Mechanosensitive Ion Channels, Part A, Volume 58 (Current Topics in Membranes)

Current Topics in Membranes, Volume 58 Series Editors Dale J. Benos Department of Physiology and Biophysics University ...

Author: Owen P. Hamill | Sidney A. Simon | Dale J. Benos

118 downloads 641 Views 8MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Current Topics in Membranes, Volume 58 Series Editors Dale J. Benos Department of Physiology and Biophysics University of Alabama Birmingham, Alabama

Sidney A. Simon Department of Neurobiology Duke University Medical Centre Durham, North Carolina

Contents Contributors xi Foreword xv Previous Volumes in Series

CHAPTER 1

xvii

Structures of the Prokaryotic Mechanosensitive Channels MscL and MscS Stefan Steinbacher, Randal Bass, Pavel Strop, and Douglas C. Rees

I. Overview 1 II. Introduction 2 III. Conductances of MscL and MscS: General Considerations 3 IV. Structure Determination of MscL and MscS 6 V. MscL and MscS Structures 11 VI. The Permeation Pathway in MscL and MscS 15 VII. Disulfide Bond Formation in MscL 17 VIII. Concluding Remarks 18 References 20

CHAPTER 2

3.5 Billion Years of Mechanosensory Transduction: Structure and Function of Mechanosensitive Channels in Prokaryotes Boris Martinac

I. Overview 26 II. Introduction 26 III. Discovery, Mechanism, and Structure of MS Channels in Prokaryotes 28 IV. Pharmacology of Prokaryotic MS Channels 44 V. Families of Prokaryotic MS Channels 45 VI. Early Origins of Mechanosensory Transduction 46 VII. Concluding Remarks 50 References 50

v

Contents

vi

CHAPTER 3

Activation of Mechanosensitive Ion Channels by Forces Transmitted Through Integrins and the Cytoskeleton Benjamin D. Matthews, Charles K. Thodeti, and Donald E. Ingber

I. II. III. IV.

Overview 59 Introduction 60 Conventional Views of MS Channel Gating 63 Tensegrity-Based Cellular Mechanotransduction 66 V. Force Transmission Through Integrins in Living Cells 70 VI. Potential Linkages Between Integrins and MS Ion Channels 73 VII. Conclusions and Future Implications 77 References 78

CHAPTER 4

Thermodynamics of Mechanosensitivity V. S. Markin and F. Sachs

I. II. III. IV. V.

Overview 87 Introduction 88 Area Sensitivity 91 Shape Sensitivity 99 Length Sensitivity and Switch Between Stretch-Activation and Stretch-Inactivation Modes 103 VI. Thermodynamic Approach and Detailed Mechanical Models of MS Channels 111 VII. Conclusions 114 References 115

CHAPTER 5

Flexoelectricity and Mechanotransduction Alexander G. Petrov

I. Overview 121 II. Introduction 121 III. Flexoelectricity, Membrane Curvature, and Polarization 122 IV. Experimental Results on Flexoelectricity in Biomembranes 131 V. Flexoelectricity and Mechanotransduction 143 VI. Conclusions 147 References 148

Contents

CHAPTER 6

vii

Lipid Effects on Mechanosensitive Channels Andrew M. Powl and Anthony G. Lee

I. Overview 151 II. Intrinsic Membrane Proteins 152 III. EVects of Lipid Structure on Membrane Protein Function 152 IV. How to Explain EVects of Lipid Structure on Membrane Protein Function 155 V. What Do These General Principles Tell Us About MscL? 171 References 174

CHAPTER 7

Functional Interactions of the Extracellular Matrix with Mechanosensitive Channels Anita Sengupta and Christopher A. McCulloch

I. Overview 179 II. Mechanotransduction 180 III. Mechanosensitive Channels in Connective Tissue Cells 182 IV. The Extracellular Environment of Cells 184 V. Force Transmission from Matrix to Cytoskeleton 187 VI. Experimental Models of Force Application to Connective Tissue Cells 189 VII. EVects of Force on Cell Surface Structures 193 VIII. Future Approaches 194 References 195

CHAPTER 8

MscL: The Bacterial Mechanosensitive Channel of Large Conductance Paul Blount, Irene Iscla, Paul C. Moe, and Yuezhou Li

I. Overview 202 II. Introduction and Historical Perspective 202 III. A Detailed Structural Model: An X-Ray Crystallographic Structure from an E. coli MscL Orthologue 207 IV. Proposed Models for How the MscL Channel Opens 212 V. Physical Cues for MscL Channel Gating: Protein–Lipid Interactions 223

Contents

viii

VI. MscL as a Possible Nanosensor VII. Conclusions 228 References 229

CHAPTER 9

227

The Bacterial Mechanosensitive Channel MscS: Emerging Principles of Gating and Modulation Sergei Sukharev, Bradley Akitake, and Andriy Anishkin

I. II. III. IV. V. VI.

Overview 236 Introduction 236 MscS and Its Relatives 238 Structural and Computational Studies 242 Functional Properties of MscS 249 What Do the Closed, Open, and Inactivated States of MscS Look Like? 256 VII. Emerging Principles of MscS Gating and Regulation and the New Directions 260 References 263

CHAPTER 10 Structure–Function Relations of MscS Ian R. Booth, Michelle D. Edwards, Samantha Miller, Chan Li, Susan Black, Wendy Bartlett, and Ulrike Schumann


Overview 269 Introduction 270 The Structure of MscS 276 MscS Mutational Analysis 282 Structural Transitions in MscS 284 Conclusions and Future Perspective 291 References 292

CHAPTER 11 The MscS Cytoplasmic Domain and Its Conformational Changes on the Channel Gating Piotr Koprowski, Wojciech Grajkowski, and Andrzej Kubalski

I. Overview 295 II. MscL and MscS: Primary Gates and Similarities in Activation 296 III. The MscL Cytoplasmic Regions and Functioning of the Channel 299 IV. The MscS C-Terminal Chamber: The Cage-Like Structure and Kinetics 300

Contents

ix

V. Structural Alterations of the MscS Cytoplasmic Chamber on Gating 303 VI. Conclusions and Perspectives 305 References 306

CHAPTER 12 Microbial TRP Channels and Their Mechanosensitivity Yoshiro Saimi, Xinliang Zhou, Stephen H. Loukin, W. John Haynes, and Ching Kung

I. Overview 311 II. A History TRP-Channel Research 312 III. The Mechanosensitivity of Animal TRP Channels 313 IV. Distribution and the Unknown Origin of TRPs 314 V. TRPY1: The TRP Channel of Budding Yeast 317 VI. Other Fungal TRP Homologues 321 VII. Sequence Information Does Not Explain TRP Mechanosensitivity 322 VIII. Conclusions 323 References 324

CHAPTER 13 MscS-Like Proteins in Plants Elizabeth S. Haswell


Overview 329 Mechanosensation and Ion Channels in Plants The Eukaryotic Family of MscS-Like Proteins The Arabidopsis MSL Genes 345 Outstanding Questions 351 Conclusions 353 References 353

CHAPTER 14 Delivering Force and Amplifying Signals in Plant Mechanosensing Barbara G. Pickard

I. II. III. IV.

Overview 362 Introduction 362 Focusing Force 365 Transduction and Ensuing Events in Thigmotropism 378

330 337

Contents

x

V. Early Events in Gravitropism 379 VI. From Primary Transduction Pulse Forward: Facilitative and Vectorial Gravitropic Reception 385 VII. What Comes Next 389 References 390

CHAPTER 15 MS Channels in Tip-Growing Systems Mark A. Messerli and Kenneth R. Robinson

I. II. III. IV. V. VI. VII.

Overview 393 Introduction 394 Lilium longiflorum Pollen Tubes 395 Saprolegnia ferax Hyphae 400 Silvetia compressa Rhizoids 402 Neurospora crassa Hyphae 405 Is Turgor Necessary for Activation of MS Channels? 406 VIII. Conclusions 407 References 409

Index 413

Contributors Numbers in parentheses indicate the pages on which the authors’ contributions begin.

Bradley Akitake (235), Department of Biology, University of Maryland, College Park, Maryland 20742 Andriy Anishkin (235), Department of Biology, University of Maryland, College Park, Maryland 20742 Wendy Bartlett (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Randal Bass1 (1), Division of Chemistry and Chemical Engineering 114‐96, Howard Hughes Medical Institute, California Institute of Technology, Pasadena, California 91125 Susan Black (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Paul Blount (201), Department of Physiology, University of TexasSouthwestern Medical Center, Texas 75390 Ian R. Booth (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Michelle D. Edwards (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Wojciech Grajkowski (295), Department of Cell Biology, Nencki Institute of Experimental Biology, 02‐093 Warsaw, Poland Elizabeth S. Haswell (329), Division of Biology, 156–29, California Institute of Technology, Pasadena, California 91125 1 Present address: Department of Analytical Sciences, Amgen Inc., 1201 Amgen Court West, Seattle, Washington 98119.

xi

xii

Contributors

W. John Haynes (311), Laboratory of Molecular Biology, Department of Genetics, University of Wisconsin, Madison, Wisconsin 53706 Donald E. Ingber (59), Vascular Biology Program, Departments of Pathology and Surgery, Children’s Hospital and Harvard Medical School, Boston, Massachusetts 02115 Irene Iscla (201), Department of Physiology, University of TexasSouthwestern Medical Center, Texas 75390 Piotr Koprowski (295), Department of Cell Biology, Nencki Institute of Experimental Biology, 02‐093 Warsaw, Poland Andrzej Kubalski (295), Department of Cell Biology, Nencki Institute of Experimental Biology, 02‐093 Warsaw, Poland Ching Kung (311), Laboratory of Molecular Biology, Department of Genetics, University of Wisconsin, Madison, Wisconsin 53706 Anthony G. Lee (151), School of Biological Sciences, University of Southampton, Southampton SO16 7PX, United Kingdom Yuezhou Li (201), Department of Physiology, University of TexasSouthwestern Medical Center, Texas 75390 Chan Li (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Stephen H. Loukin (311), Laboratory of Molecular Biology, Department of Genetics, University of Wisconsin, Madison, Wisconsin 53706 V. S. Markin (87), Department of Anesthesiology and Pain Management, UT Southwestern, Dallas, Texas 75235 Boris Martinac (25), School of Biomedical Sciences, University of Queensland, St. Lucia, Brisbane, Queensland 4072, Australia Benjamin D. Matthews (59), Vascular Biology Program, Departments of Pathology and Surgery, Children’s Hospital and Harvard Medical School, Boston, Massachusetts 02115; Department of Pediatrics, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts 02114 Christopher A. McCulloch (179), CIHR Group in Matrix Dynamics, University of Toronto, Toronto, Ontario, Canada M5S 3E2

Contributors

xiii

Mark A. Messerli (393), BioCurrents Research Center, Program in Molecular Physiology, Marine Biological Laboratory, Woods Hole, Massachusetts 02543 Samantha Miller (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Paul C. Moe (201), Department of Physiology, University of Texas-Southwestern Medical Center, Texas 75390 Alexander G. Petrov (121), Institute of Solid State Physics, Bulgarian Academy of Sciences, 72 Tzarigradsko Chaussee, 1784 Sofia, Bulgaria Barbara G. Pickard (361), Gladys Levis Allen Laboratory of Plant Sensory Physiology, Biology Department, Washington University, St. Louis, Missouri 63130 Andrew M. Powl (151), School of Biological Sciences, University of Southampton, Southampton SO16 7PX, United Kingdom Douglas C. Rees (1), Division of Chemistry and Chemical Engineering 114‐96, Howard Hughes Medical Institute, California Institute of Technology, Pasadena, California 91125 Kenneth R. Robinson (393), Department of Biological Sciences, Purdue University,West Lafayette, Indiana 47907 F. Sachs (87), Physiology and Biophysical Sciences, SUNY Buffalo, New York 14214 Yoshiro Saimi (311), Laboratory of Molecular Biology, Department of Genetics, University of Wisconsin, Madison, Wisconsin 53706 Ulrike Schumann (269), School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom Anita Sengupta (179), Department of Anatomy, University of Bristol, School of Medical Sciences, University Walk, Clifton, BS8 1TD Bristol, United Kingdom

xiv

Contributors

Stefan Steinbacher2 (1), Division of Chemistry and Chemical Engineering 114‐96, Howard Hughes Medical Institute, California Institute of Technology, Pasadena, California 91125 Pavel Strop3 (1), Division of Chemistry and Chemical Engineering 114‐96, Howard Hughes Medical Institute, California Institute of Technology, Pasadena, California 91125 Sergei Sukharev (235), Department of Biology, University of Maryland, College Park, Maryland 20742 Charles K. Thodeti (59), Vascular Biology Program, Departments of Pathology and Surgery, Children’s Hospital and Harvard Medical School, Boston, Massachusetts 02115 Xinliang Zhou (311), Laboratory of Molecular Biology, Department of Genetics, University of Wisconsin, Madison, Wisconsin 53706

2 Present address: Proteros Biostructures GmbH, D-82152, Am Klopferspitz 19, Martinsried, Germany. 3 Present address: Howard Hughes Medical Institute, Department of Molecular and Cellular Physiology, E300 Clark Center, Stanford University, Stanford, California 94305.

Foreword Mechanosensitive Ion Channels, Part A Owen P. Hamill Department of Neuroscience and Cell Biology, University of Texas Medical Branch, Galveston, Texas

Mechanosensitive (MS) ion channels represent the third major class of gated membrane ion channels after voltage‐ and receptor‐gated channels. MS channels are formed by membrane proteins that are able to sense and transduce mechanical force into electroosmotic/signaling events that increase cell survival in a dynamic and sometimes hostile mechanical environment. On the other hand, when specific MS channels operate inappropriately, the consequences can lead to reduced cell viability and the development of several clinically relevant human pathologies. Over the last two decades, research on MS channels has expanded and diversified from the early electrophysiological studies of specialized mechanosensory nerve endings to include studies of cell types across the full evolutionary spectrum and to involve the new disciplines of structural biology, molecular genetics, drug discovery, and biotechnology. As a result there has been flood of new information with the potential for even greater breakthroughs in the near future. To highlight the excitement of the field, Current Topics in Membranes has compiled two volumes on MS channels that include chapters written by many of the leading researchers studying MS channels. Part A of this volume is organized into three sections. The first section covers topics on the atomic structure of two diVerent bacterial MS channel proteins MscL and MscS, the physical and thermodynamic principles that underlie mechanical and electromechanical activation of membrane proteins, and the cellular aspects that determine how mechanical forces are conveyed to membrane proteins via lipid–protein and protein–protein interactions. The second section provides an update from several laboratories on the molecular gating dynamics and the structure–function relations of the channels MscL and MscS. The third section covers MS channels in fungi and plant cells describing the identification of a transient receptor potential xv

xvi

Foreword

MS Ca2þ channel in yeast and an MscS‐like channel in plant cells, as well as providing new insight into the special role of mechanical force and MS channels in growing plants. I would like to thank Dale Benos for his invitation to submit the original proposal to Elsevier. I would also like to thank all those involved in the production of the volumes and, in particular, Phil Carpenter for his continual and patient eVorts during the compilation phase. Finally, I would like to thank all the scientists for presenting their discoveries regarding MS channels.

Previous Volumes in Series Current Topics in Membranes and Transport Volume 23 Genes and Membranes: Transport Proteins and Receptors* (1985) Edited by Edward A. Adelberg and Carolyn W. Slayman Volume 24 Membrane Protein Biosynthesis and Turnover (1985) Edited by Philip A. Knauf and John S. Cook Volume 25 Regulation of Calcium Transport across Muscle Membranes (1985) Edited by Adil E. Shamoo Volume 26 NaþHþ Exchange, Intracellular pH, and Cell Function* (1986) Edited by Peter S. Aronson and Walter F. Boron Volume 27 The Role of Membranes in Cell Growth and Differentiation (1986) Edited by Lazaro J. Mandel and Dale J. Benos Volume 28 Potassium Transport: Physiology and Pathophysiology* (1987) Edited by Gerhard Giebisch Volume 29 Membrane Structure and Function (1987) Edited by Richard D. Klausner, Christoph Kempf, and Jos van Renswoude Volume 30 Cell Volume Control: Fundamental and Comparative Aspects in Animal Cells (1987) Edited by R. Gilles, Arnost Kleinzeller, and L. Bolis Volume 31 Molecular Neurobiology: Endocrine Approaches (1987) Edited by Jerome F. Strauss, III, and Donald W. Pfaff

*Part of the series from the Yale Department of Cellular and Molecular Physiology. xvii

xviii

Previous Volumes in Series

Volume 32 Membrane Fusion in Fertilization, Cellular Transport, and Viral Infection (1988) Edited by Nejat Du¨zgu¨nes and Felix Bronner Volume 33 Molecular Biology of Ionic Channels* (1988) Edited by William S. Agnew, Toni Claudio, and Frederick J. Sigworth Volume 34 Cellular and Molecular Biology of Sodium Transport* (1989) Edited by Stanley G. Schultz Volume 35 Mechanisms of Leukocyte Activation (1990) Edited by Sergio Grinstein and Ori D. Rotstein Volume 36 Protein–Membrane Interactions* (1990) Edited by Toni Claudio Volume 37 Channels and Noise in Epithelial Tissues (1990) Edited by Sandy I. Helman and Willy Van Driessche

Current Topics in Membranes Volume 38 Ordering the Membrane Cytoskeleton Trilayer* (1991) Edited by Mark S. Mooseker and Jon S. Morrow Volume 39 Developmental Biology of Membrane Transport Systems (1991) Edited by Dale J. Benos Volume 40 Cell Lipids (1994) Edited by Dick Hoekstra Volume 41 Cell Biology and Membrane Transport Processes* (1994) Edited by Michael Caplan Volume 42 Chloride Channels (1994) Edited by William B. Guggino Volume 43 Membrane Protein–Cytoskeleton Interactions (1996) Edited by W. James Nelson Volume 44 Lipid Polymorphism and Membrane Properties (1997) Edited by Richard Epand Volume 45 The Eye’s Aqueous Humor: From Secretion to Glaucoma (1998) Edited by Mortimer M. Civan

Previous Volumes in Series

xix

Volume 46 Potassium Ion Channels: Molecular Structure, Function, and Diseases (1999) Edited by Yoshihisa Kurachi, Lily Yeh Jan, and Michel Lazdunski Volume 47 AmilorideSensitive Sodium Channels: Physiology and Functional Diversity (1999) Edited by Dale J. Benos Volume 48 Membrane Permeability: 100 Years since Ernest Overton (1999) Edited by David W. Deamer, Arnost Kleinzeller, and Douglas M. Fambrough Volume 49 Gap Junctions: Molecular Basis of Cell Communication in Health and Disease Edited by Camillo Peracchia Volume 50 Gastrointestinal Transport: Molecular Physiology Edited by Kim E. Barrett and Mark Donowitz Volume 51 Aquaporins Edited by Stefan Hohmann, Søren Nielsen and Peter Agre Volume 52 Peptide–Lipid Interactions Edited by Sidney A. Simon and Thomas J. McIntosh Volume 53 CalciumActivated Chloride Channels Edited by Catherine Mary Fuller Volume 54 Extracellular Nucleotides and Nucleosides: Release, Receptors, and Physiological and Pathophysiological Effects Edited by Erik M. Schwiebert Volume 55 Chemokines, Chemokine Receptors, and Disease Edited by Lisa M. Schwiebert Volume 56 Basement Membrances: Cell and Molecular Biology Edited by Nicholas A. Kefalides and Jacques P. Borel Volume 57 The Nociceptive Membrane Edited by Uhtaek Oh

CHAPTER 1 Structures of the Prokaryotic Mechanosensitive Channels MscL and MscS Stefan Steinbacher,1 Randal Bass,2 Pavel Strop,3 and Douglas C. Rees Division of Chemistry and Chemical Engineering 114‐96, Howard Hughes Medical Institute, California Institute of Technology, Pasadena, California 91125

I. II. III. IV.

V. VI. VII. VIII.

Overview Introduction Conductances of MscL and MscS: General Considerations Structure Determination of MscL and MscS A. General Considerations in Membrane Protein Crystallography B. Crystallographic Analysis of MscL and MscS MscL and MscS Structures The Permeation Pathway in MscL and MscS Disulfide Bond Formation in MscL Concluding Remarks References

I. OVERVIEW The prokaryotic mechanosensitive channels of large (MscL) and small (MscS) conductance respond directly to tension applied to the bacterial membrane. Crystal structures of the Mycobacterium tuberculosis MscL and 1 Present address: Proteros Biostructures GmbH, D-82152, Am Klopferspitz 19, Martinsried, Germany. 2 Present address: Department of Analytical Sciences, Amgen Inc., 1201 Amgen Court West, Seattle, Washington 98119. 3 Present address: Howard Hughes Medical Institute, Department of Molecular and Cellular Physiology, E300 Clark Center, Stanford University, Stanford, California 94305.

Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58001-9

2

Steinbacher et al.

˚ resolutions, Escherichia coli MscS were initially reported at 3.5‐ and 3.9‐A respectively. In subsequent refinements described in this chapter, sequence register errors have been corrected to produce improved models for both channels. The pentameric MscL and heptameric MscS are each organized into transmembrane and cytoplasmic domains, although their detailed architectures are distinct in terms of polypeptide folds and oligomeric states. The basic structural framework of the MscL and MscS transmembrane domains is provided by ‐helices; each subunit of MscL has 2 helices for a total of 10, whereas MscS has 3 helices per subunit for a total of 21. In contrast to the common architectural theme of helix packing evident in the transmembrane domains, the cytoplasmic domains of MscS and MscL are markedly diVerent in terms of both overall size and polypeptide fold. The permeation pathways in both structures are formed by the right‐handed packing of helices that create funnel‐shaped pores constricted near the cytoplasmic side by the side chains of hydrophobic residues. From considerations of the relationship between pore geometry and conductance, it is likely that both channel structures represent closed states.

II. INTRODUCTION Membrane integrity is vital to cellular growth and survival. Among the insults that may be experienced by organisms are changes in external osmolarity; concentration diVerences of only 10 mM can generate osmotic pressure diVerences of 0.2 atm that may rupture membranes of radii 3 mm (Hamill and Martinac, 2001). Cells immersed in environments that can encounter even modest osmolarity changes must consequently be able to respond on a suYciently rapid timescale to prevent lysis. Osmotic downshock conditions, such as the sudden exposure of a bacteria to rain or other source of freshwater, represent a particularly challenging situation (Booth and Louis, 1999; Poolman et al., 2002). Without safety‐value mechanisms to release cellular contents (Britten and McClure, 1962), cells would not be able to withstand the resultant turgor pressures of tens to hundreds of atmospheres associated with the influx of water. Through the pioneering eVorts of C. Kung and coworkers (Martinac et al., 1987; Sukharev et al., 1994, 1997), the proteins in bacteria responsible for sensing the increase in membrane tension accompanying osmotic downshock have been identified. These proteins form high‐conductance channels in the inner membrane that can open and close in direct response to tension applied to the bilayer. Such properties are consistent with a biological role for these channels in responding to sudden increases in turgor pressure to jettison water and other cellular contents to prevent cell lysis during hypoosmotic shock. To date, two

1. Crystal Structures of MscL and MscS

3

general families of these channels have been identified, the mechanosensitive channel of large conductance (MscL) (Sukharev et al., 1994) and of small conductance (MscS) (Levina et al., 1999). Reviews of these channels have appeared (Perozo and Rees, 2003; Strop et al., 2003; Sukharev and Corey, 2004; Blount et al., 2005; Booth et al., 2005; Sukharev et al., 2005) that emphasize diVerent aspects of these channels. Although they are relatively simple, intrinsically stretch‐activated systems, the basic principles of how MscL and MscS sense forces applied to the lipid bilayer likely reflect the mechanisms underlying such diverse cellular phenomena as touch, hearing, gravity, and pressure (Kung, 2005). From a structural perspective, MscL and MscS represent fascinating targets as they provide an opportunity to explore the coupling between protein conformation and the membrane environment responsible for channel gating. Tension and pressure sensitive systems, such as MscL and MscS, have the added attraction that these environmental properties are energetically coupled to changes in protein area and volume, respectively, which may be directly quantitated from structural models. The crystallographic analyses of the M. tuberculosis MscL (Chang et al., 1998) and the E. coli MscS (Bass et al., 2002) described in this chapter were motivated by these considerations to provide the structural frameworks essential for a mechanistic understanding of mechanosensitive systems at the molecular level.

III. CONDUCTANCES OF MscL AND MscS: GENERAL CONSIDERATIONS The conductance of a channel, g, describes the coupling between the current flow through the channel, I, and the driving force across the membrane, V, in the Ohm’s Law expression: gV ¼ I

ð1Þ

where g is the inverse of the channel resistance. When I and V are expressed in amperes and volts, respectively, the units of g are siemens (S) which are equivalent to reciprocal ohms. The conductances of MscL and MscS have been reported as 3 and 1 nS, respectively (Sukharev et al., 1997, 1999; Levina et al., 1999), when measured in solutions containing 200‐mM KCl and 40‐ to 90‐mM MgCl 2. With a potential diVerence of 100 mV, a conductance of 1.6 nS equals 160 pA, which is equivalent to the flow of 109 ions/ s across the membrane. These are quite high‐conductance channels; for comparison, Kþ channels and the acetylcholine receptor have conductances that are 100 times smaller than MscL (Hille, 2001). While these conductances reflect the properties of the fully open state, subconductance states have been reported for both channels (Sukharev et al., 1999; Shapovalov and

Steinbacher et al.

4

Lester, 2004; Akitake et al., 2005). An inactivated state of MscS has also been described (Akitake et al., 2005). From conductance measurements in the presence and absence of diVerent size molecules, the pore radius in the open ˚ (Cruickshank et al., state of MscL has been estimated in the range of 15–20 A 1997; Sukharev et al., 1999, 2001b). As anticipated from the high conductances, both channels are essentially nonselective, although MscS does exhibit a slight preference for anions (Martinac et al., 1987; Sukharev, 2002). The conductance of a channel reflects, among other factors, the geometry of the permeation pathway (Hille, 2001). Viewed as a continuum model, the conductance of a channel will vary linearly with the cross‐sectional area and inversely with the length of the permeation pathway. The conductance of real channels will, of course, depend on molecular details that cannot be captured in a continuum model. One interesting example is the observation in molecular dynamics simulations that hydrophobic pores of diameter ˚ are closed to water and ions, respectively—that smaller than 4.5 or 6.5 A is, nonconducting channels need not be geometrically closed (Beckstein and Sansom, 2004). Consequently, even qualitative analyses of low conductance, selective ion channels will likely require detailed calculations. For high conductance, nonspecific channels, however, such as MscL with estimated ˚ in the open state, macroscopic, continuum pore radii greater than 10 A models may not be a bad approximation, at least for a qualitative understanding of the dependence of channel conductance on the geometry of the permeation pathway. It is in this spirit that the following analysis is presented by way of background to help functionally interpret the MscL and MscS structures. In the simplest continuum model for the conductance of a cylindrical channel of radius r and length l, the conductivity is given by (Edmonds, 2001; Nelson, 2004): g¼

cDq2 pr2 kB T l

ð2Þ

where c, q, and D denote the concentration, electronic charge, and diVusion coeYcient for the univalent permeant ion, respectively; kB is the Boltzmann constant and T is the temperature. When c is expressed in molar concentra˚ ngstroms, and taking D ¼ 2 105 cm2/s for KCl at tion units, r and l in A 298 K (Robinson and Stokes, 1959), this expression reduces to g ¼ 2:4

cr2 ðnSÞ l

ð3Þ

Macroscopic models of this type have been found to overpredict the experimentally measured conductance of channels by factors of 5–6 (Smart et al., 1997), resulting in an empirical expression for geV:

5


geff 0:4

cr2 ðnSÞ l

ð4Þ

˚ , l ¼ 30 A ˚ , and c ¼ 0.3 M, geff 0.9 nS. with r ¼ 15 A Macroscopic models may also be used to estimate the volume of water and other cellular contents flowing through an open mechanosensitive channel driven by an osmotic pressure gradient. The volume per unit time, J, of a fluid of viscosity flowing through a macroscopic pipe of radius r and length l, under a pressure diVerential P is given by the Hagen–Poiseuille equation (Denny, 1993): J¼

pr 4 P 8 l

ð5Þ

˚ and l ¼ 30 A ˚ , exFor a mechanosensitive channel, approximated as r ¼ 15 A 5 periencing an osmotic pressure gradient of P ¼ 0.1 atm ¼ 1.013 10 dyne/cm2, and given the viscosity of water as ¼ 0.01 g/cm s, the volume flow through the pipe may be calculated from Eq. (5) as J ¼ 6.7 1015 cm3/s, which corresponds to 2.2 108 waters/s (from the average molecular volume for ˚ 3 ¼ 3 1023 cm3). For comparison, the permeation rate through water 30 A the aquaporin water channel is 109 waters/s (Zeidel et al., 1992). For a cell experiencing osmotic downshock, the volume flow of water per unit time (JV) across a membrane of surface area A in response to an osmotic pressure gradient, P, is given by (Weiss, 1996): JV ¼ LV AP

ð6Þ

where LV is the hydraulic conductivity. An approximate value of LV for biological membranes is 105 cm/s atm. If an osmotic pressure gradient is imposed across a membrane containing NC channels that support a volume flow given by the Hagen–Poiseuille equation, then the equilibrium condition where the flow across the membrane [Eq. (6)] is balanced by the flow through the channels [Eq. (5)] becomes: LV AP ¼ NC

pr 4 P 8l

NC 8LV l ¼ A pr4

ð7Þ

With the parameters defined as above, NC/A 5 107 channels/cm2. For a spherical cell of radius 104 cm (approximating E. coli), A 107 cm2 or NC 5 channels per cell. Experimental estimates from electrophysiological analyses suggest there are 4–5 MscL and 20–30 MscS channels per E. coli (Stokes et al., 2003). An important conclusion from this analysis is that ˚ in the open state are anticipated mechanosensitive channels with r 15 A

6

Steinbacher et al.

to have approximately nanoSiemens conductances and support volume flows that are consistent with the observed biological properties of these channels.

IV. STRUCTURE DETERMINATION OF MscL AND MscS A. General Considerations in Membrane Protein Crystallography While the details vary, the basic steps in the structure determination of MscL and MscS generally mirror those employed in the crystallographic analysis of most membrane proteins, starting with the bacterial photosynthetic reaction center (Deisenhofer et al., 1985). These steps may be somewhat arbitrarily classified as: (i) obtaining a suitable source for the membrane protein of interest, (ii) solubilization from the membrane, (iii) purification, (iv) crystallization, and (v) structure solution. A brief overview of these steps, noting aspects relevant to MscL and MscS, follows. A more detailed description of the MscL structural analysis can be found in Spencer et al. (2003). 1. Source of Membrane Protein It is no coincidence that the membrane proteins whose structures were first determined occurred naturally at high abundance in appropriate sources. Indeed, the first structures of recombinantly expressed membrane proteins, the prokaryotic channels KcsA (Doyle et al., 1998) and MscL (Chang et al., 1998) were determined in 1998, some 13 years after the reaction center. Since most membrane proteins are found in very low abundance, recombinant methods are essential for their structure determination. Not only do recombinant methods allow overproduction in increased quantities, but they also enable membrane proteins from many diVerent species to be produced and over‐expressed. In addition, they can be prepared with diVerent aYnity tags to facilitate rapid and eYcient purification. A wide variety of over‐ expression systems have been utilized or proposed for membrane proteins (Grisshammer and Tate, 1995, 2003). However, the most successful expression system has been based on E. coli (Drew et al., 2003), which has been almost exclusively utilized for crystal structure analyses of recombinant prokaryotic membrane proteins, including MscL and MscS. Prokaryotic systems enjoy a number of significant advantages for membrane protein work; in particular, the cells can be easily grown on a large scale, and the growth conditions and induction strategies can be manipulated to obtain suYcient quantities of purified protein to facilitate screening and refinement of crystallization conditions.


7

2. Solubilization Solubilization of integral membrane proteins from the phospholipid bilayer for crystallization studies is typically achieved through detergent extraction (Michel, 1991; Iwata, 2003). This results in the formation of protein–detergent mixed micelles where the hydrophobic region of the protein is solvated by the nonpolar component of the detergent, while the polar component interacts favorably with the aqueous solution. The choice of detergent is crucial; some detergents are too vigorous and will either dissociate oligomeric proteins into subunits or, as in the case of sodium dodecyl sulfate, denature the protein, while others are too mild to eYciently extract the protein from the membrane. Although certain detergents have been used repeatedly in structural analyses (such as dodecylmaltoside used with MscL), identification of the ‘‘correct’’ detergent is still an empirical process involving the screening of several dozen detergents and must be established for each protein. While detergents are extremely useful for the membrane extraction and homogenous preparation of membrane proteins for structural studies, they are not completely faithful mimics of the bilayer environment. In particular, detergent micelles are typically spherical, in contrast to the planar structure of the bilayer. Among other factors, these distinctions will perturb the lateral pressure profile (Cantor, 1999) experienced by the protein in detergent or the bilayer, which is reflected in a general destabilization of membrane proteins in detergent solutions (Stowell and Rees, 1995; Bowie, 2001; Odahara, 2004; Lee et al., 2005). In addition to these general eVects, there can be specific phospholipid–protein interactions that are important for protein stability and function, that can also have a significant impact on crystallization (Zhang et al., 2003; Long et al., 2005; Guan et al., 2006). This is an area that will clearly benefit from more careful quantitative analysis of the lipid content of membrane protein preparations. 3. Purification The use of aYnity tags has revolutionized the process of protein purification, including membrane proteins (Kashino, 2003). Ideally, the detergent‐ solubilized material can be adsorbed to column material with a high aYnity for the tag, which after washing can be eluted by an increased concentration of the appropriate ligand. Most frequently, aYnity tags composed of multiple histidines, typically 6–12, have been employed for purification with immobilized metal aYnity columns and subsequent elution with imidazole; in the cases of MscL and MscS, the tags contained 10 histidines. Other aYnity tag purification systems are also available (Terpe, 2003). Additional purification steps such as ion exchange or gel filtration chromatography can improve purity, although sometimes these steps have a negative eVect on

8

Steinbacher et al.

crystallization, presumably due to loss of phospholipids. Gel filtration chromatography and light scattering can be very helpful in characterizing the monodispersity of a sample. Detergent solubilized material can form a variable extent of high‐molecular‐weight species (soluble aggregates) that are not suited for crystallization; this was a particular problem with MscL due to the apparent formation of MscL–MscL dimers (i.e., a species with a total of 10 subunits). 4. Crystallization The strategy for production of three‐dimensional crystals of membrane proteins closely mirrors that for water‐soluble proteins; sets of conditions with varying concentrations and types of precipitants, salts, and buVers (pH) are screened, typically by vapor diVusion methods, for initial leads that are then optimized. An informative analysis of membrane protein crystallization is presented in Iwata (2003). With a few notable exceptions, crystals of integral membrane proteins are typically of modest diVraction quality. This is likely a consequence of two factors: the generally high solvent content of membrane protein crystals and the destabilization of membrane proteins in detergent solutions. The high solvent content reflects the presence of the detergent micelle surrounding the apolar regions of the protein surface, which inhibits participation of this region in lattice contacts. For the membrane protein data presented in Fig. 1, the average Matthews coeYcient (Vm) (Matthews, 1968) for a ˚ 3/Da, which corresponds to 70% solvent; this membrane protein is 4 A ˚ 3/Da for water‐soluble proteins may be compared to the average of 2.6 A (KantardjieV and Rupp, 2004), which corresponds to 50% solvent. In both cases, there is a trend that the diVraction resolution is correlated with Vm. Of ˚ 3/Da, which indicates particular note, both MscL and MscS have Vm 6 A these crystals have high solvent contents, even for membrane proteins, and this may be correlated with the observed ‘‘modest’’ diVraction quality of the crystals. 5. Structure Determination The crystal structure determination of membrane proteins involves the same general considerations as for any other sample. Other than data collection, the main barrier is phase determination; for MscL, this involved preparation of suitable heavy atom derivatives for the method of multiple isomorphous replacement, while it was possible to incorporate selenomethionine into MscS which permitted phase determination using the multiple wavelength anomalous diVraction phasing method. The main technical issues in the crystallographic analyses of membrane proteins reflect their typically (but not exclusively) modest resolution and associated high B factors (see Section IV.B). Advances


9

FIGURE 1 Distribution of the Matthews coeYcient (Vm) and the crystallographic resolution for integral membrane proteins of known structure. Entries were taken from the database of S. White (http://blanco.biomol.uci.edu/membrane_proteins_xtal.html; White, 2004). A general trend is evident that structures with higher Vm (greater solvent content) tend to diffract more poorly than structures with lower Vm and consequently a lower solvent content. The points corresponding to the MscL and MscS structures are indicated.

in synchrotron beam lines, including automated sample handling and data collection, and refinement algorithms have significantly enhanced the ability to extract useful data from weakly diVracting crystals.

B. Crystallographic Analysis of MscL and MscS The M. tuberculosis MscL and E. coli MscS structures were originally reported in 1998 and 2002 (Chang et al., 1998; Bass et al., 2002), and deposited in the Protein Data Bank (http://www.rcsb.org/pdb; Berman et al., 2000) as entries 1MSL and 1MXM, respectively. The diVraction data for both structures were characterized by a rapid decrease in intensity with resolution, associated with large values for the overall temperature factors of ˚ 2. As a result, while the diVraction data were very strong at low 100 A resolution, the average intensity quickly decreased with increasing resolution

10

Steinbacher et al.

