessly (Pouchard and Walker, 2001).
SYSTEM REQUIREMENTS potential
CD
ArIon (12; 488,514 .. ) HeNe (1;534) HeNe (10;633)
available
0 calibrated
ArIon (12; 488,514 . . )
G)
ArIon (2.5;488)
@ used ArIon (0.5;488)
HeNe ( 10;633)
instrument
experiment
FIGURE 50.2. The four instrument metadata levels. The most generic level (I) holds datasheet-like information on potential components that may be used
in a certain microscope (e.g., three different lasers: a 12mW argon-ion laser with several wavelengths, a 534nm I-mW He-Ne laser, and a 633nm lOmW He-Ne laser). The next level (2) is for the documentation of the parts actually available in the microscope (only two of the laser500 ns
22mm
4096 x 4096
MPRS
Ti-Sapph
Yes, EOM
SM-PP
Laser-merge
Adjustable
Rotatable k-scan
4k,or8k in bidirect
>500 ns
22mm
4096 x 4096
Cl-plus
Up to 3, 408--638 nm
AOM (opt)
SM-PP
Laser-merge
Fixed
2 close gal vas
500,lk in bidirect
>1.68us @512x 512
17mm
2048 x 2048
Clsi
Up to 3, 408-638 nm
AOM
SM-PP laser input fiber.MM
Laser-merge
Fixed
2 close galvos
500,lk in bidirect
From
17mm w/scan rotation
Up to 512 x 512 in spectral mode
TriMScope'
Biotech
Leica
Lasers/Arc
Retrace protectfLaser atten
emission
fiber
4.08us at 512 x 512 in spectral mode
Olympus
FV 300
FV 1000
Many, 405--633 nm IR port Many,
Yes, AOTF
SM-PP
Laser-merge
Fixed
2 close gal vas
I k or 2k bi-direct
>2J.ls
20mm
2048 x 2048
Yes, AOTF
SM-PP
Laser-merge
Fixed
2 pair of close gal vas, separate imagelbleach
2k or 4k bi-direct
>2J.ls
18mm
4096 x 4096
3 krpm, 15 fps
>1 J.ls
18mm
CCD
2kHz
2 J.ls
17
CCD
20-125 ns
active
DSU
stabilizer, 351-633 nm, IR port Hg arc
scanners,
NA/intcnsity controlled by
NA
NA
NA
SM-PP
Laser-merge
NA
aperture
Visitech
Yokogawa'
VT-infinity
Many, 405-647 njm
Yes, AOTF
VT-Eye
Many, 351--647 nm
Yes, AOTF
SM-PP
Laser-merge (opt)
Fixed
I galva, 1 AOD
50kHz
CSU 10
2 or 3 lines'
NAtAOTF
SM 3.5J.lm
Laser-merge
Fixed
Double Petran
1800 rpm, -1 J.ls 360 fps
13 x 9.5 mm CCD
(opt)
CSU22
3 or 4 laser lines
NAtAOTF
SM 3.5J.lm
Single galvo scans an array of point sources
Disk w/micro-
1024 x 1024
lenses
core
Laser-merge
Fixed
Double Petran Disk w/microlenses
Variable, to 5 k rpm, I k fps
-I J.ls
13 x 9.5 mm CCD
core
Zeiss
circular bleach Interchangeable single-sided slitpattern disk, 3 krpm
LSM510META
Many, 351--633 nm (Ti: Sapph)
Yes, AOTF 0.05-100% (AOM)
SM-PP
Laser-merge
Adjustable
2 close gal vas
1.3 k or 2.6 kin bi-direct
640 ns2.3 ms
18mm
2048 x 2048
LSM 5 Pascal
Many, 405--633 nm
SM-PP
Laser-merge
Adjustable
2 close gal vas
1.3 k or 2.6 kin bi-direct
640 ns2.3 ms
18mm
2048 x 2048
LSM5-LJVE
Many, 405--635 nm
Nol Mechanical attenuator 0.05-100% Yes, AOTF 0.05-100%
SM-PP
Laser-merge
Adjustable Cylindrical
I galva
(>60k) 16 J.ls120 fps, 20ms 512 x 512, 1010 fps, 512 x 50
18 mrn
1024 x 1024
line-scan
Record transmitted light through disk. As the TriMScope is actually a multi-focus multiphoton fluorescence illuminator with widefield detection onto a CCD, its performance depends a great deal on the performance of this device. 3 These numbers assume that the tube mag is Ix. I
2
Light Paths of the Current Commercial Confocal Light Microscopy Used in Biology • Appendix 2
909
Table A2.1. (Continued)
Pinhole alignment
range
Tube mag
Beam dump
NA
l.lx
NA
5-place dichroic wheel
Selfaligning
ROI for CCD mode
Ix
NA
Short-pass dichroic
Zoom
32: I
Ycs
Bearn-
splitter
Acousto-
Yes
Acousto-
selection
Photodetector
Reflected! transmitted
Digitizer
z-motion
8-place filter wheel
CCD or EMCCD
2-camera port
No/yes'
CCD
Piezo ±100nm (opt)
NA (multiphoton
8-place filter wheel,
3-camera port
Yes/no
spectrometer
only)
(opt)
CCD, (opt ion, PMT, 12-bit)
Stepper
excited
CCD or EMCCD, 1~32 PMT array 4PMT, cooling option, APD option 4PMT, cooling option, +2
8
Yes/yes
12-bit
Galvo, ±40nm
8
Yes/yes
12-bit
Galvo, ±40nm
Preset
Optic
32: I
Channels
Spectral Pinhole range
Fixed
optic
Fixed 70/.lm, 180 /.l spacing
Common pinhole, adjust 20~800 /.lm
Prism, motorized
Common
Prism,
pinhole, adjust 20~800 /.lm
motorized
mirrors
mirrors
motor,
(peizo opt)
noodescanned infinite
3.8
Yes
Dichroic,
changes with cube infinite
Dichroic,
changes with cube
10: I (infinity)
3.42x
50: I (infinity)
3.82x
Yes
Yes
Dichroic, cubes, 2 positions Dichroic wheel,6 positions
NA
Ix
No
Filter cube
Fixed w/focusable. alignable pinhole lens Fixed w/focusable, alignable pinhole lens
Common
Common
pinhole, 30,60, 100, 150/.lm3 Common
pinhole, 30,60, 100, 150/.lm4
5 sizes
pinhole, alignable Common
pinhole, alignable
Selfaligning
adjust 50--800 or 50--300 on spectral Vert & horiz slits,
Replaceable filter cubes
3 side-window fiber-coupled PMTs
4
Yes/yes
12-bit
Stepper, ±SOnm
3 diffraction gratings for 2.S nm, 5 nm, and 10 nm channel width
32 element multianode PMT
32 acquired simultaneously
Yes/yes
12-bit
Stepper, SOnm
Dichroic filter cube
3 PMTs, 2
3, 2 fl, I trans
Yes/yes
12-bit
Stepper, ±lOnm
2 diff-grating channels,
5, 4 fl, I trans
Ycs/yes
12-bit
Stepper, ±lOnm
slits
5 PMTs, 2 spectral, I trans, Photon counting mode
Dichroic filter cube
CCD or EM-CCD
No/yes
CCD
Stepper, ±lOnm
Yes/no
CCD
motorized
increments
fluor, 1 trans
5 sizes
NA
No
50: I
NA
No
Ix
NA
Dichroic, 4 positions
preset, reet. array
~I k fixed, 50/.lm
Dichroics/ filters
CCD
Dichroic, 6 positions
adjustable preset! adjustable
5 slits,
Dichroics/ filters
4 hi-QE PMTs
50/.lm, 20 k on disk, ~I k /FOV
Dichroics/ Filters,
CCD or EM-CCD
2 -camera port
No/no'
CCD
2-camera port
No/no'
CCD
Yes/yes
8~12-bit
I dichroic, exchangeable by user
Selfaligning
1O~100
/.lm
CCD
Piezo,
±100 nm 4
Yes/ycs
10 bits
Piezo,
±100 nm
3 emisson,
3 barrier
NA
Ix
NA
Dichroic, 3 positions
Selfaligning
50/.lm, 20 k on disk, ~I k /FOV
Dichroics/ Filters, 3 emisson, 3 barrier
CCD or EM-CCD
0.7-40x
0.84x
Yes
Dichroic, 4 positions
4 x,y,(z), diameter
3, 200 steps, 0.1~13 Airy Units 10--1 k /.lm
3 dichroics, 6 positions, + spectral detector, \0 nm !channel
3/4 filtered PMTs;&/or diff. Grating w/32/.lPMTs
2 dichroics, 6 positions
2 filtered PMTs, trans PMT
4
detector
512 x I linear CCD
2
adjustable
0.7-40x
0.5~2x
0.84x
1.18x
Yes
Yes
Dichroic,
2 x,y,
I, 200 steps,
2 positions
diameter
0.1~13
adjustable
Units 10--1 k /.lm 17 slits, 0.5~ 10 Airy units
Achrogate,
2
line-mirror on clear
adjustable
blank
4 These
Airy
dichroic, 12 positions, 8 position bamers
Microscope IOnm, Piezo,
±5 nm
array/
No/yes
No/no
8~12-bit
8~12-bit
Microscope
10nm, Piezo, ±S nm Microscope 10nm, Piezo, ±5 om
numbers assume that the tube mag is Ix. Yokogawa scanners are manufactured by Yokogawa Electric (Tokyo, Japan), but retailed by a number of companies including, Andor Technologies (Belfast, UK), Solamere Technology (Salt Lake City, UT), PerkinElmer (Downer Grove, II), Visitech (Sunderland, UK). oIt is possible to use 4 lasers with a quad, dichroic beamsplitter. 'Transmission PMT and 4-channel non-descanned PMT detector also available. 5
- --CD
--- ®
o
:' ~ @ - - -··::: ...... ···r "'" ,~ ---
r"' 0 ·- '-Q'--_----9S% of the signal light to the detectors. The size of the raster on the specimen is controlled by a 0.S-2x zoom optic (6), that feeds the light to the y-scanning mirror (7), through the scan lens (8), the objective lens (9) and on to the specimen , ( 10). Returning signal follows the same path but mostly misses the reflecti ve strip in the Achrogate and proceeds through a wheel of secondary dichroic beam-splitters (11) to one of 2 tube-lenses (12) that each focuses the line illuminated in the specimen onto a 17-position, slit aperture plate (13). Light pass ing the slits is first filtered by emission filters (14) and then detected by a I x S12 linear CCD detector (IS) (see also Fig. 9.6).
A
Spectral Imaging
Imaging
L CL M BC MDBS
laser collimator lens mirror beam combiner main dichroic beam splitter
SCXY S PH
scanner X/Y objective sample variable pinhole
DBS EF PMT G PMTA
dichroic beam splitter emission filter photo multiplier tube grating PMT array (META)
NDD FO EPD
non-descanned detector fiber out external photodetector
0
PMT L
L
B FIGURE A2.9B. Optical beam path of the Zeiss LSM 510 META. A unique scanning module is the core of the LSM 510 META. It contains motorized dichroic mirrors and barrier filters, adjustable collimators, individually adjustable and alignable pinholes for each of 3 (or even 4) detection channels, as well as scanning mirrors, and highly sensitive PMT detectors including the 32 micro-PMTs of the META spectral detector. All these components are arranged to ensure optimum specimen illumination and efficient collection of reflected or emitted light. The highly optimized optical diffraction grating in the META detector provides an innovative way of separating the fluorescence emission spectrum to strike 32 separate, micro-PMTs, each of which covers a bandwidth of -10 nm. Thu s, a spectral signature is acquired at each pixel of the scanned image. Such a dataset can subsequently be digitally "unmixed" to separate signals from dyes with overlapping emission spectra. The Beam Path: (1) Optical Fibers, (2) Motorized collimators, (3) Beam combiner, (4) Main dichroic beamsplitter, (S) Scanning mirrors, (6) Scanning lens, (7) Objective lens, (8) Specimen, (9) Secondary dichroic beamsplitter, (10) Confocal pinhole, (11) Emission filters, (12) Photomultiplier, (13) META detector, (14) Neutral density filter, (1S) Monitor diode, (16) Fiber out.
Light Paths of the Current Commercial Confocal Light Microscopy Used in Biology • Appendix 2
917
VIS Fiber
Tube Lens
Eyepiece
Q;
03
j j Ql
Z
I
I
Ql
Scan Module on Side Port
Laser Module UV
Ql
Z
Ql
Laser Module VIS
L-----------~--~~APD2
FCS on Base Port
APD Unit
c FIGURE A2.9C. (Continued) Schematic diagram of Zeiss LSM FCS showing how the fluorescence-correlation spectroscopy (FCS) unit is attached via the base port of the Axiovert 200M microscope while the LSM 510 META is attached to the side camera port. All figures kindly provided by Carl Zeiss Inc. (Jena, Germany).
Appendix 3
More Than You Ever Really Wanted to Know About Charge-Coupled Devices James B. Pawley
INTRODUCTION The electronic structure of crystalline Si is such that electromagnetic waves having the energy of light photons (1.75-3.0 electron volts) can be absorbed to produce one free or "conduction" electron. If an image is focused onto a Si surface, the number of the photoelectrons (PE) produced at each location over the surface is proportional to the local light intensity. Clearly, all that is needed to create an image sensor is a method for rapidly converting the local PE concentration into an electronic signal. After almost 40 years of NASA and DOD funding, the slow-scan, scientific-grade, charge-coupled device (CCD) camera is now an almost perfect solution to this problem. Success in modem biological light microscopy depends to an ever-increasing extent on the performance of CCD cameras. Because such cameras differ widely in their capabilities and are also items that most biologists buy separately, rather than as part of a system, some knowledge of their operation may be useful to those practicing biologists who have not yet found it necessary to be particularly interested in "electronics." Although the basics of CCD operation are described in many other chapters (particularly, Chapters 4, 10 and 12) this Appendix describes the operating principles of these devices in greater detail and also discusses the ways that they "don't work as planned." It then covers the operation of the electron-multiplier CCD (EM-CCD), a new variant that reduces the read noise almost to zero, although at the cost of reduced effective quantum efficiency (QEeff).' The second section, How to choose a CCD, is a review of CCD specifications with comments on the relevance of each in fluorescence microscopy.
PART I: HOW CHARGE - COUPLED DEVICES WORK The first step is to imagine a rectangular area of the Si surface as being divided into rows and columns, or more usually, lines and pixels. Each pixel is between 4 x 4 flm and about 24 x 24 flm in size and the location of any pixel of the surface can be defined in terms of it being x pixels from the left side, on line y. To construct an actual system like this, start with a smooth Si surface; cover it with a thin, transparent, insulating layer of Si0 2 ; deposit onto the Si0 2 , a pattern of horizontal strips, made out of a transparent conductor called amorphous silicon (or poly-silicon), so that the strips cover the entire image sensor area. Although, viewed from the top, these strips partially overlap each other, they
I
This loss can be avoided if the system is used in photon-canting mode.
are kept electrically separate from their neighbors by additional layers of SiOz. Every third stripe is connected together to form three sets of interdigitating strips that we will refer to as Phases I, 2 and 3 (" 2, 3, Fig. A3.1). Taken together, all these phases constitute the vertical register (VR) and, after the assembly has been exposed to a pattern of light, they are used to transfer the photoinduced charge pattern downwards, one line at a time. The pixels along each line are separated from each other by vertical strips of positively doped material injected into the Si. These positive "channel blocks" create fields that prevent charge from diffusing sideways without reducing the active area of the sensor. Any photon that passes through the stripes and the Si02 , is absorbed in the Si, producing a PE. If a small positive voltage (-15 volts) is applied to the , electrodes, any PE produced nearby will be attracted to a location just below the nearest , strip (Fig. A3.2). As additional PEs are produced, they form a small cloud of PEs referred to as a charge packet. The number of PEs in the packet is proportional to the local light intensity times the exposure period and the problem now is to convey this packet to some location where its size can be measured, and to do this without changing it or losing track of the location from which it was collected. This will be achieved by using the overlying electrodes to drag the charge packet around in an orderly way until it is deposited at the readout node of the charge amplifier.
Charge Coupling The dragging mechanism operates in the following way: First 2 is also made positive so that the cloud diffuses to fill the area underneath both , and 2' Then , is made zero, forcing the packet to concentrate under 2 alone (Fig A3.2). So far, these 3 steps have succeeded in moving the charge packets that were originally under each of the , electrodes downwards by one phase or 1/3 of a "line" in the x-y raster. If this sequence is now repeated, but between 2 and 3 and then again between 3 and the , belonging to the next triplet of strips, packets will have moved down by the one entire raster line. PEs created within a particular pixel of each horizontal stripe remain confined by the channel stops as they are transferred to the next line below. A pixel of the image is therefore defined as the area under a triplet of overlying, vertical charge-transfer electrodes and between two neighboring channel blocks. The pixels on scientific CCDs, are usually square, 4 to 30 flm on a side while those on commercial, video CCDs are likely to be wider than they are high, to conform with the reduced horizontal resolution of commercial video standards. Only square pixels can be conveniently displayed in a truly digital manner. Larger pixels have more leakage current (dark-current), but are also able to store more charge per pixel (see Blooming, below).
James B. Pawley. University of Wisconsin, Madison, Wisconsin 53706 918
Handbook of Biological Confocal Microscopy. Third Edition, edited by James B. Pawley. Springer Science+Business Media, LLC. New York, 2006.
BASIC CCO ARRAY Vertical phase
-_.-
3
!
Ir
I-< t - t -
•
l- i-- charge amp. Interline transfer: (shows top· leN corner)
Full frame transfer
Full frame transfer w/electron multiplier
Gai~ "'t:'~~';~"--
4 phases,
70% over the visible range. • More expensive because of the extra fabrication. • Slightly less resolution and more fixed-pattern noise, caused by imperfections in the thinning operation, and the presence of two sets of surface states. Color Chips • One-chip color sensors employ a pattern of colored filters, one over each pixel. Light stopped by any such filter cannot be detected and is therefore lost. The QE of such sensors is therefore at least 3x lower than for an otherwise comparable monochrome chip. • 3-chip color sensors use dichroic mirrors to separate the "white" light into three color bands, each of which is directed to a separate monochrome CCD sensor. While this would seem to ensure that "all photons were counted somewhere," because such systems seldom employ microlenses, their effective QE is not much better than the I-chip color sensors and alignment of the signal light is important. '7
1. Quantum Efficiency (QE): QE is the ratio of photons striking the chip to electrons kicked into the conduction band in the sensor. It should be at least 40% and
17 10
There is no multiplicative noise because any spike above the FET noise floor counts as one electron. no matter how much it has been amplified.
927
While the QE is not much better. the resolution of the 3-chip camera is the same as that of each chip, without the interpolation needed to disentangle the 3 colored images from the output of a l-ehip color sensor.
928
Appendix 3 • J.B. Pawley
• Color can be detected by making sequential exposures of a monochrome chip through colored glass or LCD filters. This produces the same QE losses as the patterned filter but has the advantage that it can be removed when higher sensitivity is needed. This design is not suited for imaging moving objects.
2. Readout Noise: This spec is a measure of the size of the pixels and the quality of the circuitry used for measuring the charge packet in each pixel. It is measured in "±RMS electrons of noise" (i.e., 67% of a series of "dark" readings will be ± this much). • A good scientific CCD camera should have a noise level of sqrt of the number of detected photons = sqrt # electrons). Consider the signal levels that you plan to use. Will the darkest important part of your image have zero signal or do you expect some background signal from diffuse staining or out of focus light? If the dimmest pixel in your image represents -100 electrons, then the Poisson or statistical noise on this background signal will be ±1O electrons. "Adding" an additional ±1O electrons of readout noise will not make much difference to a measurement of this background signal and it will be even less significant when added to the even greater Poisson noise present in pixels where the stained parts of the image are recorded. This is especially true because RMS noise signals add as the "sqrt of the sum of the squares" (i.e., the total noise from ± 10 electrons of readout noise and ± 10 electrons of Poisson noise is only sqrt (l00 + 100) = ±14 electrons). On the other hand, if you are really trying to keep those cells alive and you find that 2,000 electrons in the bright areas is enough, the dark areas may now be only 50 electrons. As the sqrt of 50 is about ±7, an additional ±1O electrons of readout noise may no longer be acceptable, but only if you have to make measurements in the dark areas on your image. In this case, the obvious choice is a slower, quieter CCD or an EM-CCD. While in widefield fluoresecence, the background stain level is seldom so low that the sqrt of the signal recorded is lower than the read noise, the disk-scanner does provide such an image (Chapter 10). As one of the main advantages of disk-scanning is that one can scan an entire image plane very rapidly, the fact that one can read out the EM-CCD very rapidly without increasing the read noise makes it the ideal detector for this type of scanner (or, indeed for high-speed line scanning confocal microscopes).
TABLE A3.2. Dynamic Range and Pixel Size
Pixel Size Full Well Least significant bit Implied noise level
=
l2-bit camera w/small pixels
14-bit camera w/large pixels
6.7 x 6.7J.!m 27,000 6.5 electrons ±13 electrons
24 x 24 345,000
21 electrons ±42 electrons
on the chip determines the total specimen-to-chip magnification needed! Two examples: a. 1.4 NA 100x objective and a I x phototube. • The Abbe Criterion resolution @ 400 nm is about 0.22/lm. Magnified by a total magnification of 100x, this becomes 22/lm at the CCD. • A CCD having 8 x 8/lm pixels samples such an image adequately (-2.8 pixels/resolution element). b, 1.3 NA 40x objective and a Ix phototube. • The Abbe Criterion resolution @ 400nm IS now 0.25/lm. Magnified 40x this becomes, 10 /lm. • A CCD having 8 x 8/lm pixels is inadequate to sample this lower-mag, high-resolution image. If you must use this objective, you need either a higher mag phototube (2.5x) or a chip with 3 x 3/lm pixels or (as CCD pixels are seldom this small), some combination, • Saturation signal level: The maximum amount of signal that can be stored in a pixel is fixed by its area. The proportion is 600 electrons/square /lm, so a 10 /lm x 10 /lm pixel can store a maximum of 60,000 electrons before they start to bleed into neighboring pixels. In practice, as fluorescent micrographs of living cells seldom produce signals this large, large pixels are usually unnecessary, However, the saturation level also represents the top end of another spec, the dynamic range. This is usually quoted as 12-bit (4000: 1) or 14-bit (16,000: I) etc., and represents the ratio between the full-well saturation level and the readout noise. Therefore, a camera with relatively high readout noise can still look good in terms of dynamic range if it has large pixels and hence a high full-well capacity. Conversely, a 12-bit camera with small pixels can have less actual noise-per-pixel intensity measurement than a 14-bit camera with large pixels. In this case, the noise level of the 14-bit camera is >3x that of the 12-bit camera. Your signal/pixel would have to be 3x larger in order to be "seen" when using this particular 14-camera.
4. Array Size: 19 The argument for small • Assuming 0.1 /lm pixels (referred to the object plane), a 512 x 512 pixel chip will image an area of the specimen that is about 51 x 51 microns. If this is enough to cover the objects you need to see, this small chip has a lot of advantages over chips that are 1024 x 1024, or larger. • Lower cost
3. Pixel Size: • Nyquist sampling: The size of a pixel on the CCD is, in itself, not very important BUT one must satisfy the Nyquist criterion: The pixels on the chip must be -4+ smaller than the smallest features focused onto it lS (see Chapter 4): Pixel size
" Of two times smaller than the "resolution," as defined by Rayleigh, or Abbe. " The array size refers to the number of lines and pixels in the sensor, not to its total area.
More Than You Ever Really Wanted to Know About Charge-Coupled Devices • Appendix 3
• 4x fewer pixels to read out, meaning either: 4x slower readout clock, giving 2x lower readout noise. Same clock speed and noise level but 4x faster frame time. (Easier to scan specimen to find the interesting part! Time is money!) • 4x less storage space needed to record data.
929
if the signal level is so low that 1 s/frame is required to accumulate enough signal to be worth reading out, then reducing the read time much below 0.1 s loses some of its appeal. Faster readout speeds are particularly important for moving specimens, especially when doing widefield/deconvolution or when following rapid intracellular processes, such as vesicle tracking or ion fluxes.
The argument for big: • Manufacturing improvements are reducing readout noise levels at all readout speeds, and CCDs with more pixels often also have smaller pixels which can lead to lower read noise. If your labels are bright, having a larger chip allows you to see more cells in one image (as long as they are confluent!). Assuming that Nyquist is met in both cases, a large print of an image recorded from a larger sensor always looks sharper than one from a smaller array. • Binning: Binning refers to the process of summing the charge from neighboring pixels before it is read out. This increases the size of each charge packet read (making it look brighter) and reduces the number of pixels. For example: 2 x 2 binning allows the owner of a 1024 x 1024 chip to obtain the speed/noise performance similar to the smaller chip (512 x 512) and to do so in a reversible manner. However, the optical magnification may need to be increased to preserve Nyquist sampling. Before deciding that you need a larger chip, compare what you would get if the same money were spent on another scope/CCD/graduate student!
Bottom line: • If more pixels means smaller pixels, they will each catch fewer photons unless the magnification is reduced proportionally. More pixels at the same frame 20 rate mean somewhat higher read noise because the pixel clock must go faster.
5. Readout Speed: Although readout speed has been discussed above, we haven't mentioned that some good CCD cameras have variable speed readouts and the new EM-CCDs impose no high read speed penalty (Table A3.3). It is convenient to be able to read out the chip faster when searching and focusing as long as one can then slow things down to obtain a lower read noise in the image that is finally recorded. However, the read speed is only one limitation on the frame rate: TABLE A3.3. CCD Specifications Array size
Pixel Clock Rate
Noise level*
Frame time
640 x 480
13MHz (video rate) 100kHz IMHz 5MHz 100kHz IMHz 5MHz
200e/pixel**
0.033
512x512
1024 x 1024
5e/pixel 15e/pixel 35e/pixel 5e/pixel 15e/pixel 35e/pixel
2.5 sec 0.25 sec 0.05 sec 10sec I sec 0.2 sec
7. User-friendliness: State-of-the-art cameras often seem to have been designed to make sure that no one unwilling to become a devotee of "CCD Operation" can possibly use them efficiently! Start off by asking to see an image on the screen, updated and flat-fielded at the frame-scan rate and showing as "white" on the display screen, a recorded intensity that is only -5% of the full-well signal. This is where you should do most living-cell work. Then ask the salesman to help you to save time-series of this image. Increase the display contrast until you see the noise level of the image, both before and after "flat-fielding." Put a cursor on one pixel in the top frame of the stack and plot its intensity over the series.
8. "The Clincher" (Well, at least sometimes ... ): Ask him/her what the intensity number stored in the computer for some specific pixel means, in terms of the number of photons that were recorded at that location, while the shutter was open. To answer, the salesperson will have to know the QE, the fill factor and the conversion factor between the number of electrons in a pixel and the number stored in the computer memory (sometimes called the gain-setting). To help them out, any "real" scientific CCD camera has the latter number written, by hand, in the front of the certification document (usually a number between 3 and 6). If the salesman doesn't understand the importance of this fundamental number, what hope is there for you? (Hint: It is important because the Poisson noise is the sqrt of the number of electrons in the well, not the sqrt of some arbitrarily proportional number stored in your computer.)
Frame rate/s 30 0.4 4 20 0.1 5
* Assumes conventional FET circuits. ** The readout noise is relatively highcr at video rate because the higher speed oftcn precludes the use of various techniques, such as correlated double sampling, that reduce readout noise,
111
6. Shutter Stability: Though not strictly a CCD spec, electronic (LCD) or mechanical shutters are often built into modern CCD cameras. 21 The latter have the disadvantages of producing vibration and having a limited lifetime but the advantage that they transmit all of the light when they are open (even an "open" LCD can absorb >50% of the light, other electronic shutters may be better). There seems little point in having a camera capable of recording (say) 40,000 electrons/pixel with an accuracy of ±200e (or 0,5%!) if the shutter opening time is only accurate, or even reproducible, to ±10%. If one shutters the light source instead of the camera, similar limitations apply.
The readout speed of a 2 x 2 binned 1024 x 1024 is a bit slower than an actual 512 x 512 because twice as many vertical clock cycles are needed, and one still needs to read out pixel by pixel in the horizontal direction.
B. Things That Are (Almost!) Irrelevant When Choosing a Charge-Coupled Device for Live-Cell Microscopy 1. Dynamic Range: This is the ratio of the "noise level" to the "full-well" (or maximum) signal. Although 16-bit may sound a lot better than 12bit, you need to think before you are impressed. The noise level should not be more than 5 electrons/measurement. Period! 11
Often the same advantage can be gained by shuttering the light source. This may become more common as pulsed laser or light-emitting-diode light sources are introduced (see Chapters 5 and 6).
930
Appendix 3 • J.B. Pawley
Twelve bits is 4,000 levels. If the first level represents 5 electrons (in fact, it should represent half the noise or 2.5 e), then the 4,000th represents 20,000 electrons or (assuming a QE of 50%), about 40,000 photons/pixel/measurement. How often do you expect to be able to collect this much signal from an area of a living cell only 100 x 100 nm in size? You should be able to get a good, 8-bit image using only 6% of the dynamic range of a 12-bit CCD (Fig. A3.8).22 As the "full-well" signal is only proportional to the area of the pixel on the chip (area in sq. !lm x 600), the dynamic range is only really impressive if it is high AND a chip has small pixels. Then it means that the readout noise is low. A test for actual dynamic range is described below. Bottom line: For disk-scanning confocal microscopy, a large dynamic range is only important if it reflects a low readout noise level. Easier to just check the readout noise!
2. High Maximum Signal (high, full-well number, because of large pixels): On living cells, you will probably never have enough light to reach a full-well limit of even 20,000 electrons. Even if you do, there are better ways to use it (more lower-dose images to show time course?).
If 16k electrons (full well) 12 bits
=12 bits, each digital level =4 electrons.
16k
electrons
If the read noise Is ~8 electrons, or 2 gray levels, one can obtain a useful," 8-bit" (256 levels) image by using only 6% of Its dynamic range.
... 11 bits
1 0 bits
. ..
9 bits
8 bits 7 bits
.250
..
. . .. . . ..
256
128
64
o
FIGURE A3.B. Not using the full dynamic range of a CCD. As most scientific CCOs have more dynamic range than one "needs" in live-cell fluorescence microscopy, the excitation dose to the specimen can be reduced if one sets up the CCO control program to display an 8-bit image using only the bottom 1,024 levels of a l2-bit image. Such an image is more than adequate for many functions in live-cell biological microscopy (particularly when other factors such as dye-loading etc., may cause larger errors) and will require only 6% as much signal as would a "full-well" image.
22
Remember, given optical and geometrical losses, you can collect no more than about 3-10% of the photons produced, and, each fluoroscein molecule will only produce perhaps 30,000 excitations before "dying."
3. "Imaging Range" "Sensitivity" (or anything measured in LUX): Stick to something you (and I?) understand: Photons/pixel or electrons/pixel. The other conversions are not straightforward.
4. "Neat Results": Unless you know how well stained the specimen is, you cannot evaluate an image of it in a quantitative manner. (Though you may not want to admit this!) By all means, view your own specimens, but viewing "test specimens" that are not expected to fade and have a known structure (fluorescent beads in some stable mounting medium?) facilitates AlB comparison. If you do use your own test specimens to compare cameras, be sure to view them on the same scope, and with the same conditions of pixel size and readout time etc. Better still ...
C. A Test You Can Do Yourself!!! Set up each camera that you want to evaluate on a tripod, add a C-mount lens, and an ND 3 or ND 4 filter. Hook up a monitor or computer and view some scene in your laboratory under ordinary illumination (avoid light from windows which may vary from day to day). Close the lens aperture down until you can no longer discern the image (see Fig. 4.20). This is the "noise-equivalent light level": the signal level at which the electron signal (i.e., photons/pixel x QE) just equals the total noise level. Your measure is the aperture at which the image disappears.23 Because it is sensitive to both QE and readout noise level, this is a very useful measure of what we all think of as the "sensitivity." Of course, the signal level depends not only on the light intensity but also on the exposure time and the pixel area, so make sure to keep the former constant and make allowances for the latter. If you do not have even these meager facilities (a C-mount lens, an ND filter, a tripod and some time), take an image of nothing. Look at "no light" for one second, and for 100 seconds. Ask to see a short line profile that plots intensity vs. position along a line short enough that one can see the intensity of each individual pixel. The difference in the average intensity between the short and long exposure is a measure of the leakage. 24 With a little calibration from the published full-well specs (a spec less open to "interpretation" than "noise"), you can even get a direct measure of the read noise level from these dark images. (It should be the standard deviation of the values as long as they are counted in electrons, not "magic computer units" and as long as fixed-pattern noise is not a factor.) And just trying to work it all out will give you some idea if the salesman knows anything ...
D. Intensified Charge-Coupled Devices Intensified CCDs (ICCDs) are just that: the mating of an "image intensifier" to a CCD. The idea is that the photon gain of the intensifier (can be 200-2000x) will increase the signal from even a
If the lens doesn't have a calibrated aperture ring, you can open the aperture all the way and reach the "threshold" exposure level by reducing the exposure time and adding NO filters. Remember to also correct for pixel area. Larger pixels intercept more photons. 24 Wilh a good EM-CCO, this measurement can be done using a short exposure and high EM gain, then counting the number of amplified dark charge/CIC spikes across a typical line of the raster.
23
More Than You Ever Really Wanted to Know About Charge-Coupled Devices • Appendix 3
single photoelectron above the read noise of the CCD. This occurs, and can be particularly useful where fast readout is needed such as when measuring ion transients. Finally, pulsing the voltage on the intensifier section makes it possible to shutter ("gate") the camera on the ns time scale, making the rCCD useful for making fluorescence lifetime measurements (Chapter 27, this volume). However, ICCDs do not have the photometric accuracy of normal CCDs for a number of reasons: • The relationship between number stored in memory and the number of photons detected is generally unknown and variable. • The intensifier photocathode has low QE 25 (compared to that of a back-illuminated CCD). • The "resolution" is generally only dimly related to CCD array size because of blooming in the intensifier. To check this, reduce light intensity until you can see the individual flashes produced by single photoelectrons. See how many lines wide they are. (They should be one line wide.) • They have additional noise sources: phosphor noise, ions in intensifier section create flashes, high multiplicative noise in the intensifier section greatly decreases QE elfo etc.
" And the GaAsP photocathode with better QE, have to be cooled, making the assembly very expensive.
931
• Photocathode resistivity can produce "dose-rate" effects: nonlinearities in which the recorded intensity of the brightest areas may depend on (and affect) the brightness of nearby features. Because I expect that EM-CCDs such as those mentioned above will soon supplant rCCDs except where fast gating is needed, I have not gone into more detail here. For more info, go to: http://www.stanfordphotonics.com!
ACKN OWLEDG EMENTS The author would like to thank Dr. J. Janesick, formerly of the Jet Propulsion Lab (California Institute of Technology, Pasadena, CA), for many conversations about CCD operation and for the original sketches for Figures A3.!, A3.4, and A3.5 and to Colin Coates, CAndor Technologies, Belfast, UK) for his helpful comments on the manuscript and for Figure A3.6.
REFERENCES Inoue and Spring, 1997, Video Microscopy, Second Edition, Plenum, New York, 1-741, particularly Chapters 5-9. Pawley, J .B., 1994, The sources of noise in three-dimensional microscopical data sets, Three Dimensional Confocal Microscopy: Volume Investigation of Biological Specimens, (J. Stevens, ed.), Academic Press. New York, 47-94.
Index
I suppose it is inevitable that indexes are compromises: If one includes every mention of every entry, the index becomes as long as the book. There is also the time dimension: As one cannot start writing the index until the book has been paginated, every day spent on the index directly delays the publication date. For the Second Edition, I prepared the index somewhat in parallel with the page proofs and it took most of a semester. For this Third Edition, a professional indexer was used to compile the initial index. We then expanded the level of cross-referencing through a series of digital searches. The final result may show its mixed parentage. As you use this index, please consider the following. I confess that many entries contain far fewer referents when they appear as sub- or sub-sub-heads than when they appear as capitalized headings. In addition, some See alsomarkers use acronyms and it is also true that these can get confused with the real title of the entry. In compensation, have tried to put in bold type those page numbers on which lone would find the more comprehensive discussions of the topic we have added a period at the end of the major heads to distinguish them from sub-heads. My apologies for any errors. My thanks to Helen Noeldner for her calm and competent assistance during this long and laborious process. Please use the Feedback page at http://www.springer.com/387-25921-X to bring errors to our attention so that they can be corrected in future printings. Remember that this Handbook has always been a community project. Good hunting! JP,2/21106 Numbers 20 imaging, blind deconvolution approach, 476-477. 20-time V.I'. 30-time, embryo, 762-764. 20 pixel display space, 291. 20CHO dataset, 818. 20HeLa dataset, 818. 2-photon, (2PE). See Two-photon excitation. 30 Constructor, 282. 30 imaging, alternative approaches, 475-476, 607-624. See also, Confocal topics; Multidimensional microscopy topics. episcopic fluorescence image capture (EFIC), 607-608 light sheet microscopy (SPIM), 613 magnetic resonance microscopy (MRM), 618-624 amplitude modulation of RF carrier, 620 applications, 623-624 basic principles, 618-619 botanical imaging, 624 developmental biology, 624 Fourier transform, image formation, 620 future developments, 624 hardware configuration, 621, 622 histology, 623, 624 image contrast, 622-623 image formation, 619-621 Larmor frequency, 620 phenotyping, 623 schematic, 619 strengths/limitations, 622 micro-computerized tomography
(Micro-CT), 614-618 contrast/dose, 614-615 CT scanning systems, 615-618
dose vs. resolution, 616 layout, 614 living mouse, 615, 617 mouse femur, 616 operating principle, 614 tumor-bearing mouse image, 617 optical coherence tomography (OCT), 609-610 human retina 609 schematic, 610 Xenopus laevis embryo, 610 objectives on a tandem scanner, 154, 304 optical projection tomography (OPT), 610-613 lamprey larva, 612 mouse embryo, 612 plants, 774-775 setup, 611 real-time stereo imaging using LLCO related methods, 607-625 selective plane illumination microscopy (SPIM),613 Medaka heart, 614 surface imaging microscopy (SIM), 607-608 30 Scanning Light Macrography, 672. 30 for LSM, 282. 30 methods compared, 448-451, 644-647. table, 647 30 multi-channel time-lapse imaging (40/50). See also, Time-lapse imaging. table, 384. 303T3 high-content screening dataset, 820, 821.
30HeLa high-content screening dataset, 820, 821. 3PE. See Three-photon excitation.
40 imaging. See Four dimensional imaging. 4Pi microscopy, 561-570. 4Pi-PSF, 570 axial resolution, 563 ISM, 561, 569-570 OTF, 569-570 living mammalian cell imaging, 564-565 Golgi apparatus, image, 566 lobe-suppression techniques, 561 interference of excitation and detection, 561 confocal detection, 561 two-photon excitation (2PE), 561 MMM-4Pi microscopy, 554, 556, 563-564 basics, 565 scheme, 563 optical transfer function (OTF), 562, 563 outlook, 568-569 point spread function (PSF), 562-563 signal-to-noise ratio, 561 space in variance of PSF, 457, 490, 564 theoretical background, 562-563 type C, with Leica TCS, 4Pi, 565-568 imaging of living cells, 568 lateral scanning, 567 mitochondrial network, image, 568 optical transfer function (OTF), 567 resolution, 567 sketch,566 thermal fluctuations minimized, 567 z-response, 563 50 image space, display, 291-294. 20 pixel display space, 291 animations, 292-293 color display space, 291
efficient use, 292 image/view display options overview, table, 293 933
934
Index
5D image space, display (cont.) multiple channel color display, 292 optimal use, 293-294 pseudo-color, 173-175, 190, 291 stereoscopic display, 293 true color, 291 A Abbe, Ernst, 1,5. Abbe refractometer, 377. Abbe resolution criterion, 36, 37, 60, 61, 65-68,574,575,631-636,928. See also, Rayleigh criterion. breaking the Abbe limit, 573 calculation, 65-66 individual point features separated by, 68 pixel size, 62, 65, 634-635, 784, 928 Abbe sine condition, 151,239. Abbreviations, list, 125. Aberrations, 109,146-156,241,411-412, 471,480-481,542,629,640-641, 654,655,657-659, 747. See also, Chromatic aberrations; Refractive index mismatch; Spherical aberration. astigmatism, 145, 151-152,245-247, 249,483, axial, 242, 505, 542, 630 chromatic, 152-156,160,177-178,209, 242-243, 641, 659 in 2-photon disk scanning, 542, 550, 554 of AODs, 56 axial chromatic registration, 287, 658 chromatic registration, 657-658 of collector lenses, 657-658 intentional, for height measurement, 224 magnification error, 155,287,331,493, 542,641,657,883,904 measurement of, 243-244, 654, 659 multi-photon microscopy, 542 of optical fibers, 504, 507 signal loss, in confocal, 156, 178, 542, 641 standards, table, 157 coma, 145,151-152,245-246,249,483, 630 detecting, 241 monochromatic, 147-152,542. See also, Spherical aberration optical, avoiding with thin disk lasers, 109 of refractive systems, 146-156 signal loss, 156, 178, 542, 641 spherical, 15,34,147-149, 151, 160, 192, 208,241,244,247,330,395, 404-413,454-455,463,466,480, 542, 629, 640, 654-655, 657, 658, 728, 772, 774. See also, Spherical aberration; Mismatch, refractive index
blind deconvolution to remove, 480-481 cause of signal loss, 330, 389, 395, 413,457,542,661 chapter, 404-413 correction for refractive index mismatch, 192, 287, 411-412, 542 corrections for, 145,411-412,654-655 corrector optics, 192,395,398,411, 477,640,655,657 deconvolution, 463, 466, 468, 469, 471, 480,498-499,784 generated by specimen, 192, 418, 454-455, 654, 658, 747, 772, 775 of GRIN lens, 108 for IR wavelengths, 160 measurement using small pinholes, 145,407 monochromatic, 147-151 multi-photon excitation, 542, 407-410 PSF, 148,407,455,471,481,492,657 secondary, 247, 249 in thick embryo imaging, 747 Zernicke coefficients, 247, 248 wave-front, measuring performance, 145 Ablation,2-photon, 107,764-765. Absorber, saturable-crystal, 107, Ill, 112. to cover gap in titanium:sapphire lasers, 112 indium-gallium arsenide, InGaAs, III Absorption, 25, 163,309-312,338-339, 341,514-518,542,550,613,704. 2-photon, 405, 535-536, 541, 545, 550, 552,705,719,764,884 caged compounds, 543, 544 CARS, 595-596, 599 contrast, 162-165,211,595,610,613, 770,779 cross-section, 189, 426 energy levels, 514, 517, 682, 697, 705, 792 excited state, 544, 692 fiber optics, 501, 502 filters, 552 of fluorescent dyes, table, 345 fluorescent excitation, 45, 88 FRET, 184 and heating, 21, 218, 252, 539, 685 of incident light, 163, 177,427 by ink, 73 and laser operation, 82, 108, 110, 116 light lost by, 25, 166, 414-418, 457, 654 lighting models, 283, 285, 309-312 molar extinction, 80-81, 343, 353, 357, 793 nonlinear, 188,416,427,680,704, 709-710 of optical materials, 158 and photodamage, 22, 685-686, 690, 750 in photodetectors, 253 photon, 550, 749 quantum dots, 221, 343, 357-358, 696, 759, 801
RESOLFT/STED, 573 self-absorption, 490 spectra, 217, 267, 338-339, 345, 355, 390,415-416,421,538-539, 681-682, 706 in UV, 195 Absorption coefficient, complex specimen, 164. Absorption contrast, 164-167, 195, 427. equations, 164, 539 heating, 539, 685 Accuracy. biological vs. statistical, 24, 36-37, 68, 312 position, 39-41 Acetoxymethyl ester indicators, 726. deposits formaldehyde, 738 derivatization, 738 formula/reaction, 359, 738 loading method (AM ester), 358-359, 361, 726, 738-739, 744 painting brain slices with, 726-737 Achrogate beam-splitter/scan mirror, 50, 212,231-232,916. operation, 50, 232, 916 Zeiss LSM5 line scanner, 212, 231-232, 916 Achromat, 152, 153,244. chromatic correction of, 153 flatness of field and astigmatism, 152 longitudinal chromatic correction, 153 measurement, 244 Acousto-optical beam splitters (AOBS), 45, 55-57,88, 102,211,218,395. to select wavelength and intensity, 88, 102 to separate illumination and emission, 45, 218 Acousto-optical components, 43, 54-57. tellurium oxide crystal, 55 thermal stability, 56, 57, 219 Acousto-optical deflectors (AOD), 25, 33, 54-56,88,447,519,543,664,762, 908. as beam-splitters, to reduce loss, 33 to gate light source, 25. See also, AOM group velocity dispersion due to, 88, 540, 646 multi-photon excitation, 88, 540, 543, 646 multi-tracking, 664performance, 55 problem descanning fluorescent light, 56, 447 Acousto-optical modulators (AOM), 11, 55-57,88,231,519,540,543. FRAP experiments, for controlling laser, 56 group velocity dispersion, 88 Acousto-optical tunable filters (AOTF), 43, 55-56, 88, 102,219,237,346,543, 651,660,673,806,908. for selecting CW laser lines, 88, 102 blanking, 54, 55, 237, 389, 543, 628, 651
Index
leakage, 660 to regulate light intensity, 43 to spectrally filter light, 55 thermal sensitivity, 56-57, 219 Acridine Orange, 23,344,531,665-667, 691,774,874. bleaching, 693-694 Acronyms, list, 125. Actin filament, 7, 236, 372, 378, 383, 692, 696, 714, 719, 748-749, 753, 756, 759-760,773,781,804,819, 824-825, 854, 856. widefield source suitability, 142 Active laser medium, defined, 81. Active mode-locked, pulsed laser, 111. Actual focal position (AFP) defined, 405. Actuator, galvanometer, 52. Acute neocortical slice protocol, 723. Adams, Ansel, zone system, 71-72. Adaptive optics, 892. ADC. See Analog-to-digital converter. Adipocyte cells, CARS imaging, 604. Adjacent fields, automated confocal imaging, 810. ADU, analog digital units, 74-77, 630, 925. Advanced Visual System. See AVS. Aequorea victoria, biofilms, 348, 356, 736, 794, 873-874, 877. variants, table, 873, 874 Aequorin, Ca2+ reporter,736-737, 739, 741, 802. developmental cellular application, 736 ion binding triggers light emission, 737 Ca++ signal detection, 737 AFP. See Actual focal position. AIC. See Akaike Information Criterion. Airy aperture, optimum for NA, 28. Airy disks, 4, 24, 65, 131,145-146, 151, 156,210,443,444-449,454-456, 463-465,474,485,492-493,562, 567,630,655-657. Abbe criterion resolution, 65-66, 225 defined, 146, 444 diameter in image plane, 210, 225 four-lobed, from astigmatism, lSI image, 38, 146,225 intensity ratios, 28, 145-146 inverse, 11 and line spacing, 24 radius and pixel size, 4, 24, 38, 39, 60, 65-67,227,485 vs. NA and wavelength, 1,4, 146 Airy figure image, 38, 75, 79, 146, 147, 225,479,486-487,562. FWHM as optimal pinhole/slit size, 28, 36,225,232,443,454,463-465, 564,567-568,630-631,633, 655-657 and resolution, 65-67 size, and Nyquist criterion, 38, 39, 60 Airy unit, 28, 36, 41,210,222,227,232, 274,443-451,632,775,779. Akaike Information Criterion (AIC), 825.
AlexaFluor dyes, 81, 103, 184-185, 190, 192,236,330,342-344,353-357, 360,363,393,395,416,533,540, 694,726,731,749,794,799,804, 810,814,854,878,880,905. fluorescence excitation, 355 living cells rapid assessment, table, 360 structure, 356 Alexandrite (Cr)+ in BeAh04), tunable laser, 109. Alga. autofluorescence, 357 autofluorescent image, 173, 175, 192, 194-195,438-439,528,585,785, 870,881-885 biofilm, 870, 881-885 cell chamber for, 429 in laser cooling water, 116 Aliasing, 38-39, 271, 291, 293, 448, 588, 590-592, 640, 830, 833-834, 836-839, 903. and Nyquist criterion, 38-39, 448 temporal, 39, 41, 391, 836-837, 839 Alignment, 25, 85, 134-135, 157,505, 629-631, 651. of laser systems, to reduce instability, 85 of optical coherence tomography, 610 of optical system, thermal stress, 85 importance, 25, 630 and PSF, 646 of source, 134-135, 629-631 Alkali vapor lasers, diode-pumped, 103-105. Allium cepa. See also, Onion epithelium. Alpha blending, 302, 304. Alumina (AlzO) ceramic tubes for lasers, 102. Amira, 282-283, 286, 296, 302, 308, 312, 775-778. Amoeba pseudopod, detail, 168. Amplifier rods, maintenance, 116. Analog digitization, for photon counting, 29, 33-37,41,65,74,78,251,254, 258-261,263-264,404,460,495, 522, 525-526, 542, 634, 766. Analog-digital unit (ADU), to calibrate CCDs, 74, 77, 630, 925. Analog-to-digital converter (ADC), 31-34, 64-66, 70, 72, 74-75, 258-259, 261, 263,286,521,630-632,924-925. Analyze (software), 281-282, 288, 290, 301-304,312,651. Analyzer, in pol-microscopy, 25, 157,229. Analyzer, spectrum, 901-902. Anemonea majano, suicata, 874. Angular deflection, distortion, 211. Aniline Bluc stain, 430-432, 435, 438, 774. Animations, 281, 283-285, 289-290, 292-293,295,299,308,312,764, 829, 835-839, 841-844. Anisotropic crystals, 114.
935
Anisotropic sampling, 287-288. when resampling, 833-835 Anisotropic specimens, 163, 286, 320, 329, 420,623,675,678,690,710,793. Anisotropy analysis, chimeric proteins, 794. Anisotropy of fluorescence, 742, 794. Anisotropy of interference filters, 49. Annular aperture, 4, 9, 20, 211, 889. 3D pattern of point-source from lens, 4-20 in specimen-scanning confocal microscope, 9 Anti-bleaching agents, 36, 340, 363, 368, 375, 499, 694. Antibody stains, 292, 339, 342-343, 348, 357-360, 375, 528, 576-578, 582, 610,612,664,696,731,748,760, 789,802-804,812,852-855, 877-880. artifacts, 664 biofilms, 877-880 FRET, 790-791 high-content screening, 812-815, 818 in situ, 612 penetration, 387 preparation, 369, 371-372, 375-377,878 and TEM, 852-855 Antifade agent, 36, 340, 363, 368, 375, 499, 694. See also, Antioxidants. Antiflex optics, to reduce reflections, 158, 171,507,513. Antioxidants, living cell imaging, 341-342, 363, 389, 390, 729, 794. Anti-reflection (AR) coatings, 1, 8-9, 25, 49,117,139, 145, 151, 158-159, 212, 505-506, 901. color effect, 139 of optical fibers, 506 AOBS. See Acousto-optic beam-splitter. AOTF. See Acousto-optic tuning filter. APD. See Avalanche photodiode. Apochromat, 15, 147-148, 151, 153-155, 158,240-245,409-410,454-455, 655,659,771. chromatic correction, 153 compared with fluorite objective, 154 longitudinal chromatic correction, 153 Apodization, high-NA objective lenses, 240, 243, 249-250, 272, 567, 889. Applied Precision Instruments (API), 131, 137,282,388,651. APSS up-converting dye, saturation, 165. AR. See Anti-reflection. A. thaliana, 169,173, 174,175,193,196, 202,416,420-421,423,425,426, 427,431,771,772,773,775,778, 779,780. attenuation spectra, 416 birefringent structures in cells, 420-421. See also, Anisotropic specimens bleaching, 203 double imaging, 169 fluorescence spectra, 421, 423, 425
936
Index
A. thaliana (cont.) GFP protein fusion, 773 limitations for imaging, 772 mesophyll protoplasts, 196, 426 optical sectioning, 772, 775 protoplasts, 195-196,203,416,421, 425-427,429-430438-439,693 root tip fluorescence spectra, 173-175 seedling, autofluorescent image, 202 three-dimensional reconstruction, 190, 193, 771, 775, 777-778, 781 two-channel confocal images, 169, 175, 193, 196, 203, 427, 431, 772 two-photon excitation, advantages, 779, 780 two-photon fluorescence image, 427, 780 two-photon fluorescence spectra, 425, 426 Arc lamps, 132, 136-138. current/stability of plasma, 138-139 monitoring during exposure, 137 radiance, 137-138 sensitivity to environmental variation, 136 shape of discharge, 132 shift of wavelength with temperature, 137 stability of, vs. filament lamps, 137 Area of interest. See also, Region of interest. identifying, 201-202 Argon-ion laser, 85-86, 90-102, 107, 109-110, 112, 119, 124,203,338, 341,346,353,355,375,540-541, 655,657. CW,90-103, 107, 109-110, 112, 119, 124 emission stability, 86, 102 references, 124 Argon-krypton mixed-gas laser, 90, 92, 93, 102, 108, 119, 203, 343, 375, 748, 798, 811. Artificial contrast, vibration and ambient light, 201-204. Artificial lighting, image display, 306-312 Astigmatism, 145, 151-152, 245, 247, 249, 483, 505, 542, 630. of AOD, 914 and flatness of field, 152 and intensity distribution, 152, 246, 630 laser optics, 89, 106-107,505 measuring subresolution pinholes, 145 at off-axis points, 151,245,247,249 ATP-binding cassette, 362. ATP-buffer, 802-803, 812. ATP-caged, 544. ATP-gated cation channels, 359. Attenuation of light. by specimen, 164,287,298,304,320, 321,414-418,428,439,538,558, 706, 779, 782 plots, 415, 706 of laser beams, 85, 87, 354,415,904 modeling, 309, 311, 320-321, 330 ofPSF, 456, 462-463, 466,494 x-ray, 614-615
Atto Bioscience CARY confocal microscope, 215, 229, 230, 907. Autofluorescence, 44, 81, 90, 173, 175, 195, 202,339-340,360-361,369-370, 387,414,416,421-434,442-445, 447-449,451,509-510,528,530, 545,607,612,663,667-670,678, 682,690,698,706,710-711,713, 729, 742-743, 745, 764-765, 769-773, 779, 781-782, 785, 798, 815, 874, 876, 881-885. of alga chloroplast, 168, 172-176, 202, 429-435, 556, 785 A. thaliana seedling, 202, 303, 307, 772 bleaching, 202, 698, 729 cell wall, 303,431, 438, 770 emission spectra in plants, 176,421-423 extracellular matrix, 311 fixation, as a cause, 358, 369-370 fluorescent probes, 339-340, 360-361 harmonic signals. See Harmonic signals lamprey larvae, 612 multi-photon microscopy (MPM) See also, harmonic signals, 545 optical materials, 45, 158 plants, 190, 193-195,421-428,770-772 plots, 176, 421-423 removal using spectral unmixing, 192, 382, 664-667 examples, 665-666 removal on basis of fluorescence lifetime, 345-346, 348, 349, 528 UV excitation, 347 Automated 3D image analysis methods, 316-335. See also, Automated interpretation of subcellular patterns. biological objects, 319 blob segmentation example, 322-324 gradient-weighted distance transform, 323 model-based object merging, 323-325 watershed algorithm, 322-325, 777, 822 combined blob/tube segmentation, 328-330 data collection guidelines, 319-320 defined, 316,328 future directions, 334 hypothesis testing, 318 illustrations, 317 image preprocessing, 320-321 background subtraction, 320 morphological filters, 320 signal attenuation-correction, 320-321 VS. manual, 316-317 montage synthesis, 282, 293, 312, 328-332, 748, 753, 851-852, 855, 858-859 defined, 329-330 examples, 330-332, 780-781 neuron, 330
scanning electron micrographs, 851-852, 855 TEM implementation, 858-859 neurobiology example, 320 quantitative morphometry, 331 rationale, 316 registration synthesis, 328-331 defined, 328 landmark-based,328-329 multi-view deconvolution, 291, 330, 675-677 segmentation methods, 321-322 bottom-up, 321 hybrid, bottom-up/top-down, 322 integrated, 322 intensity threshold-based, 321 region-based, 321-322 top-down, 322 segmentation testing methods, 333-334 manual editing, 333-334 specimen preparation, 319-321 imaging artifacts, 320 stereology,316 time series in vivo images, 319 tube-like object segmentation example, 324-328 mean/median template response, 328 skeletonization methods, 324-325 vectorization methods, 324, 326, 327 types, 318-319 Automated fluorescence imaging, 814. endpoint translocation assays, 814 Automated interpretation of subcellular patterns, 818-828. See also, Automated 3D image analysis methods 2D dataset analysis. automated 2D analysis methods, 818 2D subcellular location features, 819-820 2DHeLa dataset images, 819 CHO cell dataset, 818, table, 820 Haralick features, 818-820 HeLa cells 2DHeLa dataset, 818 Zernike moments, 818-820 automated 3D analysis methods, 824 classification results, 824 feature normalization, 824 feature selection, 824 automated classification of location patterns, 824-825 classification accuracy, 826 confusion matrix for 3DHeLa images using SLFlO, table, 824 confusion matrix for 3DHeLa images using SLFI7, table, 825 features in SLFI7, table, 825 measured classification accuracy, table, 825 clustering of location patterns with clustering consistency, table, 826 exclusion of outliers, 825 methods, 826
Index
optimal clustering determination, 825-826 optimal consensus tree, 827 clustering of location patterns, 825-826 downsampled images, different gray scales, 824-825 future directions, 827-828 high-resolution 3D datasets, 820-822 3D3T3,820 3DHeLa, 820 color images from 3DHeLa, 821 image acquisition requirements, 821-822 images from 3D3T3, 821 image database systems, 827 image processing/analysis, 822-823 3D SLF, 822-823 edge features, 823 feature calculation process, 822 morphological features, 823 segmentation of multi-cell images, 822 texture features, 823 protein subcellular location, 818 statistical comparison of patterns, 826-827 AutoMontage software, 282, 293, 304. Avalanche photodiode (APD), 77, 233, 252-255,404,527,542,558,567, 698. array, for multi-beam sensing, 558 noise currents, 256 pulse pileup, 253, 527 unsuitability for non-descanned detection, 542 vacuum ADP, 254-255 Average intensity, 66, 110,516,556,668, 684,695, 747, 763-764, 816, 838, 930. equation, 302, 309, 668 AVS (Advanced Visual System), 282-283, 286,300,311-311,862,863. Axial chromatic aberration, 155, 658-659. Axial chromatic registration, 154,658. Axial contrast. See z-contrast. Axial edge response, 409-410, 654. calculations for glycerol, table, 409 calculations for water, table, 409 Axial illumination, 60-61, 134. Axial laser modes, 82, 1l0. Axial minimum, 3D diffraction pattern, 4, 147. Axial rays, spherical aberration, 148. Axial resolution, 3-4, 6, 172, 182, 209, 211, 225-228, 230, 240-241, 243-244, 320,370,395,407-411,413, 444-446,489,493,499,511,513, 551-553,559,561-568,571-577, 610-611,613,649,651,654, 656-657, 659, 674, 704, 747, 750-751,822. 4Pi microscopy, 561-568 coding, display, 305 defined, 3-4, 240,444-446
focus shift, 243, 407-410 as function of pinhole diameter, 656 magnification, 215 measurement, 194, 656-657, 659 multi-photon, 750 multiview, 678 near focal plane, slit-/point-scan confocal microscopes, 225-228 SHG,704 SPIM, 614, 674, 751 STED,571-577 tandem-scanning confocal microscope, 6, 225 tomography, 610-611 using mirror, 656-657
B Back-focal plane (BFP), 34, 50-51, 58, 61-62,84,126-128,166,208-210, 225,239,268,487,509,627,629, 708. Background light, from transmission illuminator, 201-202. Background noise, 260-262, 275. Background signal, 12,26,28, 37, 68-69, 71-72,88,90, 112, 115, 158, 162, 168,172-173,175,184,188, 201-202,221-225,227,232,235, 248,251,257,266-275,278-279, 283,287,-288,290, 301-302, 305, 312,321,326,339-340,343,345, 348,360-362,375,421,423, 428-429,432-433,442-451,462, 465,472-477,486,493,497,506, 509-510,518-519,535,541,543, 553, 559, 582, 584-585, 595, 598-600,602,604,621,633,656, 663-370, 676, 694, 697, 698, 707, 713, 727, 733-734, 736, 747, 755-757, 760, 798, 801, 803, 809, 813,815,818,822,830,836,839, 851. Background subtraction, 284, 301,320,473, 510. Back-illuminated CCD, 31, 77, 222, 232, 234,754. Back-propagation neural network (BPNN), 818. Backscattered light (BSL), 22-23, 57, 83-84,130,141,145,165,169-170, 180-182,191,196,202,212,221, 228,240,376,378,416,430,436, 442,631, 879. access to, antiflex optics, 6, 57, 141, 212, 229,507,513,609,631,704,707, 854, 879, 990 biofi1m, image, 880 contrast, effect of specimen absorption, 165 effect of coherence on, 130-131, 170 images made using, 22-23, 154,436-438, 513, 638, 855, 880 Amoeba pseudopod, 168-170, 191
937
cheek cells, 22, 23 diatom, 145,438,638-640,881 latex bead, 182, 196, 197,653 transparent ciliate protozoa, 141 LLLCD objectives/3D color-coded BSL as a noise signal, 663 optical coherence tomography, 609 practical confocal microscopy, 631 from specimen, 202 unmixing, 192, 382, 664-667 Back-thinned CCD, 31, 77, 222, 232, 234, 754. QE plot, 29 Bacteria. See Biofilms. Ballistic microprojectile delivery, 360, 726, 803. Ballistic photons, 418, 427, 538. Ballistic scans, 40, 41. Balloon model segmentation methods, 776. Bandpass, optical filters, 43-44, 46, 48, 49, 51,76,87,132,141,173,204,341, 528, 708, 798. for CARS, 598-599 coupling short and long-pass filters, 46 excitation and emission, 48, 141,217, 341, 708, 757, 798 laser, 106-107 liquid crystal, 425 to select range of wavelengths, 43-44 spectral detector, 203-204, 662-663, 666-667 Bandwidth, 32, 64, 69. 3 dB point, definition, 59, 65 of AOBS, 57 electronic/optical, digitization, 32, 34, 70, 238 head amplifier, 251 limiting, to improve reconstruction, 69 Nyquist reconstruction, output, 64, 69, 70, 238 Bead, fluorescence emission, 181, 182, 196. fluorescent, 454, 477, 493, 499, 527, 652, 653,656,659,784,900,904,930 image, 656 table, 653 glass, in water, 181, 198-199 latex, fluorescence image, 196,407, 455-457,463,471,656 in water, confocal serial sections, 182 Beam blanking, 54, 55, 237, 389, 543, 628, 651. Beam collimation, 728. for fiber delivery, 506 Beam delivery, with fiber optic coupling, 85-88, 107,216,503,506-508. Beam deviation, unintentional, 15-16. Beam expander, 8, 84, 124,208,212-214, 231,650,682,708,728,907. advantages, 213 Beam pointing, lasers, 85, 103, 107,201, 250. active cavity stabilization, 87 Beam quality, of diode lasers, 107.
938
Index
Beam shift, vignetting due to, 211. Beam-splitter, 33, 46--48, 50-51. See Dichroic mirrors. Achrogate, 50, 212, 231-232, 916 AOBS, 56-57 broadband, 346 dichroic, 25, 33, 35, 43-51, 56-57, 83-84,88,139,132,135,143,151, 162,203-204,207-208,211-214, 217-218,229,231-232,266,339, 341,346,375,386,424,469, 503-504,552,563-564,599, 630-632,647,650,657-658,664, 667,691,707-708,747,771-772, 810,846,879,910,907 table, 799 fiber-optic, 503-504 forty-five degree, performance, 47 fused-biconic coupler, 503-504 long-pass cut-off, 43, 46, 51, 175, 204, 564, 801, 875 multi-photon, 540-541 polarizing, 13,50,57,85,87, 100,217, 513,631 spectral problems, 50-51 triple dichroic, 33, 46, 48, 217-218, 658, 783 losses due to, 33 performance, 46--48 Beam scanning, along optical axis, 215, 555. Beam-scanning confocal microscope. See Confocal entries; Flying spot ultraviolet (UV) microscope. chromatic correction, 177 Beam-scanning systems, 6, 7, 16, 132, 146, 151,156,166,177,214-215,218, 381,554,562,564,567,568,599. coma in, 151 off-axis aberrations affecting, 156 Before-bleach/after-bleach ratio, FRET, 794. Benchtop fiber-optic scanning confocal microscopes, 507-508. Bertrand lens, 61, 157, 412, 643. Beryllium oxide (BeO), for laser tubes, 102. Beta barium borate (BBO), non-linear crystal for frequency doubling, 100, 109, 114-115, 125. BFP. See Back-focal plane. Bibliography, annotated, 889-899. adaptive optics, 892 books on 3D light microscopy, 889 differential phase contrast, 892 display methods, 892-883 fiber-optic confocal microscopes, 883 general interests, 891 historical interests, 889-890 index mismatch, 893-894 multiplex, 894 non-linear, 894 point spread function, 895-896 polarization, 894-895 profilometry, 895
pupil engineering, 896 review articles, 889 technical interests, 891-892 theory, 890-891 thickness, 896 turbidity, 896-897 variants on main theme, 897-899 Binding equation, for fluorescent indicators, 740. Biocytin, 730, 731. EM imaging of brain cells labeled, 731 protocol, 730 Biofilms, 287, 688, 529, 530, 624, 779, 870-887. 2-photon imaging, 530, 882-885 dual-channel imaging, 884 limitations of CLSM and 2-photon, 884 single-photon/2 photon comparison, 883 thick environmental biofilms image, 885 autofluorescence, 545 backscattered light, 880 fluorescent proteins for, table, 874 future directions, 887 GFP variants for, table, 873 imaging extracellular polymeric substances (EPS), 879-882 lectin-binding analysis, figures, 881, 882 lifetime imaging, 530 magnetic resonance microscopy, 624 making bacteria fluorescent, 873-874 pH imaging, 530, 739-745 sample mounting, 870-873 flow chamber system setup, 872-873 perfusion chambers, 870-872 pump selection, 871 upright vs. inverted microscopes, 870, 872 water-immersible lenses 149. 161,209, 411,429,568,613,727,737,870, 872.
stains for, 874-879, 875 Acridine Orange, 23, 344, 531, 665-667,691, 774, 874 antibodies, 877-878 biofilm community on tooth, 879 DAPI, 874. See also, DAPI effect of antibiotic treatment, 877 embedding for FISH, 876-877 FISH with fluorescent protein, 875-876, 878 imaging bacteria, backscattered light, 879 live/dead stain, Streptococcus gordonii, 876 nucleic acid, 874-875 preparing labeled primary antibodies, 878 SYTO, 874--875 temporal experiments, 885-886 multi-cellular biofilm structures, 886
time-lapse confocal imaging, 885-886 transmitted laser light image, 880 Bioimagers, kinetics, endpoint analysis, 816-817. Biolistic transfection, 360,724--726,803. Biological accuracy, vs. statistical accuracy, 24,36-37,68,73,312. Biological reliability, of measurements, 24, 36-37,68, 73, 312. Biological specimens, 6, 11, 12-13. See also, Plant cell imaging, Biofilms, Specimen preparation, and entries under specific equipment and cell/tissue type. backscattered light images, 22-23, 25, 167-168,170,880 CARS imaging, 603-604 adipocyte cells, 604 epithelial cells, 603 erythrocyte ghosts, 603 distortions caused refractive index inhomogeneity, 40--41, 181, 182, 198-199,419 tandem scanning systems for, 6, 11 Yokogawa CU-lO, 12-13 Biophotonic crystals, 188, 428. Bio-Rad, 25, 33, 35-36, 70, 113, 214, 260, 630, 638-640, 657, 748-752, 757, 759-762, 858, 889. 1024ES, 710-711, 714, 718-719 data storage, 585 using white light source, 113 MRC 1024, photon counting, 33 photon efficiency, 25, 32, 261, 748-752 MRC-600 scanner, full-integration digitizer, 70 PMT,260-261 Radiance-2100, 23, 185 resolution, 657 Biosensors, fluorescent, 33-8348, 799, 805. See also, Dyes, Fluorophores, and Chapters 16 and 17. future, 805 mitotic clock measurements, 799 Birefringence,6, 15,54,83, 103, 109, 113, 116, 162-164, 188, 189,414, 420--421,431,434,436,438,479, 503,710-711,714,717,894. acousto-optics, 54, 55 collagen fibers, 164, 188,717 contrast, 15, 162-164,188,414--428, 431--438,710-711,714,717,719, 894 deconvolution, 479--480 defined, 163, 188 in fiber-optics, 503 harmonic generation from, 428, 431--438 images of Cymbopetalum baillonii, 189 in laser components, 85, 103, 109, 113, 116 quarter-waveplate, 6 table, 715
Index
Birefringent crystals, 188, 420-421. optical effects of acoustic fields on, 54, 55 Black-body radiation, 44, 135-136. from incandescent lamps, 44, 126, 135-136 spectrum, 136 Bleaching, 10, 12-13, 20, 24, 44, 63-64, 90, 142,186-187, 194,202-203,210, 218,220,222,340,382-387,442, 539-540, 690-702, 797, 905, 907. 2-photon excitation, 539-540, 680-689, 905 acceleration, 341 of acceptor in FRET, 184-187 anti-bleaching agents, 36. See also, Antibleaching agents bleach patterns, 3D, 538, 628, 693 beam blanking, to reduce, 53-54 before/after ratio, for donor/acceptor pair, 794 chapter, 690-702 combining fluorescence with other, 383-386 in dye lasers, 103 dynamics, 202-203 fluorescence correlation spectroscopy, 383, 801 fluorescence lifetime, 382-383 fluorescence recovery after photobleaching, 51, 54, 56, 80, 90, 187, 210, 218, 224, 229, 237, 362, 382,684,390,691,759,801,805, 850 FRET, 186,382,794-798,800 fluorescence speckle microscopy, 383 in four-dimensional imaging, 222 improvement, recent, 36 laser trapping, 383 linear unmixing, 192, 382, 664-667 of living cells, 212, 220, 382, 797. See also, FRAP, FLIP intensity dependence, 341, 363 mechanism, 222-223 of non-specific fluorescence, 27, 44, 74 optical tweezers, 383, 385 performance limitations, 221, 224, 232, 381, 448-450, 556, 693. See Chapter 39 photoactivation, 187, 224, 383, 385, 541, 544-545, 693, 759 photo-uncaging, 383. See also, Photouncaging and signal per pixel, 63-64 spectral unmixing, 192, 382, 664-667 table, 384-385 techniques, 125 temperature as a variable, 696-698 time-lapse fluorescence, 382 Bleedthrough fluorescence, 185, 203, 664, 904. multi-tracking, reduces bleed-through, 664
Blind deconvolution, 190,468-487. See also, Deconvolution. 2D approach, 476-477 3D approach, 475-476 advantages/limitations, 468-472 algorithms, 472-474 of A. thaliana seedling image, 190 confocal stack, 470 data collection model, 472 data corrections, 477 defined, 469 DIC schematic, 475 DIC stack example, 470 different approaches, 475-477 deblurring algorithm, 476 Gold's ratio method, 476 inverse filter algorithm, 476 iterative constrained algorithms, 475-476 Jansson-van Cittert algorithm, 476 nearest-neighbor algorithm, 476 no-neighbor algorithm, 476-477 processing times/memory table, 476 Richardson-Lucy, 497, 568 TIRF microscopy, 477 differential interference contrast (DIC), 473-475 examples, 469, 470, 481, 482, 483 flowcharts, 473, 474 future directions, 483 Gerchberg-Saxton approach, 472 hourglass widefield PSF, 474 light source/optics alignment, 478 maximum likelihood estimation (MLE), 472-477, 669-670 new developments, 478-480 live imaging, 480 polarized light microscopy, 479 subpixel imaging, 478-479 optical sectioning schematic, 469 OTF frequency band, 474 simulated example, 481, 482 speed, 482-483 spherical aberration correction, 480-481, 471 spinning-disk confocal example, 481, 482,482 transmitted light, bright-field (TLB), 472, 477 two photon example, 481, 483 widefield simulated example, 481, 469 WWF stack example, 469 Blind spots, due to sampling with large pixels, 38. Blue Sky Research, ChromaLase 488, 107. Boar sperm cells, 557. BODIPY dye, 142,342-343,353-356,389, 692, 749, 760-762. BODIPY TR, methyl ester dyes, 760-762. Bolus injection protocol, 360, 726, 728, 731. Bone, reflectance, 167. Books on 3D LM, listing, 889.
939
Borohydride, to reduce glutaraldehyde autofluorescence, 374, 770. Botanical specimens, 414-439, 624, 784-785. See also, Plant cell imaging, and Chapters 21 and 44. birefringent structures, 420-421. See also, Birefringence deconvolution, 784-785 effect of fixation on, 195, 428 Equisetum, 774 fI uorescence properties, 421-428 emission spectra, 421-423 micro spectroscopy, 421-426 fluorescence resonance energy transfer, 425. See FRET, 425 harmonic generation properties, 428, 711-715 light attenuation in plant tissue, 414-418 absorption spectrum, 415 A. thaliana example, 416 maize stem attenuation spectra, 417, 418 M. quadrijolia attenuation spectra, 416 M. quadrifolia optical sections, 419 Mie scattering, 162-163, 167,417-418 nonlinear absorption in, 416-417 Rayleigh scattering, 162-163, 167,417, 703 light-specimen interaction, 425-428 living plant cell, 429-439 caIcofluor staining procedure, 424, 438 callus, 429 cell walls, 168-169, 188-189,303, 306,416-417,420-421,428-431, 435-136,438,439,710-711, 713-715,769-776,779-781 chamber slides, use, 429 culture chamber, 429 cuticle, 434-437, 715, 717, 779 fungi, 438-439, 624, 782, 870 hairs, 431 , 434-436, 772 meristem, 168,420,430,770,776-778, 783 microsporogenesis, 431-432 mineral deposits, 163, 420, 436-438, 703 pollen germination, 420, 433-434, 781, 783 pollen grains, 202, 305, 313, 420, 431-433,553,558,781,783 protoplasts, 195-196,203,416,421, 423-427,429-431,438-439,693 root, 172, 174,303,307,421,429, 430-431,438,464-465,556, 772-773, 775, 777, 779-783 starch granules, 202, 420-421, 428, 432-433,435,703,710-712,715, 719 stem, 168, 172, 180,417-419,421, 424,429,556,707,710-711, 713-714 storage structures, 435-436
940
Index
Botanical specimens (cant.) suspension-cultured cells, 189, 429-430 tapetum, 433-434, 779 waxes, 420, 428, 434-435, 714-715 point spread function in, 784 refractive index heterogeneity, 192, 418-420 maize stem, 419 Bovine embryo, 750. Boyde, Alan, 2, 6, 141, 154,224. See also, Stereoscopic images. BPNN. See Backpropagation neural network. Bragg grating, tuning diode, 107. Brain slices, 392-398, 722-734. 686. beam collimation, 728 choice of objectives, 395, 727-728 future directions, 929 image processing for, 732-734 algorithms, 733 alignment, center of mass in, 732-733 alignment, based on image overlap, 732 automatic detection of neurons, 733-734 drift/vibration compensation, 396, 732 image de-noising using wavelets, 734 image processing/analysis, 330-331, 395-396, 730-732 biocytin protocol, 730 classified using cluster analysis, 731-732 correlated electron microscopy, 731 montaging, 331 neuron reconstruction, 330-331, 730 protocol for PCA/CA, 731-732 spectral imaging, 382 two-photon/neurolucida system, 316 image production, 729 2-photon excitation, 727 deep imaging, 395 living neurons, 725 maintaining focus, 395, 732 microglia, 397-398 neuronal ensembles, 726 objective lenses, choice of, 727-728 second harmonic imaging, 729-730 in vivo observations, 387 preparation, 387 labeling cells, 394-396, 724-727 biolistic transfection, 724-725 bolus injection, 726 calistics, 726 choice of dyes, 729 diolistics, 726 dye injection/patch clamp, 726 genetic manipulation, 725-726 GFP transgenic mice, 726 Helios Gene Gun, 724 live-dead staining, 393 painting with AM-ester indicators, 726-737
photoactivation, 383 slice loading, 726 linear unmixing, 192, 382, 664-667 making brain slices, 393, 722-724 acute slices, 722-723 autofluorescence, 383 cultured slices, 724 mouse visual cortex, 723 primary visual cortex, 724 protocols, 731 thalamocortical slice, 724 photodamage, 729 pulse broadening, 728 reducing excitation light, 390-391 resolution, 729 second harmonic imaging (SHG), 729-730 silicon-intensified target (SIT) camera use, 730 slice chamber, 394 protocol, 727 speckle microscopy, 383 useful techniques, table, 384-385 time-lapse, 382 two-photon imaging, 727 calcium imaging, 729 z-sectioning, 729 Breakdown. electrical, in PMTs, 263, 660 optical, high power density, 198, 680, 682,685,687,703,705 Brewster surfaces, 83. Brewster windows, 83, 102-103, 115. Bright-field microscopy, 6, 127, 130, 20 I, 224,229,448,468,649,728. CCD for, 127, 483 deconvolution, 468, 472-473 depth of field, 4 low coherence light for, 130, 134-135, 139-140 optical projection tomography, 610-612 Brightness, source, 21, 26, 126-127, 129-130, 141-14~ 215. and exposure time, 141-142 gray levels, 71-73 as limitation of disk-scanners, 21, 215 of non-laser light sources, 126-127 of sun, 127, 135 Brillouin background, in glass fibers, 88. Brillouin effect, reduction, 110. Brownian motion, microtubules, 11. BSL. See Backscattered light. Buffering, fluorescent ion measurement, 740. Bulk labeling, in living embryos, 761.
c C. elegans, 746, 748, 766, 856, 857-858.
cryopreparation, 857-858 FRET imaging, 766 as model system, 746, 748 TEM images, 856, 857 Ca2+ imaging, see Calcium imaging.
Ca2+ indicators, 346-347,738,742-743. See also, Ca2+ sparks, 737-738, 742. discovery, 737, 738 Caenorhabditis elegans. see C. elegans. Caged compounds, 759-760. multi-photon excitation, 543-544 Calcein AM dye, 355, 360, 362-363, 426-427,430,685,804,812. Calcium imaging, 529, 545, 584, 736-737, 812. calibration, 742-743 data compression, 584 intensity image, 529 introduction, 736 multi-photon excitation, 545 ratiometric, 189 signal-to-noise ratio, 737 single-cell kinetic, 812 TIRF for measuring, 180 very fast imaging, 237 Calcium ion dyes, 183, 189,237, 736, 737, 741-743. See also, fura-2, Fluo-3 and Indo-I. Fluo-3 and Fura Red indicator system for determining, 183 Fluo-3 indicator system for determining, 737 fura-2 reactions, 741-742 Indo-1 and Fura-2 indicator system for Calcofluor, 424, 438. staining procedure, 438 Calibration, 34, 75-76, 742-745. Ca2+ sparks, 742 of CCD to measure ISF, 75-76 confocal microscopy, 742 errors in, 744 of ion concentrations, 742-745 ion interference, 745 of effective pinhole size, 34 in vitro, 742 Calistics, 726. Callus, 429. Cambridge Technology, galvanometers, 54. cAMP indicators, 347. Canna, 422, 710. fluorescence spectra, 422 as function of excitation intensity, 165 nonlinear absorption, 710 Carbon arc lamps, 136. CARS. See Coherent anti-stokes Raman scattering. CARS correlation spectroscopy (CS-CARS), 602. Raman spectra, 602 CARY disk-scanning confocal microscope, 215,226,229,230,907-908. diagram, 230, 907 CAT. See Computed axial tomography. Cathode-ray tube (CRT), 5-6, 53, 67, 72-73,291,293,588-589. gamma, compensation, 73 Cavities, of dielectric coatings, 46, 47.
Index
Cavity-dumped lasers, Ill, 114. for FUM imaging, 114 CCD. See Charge-coupled devices. CD. See Compact disks. cDNA-GFP fusion, in plants, 773. Cedara,281-282, 288, 302, 308. Cell adhesion imaging with TIRF, 90. Cell autofluorescence, 742. Cell chambers, 11,22,191,219,370-371, 386-387,394,429-430,564, 610-611. for 4Pi confocal, 564 for biofilms, 870-873, 875, 877, 880, 885 brain slice, 394, 723, 727, 729 for epithelial cells, 370-371, 377, 386 finder chamber, 683 flow chamber, 870-873, 875, 877, 880, 885 for high-content screening, 810 for optical projection tomography, 610-611 perfusion, 394 for plant cells, 191, 429-430 simple, 22, 394 for SPIM, 613, 625, 673 table of required functions, 380 table of suppliers, 388-389 test chamber/dye, 654, 661 Cell cycle, 790, 791. Cell damage, 2-photonmicroscopy, 680-688 See also, Bleaching; Photodamage. absorption spectra of cellular absorbers, 681 intracellular chromosome dissection, 688 mitochondria, 686 nanosurgery, 219, 686-687 one-photon vs. multi-photon, 680-689 by optical breakdown, 198, 680, 682, 685, 687, 703, 705 photochemical, 682-685 absorbers/targets, 682 beam power sensor, 683 impact on reproduction, 686, 685 laser exposure parameters, 682-683 NIR-induced DNA strand breaks, 683-684 NIR-induced ROS formation, 683 photodynamic-induced, 684 spectral characteristics, table, 682 photothermal, 685 reproductive effect, short NIR pulses, 682, 686 ultrastructure modifications, 685-686 Cell microarray (CMA), 815-816. Cell motility, 757. Cell nuclei, optical effects, 23. Cell pellet, three dimensional, 815. Cell surface targeting assays, 813. Cell walls of plants, 168-169, 188-189, 303,306,416-417,420-421, 428-431,435-436,438,439, 710-711,713-715,719,769-776, 779-782.
labeling, 775 viability, 780 Cell-by-cell analysis, 817. Cell-cell signaling, 778. Cellular structures, optical effects, 22-23. Center-of-mass alignment protocol, 733. Center pivot/off-axis pivot mirrors, I, 214. Cerium, doping of quartz lamp envelope, 116. CFP and YFP molecules, in FRET pair, 798-800. Chambers for living cell imaging, 388-389. commercial suppliers, table, 388-389 Charge amplifiers, 923-924. defined, 923 destructive readout, 923 FET amplifier performance, 923 non-destructive (skipper), 923 Charge-coupled device (CCD), 26-28, 30-31,39,61-62,65,70,74-78,88, 127, 137, 142,215,233,254, 458-459,460-461,482,552,558, 644, 754-755, 784, 918-931. See also, Electron-multiplier CCD. bit depth, 75 camera, 918-931 advances in, for speed, 754-755 bright-field imaging, 127 for disk scanner systems, 78, 205, 215, 220,233-235,349,459,754-755 pixel size, 62, 65, 634-635, 784 specifications, table, 929 time for sampling, 70 choosing, color, 927 computer-assisted pulse shaper, 88 confocal imaging, 458-459 cooled, advantages and limitations, 30-31 quantum efficiency, 26-28 spatial quantization of signal, 39 digital camera, 75 digital vs. video camera, 61-62 electron multiplier-CCD, 30-31, 76-77, 233-235,262,459-461,482, 925-926 multiplicative noise, 77, 234, 257, 262, 926 result, 205, 234, 755 table, 233, 459 evaluating, 927-931 array size, 928-929 "the clincher," 929 comparison, CCDIEM-CCD, table, 233,459 dynamic range vs. pixel size, table, 928 maximum signal, 930 quantum efficiency, 927-928 readout noise, 928 readout speed, 928-929 self test, 930 sensitivity, 930 shutter stability, 929
941
specifications, 927, table, 233, 929 user-friendliness, 929 gain-register, 76-78, 460-461 intensified, 930-931. See also, Intensified CCD monitoring during exposure, 137 multi-focal multi-photon microscopy, 552, 558 noise sources, 256, 924-925 charge amplifier, 925 clock-induced charge (CIC), 234, 926 fixed pattern noise, 924-925 multiplicative noise, 77, 234, 257, 262 noise vs. pixel dwell time, 922 table, 256 operation, 254, 918-927 blooming, 923 charge amplifiers, 923-924 charge coupling, 918-920 charge loss, 921 dark charge, 921-922 destructive readout amplifiers, 924 edge effects, 921 electron multiplier, 926-927 FET amplifier performance, 253, 922, 924 frame transfer readout, 920 full-frame readout, 920 gain register amplifier, 925-926 incomplete charge transfer, 923 interline transfer readout, 920 leakage, 921-922 non-destructive (skipper) amplifiers, 923-924 possible problems, 920 quantum efficiency vs. wavelength, 922 quantum efficiency, 920-921 readout methods, 920 signal level representing zero photons, 925 storage array, 920 performance, table, 256, 459, 923 piezoelectric dithering, increases resolution, 70 pixel size, 62, 65, 634-635, 784, 928 quantum efficiency and noise, 29, 644, 920,922 measuring, 74-76, 926 sensors size, parallel data collection, 142 snapshot camera, 65 specifications, described, 927-930 testing, 930 Cheek cells, backscattered light image, 22-23. Chemical environment probe, 517. Chimeric fusion proteins, 794, 801-802. anisotropy analysis, 794 cloning for FRET, 801-802 overexpression, 802 Chinese hamster ovary cell, 197, 556. 684 +, 818. Chirp, pre-compensation, 88, III, 602, 907.