˚ for MscL and MscS, as reflected in the limiting resolutions of 3.5 and 3.9 A respectively. While the experimentally phased electron density maps fortunately appeared to be of much better quality than would be expected for this overall temperature factor, it was diYcult to obtain well‐refined models. In the case of MscL, we initially were unable to refine any single model to values of R and Rfree below 0.40 and 0.42, respectively, and consequently a ‘‘multiple structures’’ strategy was implemented where nine copies of the structure were refined again the native data to final R and Rfree values of 0.26 and 0.35. The refinement of MscS was better behaved in that a single structure could be refined against the native data to R and Rfree values of 0.33 and 0.36, respectively, although these values are still higher than desirable. Despite extensive eVorts, we have been unable to improve the diVraction quality of MscL and MscS crystals beyond those originally described. Consequently, we decided to reanalyze the original data to see if improved structural models could be produced. Following reprocessing of the diVraction ˚ resolution for MscL and MscS, respectively, electron data to 3.5‐ and 3.7‐A density maps for model building and refinement were generated by calcu˚ resolution. Using nonlating phases from the PDB coordinates to 16‐A crystallographic symmetry averaging (five‐ and sevenfold for MscL and MscS, respectively) and solvent flattening, the phases were refined and incrementally extended over 200 steps to their limiting resolutions. As a consequence of the high degree of noncrystallographic symmetry and high solvent content, this procedure led to a significant improvement in the quality of the final electron density maps that facilitated rebuilding of the models. While the overall folds of the polypeptide chains were unchanged, register errors were detected and corrected in both MscL and MscS. A critical aspect in the refinement was the parameterization of the bulk solvent correction that is crucial for the proper scaling between observed and calculated structure factors at very low resolution. As these reflections are quite strong, small diVerences in scaling can significantly influence the resulting electron density maps, particularly in regions of the protein surface. The revised models have been refined to final values of R and Rfree of 0.319 and 0.338 for MscL, and 0.293 and 0.321 for MscS; while still high, they do represent a significant improvement over the original models and are a reasonable reflection of the quality of structural models at these resolutions. The Protein Data Bank entries for the revised coordinates of MscL and MscS are 2OAR and 2OAU, respectively. Relative to the 1MSL coordinate set, the most significant changes in the revised MscL structure are the modeling of the first 12 residues (missing in the original model) as an ‐helix, the complete rebuilding of the periplasmic


11

loop, and correction of a six‐residue register error in the cytoplasmic helix. When structurally equivalent residues are compared between the original and revised models, the overall root mean square (rms) deviation in C ˚ . If the sequence register corrections are not taken into positions is 1.5 A ˚ for all residues (10–118) in the account, the rms deviation increases to 8.7 A initial model (primarily due to the 6‐residue register shift in the cytoplasmic ˚ for residues in the membrane‐spanning helix), with an rms deviation of 1.8 A helices (15–43 and 69–89). Relative to the 1MXM model, the most significant changes in the revised MscS structure are adjustment for the deviations from exact noncrystallographic symmetry in the membrane‐spanning region, including rebuilding of the loop between TM2 and TM3, and correction of the sequence register for residues 160–195 and 226–244. One consequence of the deviations from sevenfold symmetry in the membrane‐spanning domain of MscS is that the permeation pathway more closely resembles a pinched tube. When all structurally equivalent residues in the two MscS models are ˚ ; without superimposed, the overall rms deviation in C positions is 1.4 A ˚ for all correcting for the register errors, the rms deviation increases to 3.4 A ˚ for residues in the second and residues (27–280) in the model, and is 2.2 A third transmembrane helices. There is relatively little change between the original and revised models in the permeation pathways of both MscL and MscS.

V. MscL AND MscS STRUCTURES MscL and MscS exist as pentamers and heptamers, respectively, with the permeation pathway surrounding the axis of rotational symmetry in the center of each channel. Although they share a common organization of an N‐terminal transmembrane domain and a C‐terminal cytoplasmic domain, the overall arrangements of the polypeptide folds are distinct, indicating that they do not share a common evolutionary ancestor. In the subsequent discussion, unless otherwise specified, residue numbers refer to the M. tuberculosis MscL or E. coli MscS sequences as appropriate. The polypeptide fold of an MscL subunit (Fig. 2A) exhibits a simple topology containing two membrane‐spanning helices (TM1 and TM2) and a cytoplasmic ‐helix near the C‐terminus. Starting from the conserved cytoplasmic N‐terminus, residues 1–12 of each subunit adopt an ‐helical conformation that would likely be positioned at the cytoplasmic surface of the membrane. TM1 (residues 13–47) crosses the membrane toward the periplasm and creates the permeation pathway through formation of a right‐ handed helix bundle with the symmetry related TM1s from the other

12

Steinbacher et al.

FIGURE 2 Ribbons diagram representations of the MscL channel. (A) The polypeptide fold of an individual MscL subunit viewed from the plane of the membrane, with the positions of the two membrane‐spanning helices, TM1 and TM2, indicated. The termini of the channel are designated by ‘‘N’’ and ‘‘C,’’ respectively. (B) The MscL pentamer viewed in the same orientation as (A). (C) The membrane‐spanning region of MscL viewed down the membrane normal from the periplasm. The side chains of Leu17 and Val21 that constrict the permeation pathway are shown as black CPK models. The subunit illustrated in (A) is shaded dark, while the remaining subunits are light in these panels. Ribbons representations in this chapter were prepared with MOLSCRIPT (Kraulis, 1991).

subunits (Fig. 2B and C). Residues 48–68 form an extended loop with poor density in the periplasm that approximates two antiparallel b‐strands. TM2 (residues 69–101) then returns to the cytoplasm along the exterior of the channel to complete the membrane‐spanning domain of MscL. Within the transmembrane domain, each TM1 helix interacts with four surrounding helices: two TM1 helices from adjacent subunits (helix crossing angle 42 ) and two TM2 helices, one from the same subunit (crossing angle 134 ) that contacts the periplasmic half of TM1 and the other from an adjacent subunit (crossing angle 175 ) that contacts the more cytoplasmic half of TM1. Although there are no contacts between pairs of TM2 helices, the subunits are further interconnected in the transmembrane domain by the threading of the N‐terminal helix of one subunit between the TM1 and TM2 helices of a neighboring subunit. The membrane and cytoplasmic domains are connected by a short linker leading to a cytoplasmic helix (106–125) in each subunit that associates with the symmetry related mates to form a left‐handed (crossing angle þ20 ), five‐helix bundle. Of particular note, the periplasmic loop and the cytoplasmic helix in the revised


13

crystal structure now more closely resemble the model proposed by Guy and Sukharev in their analysis of MscL gating (Sukharev et al., 2001a,b). While the helix packing arrangement in the membrane‐spanning domain of MscS is also relatively simple, the overall topology of the polypeptide fold in the cytoplasmic domain is considerably more complex than observed for MscL (Fig. 3A and B). Each subunit of MscS contains three membrane‐ spanning helices. In contrast to MscL, the N‐terminus of MscS is periplasmic, with TM1 (residues 27–60) crossing the bilayer on the exterior of the channel, TM2 (residues 63–90) forming a central layer, and TM3 (93–128) returning to the cytoplasm through the channel interior. The TM1 and TM2 helices within one subunit are packed together in an antiparallel fashion (crossing angle 165 ) that buries an extensive interface, but makes few contacts with other helices in the membrane‐spanning domain and has relatively weak density. The permeation pathway (Fig. 3C) is formed by the packing of adjacent TM3 helices to form a right‐handed helix bundle (packing angle 22 ). A pronounced kink is present in the TM3 helix near Gly113 which results in the axis of the C‐terminal end of this helix being

FIGURE 3 Ribbons diagram representations of the MscS channel. (A) The polypeptide fold of an individual MscS subunit viewed from the plane of the membrane, with the positions of the three membrane‐spanning helices, TM1, TM2, and TM3, indicated. The termini of the channel are designated by ‘‘N’’ and ‘‘C,’’ respectively, as are the ‘‘middle b’’ and ‘‘COOH‐terminal’’ domains of the cytoplasmic region of MscS. (B) The MscS heptamer viewed in the same orientation as (A). (C) The membrane‐spanning region of MscS viewed down the membrane normal from the periplasm. The side chains of Leu105 and Leu109 that constrict the permeation pathway are shown as black CPK models. The oval shape of the pore reflects deviations from exact sevenfold symmetry in this region. The subunit illustrated in (A) is shaded dark, while the remaining subunits are light in these panels.

14

Steinbacher et al.

FIGURE 4 The permeation pathways of MscL (left) and MscS (right), defined by the N‐terminal helix and TM1 of MscL and TM3 of MscS, as viewed from the membrane. One subunit is depicted in ribbons representation for each structure, while the symmetry‐related structures are shown as C traces. This figure illustrates how the pore‐forming helix in both structures connects directly to a helix likely positioned in the cytoplasmic surface of the membrane. The locations of Arg88 that may contribute to the anionic preference of MscS are shown as gray ball‐and‐stick models on the periplasmic surface.

oriented nearly perpendicular to the membrane normal. It is likely that this region of TM3 extends out of the membrane and is positioned at the headgroup‐aqueous interface. Intriguingly, although the polarity of the polypeptide chain is reversed, this feature resembles the N‐terminal and TM1 helices of MscL in that a helix positioned at the cytoplasmic membrane surface is directly connected to the helix lining the permeation pathway (Fig. 4). The minimal interaction between the TM1–TM2 helical hairpin and the rest of MscS, particularly the permeation pathway formed by TM3, is somewhat surprising, and suggests the possibility that this more loosely packed state may be stabilized by detergent (foscholine‐14). As a consequence of the lateral pressure profile in the bilayer, a reasonable expectation for when MscS is embedded in a membrane is that the TM1–TM2 helices would be tightly packed against the permeation pathway, which is not observed. This situation is reminiscent of the behavior of the voltage sensor in the KvAP Kþ channel structure (Jiang et al., 2003; Lee et al., 2005); indeed, there are intriguing similarities between these two elements of MscS and Kþ channels (Bass et al., 2003), including the presence of multiple arginine residues in a membrane embedded helical hairpin. In view of the initial report that MscS was voltage sensitive (Martinac et al., 1987), it was


15

suggested that the arginine residues present in TM1 and TM2 might be responsible for this dependence (Bass et al., 2002). Subsequent studies have indicated that the voltage dependence of MscS is considerably more complex than originally envisioned and likely has little eVect on the closed to open transition (Akitake et al., 2005). Nevertheless, these common features between the Kþ channel voltage sensor and the TM1–TM2 hairpin of MscS are suggestive that arginine‐rich helices are conformationally sensitive elements that are energetically poised to move relative to the membrane in response to environmental changes (Hessa et al., 2005). The cytoplasmic domain of MscS is quite extensive relative to MscL, with ˚ the most notable structural feature being the large interior chamber of 40‐A diameter that connects to the cytoplasm through multiple openings. The protein framework that encloses this chamber is generated by two domains of each subunit, termed middle‐b and COOH‐terminal. The middle‐b domain is organized around a b‐sheet that connects with those from other subunits to form a continuous b‐sheet structure that extends around the entire protein. As noted by Mura et al. (2003), this arrangement exhibits a striking resemblance to the heptameric Sm proteins involved in mRNA processing in terms of both the subunit fold and oligomeric organization. The b‐sheet in this domain is twisted 180 ; as a consequence of the odd number of subunits in MscS, the continuous sheet forms the equivalent of a molecular Mobius strip since it twists 3.5 times in one complete cycle around the cytoplasmic domain of MscS. The cytoplasmic domain is completed by the COOH‐terminal domain that exhibits a mixed /b‐structure assigned to the ferredoxin fold family in the SCOP database (Murzin et al., 1995). The entire heptameric assembly is linked at the C‐terminus by a seven‐stranded parallel b‐barrel that contains one strand from each subunit.

VI. THE PERMEATION PATHWAY IN MscL AND MscS The permeation pathway across the membrane is dominated by the packing of symmetry related helices, either TM1 in MscL (Fig. 2C) or TM3 in MscS (Fig. 3C), into a right‐handed bundle. The striking pattern of conserved Gly and Ala noted in MscL and MscS (Levina et al., 1999) corresponds to residues localized at these helix–helix packing interfaces. The lining of the pore with a right‐handed helical arrangement is not a unique feature of mechanosensitive channels and a similar organization has been observed in other systems, including members of the Kþ channel (Doyle et al., 1998) and aquaporin families (Fu et al., 2000; Murata et al., 2000). The general shape and size of the helical framework surrounding the permeation pathway reflects the molecular symmetry, the helix–helix crossing angle ()

Steinbacher et al.

16

and the tilt of the helix axis () with respect to the membrane normal (taken to coincide with the symmetry axis). With ideal helices and exact N‐fold rotational symmetry, these parameters are not all independent but are related by the expression (Spencer and Rees, 2002) cos a ¼ cos2 þ sin2 cos y

ð8Þ

where y ¼ 2p/N. Values of these parameters for the helices surrounding the pores in MscL, MscS, and KcsA are provided in Table I, and it may be seen that the helix–helix crossing angles calculated from this equation are in good agreement with the values observed in these structures. An interesting question concerns how the oligomeric state is specified; for example, the crossing angles of MscL and KcsA agree to within 3 , and the tilt angles within 4 , yet the former is a pentamer while the latter is a tetramer. Presumably this behavior reflects the pattern of residues, particularly Gly and Ala, along the helix packing interface, although this code cannot yet be deciphered. The permeation pathways of both MscL and MscS are roughly funnel shaped with the larger opening facing the periplasmic surface of the membrane and the narrowest point near the cytoplasm. A more quantitative analysis of the pore geometry by the program HOLE (Smart et al., 1996) ˚ at the periplasm, reveals that the pore in MscL varies from a radii of 15 A ˚ to the narrowest point of 1

ND

ND

Li et al., 2002; Blount et al., 2005

MscA1

Archaea

0.38 (þve) 0.68 (ve)

11

ND

ND

15

Le Dain et al., 1998

MscA2

Archaea

0.85 (þve) 0.49 (ve)

17

ND

ND

29

Le Dain et al., 1998

MscMJ

Archaea

0.27

9

PKþ =PCl 6

CPZ, TNP

5

Kloda, 2001; Kloda and Martinac, 2001b

MscMJLR

Archaea

2.2 (þve) 1.7 (ve)

27

PKþ =PCl 5

Not aVected by CPZ or TNP

18

Kloda, 2001; Kloda and Martinac, 2001c

MscTA

Archaea

2.8

ND

Nonselective

TNP

35

Kloda and Martinac, 2001d; Kloda and Martinac, 2002

MSL2

Plants

ND

ND

ND

ND

ND

Haswell and Meyerowitz, 2006

MSL3

Plants

ND

ND

ND

ND

ND

Haswell and Meyerowitz, 2006

a

ND indicates that a particular property has not been determined. Abbreviations: CPZ, chlorpromazine; TNP, trinitrophenol; LPC, lysophosphatidylchlone.

32

Boris Martinac

A

O2 O1 C

MscS −40 mmHg

80 pA 1 sec

O3

O2 O1

MscL −60 mmHg

C

MscA1 −50 mmHg

O1 C

O1 C

MscA2 −60 mmHg

O1 C

MscMJ −60 mmHg

O1 MscMJLR −30 mmHg

C O1

MscTA −80 mmHg

C

B Solubilized or recombinant protein

Lipids −, +H2O Detergent Reconstituted proteoliposome

2. MS Channels of Prokaryotes

33

heterologous as well as in vitro transcription/translation system demonstrated that the mscL gene alone is necessary and suYcient for the MscL activity. Since its discovery, genes homologous to mscL were found in various Gram‐negative and Gram‐positive bacteria, archaea and a single fungal genome (Kloda and Martinac, 2002; Martinac and Kloda, 2003; Pivetti et al., 2003). MscS and MscK were cloned by Booth and coworkers (Booth and Louis, 1999; Levina et al., 1999), who identified two genes on E. coli chromosome, yggB and kefA. Deletion of the two genes led to the abolishment of the activity of the MS channels of small conductance, which was originally described as a single type of MS channel in bacterial spheroplasts (Martinac et al., 1987). The MS channels aVected by the kefA and yggB null mutations correspond to MscK and MscS, respectively. The MscS channel activity is characterized by a large number of channels gating simultaneously encountered in almost 100% of spheroplast patches. MscS inactivates rapidly on sustained application of pressure (Koprowski and Kubalski, 1998). The activity of the KefA channels is less frequently encountered (70% of the patches). It is characterized by fewer channels, which do not inactivate on prolonged application of pressure to the patch pipette. YggB is a small membrane protein of 286 amino acids. In contrast, KefA is about five times larger, multidomain membrane protein of 1120‐amino acid residues. The primary amino acid sequence of YggB resembles highly the sequence of the last two domains of the KefA protein. D. Molecular Identification of MS Channels in Archaea Archaea, formerly referred to as archaebacteria, are prokaryotes like bacteria. They exist in extreme environments found on Earth (Barinaga, 1994) and constitute a separate domain on the phylogenetic tree (Fig. 1) (Woese, 1994; Pace, 1997). The existence of MS channels in archaea was first documented in the halophilic archaeon Haloferax volcanii (Le Dain et al., 1998) followed by molecular identification and functional characterization

FIGURE 3 Multiplicity of MS channels in prokaryotes. (A) Shown are current traces E. coli MscL and MscS, followed by traces of MscA1 and MscA2 of Haloferax volcanii and MscMJ and MscMJLR of Methanococcus jannaschii recorded from channels reconstituted into liposomes. The last current trace represents activity of MscTA, the channel of T. acidophilum. All traces were recorded at þ40 mV at negative pipette pressures indicated on the left of each trace. C denotes the closed state and On denotes open state of n channels. (B) A scheme of a dehydration/rehydration method used for liposome reconstitution of MS channels. Note: 1 mm Hg ¼ 133 Pa. Reproduced from Martinac and Kloda (2003).