942
Index
Chlorophylls, autofluorescence, A. thaliana, 175,194,203,425-426,528,711, 714,779, 782, 881. bleaching, 203 FUM,528 Cholera toxin transport, 790-791, 796-797, 802. FRET, 796-797, 802-803 Chromatic aberrations, 134, 152-156, 178, 242-245, 657-658, 659. apparatus for measuring, 243 axial chromatic registration, 243-345, 658, 657-659 of incandescent and arc lamps, 134 intentional, for color/height encoding, 154 lateral chromatic registration, 657-658 fluorescent latex bead labeled, 178 linear longitudinal chromatic dispersion, 154, 659, 664 measuring, 242-245 Chromatic corrections, 157, 177. excitation/emission wavelength, 177 tube length, table, 157 Chromatic magnification difference. See Lateral chromatic aberration. Chromatin, 385, 390, 684, 693-695, 812. Chromophores, 338-348, 543-544, 803-804. See also, Dyes; Fluorophors; Fluorescent probes etc. cellular introduction methods electroporation, 359-360, 803 microinjection, 360-361, 388, 739, 748, 755, 795, 803-804 table, 344-345, 803 transfection reagents, 358, 360, 362, 556, 682, 790-791, 795, 803 multi-photon excitation, 543-544 CIC, clock-induced charge, EM-CCDs, 234, 926. Circular exit pinhole, 9. Circular laser beam, corrective optics, 106. Classification, pattern. See Automated interpretation of subcellular patterns. Clathrin-GFP dynamics, 236. Clearing agents. See also, Mounting media. optical projection tomography, (OPT) 610,624 plant material, 166,417-420,439, 774-775 Clock, role in digitizing and reconstructing analog signal, 64. Clock-induced charge, in EM-CCDs, 234, 926. Closterium, 192-194. chloroplast autofluorescence, 192-195 signal variation with depth, 194 CLSM. See Confocal laser-scanning microscopy. Cluster analysis (CA), 731-732, 826. neurons classified using, 731-732 protocol with PCA and, 731-732 subcellular patterns, 826 CMA. See Cell microarray.
CNS, (central nervous system), 392-393, 395. See also, Chapters 19 and 41. Codecs, image processing, 831, 836, 840-841. Coefficient of variation, 660, 661. Cohen's k statistic, 826. Coherence length, 7-8, 84. defined, 7-8, 84, 130-131 reducing, for laser light, 84 Coherence surface, 84. Coherence volume, 84. Coherent anti-stokes Raman scattering (CARS), 90, 204, 550, 595-605. advantages, 204, 596 correlation spectroscopy, 602-603 defined, 595 energy diagram, 596 epi-detected, 597-599 forwardlbackward detected, 597-599 Hertzian dipole radiation pattern, 598 history, 595-596 imaging of biological samples, 603-604 adipocyte cells, 604 artificial myelin, 204 epithelial cells, 603 erythrocyte ghosts, 603 intensity distribution, 597 mapping intracellular water, 90 microscope schematic, 599 mUltiplex CARS microspectroscopy, 601-602 non-resonant background suppression, 600-601 energy diagram for multiplex CARS, 601 epi-detection, 600 phase control of excitation pulses, 600 picosecond vs. femtosecond pulses, 600 polarization-sensitive detection, 600 time-resolved CARS detection, 600 optimal laser sources, 599-600 pumped optical parametric oscillator (OPO) systems, 600 perspectives on, 604-605 unique features under tight-focusing, 596-597 Coherent illumination, 1,83-84. properties of laser light, 83-84 and resolution, I CoHagen fibers, 164, 188,313,361,393, 514,703-704,715. autofluorescence, 545 birefringence, 164, 188, 717 gels, 393 polarization microscopy, 164, 188 second harmonic image (SHG), 703-704, 715 Collector optics, elliptical and parabolic, 129. Colliding-pulse, mode-locked laser (CPM ), 540.
Colloidal gold labels, 167,241,846-859. contrast, 167 electron microscope markers, 846-857 correlative, 850, 852, 855 SEM,850 TEM FluoroNanoGold, 854 GFP related, 854-855, 857-858 measuring resolution, 241 quenches fluorescence, 854 Rayleigh scattering, 167 Colocalization, 517, 650, 667-670, 794, 813,881. FRET, FRET, 519 erroneous, 581 Color display, 291, 292. display space, 291 multiple channel display, 292 palette, 291 pseudo, 173-175, 190, 291 resolution, 291 true, 291 Color centers, in optics, avoidance, 116. Color filters, 43-52. See also, Filters. long-pass, 43-46, 175, 203-204, 212 short-pass, 45, 46 bandpass, 44, 45 Color print images, 592. Color reassignment, 173-175, 190,291. Coma, 145, 151,245,247, 249,483, 630. distortion away from optical axis, 151 observation using point objects, 145, 246 Commelina communis, images, 712. Commercial confocal light microscopes, 906-917. BD-CARY II, 230, 907 La Vision-BioTec TriM-Scope, 907 Leica, TCS SP2 AOBS, 910 Leica MP RS Multi-photon, 910 Nikon Clsi, 911 Olympus DSU, 913 Olympus Fluoview-1000, 912 optical parameters of current, table, 908-909 Visitech VT Infinity, 914 Visitech VT-eye, 914 Yokogawa CSU 22, 231, 915 Zeiss LSM 510 META optical, 916-917 Zeiss LSM-5-UVE Fast Slit Scanner schematic, 232, 916 Compact disks (CD) for data storage, 499, 586-587, 588, 731. Compact flash cards, 588. Components, of confocal fluorescence microscopes, 43-58, 207-208. acousto-optical devices, 54-57 chapter, 43-58 electroptical modulators, (Pockels cells), 25,54,57,87, 116,543, 701, 903-904 filterslbeam-splitters, 44-51
Index mechanical scanners (galvanometers), 51-54 polarizing elements, 58 Computed axial tomography (CAT), 610-611. Compression, data see, Data compression. Condenser lens, size, 129. magnification, 128-129 Configuration of pixels in image plane, 62. ConfMat. See Confusion matrix based method. Confocal disk-scanning microscope. See also, Disk-scanning confocal microscopy. Confocal fluorescence microscope, 73, 207, 404-413. See also, Confocal microscopy; Confocal laser-scanning microscopy. basic optical layout, 207 limitations due few photons, 73, 459 refractive index mismatch, 404-413 See also, Refractive index Confocal imaging, 4-5, 232, 235-236, 737, 738,746-766,809-817. See also, next major head and Chapters 35 and 36. 4Pi. See 4Pi microscopy automated for cytomics chapter, 809-817 of micro array slide, 816 platforms used for, 810 real-time, 810 temperature control, 810 types of assays for, 811, 813-814 workstations, 814 of biofilms, Chapter 50 deconvolution, 753. See Deconvolution by disk-scanning confocals, 232 fast, 235-236 of fluo-3 loaded cardiac myocyte, 737 fluorescent indicators for, 738 high-resolution datasets, cell arrangements, 776 of Ii ving cells, 813 of living embryos, chapter, 746-766 methods compared, 459, 644-647. See Chapters, 22, 23, and 24 of plants, 773. See also, Chapters 21and 43 VS. non-confocal, 746 time-lapse. See Time-lapse imaging Confocal laser-scanning microscopy (CLSM), 9-15, 32, 38, 81, 89, 118, 222-224,408,518,678,690,697, 750-751,754,884-885. See also, next major head advantages and limitations, 11-12, 222-223,644-647,884-885 alternatives to, 644-647, 754 comparisons, 644-647 disk-scanning and scanned slit, table, 224 digitizer employing full integration for, 32 edge response, 408
fluorescence lifetime imaging, chapter, 518 laser power required, 81 laser requirements for, 89 vs. multi-photon laser-scanning microscopy, 750-751. See Chapters 22, 23, 24 photobleaching, 690, 697 vs. selective plane illumination microscopy, (SPIM), 678 stage-or object-scanning, 13-15 TEM mode, 118 zoom magnification and number of pixels, 38 Confocal microscopy, 90, 141,265, 381-399,444-447,453-467, 650-670, 742, 770, 774, 779, 810, 811,815,870-887. See also, preceding major head and Chapters 35 and 36. art of imaging by, 650 automated, platforms used for, 810 balancing mUltiple parameters for, 650 of biofilms, 870-887 calibration of, 742 cell microarray and, 815 colocalization, 667-670 effect of MLE and threshold, table, 669 fluorogram analysis, 669 image collection, 667-668 nerve fiber, 669 quantifying, 668 setting thresholds, 668 spatial deconvolution in studies, 668-670 vs. deconvolution, 644-647, 453-467. See also, Chapters 22, 23, 24 CCD/confocal imaging combination, 458-459 deconvolving confocal data, 461-464, 466,488-500 fluorescence excitation, 459 fluorescent light detection, 459-460 gain register CCDs, 460-461 image sections, figures, 455, 456, 462 imaging as convolution, 453-457 integration of fluorescence intensity, 459 limits to linearity, 457 model specimens, 461 noise, 459-463 out-of-focus light, 461 point spread function, 453-457 practical differences, 458, 463-466 resolution, 459-463 same specimen comparison, 465 sensitivity, 459-463 shift invariance, 457, 490, 564 single point imaged, 454 summary of pros/cons, table, 459 temporal resolution, 458 focus positioning, 651-652 getting a good confocal image, 629-631
943
alignment of optics, 629-630 back-focal plane (BFP), 210, 509, 629, 633 focus, 629 low signal, 631 mirror test specimen, 630 no signal, 631, 660 simultaneous BSLJfluorescence, 631 high-content screening systems, table, 811 illumination sources, 126-144,650-651 See also, Lasers; Non-laser sources acousto-optic tuning filter (AOTF), 651. laser sources, chapter, 80-125 laser stability, 651 power measurement, 650-651 living cells, 381-399. See also, Living cells Minsky first confocal design, 2, 4-6, II, 141, 216, 890 monitoring instrument performance, 650-663 illumination source, 650-651 optical performance, 652-660 photon efficiency, 14-15,24 scan raster/focus positioning, 651-652 signal detection, 660-663 with non-laser light, 141 objective lens, 652-660. See Chapter 7 optical performance, 652-660. See also Chapters 7, 11 axial chromatic registration, 658-659 axial resolution vs. pinhole, 656-657. See also, Axial resolution contrast transfer function, 656. See CTF coverslip thickness and RI, table, 654 field illumination, 658 flatness of field, 659 Focal Check™ beads, 657-659 lateral chromatic registration, 657-658 lateral resolution, 655 refractive index, 654. See Chapter 20 resolution test slides, 656 self-lensing artifacts, 659 spherical aberration, correction, 654, 655 subresolution beads, 655-656 x-y and z resolution using beads, 656 optimizing multi-labeling, 663-667 bleed-through between channels, 663 control samples, establishing limits, 663 measuring autofluorescence, 663 multi-tracking, reduces bleed-through, 664 positively labeled sample, 664 reflected light contribution, 663 secondary conjugate contribution, 664 photon efficiency, 24, 26, 28, 30 33-34, 36 polarizing elements, 57 scan raster, 651-652
944
Index
Confocal microscopy (cant.) malfunctioning system, 653 phototoxicity from uneven scan speed, 651 sources of fluorescent beads, table, 653 well-calibrated system, 652 x and y galvanometers, 651-652 z-drive mechanism, 652 z-positioning calibration, 654 z-positioning stability, 652 separating signal by spectral regions for, 664 sequential collection reduces bleedtrough,664 signal detection for, 660-663 coefficient of variation, 660-661 instrument dark noise, 660 PMT linearity, 661-662 signal-to-noise ratio, 660 spectral accuracy, 662 spectral detector systems, 662 spectral resolution, 662-663 wavelength response, 663 signal level, 444-445 signal-to-noise ratio, 444-447 spectral analysis, of plants, 770 spectral unmixing, 192, 382, 664-667 limitations to, 667 overlapping fluorophores separation, 664-667 removing autofluorescence, 667 stage-scanning, 9 staining plant cells, 774 vs. structured-illumination methods, 265 vs. two-photon excitation, 779 Confusion matrix based method (ConfMat), 826. Constant output power laser stabilization, 86. Continuous wave (CW) laser, 87, 88, 90-118. beam intensity stabilization, 86-87 diode (semiconductor), 105-\ \0 output power/cooling, \08 pumped solid-state, table, 94, 95 dye lasers, 86, 103, 112, 114, 124, 540-541 fiber up-conversion, 109-110 gas lasers Argon-ion, 90, 102 Krypton-ion, 102 HeNe,102 HeCd,103 cesium and rubidium vapor, 103-105 table, 92-93 titanium-sapphire, 109 Contrast, 7, 11, 16, 37, 39, 49, 59-62, 68, 159,162-204,248,421,473,488, 542, 599-600, 607, 622, 656, 657, 675. See also, Rose criterion and CTF. absorption, equations, 164 chapter, 162-206
defined, 162 flare, 649 formation of, chapter, 162-206 fluorescence. See Dyes, and Fiuorophores as function of feature size, 16, 61-62, 37, 634 intrinsic, 633 measuring, 16,59 polarization. See Polarization microscopy second harmonic generation. See SHG and statistics, 633 third harmonic generation. See THG Contrast medium, and laser power, 80-81. Contrast method, defines signal required, 126. Contrast transfer function (CTF), 16, 35, 37-39,59-62,656,747. in confocal vs. non-confocal microscopy, 16. See Chapter 11 as function of grating period, 16 of microscope optical system, 35 relationship with objective BFP, 61 and spatial frequencies, 16, 37 and stages of imaging, 62 Control, of non-laser light sources, 138-139. Convalaria majalis, 425, 556. fluorescence microscopy of rhizome, 425 multi-focal multi-photon imaging, 556 Conversion techniques, 259-260. analog-to-digital, 259 digital-to-analog, 259-260 Convolution, a primer, 485-487. 3D blurring function, 486 Fourier transforms, 487 geometrical optics, 487 out-of-focus light, 486-487 Cooling water, checking/maintaining, 1\6-117. Cork microstructure, 770. Correction collar, (spherical aberration), 15, 145-149,158, 160-161,178, 241-242,247,377,407,410-412, 471, 654-655, 657. adjustment, 377, 407, 471, 499, 654-655 dry objectives, 410 multimedia, 640 Correctors, 70, 147. spherical aberration, 15, 151, 147, 192, 411-412 Intelligent imaging innovations, 78-79, 151,192,395,411,654 to stored data, second Nyquist constraint, 70 Corrective optics, for diode lasers, 107-108. Correlational light microscopy/eJectron microscopy, 731, 434, 436-437, 846-860. BSL image, 855 brain slices, 731then CLSM, 856-857 cryopreparation of C. elegans, 857-858 DIC image tracking, 849 DIC image/UV fluorescence image, 850
different requirement of LMIEM, 846-850 early 4D microscopy, 846 fluorescencelTEM to analyze cytoskeleton, 854 fluorescent micrographs, 851 FluoroNanoGold for cryosections, 854 GFP, 854. See also, Green fluorescent protein HVEM stereo-pair, 848-849 immuno-stained bovine aorta, 852 LVSEM of FRAPed microtubules, 849, 850 phalloidin as correlative marker, 235-236, 344, 376, 378, 694, 696, 756, 804, 854-856 phase-contrast imaging, 851 postembedding, 855 quantum dot labeling, 853 same cell structure LM/SEM, 850-852 same cell structure LM/TEM, 852-856 SEM images at 5kV and 20kV, 847, 848 TEM cross-section of C. elegans, 856 TEM longitudinal section of C. elegans, 857 tetracysteine tag labeling, 221, 348, 357, 853 tiled montage TEM images, 858 time-series DIC images, 847 Correlative LMIEM. See Correlational light microscopy/electron microscopy. Coumarin dye, 114,339, 344-345, 353, 355, 654-655, 661, 693. Counting statistics, 20, 30. See Poisson statistics. Cover glass. See Coverslip. Coverslip, and spherical aberration, 15, 147-150,201. See also, Spherical aberration. CPM laser. See Colliding pulse mode-locked laser. Crane fly spermatocyte, metaphase spindle, 15. Creep, in piezoelectric scanners, 57. Cr:Fosterite, femtosecond pulsed laser, 109, 114,415,541,706-709,712-714. Critical angle, for reflection of incident light surface of refracting medium, 167, 502. Critical illumination of the specimen, 128-129. Crosslinking fixatives, 369. Crosstalk. between fluorescence channels, 203, 424, 882 between disk pinholes, 227 between excitation foci, 553-556, 558-559,564 CRT. See Cathode-ray tube. Crystal Fiber AlS, HC-800-01 bandgap fiber, 88. CSU. See Confocal scanning unit. CTF. See Contrast transfer function.
Index Curtains, laser, safety, 118, 904. Cuticles, plant, 434-437,715,717,779. insect, 166 maize, 436 CWo See Lasers, continuous wave. Cyan fluorescent protein (Cyan), 221-222. Cyanine dyes, 339, 342, 344, 353-355, 362-363,374,443,540,587,760, 854, 874. Cytomics, 810, 811. automated confocal imaging, 810 automated confocal imaging, table, 811 Cytoskeletal structures, 24, 188, 190, 328-329, 368, 370, 372, 378, 383, 461-462,577,703,715,719, 773-774,813,846-848,852,854. LM-TEM analysis, 846, 854 stabilizing buffer, 852 Cytosolic markers, 757. Cytotoxicity, reducing, 36-37. See also, Bleaching; Phototoxicity. D DAC. See Digital-to-analog converter. Damage threshold, LED sources, 139. DAPl, 140,344-345,355, 358, 376, 431. plants, 431 use of, 376 Dark current, 29, 76, 234. fixed-pattern noise due to, 76 of photomultiplier tube, 29, 660 reducing, 234 Dark noise, defined, 232. Darkfield microscopy, 5, 7, 172,474,672. depth of field, 4 Data, 11-12,33,64,76,237. conversion from ADU to electron data, 76 degradation by multiplicative noise and digitization, 33 reconstructing, 64 speed of acquisition, 11-12 storage of volume of data, 237 Data collection guidelines, 319-320. Data collection model, blind deconvolution, 472. Data compression, 288-289, 292-293, 295, 319,499,580-585,762,764,819, 835-836. algorithms, 580 discrete cosine transform (OCT), 581 Huffman encoding, 580 Lempel-Ziv-Welch (LZW), 580 run-length encoding (RLE), 580 archiving systems, 580 gzip, 580 PKzip, 580 WinZip,580 calcium imaging, 584 color images, 581 different techniques, table, 581 Dinophysis image, 585 effects on confocal image, 583 examples, 583-585, 592, 834-837
file formats for, 580-588 fractal compression, 581-582 GIF (graphics interchange format), 580 JPEG (Joint Photographic Experts Group),581-584 MPEG, 836-839, 840-841 PNG (portable network graphic), 581, 584 QuickTime, 829, 831, 836-837, 840-844 TIFF (tagged image file format), 580 wavelet compression, 581-584, 819 movies, 836-842 artifacts, 839 compression ratios, 842-843 entrope, 841 MPEG formats, 840-841 Up-sampling, 838 pixel intensity histograms, 584 testing, 830, 835 time required, table, 581for WWW use, 816 useful websites, 844-845 Data projectors, 590. Data storage, 106. See also, Mass storage. Data storage systems, 287, 395. 580, 594, 764. chapter, 580-594 characteristics of 3D microscopical data, 287 databases, 861-869. See Databases random access CDR, CDRW, 586-587 DVD,587 Magnetic disks, 586 semiconductor, FLASH memory, 588 for remote presentation, 842 role for STED, 577 Databases, 20/30 biology images, 827, 861-869. benefits, 863-864 fast, simple machine configuration, 863 improved analysis and access, 863 performance, 863 remote monitoring, 863 repeatability of experiments, 863 submissions to other databases, 863 criteriairequirements, 866-867 digital rights management, 867 metadata structure, 867 query by content, 866-867 user interface, 866 dataimetadata management, 861-862 future prospects, 867 image database model, 864-865 image information management, 862 image management software, table, 865, 868 instrument database model, 864 laboratory information management systems (LIMS), 862 microscopy dataimetadata life cycle, 863
945
modern microscopes design aims, 862-865 projects, 865-866 BioImage, 865-866 Biomedical Image Library (BIL), 866 Scientific Image DataBase (SlOB), 866 recent developments, 86 I -862 MPEG-7 format, 862 relational database management systems (RDBMS), 862 TIFF format, 861 software for, 868-869 ACDSee, 868 Aequitas, 868, 869 Cumulus, 868 Imatch, 868, 869 iView, 868, 869 Portfolio, 868 price, 868 Research Assistant, 868 ThumbsPlus, 868, 869 system requirements, 864 DBR. See Distributed Bragg reflector. OCT. See Discrete cosine transform. Deblurring algorithm, 476. Deconvolution, 7, 26-28, 39,40,66, 189-190,222-223,278,456-458, 464,468,488-500,542,564,736, 746, 751-753, 778, 784-785, 828, 864, 900, 929. See also, Blind deconvolution. of 2-photon images, 498 and 3D Gaussian filtering, 70, 281, 285, 323, 392, 395, 667. See also, Gaussian 4Pi lobe removal, 562, 565 advantages and limitations, 458, 475 algorithms, 472-476, 490, 495-497, 751, 778 comparison, 497-498 iterative constrained Tikhonov-Miller, 497 Jansson-van Cittert, 496 nearest neighbor, 495-496 non-linear constrained iterative, 496-497 Richardson-Lucy, 497, 568 Weiner filtering, 496 background history, 488-490 blurring process contributions, 488 equation showing restoration possible, 489 image formation, 489-490 schematic diagram of convolution, 489 blind, 189-190, 431, 463, 469, 472-473, 478,486,492,496-497,646 chapter, 468-487 maximum likelihood estimation, 472-477,483.669-670 blurring process contributions, 488 confocal data, 39, 40, 453-467, 488-500, 753, 778. See also, Confocal microscopy, vs. deconvolution.
946
Index
Deconvolution (cont.) of simulated confocal data, 40 CARS data cannot be deconvolved, 397, 399 chapter, 453--467 colocalization, 668-670 comparison of methods, 66, 453, 467, 475--477, 497-499,644-648 convolution primer, 485--478 convolution and imaging, 490--491 Fourier transform of PSF, 489, 490 linearity, 490 optical transfer function, 490--491 shift invariance, 457, 490, 564 and data compression, 584-585 test results, 401, 461, 464--466, 481--482, 483 defined, 189-190,468 display of data, 301, 830, 835-836 examples, 40, 190, 392, 411, 462, 466, 471,488--498,510 4Pi, 468, 565 botanical specimens, 784-785 brightfield, 411, 475, 478 cardiac t-system, 498 confocal, 470 DIC,470 polarized, 479 of simulated confocal data, 40 STED, 574-576 flatfielding the data, 477 black reference, 76 white-reference, 76 fluorescence lifetime imaging, 521 four dimensional deconvolution, 391-392, 752 Fourier transform of PSF, 489, 490 future directions, 483, 766 and image formation, 490--492 linearity and shift-invariance, 457, 564 live imaging, 480, 564, 751-754 missing cone problem, 494 model specimens, 461, 464--466, 481--482, 483 multi-photon, 488-500, 542 multi-view montaging, 330, 677 ion imaging, 736 noise, 495, 635 and Nyquist reconstruction, 59, 65, 67, 68, 222-223, 635 suppressing Poisson noise, 39 optical sectioning, 752 out-of-focus light, 26-28, 431, 487, 644 and pinhole, 26, 487 point-spread function (PSF), 223, 241, 247,453,463,471,489--492,635, 655 approximations, 493 measuring PSF, 492--494 and Poisson noise reduction, 320 pre-filtering, 281, 497, 581 problem with specimen heterogeneity, 22, 648
purpose, 468 requirements and limitations, 489--494 diagram demonstrating convolution, 489 linearity, 490 missing cone problem, 494 noise, 495 optical transfer function, 490--491 point spread function, 489, 492--494 shift invariance, 457, 490, 564 sampling frequency, 635 spherical aberration, 471, 480--481 stain sparsity, 28 structured illumination, comparison, 265-279 subpixel refinement, 478--479 temporal/spatial, 392, 458, 753 transmitted light imaging, 472, 475, 478 of wavelength spectra, 382, 663-667, 771-772 limitations, 667 Deconvolution lite, 68-70. Deflector, acousto-optical. See Acoustooptical deflector. Defocusing, size and intensity distribution, 146. Degree of modulation, 268-270. locally calculated, 268-270 absolute magnitUde computation, 268-269 equations, 269 homodyne detection scheme, 268-269 max/min measured intensity difference, 268 scaled subtraction approach, 269-270 square-law detection, 268-269 synthetic pinholes, 268 Delamination, and interference fringes, 168-170. Delivery, dye, 355, 357-360, 810. Deltavision, 132, 282. Demagnification, and numerical aperture, 127. Depth discrimination, in LSCM. See Axial resolution. Depth of field, 4, 9, 13. extended-focus images, 9 fluorescence microscopy, 4 phase-dependent imaging, 13 Depth-weighting, projection images, 304, 306. exponential, 304 linear or recursive, 304 Derived contrast (synthetic contrast), 188-201. Descanned detection, 166, 208, 212, 537, 540-542, 754, 904. Design of confocal microscopes, 43, 145, 166,207-211,237. See also, Commercial confocal light microscopes. 4Pi, 563, 566
of confocal fluorescence microscope, 208 efficiency in, 43 fast-scanning confocal instruments, 237 intermediate optical system, 207-209 of microscope optics, 145 MMM, 552, 555 practical requirements, 210-211 of transmitted confocal microscope, 166 Detection efficiency, 34, 35, 210-211. measurement, 34-35 practical requirements, 210-211 Detection method, multi-photon, 541, 542. descanned, 542 Detectors, 9, 11, 25, 28, 251-264. See also, Photomultiplier tube; Chargecoupled device, etc. area detectors. See Image detectors assessment of devices, 260-262 CCD,254 noise vs. pixel dwell time, 922 comparison, table, 255-256 conversion techniques, 259-260 descanned, 208, 212, 537,540-542, 774, 904 direct effects, 252 errors, 211-212 evaluation, 211-217 future developments, 262-264 history, 262-264 image dissector, 254-255 image intensifier, 13,232-233,519-520, 524,555-556 gated, 233, 519-522, 524, 555-556 intensified. See Intensified CCD MCP-PMT. See Microchannel plate microchannel plate, 232-233, 255, 262 MCP-CCD, 262 gated, 519, 523-524, 527, 532 noise internal, 256-259 internal detection, 256 noise currents, table, 256 photoemissive devices, 256-257 photon flux, 257-258 pixel value represented, 258-259 non-des canned, 185,201,218,381,447, 456,507,542,552,559,643,646, 727, 750, 779, 904, 909-910 phase-sensitive, 518-520, 619 photoconductivity effects, 252, 253 photoemissive, 254 photography. See Photographic systems photomultiplier tube, 9, II. See also, PMT photovoltaic effect, 252-253 photon interactions in, 252-256 point detectors, 260-261 quantal nature of light, 251-252 quantum efficiency (QE) vs. wavelength, 25 for second harmonic detection, table, 707 silicon-intensified target (SIT) vidicon, 730 spectral, 203-204, 662-663, 666-667
Index TCPSC, 518, 520-523, 526 time-gated, 522 thermal effects, 252 work functions, table, 252-253 vacuum avalanche photodiode, 254, 255 Developmental biology, 545, 624. multi-photon microscopy (MPM), 545 Dextran labeling, 173-174,292,512,757. DFB. See Distributed feedback. 4',6-diamidino-2-phenylindole, 140, 344-345,355,358,376,431. See DAPI. plants, 431 use of, 376 Diatom, 438, 638-640, 881. as standard for measuring objectives, 145 test specimen, 638-640 Ole. See Differential interference contrast. Dichroic filters, 212. intensity loss, 212 transmission, 212 Dichroic mirrors (beam-splitters), 44, 45, 50-51,129,211,217-218. coating for collection mirrors, 129 double and triple, 217-218 effect of deflection angle, 211 separating emission/excitation, 44--45, 50-51 Die, of light-emitting diode, 133, 134. Dielectric butterfly, galvo feedback, 54. Differential interference contrast (DIC) imaging, 10,14,76,127,146,171, 453,468,473-475,846. blind deconvolution, 473-475 converting phase shifts to amplitude, 171 narrow bandpass filter use, 76 Nomarski DIC contrast, 2, 268, 746, 892 photon flux reduction, 127 schematic for, 475 three dimensional, 470 Wollaston prism, 156,468,473,475 Diffraction, 61, 65. contrast transfer function, 16, 35, 37-39, 59-62, 656, 747 and sharpness of recorded data, 65 Diffraction limit, 210-211. See also, Rayleigh criterion. defined, 210 point-spread function, 146 practical requirements for, 210-211 Digital light processor (DLP), projectors, 590. Digital memory system, 64. Digital microscopy, optics/statistics/ digitizing, 79. Nyquist sampling, 146 Digital printers, 591-593. Digital processing, in disk-scanning confocal, 12. Digital projectors, 590. Digital rights management (DRM), 830, 844. Digital video disks (DVD), 587-588.
Digital-to-analog converter (DAC), 64, 259-260. operation, 64 Digitization, 25, 31-32, 36, 38-39, 59, 62-63,66, 72, 75, 79, 259, 261, 286, 460, 495, 639, 911. aliasing. See Aliasing blind spots, 38 and Nyquist criterion, 38-39 precision, 25 and pixels, 62-63 of voltage output of photomultiplier rube, 31-32 Oil derivatives, 760. Dimethylsulfoxide (DMSO), 697, 726-727, 760,875. handling, 739 DIN standard, microscopes, 156. Dinophysis image, 585. Diode injection lasers, 105-108. Diode lasers, 86, 87, 107, 112, 116. distributed feedback, 107 emission stability, 86 intensity, 87 maintenance, 116 modulated, 112 noise sources, 86 physical dimensions, 106 violet and deep blue, 107 visible and red, 107 wavelength stabilization, 87 Diode-pumped alkali lasers (OPAL), 103-105. Diode-pumped lamp (DPL), 108-109. Diode-pumped solid-state lasers (DPSS), 108-109,111,112. kits, companies offering, 109 passively mode-locked, 111 ultrafast, 112 Diolistics, ballistic gene transfer, 726. Dipping objective, 149. 161,209,411,429, 568,613, 727, 737, 870, 872. Direct permeability. 358-359. Discrete cosine transform (DCT), 581. Disk-scanning confocal microscopy, 215-216, 224, 225, 228-229, 234-235 754, 755. advantages and limitations, 223-224 for backscattered light imaging, 228-229 chapter, 221-238 commercial instruments, 907, 913, 915 comparing sing1e- vs. multi-beam, 224 table, 226 and electron multiplier CCDs, 78, 205, 215, 220, 233-235, 349, 459, 754-755 embryo, 754 high-speed image acquisition, 216, 222-224, 754 image contrast in, 168-171 microscopes. table, 224 optical sectioning, 235 types, 228-232
947
Dispersion, optical, 56, 88, 152, 154,242, 411,542-543,609,683. in acousto-optical devices, 3, 15,55-56, 88 CARS signal generation, 728 compensation, 566-567 defined, 152 in fiber lasers, ultra-fast pulses, 88, 110, 113 by filter blank material, 211 generates third harmonic signal, 704-705 group delay dispersion, 537-538, 543 group velocity dispersion, 88, 111, 210, 537,609,903 in optical coherence tomography, (OCT), 609 in optical fibers, 502, 504, 507 and temperature, 15,411 for multi-channel detection, 51 using to correct for chromatic aberration, 153 Display software. See Presentation software. Displays, 580, 588-590, 594, 892. cathode ray tube (CRT), 5-6, 53, 67, 72-73, 291, 293, 588-589 data projectors, 590 digital light processor (DLP), 590 halftoning vs. dithering, 589 international television standards, 589 liquid crystal (LCD), 589-590 supertwisted nematic (STN), 589 thin-film transistor (TFT), 589 monitors, 588-589 Distortion, 39-41, 152. and resolution, practical, 39-41 Distributed Bragg reflector (DBR) diode laser, 107. Distributed feedback (DFB) diode laser, 107,113. ultrafast, 113 Dithering vs. halftoning display, 589. DLP. See Digital light processor. DMSO, 697, 726-727, 760, 875. handling, 739 DNA damage, 390, 517, 539, 680, 682-684, 812. DNA probes, 273, 317, 339, 343, 354, 358, 360,362,369,393,396,459,520, 531-532, 539-540, 691-695, 774, 779,782,812,818-825,828,874. DAPI, 140, 344-345, 355, 358, 376, 431 DRAQ5,343 Hoechst, DNA dye, 136, 339, 344, 360, 362, 520, 565-566, 683, 782 DNA sequencing, constructs for, 801-802. DNA transfer, 724, 756, 760, 773, 790, 802-804. Dominant-negative effects, 755. Donor/acceptor pair (FRET), 790, 792-794, 796-797. See also, FRET. before bleach/after bleach ratio, 794 equations, 790, 792-794
948
Index
Donor/acceptor pair (FRET) (cant.) fluorescence, 796-797 fluorophores, 794 separation in nm, table, 793 Double-image, diagram and example, 169. Double-label, 375. Down-conversion, parametric, 114. DPAL. See Diode-pumped alkali lasers. DPL. See Diode-pumped lamp. DPSS. See Diode-pumped solid-state lasers. Drift, 386-387, 652, 655, 732. CCD read amplifier, 76 compensation for, 392-393, 732-733, 886 focus, 16,40, 115, 190, 219, 386, 489, 567, 652, 720, 729, 886 compensating, 396, 732 lasers, 85-86, 115 DRM. See Digital rights management. Drosophila, 273, 675-676, 747-748, 751-752, 754, 756, 759, 804, 810. living embryo, 675-676, 752 salivary chromosomes, 273 SPIM, image, 675-676 Duty cycle, laser, defined, llO. DVD, 587-588. Dye lasers, 86, 103, 112, 114, 124,540-541 in cancer treatment, 112 colliding-pulse mode-locked, 112 with intra-cavity absorbers, 112 noise and drift, 86 references, 124 as wavelength shifters, 103 Dye-filling, studying micro-cavities, 173-174. Dyes, 22-23, 36,44,90-102, 109, 116, 118, 165,173,183,212,222,342-346, 353-358,360,430,461,462,527, 528, 575, 726, 736-738, 740-745, 748, 749, 755, 759-760, 774, 775, 782, 804. See also, Green fluorescent protein (GFP); Rhodamine dyes; Fluorescein. affect on living cells, 391, 748 AlexaFluor, 353-355 Aniline Blue, 430-432, 435, 438, 774 APSS and Canna yellow, non-linearity, 165 bandwidth of emission, 44 BODIPY TR, methyl ester, 760 BOPIDY, 142, 342-343, 353-355, 389, 692, 749, 760-762 Calcein AM, 355, 360, 362-363, 426-427,430,685,804,812 calcium dyes, 346-347 cAMP, 347 characteristics of probes/specimen, table, 344-345, 354-355 coumarin, 114, 339, 344-345, 353, 355, 654-655,661,693 cyanine, 339,342,344, 354-355, 362-363,374,443,540,587,760, 854, 874
diI derivatives, 355, 362, 389, 726, 760 donor acceptor pair, 794. See also, FRET DNA probes, 343-344, 531-532, 818-825, See also, DNA probes DRAQ5,343 dyes vs. probes, 353, for embryos, 748, 761 exciting efficiently, 44 fade-resistant, 36 See also, Antifade; Bleaching for fatty acid, 347. See also, FM4-64, below Feulgen-stained DNA, 166, 200, 298, 433,437 Fluo-3 and Fura Red, for calcium, 180, 183,345,434 Fluo-3 for calcium, 737 fluorescein, 353, 355. See Fluorescein; FITC fluorescence lifetime, 517, 527-528 FluoroNanoGold,854 FM4-64, FMI-43, lipophilic dyes, 236, 355, 359-360, 389, 556, 755, 760-761 fura-2, 103, 189, 234, 257, 345, 346, 348, 358-359, 361, 529, 531, 726-727, 730, 733, 741-743, 810, 812, 846, 850 Fura Red, 180, 183, 345, 454 future developments, 348-349 genetically expressed, 348 Glutathione, 342, 358, 545, 694, 779, 782 hazards in using, 116, 118 for ion concentration, 346-347 ion-sensiti ve probes, table, 531 kinetics, 741-742 lanthanum chelates, 345-346 laser/filter configuration, table, 799 lineage tracers, 461 lipid dyes, 236, 355, 359-360, 389, 556, 755, 760-761 living cells, rapid assessment, table, 360 loading, uniformity, 749. See also, Loading LysoTracker Red DND-99, 360 membrane labels, 344-345 membrane potential, 205, 346 microinjection, 360-361, 388, 739, 748, 755, 795, 803-804 MitoTracker Red, 142, 170,353,358, 360,430-431,692,750 multi-photon excitation, 543-544 nano-crystals, 343, 345. See also, Quantum dots Nile Red, 435, 528, 575, 774, 782 organic, 342-343, 353-356 oxygen sensor, 347 patch clamp loading, 360, 726, 734, 738-740 pH indicator, 346, 739-745. See also, pH imaging
photoactivatable, 187,210,224,383, 385, 541,544-545,693,729,759-760, 912 Kaede, 187,383,385 Kindling, 574, 760 PA-GF~ 187,383,385,752,759-760 photodestruction, 340-341. See also, Bleaching and Chapter 39 photophysical problems, 338-340 absorption spectra, 339 autofluorescence, 339-340 contaminating background, 339-340 optimal intensity, 340 RayleighlRaman scattering, 339 singlet state saturation, 338-339. See saturation, below triplet state saturation, 339 phycobiliproteins, 338, 341, 343, 355-357,693 for plants, 774-775. See Chapters 21 and 44 two-photon, 782 propidium iodide, 344, 355, 360, 426, 651,693-695,773,778-779,782, 812, 875, 877 quantum yield, 172, 180, 184, 338-845, 347,353-354, 360, 363, 383, 421, 543-544, 574, 661, 683, 690-692, 710, 737, 792, 794-795 ratio methods, 346-348, 742-743 rhodamine, 353, 355. See also, Rhodamine excitation, 109 saturation, 21-22, 41, 142, 222, 265, 276,338-340,448,643,647,899 Schiff-reagent, 262, 369, 770, 774-775, 778 selection criteria for, 353-358 signal optimization strategies for, 341-342 SNARF, 345-346, 531, 739, 744-745 specimen damage, 340-341 spectral properties, 212, 342, 344-345 spectral unmixing, 192,382,664-667 for STED, table, 575 Dynamic Image Analysis System (DIAS), 396-397, 783-784. living cells of rodent brain, 396 of plant cells, 783-784 Dynamic range, 929-930.
E e2v Technologies, EM-CCDs, 76-77, 233-234,237,262,460,925-926. E-CARS. See Epi-detected CARS. ECL. See Emitter-coupled logic. Edge detector (software), 309, 322, 327, 396, 823-826. Edge effect, self-shadowing, 172. Edge-emitting diode laser, 89, 106. corrective optics for, 89 cross-section through, 106
Index
Efficiency, laser, 102, 105-106. See also, Quantum efficiency (QE); Photon efficiency. of diode injection lasers, 105-106 wall-plug, of argon-ion lasers, 102 EFIC. See Episcopic fluorescence image capture. EGS. See Ethylene glycol-bis-succinimidyl. E-h. See Electron-hole. Electro-magnetic interference, in electrooptical modulators, 57. Electron microscopy, 167. brain slices, 730-731 chapter, 846-860 cryo-techniques, 854 fixation, 167, 368-369 immuno-stained, 371-372, 852 micrographs, 479, 847-853, 855-858 tomography (EMT), 610-611 Electron-multiplier CCO (EM-CCO), 30-31, 74-75,78,142,233-235,262, 466-467,482,647,678,737, 753-754, 784, 923-926. advantages and disadvantages, 30-31, 220,228,233-235,237,459-460, 647, 737, 909, 923-926 CIC, clock-induced charge, 234, 926 and disk-scanners, 76, 205, 215, 220, 270 frame-transfer, 262, 234 gain-register amplifier, 76-77, 258, 753, 925 interline-transfer, 233-234 mean-variance curves, 78 multiplicative noise, 77 noise currents, 256 parameters, vs. normal CCO, table, 233 QE(effective), 78, 927 readout amplifier, 76-77, 258, 753-754, 925 results, 235, 237, 755 Electron-beam-scanning television, 6-7. Electron-hole (e-h) pairs and photon counting, 29. Electronic bandwidth, 64-65. See also, Bandwidth. Electronic noise, defined, 232. Electronik Laser Systems GmbH, VersaOisc, 109. Electrons, interaction with light, 129-130. Electro-optical modulators (EOM), 25, 54, 57,87, 116,543, 701, 903-904. Electroporation, 359-360, 795, 803. for chromophores, 803 Ellis, Gordon, 2, 3,7,8,13,14,84,129, 131,478,507. Embryo imaging. See Living embryo imaging. Embryos, 761-766. bulk labeling, with dyes, 761 depiction, in time and space, 762-764 dyes, for multi-wavelength analysis, 756 FRET, 764-766
labeled proteins, 756 photobleaching, 759 transcriptional reporters, 756 EM-CCO. See Electron-multiplier CCO. Emission filter. See Filters. Emission spectra, of arc sources, 136, 176. Emission spectra, fluorophores, 1- VS. 2photon excitation, 421. Emitter-coupled logic (ECL), 259. EMT. See Electron microscopy tomography. Endomicroscopy, 511, 513, 514. distal tip for, 514 fiber-optics, 513 human cervix image, 513 human gastrointestinal track image, 514 miniaturized scanning confocal, 511 Endoplasmic reticulum, 374, 770, 819. and OiOC6, 390 FLIP, 382 FRET, 795 genetic fluorescent probes, 771,783 and harmonic signal generation, 703 in ion-imaging, 738 and phototoxicity, 685 table, 363 Endpoint data analysis, 816-817. Endpoint translocation/redistribution assays, 814. Energy diagram, lasers, 102, 105, 106. argon-ion laser, 102 helium-cadmium laser, 105 helium-neon laser, 105 semiconductor laser, 106 titanium:sapphire four-level vibronic laser, 109 Energy, of single photon, 35, 127. Energy transfer rate, for FRET, 790, 792. Entrance aperture. See Back-focal plane. EOM. See Electro-optical modulators. Epi-detected CARS (E-CARS), 597-599. erythrocyte ghosts, 603 Epi-fluorescence microscopy. See Fluorescence microscopy, 44, 166, 172-173, 195,202,235. Epi-illuminating confocal microscope, 9, 166. See also, Confocal laser scanning microscopy; Confocal microscopy. Episcopic fluorescence image capture (EFIC), 607-608. mouse embryo image, 608 Epithelial cells, 14-15,603. CARS image, 603 oral, optical sections, surface ridges, 14-15 EPS. See Extracellular polymeric substances. Erythrocyte ghosts, CARS imaging, 603. Ester-loading technique. See Acetoxymethyl esters loading method. Ethylene glycol-bis-succinimidyl (EGS), 369. Euphorbia pulcherrima, spectrum, 710.
949
European Molecular Biology Laboratory (EMBL), 53, 212. compact confocal camera, 212 Evanescent waves, 90, 177, 180,245,503, 801. defined, 90, 180 optical fibers, 503 resolution measurement, 245 Excess light. See Stray light. Excimer lasers, 112, 116. maintenance, 116 for tissue ablation, 112 Excitation efficiency, multi-focal multiphoton microscopy, 552. Excitation filter, requirements, 44. See also, Filters. Excitation source, laser. See Lasers; Nonlaser sources. Excitation wavelength change, contrast, 173. Explants, for imaging living embryo, 748-749. Exposure time, 62, 65, 71-76, 81, 127, 137, 141-142, 176,212,219,224,226, 231-236, 267, 270, 276, 346, 363, 392-393,423,427,459-460,477, 495,556,613,627-628,651,655, 681-686,692-697,708,746-747, 753-755, 760-764, 783-784, 816, 822, 850-851, 873. for CCOs and EM-CCOs, 127, 137, 141-142,231-236, 267 disk scanners, 231-235 laser, safety, 117-118, 839, 900, 903-904 reducing, 753-755 and source brightness, 141-142 total, comparison of methods, 442, 449 UV,116 x-ray, 614-616 External laser optics, maintenance, 117. External photoeffect. See Photoemissive effect. External Pockels cell, 25, 54, 87, 116, 543, 701,903-904. External-beam prism method, laser control, 90. Extracellular polymeric substances (EPS), 183,311,358,376,703-704,717, 760, 783, 870, 879-880. See also, Collagen. bleaching, 693 damage, 685 dye, 361 lectin-binding in biofilms, 870, 879-880 matrix, 760 negative contrast, 173 in optical projection tomography, 612 plants, 438, 783 preparation, 376 Extrinsic noise, reduction, 21. F
Fabry-Perot interferometer, optical cavity, 81-82.
950
Index
Fast Fourier transform, 487. to identify interference fringes, 202 Fast line scanner, 231-232. Fatty acid indicator, 347. FBG. See Fiber Bragg Grating. FBR. See Fiber Bragg Reflector. FBTe. See Fused biconical taper couplers. F-CARS. See Forward-detected CARS. FCS. See Fluorescence correlation spectroscopy. Feedback, 136, 139. for control of light-emitting diode, 139 to increase source stability, 136 Femtosecond pulsed lasers. See Ultrafast lasers. Feulgen-staining, DNA, 166, 200, 298, 433, 437 Fianium-New Optics, Ltd., FemtoMaster1060 fiber laser, 113-114. Fiber Bragg Grating (FBG), laser stabilization, 87. Fiber Bragg Reflector (FBR), stabilizes laser, 87. Fiber lasers, 85, 101, 109-110, 113-114, 124. defined, 109-11 0 temperature sensitivity, 85 tutorial reference, 124 ultrafast, 101, 113-114 Fiber optics. See Chapter 26. beam-splitters, 503-504 Bow-tie, pol-preserving fiber, 503 cable, for delivering ultrafast pulses, 88 laser output, 106 pigtail, 106 Fiber optics used in microscopy, 501-507. evanescent waves in optical fibers, 503 fiber image transfer bundles, 504-505 fiber-optic beam-splitters, 503-504 fused biconical taper couplers, 503-504 glass made from gas, 501 gradient-index optical fibers, 501-502 key functions of fibers, 505-507 delivering light, 505-506 detection aperture, 506 diffuse illumination, 507 for femtosecond laser pulses, 507 large-area detection, 507 large-core fibers, as source/detection apertures, 507 same fiber for source and detection, 506 single-mode fiber launch, 505 SMPP optical arrangement, 216 managing insertion losses, 506 angle polishing of fiber tips, 506 anti-reflection coating of fiber tips, 506 index matching of fiber tips, 506 microstructure fibers, 504 modes in optical fibers, 502 polarization effects in optical fibers, 503 polarization-maintaining fibers, 503
step-index vs. gradient index, 502 step-index optical fibers, 501-502 transmission losses in silica glass, 502 Fiber-optic confocal microscopy, 501-515, 893. benchtop scanning microscopes, 507-508 clinical endomicroscopy, 513 distal tip, 514 human cervix image, 513 human gastrointestinal track image, 514 image transfer bundles, 504-505 managing insertion losses, 506 miniaturized scanning confocal, 508-512 bundle imagers for in vivo studies, 509 with coherent imaging bundles, 508-509 imaging heads, 508-512 objective lens systems, 509 optical efficiency, 509 optical schema, 508 resolution, 509 rigid endoscope, 511 vibrating lens and fiber, 510-511 in vivo imaging in animals, 510-514 Fiber-optic interferometer, 240-241, 504, 609. diagram, 241 for measuring point spread function, 240-241 Fiber-optic light scrambler, 8, 13, 131-132, 143. Fibroblasts, 292, 361, 691, 798, 803, 852. Field diaphragm, 34-35,127-128, 139,461, 627,648-649. Field effect transistor (FET) CCD amplifier, 30-31,77,922-927,929. noise vs. pixel dwell time, 922 Filament-based lamps, 34,44, 126-132, 135-138, 346, 507, 648, 663. fiber optic, 507 image, 100 W halogen bulb, 135 size, 126-127 spectrum, 44, 136 stability, 34, 137 File formats, multi-dimensional images, 288-289. Fill factor. of CCD, 920-921, 927, 929 disk-scanning microscopes, 224-228, 233, 552 Filtering, digital, 281, 810. See also, Deconvolution. Gaussian, 41, 65. See also, Gaussian filters multi-dimensional microscopy display, 281 nonlinear, deconvolution, 190 sets, for automated confocal imaging, 810 smoothing, effect on contrast, 59 to reduce "noise" features, 70 Filters, optical, 43-51, 70, 89, 162, 190, 212,753. See also, Heat filters.
conventional, 45 hard vs. soft coatings, 45-49 intensity loss, 212 interference, 45-51 conventional and hard coatings, 46 multi-channel detection, 51 ND filters, 43, 89 notch and edge, 50 tuning with angular dependence, 50 to select image contrast features, 162 short-pass, interference type, 46 transmission vs. laser line, 212 types, 46 wavelength selective, 43-51 FiRender,281-282. First or front intensity, projection rule, 302, 304. FITe. See Fluorescein isothiocyanate. Fixation, specimen, 368, 378, 428, 852, 854, 856. antibody screening with glutaraldehyde fix, 377 artifacts, 195,369-373,428,624,815, 854, 857 autofluorescence, 358, 663 borohydride to reduce autofluorescence, 374,770 chapter 368-378 characteristics, 368-370 chemical fixatives, 369 crosslinking fixatives, 369 freeze substitution, 369, 769, 854-856 microwave fixation, 369 protein coagUlation, 369 cryo-fixation, 854 dehydration, 166, 368,417-418,481, 611,623-624,815,849,854-855 effect on plants, 428 for electron microscopy, 167,368-369, 372,479,731,851-860 ethylene glycol-bis-succinimidyl, 369 evaluation, 371-374 cell height to measure shrinkage, 371-373 MDCK cell example, 372, 373 formaldehyde, 369-370, 373 general notes, 374-378 geometrical distortion, 372-373, 815 GFP, 854, See also, Green fluorescent protein arsenical derivatives, 348 glutaraldehyde, 369, 370 high-content screening, 815 immunofluorescence staining, 371, 372, 852 improper mounting, 376 microwave, 377-378 mounting methods, 370-374 critical evaluation, 371-374 media refractive index, table, 377 technique, 371 optical properties of plants, 428 pH shift/formaldehyde, 370-371, 373
Index
plants. See also, Botanical specimens, Plant cells, 428, 769-770, 773-774 refractive index of mounting media, table, 377 optical effects, 428 refractive index of tissue/organs, table, 377 shrinkage, 369-373, 624, 815, 854 staining, 370-371 tissue preparation, 376 Fixed wavelength lasers, table, 119-120. Fixed-pattern noise, 74-76, 278, 924, 927, 931. Flare, out-of-focus light, 6, 132,157-158, 172,395,456,465-466,469,471, 481,649,731. Flatness of field, 145, 151, 154,418,457, 639. measurement! small pinholes, 145, 457, 639 objectives, to improve, 151-152 Flat-fielding CCO data, 76,477. black reference, 76 white-reference, 76 Flexible scanning, 51-52. FUM. See Fluorescence lifetime imaging microscopy. FUP. See Fluorescence loss in photobleaching. Flip mirrors, to control laser, 58. Floppy disks, 586. Fluorescein,48, 80-81, 88, 203, 261, 353-355,375,443,582,697,781, 794,930. arsenical derivatives, 348 calculating laser power needed, 80-81, 443 derivatization, diagram, 354 double-labeling, 375 filters for, 48 photobleaching quantum yield, 363 rhodamine and, FRET between, 794 Fluorescein isothiocyanate, 88, 198,203, 261, 263, 335, 375, 394, 397-398, 511-512,527-528,582-583, 693-694,781,794,799,884,885. See also, Fluorescein. 2-photon, 781 biofilms, 884-885 dextran, 292, 512 filter sets, 48-49 FRET, 794, 799 lifetime, 527-528, 532 photobleaching quantum yield, 363 toxicity, 391, 693-694 Fluorescence anisotropy measurements, 742. Fluorescence contrast, 172-173. Fluorescence correlation spectroscopy (FCS), 5, 363,383, 385,602, 801, 803,805,917. and CARS, 602 FRET,801
laser requirements, 81 table, 385 Fluorescence emission, botanical specimens, 425-428. 1- VS. 2-photon excitation, 421 Fluorescence imaging, deconvolution vs. confocal, 459-460, 644-648. Fluorescence in situ hybridization (FISH), 316-317,319,323,331,333-334, 343, 875-878. biofilms stains, 875-878 with fluorescent protein, 878 Fluorescence ion measurement, 736-738, 740-745. See also, Calcium imaging, pH, etc. calcium imaging, 736-737 concentration calibration, 742-745 indicator choice, 738 interpretation, 740-741 pH imaging, 346, 739-745 water-immersion objectives, 737 Fluorescence lifetime imaging microscopy (FUM), 108, III, 114, 139, 204, 233, 382-383, 385, 516-533, 799-801. advantages, 766, 800 alternatives to, 766 analysis, 251 applications, 516-518, 527-532 calcium imaging, 529 chemical environment probe, 517 FRET, 517-518 ion concentration, 517, 528-530 multi-labeling with dyes, 517, 527-528 pH imaging, 529-530 probes, 517 table, 530-532 comparison of methods, 523-527 acquisition time, 525-526 bleaching, 524 cost, 526 detector properties, 526-527 mUlti-exponential lifetime, 523-524 photon economy, 524-525 pile-up effect on detection efficiency, 526 shortest lifetime, 523 table, 526 decay process of excited molecule, table, 518 frequency domain methods, 518-520 disk-scanning implementations, 520 phase fluorometry method, 518-519 point-scanning implementations, 520 widefield, spinning-disk, 519-520 frequency-domain, 108 reducing repetition rate, III FRET, 799-801 history, 516 Jablonski diagram, 516, 517, 697, 792 with light-emitting diode sources, 139 limitations, 800 living cell images, 204
951
methods, 518-527 comparison, 523-527 frequency domain, 518-520 time domain, 520-523 multi-focal multi-photon microscopy, 555-556 quantitative fluorescence, 517-518 quantum efficiency, 516 spectroscopy, 516 table, 385 time domain detection methods, 520-523 point-scanning, 522 streak camera, 520 TCSPC FUM, 522-523 time-gated FUM, 523 use of intensified CCOs for, 233 Fluorescence loss in photobleaching (FUP), 187, 382, 384, 801. FRET, affected by, 801 table, 384 Fluorescence microscopy, 4, 9. 13,43-44, 154,166, 172-173,195,202,235, 251, 448-451, 809-810 See also, Widefield (WF) fluorescence microscopy. chromatic correction, 154 compared to disk-scanning microscopes, 235 vs. confocal imaging, 13 depth of field, 4 filters for selecting wavelengths for, 43-44 folded optical path, 166 increase contrast with less intensity, 172-173 signal-to-noise ratio comparative, 448-451 bleaching-limited performance, 448-450 configurations of microscope, 448, 449 disk-scanning microscope, 449 line illumination microscope, 449 saturation-limited performance, 450 scanning speed effects, 450-451 SIN ratios, table, 450 wide field (WF) microscope, 450 spectral problems, 44 Fluorescence, quenched by colloidal gold, 854. Fluorescence recovery after photobleaching (FRAP), 51, 54, 56, 90, 187,210, 218,224,229, 237, 362, 382, 384, 390,691,759,801,805,850. in biofilms, 874 damage to cellular structure, 341, 859-851 damage to microtubules, 341, 850-851 efficiency of illumination light path, 210 related to TEM of same specimen, 850-851 setups for, 218, 907 table, 384 using CARY2 disk-scanner, 229, 907
952
Index
Fluorescence resonance energy transfer (FRET), 26-28, 34, 184-187, 204, 218,221-222,382,384,425, 517-518,556,650,691,741-742, 764-766, 788-806, 796-797. based on protein-protein interactions, 800 based sensors, 798-799 botanical specimens, 425 C. elegans, 766 chapter, 778-806 cloning and expression of fluorescent constructs for, 801-804 donor/acceptor pair, 790, 792-794 donor, 796-797 efficiency, 792 experimental preparation, 795 FCS and, 801 FUM and, 799-801 between fluorescein and rhodamine, 794 fluorescence lifetime imaging, 517-518 fluorescent proteins, 794-795 FRAP and, 801. See also, Fluorescence recovery after photobleaching future perspectives, 805 induced by cholera toxin transport, 797 intramolecular, 765 kinetics, 741-742 in living cells, 195-186, 204 chapter, 788-806 in living embryos, 764-766 MMM,797-798 nanobioscopy of protein-protein interactions acceptor bleach for, 797-798 donor fluorescence for, 796-797 measurement methods for, 795 sensitized emission of acceptor, 795-796 photobleaching, 691 practical measurements, 792 probes, 221-222 quantum dots, 801 setups, 218 small molecules, 794-795 spatial orientation factor, 792-793 spectrofluorimetry, 793 spectroscopic properties used for, 795 standards for, 34 table, 384 theory, 790-794 TIRF and, 80 I total, measured with widefield, 26-28 in transgenic animals, 765 wavelength depiction, 793 Fluorescence saturation, singlet-state, 21-22, 41, 142,265,276,339,448,643, 647,899. Fluorescence speckle microscopy (FSM), 13, 383, 385, 889. table, 385 Fluorescent biosensor, 799, 805. future, 805 mitotic clock measurements, 799
Fluorescent constructs for FRET, 801-802. cloning of fluorescent chimeras, 801-802 expression and over-expression, 802 functional activity of expressed, 802 Fluorescent dyes. See Dyes; Fluorescent indicators; Fluorescent probes. Fluorescent efficiency, 34. Fluorescent emission, incoherence, 130. Fluorescent indicators, 346-348, 736-743. See also, Fluorescent probes, and particular ions. binding equation, 740-741 buffering, 740 calcium imaging, 736-737 calibration, 742-743 indicators, 738 cellular introduction, 738-739. See also, Loading cellular trapping, 738 choice, 738 concentration, 741-742 dialysis, 740 free diffusion, 741 genetically expressed intracellular, 348 green fluorescent protein, 348 ion indicators, 348 ligand-binding modules, 348 handling, 739-740 inaccurate measurements, 740-741 intracellular parameters imaged, 346-348 Ca2+, 346-347 cAMP, 347 fatty acid, 347 ion concentrations, 346-347 membrane potentials, 346 other ratioing forms, 347-348 oxygen, 347 pH, 346, 739-745 wavelength ratioing, 346 positive pressure, 740 selectivity, 743 Fluorescent intensity (IF), TIRF, 180. Fluorescent labels, 342-346, 530-532, 761, 775. See also, Dyes; Fluorescent probes; Chapters 16-17, and by name of dye. Fluorescent probes. 353-364, 387-389, 517, 530-532, 736-737, 739-740, 755, 769,771,773,783,806,810,811. See also, Dyes, Fluorescence indicators and by name of dye, Chapters 16, 17. automatic living cell assays, 811 bound, 737 care, 739-740 characteristics, table, 344-345, 354 development, 736 dye criteria for, 353-358 AlexaFluor dyes, 353-355 BOPIDY dyes, 353-355, 749, 760-762 coumarin dyes, 353, 355 cyanine dyes, 353, 374, 587, 760, 854, 874
dye classes, table, 355 dye vs. probes, 353 fluorescein, 353, 355. See also, Fluorescein fluorescent proteins, 355-357 GFP, 355-357. See, Green fluorescent protein indicators of intracellular sate, 346-348 Ca 2+ indicators, 346-347 protein multi-photon excitation, 357-358 phycobiliproteins, 355-357 probes/specimen characteristics, table, 354 quantum dots, 357 rhodamine, 342-345. See also, Rhodamine excitation, 737, 344-345 for fluorescence lifetime imaging, 517, 530-532 genetically encoded, for plant imaging, 769, 771, 773, 783. See also, Transcriptional reporters; Transfection agents for high-content screening, 810 high specificity/high sensitivity, 806 living cell imaging, 387-389 rapid assessment by, table, 360 loading methods, 358-360. See also, Loading acetoxymethyl esters, 359 ATP-gated cation channels, 359 ballistic microprojectile delivery, 360, 724-725,802-803 direct permeability, 358-359 electroporation, 359-360, 795, 803 microinjection, 360-361, 388, 739, 748, 755, 795, 803-804 osmotic permeabilization, 359 peptide-mediated uptake, 359 transient permeabilization, 359 whole-cell patch pipet delivery, 360, 726-727, 734, 738-740 photoactivatable, 210, 224, 383, 385, 541, 544-545,693, 759-760, 912 Kaede, 187,383,385 Kindling, 574, 760 PA-GFP, 187,383,385,752,759-760 photobleaching, 362-363. See also, Bleaching phototoxicity, 363-364 See also, Phototoxicity factors influencing, table, 363 specimen interactions, 361-362 cytotoxicity, 362 localization, 361-362 metabolism, 361-362 perturbation, 362 target abundance/autofluorescence, 360-361 tissues, 360 Fluorescent proteins, 187,355-357,739, 794-795.
Index
emission change after photo damage, 187 FRET, 794-795 genetically engineered variants, 739 ion binding regions, 739 Fluorescent lights, stray signal, 20 I, 632, 904. Fluorescent staining, 371, 393, 438, 774. See also, Dyes; Staining. immunofluorescence, 371, 372, 852 living cells, 393 microglia, 319-320, 393-398 nuclei of living or dead cells, 393 Fluorite (CaF2), optical to reduce chromatic aberration, 153. FluoroNanoGold, cryosections, 854. Fluorophores, 44, 338-349, 543-544, 664-667, 748, 794, 799. See also, Dyes, Fluorescent labels. Flying spot detector for measuring photon efficiency, 34-35. Flying spot ultraviolet (UV) microscope, 6-7. Fly's-eye lenses, for diode lasers, 107-108. FM4-64, FM 1-43, and other lipophilic membrane dyes, 236, 355, 359, 360, 389, 556, 775, 760-761. Focal Check™ beads, 657-659. Focal-plane array detection, 2-photon, 542. Focal shift for mismatched RI, 405, 407-410,553. defined, 405 dependence, 410 for glycerol, table, 409 for water, table, 409 Focus, 3-4, 13, 36, 197. for confocal microscope, 36 displacement, by living cell specimen, 22-23 effect of coverslip, 197 extended, 9 in phase-dependent imaging, 13-14 planes, diagram, 27 position, confocal microscopy, 651-652 Focused spot. See Point spread function. Folded optics, for trans-illuminated confocal microscopy, 166. Formaldehyde, 369-370, 373-377, 428, 738. AM-loading releases formaldehyde, 738 fixation protocol, 371 permeabilization agents for, 375 pH shift method, 370-371, 373 for plants, 428 stock solutions, 370-371 Ftirster distance, defined, 184, 790, 792, 793. Forster equation, 184, 790, 793. Forster resonance cnergy transfer. See also, Fluorescence resonance energy transfer.
Forsterite laser (Cr~ in MgSi04 ), 109, 114,415,541,706,707-709, 712-713. second/third harmonic generation, 114 tunable, 109 Forward-detected CARS(F-CARS), 597-599, 603. erythrocyte ghosts, 603 Foundations of confocal LM, chapter, 1-19. Four-dimensional images, 746-749, 752, 761-764. advantageous techniques, 746-747 automatic image analysis, 321 deconvolution, 495 embryogenesis visualization strategies, 761-764 living cells, 393 of living embryos cellular viability, 747-748 challenges, 762 dataset display strategies, 393, 763-764 deconvolution, 752 for large thick specimen, 746-747 photobleaching during, 747-748 photodamage during, 746 required datasets for, 746-747 multi-photon, 535 structured illumination, 482 SPIM,676 Fourier analysis. 4Pi microscope, 563, 576 analogy with image reconstruction, 69 of blind deconvolution, 472-476, 478 and convolution, 485-487 of image formation, 446, 454, 456-457 MRM, 618-620 of periodic test specimen, 638-639 of short laser pulses, 88, 728 SPIM multiview processing, 675-677 STED,574 of structured-illumination images, 268, 270-273 and wavelet processing, 734 Fourier plane. See Back-focal plane, 20 I, 245,509. Fourier space, 270-271. Fourier transform, 201, 202, 271, 487, 489, 490-492, 620. of AC interference in image, 201-202, 651 and convolution, 487 and deconvolution, 487, 490-492 for detecting stray light into detector, 201 identifying interference fringes, 202 of microtubule TIRF image, 183 missing cone problem, 494 MRM image formation, 620 of point spread function, 489, 490 Fractal compression, 581-582. Frame rate. See also, Speed in confocal microscopy, II matching, 838-839
953
FRAP. See Fluorescence recovery after photobleaching. Free diffusion, of fluorescent indicators, 741. Free-ion concentration, 742. Freeze thawing, 731, 739. Frequency, 52, 65, 82. laser vs. pumping power, 82 of resonant galvanometer, 52 of sampling clock, 64 Frequency doubling. See Second harmonic generation. Frequency-resolved optical gating (FROG) for pulse length measurement, 115. FRET. See Fluorescence resonance energy transfer. Frustrated total internal reflection, defined, 177. FSM. See Fluorescence speckle microscopy. Full-well of CCD pixel, defined, 75. Full-width half maximum (FWHM) resolution. 4Pi, 562, 567 of beams in scanning disk, 554 of CARS, 597, 599 of confocal performance, 656-657, 661-662 of emission wavelength LED, 136 quantum dots, 343 of interference filters, 44 laser bandwidth, 93, 95, 100, 101 laser pulse length, 109, 112,507,537, 538, 902 micro-surgery precision, 219, 687 multi-photon, 682-683, 901-902 objective resolution (PSF), 149,209,225, 444-445,456,492,509,552,571 PMT rise time, 225 resolution, with spherical aberration, 407 table, 409 SPIM,675 STED, 572, 576-578 z-resolution, measured, 194 Fundamental limits, chapter, 20-42. Fungi, 438-439, 624, 782, 870. Fura-2 [calcium ion] indicator dye, 103, 189,234,257,345,346,348, 358-359,361,529,531,726-727, 730, 733, 741-743, 810, 812, 846, 850. Fused bi-conical taper couplers (FBTC), 503-504. Future, 143-144, 160, 192,219-220,234. of EM-CCD with interline transfer, 234 of laser-scanning confocal microscopes, 219 of non-laser light sources, 143-144 spherical-aberration corrector, IS, 147, 151,192 of tunable objective, 160 FWHM. See Full-width half maximum.
954
Index
G Gain, 31, 232. of image intensifier, 232 photomultiplier tube, from collisions at first dynode, diagram, 31 Gain register, (EM-CCD) 76-78, 233-234. CCD (CCD), 76-78 of electron multiplier-CCD, 233-234 Gain setting, 75, 115. defined,75 effect of bandwidth on, 115 GAL4 genes, 773. Gallium arsenide (GaAs). diode laser, 107, 111 InGaAs photodiode, 707-708 LEDs, 133, 138, 143 PMT photocathode, 4, 28-29, 232, 252, 255,263,464,527,931 Galvanometer, 11,25,36,40,51-54,56,57, 63,211,215,223,231-232,513, 543,552,558,599,651-652,753, 806,907,910-911,914,931. See also, Linear galvanometers. defined, 52-54 distortion, 211 electromechanical properties, 40 errors, 40 in fiber-optic micro-confocal, 513 figure, 63 line-scanner, 231-232 linear, 52, 53, 223 measurement, 651-656 multi-focal, 554 multi-photon, 543 resonant, 25, 52-54, 56-57, 223, 447, 510, 539, 543, 552, 558, 910 specifications for, 214, 543 ultra-precise, 211 x-y scanners, 213-215, 223, 651-654, 806,907,910-911,914 Gamma, brightness non-linearity, 72-73, 287,832-833. data projector, 590 display, 582-583, 589, 832-833 Gas lasers, 86,90-105. See also, CW lasers; Pulsed lasers. continuous wave, 90-105 maintenance, 116 noise sources, 86 pressure, 102 Gating, intensified CCD, 25, 233, 262, 522, 555. Gaussian beam profile, lasers, 80-81, 83-84, 108-109, 111, 113, 116,231,269, 338, 456, 496, 502, 538-539, 554, 891. in CARS, 597 converted into line, 231, 916 fiber optic, 502, 505, 506 filling back-focal plane, 210, 509, 629, 633
"Gaussian-to-flat-wavefront" converter, 554 Kerr effect produces self-focusing, 111 laser beam profile, 538-539, 554, 597, 635-636 noise, 473, 497, 925 from optical fiber, 502, 505-506 optical tweezers, 89. See also, Laser trapping spatial filter, 89, 729 Gaussian filters, digital, 39, 41, 65, 70, 89, 281,285,301,323,338,391-392, 399,497,499,510,650,667-668, 676, 729, 734, 753, 764, 830. of 3D data to reduce Poisson noise, 39, 41,65,69-70,269,281,285,323, 391-392,399,499,510,635-636, 650, 667-668, 676, 764, 830 "Gaussian blob," 635-636 and Nyquist reconstruction, 65 in presentation displays, 830 results, 285, 676, 733, 835-837 Gaussian laser pulses, 536-536, 902. Gaussian noise, 473, 497, 925. Gaussian norm statistical tests, 830, 835, 837. GDD. See Group delay dispersion. Gene gun, 360, 724-724, 730. Geometric contrast, 180-187. Geometric distortion, 6, 23, 36, 39-41, 53, 152, 211, 215-216, 265, 297, 329, 372-373,448,480,590,641, 653-654, 741, 835. kinetic, 741 measurement, 651-656 projector, 590 of specimen preparation, 372-373, 815, 872 Gerchberg-Saxton algorithm, deconvolution, 472. GFP. See Green fluorescent protein. Ghost images, from transmission illuminator, 201-202. GIF (Graphics interchange format), 580. Gires-Toumois interferometer (GTI), to reduce GVD, 88. Glan-Taylorpolarizer, 85, 87,100,171. in single-sided confocal microscope, 171
Glan-Thompson polarizer, attenuator, 85, 904. Glutaraldehyde, fixative, 369, 369-374, 377-378,428,438,731,852. antibody screening with, 377 autofluorescence of, 374, 428, 770 fixation protocol, 370 stock solutions for, 370 Glutathione (GSH), 342, 358, 545, 694, 779. visualization, in plant cells, 782 Glycerol, immersion/mounting medium, 404,407,409-410,435,563,654, 698, 785. clearing, 198, 200 diffusion in, 698
immersion objective lenses, 412, 563, 567 example, 785 mounting media, 371, 373, 375, 377-378, 420 RI-mismatch, table, 409, 410 Goggles, laser, for eye protection, 118. Gold's ratio method, 476. Golgi receptor, 374, 389, 556, 564-566, 791. Golgi stain, 107, 283, 298. Gourard shading, 308, 309, 311. Gouy phase shift, 597. Graded index (GRIN) lenses, 84. in diode lasers, 108 Gradient index optical fibers, 501-502. Gradient-weighted distance transform, 323. Graphics interchange format. See GIF. Grating, periodic. GVD compensator, 88, 504, 538, 686 laser tuning, 90, 103, 106-107, 111 minimum spacing, 1, 16,652 OCT phase-delay, 609 pulse compressor, 113 spectral detector, 87, 346, 422, 664, 772 structured illumination, 266-267, 273 Gray levels, 71-76. intensity spread function, 74-76 printer, 592 Green fluorescent protein (GFP), 90, 174, 221-222,348,355-357,429, 478-479, 556, 568, 571, 612, 614, 625, 675-676, 690, 692, 698-699, 724-725, 727, 731, 741, 747-752, 755, 756-763, 766, 769-773, 781-785,798-806,812-815, 820, 854-859, 862, 873-875, 877-879, 885. See also, Transfection reagents; Transcriptional reporters. biofilms labeling, 873 or CFP molecules, as FRET pair, 798 constructs, in embryos, 756 EM imaging, brain cells, 731, 854-859 FRET, 793-795, 798-803 image contrast, 174 limitations, 760 membrane localized, 749 methods with Correlative LMIEM, 854 mice, 727 photoactivatable, 187,383,385,752, 759-760 photobleaching, 690, 692, 698 for plant imaging, 424, 429-430, 769-773, 781-785 direct visualization, 773
genetic fusions, 773, 783 genetic marking, 773 two-photon excitation, 782-783 protein fusions/cytoskeleton, 773-774, 801 tagged proteins, 758 TIRF,90
Index
FRET,794 Grey levels, 71-76. printer, 592 GRIN. See Graded index. Ground state depletion (GSD), 573. Group delay dispersion (GDD), 537-538, 543. Group velocity dispersion (GVD), 88, III, 210,537,606,609,903. in optical coherence tomography, 609 pulse broadening due to, 88, 111,210, 537-538,543,606,609,728, 903 GSD. See Ground state depletion. GTI. See Gires-Tournois interferometer. Guinea-pig bladder, calcium sparks, image, 237. GVD. See Group velocity dispersion. Gzip, 580. H
Hairs, plant, 431, 434-436, 772. Halftoning vs. dithering, 589. Halogen lamps, 126-127, 132, l36-139, 143, 159,663. brightness V.I'. temperature, 136 filaments, 132 image, 135 lifespan, 136 power available, 126-127 stability plot, 137 Haralick features, 818-820. Hard coatings, for interference filters, 45, 48. Hard copy, 580, 590-594. photographic systems for, 590-591 printers, 591-593 aliasing, 592 color images, 592 digital,591-593 grey levels, 592 ink jet, 593 laser, 593 posterizing, 591 scaling techniques, 592 Harmonic signals, 2, 49, 80, 90, 100, 109, 113-114, 162-163, 174,179-180, 188,243,361,414,428,535,545, 550,556,577,596-597,682, 703-704, 708-719, 722, 729, 734, 894 See also, Second harmonic generation; Third harmonic generation see Structured illumination. chapter, 703-721 contrast, 179-180, 188 descanned detection, 56 in lasers, 109, 113, 114, 115 plants, 428 second and higher, 114 Haze, from out-of-focus light, 227. HBO-50 mercury-arc bulb, 126. HCS. See High content screening.
Heat, 84-85, 89-90,109,129,133. filtering, dichroic filters, 43-44, 129, 132 heat sink for LED light source, 133 from laser cooling, 84-85, 109 of optical trap, 89-90 placing system components, 129 Heat filters, to exclude IR light, 43-44, 129, 132. liquid, 132 Heating. See also, Thermal variables. detectors, 252 microwave fixation, 377 in magnetic resonance imaging, 621-622 multi-focal, multi-photon, 551, 556, 685, 903 specimen, by the chamber, 387-389, 394, 732 specimen, by the illumination, 43, 89, 132,211,218,341,536,539,544, 556,621-622,681,685, 884, 903 calculation, 89, 685, 904 stability, 652 HeLa cells, 391-392, 693, 799, 812, 814, 820, 828, 854. Helios Gene Gun System, 724. Helium-cadmium (He-Cd) laser, 83, 86, 90, 93,103, 105, 115. operational lifetime, 115 output variation, 86 transverse electromagnetic mode, 83 Helium-neon (He-Ne) laser, 82, 84, 88-90, 93,102-103,105,107,240,241, 376, 673, 680, 798, 799, 864, 875. four state, 82, 105 Heterectis crispa, 874. Hidden-object removal, 304-305. High content screening (HCS), 809-817. for cytomics chapter, 809-817 data management/image informatics, 816-817 fluorescence analysis of cells, table, 812 multiple fluorescent probes, 810 High resolution spatial discrimination, 813. High throughput screening (HTS), 809. High voltage electron microscope (HVEM), 846. stereo images of platelets, 848-849 Hippocampal brain slices, 268, 316-317, 393, 556-557, 722, 724-725, 727. calcium imaging, 556-557 culture protocol, 724-725 damage, 341 at neurons, 205, 268, 316-317, 393 Histology, 623, 624. Historic overview of biological LM, table, 2-3. Hoechst, DNA dye, 136, 339, 344, 360, 362, 520, 565-566, 683, 782, 812. 4Pi, image, 565-566 FUM image, 521 high-content screening, 812, 814 Holey optical fiber/non-linear effects, 88.
955
Holographic diffusers, to reduce coherence, 84. Holography, holomicrography, 7-8. Hooke, Robert, image of cork, 769-770, 785. HTS. See High throughput screening. Huffman encoding, 580-581. Human endomicroscopy, confocal. cervix, 513 gastrointestinal track, 514 Human retina, viewed with OCT, 609. Huygens, 3D software, 104, 4l3, 669, 778. Huygens-Fresnel wavefront construction, 406. HVEM. See High-voltage electron microscope. Hybrid mode-locked dye laser, 540-541. Hymenocallis speciosa, fluorescence spectra, 422. Hysteresis. in Piezoelectric scanners, 57, 754 temperature cycling of lenses, 249 I
ISM, (Incoherent Illumination Image Interference Imaging), 275, 561, 569-570, 672. optical transfer function (OTF), 569-570 ICNIRP. See International Commission of Non-Ionizing Radiation Protection. ICTM. See Iterative constrained TikhonovMiller algorithm. IEC. See International Electrotechnical Commission. IF. See Fluorescent intensity. Illumination, 44, 210. See also, Structuredillumination microscopy, and Chapter 6. brightness, table, 140 errors, 211-212 evaluating, 211-217 goal in confocal microscopy, 210 path, 211-212 types of lamps, 44 vignetting caused by beam shift, 211-212 Image(s),9, 11-12,30-31,38-39,59, 145, 192,210,219,280,286-290. See also, Multidimensional microscopy images. contrast, 7, 11, 16,39,49,60-62,68, 159, 162, 165, 167, 173-175, 180, 189-190,192,201-204,248,421, 473,488,542,599-600,607,622, 656, 657, 675 chapter, 162 flare, 649 definition, 280 degradation of, measuring, 145 extended-focus, 9 motion between specimen and objective, 39 multi-dimensional microscopy, 286-290 anisotropic sampling, 287
956
Index
Image(s) (cant.) calibrating image data, 286-288 contrast transfer function (CTF), 61. See CTF data type/precision in computations, 288-289 digitization, defined, 62 dimensions, 286-288 display devices, non-linearity of, 72-73 file formats, table, 288-289 processor performance, 289-290 Voxel rendering speed, 290 real, disk- and line-scanners, 30-31 reconstructing, and noise reduction, 38-39. See also, Reconstruction; Nyquist reconstruction sharpness of vs. signal intensity, 192 of source and detector pinholes, 210 speed of acquisition, 11-12. See also, Speed as sum of point images, 59 thermal distortion, 219. See also, Thermal variables Image analysis. See Automated 3D image analysis methods; Automated interpretation of subcellular location pattern. Image dissector, 254-255. in trans-illumination mode, 10 Image enhancement. See Deconvolution, 488-499. Image iconoscope, for television, 6-7. Image intensifiers, 13, 232-233, 235, 255, 460,477,519-520,522,524, 555-556,730, 737, 784, 801, 930. Image Pro Plus, 282, 290. Image processing. See also, Automated 3D analysis methods, and Multidimensional microscopy display. for display, Chapter 14 for measurement, Chapter 15 Image resolution, 8, 9. See also, Resolution. Image substrate, automated confocal, 810. ImageJ, free software, 282, 290, 395, 732-733, 762-764, 795, 858. Imaging system, optics characterized by CTF,61. Imaging techniques, 382-386, 394-395. combining fluorescence with other, 383-386 fluorescence correlation spectroscopy, 383 fluorescence lifetime (FUM), 382, 516-532 fluorescence loss in photobleaching (FUP),382 fluorescence recovery after photobleaching, 382 fluorescence resonance energy transfer, 382 fluorescence speckle microscopy (FSM), 383 laser trapping, 383 linear unmixing, 192, 382, 664-667
multi-channel time-lapse fluorescence, 382 optical tweezers, 383 photoactivation, 187,224,383, 385, 541, 544-545,693,759 photo-uncaging, 383. See also, Photouncaging physiological fluorescence, 383 spectral, 382 table, 384-385 time-lapse fluorescence, 382 Imaris, software, 193, 281-282, 284, 287-288,290-291,299,301-303, 308,311-312, 764, 795. In vitro fertilization, mitotic apparatus, 188. In vitro preparations. 20 mixed-cell, assays, 813 antifade agents. See also, Antifade, 342 automated analysis, 318-320 backscattered light image, 513 biofilms, 870, 872, 879, 884 bleaching, 551, 851 brain slices. See Brain slices, 392-393, 725 cell maintenance, 387 cytoskeleton, 368 fertilization, 188 GFP,357 high content screening, 809, 813-816 high speed imaging, 11,237, 809, 813, 815-816 ion imaging, calibration, 742 living cell imaging, 387 micro-CT, 614, 617 micro-MRI, 618, 621, 623-625 multi-photon, 535 optical coherence tomography image, 609 photodamage, 684 In vivo (intact animal) imaging, 112, 368-377,512,545,806. 2-photon microscopy (MPM), 535, 543, 545 cell preparations, 387 comparison with fixed material, 368-377 FUM calibration, 517 labeling, 372-373 miniaturized confocal, 504, 508, 511-513 micro-CT, 614, 617 micro MRI, 618, 621, 623-625 molecular imaging, 806 photodamage, 684, 693-694, 698 "stick" lenses, 806 Incandescent lamps, 34, 126, 133-137,477, 499 See also, Halogen lamps. black-body radiation emitted by, 135-136 spectrum vs. temperature, 137 stability, 137,477 Incidence angle, 49, 50. efficiency, 143 interference filters/transmission, 49 reflectivity, diagram, 50 Incident light beam, sample interaction, 162-163.