34

Boris Martinac

of MS channels in methanogenic archaeon Methanococcus jannaschii (Kloda and Martinac, 2001b,c) and in two thermophilic archaea T. volcanium and T. acidophilum (Fig. 1) (Kloda and Martinac, 2001d). MscA1 and MscA2 are two types of MS channels found in the cell membrane of Haloferax volcanii (Le Dain et al., 1998) (Fig. 1). Both channels have large conductance, rectified with voltage, and are blocked by submillimolar concentrations of the lanthanide Gd3þ, a common blocker of MS channels (Sachs and Morris, 1998; Hamill and Martinac, 2001) (Table I). Similar to the bacterial MS channels, they are activated solely by tension in the lipid bilayer. Consequently, in patch‐clamp experiments they fully preserve their mechanosensitivity after detergent solubilization and reconstitution into artificial liposomes (Fig. 3). Two types of MS channels, MscMJ and MscMJLR, have been identified in the genome of Methanococcus jannaschii (Kloda and Martinac, 2001b,c). The primary amino acid sequence of MscMJ shares high homology with MscS of E. coli (Levina et al., 1999). The channel has conductance of 270 pS and prefers cations to anions with a selectivity characterized by PKþ =PCl 6. Its activation by membrane tension is comparable to the MscS activation (Table I). MscMJLR (i.e., MS channel of Methanococcus jannaschii of large conductance and rectifying) is a second MS channel of Methanococcus jannaschii, which was identified and functionally characterized shortly after MscMJ was described (Kloda and Martinac, 2001b). Like MscMJ, MscMJLR shares sequence homology with a large group of MscS‐like proteins identified in prokaryotic microbes as well as in eukaryotic organisms including the experimental plant Arabidopsis thaliana and fission yeast Schizosaccharomyces pombe (Kloda and Martinac, 2002; Pivetti et al., 2003). MscMJLR is cation selective with the permeability ratio PKþ =PCl 5 comparable to the selectivity of MscMJ. However, MscMJLR diVers from MscMJ in both conductive and MS properties. Comparable to MscL of E. coli, MscMJLR has a very large conductance of 2.0 nS that is approximately seven times larger than the 270‐pS conductance of MscMJ. It also requires much higher membrane tension for activation (Kloda and Martinac, 2001c) (Table I). MscMJLR is also blocked by submillimolar concentrations of Gd3þ comparable to other prokaryotic MS channels (Kloda, 2001; Kloda and Martinac, 2002). Interestingly, the amino acid sequence of the third membrane‐spanning domain TM3 of MscMJ and MscMJLR resembles the sequence of the highly conserved TM1 helix of MscL. This is important because TM1 is the helix essential for the opening of the MscL pore by membrane tension (Yoshimura et al., 1999; Ajouz et al., 2000; Betanzos et al., 2002; Perozo et al., 2002a). The presence of multiple MS channels in prokaryotic cells indicates the importance of MS channels


35

for the survival of these microbes being frequently exposed to environmental osmotic challenges. The MS channels of thermophilic archaea T. volcanium and T. acidophilum were identified using a functional approach similar to the one used for molecular ident ification of MscL (Sukharev et al ., 1994b). Twenty N‐terminal amino acid residues of the MS protein of T. volcanium match with 75% identity the start of the open reading frame of a gene encoding MscTA of the related T. acidophilum (Kloda and Martinac, 2001c). The channel is nonselective for cations and anions and has a large conductance of 2.0 nS, comparable to the conductance of MscL and MscMJLR. Similar to all currently known prokaryotic MS channels, MscTA is activated purely by membrane tension in the lipid bilayer (Kloda and Martinac, 2001d). However, membrane tension required for MscTA activation is unusually high compared to other bacterial and archaeal MS channels (Table I) although high negative pressure is also required for the activation of MscL homologues found in Synechocystis sp. and Mycobacterium tuberculosis (Moe et al., 1998, 2000). E. Molecular Structure of Prokaryotic MS Channels The structure of two prokaryotic MS channels, MscL and MscS, has been solved by X‐ray crystallography. Rees and coworkers (Chang et al., 1998) solved the 3D oligomeric structure of the MscL homologue from Mycobac˚ terium tuberculosis (Tb‐MscL). The structure of MscL obtained at 3.5‐A resolution shows a pentameric channel most likely in a closed state. The channel monomer is composed of two a‐helical transmembrane (TM) domains, TM1 and TM2, cytoplasmic N‐ and C‐terminal domains, and a central periplasmic domain (Fig. 4A). The transmembrane TM1 helices form ˚ at the cytoplasa tightly packed bundle funneling to a constriction of 2 A mic side of the channel. The hydrophobic constriction is thought to function as the channel gate. On the basis of functional studies examining permeation of large organic cations through the channel pore (Cruickshank et al., 1997) as well as on spectroscopic studies (Perozo et al., 2002b), the diameter of the MscL funnel at the constriction site was determined to vary between 2 and ˚ during the channel gating. The overall change in diameter of the 30 A ˚ (Corry et al., 2005), indicating channel protein on MscL opening is 16 A that during opening MscL is undergoing one of the largest conformational changes known in membrane proteins (Fig. 4B). TM1 and TM2 helices are connected by a periplasmic loop that is structurally not well defined. The periplasmic loop is thought to function as a spring resisting the channel

36

Boris Martinac

A Periplasm

TM1 ~35 Å

Membrane

TM2 Cytoplasm

Channel gate

Diameter of pentamer (Å) B Close

50 Å

Open

66 Å

FIGURE 4 (A) The structure of the MscL channel pentamer (left) and a channel monomer (right) from Mycobacterium tuberculosis according to the 3D structural model of a most likely closed channel (Chang et al., 1998). The thickness of the membrane bilayer (shown as solid ˚ . The channel gate is formed by a group of amino acids at the cytoplasmic end blocks) is 3.5 A of the TM1 transmembrane domain. Modified from Oakley et al. (1999). Figure based on model 1MSL in Protein Data Bank (http://www.rcsb.org/pdb/). (B) A diagram of a closed and open MscL channel indicating the scale of conformational change involved in channel gating based on a FRET spectroscopic study. Adapted from Corry et al. (2005).

opening (Ajouz et al., 2000). A molecular dynamics study of MscL embedded in a curved bilayer suggested that the periplasmic loop could be the first among the MscL domains undergoing structural changes on channel opening (Meyer et al., 2006). The secondary structure of the cytoplasmic N‐terminal domain remains unresolved at present. However, amino acid


37

deletions or substitutions in the N‐terminus were shown to severely aVect MscL gating (Blount et al., 1996; Ha¨se et al., 1997), suggesting a significant functional role for this structural domain. On the basis of results showing that disulfide coupling is occurring between several highly conserved N‐terminal residues that were replaced by cysteines, a model has been proposed in which the N‐terminus presents an integral component of the MscL‐gating mechanism. In this model, the N‐terminus forms a second gate working in accord with the ‘‘hydrophobic gate’’ in the TM1 helix bundle (Sukharev et al., 2001; Betanzos et al., 2002). Nevertheless, the precise role of the N‐terminal domain in the MscL gating still awaits to be determined experimentally. According to the 3D crystal structure, C‐terminus forms an a‐helical bundle (Chang et al., 1998). Its physiological relevance has been put in doubt given the unorthodox crystallographic conditions (pH 3.5) and the abundance of charged groups pointing at the core of the bundle thus indicating a possible instability of its structure. Spectroscopic and molecular dynamics studies, however, demonstrated that under physiological conditions (pH 7.0) the C‐terminal cytoplasmic domain also forms an a‐helical bundle, located near the fivefold symmetry axis of the channel molecule (Elmore and Dougherty, 2001; Perozo et al., 2001; Martinac, 2004). According to a model based on cysteine‐cross‐linking experiments, the charged residues of C‐terminal helices point toward the aqueous medium and the a‐helical bundle is held together by leucine–isoleucine interactions (Anishkin et al., 2003). Interestingly, deletion of the C‐terminal bundle was shown not to significantly aVect MscL mechanosensation (Blount et al., 1996; Ha¨se et al., 1997; Ajouz et al., 2000), suggesting that this structural domain does not participate in the channel gating. The role of the C‐terminus was proposed to be that of a size‐exclusion filter at the cytoplasmic side of the MscL pore, preventing loss of essential metabolites (Anishkin et al., 2003). According to this model, the C‐terminal domain is stably associated in both closed and open conformations of the channel. A study, however, showed that the stability of the domain is pH dependent (Kloda and Martinac, 2006), indicating that the cytoplasmic a‐helical bundle may not only function as a size‐exclusion filter but also influence channel gating in a pH‐dependent manner. Rees and coworkers (Bass et al., 2002) solved also the 3D crystal structure ˚ resolution, the MscS structure shows of MscS of E. coli. Obtained at 3.9‐A that the channel folds as a homoheptamer, which has a large, cytoplasmic region (Fig. 5). Each of the seven MscS subunits contains three TM domains with N‐termini facing the periplasm and C‐termini at the cytoplasmic end of the channel. According to the crystal structure, the TM3 helices line the channel pore, whereas the TM1 and TM2 helices constitute the sensors for membrane tension and voltage (Bass et al., 2002; Bezanilla and Perozo, 2002).

38

Boris Martinac 80 Å Periplasm

TM1

Membrane TM3

TM2

120 Å Cytoplasm

Middle b -domain

C-terminus FIGURE 5 The structure of MscS from E. coli showing the channel homoheptamer (left) and a monomer (right) based on the crystal structure (Bass et al., 2002) and viewed by PyMol19. Residues 27–280 were resolved. Secondary structural domains and the position of the TM3 transmembrane helix are indicated in the diagram of the monomer. Highlighted in red is a conserved structural motif of glycine and alanine residues in the pore‐lining transmembrane helix TM3 essential for gating of prokaryotic MS channels. Reproduced from Martinac (2005a).

Although on initial analysis the 3D structure of MscS was thought to be that of an open channel (Bass et al., 2002), the precise conformation of MscS in the crystal form is controversial at present. A study employing molecular dynamics simulations implied that water and ions cannot pass through the channel pore because of the hydrophobicity of the TM3 residues lining the narrowest portion of the channel pore. This suggested that the crystal structure may reflect an inactive or desensitized functional state rather than the open state (Anishkin and Sukharev, 2004). In another computer simulation study, electric fields were applied to the MscS channel to model the eVect of the membrane potential (Spronk et al., 2006). As expected, the application of a potential increased the hydration of the pore and resulted in current flow through the MscS channel. Since the calculated channel conductance was in good agreement with experiment, it was concluded that the MscS crystal structure could be closer to a conducting than a nonconducting state, which ˚ diameter of the TM pore (Bass et al., 2002). would correspond to 11‐A According to another molecular dynamics, simulation study the diameter of


39

˚ in its the highly hydrophobic MscS channel pore was measured to be 6.5 A narrowest section (Sotomayor and Schulten, 2004). Nevertheless, this study seems to support the notion of the crystallographic structure representing an open state of the channel, because the simulations reported a spontaneous closure of the MscS TM pore when it was permitted to gate spontaneously in a relaxed membrane environment. The channel could be reopened in further simulations by applying membrane tension, which allowed a detailed view of interactions and geometric transformations governing pore closing and opening. TM1 and TM2 transmembrane domains surround the TM3 helices and are in contact with membrane lipids indicating that they may constitute the sensor for membrane tension. In addition, TM1 and TM2 helices may also underlie modulation of the channel by voltage (Martinac et al., 1987; Sukharev, 2002) because of the presence of three arginine residues in their structure (Bass et al., 2002; Bezanilla and Perozo, 2002). However, the precise contribution of these charged residues to the channel voltage dependence has to be established experimentally. A large C‐terminal cytoplasmic domain is characterized by an interior ˚ in diameter, which is in contact with the cytoplasm through chamber of 40 A multiple openings. Similar to the C‐terminal domain of MscL, the cytoplasmic domain of MscS could function as a molecular sieve designed to exclude essential solutes from leaving bacterial cells during a hypoosmotic shock. F. Bilayer Mechanism and Gating by Mechanical Force The property of being activated by amphipaths, which are compounds having both hydrophilic and hydrophobic properties and were reported to reversibly change shape of red blood cells (Deuticke, 1968; Sheetz and Singer, 1974), led to a proposal that bacterial MS channels should sense directly membrane tension developed in the lipid bilayer alone (Martinac et al., 1990; Markin and Martinac, 1991). The bilayer mechanism, as this mechanism of the MS channel gating has since been named (Hamill and McBride, 1997), found further support from studies showing that bacterial MS channels preserved their mechanosensitivity after reconstitution into artificial liposomes (Berrier et al., 1989; Delcour et al., 1989; Ha¨se et al., 1995; Blount et al., 1996). This property turned out to be crucial for molecular identification of MscL, the first MS channel identified at the molecular level (Sukharev et al., 1994a, 1997). The bilayer mechanism has since been well documented not only for bacterial channels (Berrier et al., 1989; Delcour et al., 1989; Sukharev et al., 1993, 1994a,b, 1999; Ha¨se et al., 1995), but also for archaeal (Le Dain et al., 1998; Kloda and Martinac,

40

Boris Martinac

2001b,c,d) as well as for eukaryotic MS channels (Patel et al., 1998, 2001; Maroto et al., 2005). The property of some eukaryotic MS channels of being gated by bilayer mechanisms was also essential for identification of TRPC‐1 as the MscCa in vertebrate cells (Maroto et al., 2005). Activation of prokaryotic MS channels by pressure (i.e., membrane tension) follows Boltzmann distribution function of the form: ðGo t AÞ 1 1 ð1Þ Po ¼ exp½aðp1=2 pÞ ¼ exp kT where Po is the single channel open probability, a is the slope of ln [Po/(1–Po)] plotted against negative pressure, p1/2 is the negative pressure (suction) applied to the patch pipette at which the MS channel is open 50% of the time (i.e., Po ¼ 0.5), Go is the diVerence in free energy between the closed and open conformations of the channel in the absence of externally applied membrane tension, A is the diVerence in membrane area occupied by an open and closed channel at a given membrane tension, and tA is the work required to keep an MS channel open by external mechanical force at the open probability 0 < Po < 1. The conversion from negative pressure (suction) p applied to a patch pipette to membrane tension t is obtained using the Laplace’s law t ¼ p(r/2) in which r is the radius of curvature of the membrane patch. This conversion between pressure applied to the patch pipette and bilayer tension in the membrane patch is possible because it was shown that MS channels respond to mechanical forces along the plane of the cell membrane (membrane tension), and not pressure perpendicular to it (Gustin et al., 1988; Sokabe and Sachs, 1990). Membrane tensions required for half activation of MS channels are on the average in the order of several dynes/cm (103 N/m) (Sachs, 1988). Since membrane tension t is nearly proportional to the pressure within the range of pressures required for activation of a prokaryotic MS channel reconstituted into a liposome patch and, therefore, is well approximated by the Laplace’s law, multiplying p1/2 by a (Eq. 1) gives a good estimate of the free energy of MS channel activation Go (Hamill and Martinac, 2001; Martinac, 2001): MSC

¼ p1=2 a ¼

Go kT

ð2Þ

The estimates of Go obtained for MscL and MscS using Eq. (2) are 17.0 and 7 kT, respectively (Martinac, 2001), which is in a good agreement with the patch‐clamp results showing that approximately two times less negative pressure is required for activation of MscS compared to MscL in giant spheroplasts of E. coli (Berrier et al., 1996; Blount et al., 1996) (Table I).

41


G. Spectroscopic Studies Lipid bilayer is at least 10 times more compressible in area than in volume (Hamill and Martinac, 2001). Consequently, any fractional change in area is accompanied by a proportional change in membrane thickness (h) so that A h ¼ A0 h0

ð3Þ

where h0 and A0 are the unstressed membrane thickness and area, respectively. A 2–4% change in bilayer area with a thickness of 3.5 nm would thin the membrane 0.1 nm. Given that thinning of a liposome patch would produce a change in matching hydrophobic surfaces of the bilayer and a reconstituted MS channel protein, the assumption was that hydrophobic mismatch could trigger MS channel activation. This is because the energy for transferring a hydrophobic protein surface from an organic solvent to an aqueous environment is 17 mJ/m2 (Chothia, 1974). The hydrophobic surface match model derives from the original studies of the gating of gramicidin. This is a small hydrophobic peptide of 15 amino acids that forms cation‐selective channels in lipid bilayers by membrane association of one momomer from each monolayer (O’Connell et al., 1990; Harms et al., 2003). Gramicidin exhibits tension sensitivity in lipid bilayers (Elliot et al., 1983) and can switch between stretch activation and stretch inactivation depending on the thickness of the bilayer in which it is reconstituted (Martinac and Hamill, 2002). Together with the fact that prokaryotic MS channels can be activated by amphipaths known to insert preferentially in one leaflet of the bilayer (Martinac et al., 1990) the assumption that bilayer tension could aVect hydrophobic matching between the bilayer and the MS protein led to a spectroscopic and patch‐clamp study in which two potential triggers of MS channel gating by the bilayer mechanism were evaluated: (i) protein–lipid bilayer hydrophobic mismatch and (ii) bilayer curvature (Perozo et al., 2002a) (Fig. 6). In this study, structural changes in MscL induced either by hydrophobic mismatch or curving the bilayer by insertion of the amphipath lysophosphatidylcholine (LPC) were examined by combining cysteine‐scanning mutagenesis with site‐directed spin labeling (SDSL), electronparamagnetic resonance (EPR) spectroscopy, and patch‐clamp functional analysis of MscL reconstituted into liposomes. The study demonstrated that hydrophobic surface match could stabilize intermediate conformations of MscL requiring less tension to open the channel in thin bilayers (18 hydrocarbons per acyl chain), but was insuYcient to fully open the channel. However, curving the bilayer by asymmetric insertion of LPC opened MscL without applying membrane

42

Boris Martinac

FIGURE 6 Schematic diagram of two possible mechanisms of MscL activation by bilayer deformation forces. Hydrophobic mismatch and bilayer curvature are considered as deformation forces of pressure‐induced changes in the lipid bilayer causing conformational changes in MS channels. These changes were studied experimentally by reconstituting purified MscL proteins in liposome bilayers prepared from synthetic phosphatidylcholine lipids of well‐defined composition (Perozo et al., 2002a). Reproduced from Martinac (2005b).

tension (Perozo et al., 2002a). Thus, the SDSL EPR spectroscopic study by Perozo et al. (2002a) has demonstrated that the mechanism of mechanotransduction in MS channels is defined by both local and global asymmetries in the transbilayer tension profile at the lipid–protein interface, since addition of LPC to one monolayer of liposomes reconstituted with MscL channels created local stresses leading to redistribution of the transbilayer pressure profile in


43

the lipid bilayer, whereas LPC addition to both monolayers did not. The open ˚ in diameter which is lined by the state of MscL has a water‐filled pore of >25 A TM1 helices from the five subunits (Perozo et al., 2002b). This result is consistent with several studies showing that MscL undergoes a large conformational change when opening and closing (Biggin and Sansom, 2001; Gullingsrud et al., 2001; Sukharev et al., 2001; Betanzos et al., 2002; Colombo et al., 2003; Gullingsrud and Schulten, 2003). Conformational changes involved in MscL gating have also very been measured using FRET spectroscopy (Corry et al., 2005). In this study, MscL reconstituted into liposomes was also activated by LPC similar to the SDSL EPR study and the change in FRET eYciency on the channel opening was recorded using a confocal microscope. The diameter of the MscL protein was found to increase by ˚ on channel activation by LPC (Fig. 4B), which is in excellent agree16 A ment with the overall change of the channel diameter estimated by EPR spectroscopy (Perozo et al., 2002b). These key findings in bilayer‐controlled functional properties of MS channels emphasize that the lipid bilayer is much more than a neutral solvent by actively modulating the specificity and fidelity of signaling by membrane proteins (Kung, 2005). A molecular dynamics simulation study by Elmore and Dougherty (2003) reported that MscL protein–lipid interactions were clearly altered by the headgroup changes, leading to conformational diVerences in the C‐terminal region of MscL. The simulations indicated further that hydrophobic matching between MscL and the lipid membrane as well as lipid–protein interactions in general could be more important for proper MscL function and assembly than are protein–protein interactions. This notion has further been supported by another study showing that when hydrophobic residues thought to make contact with the membrane lipid near the periplasmic end of the TM1 or TM2 transmembrane domains of MscL are replaced by hydrophilic residues, MscL apparently loses its mechanosensitivity by becoming unable to open in response to membrane tension. These results suggest that the hydrophobic interaction between the membrane lipid and the periplasmic rim of the MscL funnel is important for the proper function of this channel (Yoshimura et al., 2004). H. Structural Models of Gating in MscL and MscS Despite some discrepancies in details of current models of MscL gating, all models include an iris‐like rotation and tilt of TM helices as a major structural change during opening of MscL (Betanzos et al., 2002; Perozo et al., 2002b; Anishkin et al., 2005). The TM1 helices, which are packed together to form a right‐handed bundle in the MscL pentamer, tilt with

44

Boris Martinac

respect to the membrane plane and cause the channel to flatten. Molecular dynamics simulations indicated also that MscL opening should radically reform its tertiary structure (Gullingsrud et al., 2001). Indeed, the flattening ˚ in of the TM helices leads to opening of a wide channel pore of some 30 A diameter (Cruickshank et al., 1997; Sukharev et al., 2001; Perozo et al., 2002b). This basic model of MscL gating is also consistent with the fact that specific hydrophobic mismatch levels stabilize intermediate conformational states of the channel (Perozo et al., 2002a; Elmore and Dougherty, 2003). Instrumental for the changes in helix–helix packing during the close‐to‐open transition of MscL appears to be the pattern of conserved glycine (Gly) and alanine (Ala) residues near the constriction of the channel pore formed by TM1 helices (Perozo, 2006). In MscS, similar pattern of Gly and Ala residues in the TM3 helix near the constriction point of the channel pore form also a structural motif that is essential for proper channel gating (Fig. 5). The position of the Gly‐Ala pattern on the TM3 helix faces is conserved in the MscS family of proteins (Kloda and Martinac, 2002; Pivetti et al., 2003) underlining the significance of this structural motif for gating of prokaryotic MS channels. The gating model of MscS resembles that of MscL. The channel opening is facilitated by slight iris‐like rotations and tilt of TM3 pore‐lining helices (Edwards et al., 2005). However, the structural changes in MscS are of smaller magnitude compared to that of MscL, which is consistent with approximately three times smaller MscS conductance (Table I). In addition to showing the importance of the Gly‐Ala motif, the study by Edwards et al. (2005) demonstrated also a remarkable level of plasticity that could be tolerated within MscS primary structure without impairing the channel function (Martinac, 2005a). For interested readers, further details of structural features and molecular dynamics of MscL and MscS can be found in several reviews (Perozo and Rees, 2003; Blount et al., 2005; Sukharev et al., 2005; Tajkhorshid et al., 2005; Perozo, 2006).

IV. PHARMACOLOGY OF PROKARYOTIC MS CHANNELS Prokaryotic MS channels can be blocked by submillimolar concentrations of gadolinium (Gd3þ) (Martinac, 2001), which is a common blocker of MS channels in many types of cells (Hamill and McBride, 1996). The channels that were probed by Gd3þ include MscL and MscS of E. coli (Berrier et al., 1992, 1996), MscA1 and MscA2 of Haloferax volcanii (Le Dain et al., 1998), MscMJ and MscMJLR of Methanococcus jannaschii (Kloda, 2001; Kloda and Martinac, 2002), and MscTA of T. acidophilum (Kloda and Martinac, 2001d) (Table I). In contrast to other prokaryotic MS channels, which were blocked by submillimolar concentrations, at least 1 mM of Gd3þ was required


45

to block MscTA. In this context, it is important to mention that Gd3þ does not block prokaryotic channels by aVecting the channel proteins directly but rather by modifying the mechanical properties of the lipid bilayer surrounding the MS channels (Ermakov et al., 1998). This appears to coincide with comparably larger concentrations of Gd3þ required for the MscTA block because a much higher membrane tension corresponding to its unusually high free energy of activation is required for MscTA activation compared to other bacterial and archaeal MS channels (Table I) (Kloda and Martinac, 2002). Besides by Gd3þ MscL was also probed by the spider venom peptide GsMtx‐4, a novel specific inhibitor of stretch‐activated cation‐selective MS channels in vertebrate cells (Suchyna et al., 2000; Bode et al., 2001). The peptide neither could block MscL nor did exert any eVect on its gating (Liu and Martinac, unpublished). Both MscS and MscL of E. coli are activated by amphipaths, such as chlorpromazine (CPZ), trinitrophenol (TNP), local anesthetics and lysolipids (Martinac et al., 1990; Perozo et al., 2002a), which are known activators of prokaryotic and eukaryotic MS channels (Martinac et al., 1990; Hamill and McBride, 1996; Patel et al., 1998; Kloda and Martinac, 2001a; Qi et al., 2005) (Table I). Similarly, the MS channel of the archaeon T. volcanii exhibited an increase in activation by negative pressure in the presence of TNP (Kloda and Martinac, 2001d). MscMJ could also be activated by both CPZ and TNP (Kloda and Martinac, 2001b), whereas MscMJLR was aVected by neither of the two (Kloda and Martinac, 2001c). The eVect that amphipaths exert on prokaryotic MS channels is indirect, since it is caused by diVerential insertion of these compounds into the inner and outer leaflet of the lipid bilayer (Markin and Martinac, 1991; Perozo et al., 2002a). In contrast, parabens, which are alkyl esters of p‐hydroxybenzoic acid and are a class of antimicrobial agents, were shown to open MscL and MscS of E. coli by directly interacting with the gate of these channels (Nguyen et al., 2005). V. FAMILIES OF PROKARYOTIC MS CHANNELS The finding of MS channels in prokaryotes suggests that these membrane proteins were among the first macromolecules that evolved to facilitate transport of solutes in membranes of protocells. The accessibility of a large number of genome sequences of diVerent bacterial and archaeal evolutionary groups available in various data bases has made possible the analysis of phylogenetic distribution of MS channels from these microorganisms. Multiple sequence alignments of homologues of MscL and MscS revealed that they form separate families of prokaryotic MS channels (Kloda and Martinac, 2002; Pivetti et al., 2003). It has been suggested that MscL‐like progenitor molecules might present the prototype of prokaryotic MS genes (Kloda and Martinac, 2002)

46

Boris Martinac

and that the two MS channel families may share a common ancestry. An opposite view based on the lack of statistical evidence for a link between the MscL and MscS families argued that the mscS and mscL genes might have followed separate evolutionary pathways (Okada et al., 2002; Pivetti et al., 2003). Nevertheless, whether an MscL‐like progenitor molecule gave rise to a variety of prokaryotic MS channels remains unclear at present because sequence similarity between the highly conserved pore‐lining helices in the prokaryotic MS channels, that is, TM1 in MscL and TM3 in the YggB subfamily of MscS proteins, seems to suggest an evolutionary link between MscS and MscL families (Kloda and Martinac, 2002; Pivetti et al., 2003). A. MscL Family A group of MscL relatives forms a separate family, which encompasses MS channels of Gram‐negative and Gram‐positive bacteria, as well as those of a single archaeon, Methanosarcina acetivorans, and a fungus, Neurospora crassa (Kloda and Martinac, 2002; Kumańovics et al., 2003; Martinac and Kloda, 2003; Pivetti et al., 2003). In terms of their size and sequence, the archaeal and fungal proteins are the most divergent members of the MscL family. In relation to bacterial MscL homologues, they most closely resemble those of Gram‐positive bacteria (Pivetti et al., 2003). B. MscS Family The MscS channel family is larger than the MscL family. It includes a number of representatives from bacteria, archaea, fission yeast Schizosaccharomyces pombe and plant A. thaliana, but not from animals (Kloda and Martinac, 2002; Martinac and Kloda, 2003; Pivetti et al., 2003). The MscS relatives are more diverse and vary much more in size and sequence than the MscL relatives. Nevertheless, the MscS family is not ubiquitous, since several organisms with fully sequenced genomes, including Gram‐negative chlamydias, Gram‐positive clostridia, mycoplasmas, and ureaplasmas, do not encode recognizable MscS homologues (Pivetti et al., 2003).

VI. EARLY ORIGINS OF MECHANOSENSORY TRANSDUCTION The Earth was formed 4.6 billion years ago and for most of the time since its formation, life on Earth has exclusively consisted of microorganisms (Woese, 1981) (Fig. 7). Given the obvious significance of water for existence of life, the early microbes would have required ‘‘emergency valves’’ for

47

2. MS Channels of Prokaryotes Oxygen atmosphere forming

Earth formed Stabilization of crust

Humans Horses Dinosaurs

Microorganisms

4.5

4.0

3.5

3.0

2.5 2.0 1.5 Billions of years ago

Trilobites

1.0

0.5

0

FIGURE 7 Biological time scale for the planet Earth from the time of the Earth’s formation 4.6 billion years ago to the time of human origin. The oldest microfossils of prokaryotic cells are 3.5 billion years of age (Woese, 1981). Reproduced with permission from Woese (1994).

release of osmotic stress to make their survival possible in environments of varying osmolarity. Hence, diVerent authors suggested that the MS channels could have first evolved as cellular osmoregulators (Sachs, 1988; Kung et al., 1990; Morris, 1990; Martinac, 1993; Kung and Saimi, 1995; Sackin, 1995). A. Physiological Function of MS Channels in Prokaryotic Cells Bacteria possess multiple adaptation mechanisms enabling them to grow in a wide range of external osmolarities (Wood, 1999; Sleator and Hill, 2001). MS channels, which are located in the cytoplasmic membrane of bacterial cells (Berrier et al., 1989; Levina et al., 1999; Norman et al., 2005) (Fig. 8A), participate in the response to excessive turgor pressure caused by hypotonic conditions. The large conductance and lack of ionic specificity allows the MS channels in prokaryotes to function as ‘‘emergency valves’’ for rapid and nonspecific release of solutes (Fig. 8B). As sensors and regulators of the cellular turgor, they provide a safeguard without which the bacterial cells would lyse. This has unambiguously been demonstrated for MscL and MscS of E. coli (Blount et al., 1997; Ou et al., 1997; Levina et al., 1999; Booth et al., 2005). Mutants of E. coli lacking both MscL and MscS die on transfer from a medium of high to a medium of low osmolarity (Booth and Louis, 1999; Levina et al., 1999). The third channel, MscM is insuYcient alone to protect them. Supporting evidence has been provided for marine bacterium Vibrio alginolyticus in which introduction of an mscL gene was found to alleviate cell lysis by hypoosmotic shock (Nakamaru et al., 1999). Bacterial cells lacking only MscS or MscL are, however, fully functional. A multiplicity of MS channels may be required to provide a safeguard against the deleterious eVects that sudden changes in external

48

Boris Martinac

FIGURE 8 (A) Detection of fluorescence and channel activity from MscL channels labeled by green fluorescent protein (GFP) in a giant spheroplast. Confocal image of the giant spheroplast shows that the fluorescence from MscL‐GFP is mostly detected in the membrane area, suggesting that MscL‐GFP is located in the cytoplasmic membrane. Scale bar ¼ 5 mm.


49

osmolarity could have on these microorganisms. Hence, the need for channels operating at diVerent levels of cellular turgor appears to be dictated by the diVerent environmental cues of the living habitats in which prokaryotes exist. Prokaryotic MS channels might also sense changes in turgor pressure during cell division and cell growth given that cell turgor is essential for growth and cell wall synthesis (Csonka and Epstein, 1996). Increase in cell turgor stretches the cellular envelope and causes increase in cell volume, which is required for the synthesis and the assembly of cell wall components resulting in enlargement of the envelope and growth of bacterial cells. Indeed, the expression of MscS and MscL is induced on entry into stationary growth phase when the cells undergo cell wall remodeling and need to relieve the turgor pressure (Stokes et al., 2003). The physiological role of MS channels in archaea has not clearly been established. However, the archaeal MS channels could be expected to have functions similar to those of their bacterial counterparts. Supporting indirect evidence comes from experiments in which expression of the archaeal MscMJ channel in E. coli was shown to impair growth of the bacterium. The growth was partially restored in media of high osmolarity that would cause MscMJ to remain predominantly closed (Kloda and Martinac, 2001b). Although not much is known about turgor pressure in archaea, a partial rescue of E. coli cells expressing MscMJ in media of higher osmolarity seems to suggest that cellular turgor could be higher in E. coli than in the marine Methanococcus jannaschii. Since changes in osmolarity such as ones occurring during flood, drought, or volcanic activity can also be expected to occur in the extreme environments inhabited by archaea, MS channels in these prokaryotic cells may also serve as emergency valves in cellular osmoregulation. B. Function of MscS‐Like Channels in Mechanosensory Transduction in Plants Plants respond to a number of mechanical stimuli including touch and gravity (Blancaflor and Masson et al., 2003; Braam, 2005) that cause rapid changes in proton and calcium concentration in plant cells. MS ion channels Shown below is a patch‐clamp recording of MscL‐GFP (▽) and MscS (*) channel activity in excised patch of a giant spheroplast. Pipette voltage is þ30 mV. Adapted from Norman et al. (2005). (B) MS channels in bacteria are essential to maintain cell integrity. Osmotic stress caused by a hypoosmotic shock opens MscL and MscS channels to release excessive turgor pressure. Normally, cell turgor norm of a bacterial cell is 4–6 atm. Depending on the magnitude of the hypoosmotic shock, the turgor pressure may increase well above 10 atm, which without MS channel opening would cause a cell to lyse. is osmotic pressure diVerence in atm (at 22 C), and C is concentration gradient of solutes in moles per liter (osmolarity).

50

Boris Martinac

that could mediate these rapid responses have indeed been reported in plants (Falke et al., 1988; Cosgrove and Hedrich, 1991; Ding and Pickard, 1993; Qi et al., 2004). However, none of these MS channels have been identified to date. Phylogenetic analysis of distribution of prokaryotic MS channels (Section VI) helped to identify MscS‐related proteins in the experimental plant A. thaliana (Kloda and Martinac, 2002; Pivetti et al., 2003). Out of 10 MscS‐like proteins found in this plant two of them, MSL2 and MSL3, have been characterized (Haswell and Meyerowitz, 2006). Both proteins are localized to the inner membrane of the envelope of plastids, which are plant‐ specific endosymbiotic organelles responsible for photosynthesis, gravity perception, and many metabolic reactions. According to the study by Haswell and Meyerowitz (2006), both MSL2 and MSL3 are involved in control of the plastid size, shape, and possibly division by altering ion fluxes in response to membrane tension occurring during plant morphogenesis. Finding prokaryotic‐type MS channels in plants may not be surprising given that plastids in green plants may have originated directly from a cyanobacterium‐ like prokaryote via primary endosymbiosis (Raven and Allen, 2003; Nozaki, 2005).

VII. CONCLUDING REMARKS This chapter provides a brief overview of an area of MS channel research that over the last 20 years has gone a log way from its beginnings marked by a discovery of MS channels in bacteria. Recent findings and new developments that are briefly outlined have significantly contributed to our understanding of basic principles and evolutionary origins of mechanosensory transduction in living cells. In the future, we may expect further exciting developments of this research area to continue. Acknowledgments I wish to thank Dr. Anna Kloda for useful suggestions and critical reading of the manuscript and Dr. Annette Hurst for helpful comments. Supported by the Australian Research Council.

References Ajouz, B., Berrier, C., Besnard, M., Martinac, B., and Ghazi, A. (2000). Contributions of the diVerent extramembraneous domains of the mechanosensitive ion channel MscL to its response to membrane tension. J. Biol. Chem. 275, 1015–1022. Anishkin, A., and Kung, C. (2005). Microbial mechanosensation. Curr. Opin. Neurobiol. 15(4), 397–405. Anishkin, A., and Sukharev, S. (2004). Water dynamics and dewetting transitions in the small mechanosensitive channel MscS. Biophys. J. 86, 2883–2895.


51

Anishkin, A., Gendel, V., Sharifi, N. A., Chiang, C.‐S., Shirinian, L., Guy, H. R., and Sukharev, S. (2003). On the conformation of the COOH‐terminal domain of the large mechanosensitive channel MscL. J. Gen. Physiol. 121(3), 227–244. Anishkin, A., Chiang, C.‐S., and Sukharev, S. (2005). Gain‐of‐function mutations reveal expanded intermediate states and a sequential action of two gates in MscL. J. Gen. Physiol. 2005 125, 155–170. Barinaga, M. (1994). Molecular evolution. Archaea and eukaryotes grow closer. Science 264, 1251. Bass, R. B., Strop, P., Barclay, M., and Rees, D. (2002). Crystal structure of Escherichia coli MscS, a voltage‐modulated and mechanosensitive channel. Science 298, 1582–1587. Berrier, C., Coulombe, A., Houssin, C., and Ghazi, A. (1989). A patch‐clamp study of inner and outer membranes and of contact zones of E. coli, fused into giant liposomes: Pressure‐ activated channels are localized in the inner membrane. FEBS Lett. 259, 27–32. Berrier, C., Coulombe, A., Szabo, I., Zoratti, M., and Ghazi, A. (1992). Gadolinium ion inhibits loss of metabolites induced by osmotic shock and large stretch‐activated channels in bacteria. Eur. J. Biochem. 206, 559–565. Berrier, C., Besnard, M., Ajouz, B., Coulombe, A., and Ghazi, A. (1996). Multiple mechano‐ sensitive ion channels from Escherichia coli, activated at diVerent thresholds of applied pressure. J. Membr. Biol. 151, 175–187. Betanzos, M., Chiang, C.‐S., Guy, H. R., and Sukharev, S. (2002). A large iris‐like expansion of a mechanosensitive channel protein induced by membrane tension. Nat. Struct. Biol. 9(9), 704–710. Bezanilla, F., and Perozo, E. (2002). Force and voltage sensors in one structure. Science 298, 1562–1563. Biggin, P. C., and Sansom, M. S. P. (2001). Channel gating: Twist to open. Curr. Biol. 11(9), R364–R366. Blancaflor, E. B., and Masson, P. H. (2003). Plant gravitropism. Unravelling the ups and downs of a complex process. Plant Physiol. 133, 1677–1690. Blount, P., Sukharev, S. I., Schroeder, M. J., Nagle, S. K., and Kung, C. (1996). Single residue substitutions that change the gating properties of a mechanosensitive channel preotein of Escherichia coli. Proc. Natl. Acad. Sci. USA 93, 11652–11657. Blount, P., Schroeder, M., and Kung, C. (1997). Mutations in a bacterial mechano‐sensitive channel change the cellular response to osmotic stress. J. Biol. Chem. 272, 32150–32157. Blount, P., Sukharev, S. I., Moe, P., Martinac, B., and Kung, C. (1999). Mechanosensitive channels in bacteria. Methods Enzymol. 294, 458–482. Blount, P., Iscla, I., Li, Y., and Moe, P. C. (2005). The bacterial mechanosensitive channel MscS and its extended family. In ‘‘Bacterial Ion Channels and Their Eukaryotic Homologs’’ (A. Kubalski and B. Martinac, eds.), pp. 247–258. ASM Press, Washington DC. Bode, F., Sachs, F., and Franz, M. R. (2001). Tarantula peptide inhibits atrial fibrillation during stretch. Nature 409, 35–36. Booth, I. R., and Louis, P. (1999). Managing hypoosmotic stress: Aquaporins and mechanosensitive channels in Escherichia coli. Curr. Opin. Microbiol. 2, 166–169. Booth, I. R., Edwards, M. D., Murray, E., and Miller, S. (2005). The role of bacterial channels in cell physiology. In ‘‘Bacterial Ion Channels and Their Eukaryotic Homologs’’ (A. Kubalski and B. Martinac, eds.), pp. 291–312. ASM Press, Washington DC. Braam, J. (2005). In touch: Plant responses to mechanical stimuli. New Phytol. 165, 373–389. Chang, G., Spencer, R., Lee, A., Barclay, M., and Rees, C. (1998). Structure of the MscL homologue from Mycobacterium tuberculosis: A gated mechano‐sensitive ion channel. Science 282, 2220–2226.