Indo-I, calcium indicator, 103, 189,257, 345,346,348,529,531,544,693, 697, 742-743. Infinity corrected optics, 155-157, 166,239, 405. advantages, 156-157, 166,239,405 Infinity PhotoOptical, InFocus spherical aberration corrector, 15, 151. Infinity space, generating, 157. Information, 27, 60, 64, 73-74, 179, 235, 241,243,268,270-275,278, 330, 334, 353, 369, 382-383, 396, 398, 443,448,459,468,475-476,481, 487,488-490,494,496-499,506, 512-513,517,519,522-524, 543-544, 556, 559, 570, 580-587, 596, 643, 650, 732, 715, 769, 774, 776, 779, 782, 790, 794, 800. 3-dimensional, 321, 378, 396, 747 4Pi,570 and bleaching, 222, 690-692, 705 CARS, 597-598, 602 colocalization, 668 confocal, 461, 462 contrast, see Chapter 8 and Contrast crystal orientation, 179, 188 display of, 280-281, 288-291, 293, 295-297,299-301,304-305,311 efficiency, 336, 628, 631 of electronic signal, limitations on, 64 genetic, 756, 762-763 lost signal, 25-28 matching gray levels to, 73-74 micro-CT, 615 micro MRI, 618 and Nyquist sampling. See Nyquist sampling, 38, 39, 634-637 chapter, 59-79 optical projection tomography, 612 out-of-focus light, 27, 368, 458, 461, 746, 784 parallel vs. serial acquisition, 223-224 PSF, 245, 247, 250 from second harmonic generation signal, 179 Shannon theory, 443 on source brightness, 137 spectral, 665-667 SPIM, 614, 675-378 storage, 106 chapter, 580-594 theory, 4, 64, 443 transmission, contrast transfer function, 37, 60 Index mismatch. See Spherical aberration. Infrared (IR) lasers, 89, 383, 385. See also, Ultrashort lasers; Laser tweezers. solid state lasers, 108-109 Infrared paper, to identify infrared beams for safety purposes, 118. Ink jet printers, 593. Innova Sabre/frequency-doubling crystal, 102.
Index
Insect cuticle, transparency to NIR light, 166. Installation requirements, for laser sources, 85. Instrument dark noise, 660. See also, Noise Integrated circuit (IC) chip, 9. Intelligent imaging innovations, (III), 3D imaging system supplier, 78-79, 151, 192,395,411,654. Intensified CCD, 13,232-233,460,477, 519-522, 524, 555, 556, 737, 784, 930. Intensity, light, 26, 37, 43, 58, 59, 61, 71-72, 8~ 87, 133, 136, 163, 165, 180, 189, 192, 208, 217, 222, 228, 258,270,391,413,426,459,461, 487,536,538,571-573,633,681, 693,705,810,901. of excitation light, 80, 222, 680-682 laser beam, stability, 86 losses detection path, table, 217 illumination path, table, 217 minimum needed, 392 on optical response of specimen, 165 in photons/second, 80 regulating, 43, 88 singlet-state saturation, See Saturation and visibility, 37 Intensity control. continuous wave laser, 88 non-laser, 128 Intensity distribution, 146-154. of Airy disk, 65, 146. See also, Airy disks changes with focus, 147,407,455, 463,471 effect of coverslip thickness, 149 effect of RI mismatch, 148. See also, Spherical aberration in focal spot, plots, 147-154 nonsymmetrical change with focus, 148 unit image, 147 with astigmatism, 152 with coma present, 151 with spherical aberration, 148-150, 212 Intensity loss, with spherical aberration in detection path, 148-150,212. See Spherical aberration. Intensity spread function (ISF), 74-78. CCDs and PMTs compared, table, 78 defined,75 estimating intensity measurement error, 76 and gray levels, 74-75 measuring, 75 Interference contrast. differential interference contrast, (DIC), 10, 14,76,127,146, 171,453,468, 473-475, 846, See also, Differential interference contrast. deconvolution of, 473-475
phase-contrast, 9, 171, 368, 372, 453, 506,643,649,731,851,854,890, 892. See also, Phase contrast centering the phase rings, 643. See also, Bertrand lens scanning, 9, 13 using fiber optics, 506 RI inhomomogeneity and contrast, 22-23, 41 Interference filters, 45-51, 102, 136, 212. in argon-ion laser systems, 102 continuously-graded, 137 destructive and constructive reflections, 45 transmission, 212 types, 46-49 Interference fringes, coverslip surface, 168, 170. Amoeba plasma membrane/coverslip, 170 in close proximity, 168 Interference mirrors, 46. Interference mode, coherent light, 130. Interference, speckle pattern, 8, 13, 84, 90, 130-132, 144. in backscattered light images, 448 fluorescence speckle microscopy (FSM), 13,383,385,889 Interferometer. 4Pi microscopy, 561 Fabry-Perot (laser), 81-82 fiber-optic, for testing objectives, 240-241 Gires-Tournois, 88 Mach-Zender, to measure pupil function, 245 optical coherence tomography, (OCT), 504, 609 Twyman-Green, 239 Inter-fluorophore distance, measurement, 184. See also, Fluorescence resonance energy transfer. Interfocal crosstalk, 227-228. disk scanners, 227-228, 444, 449 time multiplexing as solution to, 553-554 Interlocks, laser safety, 118. Intermediate optical systems, LSCMs, chapter, 207-220. Internal focusing elements, in objective, 157,511. International Commission of Non-Ionizing Radiation Protection (ICNIRP), 117. International Electrotechnical Commission (lEC), 117. International television standards, 589. Internet sources. See Links. lasers, 123, 124 Intrinsic noise, 21. See also, Poisson noise. Inverse filter algorithm, 476, 477. Ion-binding in Aequorin emits light, 737. Ion concentrations, 346-347, 517, 528-530, 741. chapter, 736-745 determination, 517, 528-530
957
Ion-concentration imaging, 736-738, 740-745. See also, Calcium imaging, pH, etc. calcium imaging, 736-737 concentration calibration, 742-745 indicator choice, 738 interpretation, 740-741 pH imaging, 739-745 water-immersion objectives, 737 Ion sensitive probes, optical, 348, 737. table, 53 I -532 IR. See Infrared; Near infrared. Irradiance, arc and halogen light sources, 130. table comparing, 130 ISO standard, microscope dimensions, 156. Iso-intensity surface, or arc sources, 304. Iterative constrained algorithms, 475-476. See also, Deconvolution; Nonlinear constrained iterative deconvolution algorithms. Iterative constrained Tikhonov-Miller algorithm (ICTM), 497.
J Jablonski energy diagrams, 516, 517, 697, 792. Jansson-van Cittert algorithm, 476, 496. Jitter, defined, for scanners, 54. JND. See Just noticeable difference. Joint Photographic Experts Group. See JPEG. JPEG (Joint Photographic Experts Group), 581-584. Just noticeable difference (JND), ocular response, 72-73.
K Kaede, photoactivatable fluorescent protein, emission change after photodamage, 187,383,385. example image, 187 Kalman averaging, 21, 39, 53, 304, 306, 627,638,655,750,754,781. comparison with deconvolution, in reducing intensity, 39 Kepler, Johannes, 788. Kerr cell, 516. mode-locking (KLM), Ill, 133 of titanium:sapphire lasing rod, 113 Kerr effect, defined, Ill, 179. self-focusing of pulsed laser light, III Kindling proteins, 574, 760. Kinetics, 691, 694-698, 741-742, 774, 796, 810-812,816-817. bleaching, 691, 694-698 and endpoint data analysis, 816-817 fluorescence, 262-263, 348, 383, 385, 571,578,741-742. See also, FUM FRET,796 high content screening, 810-812, 816-817
958
Index
Kinetics (cant.) ion concentration dyes, 741 and STED, 571, 578 Kino, Gordon, confocal design, 6. KLM. See Kerr lens mode-locking. Kohler illumination, 34, 127-128, 131,229, 251,627,648-649. coherence of light, 131 in disk scanner, 229 field diaphragm, 35, 127-129, 139,461, 627,645,648-649 to limit non-uniformity of illumination, 127-128 to measure photon efficiency, 34 Krypton laser, 102, 119, 346, 355. comparison with argon-ion laser, 102 wavelength, 102 Krypton/argon (KrlAr) laser, 90, 92, 93, 102, 108, 119, 203-204, 343, 375, 748,798,811. stabilization, 88 KTP. See Potassium titanium oxide phosphate. L
Labeled structures, plants, 757, 761, 775. bulk labeling, living embryos, 761 cell walls, 775 selective labeling, 757 Label-free microscopy, noise, 114. Lamp housing, 134. Lamprey. labeled axons, 235, 236 larva, optical projection tomography image, 612 Landmark-based registration synthesis method, 328-329. Lanthanide chelates, 345-346. Large mode area photonic crystal fiber (LMAPCF), 110. Larmor frequency, MRM imaging, 618-622. Laser(s), 7-9, 44, 80-83, 88, 90, 94, 112-114,119-120,131,540-543, 599-600. See a/so, Fiber lasers; Mode-locked lasers; Multi-photon ultrafast lasers; Up-conversion fiber lasers; Ultrafast lasers. Alexandrite, 109 amplifier rods, 116 attenuation of, 85, 87-88, 354, 415, 904 axial or longitudinal modes, 83 basic operation, 81-83,116 CARS microscopy requirements, 599-600 chapter, 80-125, table, 119-120 coherence, spatial and temporal, 83-84 colliding-pulse mode-locked (CPM), 540 for confocal, 7, 9-10, 77-78, 280, 535-545 continuous-wave, 90-110 control of power, 543 Cr:Forsterite, 109, 114,415,541, 706-709,712-714
excitation wavelength choice, 540-542. See a/so, Acousto-optical devices, filters femtosecond pulsed laser, 44. See also, Ultrafast lasers fiber-based lasers, 109-111, 113-118 table, 94 ultrafast, 113-114 up-conversion fiber lasers, 109-110 fiber light delivery, 107,See also, Fiberoptics GaAs, 107, III gas, 90, 91-10. See also, lasers by gas. alkali-vapor, 103 Ar-ion, 90, 101-102 Kr-ion, 102 HeNe, 102-103 HeCd,103 heat removal, 84 hybrid mode-locked dye laser, 540-541 important properties for confocal, 80 light delivery, 87-89 fiber-optic, 106 mirrors, 88 longitudinal modes, 82-83 maintenance, 115-116 acti ve media replacement, 115 cooling components, 116-117 optical resonator, 116 metal vapor, 112 microscopical uses nonlinear: 2- 3-photon, 90 Raman and CARS, 90 TIRF,90 tweezers, 89. See Laser trapping multi-photon. See Multi-photon microscopy Nd:glass, 706-708 Nd:YAG, lasers, 88-89,91,95,97,103, 107-109, Ill, 113-115, 117,218, 245,514,680,798 Nd:YLF, lasers, 89, 98, 100, 103, 109, 112-114,750,760-761 Nd:YV0 4 , lasers, 89, 95, 100, 103, 107-109, Ill, 113-114,541 NO SMOKING, 116 performance tables, 91-101 phase randomization, 8, 13,131-132, 143 pointing error, 87 active cavity stabilization, 87 polarization, 83, 88-89 power control, 543 pulse broadening/compensation, 88, 901-904 pulsed, 110-115. See also, Titaniumsapphire, Cr:Forsterite, Nd:glass,YAGIYLFIYV0 4 , etc. cavity dumped, III Kerr lens mode-locked, III modulated diode lasers, 112 pulse-length measurement, 115, 901-903 purpose, 110
saturable Bragg reflector, III ultrafast, DPSS lasers, 112 ultrafast, fiber lasers, 113 white-light continuum lasers, 113 why are they useful?, 110 pumping power requirements, 82 safety, 117-118,839,900. See also, Safety goggles, 118 screens and curtains, 118, 904 solid state, 103. See also, Solid-state lasers semi-conductor, 105-107 thin-disk lasers, 109 spectrum of light, 44 stabilization, 85-87 active, 87 titanium:sapphire laser, 82, 84-86, 88-91, 94,100-103, 105,107,109, 111-112,114,123-124,165,346, 358,415,423-424,459,541,550, 551,645-647,688,706-708,713, 727, 750, 756, 759 4Pi, 563-564, 567 brain slices, 731 CARS, 599 compared to other fast lasers, 82-83, 85,110,112-113 embryos, 731, 750, 756, 759, 764 maintenance, 116 and OPO, 114-115 plants, 415, 423-424, 706-708, 713-714, 717, 781-783 popular models, specs, table, 120 STED,575 transverse modes, 82-83, 85,110 tweezers, 89. See Laser trapping types, 90 ultrafast fiber, 113-114, See a/so, Ultrafast lasers wavelength expansion by sum-anddifference mixing, 114 optical parametric oscillators, 114-115 second/third harmonic generation, 114 white light continuum lasers, 88, 109, 113 continuum, 88, 109 He:Cd,113. Laser cavity stabilization, active, 87. Laser cutters, 686-687. integration, 218-219 Laser illumination, conditions for, 8. Laser lines, using acousto-optical tunable filters, 56. Laser media, maintenance, 115-116. Laser printers, 593. Laser rods, maintenance, 116. Laser Safety Officer, 117. Laser sources, 9, 80-125. See also, Lasers. Laser speckle, 84,90, 130-132,448. removing, 84. See a/so, Scramblers source, 130 Laser trapping, 80, 89, 110, 218-219, 383, 385, 539, 646, 680.
Index
Laser tubes, operational lifetime, 102, 115. components likely to fail, 115 Laser tweezers. See Laser trapping. LaserPix, 282. Laser flying-spot microscope, 7. Lasersharp, confocal microscopes, 282, 284, 285, 288, 292, 296, 302-306. LaserVox, 281-282. Lateral chromatic aberration (LCA), 14, 155-156,239,242-243,287,640, 657-658. correction in conventional optics, 155 measured, 657-658 Lateral coherence, 8, 84, 267. Lateral resolution, 1-4,9, 11-13,28,207, 209,222,225,230,238,270,320, 409,453,511,513,542,552,554,
563,568,651,654-656,747. See also, Resolution. 4Pi,568 CARS, 596-597, 599 confocal endoscopy, 5 I 1, 513 confocal optics, improvement, 9, 651, 654-656 of display, 292 light microscopy, 1-3 optical coherence tomography, 609-610 with pinhole and slit disks, 225 and spherical aberration, 409 SPIM, 613, 674 STED, 573-575, 578 table, 209, 409 Laterally-modulated excitation microscopy, see Stuctured-illumination. LCA. See Lateral chromatic aberration. LCD. See Liquid crystal display. LCOS. See Liquid-crystal-on-silicon. LCS (Leica Microsystems AG), 282, 312, 910. Lecithin myelin figures, CARS image, 204. LED. See Light-emitting diode. Leica, confocal manufacturer, 51-53, 56-57, 160,218,797,910. acousto-optical beam-splitter, 160, 218 objective lens transmission, 160 RS Scanner, 52-53 spectral confocal, TCS SP2, 51, 56-57, 910 tube length conventions, 157,239 Leica Microsystems AG, 282, 910. Leica TCS 4Pi, 119-120,565-568. 4Pi microscopy type C, 565-568 imaging of living cells, 568 lateral scanning, 567, 910 mitochondrial network image, 568 optical transfer function (OTF), 567 sketch, 566, 910 thermal fluctuations minimized, 567 Lempel-Ziv-Welch (LZW), 580-582, 584. Lens aberrations, 13-15. See also, Aberrations. Lens focal length, change, with wavelength, 152.
Leonardo da Vinci, early optical studies, 788-790. Leukocytes, 347, 387, 520, 815, 854. automatic analysis, 815 multi-photon, phase-based FUM, 521 Lifetime. See Fluorescence lifetime imaging microscopy. Ligand-binding modules, 256, 348, 741, 846. Light detection, general, 28-33, 251-264. See also, Detectors; specific detectors: CCDs, PMTs, etc. assessment of devices, 260-262 charge-coupled device (CCD), 254 comparison, table, 233, 255-256, 647 conversion techniques, 259-260 direct effects, 252 future developments, 262-264 history, 262-264 image dissector, 254-255 microchannel plate, 232-233, 255, 262 gated, 5 I 9, 523-524, 527, 532 MCP-CCD, 262 noise internal to, 256-259 internal detection, 256 noise currents table, 256 photoemissive devices, 256-257 photon flux, 257-258 pixel value representation, 258-259 photoconductivity, 252, 253 photoemissive, 254 photon interactions, 252-256 work functions, table, 252-253 photovoltaic effect, 252-253 point detectors, 260-261. See also, PMT quantal nature of light, 251-252 thermal effects, 252 vacuum avalanche photodiode, 254, 255 Light dose, related to pixel/raster size, 64. Light, effects, on plant cells, 770. See also, Bleaching, Phototoxicity. Light-emitting diode (LED), 34, 54, 132-133, 135-139, 143,237. aligning, 135 control by current-stabilized supply, 138-139 definition, 105 to detect galvanometer rotor position, 54 excitation wavelength for fluorophores, 136 expected cost reduction, 237 fluorescence image, 142 galvanometer position feedback, 53 lifespan, 137 to measure photon efficiency, 34 microscope illumination, 131-139, 141, 143 organic, projected development, 143 radiance, 138 spectra, 133 stability, 136 temperature effects, 137 wavelength vs. current change, 137
959
Light flux, light-emitting diode temperature, 133. Light intensity, 71, 163. Light microscopy history, 1-4. Light paths. See also, Commercial confocal light microscopes. separating excitation/emission, 44-45 Light piping by specimen vs. depth, 182. Light-sheet illumination, 672-673. Light sheet microscopy, 613. chapter, 672-679 optical setup for, 6 I 3 white-light continuum lasers, 113 Light sources, widefield, 132-139, 143. See also, Chapters 5 and 6, Arc lamps, LEDs, Lasers; Nonlaser light sources; Filaments; Halogen. commercial sources, 143 solar, 126-127, 131, 135 stand-alone, 143 table, comparative performance, 140 types, 132-139 Light transmission, II, 139, 160-161, 223-229. cummulative loss along optical path, 139 of Nipkow disk system, 11,223-229 specifications for objectives, table, 160-161 Lighting models, 3D image display, 306-312. absorption, 309-312 advanced reflection models, 309 artificial lighting, 309-312 Gourard shading, 308 gradient reflection models for voxel objects, 309 Phong shading, 308-309 Phong/Blinn models, 308 simulated fluorescence process, 310 surface shading, 310 transparency, 280, 284, 287, 300, 304, 309,311-312 Lilium 10ngiftorum, image, 783. Limitations, confocal microscopy, chapter, 20-42. fundamental, 20-42 table, 41, 647 typical problem, 21, 24 Linear galvanometers, 54. Linear longitudinal chromatic dispersion (LLCD), stereoscopic confocal image, 154. Linear unmixing. See Spectral unmixing. Line-scanning confocal microscope, 50, 51, 231-232,237, 784, 908, 916. Linearity, 72, 490. deconvolution for image enhancement, 490 display advantages and disadvantages, 72 Links (Internet addresses). 2 photon excitation spectra, 546, 727, 729, 782 brain slices, 727
960
Index
Links (Internet addresses) (cont.) CCDs, 76, 234, 927, 931 components, 58 confocal Listserve, 390, 901 deconvolution, 495 dyes, 221, 343-344, 782 fluorescent beads, 653 FRET technique, 185,803 high-content screening systems, 811 image management, 865 lasers, 104, 115, 120, 123-125 live-cell chambers, 388-389, 870 movies related to book, 235, 392 muscles, 237 non-laser light sources, 138, 143 plants, 769 safety, 900 software, 282, 376, 594, 734, 762, 764, 776, 777, 820, 824, 827, 831-833, 844, 845, 864-862, 865-867, 869 SPIM,672 Lipid dyes, 236, 355, 359-360, 389, 556, 755,760-761. Lipid receptors, 790. Liquid crystal-on-silicon (LCOS), 266. Liquid crystal display (LCD), 39, 67, 73, 291,293,589-590. digital projectors, 590 filters, 928 non-linearities, 73 shutters, 299, 929 supertwisted nematic (STN), 589 thin-film transistor (TFT), 589 Liquid crystal technology/dynamic polarization microscopy, 188. See also, Pol-scope. Lissajous pattern, circular scanning. 554. "tornado" mode, SIM scanner, 52 List servers, 125. Lithium triborate (LBO), as non-linear crystal for multiplying infrared output, 109, 115. Living cells, 80,90, 114, 136, 145-161, 167,219,221-222,381-399, 429-439,480,564-566,568, 746-766, 770, 772-773, 788-806, 811, 813. See also, Brain slices, Plants cell imaging, and by celUorganism name. 2-photon, penetration, 749-751 2D plus time, 753-754, 762-764 3D projection, 763 4D data, 746-747, 764 4Pi microscopy, 564-565, 568 acquisition speed, 222, 753-754 algorithms, 763-764 assays, 811 beauty and functionality, 790 bleaching of, 797. See Bleaching; Photodamage cell-chamber, 11,22,191,219,370-371, 386-387,394,429-430,564, 610-611
for 4Pi confocal, 564 for biofilms, 870-873, 875, 877, 880, 885 for brain slices, 394, 723, 727, 729 for epithelial cells, 370-371, 377, 386 finder chamber, 683 flow chamber, 870-873, 875, 877, 880, 885 for high-content screening, 810 for optical projection tomography, 610-611 perfusion, 394 for plant cells, 191,429-430 simple, 22, 394 for SPIM, 613, 625, 673 table of required functions, 380 table of suppliers, 388-389 test chamber/dye, 654, 661 cell-cycle effects, 790 chromatin, 385, 390-392, 684, 693-695, 812 chromatin dynamics, 390-392 CNS tissue slice preparation, 393 confocal microscopy, 381-399,746,813 difficulties, 381 future directions, 398-399 considerations, 386-390 antioxidants, 390 experimental variables, table, 386 fluorescent probes, 387-389 maintenance of cells/tissues, 387 minimizing photodynamic damage, 136,389 photon efficiency, 141-161,389-390 in vitro preparations, see In Vitro in vivo preparations, see In Vivo contrast, 747 dyes, 748. See also, Dyes; Fluorophors etc. for rapid assessment, table, 360 embryos, imaging, 746-766. See also, Living embryo imaging external membranes, SHC image, 90 fluorescent staining, 393 microglia, 393 nuclei, living/dead cell, 393 fluorophore effects, 748 FRET imaging, chapter, 788-806 future, 221-222 handling data, 395-396 imaging techniques, 382-386, 394-395 low-dose imaging, 391-392 microglial cell behavior example, 392-398 no damage from SHG imaging, I J4 online confocal community, 390 photon efficiency, 141-161,389-390 phototoxicity, 390-391 assays for, 813 plant, 429-439. See also, Plant cell imaging reflectance imaging, 167 second harmonic generation. See also, SHG
of external membranes, 90 no damage, 114 test specimen for, 390 widefield, 646-647, 751-753 working distance, 5, 9, 129, 145, 154, 157, 198, 249, 511, 568, 598, 634, 673, 678, 727-728, 747, 774, 779, 781,872 table, 158 Living embryo imaging, 749-751, 762-764. aberrations caused by, 747 apparatus, 748 C. elegans, 746, 748 deconvolution helps confocal, 751-753 developmental changes, 746 Drosophila, 273, 675-676, 747-748, 751-752,754,756, 759, 804, 810 dyes, 748 introduction of, 755 embryo size vs. speed acquisition, 753-754 explants, 748-749 future developments, 766 fluorescent probe four dimensional, 746-747, 749 cellular viability, 747-748 challenges, 762 dataset display strategies, 761-764 photodamage during, 746-748 high speed acquisition disk-scanning confocal microscopy, 754 hardware, 754-755 light scattering, 747 optimal acquisition, parameters, 753-754 refractile specimens, 747 superficial optical sections, 748 thick specimens effective strategies, 748-753, 755-761 inherent trade-offs, 747-748 selective plane illumination (SPIM), 751 "Test drives," for living embryo imaging, 752. widefield/deconvolution, 751-752 LLCD. See Longitudinal chromatic dispersion. LMA-PCF. See Large mode area photonic crystal fiber. Loading methods, fluorescent probe, 347, 358-360,430,732-734,738,739. acetoxymethyl esters, 359, 360. See also, Acetoxymethyl esters ATP-gated cation channels, 359 ballistic microprojectile delivery, 360, 726, 803 direct permeability, 358-359 electroporation, 359-360, 795, 803 ion indicators, 738-739, 742 low level, 430 membrane permeant esters, 359-360 microinjection, 360,361,388,739,748, 755, 795, 803-804
Index
neurons, 722, 726, 730, 732-734 osmotic permeabilization, 359 peptide-mediated uptake, 359 plant cells, 769 stabilizing chemicals, 341-342, 362 transient permeabilization, 359 whole-cell patch pipet, 360 Local projections, display, 305-306, 307. Location proteomics, 818. Longitudinal chromatic aberration, 152-155. Longitudinal coherence length, 7, 8, 84, 130,131. Longitudinal linear chromatic dispersion (LLCD) objectives for 3D colorcoded BSL confocal, 154. Long-pass filters, 43-44. Low-voltage scanning electron microscope (LVSEM), 846-847, 849-850, 852. LSM. See Laser-scanning confocal microscopes; Laser-scanning f1yingspot microscope. 6-7 Lucoszs formulation, 273. Luminescent nanocrystals, 343, 345. Luminous intensity vs, color, dye molecule, 138. LVSEM, 846-847, 849-850, 852. LysoTracker Red DND-99, 359-360, 709-710. rapid assessment table, 360 spectra, 7 I 0 LZW compression. See Lempel-Ziv-Welch. M Mach-Zehnder interferometry, 245. Machine learning. See Automated interpretation of subcellular patterns. Macrography, 3D light scanning, 672. Magnesium fluoride (MgF2). for anti-reflection coating, 158 Magnetic disks, 586. Magnetic resonance imaging (MRI), 618. Magnetic resonance microscopy (MRM), 618-624. amplitude modulation for RF carrier, 620 applications, 623-624 botanical imaging, 624 developmental biology, 624 histology, 623 phenotyping, 623 basic principles, 618-619 Fourier transform/image formation, 620 future development, 624 hardware configuration, 621-622 image contrast, 622-623 image formation, 619-621 Larmor frequency, 620 Schematic diagram, 618-619 strengths/limitations, 622 Magnification, 24, 35-41, 62, 131,215,443. See also, Nyquist sampling; Oversampling; Undersampling. calibrating, 653, 658 and CCD pixel-size, 62, 70
confocal, 52-53, 62-64 effect on pixel size, 24, 928 factor, 24, 28 and lateral chromatic aberration, 278 for line-scanner, 232 over-sampling, 68-70, 493, 509, 635, 729 high-content screening, 816 and pinhole size, 28 under-sampling, 68 zoom magnification, 11, 24, 37, 63-34, 66,70,79,317,389,493,627, 634-636, 731 Maintenance. cell viability, 387 dye lasers, 114 lasers, 115-117, 124 remote logging of, 864 troubleshooting reference, 124 Maize (Zea mays), 167-168, 172, 179,202, 417-424,428,438,710-711, 713-714. 2-photon, time-lapse microspectroscopy, 423 abnormal vasculature, 437 anther, 420, 433 attenuation spectrum, leaf, 418 cross-sections, stem, 172, 707 emission spectrum, 710, 711, 713 fluorescence spectra, 422-424 leaf, attenuation spectrum, 418 optical section, 172, 179 reflectance, 167 surface, 436 meristem, 420, 430-432, 707 multi-photon excited signals, 422-424 polarization microscopy, 707, 711 pollen grain, 202, 433-434 protoplast, 424 root, 432 second harmonic imaging, 707, 711 silica cells, 428, 437, 707 spectrum, 422,423, 710 starch, 420, 435-436, 707, 711 stem attenuation spectra, 417, 418, 713 optical sections, 419, 714 storage structures, 420, 435-436, 707, 711 Manufacturers. See also, Commercial confocal light microscopes; Appendix 2. listing with web addresses, table, 104-105. Mapping conventions, in image processing, 294-296, 300-304. data values, 300-304 choosing data objects, 300-301 object segmentation, 301-302 projection rules, 302-304 scan conversion, 301-302 table, 300 visualization, 300
961
multi-dimensional image display, 294---296 G function, 294 image/space view, 296 orthoscopic view, 294 reducing geometric dimensions, 294 rotations, 294---296 visualization process, 294 MAR. See Mark/area ratio. Marching cubes algorithm, 301-302, 304, 776. Marconi, CAM-65 electron multiplier CCD camera, 76. See also, EM-CCD. Mark/area ratio (MAR), 279. Marsilea quadrifolia, 416, 419. attenuation spectra, 416 optical section, 419 Mass balancing, to reduce scanner vibration, 54. Mass storage, 580-588, 593-594. data compression for, 288-289, 292-293, 295,319,499,580-585. See also, Data compression algorithms, 319, 580 archiving systems, 580 color images, 581 file formats, 580-588 removable storage media, 585-588. See also, Removable storage media random-access devices, 586-588 sequential devices, 585-586 solid state devices, 588 time required, table, 581 Materials, silicon, fused quartz, beryllium, 52. Mathematical formulas, for confocal microscope performance, table, 209. Maximum intensity projection, 180, 284-285,292,294,298,302-304, 307,313-314,319, 325-326, 330-331,585,755,763-764,770, 774, 881, 884. local, 305 Maximum likelihood estimation (MLE), 472-475,495,497-498,669. blind deconvolution, 472-475, 498, 784 effect on colocalization, table, 669 M-CARS. See Multiplex CARS micro spectroscopy. MCP. See Microchannel plate, 232-233, 255,262. MCP-CCD,262 Gated intensified, 519, 523-524, 527, 532 MCP-PMT. See Microchannel plate photomultiplier. MDCK cell, 372-374. actin cytoskeleton, 374 Golgi apparatus, image, 374 morphologic changes, 374 stereo image, 373, 374 vertical sections, image, 372
962
Index
Measurements, 20, 33-36, 76, 139-141, 159. achromat perfonnance, 194 buffering of, ion measurement, 738, 740 field flatness, 26-28 geometric distortion, 653-654 laser pulse length, 109, 112, 115,507, 537,538,902-903 light throughput, 139-141 limits on confocal intensity, accuracy, 20 photon efficiency, 33-36 pinhole, effective size, 34 intensity spread function histogram, 74-78 resolution, 241-245, 657, 658 shrinkage, specimen preparation, 371-373 spectral transmission of objective, 159 spherical aberration, 145, 407 surface height, using LLCD BSL confocal, 224 z-resolution, 194 Mechanical scanners, 51-54. Melles Griot catalog, real lens, perfonnance, 210. Membrane penneant esters, 361, 358-359, 361, 726, 738-739, 744. Membrane potentials, 179, 188, 204-205, 346,353,383,517,743,811-813. Memory stick, 588. Mercury arc lamp, 37, 44, 132, 135-138. fluorophores matching excitation, 135-136, 139 iso-intensity plots of discharges, 132 and pinhole size, 37 radiance, improvements, 137-138 wavelengths, 44 Mercury-halide arc source, 136, 138, 143-144. spectrum, 144 Mercury-iodine (Hg-I) arc lamp, radiance, 138. Mercury-xenon arc lamps, 136-138. spectral lines, 136 Meristem, 168, 420, 430, 432, 770, 776-778,782. maize, 168, 432 Merit functions, confocal scanners, 217. object-dependent, defined, 217 object-independent, defined, 217 Mesophyll cells, 169, 193, 195,417-418, 423,428,430,711-712,714,779. A. thaliana, 193, 196 photodamage, 203 protoplasts, 196, 203, 424, 425-426, 430,439 harmonic images, 711-712, 714 image, 424 spectra attenuation, 416, 418 change with 1- vs. 2-photon, 421, 423 emission, 423 Metal-halide light source, 136, 143-144, 907,908.
Metal vapor lasers, 112. Metamorph, 281-282, 290, 311, 817. Microchannel plate (MCP) image intensifier, 233,255,519,532. multiplicative noise, 233 photocathodes, 262 PMT, 255, 523, 532 Microchannel plate PMT (MCP-PMT), 255. Micro-computerized tomography (MicroCT),614-618. contrast/dose, 614-615 dose vs. resolution, graph, 616 layout, 614 mouse images, 615-617 tumor-bearing, 617 operating principle, 614 Micro-CT. See Micro-computerized tomography. Microdissection. with multi-photon IR light, 686-687 with nitrogen lasers, 112 Microelectrodes, for introducing indicator, 738. Microglial cell behavior, 392-398. Microinjection, 360-361, 388, 739, 748, 755, 795, 803-804. of chromophores, 803-804 Microlens array, 12, 134, 135, 216, 225, 231, 235. for 4Pi confocal, 563-565 for CCD, 237 for disk scanners, 6, 12, 216, 224, 226, 231,458 for light-emitting diode source, 134-135 for multi-focal, multi-photon, (MMM), 537,551-555,558 principle, 135 in Yokogawa disk-scanning confocal, 12, 224-226,231,235 Microscopes, 217, 226. See also, particular types. attachment of confocal scanner, 21 7 specification comparisons, table, 226 Microscopy laboratory URLs, 125. See also, Links. Microspectroscopy, 421-425, 426, 516. CARS, 601-602 fluorescence properties of plants, 421-425 lifetime, 516 of maize, 424 multi-photon setup, 424 Microspores, birefringence images, 189, 431-432. Microsporogenesis, 431-432. Microstructure fibers, 504. Microsurgery, 112,219,686-687,764-765. Microtubule, 11,68,80, 188,222,292,432, 582,703,714,752-753,759,773, 790, 852. See also, Cytoskeleton. birefringence, 714-715 Brownian motion of, 11 electron microscopy, 848, 850 fixation, 369, 372-375 fluorescence correlation spectroscopy, 383
GFP, 12. See also, Green fluorescent protein in mitosis, 759. See also, Mitotic apparatus polarization microscopy, 15, 173, 188, 420-421 photodamage of, 341, 850-851 stabilizing buffers, 852 STED, 576-577 stereo image, 752 TIRF, 180, 183 second harmonic generation, for tracking, 90 Microwave fixation, 377-378. Microwire polarizer (Moxtec Inc.), 85. Mie scattering, 162-163, 167,417-418. clearing with index-matched liquid, 167 comparison with Rayleigh scattering, 163 light attenuation in plant tissue, 417 by refractive structures, 162-163 MIl. See Multi-photon intrapulse interference, 88. Mineral deposits, plant, 163-420,436-437, 703. Miniaturized fiber-optic confocal microscope, 508-512. bundle imagers for in vivo studies, 509 clinical endoscope, 514 objective lens system, 509 optical efficiency, 509 optical schema, 508 resolution, 509 rigid endoscope, 511 single fiber designs, 510 vibrating lens and fiber, 510-511 in vivo imaging in animals, 512 Minolta, CS-100 radiospectrometer, 139. Minsky, Marvin, 2, 4-6, II, 141, 216, 890. Mirror coupling, pulse width and pulse shape, 88. Mirrors, 26, 48, 54, 63, 209-210, 214. galvanometer, 54. See also, Galvanometers internal, testing reflectance losses, 26 laser-line, 48 performance, 54, 63. scan angle and magnification, 63 size calculation for LSCM, 209 x-y scanning mirror orientations, 214 Mismatch, 893. probe shape/pixel, 39, 466 caused by chromatic aberration, 243 refractive index, 377, 404-412, 411, 654, 658, 747, 863, 893 4Pi,568 causing signal loss, 148-150, 408-409, 654 chapter, 404-412 corrections, 411-412 embryos, 747 film vs. CCD, 590 harmonic signal generation, 704-705 less, at long wavelength, 416
Index
measurement, 148-150, 655-656 of movie frame rate, 839 MRM, contrast agent vs. imaging time mylar flakes, 198 resolution loss measured, 192-194 vector mismatch in CARS, 596-597, 600 z-distortion, 287 Mitotic apparatus, 15, 173, 373-374, 377, 386,421,431,693,749,752,799. See also, Microtubules. damage, 693 fixation, 373-374, 377 FRET, 765 marker, cyclin-B, 790 Pol-scope, 13, 188, 432, 468, 479-480. deconvolved, 479 images, 15, 188,479,717 SHG imaging, 702, 718-719 in vitro fertilization, 188 Mitotracker stain, 142, 170, 353, 358, 360, 430-431, 692, 750. living cells rapid assessment, table, 360 Mixing, sum or difference, to generate laser wavelengths, 114. MLE. See Maximum likelihood estimation. MMM. See Multi-focal, multi-photon microscopy. MMM-4Pi microscopy, 556. MO (magneto-optical) disks, 586. Model-based object merging, 323-325. Mode-locked lasers, 87,101,111-114,118, 124, 358, 520, 646, 728-729, 749, 901-904. active, pulsed laser class, III adjustment of, 901-904 for CARS, 599-600 colliding pulse, 112 fiber, ytterbium and neodymium, 113-114 fiber/diode, ultrafast, 113-114 FUM,520 interference with, by specimen, 171 Kerr lens, III modulator, fiber lasers, III multi-photon, 535-536, 540-541, 550-551,563-564,567,646, 728-729, 749 passive, III, 113-114 saturable Bragg reflector, III SHG, THG, 706-707 Mode-locked oscillators. See also, Modelocked lasers. nanojoule pulse energies, 111 Moire effects, 270-271, 755. ambient fluorescent room lighting, 20 I banding patterns, 755 disk-scanners and CCDs, 231, 754-755 structured-illumination methods, 268-271, 273 Molecular imaging, in vivo, 387, 618, 624, 790, 803-806. FRET, 790, 803-806 micro-CT, 618
MRM,624 Monitors, computer display, 588-589. Monkey cells, 693, 803. Monomeric red fluorescent protein (mRFP) constructs, 756, 760, 798. with CFP or GFP molecules, as FRET pair, 798 Montage synthesis method, 281, 312, 318, 328-331,748,753,851-852,855, 858-859. defined, 329-330 examples, 330-332, 780-781 scanning electron micrographs, 851-852, 855 TEM methods, 858-859 Moon, early phase measurements, 788, 789. Morphological filters, 285, 300-30 I, 316-317,320-322,730-734,817, 826. high-content screening, 812, 819, 826 Morphometry, 145,316,319,331,726,728. group properties, 331 intensity/spectral measures, 331 interest points, 33 I invariants, 33 I location/pose, 331 shape measures, 331 size measures, 33 I texture measures, 331 topological measures, 331 Mosaicing. See Montage synthesis method. Mounting medium, 166, 198,342,370-371, 373-377,404-413,418,454,457, 473,493-794,499,564,631,642, 652, 655, 696, 730, 774 See also, Clearing agents. brain slices, 730 chapter, 404-413 clearing solutions, 166, 417-418, 420, 439,610, 624, 774-775 effect of glass bead, 199 plant specimens, 418, 431, 774 refractive index, tables, 377 selection, 198, 631 Mouse, 192, 376, 393, 608, 612, 615+, 723, 726. confocal colonoscopy, 509, 512 embryo optical projection tomography image, 612 SIMIEFIC image, 608 GFP transgenic, 726 hippocampus, 393 micro-CT image, 614-615, 617 femur, 616 tumor, 617 spectral unmixing image, 192, 382, 664-667 examples, 665-666 visual cortex brain slices image, 723 Movement contrast, 190. Movie compression, 836-840. Moving-coil actuators, galvanometer, 52. Moxtec Inc., Microwire polarizer, 85.
963
MPA. See Multi-photon absorption. MPE. See Multi-photon excitation. MPEG display formats, 836-84 I. MPEM. See Multi-photon microscopes. MPLSM. See Multi-photon laser-scanning microscopy. MPM. See Multi-photon microscopy. MQW. See Multiple quantum wells. mRFP. See Monomeric red fluorescent protein. MRI. See Magnetic resonance imaging. MRM. See Magnetic resonance microscopy. Multi-channel experiments, 813. filters and dispersive elements, 51 time-lapse fluorescence imaging, 382, 384 toxicity, 755 Multi-dimensional microscopy, display, 280-314. See also, Automated 3D image analysis methods. 2D pixel display space, 291 efficient use, 292 animations, 292-293 artificial lighting, 306-308 CLSM images, 286-290 anisotropic sampling, 287 calibrating image data, 286 data type/computational precision, 288-289 dimensions available, 286-287 file formats for calibration/interpretation, 288-289 image data, 286 image size available, 287 image space calibration, 287-288 image/view dimension parameters, table, 288 processor performance, 289-290 storing image data, 286 voxel rendering speed, 290 color display space, 291-292 commercial systems, tables available systems, 282-283 desirable features, 288-289 display options, 293 geometric transformations, 295 projection options, 300 realistic visualization techniques, 307 criteria for choosing visualization, 281 data values, definition, 222, 280 dimensions, 280, 323 degrees of freedom, optical image, 8-9 depth-weighting, 304, 306 exponential, 304 linear, 304 recursive, 304 display view, definition, 280 hidden-object removal, 304-305 local projections, 305-307 z-buffering, 304-305 highlighting previously defined structures, 284 image, definition, 280
964
Index
Multi-dimensional microscopy, display (cont.) image/view display options, table, 293 geometric transformations, table, 295 intensity calibration, 304 iso-intensity surface, 304 laser-scanning microscopy, 280 lighting models, 306-312 absorption, 309-312 advanced reflection models, 309 artificial lighting, 309-312 Gourard shading, 308-309 gradient reflection models/voxel objects, 309 Phong shading, 308-309 PhonglBlinn models, 308 simulated fluorescence process, 310 surface shading, 310 transparency, 309-312 living cells of rodent brain, 392-398 mapping data values, 300-304 choosing data objects, 300-301 object segmentation, 302 projection rules, 302-304 scan conversion, 301-302 segmenting data objects, 30 I visualization model, 300 mapping into display space, 294-296 G function, 294 image/space view, 296 orthoscopic view, 294 reducing geometric dimensions, 294 rotations, 294-296 visualization process, 294 measurement capabilities See also, Chapter 15 reconstructed views, 312-313 results, 284-285 objective vs. subjective visualization, 281 prefiltering, 281 principle uses, 281-285 projection/compositing rules, 302-304 alpha blend, 302, 304 average intensity, 302 first or front intensity, 302 Kalman average, 304 maximum intensity, 302 pseudo color, 173-175, 190, 291 purpose, 281-285, 293-295 realism added to view, 306-308 techniques for, table, 307 reconstructed view generation, 290-312 5D image display space, 291-294. See also, 5D image display space choosing image view, 291-294 subregion loading, 290-291 reconstruction, definition, 280 reflection models, 306-308 rendering, definition for, 280 software packages, table, 282-283 stereoscopic display, 293, 296-299 color space partitioning, 297 interlaced fields of frame, 297
pixel-shift/rotation stereo, 297 stereo images example, 298 synchronizing display, 297 true color, 291 unknown structure identification, 281-284 viewing data from, 283 visualization parameters, table, 285 z-coordinate rules, 304 z-information retained by, 296-300 non-orthoscopic views, 299 stereoscopic views, 296-299 temporal coding, 299-300 z-depth, 299-300 Multi-fluorescence, systems for utilizing, 217+. Multi-focal, multi-photon microscopy (MMM), 221, 276, 550-559, 797. 4Pi-MMM, 563-564 basics, 565 scheme, 563 alternative realizations, 554-555 background, 550 beam subdivision approaches, table, 558 current developments, 558-559 experimental realization, 551-555 FRET, 797 imaging applications, 556 boar sperm cells, 557 Convallaria, 556 FRET, 556 hippocampal brain slices, 557 pollen grains, 556 Prionium, 556 interfocal crosstalk, 553-554, 556 time-multiplexing, 553-555 limitations, 556-558 localization, 538 Lissajous pattern of scanning foci, 554 "tornado" mode, SIM scanner, 52 Nipkow-type micro lens array, 551-552 optimum degree of parallelization, 550-551 resolution, 552-553 schematic diagram, 552 time multiplexing, 553-554 variants, 555-556 FUM, 555-556 MMM-4Pi, 556 SHG,556 space multiplexing, 555 Multi-length fiber scrambler, 8. See also, Scramblers, light. Multi-photon absorption (MPA), 535. Multi-photon excitation (MPE), 356-358, 535-545, 894. See also, Multi-focal multi-photon microscopy. absorption, 705-707 advantages/disadvantages, 644-647, 749-751 autofluorescence, plants, 424, 427 background from SHGITHG, 361, 708-709, 728 backscattered light imaging, 429
bleaching, 218, 338, 539-540, 680-689, 692-693, 905. See also, Bleaching; Chapter 38 caged compounds, 187, 383, 543-544, 692,729,912 cell viability during imaging, 544-545 chromophores for, 543-544 detection, 538 duty cycle, 644 excitation localization, 538 excitation spectra, 125 FUM, 576 fluorophores for, 543-544 FRET, 797 heating, 539-540 history, 535 image formation, 535-540 instrumentation, 540-543, 900-905. See also, lasers for. See also, Ultrafast lasers Alexandrite, 109 Cr:Forsterite, 109, 114,415,541, 706-709, 712-714 Nd:glass,706-708 Nd:YAG, 88-89, 107-109,514,680, 798 Nd:YLF, 89,112-114,750,760-761 Nd:YV04 , 89, 95,107-109,113-114, 541 Ti:Sapph. See Laser, titanium-sapphire laser multi-focal, multi-photon microscopy alignment, 900-901 beam delivery requirements, 541 control of laser power, 543 CPM laser, 540 descanned detection, 166, 208, 212, 428, 537, 540-542 excitation wavelengths, 541 focal plane array detection, 542 hybrid mode-locked dye laser, 540-541 lasers/excitation wavelength choice, 540-542 non-descanned detection, 185, 201, 218,381,447,456,507,542,552, 559,643,646,727,750,779,904, 909,910 non-mechanical scanning, 543 optical aberrations, 542 power requirements, 541, 903, 904 pulse spreading due to GDD, 547, 538, 543 resonant scanning, 543 whole-area and external detection, 541-542 optical pulse length, 537-538 group delay dispersion, 537-538, 543 group velocity dispersion, 88, 111, 210, 537, 606, 903 measurement, 115,901-903 penetration, 749-750 photodamage, 539-540, 680-688, 692-693
Index
physical principles, 535-540 refractive index mismatch, 404-413 resolution, 539 SHG and THG background, 361, 708-709, 728 two-photon absorption cross-sections, 125 (URL) 543-544 wavelengths, 538-539 Multi-photon intrapulse interference (MU), 88. Multi-photon microscopy (MPM), 10-11, 56,172-177,210,535-545,681, 682, 685-688, 746-766, 894, 900-905. advantages/disadvantages, 644-647, 749-751 alignment, 901-902 autofluorescence, 425-427, 545 SHG, THG, 361, 708-709, 728 calcium imaging, 545 cell damage during, 544-545, 682, 685 I-photon vs. 2-photon excitation, 681 absorption spectra of cellular absorbers, 681 intracellular chromosome dissection, 688 mitochondria, 686 nanosurgery, human chromosomes, 686-687 by optical breakdown, 198, 680, 682, 685, 687, 703, 705 photochemical, 682-685 photothermal, 539, 545, 681, 685, 904 reproduction affected by ultrashort NIR pulses, 686 ultrastructural modifications, 685-686 cell viability, 544-545 compared with other 3D methods, 644-647, 748-751 deconvolution, 495-498 developmental biology, 545, 746-754, 757, 759-760, 764 dispersion as problem, 56. See also, GYD;GDD fluorescence, contrast, 172-177 for living embryo imaging, chapter, 746-766 need for efficient illumination light path, 210 optical layout, 540 photobleaching, 545, 680-688, 692-693 practical operation, 900-905 protein damage/interactions, 765 resolution, 552 setup/operation, 540, 900-905 schematic diagram, 540, 901-902 in vivo (intact animal) imaging, 545 ultrafast lasers, 88, 90, 109. See also, Ultrafast lasers Alexandrite, 109 Cr:Forsterite, 109, 114,415,541, 706-709,712-714
Nd:glass, 706-708 Nd:YAG, 88-89, 107-109,514,680, 798 Nd:YLF, 89, 112-114,750, 760-761 Nd:YY04, 89, 95,107-109,113-114, 541 Ti:Sapph. See Laser, titanium-sapphire laser uncaging, 545 Multi-photon-based photo-ablation, 764. Multi-slit design, for disk-scanning confocal, 229. Multi-view deconvolution, 330, 675-677. Multiple quantum wells (MQW), diode injection lasers, 106. Multiplex CARS microspectroscopy (MCARS), 601, 602. Multiplicative noise, 28-33, 51, 77-78, 224, 234,256-258,262,275,443,460, 633,661,667. of EM-CCD, 30-31, 77-78, 264, 256, 262 losses in effective QE from, 33, 234, 443 from PMT, 29, 51, 77-78, 233,256-258, 460,633,661,667 and quantum efficiency, 33, 234, 443 photon counting, 32-33, 78 pulse pile-up, 32-33, 35, 78, 521, 523, 526-527 table,256 why it is usually unnoticed in LSCM, 633, 661 Muscle, 737, 739-742. fatigue, 739-740 N
NA. See Numerical aperture. Nanobioscopy, protein/protein interactions, 795-798. acceptor bleach, 797-798 donor fluorescence, 796-797 FRET measurement, 795 sensitized acceptor emission, 795-796 Nanoscale resolution with focused light, 571-578. See also, Stimulated emission depletion (STED) microscopy. breaking the diffraction barrier, 571-573 different approaches, 573-574 ground state depletion (GSD), 573 STED, 573-574 outlook, 577 RESOLFT concept, 571-573 resolution, new limiting equation, 571 measured, 578 stimulated emission depletion (STED), 573-578 axial resolution increase, 576 compared to confocal microscopy, 576 dyes, suitable, table, 575 OTF comparison, 578 PSF comparison, 578
965
Nanosurgery, 219. with multi-photon systems, 90 NCI60 CMA, standard encapsulation, 816. NCPM. See Non-critical phase matching. ND. See Neutral-density filters. Near infrared (NIR) lasers, 10, 90, 106. See also, Lasers: titanium-sapphire; Nd:; Cr:Forsterite. Near infrared (NIR), 10, 90, 106. diode injection lasers, 106 for laser tweezers, 90 objective lenses designed, 174 Nearest-neighbor deconvolution algorithm, 476. image enhancement, 495-496 Negative contrast, for fluorescence microscopy, 173-174. Negative feedback, to correct mirror motion, 53. Neodymium glass laser, 706-708. Neodymium-yttrium aluminum garnet (Nd:YAG) lasers, 88-89, 91, 95, 97, 103,107-109, Ill, 113-115, 117, 245, 514, 680, 798. infrared range, 108 pumping non-linear crystal/green light, 114-115 Neodymium-yttrium lithium fluoride (Nd:YLF) laser, 89, 98, 100, 103, 109, 112-114, 750, 760-761. Neodymium-yttrium orthovanadate (Nd:YY0 4 ) laser, 89, 95, 100, 103, 107-109, Ill, 113-114,541. kits utilizing, 113 Nerve cells, images. Alexa stained, 330 backscattered light images, 167 eye, optic nerve, 481 Golghi-stained, 298 Lucifer-yellow, 314 microglia, 396-398 rat-brain neurons, 398 transmitted light, 475 Neutral-density filters (ND), 43, 76, 126. in fixed-pattern noise measurements, 76 to reduce source brightness, 43, 126 NFP. See Nominal focal position. Nikon, confocal manufacturer, 13, 15, 119-120, 161, 199,201,507, 638-640,657, 750, 910. C I confocal microscope, 119-120, 507 C 1si spectral confocal microscope, 908, 910 CF objectives, 154-156,217,669,779 confocal X-Z, BSL image, 22 Plan Apo objective, 13, 15, 638 resolution, measured, 16, 638-640, 657 water-immersion lenses, 15 high-content screening, 810 tube length conventions, 157,239 Nile Red, dye, 435, 528, 575, 774, 782
966
Index
Nipkow disk scanning, 2, 5-6, 11, 12,41, 215,223,231,276,551,754, 783-784, 810, 894. See also, Yokogawa; Disk-scanning confocal microscopy. commercial systems, 907, 913, 915 compared to single-beam scanning, 458 for high-content screening, 810 micro-lens system, 6, 12,216,224-226, 231,234,237,551-552 multi-photon, 537, 551-558, 563-565. See also, Multi-focal, multi-photon microscopy rotation, 754 for single-sided confocal, 6, 141,223, 229 source brightness, 141 speed of image acquisition, 11, 220, 222-226,227,231 for tandem-scanning, 141, 215 visualization, of cells, 458, 667, 754, 784 Nipkow, Paul, 5-6, 109 NIR. See Near infrared. Nitrogen lasers, 112. nanosurgery using, 219 NLO. See Non-linear optical effects. NMR. See Nuclear magnetic resonance. Noise, 21, 28, 74-77, 83, 87,114,190,232, 256-259, 442-444, 495. See also, Signal-to-noise ratio; Poisson noise; Quantum noise. background, 443-444 ofCCD detectors, 30-31, 77-78, 232-233, 256, 262 equations, 256 table,256 vs. photomultiplier tube detectors, 74, 77 CIC, clock-induced charge, EM-CCDs, 234,926 in counting quantum-mechanical events, 21 deconvolution reduced noise, 39-40, 114, 392,495,498,667,783,835-836 detector, 28 fixed-pattern, 74, 76, 278, 924, 927, 931 in fluorescence microscopy, defining, 74-75 in lasers, sources, 85-86 reducing, 87 limits grey levels, 443 measurement, 74-75 multiplicative, 28-33, 51, 77-78, 224, 234,256-258,262,275,443,460, 633,661,667 in photon detectors, 256-259 noise currents table, 256 photo flux, 257-258 photoemissive devices, 256-257 pixel value represented, 258-259 Poisson. See Poisson noise polarization, in laser systems, 83 read, and readout speed, 77
shot, 442-443. See also, Poisson noise single-pixel, 65, 67, 190,635, 832, 835-836 deconvolving, to reduce, 39-40, 392, 498,667,784,835-836 reducing, 39,40, 190,41,65,392,498 sources of, 442-444 wavelet transform to reduce, 733-734, 819-820 Nomarski DIC contrast, 2, 368, 746, 892. See also, Differential interference contrast. Nominal focal position (NFP), 405, 408, 409. calculations for glycerol, 409 calculations for water, 409 z-responses, diagram, 408 Non-confocal microscopy vs. confocal, 746. high content screening, table, 811 Non-critical phase matching (NCPM), 114-115. Non-descanned detection, for MPM, 185, 201,218,381,447,456,507,542, 552, 559, 643, 646, 727, 750, 779, 904,909,910. for CARS, 559 No-neighbor algorithm, 476-477, 496. Non-laser light sources, chapter, 126-144. arc sources, 130, 132, 140 commercial systems, table, 143 comparative performance, table, 140 control, 138 for disk -scanning confocal, 141 filament sources, 135-136 LEOs, 132-133, 135, 138-139, 143 light scramblers, 131-132 measured performance, 139-141 results, 142 solar, 126-127, 131,135 stability, 136-137 Nonlinear constrained iterative deconvolution, 68, 458, 475-476, 496-497,499,520,568. Nonlinear conversion, tunable laser, 114. Nonlinear crystals, frequency multiplying, 109. Nonlinear optical (NLO) effects, in microscopy, 90, 114, 163, 165, 177, 179,188,190,195,416-417, 426-427,430,442,504,535,507, 703-720,728,741,751. See also, Multiphoton/microscopy; Harmonic signals; SHG, THG. absorption, 188,415-418,426-427,430, 705 bleaching, 536, 550, 558, 645, 680-685, 693,697, 707, 729. See also, Bleaching; Photodamage CARS, 595-598, 600 DIC, 473-474. See also, Differential interference contrast fluorescence, 172, 179 focus shift with spherical aberration, 409
harmonic generation, 704-705 emission, 710-711 energy state diagram, 705 multi-photon absorption/fluorescence, 705 second harmonic generation (SHG), 704-705 setup, 708-709 third harmonic generation (THG), 705 light sources/detectors, 706-708 light attenuation spectra in plants, 706 photodetector characteristics, 707 pulsed-laser, table, 706. See also, Ultrafast lasers in optical fiber, 504-508 optically active animal structures, 714-717 man-made collagen matrix, 717 signal-producing structures, table, 715 spindle apparatus, 717-718 zebrafish embryo, 716, 718 optically active plant structures, 710-714 Canna, 710 Commelina communis, 712 emission spectrum of maize, 710, 711 maize stem, 711, 714 potato, 712 rice leaf, 712, 715 polarization dependence of SHG, 717, 719 setup for, 708-710 spectra, 415, 417, 435 Euphorbia pulcherrima, 710 maize leaf, 710 Pyrus serotina, 711 STED microscopy, 571-579. See also, STED microscopy. structured illumination, 270, 276 Non-radiative dipole-dipole interactions, 790. Non-specific staining, 27, 44, 74, 303, 345, 357-358,442,467,472,617,660, 667-668, 760, 820, 878, 882. See also, Background. Non-tunable solid-state laser, 103. Normal, free-running, pulsed laser, III. Northern Eclipse, software, 282. Notch filter, to transmit laser line, 49. Novalux Inc., Protera 488 laser system, 107. NSDC. See Nipkow spinning-disk confocal. Nuclear import analysis, 802. Nuclear magnetic resonance (NMR), 618. Numerical aperture (NA), 1,4,24,28,61, 126,141,145,148,168, 180, 195, 198, 239-250. affects surface reflection contrast, 180 defined, I determining axial resolution, 4, 241-242, 657 determining lateral resolution, I, 241-242, 656 diffraction orders accepted by, 61 effect on self-shadowing, 168, 198
Index in fiber-based mini-confocal endoscopes, 509 image brightness, 126 matching to CCD pixel size, 62, 928 objective lenses with high, 145, 239-250 empty aperture, 248 with oil-immersion vs. water objective, 148 pinhole size as function, 28 and refractive index mismatch, 147-148. See also, Spherical aberration in tandem scanning confocal microscopy, 141 vertical shadowing, 195 and zoom setting optimal, 24 Nyquist criterion, and digitization, 38-39, 64-68. Nyquist digitizing, 65, 67. Nyquist filtering, 70-79, 281. Nyquist frequency, 64, 301. See also, Shannon sampling frequency. Nyquist, Harry, 64. Nyquist noise, 256. Nyquist reconstruction, limit output bandwidth, 59, 66-67, 69, 70, 173, 235-236,280-315,458,468-469, 474-475,496-497,563,585,603, 607, 610, 615, 635, 672, 675, 677-678,690,772,730-731,762, 77, 774-776, 778, 784, 883. Nyquist sampling, 24, 37, 39,40,53,60, 64-70, 73, 75-76, 78-79, 142, 146, 152, 205, 222, 258, 271, 273, 289, 386,391,448,635-636. blind spots, 38 for CCD camera, 70, 233, 273, 928 and deconvolution, 59, 65, 67-68, 222-223, 635 diagram, 60 optimal, results of deviating from, 24 practical confocal microscopy, 448, 635-636 reconstruction, see Nyquist reconstruction. relationship with Rayleigh-criterion and PSF, 39, 60, 64, 66 signal-to-noise ratio, 67, 448 subpixel, resampling, 478-479
o Object scanners, image quality, 216. Objective lenses, 13, 15,25-26,34,49, 145, 152, 156, 239-250, 652-660. See also, Aberrations. apodization, 250 axial chromatic registration, 287, 658 axial resolution measurement, 656-657 vs. pinhole size, 656 chromatic aberrations 14, 145, 152-156, 160,177-178,209,242-243,641, 659 apparatus in measuring, 243-244, 654, 659
axial shift, 243-245, 657-658 chromatic registration, 657-658 cleaning, 642 confocal performance, 145-161, 652-660 contrast transfer function (CTF), 16,35, 37-39, 59-62, 656, 747 coverslip thickness, table, 654 dipping lenses, 161,209,411,429,568, 613, 727, 737, 870, 872 dry, high-NA, aberrations, 15 field illumination, 34-35, 127-128, 139, 461, 627, 658 flatness of field, 145,151-152, 154,418, 457, 639, 659 Focal Check™ beads, 657 high-NA planapochromat, 13, 145, 239-250 infinity correction, 155-157, 166, 239, 405 advantages, 49 lateral chromatic registration, 657-658 lateral resolution. See CTF light, vector nature, 267 mounting media. See Mounting media photon efficiency losses, 25-26 plan objectives, table, 152 point spread function of high NA, 239-250 measuring, 240-242,455,462,471, 656 polarization effects, 249-250 pupil function, measured, 245-248 3D point spread function restored, 247-248 empty aperture, 248 Mach-Zehnder interferometry, 245 phase-shifting interferometry, 245 Zernike polynomial fit, 245-247 table, 247 resolution test slide, 169, 656 spherical aberration. See Spherical aberration correction, 654-655 sub-resolution beads, 181-182, 196, 454, 477,493,499,527,652-656,784, 900, 904, 930 images, 656 table of suppliers, 653 temperature variations, 248-249 transmission, optical, 154, 158, 159-161. See Transmission, objective table of objecti ve lenses, 159-161 water-immersion, 145, 149-150 dipping objectives, 161,209,411,429, 568,613,727,737,870,872 use and limitations, 15 working distance, 5, 9, 129, 145, 154, 157, 198, 249, 511, 568, 598, 643, 673, 678, 727-728, 747, 774, 779, 781,872 x-y and z resolution using beads, 656 OCT. See Optical coherence tomography. OLED. See Organic light-emitting diodes.
967
Olympus, confocal manufacturer, 52-53, 54, 119-120, 161, 184, 187,204,229, 230,234-236,419,421,427,557, 708-709, 727-730, 797, 908, 912. Fluoview-1000, 119-120, 184, 187, 204, 908, 912 DSU disk-scanning confocal microscope, 229-230, 234-235, 908, 913 FRAP system, 210 FRET, 797 high content screening, 811 objectives, 557, 727-730 stick, in vivo objectives, 806 TIRF objectives, 183 transmission, table, 159, 161 SIM scanner, 52-54 tube-length conventions, 157, 239 On-axis reflections, artifact, 171. Onion epithelium (Allium cepa), 390. Online confocal community, Listserv, 390. OPA. See Optical parametric amplifiers. OpenLab, 282. Operational lifetime, of laser tubes, 102. OPFOS, Orthogonal-plane fluorescence sectioning, 672-673. OPO. See Optical parametric oscillators. OPT. See Optical projection tomography. Optical aberrations, 109, 542. See also, Aberrations. thin-disk laser optics, 109 Optical layout of confocal microscopes, 212-213. See also, Optical paths by class, 213 evaluation, 212-213 class I systems, 212 class 2 systems, 212-213 class 3 systems, 213 Optical bandwidth/electronic bandwidth, 32. See also, Bandwidth. Optical breakdown, 198,680,682,685,687, 703,705. Optical coatings, maintenance, 116. Optical coherence tomography (OCT), 609-610. of human retina, 609 schematic, 610 Xenopus laevis embryo, 610 Optical components, chapter, 43-59. Optical density (OD), 71, 81,416. filters, 43, 49-50 Optical disks, 586. Optical efficiency, improvements, 143-144, 216. See also, Photon efficiency. of disk scanners, 216 of light-emitting diodes, 143-144 Optical elements, 43-58, 128,211. confocal Iight beam affected by, 211 of Kohler illumination components, 128 light beam characteristics affected by, 211 chapter, 43-58
968
Index
Optical excitation, diagram, 82. Optical fiber. See Fiber optics. Optical fiber, for scanning by moving fiber tip, 213-214. Optical heterogeneity, specimen, 22-23. reflection, refraction, scattering, 192-197 Optical images, electronic transmission, 5-6. Optical materials, 158, 501. thermal properties, 158, 248-249 Optical parametric amplifiers (OPA), 100-101,112,114-115,118,124.
components, 115 table, 101 Optical parametric oscillators (OPO), 100-102,111-112,114-115,118,
541,600. for CARS microscopy, 600 cavity dumped, to increase white light, 113 tunable, 114-115 table, 101 Optical path. of. 4Pi, confocal, 563 commercial, 566 acousto-optical device, 55 compound light microscope, 156-157 CARS, 599, 601, 907 CARV-2 disk scanner, 230 confocal, 10,208-209,212,632,681 beam-splitter, 213 disk-scanner, 12,216 folded, 166 scanning systems, 214 fiber-optic confocal, 508 interferometers, 243, 245 Kino single-sided disk scanner, 229 LaVision-Biotec, Trimscope, 907 Leica, TCS AOBS, 910 magnetic resonance imaging, 621 Minsky confocal, 5, 25 for measuring photon efficiency, 34 multi-photon, 540, 681, 708-709 multi-focal, 552, 555 spectrometer, 424 Nikon Clsi, 911 Olympus DSU disk-scanner, 230 Olympus Fluoview-1000, 912 optical coherence tomography, 610 optical projection tomography (OPT), 611 Petran tandem scanner, 228 selective plane illumination (SPIM), 613, 673 or simultaneous BSL and fluorescence, 128 surface 3D imaging, SIMIEFlC, 608 surface spherical aberration, 405-406 STED,573 structured-illumination, 266 Visitech VT-Infinity and VT-eye, 914 Yokogwawa dual-disk-scanner, 231, 915 Zeiss LSM-51O, META, 916-917 Zeiss LSM-5-Live, 50, 232, 916
Optical performance, practical tests, 652-660. axial chromatic registration, 658 axial resolution using mirror, 656-657 chromatic aberration, 659 chromatic registration, 657-658 contrast transfer function (CTF), 656 coverslip thickness vs. RI, table, 654 field illumination, 658 flatness of field, 659 Focal Check™ beads, 657, 658 lateral resolution, 655 resolution test slides, 655-656 specimen self-lensing artifacts, 659 spherical aberration correction, 654-655 Optical power, specimen plane, table, 140, 644. Optical probes, 737. See also, Dyes; Fluorescent indicators; Fluorophors; Fluorescent labels. Optical projection tomography (OPT), 610-613. lamprey larva, 612 mouse embryo, 612 refractive index, 613 setup, 611 Optical pulse length, 537-538. See also, Pulse broadening. group delay dispersion, 537-538 group velocity dispersion, 537 measurement, 115,901-903 Optical resonator in laser, 81-82, 116. laser, 81-82 maintenance, 116 Optical sectioning, 9-10, 13, 180, 182,222, 223,236,268-270,469,748, 763-764, 772, 774, 775, 784. See also, Deconvolution, Confocal, etc. algorithms for widefield, 763-764 of A. Thaliana root, 772, 775 with confocal laser-scanning microscope, 9-10 example, 182,463,471,492,656 dynamic imaging, 784 improvement, with deconvolution, 752 latex bead, 3D image, 196 limiting excitation, 223 near surface of living embryo, 748 near to refractive index interface, 180 selective plane illumination, 748 structured illumination, 268-270 with widefield phase-dependent imaging, 13 Optical system, losses, 25-32, 217. Optical transfer function (OTF), 164-165, 490-491,562,563,567,569-570, 578. See also, Point-spread function; Contrast transfer function. 4Pi microscopy, 562, 563, 567 contrast, 164-165 deconvolution for image enhancement, 490-491 15 M, 569-570
point spread function, 490-491. See also, Point-spread function STED comparison, 578 Optical tweezers, 89-90, 110, 218, 383, 385. setups for integrating, 218 table, 385 trapping wavelength, 89-90 Optics, general, 12, 125, 156-157. finite vs. infinity, 156-157 Optiscan confocal endoscope, 213-214. Organic dyes, 109, 203, 342-343, 353-356. See also, Dyes; Fluorophores; Fluorescent labels; Fluorescent probes. AlexaFluor, 353-355 BOPIDY,353-355 classes, table, 355 coumarin, 353, 355 cyanine, 353-355 fluorescein, 353-355 rhodamine, 109, 203, 353, 355 Organic light-emitting diodes (OLED), 143. Orthogonal-plane fluorescence sectioning (OPFOS), 672-673. Oryza sativa. See Rice. Oscillating-fiber scrambler, 8. Osmotic permeabilization, 359. OTE See Optical transfer function. Out-of-focus light. deconvolution vs. confocal microscopy, 461. information, 26, 32, 487, 644-646. Output amplifier, reconstructing analog signal, 64. Output modulation, of semiconductor lasers, 108. Overheating, of filters, 43. See also, Thermal variables. Overlap alignment protocol, montaging, 732. Over-sampling, 60, 70, 728. vs. duplicate-and-smooth process, 70 reasons for, 68 subpixel, resampling, 478-479 Oxygen sensor, 45, 347. p Pack-and-go mode, Power Point, 842, 844. Paeonia suffruticosa, 421. Panda pattern, polarization-preserving fiber, 88. PAS. See Periodic-acid Schiff. Passively mode-locked lasers, Ill. Patch clamp, for loading dye, 360, 726-727, 734, 738-740. Patch pipette, 738. Pattern analysis. See Automated interpretation of subcellular patterns. Patterned-illumination microscopy, see Structured illumination microscopy PC. See Personal computer. PCA. See Principal component analysis. P-CARS, Polarization-sensitive detection CARS.
Index
PCF. See Photonic crystal fiber. PE. See Photoelectrons. Pear (Pyrus serafina), spectrum, image, 711. Pearson's correlation coefficient, 668. Pellicle beam-splitter, 216, 228-229, 231, 346. Peltier cooling. CCDs, 234,447 cell chamber, 387-389 lasers, 85,106-108,111,117 Penetration depth, 177,343,643,672,731, 765. of dyes, 360, 387, 731, 739,882,874 of fixative, 369-370, 376, 857 FRET sensors, 798-799 long laser wavelengths, 109,416,418, 427-428 multi-photon, 381,418,433,435,439, 543,545,558,646,684,708,714, 728, 749, 904 in plant imaging, 779 in scanning electron microscopy, 847 in SPIM, 613, 675-678 TIRF, 177-178 Peony flower, autofluorescent petals, 173-174,176,421,423. Peptide-mediated uptake, 359. Perfusion. chambers, 381, 386--389, 394, 726, 729, 769, 870-873 fixation, 376 Periodic grating. See Grating. Periodic-acid Schiff (PAS) reagent, 262, 369, 770, 774-775, 778. maize pollen grain, 202 Periodically poled (PP) waveguides, 114-115. Perrin-J ablonski diagram, 516, 517, 697, 792. photo bleaching, 697 Personal computer (PC), performance needed for image processing, 289-290. Perspectograph, early studies, 789. Petnin disk, 2, 6, 11, 135, 141,215, 223-224,228,251,265,381,387, 447,458,554. Petran, Mojmir, 2, 6, II, 215, 223, 228. pH imaging, 188-189,221,346,348,359, 386,421,517,529-530,664, 739-740,743, 744. calibration, 421, 530, 745 display, 287 intensity image, 529, 530, 739, 740, 744 lifetime image, 530 pH indicators, 346, 739-742. pH shift/formaldehyde fixation, 370-371, 373. Phalloidin, as correlative marker, 235-236, 344, 376, 378, 694, 696, 756, 804, 854-856. Pharmacological screening, 813-814.
Phase and intensity determination from correlation and spectrum only (PICASO), 115. Phase contrast, 9, 171, 368, 372, 453, 506, 643,649, 731, 851, 854, 890, 892. coherent light for, 130 depth of field, 13 and holography, 7 scanning, 9, 13, 386 Phase fluorometry, 518-519, 526. comparison of FUM methods, table, 526 excitation/emission signals, 519 fluorescence lifetime imaging, 518-519 Phase randomization, to scramble light, 8, 13,84,131-132,143,507. Phase-dependent imaging, depth of field, 13. Phase-shifting interferometry, 245. Phenotyping, 623-624. Phong shading, 308-309. PhonglBlinn models, 308. Phosphoinositide signaling, 799. Photo efficiency. See Photon efficiency. Photoactivatable dyes. See Photoactivation. Photoactivation, 187,224,383, 385, 541, 543-545, 693, 759. example, 759 genetically encoded Kaede, 187,383, 385 Kindling, 574, 760 PA-GFP, IS7, 383, 385, 752, 759-760 table, 385 Photobleaching, 174, 218, 224, 275, 341-342,362-363,545,690-700, 729, 747-748, 759. See also, Bleaching, and Chapter 39. autofluorescence, 698 defined, 218, 691 dynamics, as a source of contrast, 202-203 effect on contrast, 174 fluorescence intensity loss, 691, 694, 696, 698+ fluorescent image of single protein, 699 fluorescent probes, 362-363 fluorescent recovery vs. irradiation time, 699 fluorophores signal optimization, 341-342 choice of fluorophore, 342 fluorophore concentration, 342 light collection efficiency, 217, 341 protective agents, 36, 341-342, 363, 368,375,499,694 spatial resolution, 341 in four-dimensional imaging, 747-748 green fluorescent protein (GFP), 690, 692, 698 intentional See Fluorescence recovery after photobleaching (FRAP) kinetics, 695 mechanisms, 340, 691-693 FRET, 691 multi-exponential fluorescent bleaching, 697
969
multi-photon microscopy (MPM), 545 Perrin-Jablonski diagram of bleaching, 697 photocycling, fluorescent proteins, 698 propidium bound to DNA, plot, 695 reactive oxygen species, 341-342, 362-363, 390, 544, 682-684, 691, 693-694, 852-853 reduction in, 693--696 antifade agents, 36, 341, 368, 375, 499, 694 disk-scanning microscopy, 224 quantum dots, 694 results, in living embryos, 759 of single molecules, 696-698 structured-illumination methods, 275 two-photon excitation microscopy (TPEM), 690, 697 Photocathode, PMT, 28-29, 232-233. quantum efficiency, 232-233 to reduce transmission losses, 28-29 Photoconductivity, in photodetectors, 252, 253. Photocycling, fluorescent protein molecules, 698. Photodamage. See Phototoxicity. Photodetector. See Detectors; Light detectors; CCD; EM-CCD; PMT etc. Photodiode, 134-135,253-255,610, 707-708. feedback, to stabilize laser, 87, 682 feedback, to stabilize arc/filament, 134-135, 137 in hybrid PMT, 29, 30 infrared sensitive for IR lasers, 707 photometer sensor, air space, 26 quadrant, for alignment, 87, 134 of self-aligning source, 134-135 for testing display software response, 830 vacuum avalanche, 254, 255 Photoelectric effect, and LED operation, 137. Photoelectrons (PE), 29, 30, 62-63, 77, 232-234, 254-255, 257, 259-264, 339, 633, 863. amplification of, 62-63 in the CCD, 232-234,495,918,931 production in PMT, 30 single-PE pulse-height spectrum, 29, 77 secondary electrons, as source of PMT multiplicative noise, 77 Photoemissive devices, 256-257. Photoemissive effect, 254. Photographic recording systems, 6-7, 11-12,20,22, 30,71-72, 132, 139, 141,162,207,217, 263, 280,48~ 581+,588,590-591,593-594,607, 613,628-629,633,640,643,712, 829,862, 865-867. "toe" response, quadratic, 71 Photometer paddle, to measure light beam, 26,35,139-140,159,391,650-651, 665.
970
Index
Photometric response, and HO curves, 71. Photomicrography (Loveland), 139. Photomultiplier tube (PMT), 9, 28-31, 35-36,51,62-63,74-75,222,232, 251,254,255,258-261,443,527, 661-662. after pulsing, 257 Bio-Rad,260-261 as confocal detectors, pros/cons, 222 for epi-fluorescence confocal microscope, 9 functioning, 62-63 GaAs photocathode, 28-29, 232, 252, 255,263,527,931 gain from collisions at first dynode, 31 grey levels, 443 hybrid, single-pixel signal levels, 31, 254-255 linearity, 661-662 microchannel plate, 232-233, 255, 262 mini-PMT arrays, 51, 667 multiplicative noise, 28-30, 77, 633, 677,926. See also, Multiplicative noise in multi-channel detection systems, 51 noise and gain, 74-75 number of photons striking per unit time, 35-36 optical enhancer to increase QE, 28-30 photon counting, 21, 29-30, 32-35, 258-259, 260-263, 542 quantum efficiency, 527 vs. cooled CCO, 26-28 signal variation with time, 232 transit time spreads, 527 Photon(s), 20-21, 30, 33-36, 63-64, 132. counting precision, 20-21 uncertainty, 63-64 interactions with photomultiplier tube, 30 lost, 33-36 Photon counting, 21, 29-30, 32-35, 258-259, 260-263, 542. circuits, 33-34, 258, 521 digital representation of optical data, 32-33 effects, 34-35 examples, 35, 263 hybrid PMT, 29-30 pile-up losses, 32-33, 35, 78, 521, 523, 526-527 with PMT, 29-30, 32-35, 258-259, 260, 263 Photon detector types. See Oetectors and entries by each detector type. CCO,254 direct effects, 252 image dissector, 254-255 microchannel plate, 232-233, 255, 262 MCP-CCO, 262 gated, 519, 523-524, 527, 532 photoconductivity effects, 252, 253 photoemissive, 254 photovoltaic, 252-253
thertnal effects, 252 vacuum avalanche photodiode, 254, 255 work functions, table, 252-253 Photon efficiency, 24-36, 215, 217, 341, 631. defined, 24 as a limitation of confocal systems, 24, 223 measuring, 26, 33-36, 217 practical confocal microscopy, 631 of scanners, 215 table listing photon losses, 217 Photon flux, statistics, 256-258. Photon interactions, 252-256. Photon (shot) noise, 660-661. See also, Poisson noise. Photonic crystal fiber (PCF), laser delivery, 1,88,109-110,113,504,541. for white light source, 113 Phototoxicity, 112,363-364,390-391,651, 729,746, 770. chapter, 680-689 in brain slices, 729 damage is higher to either side of raster, 54 factors influencing, table, 363 fluorescent probes, 363-364 live celis, 390-391 reduction, 391 from uneven scan speed, 651 Photo-uncaging, 187,210,383,385,541, 544-545,692,729,760,912. See also, Photoactivation. Photovoltaic effect, 252-253. Phycobiliproteins, 338, 341, 343, 355-357, 693. Physical limitations, 20, 24, 63-64. on accuracy and completeness of data, 20 Poisson noise, 63-64. See also, Poisson noise Physiological fluorescence imaging, 383, 385. PICASO. See Phase and intensity determination from correlation and spectrum only. Piezoelectric effect, defined, 57. Piezoelectric focus controls, 166,215,219, 222,231,241,245,268,468,754, 909. Piezoelectric scanning systems, 57,215, 238,510.555,610. Piezoelectric devices. AOO driver, 54-55, 57 acousto-optical components, 54-55, 57 to align objective, 166 dithering to increase CCO resolution, 70 effect described, 57 to focus objective, 166,215,219,222, 231,241,245,268,468,754,909 laser alignment, 1;7 light scrambler, 84
to move optical fiber, 84 to move scanning mirror, 57, 215, 238, 510,555,610 to move stage, 215, 567 phase-shifter in 4Pi confocal, 609 in structured illumination, 268 optical coherence tomography, 609-610 stretching optical fiber, 609 Pile-up, of pulses. in avalanche photodiode, 253 in photomultiplier tube output, 32-35 measuring risk of, 34-35 p-i-n diode, 253. Pinhole, 26-28, 33-35, 149, 150, 154,201, 210,213,215,224-228,395, 631-632. advantages and disadvantages, 26-28 calibrating diameter, 33-34 confocal, proper use, 28 disk-scanning, 224-228 mini-image detection, 32 optical fiber as, 506-507 optimal size, 226-227, Chapter 22 Fraunhofer formula, 225 position in confocal microscope, 210 practical, in confocal, 631-632 radius, effective, 35 ray paths, different sizes, 226-227 single-mode polarization preserving fiber, 213 small pinholes, effect, 225 of tandem scanners, 215 vibration shifts relative positions, 201 Pinhole disks, critical parameters, 224-228. Pinhole energy, with spherical aberration, 149, 150, 154,631-632. penetration into water, 149, 150 defocus and NA, 150 defocus and wavelength planapochromat, 154 Pixel clock, digitization, 62, 64-65, 201, 234-235,258,903,923,929. CCO, table, 929 Pixels, 38-39, 60, 62-63, 65, 258-259. defining, 60 digitization, 62-63 optimal, 63-64, 66 representing intensity, 258-259 and resolution, 38-39 and Abbe criterion resolution, 38-39, 65 PKzip, 580. Plan objectives, Zeiss, field diameter, table, 152. Planapochromat, 152, 155. See also, Objectives. flatness of field and astigmatism, 152 lateral chromatic aberration, 155 Plancks law, energy of photon, 35, 137,252, 424. Planar illumination, SPIM, optical sectioning, 751.