52

Boris Martinac

Csonka, L. N., and Epstein, W. (1996). Osmoregulation. In ‘‘Escherichia coli and Salmonella: Cellular and Molecular Biology’’ (F. C. Neidhardt, R. Curtiss, III, J. L. Ingraham, E. C. C. Lin, K. Brooks Low, B. Magasanik, W. S. ReznikoV, M. Riley, M. Schaechter, and H. E. Umbarger, eds.), 2nd edn., pp. 1210–1225. ASM Press, Washington DC. Chothia, C. (1974). Hydrophobic bonding and accessible surface area in proteins. Nature 248, 338–339. Colombo, G., Marrink, S. J., and Mark, A. E. (2003). Simulation of MscL gating in a bilayer under stress. Biophys. J. 84, 233–237. Corry, B., Rigby, P., and Martinac, B. (2005). Conformational changes involved in MscL channel gating measured using FRET spectroscopy. Biophys. J. 89, L49–L51. Cosgrove, D. J., and Hedrich, R. (1991). Stretch‐activated chloride, potassium, and calcium channels coexisting in plasma membranes of guard cells of Vicia faba L. Planta 186, 143–153. Cruickshank, C. C., Minchin, R., Le Dain, A., and Martinac, B. (1997). Estimation of the pore size of the large‐conductance mechanosensitive ion channel of Escherichia coli. Biophys. J. 73, 1925–1931. Cui, C., Smith, D. O., and Adler, J. (1995). Characterization of mechanosensitive channels in Escherichia coli cytoplasmic membranes by whole cell patch clamp recording. J. Membr. Biol. 144, 31–42. Delcour, A. H., Martinac, B., Adler, J., and Kung, C. (1989). Modified reconstitution method used in patch‐clamp studies of Escherichia coli ion channels. Biophys. J. 56, 631–635. Deuticke, B. (1968). Transformation and restoration of biconcave shape of human erythrocytes induced by amphiphilic agents and changes of ionic environment. Biochim. Biophys. Acta 163(4), 494–500. Doyle, D. A., Morais Cabral, J., Pfuetzner, R. A., Kuo, A., Gulbis, J. M., and Cohen, S. L. Chait, B. T., and MacKinnon, R. (1998). The structure of the potassium channel: Molecular basis of K þ conduction and selectivity. Science 280, 69–77. Ding, J. P., and Pickard, B. G. (1993). Modulation of mechanosensitive calcium‐selective cation channels by temperature. Plant J. 3, 713–720. Dutzler, R., Campbell, E. B., and MacKinnon, R. (2002). Gating the selectivity filter in ClC chloride channels. Science 300, 108–112. Edwards, M. D., Li, Y., Kim, S., Miller, S., Bartlett, W., Black, S., Dennison, S., Isla, I., Blount, P., Bowie, J., and Booth, I. R. (2005). Pivotal role of the glycine‐rich TM3 helix in gating the MscS mechanosensitive channel. Nat. Struct. Mol. Biol. 12(2), 113–119. Elliot, J. R., Needham, D., Dilger, J. P., and Haydon, D. A. (1983). The effects of bilayer thickness and tension on gramicidin single‐channel lifetime. Biochim. Biophys. Acta 735, 95. Elmore, D. E., and Dougherty, D. A. (2001). Molecular dynamics simulations of wild‐type and mutant forms of the Mycobacterium tuberculosis MscL channel. Biophys. J. 81, 1345–1359. Elmore, D. E., and Dougherty, D. A. (2003). Investigating lipid composition effects on the mechanosensitive channel of large conductance (MscL) using molecular dynamics simulations. Biophys. J. 85(3), 1512–1524. Ermakov, A., Averbach, A. Z., Arbuzova, A. B., and Sukharev, S. I. (1998). Lipids and cell membranes in the presence of gadolinium and other ions with high aYnity to lipids. 2. A dipole component of the boundary potential on membranes with diVerent surface charge. Membr. Cell Biol. 12, 411–426. Falke, L. C., Edwards, K. L. Pickard, B. G., and Misler, S. (1988). A stretch‐activated anion channel in tobacco protoplasts. FEBS Lett. 237, 141–144.


53

Gullingsrud, J., and Schulten, K. (2003). Gating of MscL studied by steered molecular dynamics. Biophys. J. 85, 2087–2099. Gullingsrud, J., Kosztin, D., and Schulten, K. (2001). Structural determinants of MscL gating studied by molecular dynamics simulations. Biophys. J. 80(5), 2074–2081. Gustin, M. C., Zhou, X.‐L., Martinac, B., and Kung, C. (1988). A mechanosensitive ion channel in the yeast plasma membrane. Science 242, 762–765. Ha¨se, C. C., Le Dain, A. C., and Martinac, B. (1995). Purification and functional reconstitution of the recombinant large mechanosensitive ion channel (MscL) of Escherichia coli. J. Biol. Chem. 270, 18329–18334. Ha¨se, C. C., Le Dain, A. C., and Martinac, B. (1997). Molecular dissection of the large mechanosensitive ion channel (MscL) of Escherichia coli: Mutants with altered channel gating and pressure sensitivity. J. Membr. Biol. 157, 17–25. Hamill, O. P., and McBride, D. W., Jr. (1996). The pharmacology of mechanogated membrane ion channels. Pharmacol. Rev. 48, 231–252. Hamill, O. P., and McBride, D. W., Jr. (1997). Induced membrane hypo/hyper‐ mechanosensitivity: A limitation of patch‐clamp recording. Annu. Rev. Physiol. 59, 621–631. Hamill, O. P., and Martinac, B. (2001). Molecular basis of mechanotransduction in living cells. Physiol. Rev. 81, 685–740. Hamill, O. P., Marty, A., Neher, E., Sackmann, B., and Sigworth, F. J. (1981). Improved patch‐ clamp techniques for high‐resolution current recording from cells and cell‐free membrane patches. Pflu¨gers Arch. 391, 85–100. Harms, G. S., Orr, G., Montal, M., Thrall, B. D., Colson, D., and Lu, H. P. (2003). Combined patch‐clamp and single‐molecule imaging: Microscopy study of single‐molecule ion channel dynamics. Biophys. J. 85, 1826. Haswell, E. S., and Meyerowitz, E. M. (2006). MscS‐like proteins control plastid size and shape in Arabidopsis thaliana. Curr. Biol. 16, 1–11. Iscla, I., Levin, G., Wray, R., Reynolds, R., and Blount, P. (2004). Defining the physical gate of a mechanosensitive channel, MscL, by engineering metal‐binding sites. Biophys. J. 87(5), 3172–3180. Jiang, Y., Lee, A., Chen, J., Cadene, M., Chait, B. T., and MacKinnon, R. (2002). Crystal structure and mechanism of a calcium‐gated potassium channel. Nature 417, 515–522. Kloda, A. (2001). Mechanosensitive channels in prokaryotes. Molecular cloning and characterization of three types of archaeal MS channels. Ph.D. thesis, The University of Western Australia. Kloda, A., and Martinac, B. (2001a). Mechanosensitive channels in Archaea. Cell Biochem. Biophys. 34, 349–381. Kloda, A., and Martinac, B. (2001b). Molecular identification of a mechanosensitive ion channel in Archaea. Biophys. J. 80, 229–240. Kloda, A., and Martinac, B. (2001c). Structural and functional similarities and diVerences between MscMJLR and MscMJ, two homologous MS channels of M. jannashii. EMBO J. 20, 1888–1896. Kloda, A., and Martinac, B. (2001d). Mechanosensitive channel in Thermoplasma a cell wall‐ less Archaea: Cloning and molecular characterization. Cell Biochem. Biophys. 34, 321–347. Kloda, A., and Martinac, B. (2002). Mechanosensitive channels of bacteria and archaea share a common ancestral origin. Eur. Biophys. J. 31, 14–25. Kloda, A., and Martinac, B. (2006). C‐terminal charged cluster of MscL, RKKEE, functions as a pH sensor. Biophys. J. 90, 1992–1998. Koprowski, P., and Kubalski, A. (1998). Voltage‐independent adaptation of mechanosensitive channels in Escherichia coli protoplasts. J. Membr. Biol. 164, 253–262.

54

Boris Martinac

Kubalski, A., and Martinac, B. (2005). ‘‘Bacterial Ion Channels and Their Eukaryotic Homologues.’’ ASM Press, Washington DC. Kumańovics, A., Levin, G., and Blount, P. (2003). Family ties of gated pores: Evolution of the sensor module. FASEB J. 16, 1623–1629. Kung, C. (2005). A possible unifying principle for mechanosensation. Nature 436, 647–654. Kung, C., and Saimi, Y. (1995). Solute sensing vs. solvent sensing, a speculation. J. Eukaryot. Microbiol. 42, 199–200, 1995. Kung, C., Saimi, Y., and Martinac, B. (1990). Mechanosensitive ion channels in microbes and the early evolutionary origin of solvent sensing. In ‘‘Current Topics in Membranes and Transport’’ (T. Claudio, ed.), Chapter 36, pp. 9451–9455. Academic Press, New York. Le Dain, A. C., Saint, N., Kloda, A., Ghazi, A., and Martinac, B. (1998). Mechanosensitive ion channels of the archaeon Haloferax volcanii. J. Biol. Chem. 273, 12116–12119. Levina, N., Totemeyer, S., Stokes, N. R., Louis, P., Jones, M. A., and Booth, I. R. (1999). Protection of Escherichia coli cells against extreme turgor by activation of MscS and MscL mechanosensitive channels: Identification of genes required for MscS activity. EMBO J. 18, 1730–1737. Li, Y., Moe, P. C., Chandrasekaran, S., Booth, I. R., and Blount, P. (2002). Ionic regulation of MscK, a mechanosensitive channel from Escherichia coli. EMBO J. 21, 5323–5330. McLaggan, D., Jones, M. A., Gouesbet, G., Levina, N., Lindey, S., Epstein, W., and Booth, I. R. (2002). Analysis of the kefA2 mutation suggests that KefA is a cation‐specific channel involved in osmotic adaptation in Escherichia coli. Mol. Microbiol. 43, 521–536. Markin, V. S., and Martinac, B. (1991). Mechanosensitive ion channels as reporters of bilayer expansion. A theoretical model. Biophys. J. 60, 1120–1127. Maroto, R., Raso, A., Wood, T. G., Kurosky, A., Martinac, B., and Hamill, O. P. (2005). TRPC1 forms the stretch‐activated cation channel in vertebrate cells. Nat. Cell Biol. 7(2), 179–185. Martinac, B. (1993). In ‘‘Thermodynamics of Membrane Receptors and Channels’’ (M. B. Jackson, ed.), pp. 327–351. CRC Press, Boca Raton FL. Martinac, B. (2001). Mechanosensitive channels in prokaryotes. Cell. Physiol. Biochem. 11, 61–76. Martinac, B. (2004). Mechanosensitive ion channels: Molecules of mechano‐transduction. J. Cell Sci. 117, 2449–2460. Martinac, B. (2005a). Structural plasticity in MS channels. Nat. Struct. Mol. Biol. 12(2), 104–105. Martinac, B. (2005b). Force from lipids: Physical principles of gating mechano‐sensitive channels by mechanical force revealed by chemical manipulation of cellular membranes. Chem. Educ. 10(2), 107–114. Martinac, B., and Hamill, O. P. (2002). Gramicidin A channels can switch between stretch‐ activation and stretch‐inactivation depending upon bilayer thickness. Proc. Natl. Acad. Sci. USA 99, 4308. Martinac, B., and Kloda, A. (2003). Evolutionary origins of mechanosensitive ion channels. Progress Biophys. Mol. Biol. 82, 11–24. Martinac, B., Buechner, M., Delcour, A. H., Adler, J., and Kung, C. (1987). Pressure‐sensitive ion channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 84, 2297–2301. Martinac, B., Adler, J., and Kung, C. (1990). Mechanosensitive ion channels of E. coli activated by amphipaths. Nature 348, 261–263. Martinac, B., Delcour, A. H., Buechner, M., Adler, J., and Kung, C. (1992). Mechanosensitive ion channels in bacteria. In ‘‘Comparative Aspects of Mechanoreceptor Systems’’ (F. Ito, ed.), pp. 3–18. Springer Verlag, Berlin.


55

Meyer, G. R., Gullingsrud, J., Schulten, K., and Martinac, B. (2006). Molecular dynamics study of MscL interactions with a curved lipid bilayer. Biophys. J. 91, 1630–1637. Moe, P. C., Blount, P., and Kung, C. (1998). Functional and structural conservation in the mechanosensitive channel MscL implicates elements crucial for mechanosensation. Mol. Microbiol. 28, 583–592. Moe, P. C., Levin, G., and Blount, P. (2000). Correlating a protein structure with function of a bacterial mechanosensitive channel. J. Biol. Chem. 40, 31121–31127. Morris, C. E. (1990). Mechanosensitive ion channels. J. Membr. Biol. 113, 93–107. Nakamaru, Y., Takahashi, Y., Unemoto, T., and Nakamura, T. (1999). Mechanosensitive channel functions to alleviate the cell lysis of marine bacterium, Vibrio alginolyticus, by osmotic downshock. FEBS Lett. 444(2–3), 170–172. Nguyen, T., Clare, B., Guo, W., and Martinac, B. (2005). The eVects of parabens on the mechanosensitive channels of E. coli. Eur. Biophys. J. 34, 389–396. Norman, C., Liu, Z.‐W., Rigby, P., Raso, A., Petrov, Y., and Martinac, B. (2005). Visualisation of the mechanosensitive channel of large conductance in bacteria using confocal microscopy. Eur. Biophys. J. 34, 396–403. Nozaki, H. (2005). A new scenario of plastid evolution: Plastid primary endosymbiosis before the divergence of the ‘‘Plantae,’’ emended. J. Plant Res. 118, 247–255. Oakley, A. J., Martinac, B., and Wilce, M. C. J. (1999). Structure and function of the bacterial mechanosensitive channel of large conductance. Protein Sci. 8, 1915–1921. O’Connell, A. M., Koeppe, R. E., and Andersen, O. S. (1990). Kinetics of gramicidin channel formation in lipid bilayers: Transmembrane monomer association. Science 250, 1256. Okada, K., Moe, P. C., and Blount, P. (2002). Functional design of bacterial mechanosensitive channels. J. Biol. Chem. 277, 27682–27688. Ou, X., Blount, P., HoVman, R., Kusano, A., and Kung, C. (1997). Random mutagenesis of a mechanosensitive channel identifies regions of the protein crucial for normal function (Abstract). Biophys. J. 72, A139. Pace, N. R. (1997). A molecular view of microbial diversity and the biosphere. Science 276, 734–740. Patel, A., Honore´, E., Maingret, F., Lesage, F., Fink, M., Duprat, F., and Lazdunski, M. (1998). A mammalian two pore domain mechano‐gated S‐like K þ channel. EMBO J. 17, 4283–4290. Patel, A. J., Honore, E., and Lazdunski, M. (2001). Lipid and mechano‐gated 2P domain Kþ channels. Curr. Opin. Cell Biol. 13, 422–428. Perozo, E. (2006). Gating prokaryotic mechanosensitive channels. Nat. Rev. Mol. Cell. Biol. 7, 109–119. Perozo, E., and Rees, D. C. (2003). Structure and mechanism in prokaryotic mechanosensitive channels. Curr. Opin. Struct. Biol. 13, 432–442. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2001). Site‐directed spin‐labeling analysis of reconstituted MscL in the closed state. J. Gen. Physiol. 118(2), 193–206. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2002a). Physical principles underlying the transduction of bilayer deformation forces during mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703. Perozo, E., Cortes, D. M., Sompornpisut, P., Kloda, A., and Martinac, B. (2002b). Structure of MscL in the open state and the molecular mechanism of gating in mechanosensitive channels. Nature 418, 942–948. Pivetti, C. D., Yen, M. R., Miller, S., Busch, W., Tseng, Y. H., Booth, I. R., and Saier, M. H., Jr. (2003). Two families of mechanosensitive channel proteins. Microbiol. Mol. Biol. Rev. 67, 66–85.

56

Boris Martinac

Qi, Z., Kishigami, A., Nakagawa, Y., Iida, H., and Sokabe, M. (2004). A mechanosensitive anion channel in Arabidopsis thaliana mesophyll cells. Plant Cell Physiol. 45, 1704–1708. Qi, Z., Chi, S., Naruse, K., and Sokabe, M. (2005). Activation of a mechanosensitive BK channel by membrane stress created with amphipaths. Mol. Membr. Biol. 22, 519–527. Raven, J. A., and Allen, J. F. (2003). Genomics and chloroplast evolution: What did cyanobacteria do for plants. Genome Biol. 4(3), 209. Ruthe, H. J., and Adler, J. (1985). Fusion of bacterial spheroplasts by electric fields. Biochim. Biophys. Acta 819(1), 105–113. Sachs, F. (1988). Mechanical Transduction in biological systems. CRC Crit. Rev. Biomed. Eng. 16, 141–169. Sachs, F., and Morris, C. E. (1998). Mechanosensitive ion channels in nonspecialized cells. Rev. Physiol. Biochem. Pharmacol. 132, 1–77. Sackin, H. (1995). Mechanosensitive channels. Annu. Rev. Physiol. 57, 333–353. Sheetz, M. P., and Singer, S. J. (1974). Biological membranes as bilayer couples. A molecular mechanism of drug‐erythrocyte interactions. Proc. Natl. Acad. Sci. USA 71(11), 4457–4461. Sleator, R. D., and Hill, C. (2001). Bacterial osmoadaptation: The role of osmolytes in bacterial stress and virulence. FEMS Microbiol. Rev. 26, 49–71. Sokabe, M., and Sachs, F. (1990). The structure and dynamics of patch clamped membrane: A study using diVerential interference contrast microscopy. J. Cell Biol. 111, 599–606. Sotomayor, M., and Schulten, K. (2004). Molecular dynamics study of gating in the mechanosensitive channel of small conductance MscS. Biophys. J. 87, 3050–3065. Spronk, S. A., Elmore, D. E., and Dougherty, D. A. (2006). Voltage‐dependent hydration and conduction properties of the hydrophobic pore of the mechanosensitive channel of small conductance. Biophys. J. 90, 3555–3569. Stokes, N. R., Murray, H. D., Subramaniam, C., Gourse, R. L., Louis, P., Bartlett, W., Miller, S., and Booth, I. R. (2003). A role for mechanosensitive channels in survival of stationary phase: Regulation of channels expression by RpoS. Proc. Natl. Acad. Sci. USA 100, 15959–15964. Strop, P., Bass, R., and Rees, D. C. (2003). Prokaryotic mechanosensitive channels. In ‘‘Advances In Protein Chemistry: Membrane Proteins’’ (D. C. Rees, ed.), Vol. 63, pp. 177–209. Academic Press, New York. Suchyna, T. M., Johnson, J. H., Hamer, K., Leykam, J. F., Gage, D. A., Clemo, H. F., Baumgarten, C. M., and Sachs, F. (2000). Identification of a peptide toxin from Grammostola spatulata spider venom that blocks cation‐selective stretch‐activated channels. J. Gen. Physiol. 115, 583–598. Sukharev, S. (2002). Purification of the small mechanosensitive channel of Escherichia coli (MscS): The subunit structure, conduction, and gating characteristics in liposomes. Biophys. J. 83, 290–298. Sukharev, S. I., and Corey, D. (2004). Mechanosensitive channels: Multiplicity of families and gating paradigms. Science’s STKE www.stke.org/cgi/content/full/sigtrans; 2004/219/re4, pp. 1–24. Sukharev, S. I., Martinac, B., Arshavsky, V. Y., and Kung, C. (1993). Two types of mechanosensitive channels in the E. coli cell envelope: Solubilization and functional reconstitution. Biophys. J. 65, 177–183. Sukharev, S. I., Blount, P., Martinac, B., Blattner, F. R., and Kung, C. (1994a). A large mechanosensitive channel in E. coli encoded by mscL alone. Nature 368, 265–268. Sukharev, S. I., Blount, P., Martinac, B., and Kung, C. (1994b). Functional reconstitution as an assay for biochemical isolation of channel proteins: Application to the molecular


57

identification of a bacterial mechanosensitive channel. Methods: Companion Methods Enzymol. 6, 51–59. Sukharev, S. I., Blount, P., Martinac, B., and Kung, C. (1997). Mechanosensitive channels of Escherichia coli: The MscL gene, protein, and activities. Ann. Rev. Physiol. 59, 633–657. Sukharev, S. I., Sigurdson, W. J., Kung, C., and Sachs, F. (1999). Energetic and spatial parameters for gating of the bacterial large conductance mechanosensitive channel, MscL. J. Gen. Physiol. 113, 525–539. Sukharev, S., Betanzos, M., Chiang, C. S., and Guy, H. R. (2001). The gating mechanism of the large mechanosensitive channel MscL. Nature 409, 720–724. Sukharev, S., Anishkin, A., Chiang, C.‐S., Betanzos, M., and Guy, H. R. (2005). MscL, a bacterial mechanosensitive channel. In ‘‘Bacterial Ion Channels and Their Eukaryotic Homologs’’ (A. Kubalski and B. Martinac, eds.), pp. 259–290. ASM Press, Washington DC. Tajkhorshid, E., Cohen, J., Aksimentiev, A., Sotomayor, M., and Schulten, K. (2005). Towards and understanding of membrane channels. In ‘‘Bacterial Ion Channels and their Eukaryotic Homologs’’ (A. Kubalski and B. Martinac, eds.), pp. 153–190. ASM Press, Washigton DC. Woese, C. R. (1981). Archaeobacteria. Sci. Am. 244(6), 94–106. Woese, C. R. (1994). There must be a prokaryote somewhere: Microbiology’s search for itself. Microbiol. Rev. 58(1), 1–9. Wood, J. M. (1999). Osmosensing by bacteria: Signals and membrane‐based sensors. Microbiol. Mol. Biol. Rev. 63, 230–262. Yoshimura, K., Batiza, A., Schroeder, M., Blount, P., and Kung, C. (1999). Hydrophilicity of a single residue within MscL correlates with increased mechanosensitivity. Biophys. J. 77, 720–1972. Yoshimura, K., Nomura, T., and Sokabe, M. (2004). Loss‐of‐function mutations at the rim of the funnel of mechanosensitive channel MscL. Biophys. J. 86(4), 2113–2120. Zoratti, M., and Ghazi, A. (1993). Stretch‐activated channels in prokaryotes. In ‘‘Alkali Transport Systems in Prokaryotes’’ (E. P. Bakker, ed.), pp. 349–358. Boca Raton, CRC Press.

CHAPTER 3 Activation of Mechanosensitive Ion Channels by Forces Transmitted Through Integrins and the Cytoskeleton Benjamin D. Matthews,*,{ Charles K. Thodeti,* and Donald E. Ingber* *Vascular Biology Program, Departments of Pathology and Surgery, Children’s Hospital and Harvard Medical School, Boston, Massachusetts 02115 { Department of Pediatrics, Massachusetts General Hospital, Harvard Medical School, Boston, Massachusetts 02114

I. II. III. IV. V. VI. VII.

Overview Introduction Conventional Views of MS Channel Gating Tensegrity‐Based Cellular Mechanotransduction Force Transmission Through Integrins in Living Cells Potential Linkages Between Integrins and MS Ion Channels Conclusions and Future Implications References

I. OVERVIEW Mechanosensitive (MS) ion channels play a central role in the process of cellular mechanotransduction by which living cells convert mechanical signals into chemical and electrical responses. Current views of the mechanism of MS channel gating focus almost entirely on local modulation by plasma membrane tension or by ‘‘gating springs’’ within the underlying submembranous cytoskeleton (CSK). However, cells within many solid tissues commonly experience mechanical stresses that are transmitted over extracellular matrix (ECM) scaVolds to specific transmembrane integrin receptors. Integrins physically anchor the ECM to both the submembranous CSK and the deeper CSK (i.e., the microfilament–microtubule–intermediate filament lattice), and thus, they also constantly experience forces that are generated Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58003-2

60

Matthews et al.

within cytoskeletal contractile actomyosin filaments and exerted on these same adhesion sites. We have previously proposed that cells use tensegrity architecture to mechanically stabilize their shape by maintaining prestress in the interconnected ECM–integrin–CSK lattice and that the activity of certain MS channels may be modulated through tension transfer inside the cell via this tensed structural network, rather than through direct lipid bilayer distortion. Here, we review this tensegrity‐based mechanism for MS channel regulation in light of recent work which confirms that integrins provide a specific path for stress‐dependent activation of MS channels. We also discuss potential molecular mechanisms that might mediate this tensegrity‐based mechanotransduction mechanism for both short‐ and long‐range force transfer through living cells.

II. INTRODUCTION Cellular mechanotransduction—the process by which cells convert mechanical signals into changes in intracellular biochemistry—plays a central role in tissue development, as well as in many disease processes (Ingber, 2003a, 2006). Cells within all living tissues encounter mechanical forces continuously within a changing dynamic environment, and they have evolved an exquisite mechanosensory system to perceive and respond to these forces (Ingber, 1997b, 2006). Cells first sense mechanical stresses, like other external signals, when they impinge on the cell surface. Early work on the mechanism of mechanosensation revealed that virtually all cell types express MS ion channels on their plasma membranes that alter their activity (i.e., either become activated or deactivated) when mechanically stressed (Hamill and McBride, 1997; Sukharev and Corey, 2004). In fact, the earliest recorded response to force application is a change in electrical activity which results from opening of MS channels and occurs within milliseconds (Sachs, 1992). Due to their fast response, MS channels mediate specialized sensory functions such as hearing, touch, and vestibular function (Hamill and Martinac, 2001; Sukharev and Anishkin, 2004; Sukharev and Corey, 2004). However, MS channels are also present in nonsensory cells such as endothelium, smooth muscle, and heart cells. In these parenchymal cells, MS channels regulate a number of biochemical and physiological responses, including strain‐induced endothelial cell orientation, activation of protein kinases, and secretion of inflammatory mediators (Naruse et al., 1998a,b,c). Initially, it was assumed that forces applied to the cell surface activate MS channels as a result of localized distortion of the plasma membrane which results in increased tension in the lipid bilayer that is transmitted directly to

3. Integrins and Mechanosensitive Channels

61

the channel molecule (Fig. 1A) (Martinac et al., 1990; Hamill and McBride, 1993, 1994; Sukharev et al., 1994; Hase et al., 1995; Hamill and Martinac, 2001; Kloda and Martinac, 2001a,b,c; Martinac, 2001; Perozo and Rees, 2003). This appears to be true for certain MS channels; however, many others appear to require interactions with elements of the submembranous CSK for their activation or regulation (Corey and Hudspeth, 1983; Guharay and Sachs, 1984; Hamill and McBride, 1996, 1997; Gillespie and Walker, 2001; Cho et al., 2002). In general, this is still thought to be a local phenomenon that results from membrane distortion‐dependent alteration of the underlying submembranous CSK that then indirectly triggers MS channel activity via internal CSK‐associated ‘‘gating domains’’ (Fig. 1B). Meanwhile, separate studies on how cells sense and respond to mechanical stresses transmitted through ECM have revealed that these forces are preferentially transferred into cells via transmembrane integrin receptors

FIGURE 1 The bilayer (A) and tethered (B) models that have been proposed to explain the mechanosensitivity of MS channel gating. (A) Diagrammatic representation of an MS ion channel that alters its conformation and changes its opening and closing rates when the membrane bilayer distorts, thereby exerting tensional forces (T; arrow indicates direction) directly on the channel molecule. Lipid bilayer distortion may be produced by surface shear forces or potentially by cytoskeletal distortion of transmembrane integral membrane molecules that tightly associate with the lipid bilayer (black ovals). (B) In the tethered model, an MS ion channel experiences tensional forces that are transmitted directly from the internal CSK. These forces stimulate ion flux by tugging on the cytoplasmic portion of the channel that acts as a ‘‘gating spring’’ and opens the pore when tensed. Adapted from Ingber (2006).

62

Matthews et al.

that cluster together within specialized multimolecular anchoring complexes, known as focal adhesions (Wang et al., 1993). Integrins and tightly associated focal adhesion molecules, such as talin, vinculin, and paxillin, also mediate outward transfer of cell tensional forces from the contractile actin CSK to external ECM adhesions on the outside surface of the cell (Geiger et al., 2001). Thus, integrins are now viewed as bidirectional mechanoreceptors; however, they are not signaling receptors themselves, and thus they must partner with other transduction molecules within the focal adhesion to mediate mechanochemical conversion (Ingber, 1991). Integrins mechanically couple the ECM to the deep internal CSK (microfilament–microtubule–intermediate filament lattice) and interconnected nuclear scaVolds by forming a macromolecular complex with underlying focal adhesion proteins. Because integrins are anchored to ECM that is relatively rigid compared to the cell, they are able to resist cell‐generated traction forces and thereby maintain the cell in a state of isometric tension (i.e., tensionally ‘‘prestress’’ the cell and CSK) (Ingber, 1997b, 2003b, 2006; Stamenovic and Coughlin, 1999; Wang et al., 2001). On the basis of these and other observations, we previously proposed that living cells are organized as ‘‘tensegrity’’ structures that gain their shape stability by maintaining tensional prestress within an interlinked network of opposing tension and compression elements (Fig. 2A) (Ingber and Jamieson, 1985; Ingber, 1993, 2003b). In this type of structural system, stresses applied locally to key surface anchoring molecules result in force focusing and channeling (potentially over long distances) through the discrete network linkages that connect the elements that comprise the structure. In addition, network shape stability and the eYciency of force transfer through the lattice may be modulated by altering the level of tensional prestress in the system. Hierarchical tensegrities built from multiple tensegrity modules connected by similar rules also can be constructed; these exhibit coordinated behavior between part and whole similar to that observed in all living structures (Ingber, 1997b, 2003b). Importantly, experimental work has confirmed that many diVerent types of cells use tensegrity to stabilize their shape (Wang et al., 1993; Ingber, 1997b, 2003b, 2006; Komulainen et al., 1998; Ralphs et al., 2002; Brangwynne et al., 2006; Kumar et al., 2006), and that forces transmitted over integrins are preferentially channeled throughout cytoskeletal filaments in the cytoplasm, resulting in stress concentrations at distant sites (e.g., on the nucleus and membrane at the opposite pole of the cell) (Maniotis et al., 1997; Wang et al., 2001; Hu et al., 2003, 2004, 2005). The dependence of cell, tissue and organ mechanics, as well as many biological responses, on cytoskeletal prestress also has been confirmed experimentally (Ingber, 2003b, 2006). Most importantly in the present context, the tensegrity model suggests that certain MS channels might be activated by external forces that are

63

3. Integrins and Mechanosensitive Channels A

Com pr

essio n

C

Actin protofilament sion

Ten Tension

B

Tension ressio n

Te ns ion

Comp

ECM

MT MFs IFs

on nsi Te Tensegrity Force Balance

Spectrin Compression

Incompressible lipid bilayer

Compression

FIGURE 2 Tensegrity architecture and its use by living cells at diVerent size scales in the CSK. (A) A photograph of a tensegrity structure with isolated compression struts (aluminum bars) and tension elements (metal wires) labeled to visualize the tensegrity force balance based on local compression and continuous tension which prestresses (and thereby mechanically stabilizes) the entire structural network. (B) A schematic diagram of the complementary force balance between microfilaments (MFs) and intermediate filaments (IFs) that transmit tension to compressed microtubules (MTs) and relatively noncompressible regions of the underlying ECM that balance these forces, and thereby establish a tensegrity force balance in the whole CSK (the submembranous CSK is not shown in this view). A, B (Ingber, 2003b), reproduced with permission from ‘‘The Company of Biologists.’’ (C) A tensegrity force balance is established in the submembranous CSK at a smaller size scale as a result of relatively rigid actin protofilaments and noncompressible regions of the lipid bilayer resisting tensional distortion of flexible spectrin cables. Based on work of Sung and Vera (2003).

transmitted across integrins and associated cytoskeletal molecules within focal adhesions or the deeper CSK, rather than through direct lipid bilayer distortion alone (Ingber, 1997b). In this chapter, we describe this potential mechanism for integrin‐dependent activation of MS channels in greater detail, discuss potential molecular mediators of this response, and review recent advances in this area. III. CONVENTIONAL VIEWS OF MS CHANNEL GATING MS channels range from simple two transmembrane‐spanning domain‐ containing proteins in bacteria to large multiprotein complexes in higher organisms (Hamill and Martinac, 2001). The MS channels that have been

64

Matthews et al.

identified and cloned in prokaryotes and eukaryotes are divided into diVerent families, including DEG/ENaC (degenerins/epithelial sodium), TRP (transient receptor potential), KþCa, Kir (inward rectifier potassium), two pore Kþ, MscS (small conductance MS), MscL (large conductance MS), and archeal MS channels (Hamill and Martinac, 2001; Sukharev and Corey, 2004). Two potential mechanisms have been suggested to explain the mechanosensitivity of MS channel gating: the bilayer model and the tethered model (Hamill and McBride, 1997; Hamill and Martinac, 2001; Martinac, 2004). In the bilayer model, local tension in the lipid bilayer of the cell’s surface membrane is alone suYcient to alter MG channel activity directly (Fig. 1A) (Martinac et al., 1990). This possibility is supported by the finding that certain prokaryotic MS channels, such as MscL and MscS, retain their mechanosensitivity when they are purified and reconstituted into pure lipid vesicles (Martinac et al., 1990; Sukharev et al., 1993, 1994; Opsahl and Webb, 1994; Hase et al., 1995; Hamill and McBride, 1997; Kloda and Martinac, 2001a,b,c; Martinac, 2001; Perozo and Rees, 2003), although whether this mechanism is used in living cells still remains unclear. In contrast, the tethered model suggests that a portion of the MS channel functions like a ‘‘gate’’ that is tethered to molecular elements in the CSK inside the cell (or to components of the ECM outside the cell) which actually sense the mechanical stress and transmit it to the channel. Hence, these connecting proteins act like gating springs that change MS channel gate opening and closing kinetics when mechanically stressed (Guharay and Sachs, 1984; Howard et al., 1988; Hudspeth and Gillespie, 1994; Huang et al., 1995; Hamill and McBride, 1996). In this gated model, local stress‐ induced displacement of the channel with respect to the CSK, such as might occur from generalized distortion of the membrane relative to its underlying submembranous CSK (or overlying ECM) due to fluid shear, would cause channel gating. Thus, the mechanosensitivity of a membrane channel protein might be aVected by local mechanical properties of the adjacent lipid bilayer, submembranous CSK, or adjacent ECM. The submembranous CSK is a specialized part of the cortical CSK composed of an actin–spectrin–ankyrin network that structurally supports the fluid bilayer and provides the cell membrane with a shear rigidity that is lacking in simple bilayer vesicles (Takakuwa and Mohandas, 1988; Mohandas and Evans, 1994). In mammalian cells, this relatively flexible submembranous CSK allows the cell to maintain excess membrane surface area beyond that required to enclose its volume when the cell is not fully spread (i.e., it buckles when cells round like when fabric is pulled together with a purse string). This additional surface area serves as an immediate membrane reserve (Evans, 1992; Mohandas and Evans, 1994), such that when a mammalian cell experiences


65

rapid mechanical deformations (e.g., inflation, stretching) the excess membrane will first unfold and smooth out before significant tension develops in the lipid bilayer (Knutton et al., 1976; Solsona et al., 1998; Raucher and Sheetz, 1999; Zhang and Hamill, 2000). This view is supported by the finding that the level of membrane tension measured in animal cells using optical tweezers is 1000 times less than that which is required to activate MS channels in the membranes of bacterial cells (Vogel and Sheetz, 2006). Other studies suggest that the plasma membrane of animal cells cannot be stretched; instead new membrane needs to be exocytosed or membrane lipids need to flow from one place to the other during cell spreading or major cell shape changes to prevent membrane tearing (Sheetz et al., 2006). For these reasons, it has been suggested that MS channels are not directly activated by changes in lipid bilayer tension in mammalian cells (Sheetz et al., 2006; Vogel and Sheetz, 2006). However, the mechanism of MS channel activation in higher cells remains controversial. For example, membrane tension is likely not uniform throughout the surface membrane, and it is possible that the low membrane tension measured in animal cells is an artifact of the optical tweezer measurement technique which cannot eVectively measure forces >100 pN (Hamill, 2006). Migrating cells also can generate traction forces at their leading edge that not only induce elongation and stretching of the whole cell, but also cause ripping of the plasma membrane at the cell’s rear trailing edge (Mayer et al., 2004). Many cells also exhibit membrane tearing under physiological conditions in vitro and in vivo (e.g., in intestinal cells during normal peristalsis) (McNeil and Ito, 1989) and thus, membrane tension can clearly reach lytic levels in animal cells (Hamill, 2006). Importantly, the cellular tensegrity model provides a way to reconcile these ostensibly conflicting findings, as will be described below. Although cytoskeletal and ECM components have been suggested to contribute to MS channel gating, their molecular identity remains unknown. Cytochalasin, which disrupts the connectivity of the internal actin CSK without depolymerization of F‐actin in living cells (Schliwa, 1982), increases the mechanosensitivity of many MS channels in animal cells (Guharay and Sachs, 1984; Sokabe et al., 1991; Small and Morris, 1994; Hamill and McBride, 1996). This observation suggests that actin microfilaments may normally suppress the level of tension transmitted to MS channels through the bilayer or the underlying submembranous CSK. This may occur either through microfilament‐binding interactions that produce molecular conformational changes or by altering the level of preexisting isometric tension (prestress) maintained in this molecular network. Microtubules also contribute to touch sensitivity in nematodes and insects (Thurm, 1964; Hamill and McBride, 1996; Tavernarakis and Driscoll, 1997). However, disruption of microtubules using colchicine has little eVect on MS channel activity in

66

Matthews et al.

skeletal muscle and in Xenopus oocytes measured using the patch‐clamp technique (Guharay and Sachs, 1984), it also does not significantly alter tactile sensation in the cockroach (Kuster et al., 1983). Thus, MS channel interactions with these CSK elements may vary considerably between diVerent cell and tissue types. Direct evidence relating to how certain MS channels are activated by molecules in the CSK and/or ECM is lacking (Hamill and Martinac, 2001). The basic idea of the tethered model is that proteins within these relatively rigid structural networks directly interact with the MS channel and that specific consensus regions or domains in the membrane channel mediate these interactions. For example, C‐terminal cysteine‐rich regions (Kanzaki et al., 1999) and N‐terminal repetitive ankyrin repeats (Walker et al., 2000) in the intracellular domain of MS channels may mediate interactions with adaptor proteins of the CSK or with other membrane components. Alternatively, these domains might serve to localize or cluster MS channels at particular sites where mechanotransduction occurs, rather than to mechanically gate the channel directly. Another possibility is that the main function of cytoskeletal or ECM linkages is to alter the level of tension experienced by the MS channel within the lipid bilayer by absorbing mechanical stresses, and thereby modifying the time‐dependence of channel adaptation (Ingber, 1997b; Hamill and Martinac, 2001).