Index
Plane of focus, distortion, 16,23. by beam deviations, 16 by refractile cellular structures, 23 Plant cell imaging, 769-785. autofluorescence, 770-772 birefringent structures, 162-164. 420-421. See also, Birefringence chamber slides for plants, 429 clearing intact plant material, 166, 417-418,420,439,610,624, 774-775 computer visualization methods, 778 deconvolution, 784-785 direct imaging, 772-773 dynamic imaging, 783-784 effect of fixation, 195, 428 Equisetum, 774 fluorescence properties, 421-428 emission spectra, 421-423 microspectroscopy, 421-426 fluorescence resonance energy transfer, 425. See also, FRET harmonic generation See Harmonic signals fungi, 438-439 genetically encoded probes, 769, 773, 783 green fluorescent protein fusions, 773, 783 of green tissues, 770 hairs, 434-435 history, 769 light attenuation in plant tissue, 414-418 A. thaliana, example, 416 absorption spectrum, 415 effect of fixation, 428 maize stem spectra, 417, 418 M. quadrifolia spectra, 416 M. quadr(folia optical sections, 419 Mie scattering, 162-163, 167,417-418 nonlinear absorption, 416-417 Rayleigh scattering, 162-163, 167,417, 703 light effects on, 770 light-specimen interaction, 425-428 living plant cell specimens, 429-439 calcofluor staining procedure, 424, 438 callus, 429 cell walls, 168-169, 188-189,303, 306,416-417,420-421,428-431, 435-136,438,439,710-711, 713-715, 769-776, 779-781 chamber slides, use, 429 cuticle, 434-437, 715, 717, 779 fungi, 438-439, 624, 782, 870 hairs, 431, 434-436, 772 meristem, 168,420,430,770,776-778, 783 microsporogenesis, 431-432 mineral deposits, 163,420,436-438, 703 pollen germination, 420, 433-434, 78 L 783
pollen grains, 202, 305, 313, 420, 431-433,553,558,781,783 protoplasts, 195-196,203,416,421, 423-427,429-431,438-439,693 root, 172, 174,303,307,421,429, 430+,438,464-465,556,772-773, 775, 777, 779-783 culture chamber, 429 starch granules, 202, 420-421, 428, 432-433,435,703,710-712,715, 719 stem, 168, 172, 180,417-419,421, 424,429,556,707,710-711, 713-714 storage structures, 435-436 suspension-cultured cells, 189, 429-430 tapetum, 433-434, 779 waxes, 420,428, 434-435, 714-715 new spectral tools, 770 obtaining spectral data, methods, 772 penetration values, 779 photodamage, 770 point spread function, 722, 784 refractive index heterogeneity, 192, 418-420 single-photon confocal excitation, 772-778 specific methods, 769 spectral unmixing, 770 examples, 665-666 staining, 774 technological developments, 769 textbooks, 769 three dimensional, 771 clearing agents, 166,417-418,420, 439,610,624, 774-775 deconvolution protocols, 784 reconstruction, 775-776 segmentation, 776-778 two-photon excitation, 415-419, 421, 423 advantages, 778-779 best conditions, 781 compared with one photon, 421 cell viability, 779-782 deconvolution protocols, 784 dyes, 782 green fluorescent protein, 782-783 light-specimen interaction, 425-427 microspectrometer, 424 pitfalls, 782 thick specimens, 779 in vivo, 781 Plasma membrane, microscopy. See Total internal reflection microscopy (TlRF). Plasma light sources, spectra, 44. Plasmid DNA, nick-damage, 684, 724, 802-804. See also, Microinjection; Electroporation; Biolistic transfection. Plasmodesmata, 777.
971
Plumbago auriculata, fluorescence spectra, 422. PMT. See Photomultiplier tube. p-n diode, 253. See also, Photodiode. PNG (Portable network graphic), 581, 584. Pockels cell, variable beam attenuator, 25, 54,57,87,116,543,701,903-904. Pockels effect, in crystals, 57. Point-spread function (PSF), 4, 10, 23, 27, 39,68-70, 145-146, 189-190,208, 223,239-250,271,275,330,378, 405,407,409,446,448,453-457, 485-486, 489-494, 536, 562-564, 570, 574, 578, 635, 656, 674, 750, 784, 830, 895. 3D, 68-70, 247-248 4Pi microscopy, 562-563 additional information from, 570 space invariance of PSF for, 564 apodization, 240, 243, 249, 250, 272, 567, 889 blind deconvolution, 468, 485 in botanical specimens, 772, 784 in brain slices, 729 calculations, RI-mismatch, 407 for glycerol, table, 409 for water, table, 409 CARS, 596 comparing widefield with confocal, 27, 453-457,493,644-647 confocal, 10, 12, 20s+, 212, 216, 405, 632,681 vs. deconvolution, 27, 453-457, 493, 644-647 deconvolution, 189-190,223,489, 490-494, 784. See also, Deconvolution quantifying PSF, 492-494 deformation caused by RI anomalies, 22-23 Fourier transform, 489, 490 lateral resolution. See Lateral resolution measuring, 240-242, 455, 462, 471, 656 amplitude/phase, 242 fiber-optic interferometer, 240-241 images, 246-248 high-NA objectives, 239-250, 492, 656 pupil function, 240 for 3D deconvolution, 145-146 non-linear, 552, 750 and Nyquist, 635, 636, 751, 752 optical transfer function, related to, 490-491 polarization effects, 249-250 pupil function, 245-248. See also, Pupil function Rayleigh-criterion and Nyquist sampling, 39 refractive index mismatch, 405, 407 spherically aberrated, 148-150, 407, 492 shape in telecentric systems, 208 SPIM,674
972
Index
Point-spread function (PSF) (cant.) STED, diagram, 574, 578 structured illumination see Structured illumination microscopy temperature effects, 25, 85, 248-249, 630 terminology, 405 Wiener filtering, 494 Points, defined, 59. Point-source, for measuring photon efficiency, 33. Poisson noise, 20-21, 29, 37, 63-64, 67, 69, 74-75,81, 164-165,211,232,234, 442,456,460-463,468,487,495, 497,633-636,647,651,655,660, 693,784,835,923-924,926. See also, Quantum noise, Shot noise. bleaching, 693 ofCCD charge transfer, 920 dark charges, 921-922 CT imaging, 615 and display linearity, 72-73, 588 digitization, as part of signal, 65, 69, 633-636 of EM-CCD, 233-235, 262, 927-928 and FUM, 524-525 and gray levels, 74 importance of deconvolution, 38-41, 60, 69, 189-190, 222-223, 320, 399, 471-472,481,495,751-753,835 intensity spread function, 75-78 photomultiplier tube, 74-75 affects effective QE, 31 multiplicative noise, 29, 647, 660 in photon detection, 63-64 and pixel size, 64, 68, 633-636, 928 practical effects, 67 single-pixel noise, 65, 67, 190, 635, 832, 835-836 spectral unmixing, 667, 770 examples, 665-666 structured illumination, 278 uncertainty in contrast, 74, 164-165 and visibility, 37, 667 Polarization, 13,49,57, 83, 88, 89, 211-212. attenuator, 43, 543, 907 beam-splitter, 13,50-51,57,85,87, 100, 171,217,513,631,904 to avoid spectral distortion, 49 circular or phase randomized, 211-212, 229 effect on AODs, 55 effect, of dichroic beam-splitters, 34, 49-50 Kerr cell, 111, 113,516 of laser light, 8, 83, 88-89, 113, 478, 558 optical components, 57,155,211 optical fibers, 213 Pockels cell, 25, 54, 57, 87,116,543, 701,903+ rectified DIC optics, 846
to reduce refiections, 6, 25,141,158,171, 516,229. See also, Antifiex system scramblers, 8, 84, 132, 143 Polarization effects, 211, 249-250, 503. birefringence, 188, 420-421, 431, 434, 436, 438, 480, 503. See also, Birefringence blind deconvolution, 479 and CARS microscopy, 595, 600-604 high-NA objective lenses, 249-250, 267 interaction with nucleus, 23 optical fibers, 503, 507 stereo image displays, 299, 589 Polarization microscopy, 43, 50-51, 154, 156, 162, 188, 288, 348, 438, 479-480,513,555,711,714-715, 717,719,891,894. centrifuge microscope, 8 of collagen fibers, 164, 188, 717 DIC, 10, 14, 127, 146,468,473 and FRET, 793 and harmonic generation, 179, 428, 704-706,717, 719 MFMP, 555 mitotic apparatus, 15, 717 p- and S-, and incidence angle, 50-51 Pol-scope, 13. 188, 432, 468, 480 PSF, 406-407 to regulate light intensity, 43 STED,578 Polarization noise, in lasers, 83. Polarization-preserving fiber, 49, 87, 503, 505,507. as a pinhole, 213 Polarization-sensitive detection CARS (PCARS), 600, 601, 604. adipocyte cells, 604 Polarized light, 7, 14,83-85, 146, 158, 162, 171,229,406-407,420,479,894. deconvolution, 479 image formation, 406-407 PSF,479 Polarizer, 83, 128, 188, 249, 268, 275, 420, 479, 711, 903-904. for antifiex, 6, 84, 141, 158, 229 for attenuation, 43, 85, 87-88, 543, 903-904 for CARS, 601 Glan-Taylor, 85, 87, 100, 171 Glan-Thompson, 85,904 LCD, 589, 715 micro-wire, 85 structured illumination, 264 tunable,715 Pollen germination, 433-434. Pollen grains, 202, 305, 431-433, 438, 553, 556, 558, 678, 781, 783. germination, 433-435, 783-784 multi-focal multi-photon imaging, 556 Pol-scope, 13, 188,432,468,479-480 test specimen, 195, 269, 313, 553, 556, 678
Pol-scope, 13, 188,432,468,479-480. deconvolved, 479 images, 15, 188, 479, 717 Portable network graphic. See PNG. Position, accuracy in CLSM, 40. Position sensors, galvanometer, 53-54. Posterizing, 591. Potassium titanium oxide phosphate (KTP) crystal for non-linear optical frequency conversion, 107. Potato (Solanum tuberosum) SHG signal, 712. Power requirements, for lasers, 65, 80-81. Power spectrum. See Contrast transfer function. Power supply, laser as noise source, 86. PP, Periodically poled waveguide, 114-115. Practical confocal, 2-photon microscopy, tutorial. See also, each topic as a major entry. 2-photon excitation duty cycle, 644 peak power level, 644 photodamage vs. penetration, 645 power vs. penetration, 646 3D microscopy methods compared, table, 647+ best 3D method for, 644-647 biological reliability, 631 bleaching pattern, 627-628 quantum efficiency, 628 chapter, 627-649 confocal images with few photons, 634 deconvolution, factors, 646 filling back-focal plane, 210, 509, 629, 633 focus, compensating drift, 395, 732 getting a good confocal image, 629-631 alignment of optics, 629-630 back-focal plane (BFP), 210, 509, 629, 633 focus, 629 low signal, 631 mirror test specimen, 630 no signal, 631, 660 simultaneous BSLlfiuorescence, 631 getting started, 627 Kohler illumination for transmission, 34, 127-128,131,229,627,648-649 multi-photon vs. single-photon, 646 new controls, 631-636 biological reliability, 631 pinhole size, 631-632 pixel size, 62, 634-635, 784, 928 Nyquist reconstruction/deconvolution, 635-636 over-sampling, 635 photon efficiency, 24-26, 215, 217, 341, 631 pinhole summary for, 26-28, 633
Index pixel size, 62, 634-635, 784, 928 measuring, 635 summary for, 636 poor performance, reasons, 640-643 air bubbles, 643 curvature of field, 641 dirty objective, 642-643 imaging depth, 643 under filling objective pupil, 642 optical problems, 640-641 sampling problems, 640 singlet-state saturation, 643 under-sampling, 635 schematic diagram, 632 statistical considerations, 633-634 stray light, 201, 632, 904 test specimen, 636-640 description, 636-637 diatom, 638-640 figures, 637-640 reasons for, 636 widefield vs. beam scanning, 647 Prairie Technologies, LiveScan Swept Field design, 237. Pre-amplifier, in digitizing analog signal, 64. Precompensation, in fiber optic cables, 88. Presentation software, 829-845. helpful URLS, 844-845 movies, 837-844 artifacts, 839-840 coding limitations, 838 compression of large movies, graph, 843 compression of PAL TV movies, table, 842 digital rights management, 844 entropy, 841 frame count matching display cycle, 838-839 MPEG display formats, 840-841 overlaying, 844 Pack-and-go mode, 842, 844 performance benchmarks, 841-842 region code, 844 remote use, 842-844 rules 837-838, 844 up-sampling, 838-839 very high resolutions, 841 precautions, 829-830 testing, 830-836 aliasing gallery, 834 aligning images, 835 brightness, 832 changing display size, 832-835 codecs, 831 compression, 835-836 compression artifacts, 837 cropping, 835 digital rights management (DRM), 830 down-sampling in PowerPoint, 834 fast graphics cards, 831, 832 gamma, 832-833
measuring display speed/sensitivity, 830 random color dot image, 836 reference images, 830-831 removing distortion, 835 resolution, 832-835 rotating, 835 scaling, 835 screen capture, 830 static image performance, 831 step image, 833 under-sampled image, 835 up-sampling, example, 834 viewer, 830 Preventive maintenance, lasers, 115-116. Principal component analysis (PCA), 731-732. Printers, 591-593. aliasing, 592 color images, 592 grey levels, 592 ink jet, 593 laser, 593 posterizing, 591 scaling techniques for, 592 Prionium, MMM image, 556. Probe, mismatch with pixel shape, 39. Processor performance, 3D-image display, 289+. Projection/compositing rules, 3D-image display, 302-304, 763-764. alpha blend, 302, 304 average intensity, 302 first or front intensity, 302 Kalman average, 304 maximum intensity, 302 Propidium iodide, 344, 355, 360, 426, 651,693-695,773, 778-779, 782, 812. dead cell indicator, 426, 651, 875, 877 Proteins, 195, 756, 760, 794-795, 804. See also, Green fluorescent protein, etc. chimeric fusion, 794 fluorescent, FRET, 794-795 Kaede, 187, 383, 385 Kindling, 760 microinjection, 804 PA-GFP, 187,383,385,752,759-760 tagged, 756, 758 translational fusions, 756 UV absorption, 195 Proteomics, 237, 790, 804, 809, 818, 867. location, 825 Protoplasts, 195,416,429,430,431. A. thaliana, 195-196,203,416,421, 423-427,429-431,438-439,693 Proximal tubule, labeled, 744. Pseudo color display, 173-175, 190, 291. PSF. See Point spread function. Pulse broadening, 88, III, 210, 537-538, 543, 606, 609, 728, 903. Pulse length measurement, 115,901-903. Pulse spreading. See Pulse broadening.
973
Pulsed lasers, 81, 96-100,110-114,120, 137. See also, Lasers; Ultrafast lasers. broadband tunable, table, 120 diode, table, 96-97 DPSS, table, 98 dye, table, 96 excimer, table, 96 for FUM, 537 kits, table, 98, 100 nitrogen, table, 96 scanning only region of interest, 237 for 2-photon excitation, 81 ultrafast, table, 99-100 vapor, table, 97 Pulse-counting mode, 21, 29-30, 32-35, 258-259, 260, 263. Pump sources, for dye lasers, 103. Pumping media, maintenance, 116. Pumping power vs. frequency cubed, 65, 82. Pupil function, 211,245-248. 3D point spread function restored, 247-248 4Pi, 566-567 AOD,56 empty aperture, 248 of human eye, 72, 128 intermediate optics, 211, 222, 225, 250 Kohler illumination, 34, 127-128, 131, 229,251,627,648-649 Mach-Zehnder interferometry, 245 measurement, 246-248 images, 246-248 objective, 24,155,158-159,211, 239-240,242,492,551-552,554, 566-567, 650 orthonormal Zemike polynomial for, table, 247 phase-shifting interferometry measuring, 245 polarizing effects, 249 pupil plane, 50 See also, Back-focal plane transfer lens, 728 view of pupil image, 629 Zemike polynomial fit, 245-247 Purkinje cells, Golgi-stained, 167-168. Pyrus serotina. See Pear.
Q QE. See Quantum efficiency. Q-switched pulsed laser systems, Ill, 114-115. Quantitative analysis, flying-spot microscope, 6-7. Quantization, limitations imposed by, 37-39. See also, Chapter 4. Quantum dots, 221, 343, 357-358, 360-361, 656,694,696,757,801,814,846, 853. See also, Semiconductor nanocrystals. assays for, 814 in electron microscope, 852-854 FRET,801
974
Index
Quantum dots (cant.) labeling, 853 toxicity, 357, 694 Quantum efficiency (QE), 25-30, 74-78, 222,232-234,238,251,254-255, 349,355,375,383,390,442-443, 459,516,527,575,628,646,556, 703,751, 793, 920-922. of back-illuminated CCO, 77-78 charge-coupled device (CCO), 26-28, 74-76,142,215,232,234,257-258, 261,644, 707, 751, 754, 810, 920-921 comparative among CCO cameras, 76 effect on Poisson noise, 74-75 effective QE, of photon detectors, 28, 29 of electron-multiplier CCO, 4, 30, 59, 234, 920 FLI~, 516-517, 520, 523, 526-527, 529, 530 FRET,792 of human eye, 251 and intensity spread function, 74-75 and multiplicative noise, 77 optical enhancer, to increase QE, 28-30 optimal 30 microscopy, 644 photomultiplier tube (P~T), 26-28, 51, 77,222,257,262,527,707 graph,29 table, 707 signal-to-noise ratio, 263, 442-443 variation with wavelength, 29 vs. wavelength, 922 Quantum noise, 21 63-64,69-70,468,472. See also, Poisson noise. and approximation, for reconstruction, 69-70 Quantum wells, as absorbers, 111. Quantum yield, of fluorescent dyes, 172, 180,338-345,353,360,363,383, 421,543-544,574,661,683,690, 710, 737, 792, 794-795. Quartz-halogen lamp, control, 138-139. See also, Halogen lamps. R
Rabbit, 237, 744. antibodies, 855, 877-878 kidney proximal tubule, pH, 744 Radiance, of non-laser light sources, 126, 132, 137-139, 141. measuring with radiospectrometer, 139 Table, 1140 Radiospectrometer, radiance vs. wavelength, 139. Raman background, in glass fibers, 88, 90, 162,506-507. lower in large-mode-area, fiber, 110 Raman scattering, 162-163, 167,339-340, 348, 506-507, 545, 697. and bleaching, 697 defined, 162
Raman spectroscopy, 48-49, 90, 167,254, 339-340,507,545,697. See also, CARS. CARS, 204, 550, 577, 595-605 chemical imaging, 90 hard-coating on interference filters used, 48-49 image contrast, 167 Ramp-up, for light sources, 136, 137. and long-term stability, 137 and short-time stability, 136 Rare earths, for doping fiber lasers, 11 O. Raster, 62-64. convolution, 485-486 dimensions, in specimen, 63 retrace, 25, 33, 53-54, 219, 338, 389, 543, 628, 651, 908. See also, Retrace, raster scanning shape, 63 size, vs. pixel size and light dose, 64 temporal limitations, 141 Raster scanning, 5-6, 25, 141-142, 223, 540,596. alignment, 629-630, 651 assymmetrical sampling, 38-40 bleach pattern, 30, 538, 628, 693 chromatic aberration limitations, 156, 640-641 damage is higher to either side of raster, 54 display, 830-831, 835 distortion, 40 and electronic bandwidth, 70, 238 for fast confocal imaging, 223 fiber-scanning, 214,508 galvanometer limits, 52-54, 223, 651 limitations imposed by AOOs, 56 ~PEG formats, 840 Nyquist sampling, 38, 41, 59-60, 62, 634-635 off-axis aberrations, 151, 640-64\, 659-662 pattern on Nipkow disk, 5-6, 223-225. See also, Nipkow disk scanning retrace gating, 25, 54, 56, 219, 389, 543, 628,651,908 scan angles, 209, 214 stability, 708 sampling in time and space, 141-142 timing, 33, 53, 753 zoom, raster size and magnification, 11, 24,37,63-64,66,70,79,317,389, 493,627,634-636,655-658,683, 731 Rat, cells and tissues, 205, 320, 323, 330, 398, 739, 813. brain slices, 393, 398, 686 CA 1 region, 323 cardiac muscle, 498, 529, 556 cerebellar granule neurons, 813 EOL muscle, calcium, 740 fixation, 370, 372, 393 hippocampus, 268, 317, 341
interossi muscles, SNARF-l, pH image, 739 intervertebral disk, 310-311 kidney, 511, 803 leukemia cells, 347, 520-521 FLI~ image, 521 neuron, membrane potential, 205 tooth, 667 Rate, imaging, limited by signal level, 73. Ratiometric imaging, 189, 346-347. See also, Calcium imaging, pH, etc. bleach ratio, 697-698 calcium, 736-737, 850. See also, Calcium imaging CARS, 600, 602, 604. See also, CARS concentration calibration, 742-745 to detect colloidal gold labels, 167 to determine ionic concentration, 36 FLI~, 516-532. See a/so, FLI~ FRET, 174, 184,790,794-795,797-798. See also, FRET glutaraldehyde autofluorescence assay, 369 HCS, high-content screening, 813, 823-824. indicator choice, 738 interpretation, 740-741 live/dead assay, 875 pH, 739-744. See also, pH imaging structured illumination. See Structured illumination microscopy water-immersion objectives, 737 Rayleigh criterion (Abbe criterion), 1-3, 9, 37-39,60-61,66, 129, 146,486, 703, 822, 928. breaking the Abbe/Rayleigh barrier, 571-573 Nyquist sampling, 39, 60, 66 of two point images, 1-3, 146 Rayleigh scattering, 162-163, 166, 167, 339, 342, 417, 703, 747. compared to ~ie scattering, 163 in embryos, 747 by coJloidal gold labels, 167 light attenuation in plant tissue, 417 wavelength dependency, 162-163 Rayleigh unit, 147. Reactive oxygen species (ROS), 341-343, 362-363,390,544,682-684,691, 693-694, 852-853. See also, Bleaching; Phototoxicity. as basis of correlative TE~ staining, 852-853 Readout noise, 74-75, 77, 232. See also, Noise. and readout speed, 77 Real image, disk-scanners, 224. Real-time 20 imaging, 12-13, 167-168, 215,222-224,232,235,307,496, 542. Real-time 30 imaging, 154. Receptors. cholera toxin, 790-791, 796-797, 802
Index
deconvolution, 495 EGF,533 ERD2, 791, 796 fibrinogen, 846-847, 850 high-content screening, 809, 812-814 KDEL, 790, 797 ligands, 354 lipid, 790, 791 proteins, 357 Streptococcus, 879 transferin, 819 uncaging, 545 Reconstruction, 3D. definition, 280. Nyquist and filtering/deconvolution, 59, 66-67, 69, 70, 173, 235-236, 280-315, 458, 468-469, 474-475, 496-497, 563, 585, 603, 607, 610, 615, 635, 672, 675, 677-678, 690, 772, 730, 731, 762, 77, 774-776, 778, 784, 883 Recording times, 141-142. in widefield microscopy, 141-142 using LED source systems, 141-142 Recovery curve, after bleaching, 187. Red fluorescent protein (RFP), 221-222. Reference list. historic, 889-899 lasers, 123-125 Reflected-light images, 180, 181. See also, Backscattered light. confocal, of integrated circuit, 180 of glass bead, in water, 181 Reflecting objectives, constraints, 156. Reflection contrast technique, Antiflex, 159. Reflection mode, low coherence light, 130. Reflectivity, optical surfaces, 159, 163, 167-171. anti-reflection coatings, 158 on-axis, artifact, 168-171 refractive index, 159, 163, 167 Refracting regions affect imaging beam, 15-16. Refractive index, (RI), 14-15, 23,45, 148, 152, 163, 198,377,404-413, 418-420,613,654. See also, Spherical aberration; Dispersion. anomalies in, effect on PSF, 23, 418-420 of biological structures, 163,377 table, 277 of botanical specimens, 418-420 coverslip thickness, importance, table, 654 of immersion medium, 277, 411 effect on PDF, 23, 418-420 effect on sharpness, 14-15 effect of wavelength and temperature on, 148, 248-249, 411 and intensity, and spectral broadening, III
of layers in interference filters, 45 of mounting media, table, 198, 342, 370-371,373-377
of optical glass vs. wavelength, 152 optical projection tomography (OPT), 613 self-shadowing, 198 temperature, 148, 248-249, 411 of tissue/organs, table, 377 Refractive index mismatch, effects, 404-413. See also, Spherical aberration. table for glycerol, 409 table for water, 409 calculation, 404-407 dependence of focal shift, 410 diagram, 404 dry objectives, 410-411 experiments, 409-410 water/glycerol results, table, 410 field strength calculation, 405 other considerations, 410-413 spherical aberration correctors, 15, 151, 147,192,411-412 terminology, 405 actual focal position (AFP), 405 focal shift, 405 nominal focal position (NFP), 405 theory, 404-407 Region code, for MPEG-encoded movies, 844. Region-of-interest (ROI), 835. brain slice, 726, 733 diagonal, 658 display presentation, 835 embryos, 747, 759 FRAP, 51, 187 FRET, 797, 80\ in image processing, 289, 300, 323, 330, 676 labeling, 353 must be smaller at high resolution, 577 nanosurgery, 219, 686 photobleaching, 690 preprocessing, 676 rapid acquisition, 236-237 structured illumination, example, 272 viability studies, 683 Registration synthesis method, 328-331. defined, 328 landmark-based, 328-329 multi-view deconvolution, 330 Relationships, in fluorescence microscopy, 80. energy per photon, 80 flux per pixel, 80 photons/s vs. wavelength, 80 Relative motion, objective vs. specimen, 39-40. Relaxation, in laser energetics, 82. Relay optics (tel an lenses), 145, 157,214, 455. Reliability. of 3D image, 461,517 biological, vs. damage, 24, 68, 631, 633 lasers, 80, 102, 115 living cell work, 387
975
mirror position, 40 photometric, 312 spectral detectors, 662 Removable storage media, 585-588. random-access devices, 586-588 compact disks (CD), 586-587 digital video disks (DVD), 587-588 floppy disks, 586 magnetic disks, 586 MO (magneto-optical) disks, 586 optical disks, 586 WORM (write once, read many) disks, 586 Rendering, of 3D views, 280, 285, 290, 30\, 307,309,311,377,749,762,764. definition, 280 voxel speed, 290 RESOLFT microscopy, 571-574, 577. See also, STED. breaking diffraction barrier, 571-573 concept, 571-573 different approaches, 573-574 ground state depletion (GSD), 573 STED, 573-574 outlook for, 577 resolution, new limiting equation, 571 triplet-state saturation, 573 Resolution, 1, 4, 13, 16, 24, 36-41, 59, 61, 65-67,210. See also, PSF; FWHM. adequate levels, 36-41 axial, 13 axial-to-Iateral ratio vs. NA, 4 back-focal plane diameter, table, 210 confocal vs. non-confocal, 16 and contrast transfer function, 37, 59, 61 estimating, 65-67 measured, widefield, 16 minimum resolvable lateral spacing, 1, 16 spatial and temporal, 24 sufficient, 36-37 Resolution, structured illumination. Fourier-space, 270-271 linear image reconstruction, 271 Lucosz's formulation, 273 methods, 270-276 Moire effects, 270-271 photobleaching, 275 reconstruction results, 272 standing-wavefield microscope, 275 thick samples, 274, 275, 278-279 Resolution scaling, STED comparison, 578. Resolution test slides, 16, 656. Resonant cavity, laser, 81-82, 111, 115. Resonant scanners, 52-54, 56-57, 223, 447, 543. acceleration distorts mirror shape, 53 blanking, 25, 218, 338, 389, 543, 628, 651,908 compared to acousto-optical deflector, 56 duty cycle, 52 galvanometer, 52 multi-photon excitation, 543 raster-scanning, 33, 53-54, 56
976
Index
Resonant scanners (cont.) retrace. 54, 56. See also, Retrace, below. scan speed, 54 Retrace, raster scanning shape, 25, 54, 56, 219,389,543,628,651,908. acousto-optical deflector, 56 blanking, 25, 219, 338, 389, 543, 628, 651,908 raster-scanning, 33, 53-54, 56 Review articles, listing, 889. RFP. See Red fluorescent protein. Rhodamine, dyes, 81,109,116,136,140, 203,264,292,339,342-345,353, 355,362-363,375-378,409,538, 553, 592, 693, 697-698, 762, 783-784, 794, 851, 854-856. arsenical derivatives, 348 bleaching, 697, 698 calibration plot, 661, 851 excitation of, 181, 109 fluorescence correlation spectroscopy, 693 FRET,347 photobleaching quantum yield, 363 planar test specimen, 538 power for 1-, 2-photon excitation, 81, 3 41 Rhodamine-123, 374, 389 resolution measurement, 409 stability and cost, 116 Rice (Oryza sativa), 168, 171,414,415, 712,715. absorption spectrum, 415, 706 backscattered light image, 168, 171 emissions spectra, autofluorescence, 713 leaf fluorescence images, 714-715 light attenuation in plant tissue, 414 silica deposits, 714-715, 717 Richardson-Lucy, deconvolution, 497, 568. Richardson Test Slide Gen III, 652, 656. RLE. See Run-length encoding. RNA, microinjection of, 803, 804. RNA labels, 344, 369,465,531-532,612, 691, 758, 874-875. ROI. See Region of interest. Room light, as stray signal, 201, 632, 904. Roots, plants, 172, 174,303,307,421, 430-432,438,464-465,556, 772-773, 775, 777, 779-783. maize, image, 432 mounting, 429, 431 ROS. See Reactive oxygen species. Rose Criterion, 37-38, 68, 164, 633. relationship with signal-to-noise ratio, 164 for visibility, 37 Rotating, specimen, 188, 568, 835. micro-CT, 615 optical projection tomography, 610-611 SPI~, 672-673, 676, 751 Rotor, galvanometer, detecting position, 53-54. Run-length encoding (RLE), 580.
S Safety, 83, 85, 90, 115, 117-118, 124, 132-139,900,903,904. arc sources, 132-139 beam-stop design and use, 118.903-904 classification of laser systems by hazard, 117 cleaning objectives, 642 display geometry, 297 equipment needed, 900 eye protection against Brewster surface reflections, 83 goggles, 118 with external-beam prism method, 90 fiber optics for transporting laser light, 88 hazardous materials fluorescent laser dyes, 85, 103, 116 used beryllium oxide tubes, 115 high pressure Xe lamps, 136 monitor power to avoid explosions, 138-139 in disk-scanning confocal microscope, 231 laser, 117-118, 839, 900, 903-904 installation requirements, 85 monitor power to avoid explosions, 138-139 references, list, 123 safety curtains, 117, 904 training, 118 SA~, saturable absorber mirror, 111. Sampling. See Digitization, 20, 63-64. non-periodic data, 38 optimal,63 Saponin, formaldehyde fixation, 359, 375, 856. Saturable absorber mirror, pulsed lasers, 111. Saturable Bragg reflector (SBR), 111. Saturable output coupler (SOC), 107, Ill. Saturation, singlet-state fluorescence, 21-22, 41,142,265,276,339,442,448, 450,643,647, 899. performance limitations, 81, 450, 928 SBR, saturable Bragg reflector, 111. SBT. See Spectral bleedthrough. Scaling techniques, 592, 835. Scan angle, and position in image plane, 209-210. Scan instability, detecting, 40-41. Scan raster, testing, 651-654. malfunctioning system, 653 photo toxicity from uneven scan speed, 651 sources of fluorescent beads, table, 653 well-calibrated system, 652-653 x and y galvanometers, 651-652 z-positioning calibration, 652, 654 stability, 652 Scanned-slit microscopes, table, 224. Scanner arrangements, evaluation, 213-215.
Scanners, 51-55, 57, 214-216. acousto-optical deflectors, 55. See also, AODs mirror arrangements, 214 evaluating, 215-216 mechanical, 51-54. See also, Gal vanometers piezo-electric, 57, 215, 238, 510, 555, 610 single mirror/double tilt, 215 sinusoidal, "tornado" mode, SI~ scanner, 52 Scanning electron micrographs, 428, 434, 437, 846-848, 850-852. Scanning laser ophthalmoscope (SLO), 480. Scanning fiber-optical microscopy. See Fiberoptic confocal microscope. Scatter labeling for tracing lineage, 461, 462. Scanning systems for confocal light microscopes. See also, Galvanometers; Disk-scanning confocal microscopy; Acoustooptical deflectors; Linescanning confocal microscopes; Raster. Lissajous pattern, circular scanning. 554 "tornado" mode, SI~ scanner, 52 Scattering, 162-163, 167-171,550. coherent anti-Stokes Raman (CARS), 550 elastic, Rayleigh, 162-163, 166-167,339, 342,417,703-747 Raman, 162, 167, 339-340, 348, 506-507,545 and reflection contrast, 167-171 Scattering object, viewed by TIR~ 177. See also, Backscattered light. Schiff reagents, 262, 369, 770, 774-775, 778. Schottky diode, photo detector, 253. Scientific thought, four aspects, 789-790. Scion Image, 281-282, 395, 730. Scramblers, light, 8,13,84,131-132,143, 507. Screen capture, 830. Screens, to enclose laser beams, 118. SD. See Static discharges. SDA. See Stepwise discriminant analysis. Sea urchin, S. purpuratus, 173, 198, 200. Second harmonic generation (SHG), 90, 114-115,166-167,179,188,550, 552, 556, 703-719, 729-730. See also, Harmonic signals. as autofluorescence, 361 cell chambers, 166,429,552 detectors, 706-708, 728 disk-scanning, 552, 556 double-pass detection, 166-167 table, 706-708 crystalsforSHG, 103, 107, 114-115, 188, 703 energy relations, 705 in lasers, 103, 107, 114-115 layout, 166, 191, 552, 708-709, 712
Index
light attenuation spectra, 706 light sources, 706-708 brain slices, 729-730 non-linear optical microscopy, 704-705 optically active animal structures in, 714-717 brain slice, 729-730 collagen structure, 703, 717 sarcomeres, 716 spindle in mouse zygote, 717 spindle in zebrafish embryo, 718 structures producing SHO, table, 715 table of structures, 715 zebrafish embryo, 716, 718 optically active plant structures, 428, 710-714 Canna, nonlinear absorption, 710 cell wall, 428, 711, 714 Commelina communis, 712 emission spectrum of maize, 710, 711 Euphorbia pulcherrima, spectrum, 710 mineral deposits, 436 Pyrus serotina, spectrum, 711 rice leaf, 712, 715 starch granules, 433 maize, 710-711, 713-714 emission spectrum, 710-711 leaf spectrum, 710 pol-microscopy, 711 stem, optical section, 714 stem, spectrum, 710, 713 chloroplasts, tumbling, 713 membranes of living cells, 90 mineral, deposits, 436 photo detector suitability, table, 706-707 polarization dependence, 71, 717-720 potato, as SHO detector, 712 pulsed laser suitablity, table, 706 signal generation, 179,552,597, 704-705 spectra, 706 spectral discrimination, 421 starch granules, 433 Segmentation, FUM image, 527-528. Segmentation methods, 281, 283-285, 290, 300-302,304-306,309,311-312, 316-319,321-330,333-334, 527-528, 776-778, 812. 3D, 776, 822, 828 automated, 818, 821-822, 828 background, 321 blob segmentation example, 322-324 gradient-weighted distance transform, 323 model-based object merging, 323-325 watershed algorithm, 323-324 boltom-up, 321 combined blob/tube segmentation, 328-330 foreground, 321 hybrid bottom-up/top-down, 322 integrated, 322 intensity threshold-based, 321 object, 321
for plant cells, 774-777 balloon model, 776 watershed algorithm, 322-325, 777, 822 region-based,321-322 top-down, 322 tube-like object segmentation, 324-328 mean/median template response, 328 skeletonization methods, 324-325 vectorization methods, 324-327 validation/correction, 333-334 manual editing, 333-334 Selective plane illumination microscopy (SPIM), 613, 614, 672-679, 751. 3D scanning light macrography, 672 anisotropic resolution, 678 applications, 675 axial resolution, 674-675 vs. CLSM, 678 Drosophila embryogenesis, 675-676, 747-748,751-752,754, 756, 759, 804,810 and FUM, 527 images processing, 675-678 image fusion, 676-677 pre-processing, 676 registration, 676 lateral resolution, 674 light-sheet illumination, 672-674 light sheet thickness, 674-675 Medaka,614-615 heart image, 614 embryo image, 675 multi-view reconstruction, 675-678 point spread functions (PSF), 674 schematic setup, 613, 673 thin, laser light-sheet microscope,TLLSM, 672 Self-aligning arc source, 135. Self-shadowing, 165, 174, 194, 195. in confocal optical sections, 174 spherical structure, 195 in epi-fluorescent mode, 165, 194 SEM. See Scanning electron microscope. Semi-apochromat, pros and cons, 158. Semiconductor lasers, 86, 105-108. noise sources, 86 Semiconductor nanocrystals (quantum dots), 221, 343, 357-358, 360-361. 656, 694, 696, 757, 759, 801, 814, 846, 853. as probes, 221, 757, 759 Semiconductor saturable absorber mirror (SESAM), 107, Ill. for self-starting intense optical pulse trains, III Sensiti vity, video photodetectors, 6-7. Sensitized emissions, of acceptor, 795-796. See also, FRET. Sequential devices, 585-586. Serial sampling, single-beam confocal, 20. SESAM, Semiconductor saturable absorber mirror, 107, III.
977
SFP. See Simulated fluorescence process. Shannon, Claude, 64-65. Shannon sampling frequency, defined, 64, 443. SHOo See Second harmonic generation. Shift invariance, deconvolution, 457, 490, 564. Short-pass filters, 43-44. Shot noise, 232, 256-257, 286, 442, 460-461,495,558,660-661,831. See also, Poisson noise, Quantum noise. Signal, 27, 62. See also Speed relationship to magnification, 62 Signal attenuation-correction, 320-321. Signal detection, basics, 660-663, 918-931. See also, Detectors. coefficient of variation, 660 instrument dark noise, 660 photon (shot) noise, 660-661 PMT linearity, 661-662 signal-to-noise ratio, 660 spectral accuracy, 662 spectral resolution, 662-663 wavelength response, 663 Signal levels, 16-photon peak signal, 73-74. Signal-to-background ratio, of titaniumsapphire laser, 112. Signal-to-noise (SIN) ratio, 37, 53-54, 67, 81,164,251,257,265,330,340, 386,391,442-451,470,481,495, 498-499, 528, 542, 562, 567, 582, 599, 621-622, 660, 690, 696, 699, 707, 736-737, 740, 753, 769, 772, 778-780,810,813. 3D imaging, 448-451 4Pi microscopy, 562-567 bleaching, 391, 442, 690, 696 in calcium imaging, 737 chapter, 442-451 comparative performance, 256, 448-451 bleaching-limited performance, 448-450 configurations of microscope, 448, 449 disk-scanning microscope, 449 line illumination microscope, 449 saturation-limited performance, 450 scanning speed effects, 53, 450-451 structured illumination, 265-266, 270, 275-276, 279-280 SIN ratios for, table, 450 widefield (WF) microscope for, 450 confocal microscope, 444-447, 660 calculations, 444 detectability, 446-447 methods compared, 450 noise model N 1, 444-445 noise model N2, 446-447 deconvolution, 470, 481, 495, 498-499 designs, confocal, 212-216, 447-448, 450 disk-scanners, 221 dynamic range, 2-photon, 644, 778-780 high-content screening, 810
978
Index
Signal-to-noise (SIN), ratio (cont.) improvements, 736 micro-CT, 615 magnetic resonance microscopy, 621-622 multi-photon fluorescence microscope, 112,427,447,542,779 Nyquist sampling, 67, 448 optimal excitation power, 81, 340 Rose criterion, visibility, 37-38, 68, 164, 633 saturation, 442 vs. scan rate, 53 signal level, 67, 75, 528 sources of noise background noise, 443-444 grey levels, 443 quantum efficiency, 442-443 shot noise, 442-443 sources of noise, 442-444 STED,574 and visibility, 37 Silica glass, transmission losses, 502. Silicon diodes, near infrared emission, 132. Silicon-intensified target (SIT) camera, 730. brain slices, 730 SIM. See Surface imaging microscopy. Simplicity, as design goal, 43, 66, 220, 229, 387,508,647. Simulated fluorescence process (SFP), 310. Single-cell automatic imaging, 809, 812. Single-cell calcium imaging, 812. Single-longitudinal-mode fiber laser, 1l0. Single-mirror/double tilt scanner, 215. Single-molecule, 80. biochemistry, 221-222, 575, 690, 693, 696 bleaching, 690, 693, 696, 697-698, 699 Single-photoelectron pulse heights, 30. Single-photon, energy, equation, 35. Single-photon counting avalanche photodiodes (SPAD), 527. Single-photon excitation, plant imaging, 772-778. Single-photon pulses. See Photon counting. Single-scan images measure scan stability, 40-41. Single-sided disk scanning, confocal microscopy, 132, 141-142, 168,171, 175,215-216,229,231,907,913. See also, Disk-scanning confocal microscopy. advantages and disadvantages, 215-216 basic description, 141 commercial, 907, 913 light source, 132, 141-142 Singlet state saturation, 21-22, 41, 81, 142, 265,276,338-339,442,448,450, 643,647, 899, 928. Sinusoidal bidirectional scanning, 25, 52-54. See also, Resonant scanners. duty cycle, 53, 260 Sinusoidal image, 831, 838. fiber-optic confocal, 510
Sinusoidal modulation, in FUM, 524-526. SIT. See Silicon-intensified target camera imaging. SLF. See Subcellular location features. Slice AM-dye-painting protocol, 726-727. Slice chamber protocol, 727. Slit scanning confocal, 12,25,37,50,51, 56,221-226,231+,235,238,519, 522, 664, 741, 914, 916. Achrogate, 50, 212, 231-232, 916 with AOD scanning, 56, 914 commercial, 913-914, 916 critical parameters, 224-228 optical sectioning, 228, 444-449 optimal slit size, 225 point excitation, slit detection, 914 SLM. See Spatial light modulator. SLT. See Subcellular location tree. Smart media, digital storage, 588. SMD. See Surface mount device. SNARF-I, 345, 346, 531, 739, 744-745. ratiometric pH label, 744-745 stained rat interossi muscles, 739 table of variants, 531 Snell's law of refraction, 167,654. SOc. See Saturable output coupler. Software packages, visualization, table, 282-283. SoftWorx, 3D display software, 282. Solanum tuberosum, potato, 712. Solid state memory devices, 588. compact flash cards, 588 memory stick, 588 smart media, 588 Solid-state photodetector, 30-31, 918-931. See also, CCD; EM-CCD. Solid-state lasers, 86, 103-118, 236-237. cooling, 108 noise sources, 86 thin -disk lasers, 109 tunability, 109 use, 236-237 Source brightness, measure, radiance units, 126. Source optics, reflecting and collecting light, 134. Space invariance, telecentric systems, 207-208. Space multiplexing, in MMM, 555. Spacer, material in interference filters, 46. SPAD, single-photon counting APD, 527. Spatial coherence, 84. Spatial filter, 89, 107,391,542, 708, 729. optical devices for, 89, 222-223, 729 digital, 391-392. See also, Gaussian filtering Spatial frequency, 37, 60, 65, 66. See also, CTF. and contrast transfer function, 37 and geometry, 66 response of microscope, and pixel size, 65 zero, as measure of brightness, 60
Spatial laser beam, characteristics, 89. Spatial light modulator (SLM), 266. Spatial orientation factor, for FRET, 792-793. Spatial resolution, in confocal microscopy, 24. See also, Resolution, PSF, CTF. Special setups, for CLSM, 218-219. Specifications, general, for scanner, 54. Specimen, general considerations, 192-197, 228, 361-362, 779. See also, Living cells, Living embryo imaging. fluorescent probes interactions, 361-362 cytotoxicity, 362 localization, 361-362 metabolism, 361-362 perturbation, 362 optical heterogeneity, 22, 23 plants. See Plant cell imaging; Botanical specimens Specimen chambers. See Living cell chambers. Specimen heating, in 2-photon, 539. Specimen holder, for scanning specimen, 9. Specimen preparation, for automatic 3D image analysis, 319-321. image analysis, 319-321 imaging artifacts, 320 Specimen preservation, general, 368-378. antibody screening on glutaraldehydefixed specimens, 377 evaluation, 371-374 cell height to measure shrinkage, 371-373 defined structures, distortion, 373-374 MDCK cell, stereo image, 373 MDCK cell, vertical sections, 372 fixation/staining, 370-371 fixative characteristics, 368-370 chemical fixatives, 369 cross-linking fixatives, 369 freeze substitution, 369 microwave fixation, 369 protein coagulation, 369 formaldehyde, 369-370, 373 general notes, 374-378 glutaraldehyde, 369, 370 immunofluorescence staining, 371 improper mounting, 376 labeling thick sections, 376-377 microwave fixation, 377-378 mounting methods, 370-374 critical evaluation, 371-374 mounting media, table, 377 pH shift/formaldehyde fixation, 370-371, 373 refractive index mismatch, 377 mounting media, table, 377 refractive index of tissue/organs, table, 377 tissue preparation, 376 triple labeling, 375-376 Specimen-scanning confocal microscope, 9.
Index
Speckle, from high-coherence sources, 8, 84,90,130-132,448. Speckle microscopy, 13,383,385, 889. Spectra, emission. arcs, 130 black body, 136 LEDs, 133 solar, 127 tungsten source, 153 Spectral accuracy, 662. Spectral bleedthrough (SBT), 185,203-204, 664. in intensity-based FRET, 185 Spectral confocal image A. thaliana seedling, 175. Spectral detector, 203-204, 662-663, 666-667. testing, 662 Spectral discrimination, filter bandwidths, 44. Spectral imaging, 175, 382, 384. table, 384 Spectral leakage, inter-channel signal imbalance, 185, 203-204. Spectral phase interferometry, for direct electric field reconstruction (SPIDER), 115. for pulse length measurement, 115, 901-903 Spectral properties, of filters V.I'. angle, 49. Spectral resolution, of detection system, 203-204, 662-663, 666-667. Spectral response. of CCD chips, 29, 234, 922 of eye, 153 PMT photocathodes, 29 Spectral transmission, objectives, plots, 159-161. Spectral unmixing, 190-192, 319, 361, 382, 384,386,423-425,431,664-667, 770,905. detectors for, 51, 667 examples, 665-666 limitations, 51, 382, 667 overlapping fluorophore emission, 190, 319,423-425,664-667 removing autofluorescence using, 667 Spectrofluorimetry, for FRET, 793. 795. Spectroscopic ruler, 765. Speed, in confocal imaging, 7. 11-12, 36, 41.53, 142,222-224,235-236,434, 447,450,458,460,482,526,536, 563-564, 748, 753-755, 784. See also, Temporal resolution. 4Pi-MMM, 563-564 AOD,55-56 calcium imaging, 741 CARS, 599-600, 604 charge-coupled device cameras, 77-78. 142,229,231-235,259.647.651, 754-755, 885 data compression, 581-582, 586-588 detector, in FUM, 523, 558
disk-scanning confocal, 141,216,224, 754 for display, processing, 803, 839, 841842, 862 factors affecting. 235-236, 482, 496, 753-754 of fixation, 370 FRET, 795, 805 galvanometer, 52-54, 211. 214 high-content screening, 809-810, 813 MMM, 551-555, 563-564 need for. in living cell imaging, 222, 753-754 rendering, 3D display, 831 SPIM, 613, 678 Spermatocyte. crane fly, 15. Spherical aberration, 15,34, 147-149, 151, 160. 192-197,208,241,244,247, 330,395,404-413,454-455,463, 466.471,480-481,542,629,640, 654-655, 657-658, 728, 772, 774, 893, 903-904. See also, Aberrations, spherical. blind deconvolution, 471, 480-481 chapter, 404-413 confocal microscopy performance, 654 correction of RI mismatch, 192, 287, 411, 542 correction of, figure, 145,411-412, 654-655 corrector, 92. 395, 398,411,477, 640, 654 deconvolution, 463, 466. 468-469. 471, 480, 498-499, 658, 728, 784 effect of specimen, 192-197. 418, 454, 747 index mismatch. See Index mismatch measurement, 145,407,455,471, 481, 492, 657 signal loss, 330, 389, 395, 413. 457, 661 SPIDER, Spectral phase interferometry for direct electric field reconstruction, 115,901-903. Spill-over, between detection channels. See Bleedthrough. SPIM. See Selective plane illumination microscopy. Spinning disk, 3, 5-6, 11,40,141,176,216, 223-224,231-232,235-236, 260-265.459-460,464,468, 481-483,783-784,810-811. See also, Diskscanning confocal microscopy. commercial, 907, 913, 915 FUM, 519-520, 522 high-content screening, 810-811, 820 MMM, 554, 558 performance, 449-450 systems for, cytomic imaging, 810 vs. TPE imaging, in plant cells, 783 Yokogawa CSU- 10/22, 231. 915 Spinning-disk light scrambler, ground glass, 8.
979
Spinning filter disk, digital projector, 590. Spirogyra, and depth of optical sections, 195. Spot scanning, to avoid coherence effects, 84. Spot size, full-width at half-maximum. See Pointspread function, Full-width half-maximum. Square pixels, advantage of using, 62. Stability, 86, 102, 103, 136-139, 826. algorithmic, 473 arc sources, 136-137,477 argon-ion laser vs. krypton laser, 102 disk scanners, 215 of DVDs, 587 dye. See Dyes; Bleaching from fiber-optic coupler, 505-506 galvanometer, 54 halogen sources, 136-139,346 interferometer, 240-241, 267 laser, 81, 85-89, 704 diode, 106, 108-109 fiber output, 505 helium-cadmium, (low), 103 intensity, 85-87,113,116,136,477, 903 measurement, 650-651 pointing, 87, 903 results, 86, 103 structure, 82-85, 103 thermal, I II wavelength, 106-108, 115, 118 mechanical, 39, 82, 85, 201, 267, 512, 652 objectives, 146 photostability, 363, 369, 690-702, 802. See also, Dyes; Bleaching scan, 40, 638-639, 651 shutter, CCD camera, 929 thermal, 111,219, 387, 389, 394, 539. See also, Thermal variables Stage-scanning confocal microscope, II. piezoelectric scanners, 57, 708 Staining, plants, 438, 774. See also, Dyes; Livingcells; Botanical specimens; Plant cell imaging; Fluorophors. calcofluor procedure, 438 of plant tissues, 774 Standards, ISO (DIN) microscope design, 156+. Standing-wavefield microscope, 275. Starch granules, plant, 202, 420-421, 428, 432-433,435, 703, 710-712, 715, 719. Static discharges, destroy semiconductors, 109. Statistical noise, in counting quantummechanical. See Poisson noise. STED. See Stimulated emission depletion. Stem-cells, 623, 678, 762, 790, 813. Stem, plant, 168, 172,180,417-419,421, 424,430,556,707,710-711, 713-714.
980
Index
Stentor coeruleus, backscattered light image, 168. Step index optical fibers, 501-503. Stepwise discriminant analysis (SDA), 818, 820. Stereo Investigator, software, 282. Stereology, 316, 319. Stereoscopic image, about, 6-7, 9, 11, 154, 224,298-299,317,396. biofilms, 880 cheek-cell specimen, 23 diatom, 640 Drysophila, microtubules, 752 embryo, 200 fat crystal, polarization, 479 neurons, 298, 314 Alexa stained, 330 backscattered light images, 167 eye, optic nerve, 481 Golghi-stained, 298 Lucifer-yellow, 314 microglia, 396-398 rat-brain neurons, 398 transmitted light, 475 lung, 292 MDCK cells, 373-374, 378 Milium chromosomes, Fuelgen-stained, 298 Paramecium, chromosomes, 298 pea root, RNA transcript, 465 platelet, high-voltage, EM, 848-849 sea urchin, S. Purpuratus, 173, 198, 200 skin, 298 Spirogyra, 195 tandem-scanning confocal microscope, 6 Stereoscopic views, image processing and display, 290, 292, 293, 295-299, 451,764. color space partitioning, 297 display, 293, 299 interlaced fields of frame, 297 movie projection, 838 pixel-shift/rotation stereo, 297 stereo images example, 298 synchronizing display, 297 Stick objective, for in vivo confocal, 806. Stimulated emission depletion (STED) microscopy, 3, 539, 561, 568, 571-578. axial resolution increase, 576 breaking the diffraction barrier, 571-573 challenges, 577 compared to confocal, 575-576 diagram, 573 different approaches, 573 dyes used successfully, table, 575 OTF compared to confocal, 578 outlook, 577 PSF compared to confocal, 578 RESOLFT, the general case, 572-573 results, 576, 578 triplet-state, 573
Stimulated emission of radiation, defined, 82-83, 124. Raman scattering, 167 semiconductor, 106 and stimulated-emission depletion, 573, 577 STN, supertwisted nematic, 589. Stochiometry, ion kinetics, 741. Stokes field intensity, 595, 597. Stokes laser, in CARS microscopy, 595, 597-604. Stokes shift, 44-45, 268, 338, 341, 343, 443-447, 539, 542, 690, 759, 792-793. anti-Stokes, CARS, 550, 595-604 defined, 44-45 in fluorescence resonance energy transfer, 792+ large, in 2-photon, 539, 646 of quantum dots, 694, 759 size of fluorophores, 45 Storage, digital. See Data storage. Storage structures, plant, 435-436. maize, image, 436 Stray light, 58, 632, 904. laser light, 632 non-descanned detection, 904 practical confocal microscopy, 632 room light, 201, 632 Streak camera, FUM detector, 520. Strehl ratio, measure of image sharpness, 247. S. purpuratus (Sea urchin), 173, 198, 200. embryo, 173, 198,200 first mitotic division, 173 image degradation, from top and bottom, 198 stereo-pairs of embryo, 200 Structural contrast, 188. See also, Harmonic signals. Structure, optical, 59, 68, 132-135. of light-emitting diodes (LED), 133 of microscope sources, 132-135 recognizing features in noisy images, 68 chapter, 265-279 Structured illumination microscopy, 265-279. advantages/disadvantages, 265 computing optical sections, 268-270 vs. confocal microscopy, 265 degree of spatial excitation modulation, 268-270 absolute magnitude computation, 268-269 homodyne detection scheme, 268-269 max/min intensity difference, 268 scaled subtraction, 269-270 square-law detection, 268-269 synthetic pinholes, 268, 269 experimental considerations, 265-268 illumination masks for, 266 light source for, 267
pattern generation, 266-268 schematic setup, 266 nonlinear, 276 resolution improvement, 270-276 Fourier-space, 270-271 linear image reconstruction, 271 Lucosz's formulation, 273 Moire effects, 270-271 photobleaching, 275 reconstruction steps/results, 272 standing-wavefield microscope, 275 test results, 274 thick samples, 274, 275, 278-279 Subcellular location features (SLF) in automatic image analysis, 819-820, 822-824, 828. 2D,819-820 2D SLF feature descriptions table, 819 3D SLF, 822-823 test results, table, 824 Subcellular location tree (SLT), 825. Subpixel deconvolution, 478-479. Subresolution beads, 655-656. See also, Beads. Sun, microscope light source, 126-127, 131, 135. spectrum, 127 Superficial optical sections, living embryo, 748. Supertwisted nematic (STN), 589. Surface imaging microscopy (SIM), 607-608. mouse embryo, 608 setup, 608 Surface mount device (SMD), for LED, 133. Surface orientation, affects reflected light, 181. Surface structures, distortion, 197. Surface topography, maximum intensity, 180. Surfaces, of interference filters, 47. Suspension-cultured cells, 189, 429-430. bacteria, 876, 878 image, 430 frozen, 854 Swept-field confocal microscope, 238. Synchrotron, wide-spectrum light source, 135+. Synthetic pinholes, in structuredillumination microscopy, 268, 269. images, 269 SYTO, 396, 874-876, 879-885. T Tagged image file format. See TIFF. Tandem-scanning confocal microscope (TSM),2-6, 11, 13-15,39-40, 141, 167,215-216,223-224,228-229, 447. comparison with other confocals, 13-15
Index description, 6, 141, 215-216, 228-229 development, 5-6 evaluation, 215, 216 observing ciliate protozoa, 141 rate of data acquisition, 11 real-time imaging of tooth, 167 sources of vibration, 39--40 viewing color/depth-coded, real-time, stereo images, 154, 304 Tapetum, plant, 433, 434, 779. TEC, Thenno-electrically cooled, see Peltier cooling. Telan systems, 129, 157. Telecentric plane, 208-209, 211. conjugate, 208-209 effect of angular deflection in, 211 Telecentricity, 207, 214. of closely-spaced scan mirrors, 214 defined, 207 Tellurium oxide (Te02), for use in AODs, 55 TEM. See Transmission electron microscope. TEM. See Transverse electromagnetic modes. Temperature, 29, 56, 133, 135-136,856, 885. See also, Thermal variables. Temperature tuning, of diode lasers, 108. Temperature effects on high NA objectives, 248+. Temporal aliasing, 39, 41, 391, 836-837, 839. Temporal coding, 299-300. Temporal coherence, 7-8, 82-85, 131. defined, 84 Temporal dispersion, 502. See also, Pulse broadening. Temporal displays, 292-293, 297, 836. Temporal experiments, biofilms, 885-886. Temporal pulse behavior, pulsed laser, Ill. See also, Pulse length measurement; Pulse broadening. Temporal resolution, 12, 24, 36-38, 41, 221-222,322,334,386,391,399, 458,558,577,618,620,622,651, 667, 730, 737, 746, 772, 784, 801, 809. See also, Fluorescence lifetime imaging (FUM). of photodetectors, 263 Temporal signals, 162, 286, 331, 383. 'Test drives," for living embryo imaging, 752. TFT. See Thin-film transistor. Tetracysteine, labels, 221, 348, 359, 853. Thalamocortical slice protocol, 724. Thermal lensing, pulsed lasers, 109, 113, 543. Thermal variables, 219, 856. active medium, lasers, 81 of AODs, 56-57 arcs, peak emission wavelengths, 129 automated confocal imaging, 810
cell chambers, 117,386-389,394,727, 790,810,814,885-886. See also, Cell chambers cooling, 108, 133 cryo preparation for EM, 856-857 on detectors, 29, 252, 256-257, 495 drift, 16, 115,219,386,567,489,652 compensating, 396, 732 on dye labeling, 359, 361, 738-739 effects of anti-bleaching agents, 694 effect on bleach rate, 696-689 effect fiber pinhole size, 506 fiber-optic, pol-preserving fiber, 503 filament spectra, 135-136 fixation, 369-372, 375, 377 incandescent lamp emission, 135-136 laser cavity, 34, 82, 85-88, 107, 109, 111, 541 of LED, 133, 136-138 brightness, 133 lensing, in pulsed lasers, 109, 113,543 and light-source output, 136, 138,650 noise signal, 254, 257, 232-234, 261-262,495,660,734,921,924, 925 on objective lenses, 248-249 in photography, 71 properties of ice, 856 properties of optical materials, 158, 248-249 and photomultiplier tube, (PMT), 29 on refractive index, 15,56, 145,411 immersion oil, 148-149,248-249, 411 retinal exposure, 117-118 sensors, 255-256, 727 solid-state laser, 86, 108 specimen damage, 84-85, 139,685 specimen heating, 539, 545, 681, 685, 904 temperature tuning, laser, 108, 115 thermo mechanical effects, 685 time constant, 38 Thermo-electrically-cooled, see Peltiercooled. diode lasers, 85,107-108, Ill, 117 THG. See Third hannonic generation. Thick samples, 274, 275, 278-279. See also, Living embryo imaging; Brain slices; Biofilms. background, 278 structured illumination, 274, 275, 278-279 close focus region, 279 distant focus region, 279 in focus region, 278 number of collected photons, 279 Thin disk lasers, 109-110. Thin Laser Light Sheet Microscope (TLLSM), 672. See also, SPIM. Thin-film transistor (TFT), 589. Third harmonic generation (THG), 90, 166-167,179-180,188,428,435, 550, 705-718.
981
CARS, 596-597 contrast mechanism, 166-167 deposits no energy, 361 detectors for, 421, 706-708 table,707 double-pass detection method, 166-167 intracellular inhomogeneities tracked, 90 light attenuation spectra, 706 light sources, 706-708 to make more laser lines, 109, 114 mechanism, 705 micro spectroscopy, 421 MMM, 551, 559 non-linear optical microscopy, 705 optical sectioning, 704 optically active animal structures, 714-717 collagen mat, polarization microscopy, 717 mouse zygote spindle, 717 structures producing THG, table, 715 zebrafish embryo, 716, 718 optically active plant structures, 710-714 cell walls, 438 Commelina communis, 712 Euphorbia pulcherrima spectrum, 710 maize, emission spectrum, 710, 711, 713 maize, polarization microscopy, 711 maize, stem section, 714 phytoliths, polarization microscopy, 720 potato, 712 Pyrus sera tina spectrum, 711 rice leaf, image, 712, 715, 719 photon interactions, 179 pulsed lasers suitable, table, 706 STED,577 structural contrast, 188 Three-decibel point (3dB), for bandwidth, 59,65. Three dimensional cell pellet, 815. Three dimensional microscopy, 766, 771, 804+. future perspectives, 804-805 living embryos, 766 of plant cells, 771 Three dimensional projections, embryo, 763. Three dimensional segmentation, plant, 776-778. Three-channel confocal microscopy. with 4 recombinant proteins, 190 assays for, 814 Three-dimensional diffraction image, 4, 147, 407,455,463,471,491. Three-dimensional micro-array assays, 815-816. Three-dimensional reconstruction, 775-776, 778, and Chapters 14 and 15. plant imaging, 775-776 A. thaliana, 778 Equisetum, 774
982
Index
Three-photon excitation (3PE), 88,415,447, 535,550-552,555,558,647,680, 709,876. absorption cross-section, 680 damage, 682, 686 fiber-optics, 507 resolution, 447 setup, 708-709 TIFF (Tagged image file format), 580. Tiled montage, 851, 858. Tiger, ECDL laser system, 90. Time correlated single-photon counting (TCSPC), 518, 520-523, 526. for lifetime imaging, table, 526 FUM, 520-523 FRET-FUM, 186 schematic diagram, 521 Time multiplexing, of adjacent excitation spots, to reduce flare in MMM, 553-554. Time-gated detection, FUM, 522-524, 526, 528+. diagram, 522 FUM methods compared, table, 526 FUM, image, 528-529 Time-lapse imaging, 136,222,354, 382-384, 392-399, 652, 773, 885-886. Amoeba pseudopod, 191 confocal of plant cells, 773 high-content screening, 812 illumination stability, 136 image analysis, 286, 320, 333, 732-733 mechanical stability, 219 micro spectrometry, maize damage, 424-426 rectified-DIC, of platelets, 846 SPIM,613 table, 384 three-dimensional plus time, 222 two-dimensional plus time, 222 Time-lapse recordings. Amoeba pseudopod, 191 Ascaris sperm, 846 biofilms, 885 brain slices, 725, 727, 729, 732-733 embryos, 676, 749, 752, 759, 761 meristem growth, 430 plant roots, 781, 784 rectified-DIC, of platelets, 846 two-photon microscopy, 10 TIRF. See Total internal reflection fluorescence. TIRM,177-179,477. Tissue specimens, introducing the probe, 360. Titanium:sapphire laser (Ti : Sa), 82, 84-86, 88-91,94, 100-103, 105, 107, 109, 111-112,114,123-124,165,346, 358,415,423-424,459,541,550, 551,645-647,688, 706-708, 713, 727, 750, 756, 759. See also, Lasers, titanium; sapphire and Ultrafast lasers.
4Pi, 563-564, 567 brain slices, 731 CARS, 599 compare to other fast lasers, 112-113 Cr; Fosterite, femtosecond pulsed laser, 109, 114,415,541, 706-709, 712-714 embryos, 750, 756, 759, 731, 764 emission stability, 86 four-level vibronic model, 82, 109 maintenance, 116 multi-photon excitation, 541 and OPOs, 114-115 plants, 415, 423-424, 706-708, 713-714, 717,781-783 popular models, specs, table, 120 STED,575 ultrafast, 112-113 URLs,124 TLB. See Transmitted light bright-field. TLLSM. See Thin Laser Light Sheet Microscope. Tobacco, 116, 189-190,430,693. smoke, not around lasers!, 116 suspension-cells, birefringence, 189-190 GFP expressing cells, 430 photo-bleaching, 693 "Toe" photographic response, defined, 71. Tornado mode, SIM scanner, 54. Total fluorescence signal, 742. Total internal reflection fluorescence microscopy (TIRF), 90, 160, 180-184,223,477,801. blind deconvolution, 477 vs. confocal image, 184 contrast, 180-184 cytoskeleton, image, 183 FRET, 801 limits excitation to single plane, 223 objectives, for epi-TIRF, 161 Total internal reflection microscopy (TIRM), 177-179,477. blind deconvolution, 477 evanescent wave generation, 178 TPE. See Two-photon excitation. TPEM. See Two-photon excitation microscopy. Trade-offs, 36, 68, 78-79, 221, 224, 644-648, 747-748, 825. beam power, visibility/damage, 693 blind deconvolution, 483, 488, 499 compression algorithms, 581, 840 confocal endoscopes, 508 when digitizing, 68, 78-79 embryo specimens, 747-748
high-content screening, optimal clustering, 825 living cells, 381, 693 micro-CT, dose/resolution, 616 MRM, time/resolution, 622 and pinhole size, 265, 267 processing speed/segmentation, 301
speed, SIN, sensitivity and damage, 221, 224, 232, 556, 644-648 SPIM, resolution and number of views, 613 Transcriptional reporters, embryo analysis and, 748, 755-756. FluoroNanoGold,854 mRNA, 316-317,465 plants, 773, 781 NF-kB,814 Transfection buffer, electroporation, table, 802. Transfection, cellular, 756-758, 790, 791. brain slices, 722, 724-725, 730-731 Transfection reagents, for chromophores, 358,360,362,556,682,790-791, 795,803. 2-0ST-EGFP, 566 COS7,693 EB3-GFP, 183 for FRET, CFP/YFP, 795-796, 798, 801-802 GaIT-EGFP, 566 GFP-MusculoTRIM, 184 ligand binding, 348 Transfer function, implications for image contrast, 164-165. See also, CTF. Transient permeabilization, 359, 373, 375. Trans-illumination, absorption contrast, 166. Transistor-transistor logic (TTL), 259. Transit time spreads (TTS), 527. Translational fusions, 756, 757. See also, Transfection agents. subcellular specific protein distribution, 756 Transmission, 33, 49, 159,225,231,804. AOBS,57 contrast, 163-164 disk-scanning micro-lens array, 223-226, 227-229, 231, 235 dispersion, 683 of filters. See Filters linear vs. log plots, 44-49 of glass fibers, 501-505 illuminator, 201,127-128 losses due to refractive optics, 33, 217 table, 217 of objectives, 154, 158, 159-161,641 relative, measurement, 26, 34, 36 table, transmission, 158, 159-161 of plant tissue, spectra, 416, 422 of Polaroid materials, 85 SHG signal detection, 707-709, 729-730 by small pinholes or slits, 225 Transmission electron microscope (TEM), 846. correlated LM-TEM images, 852-855, 857-859 stereo images of platelets, 848-849 Transmission illuminator, ghost images, 201-202. Transmission intensity, specimen thickness, 164.
Index
Transmittance, optical system, measured, 25-26. table, 217 Transmitted light brightfield, 468, 472-473, 477. blind deconvolution, 472-473, 477 Transparency, lighting models, 309-312. Transverse electromagnetic modes (TEM) laser, 83. Trends, in laser design, 118. Triple-dichroic, 33, 46, 48, 217-218, 678, 783. light loss due to, 33 performance, 46-48 Triplet state, 103, 338, 339-342, 348, 362-363,390,516-518,573,646, 684,691-693,697,698,704,852. saturation, 339, 573 as a RESOLFT mechanism, 573 Triton X-100, 730, 852. formaldehyde fixation, 370-372, 375-377 True color, 291. TSM. See Tandem-scanning confocal microscope. TTL. See Transistor-transistor logic. Tube length/chromatic corrections, table, 157. Tunable lasers, 91, 103, 107, 109, 120. broadband, table, 120 continuous wave dye, table, 91 diode, emerging techniques, 107 solid-state, 106, 109 solid-state ultrafast, 103 Tungsten carbide electrodes, radiance, 137-138. Tungsten halogen source, 132, 137, 153. Turnkey ultrafast laser systems, 118. Tutorials, lasers by level, 124. Tweezers, optical, 89-90,110,218,383, 385. setups for integration, 218 single-longitudinal-mode fiber laser for, 110 trapping wavelength, 89-90 Two-channel confocal images, 175-1 77 , 177, 193, 425, 522. A.thaliana, epidermaUmesophyll cells, 193,425,431-432,434-436 Amoeba pseudopod, 169 colocalization, 667 display, 311, 841 FUM,522 harmonic images, 714-716 mouse muscles, 716 montaging, 331 neurons, 332 microglia, 396-398 eye, optic nerve, 481 Golghi-stained, 298 Lucifer-yellow, 314 rat-brain neurons, 398 transmitted light, 475
of peony petal, cytoplasmic, 175-176 rat intervertebral disk, 310-311 of zebrafish embryo, 177 Two-dimensional imaging, 60, 222, 397-398. time lapse, 222, 397-398 Two-photon fluorescence excitation (2PE), 156, 160,218,535,536,750, 778-783. chapter, 535-549 chromatic correction for, 156 for plant cells advantages of, 778-779 cell viability, 779-781 vs. confocal microscopy. 779 dyes, 782 of green fluorescent protein, 782-783 pitfalls, 782 of thick specimens, 779 in vivo, 781 special objectives for, 160 visible and ultraviolet dyes, 218 Two-photon microscopy, 10-12, 195,357, 535-549,690, 697, 900-905. See also, Multi-photon excitation; Multiphoton microscopy autofluorescence, 545 basic principles, 535 of biofilms, 882-885 bleach planes, in fluorescent plastic, 193, 194 caged compounds, 544 calcium imaging, 545 chromophores, 543 2-photon absorption, 543 diagram, 540 detection, 538, 541 descanned, 542 non-descanned (whole area) detector, 541 stray light, 904 fluorescence, shadowing, 195 group delay dispersion, 5443 laser. 540-541 alignment, 900-904 monitoring, 901-903 mounting, 541 power level, 903-904 safety, 117-118,839,900, 903-904 living cell studies, review, 544-545 living animal studies, 545 minimize exposure during orientation, 905 mirror scanning, 543 optical aberrations, 542 photobleaching, 690, 697 practical tips, 900-905 beam alignment, 901 bleed-through, 904 choice of pulse length, 537, 903 pulse length, 109, 112, 115, 507, 537, 538, 902-903
983
specific specimens, see specimens by name imaging multiple labels, 904-905 neurolucida protocol, 731 resolution, 539 and speed, 12 vs. spinning disk imaging. in plant cells, 783 stray light and non-descanned detection, 904 theory, 535, 537 wavelengths, 538-541, See also, Botanical specimens U UBC 3D living-cell, microscopy course, 174,183,184,190,205,364,430, 435, 439, 805-806. Ulbricht sphere, for measuring light, 140. Ultrafast imaging, two dimensional, 222. 3D, 235 Ultrafast lasers, 88,101,103,112-114. Cr:Fosterite. 109, 114,415,541,706, 707-709, 712-713 diode-pumped solid-state (DPSS), 112 distributed feedback (DFB) diode laser, 113 fiber, 113-114 table, 101 fiber-diode, mode-locked, 113 Nd:YAG, 88-89, 91, 95, 97,103, 107-109,111,113-115,117,218, 245, 514, 680, 798 Nd:YLF, 89, 98,100,103,109,112-114, 750, 760-761 Nd: YV04 , 89, 95, 100, 103, 107-109, 111,113-114,541 solid-state, tunable, 103 spectrum, 44 titanium: sapphire, 112-113. See also, Laser, titanium: sapphire; Titaniumsapphire laser Ultrafast pulses, delivery by fiber optics, 88, 507. dispersion losses, 502 Ultraviolet (UV), argon-ion laser lines, 85,
87,90,102,339,346. other UV lasers, 111-11 7 use for micro-surgery, 218-219 Ultraviolet (UV) confocal microscopy, 109, 174,195,571. absorption, 707, 713 autofluorescence, 431-432, 434, 544 CCD response, 29, 255, 459, 921-922 correct imaging with planapochromats, 14,154 damage, 212, 290, 439, 544, 680, 686, 903 disk-scanners, 229 DNA-dyes, 782, 874. See also, DAPI; Dyes GFP excitation, 798, 873 high-content screening, 811 ion-imaging, 346, 383, 529, 738, 742
984
Index
Ultraviolet (UV) confocal microscopy (cant.)
multi-photon excitation, 535, 538, 544, 559,646,706,905 photoactivation, 759 safety, 117-118,839, 900,903-904 simultaneous with DIC imaging, 846, 850 as source of stray signal in PMT envelopes, 257 Ultraviolet performance of objective lenses, 154, 159-161, 706. Ultraviolet widefield light sources, 132, 136, 139, 143,226,542. table, 226 Ultraviolet transmission of optical fibers, 88. Ultraviolet (UV) light, effects produced by multiphoton intrapulse interference, 88. Ultraviolet scanning light microscope, 6-7. Uncaging, multi-photon microscopy, 383, 385,545,693,760-764. See also, Photoactivation. Unconjugated bodipy/ceramide dyes, 760. Under-sampling, 79, 635, 640, 652, 662, 831, 833, 836, 839, 84!. example, 640 uses, 68 Uniformity, of light source, 127-129. Unit image body, 3D Airy figure, 147. Upright vs. inverted microscope, 140, 157, 217,230,413,722,727,870-872. Unmixing. See Spectral unmixing; structured illumination. Up-conversion, fiber lasers, 110. doped ZBLAN, 1I0 dual-ion doped, 1I0 Uv. See Ultraviolet.
Video, 2, 4, 5-7, 11-14, 17,37,52-53, 61-62, 88, 219, 237, 261, 263, 346, 372,430,451,505,539,554,556, 589-590,593,604,860,885. confocal, 25, 237, 914 impact on light microscopy, 5-7, 14 results, 14 signal, 258-259 Video-enhanced contrast microscopy, imaging small features, 14,68. Vignetting, 210-211, 229, 245-247, 492, 54!. objective, off-axis performance, 245-247 Visibility, and signal-to-noise ratio, 37-38, 68. See also, Rose Criterion. Visilog/Kheops, software, 282, 301-302, 312. Visitech, confocal manufacturer, descriptions, 88, 119-120, 226, 237, 908. VT eye, 119-120,908,914 VT Infinity, 119-120, 908, 914 Visual cortex, identification of primary, 724. Visual observation, magnification for, 146. non-linearity, 72-73 Visualization, 280, 282-283. See also, Multidimensional microscopy images; Rendering. definition, 280, 292 software packages for, table, 282-283 Vitrea2IVoxel View, software, 282, 335. Volocity (software), 281, 236, 282, 295, 299,312,757,762-764. VolumeJ, software, 282, 304, 764. VolVis, 281-282. VoxBlast, 283, 301-302, 309, 312. Voxel, defined, 20. Voxel rendering, speed, 290. Voxx, software, 283, 377, 764.
V
W
Vacuum avalanche photodiode (VAPD), 31, 254,255. definition, 254 schematic, 31, 255 VAPD. See Vacuum avalanche photodiode. Vertical-cavity semiconductor diode laser (VCSEL), 108. Vibration. compensation, 732 from cooling water, 84, 102, 499 of disk scanner, 753 causing distortion, 16, 39-41, 166,201 of galvanometer mirrors, 40, 201 high-frequency, of acousto-optic devices, 55, 84 isolation, 85, 201, 219, 541 measurement, 30-41, 652 of mechanical shutters, 929 of objective lens motion, 754 optical fiber isolation, 505, 507 of optical fiber scrambler, 8, 84, 131 Vibronic laser, Ti: Sa four-level, 109.
WAD. See Whole-area; Non-descanned detection. Water, as immersion medium, 409, 410. refractive index mismatch, table, 409, 410 two-edge response curves, 410 Water-coverslip interface, spherical aberration generated at, 147. Water-immersion objectives, 15, 23, 36, 141,148-149,154,190,235, 241-242, 247, 261, 377, 386-387, 389, 395, 411-412, 513, 542, 552, 556,562,567-568,584,654-656, 708, 727-728, 737, 747, 772. See a/so, Spherical aberration. 4Pi, 562, 567-568 advantages, 149 biofilms, 870, 872 brain slices, 727-728,730, 737 chapter, 404-413 correction-colorlflatness/transmission, 154 deep imaging, 395
dipping objectives, 161,209,411,429, 568,613,727,737,870,872 in fluorescence ion measurement, 737 ion measurement, 737 living cells, 386-387, 389, 395, 398 performance measured, 47, 655-656 plant cells, 429, 433, 772 STED,576 transmission curves, 159-161 use and limitations, 15 Watershed algorithm, 322-325, 777, 822. for segmentation, plant cell images, 777 Wave optics, 4, 10. for calculating axial resolution, 4, 146, 154 Wavefront error, 217. lower, with hard coatings on filters, 45 Wavelength, 24, 28, 43-51, 62, 88, 107, 114-115,118,129-130,135-139, 165-166. calculation of Forster radius, FRET, 793 and CCD coupling tube magnification, 62 filters for selecting, 43, 44, 88 in multi-photon lasers, 165-166.415, 750 multiple, dynamic embryo analysis, 756 of non-laser light sources, 129-130, 135-136 and optimal zoom setting, 24 vs. pinhole size, 28 selecting, with interference filter, 88, 165-166 stability, in non-laser light sources, 137-139 tunability, of lasers, 107, 109. Wavelength expansion, non-linear, 114-115. Wavelength ratioing, 346. See also, FRET; FUM. Wavelength response, chromatic aberration, 663. Wavelength-selective filters, 43-51, 88. Wavelength-tunable lasers, summary, 107, 113, 116, 118,550. Wavelet compression, 581-584. Wavelet de-noising protocol, 733-734, 819-820. Waxes, plant, 420, 428, 434-435, 714-715. Website references, 123. 2 photon excitation spectra, 546, 727, 729, 782 brain slices, 727 CCDs, 76, 234, 927, 931 components, 58 confocal Listserve, 390, 901 deconvolution, 495 dyes, 221, 343-344, 782 fluorescent beads, 653 FRET technique, 185, 803 high-content screening systems, 811 image management, 865 lasers, 104, 115, 120, 123-125 live-cell chambers, 388-389, 870 movies related to book, 235, 392 muscles, 237
Index
non-laser light sources, 138, 143 plants, 769 safety, 117-118, 839, 900, 903-904 software, 282, 376, 594, 734, 762, 764, 776,777,820,824,827,831-833, 844, 845, 864-862, 865-867, 869 SPIM,672 Wedge, compensator, 566-567. Wedge, rotating, for light scrambling, 84, 131. Wedge error, in interference filters, 45-46, 151,211-212,630. in traditional filters, 45 Wedged fiber-optics, reduce reflections, 85. Well-by-well data, 817. WF. See Widefield. WFF. See Widefield fluorescence microscopy. White light continuum lasers, 88, 109, 113 continuum, 88, 109 He:Cd,113. Whole-area and external detection, 541-542. See also, Non-descanned detectors. Whole-cell patch pipet delivery, 360, 726-727. Widefield deconvolution, 751-753, 785. See also, Deconvolution. botanical specimens, 785 for living imaging, 751-753 Widefield (WF) fluorescence microscopy, 3, 22-23,26,172-173,219,453-467, 518. See also, Epifluorescence microscopy, Deconvolution. compared to confocal, 453-467, 644-647 CCD/confocal comparison, 458-459, 465 same specimen, 465, 482 compared to structured illumination, 274 deconvolution, imaging living cells, 23, 392 deconvolving confocal data, 461-464, 466 fluorescence detection, 459-460 fluorescence excitation, 459 fluorescence lifetime imaging, 518 gain-register CCDs, 460-461 images utilizing out-of-focus light, 26 imaging as convolution, 453-457 imaging thin specimens, 172-173 integration of fluorescence intensity, 459 interaction of photons with specimen, 22-23 light-emitting diode sources, 136 limits, Iinearity/shift-invariance, 457, 490, 564 model specimens, 461 noise, 459-463 optical sectioning schematic, 469 optical tweezers/cutters, 219, 89, 383, 385 out -of-focus light, 461 point-spread function, 453-457, 459-463 resolution, 3
sensitivity, 459-463 single point images, 454 pros/cons, 644-648 table, 459 temporal resolution, 458 Wiener filtering, 494, 496. See also, Gaussian filtering. image enhancement, 496 image restoration by, image, 494 Windows software, for automated confocal, 810. WinZip, 580. Wollaston prisms, DIC, 156,468,473,475. See also, Nomarski; DIC contrast. Working distance (WD) of objective lenses, 5,9, 129, 145, 154,157, 198,249, 511, 568, 598, 634, 673, 678, 727-728,747,774,779,781,872. table, 157-158 WORM disks (write once, read many), 586. X Xenon arc lamps, 44,132,137-138, 144. iso-intensity plot of discharge, 132 pulsed-operation, 137-138 shapes of electrodes, 132 spectral distribution, 144 super-pressure, spectrum, 44, 136 explosion hazard, 136 wavelengths available for detection, 44 Xenon/iodine fill arc, radiance, 137-138. Xenopus laevis, 13,610,746,748-753. blastomere, 757 confocal/multi-photon comparison, 750 embryo viewed with confocal, 748-753 viewed with OCT, 610, 749 embryo viewed with MRM, 623-264 in situ imaging, 746, 748 oocyte wound closure, 749 X- Y resolution, confocal/widefield compared, 36. y
Yellow fluorescent protein (YFP), 221-222, 429. FRET pair with CFP, 791-803 YFP, 221-222, 429 Yokogawa disk-scanning confocal system, 6, 12-13, 16,216,224-226,231, 234-237,458,754. CSU-IO/22 model, 223, 231, 236, 915 with EM-CCD. 234, 237, 755 high speed acquisition, II, 220, 222-226, 229,231,458,667,754,784 results, 236-237, 755, 783 vibration, 16 Ytterbium tungstate (Yb: KGW) laser, 108. Z ZBLAN up-conversion glass fiber, 110. Z-buffering, 304-305. Z-contrast, in confocal microscopy, 180.
985
Zea mays. See Maize. Zebrafish, 174, 176, 761. GFP image, 176, 176 autofluorescence, 174 pancreas expressing DsRed, 176 scatter labeling/lineage tracers, 761 Zeiss, confocal manufacturer, 212, 214, 217,226,231-232,655,771, 916-917. 510 META confocal microscope, 655, 908,916 Achrogate beam-splitterlLSM 5-Live, 50, 119-120,212,231-232,916 Axioimager system, 217 fluorescence correlation spectrometer (FCS), 383, 385, 602, 801, 803, 805, 917 HBO-100 source, self-aligning, 134-135 high-content screening, 811 LSM 5-Live line-scanning confocal microscope, 50,51,231-232,237, 784,908,916 META confocal spectral detector, 51, 119-120, 161, 202, 660, 663, 796, 916. mini-PMT arrays, 5 I, 667FRET, 706 tests, 663 objectives, advantages of, 155-156 Infinity Color-corrected System, 155, 217 plan objectives, table, 152 transmission specifications, 161 tube length conventions, 157, 239 working distance of objectives, table, 158 Zernike moments, 247-249, 818-820. Zernike polynomial fit, 245-247. table, 247 wavefront aberration function, 247 Zinc selenide (ZnSe) diode lasers, 106. Zirconium arc lamps, 136, 141. spectrum, 136 Zone System (Ansel Adams), 71-72. Zoom magnification, II, 24, 37, 63-64, 66, 70. See also Magnification optimal, 24 optical vs. electronic bandwidths, 70 relationship to area scanned, 63 Z-position and pinhole/slit size, 227. Z-resolution, 3-4, 22, 36, 149-150,224, 225-228, 563, 752. See also, Axial resolution. 4Pi microscopy, 563 in confocal fluorescence microscopy, 36 effect, of fluorescence saturation, 22 improvement, 752 of pinhole disks, 224 in STED, 576 Z-scanners, evaluating, 2 I 5. Z-sectioning, imaging brain slices, 729. Z-stack, 23, 754. of images of cheek-cell specimen, 23 speed acquisition constraint, 754