IV. TENSEGRITY‐BASED CELLULAR MECHANOTRANSDUCTION Cells, tissues, and organs are constructed as interconnected structural hierarchies composed of self‐stabilizing, tensionally prestressed networks known as tensegrity structures (Ingber, 1997b, 2006). The cellular tensegrity theory proposes that tensional forces are borne by cytoskeletal microfilaments and intermediate filaments and that these forces are balanced by interconnected structural elements that resist compression, most notably, internal microtubule struts and ECM adhesions; this creates a state of isometric tension or ‘‘prestress’’ that stabilizes the entire cytoskeletal lattice (Fig. 2B). However, individual filaments can have dual functions and hence bear either tension or compression in diVerent structural contexts or at diVerent size scales (e.g., rigid actin filament bundles bear compression in filopodia) (Ingber, 1997b, 2003b). The tensional prestress that stabilizes the whole cell is primarily generated actively by the actomyosin apparatus within contractile microfilaments. Additional passive contributions to this prestress come from cell distension through adhesions to the ECM and other cells, osmotic forces acting on the


67

cell membrane, and forces exerted by filament polymerization. Intermediate filaments that interconnect at many points along microtubules, microfilaments, and the nuclear surface provide mechanical stiVness to the cell through their material properties, and their ability to act as suspensory cables that interconnect and tensionally stiVen the entire CSK and nuclear lattice. This internal microfilament–microtubule–intermediate filament CSK is permeated by a viscous cytosol and enclosed by a diVerentially permeable surface membrane that is mechanically supported by a highly elastic submembranous CSK. This submembranous cytoskeletal network is itself organized as a tensegrity at the molecular size scale (Ingber, 2003b) with tensed spectrin molecules being balanced by compressed actin protofilaments and by interconnected regions of the noncompressible lipid bilayer (Fig. 2C) (Sung and Vera, 2003; Vera et al., 2005). Importantly, the eYciency of mechanical coupling between this submembranous structural network and the deeper internal CSK depends on the type of molecular adhesion complex that forms on the cell surface. Specifically, integrins that form focal adhesions and connect to the deep CSK can eYciently resist shape distortion, whereas other transmembrane receptors that only connect to the submembranous CSK distort easily when mechanically stressed (Fig. 3) (Wang et al., 1993). Thus, this cortical CSK can either act independently or in concert with the remainder of the cell and deeper microfilament–microtubule–intermediate filament lattice. As a result of this integrated hierarchical architectural arrangement, forces applied at the macroscale that mechanically strain ECMs and deform cells and their internal CSK through integrins are able to filter down to smaller size scales and become focused on specific molecular components, both locally near the site of force application and at distant sites within the cell and nucleus. These focused stresses produce structural rearrangements within a subset of these molecules at the nanometer scale that changes their biochemical activities through kinetic or thermodynamic alterations (Ingber, 2006; Kumar et al., 2006). The mechanical stability of the network, and the eYciency of force transfer, can be changed by altering the level of prestress within the lattice; this can occur at one level of structural hierarchy (e.g., internal deep CSK independently of the submembranous CSK), or throughout structures distributed throughout the whole cell depending on the pattern of structural connections that the cell forms in response to diVerent microenvironmental stimuli (e.g., whether it is bound to ECM via integrins or other cells via cadherins or selectins) (Wang et al., 2001; Ingber, 2003b). As described above, the cellular tensegrity model is now supported by a plethora of experimental evidence (Caspar, 1980; Wang et al., 1993, 2001; Pickett‐Heaps et al., 1997; Farrell et al., 2002; Hutchison, 2002; Hu et al.,

68

Matthews et al.

FIGURE 3 Schematic representations of force application to transmembrane surface proteins that connect only to the submembranous CSK (A) or to integrin receptors that form focal adhesions that physically link the ECM, membrane, and submembranous CSK to the deep CSK (B) in mammalian cells. (A) Forces dissipate locally due to the high flexibility of the submembranous CSK (e.g., through extension of spectrin molecules) and thus, MS channels do not experience levels of distortion (strain) required for their activation. (B) When forces are applied to integrins that form stiVened focal adhesions, these stresses are channeled through protein linkages to associated MS channels in the focal adhesion, thereby causing molecular distortion and MS channel activation. Disruption of deeper actin microfilament tethers would result in greater distortion of the entire focal adhesion complex, increased internal strain, and hence, greater distortion‐based activation of the associated MS channels (not shown).

2003, 2004, 2005; Zanotti and Guerra, 2003; Brangwynne et al., 2006; Kumar et al., 2006) as well as by advances in mathematical, engineering, and statistical modeling (Connelly and Back, 1998; Stamenovic and Ingber, 2002; Wendling et al., 2003; Lin et al., 2004; Liu et al., 2004; Sultan et al., 2004; Shen and Wolynes, 2005; Sitharaman et al., 2005; Vera et al., 2005). In addition, a tensegrity‐based computational model of the submembranous CSK of the red blood cell has recently been described which adds additional support for the concept of the cell being a hierarchical tensegrity structure in which this cortical CSK can function either independently or in concert with the remainder of the internal CSK depending on whether these two networks are mechanically coupled inside the cell.


69

This is important in the context of mechanotransduction because the hierarchical cellular tensegrity model predicts that key cytoskeletal anchoring molecules, such as integrins, may preferentially sense physical forces at the cell surface and transmit these mechanical signals through either linkages within focal adhesions or deeper filamentous CSK connections to other mechanochemical transducer components, such as MS channels, at various locations inside the cell. The use of tensegrity by cells would also suggest that the actin CSK may modify MS channel activity by controlling the level of isometric tension (prestress) within the regions of submembranous cytoskeletal network (i.e., when it is connected to the deep CSK), rather than through direct binding interactions between actin and the channel molecules. For example, an increase of prestress in the submembranous CSK resulting from transmission of tensional forces from the actin CSK via focal adhesion connections may locally suppress cortical membrane deformation and thereby ‘‘tune’’ MS channel activity much like increasing the level of tension in a violin string that constrains the vibrational displacement of the string and hence, alters sound propagation when the string is strummed (Ingber, 1997b). Importantly, cellular tensegrity can reconcile the reality that animal cells have excess membrane which is much more flexible than bacterial membranes, yet forces can be distributed across wide regions of the cell surface (e.g., from the leading edge to the trailing membrane). Tensegrity predicts that these forces are not transmitted over long distances through the bilayer; rather they are channeled through integrins, focal adhesion proteins, and associated cytoskeletal filaments that connect one pole of the cell to the other, and link the surface membrane to the nucleus (Fig. 4). If the force is exerted on the membrane bilayer directly, then tearing will occur locally if the force produces a large hole in the membrane because it is exerted faster than membrane flow or replenishment can take place; this can occur at the initial site of force application, or at distant sites where forces are channeled through cytoskeletal connections (e.g., trailing edge of the cell). In contrast, formation of integrin‐mediated focal adhesions may overcome the high flexibility of surface membrane by recruiting focal adhesion proteins stiVening the cortical CSK locally (Fig. 3). Forces may then be focused on MS channels that associate with the stiVened focal adhesion, or with channels located at distant sites in the cell that are connected to the focal adhesion by stiVened linkages that channel these forces (Fig. 4), as has been shown to occur in many cell types (Maniotis et al., 1997; Hu et al., 2003). The clearest example of use of the discrete CSK channeling mechanism by cells is the demonstration that force application to cell surface ECM adhesions and associated cell distortion can activate MS channels on the nuclear membrane deep inside the cytoplasm (Itano et al., 2003).

Matthews et al.

70

Forward motion

Trailing edge

Ca

Ca

Leading edge Ca

Ca

ECM FIGURE 4 Schematic of a migrating cells showing both local and long‐range activation of MS ion channels via force transfer through discrete cytoskeletal filaments. Migrating cells exert greatest traction forces on integrins within focal adhesions just behind the leading edge, and at the rear trailing edge of the cell, because tensional forces (red double‐headed arrows) are transmitted through interconnecting cytoskeletal filaments (e.g., stress fibers that can span almost the entire length of the cell). Forces also can be transmitted over microfilaments and intermediate filaments to the nucleus at the cell center which commonly distorts in a coordinated manner as cells spread and move. Channeling of forces through the cytoplasm in this manner can result in stress concentrations simultaneously at multiple sites located throughout the cell that can activate MS channels and promote Ca2þ entry in these various locations (i.e., leading edge, trailing edge, and nuclear membrane). In this manner, animal cells can transmit tensional forces over large distances of the cell membrane through internal stiVened cytoskeletal elements, rather than through the lipid bilayer or the cortical membrane which could not support this type of force transfer to due its high flexibility.

This finding clearly shows that long distance force transfer which can activate MS channels at multiple locations in cells likely does not occur via lateral transmission through the bilayer, but rather through channeling via discrete filamentous connections inside the CSK (Fig. 4), as predicted by the tensegrity model.

V. FORCE TRANSMISSION THROUGH INTEGRINS IN LIVING CELLS In living tissues, mechanical stresses are normally distributed to cells through the ECM scaVolds that hold the cells together and provide mechanical support to the tissue (Alenghat and Ingber, 2002). As described above, mechanical signals that propagate from the ECM converge on cell surface integrin receptors (Ingber, 1997b; Kumar et al., 2006) that span the plasma membrane and physically link intracellularly to the contractile actin CSK by forming specialized macromolecular complexes, known as focal adhesions, that function as dynamic ‘‘spot welds’’ that anchor the cell to the ECM (Burridge and Chrzanowska‐Wodnicka, 1996; Geiger et al., 2001). The CSK


71

responds mechanically to forces transferred over the ECM and channeled through integrins by rearranging interlinked actin microfilaments, microtubules, and intermediate filaments that comprise the lattice, thereby strengthening the whole cell against the potential deleterious eVects of mechanical distortion (Wang et al., 1993, 2001; Maniotis et al., 1997; Ralphs et al., 2002; Ingber, 2003b; Matthews et al., 2006). The ability of the elements of the cytoskeletal network to rearrange and stiVen in response to stress also depends on the level of tensile prestress in the CSK, in accordance with the tensegrity model (Ingber, 1997a, 2006; Stamenovic and Ingber, 2002; Stamenovic et al., 2003). The use of transmembrane adhesion receptors and linked cytoskeletal filament networks for force transmission provides a way for cells to channel and focus stresses applied at the cell surface so that they concentrate on local focal adhesions as well as at distant sites in the cell (e.g., mitochondria, nucleus, focal adhesions at the opposite pole of the cell) (Wang et al., 1993; Davies et al., 1994, 2003; Maniotis et al., 1997; Helmke et al., 2003; Ingber, 2003b; Matthews et al., 2004, 2006). For example, when mechanical stresses are applied to specific cell surface receptors using ligand‐coated magnetic microbeads with applied magnetic fields, the cell appears either highly flexible or extremely stiV when probed through transmembrane metabolic receptors (e.g., growth factor receptors, histocompatibility antigens) that only link to the submembranous CSK or through integrins that form focal adhesions that connect this cortical network to the deeper CSK (microfilament–intermediate filament–microtubule–nuclear lattice), respectively (Fig. 3) (Wang et al., 1993, 2001; Maniotis et al., 1997; Matthews et al., 2004). Another pertinent example is that when cells that express constitutively active myosin light chain kinase (and hence produce higher prestress in the deep CSK) were compared with control cells using this magnetic cytometry method, no diVerence in cell stiVness could be detected when probed through transmembrane molecules that only link to the submembranous CSK, whereas the more highly prestressed cells appeared much stiVer than controls when probed through integrins (Cai et al., 1998). Importantly, when the stiVness of the same cells was analyzed using the classic glass pipette ‘‘poking’’ technique, the ‘‘cortex’’ appeared stiVer in the more highly prestressed cells (Cai et al., 1998). This means that poking produces extremely large mechanical strain in the surface membrane and nonspecifically (i.e., in the absence of specific focal adhesion linkages) distorts the deeper CSK as well. If this distortion is great enough, it could potentially activate MS channels or internal Ca 2þ release mechanism by pulling on cytoskeletal elements that link to these structures from inside the cell.

72

Matthews et al.

Integrins can trigger signaling transduction cascades and induce focal adhesion formation as a result of ECM ligand binding and associated changes in integrin receptor conformation alone (Shimaoka and Springer, 2003; Springer and Wang, 2004). However, application of mechanical forces to bound integrins also can convey distinct biochemical signals to the cell (Ingber, 1997b; Geiger and Bershadsky, 2001). For example, application of force to integrins induces activation of Rho and its eVectors mDia and ROCK, which promotes actin filament polymerization and induces cytoskeletal contraction, respectively; these eVects result in focal adhesion formation (Riveline et al., 2001; Galbraith et al., 2002). Stresses applied to integrins can also regulate gene expression both transcriptionally by activating chemical signaling cascades (e.g., cAMP signaling; Meyer et al., 2000) and posttranscriptionally by modulating the formation of protein synthetic complexes at focal adhesions (Chicurel et al., 1998). In addition, mechanical stress application to cells already bound to ECM can induce a new wave of integrin activation which results in the activation of Rac and Rho that are important for lamellepodia extension, stress fiber reinforcement, and realignment of cells when these receptors bind to additional ECM molecules (Tzima et al., 2001, 2002; Tzima, 2006). Integrins also mediate many other eVects of mechanical forces on biochemistry and cellular physiology. For example, cell proliferation depends on the ability of cells to spread and generate traction stresses on the ECM (Huang and Ingber, 1999), a process which is mediated through integrin‐ dependent activation of Rho, mDia1, and ROCK (Mammoto et al., 2004). Strain‐induced activation of p38 MAP Kinase in cardiomyocytes is integrin dependent (Kudoh et al., 1998; Aikawa et al., 2002), as is the release of growth factors by cyclic strain in vascular smooth muscle cells (Martinez‐ Lemus et al., 2003). Furthermore, integrin‐dependent cell spreading and associated mechanical distortion of the nucleus appear to induce Ca2þ entry into the nucleus and turn on gene transcription through activation of MS channels on the nuclear membrane in fibroblasts (Itano et al., 2003). Thus, preferential channeling of forces through integrins, focal adhesions, and linked cytoskeletal networks that produce stress concentrations at numerous sites in the cell (Maniotis et al., 1997; Wang et al., 2001; Hu et al., 2003, 2004, 2005; Ingber, 2006) may be responsible for simultaneously distorting, and thereby activating, multiple mechanotransducer molecules throughout the cell, as suggested by the tensegrity model (Ingber, 1997b, 2006). Importantly, experimental studies confirm that the eYciency of force channel through the CSK also depends on the level of prestress in the cell, such that only local stress concentrations are observed at the site of force application when prestress is dissipated using pharmacological or genetic techniques (Hu et al., 2003).


73

VI. POTENTIAL LINKAGES BETWEEN INTEGRINS AND MS ION CHANNELS Although it is now clear that integrins mediate various forms of mechanotransduction, their role in control of MS channel activity remains unclear. The possibility that integrins may also control MS channel function has been raised based on circumstantial evidence in the past. For example, mechanical strain‐induced electrophysiological responses can be inhibited in chondrocytes by adding either soluble antibodies or RGD‐peptides (from the integrin‐binding site of fibronectin) that interfere with integrin binding, or the MS channel inhibitor, gadolinium chloride (Salter et al., 1997). Uniaxial cyclic strain also activates MS Ca2þ channels in endothelial cells (Naruse et al., 1998a,b; Sasamoto et al., 2005), and this is accompanied by increased expression of b3‐integrins (Suzuki et al., 1997). Both integrin activation and Ca2þ influx are also critical for stretch‐induced IL‐6 secretion in endothelial cells (Sasamoto et al., 2005). To explore whether forces applied to integrins induce changes in MS channel activity, investigators have applied mechanical forces directly to integrins via receptor‐bound magnetic microbeads in conjunction with applied magnetic fields. Application of tensional force to surface‐bound, collagen‐coated microbeads with a magnetic tweezer induces a global increase in cytoplasmic Ca2þ levels in fibroblasts (Glogauer et al., 1995). However, it remains unclear whether MS channels directly mediate these eVects, or if the applied stress induces release of Ca2þ from intracellular stores in these cells (Glogauer et al., 1995, 1997b, 1998). Interestingly, actin, but not vinculin, was recruited to the bead site in response to direct force application to these collagen‐coated beads, and disrupting the actin CSK with cytochalasin D increased the integrin‐mediated Ca2þ release in response to force application (Glogauer et al., 1995, 1997a). Using a similar magnetic manipulation technique, we showed that application of force directly to integrins via bound magnetic RGD‐coated microbeads that ligate cell surface integrins induces a rapid increase in intracellular Ca 2þ levels in capillary endothelial cells and that this response can be suppressed by addition of the MS channel inhibitor, gadolinium chloride (Matthews et al., 2006) (Fig. 5). In more recent unpublished studies, we have detected local Ca2þ influx at the site of bead binding to integrins within 70 ms after force administration, as well as a force‐dependent increase in Ca2þ entry as the level of stress was raised from 0.1 to 2 nN. Most importantly, application of similar forces to nonintegrin transmembrane receptors that do not promote focal adhesion formation on the surface membrane of the same cells did not produce this response, and similar responses were observed in multiple cell types. Thus, generalized deformation of the plasma membrane does

FIGURE 5 Activation of Ca2þ entry through MS ion channels by forces applied to cell surface integrin receptors using magnetic pulling cytometry. (A) Phase contrast view of an adherent capillary endothelial cell with attached RGD coated magnetic microbead (4.5 mm, white arrow). Black arrowhead denotes the position of an electromagnetic needle used to apply force to cell via the attached magnetic bead. (B) A time series of pseudocolored fluorescence images of the cell shown in (A) after mechanical stress (5 nN) was applied with the magnet. These pseudocolored images demonstrate a transient stress‐induced increase in intracellular calcium [Ca2þ]i as a brief shift in color from blue to yellow over 0–57 s, as detected using FURA‐2AM ratio‐imaging (color bar indicates [Ca2þ]i in nanomolar; the force pulse was applied at 9 s). (C) Plot of average [Ca2þ]i for control (open diamonds) and gadolinium chloride‐treated (solid circles) cells as a function of time; the inhibition of stress‐induced Ca2þ influx by gadolinium suggests that force‐induced Ca2þ entry into these cells is via stress‐dependent activation of MS channels. Black arrow indicates when the 3‐s force pulse was applied. Reprinted with permission from ‘‘The Company of Biologists’’ (Matthews et al. (2006)).


75

not appear to be suYcient to activate Ca2þ entry through MS channels in these cells; forces must be applied through activated (ligated) integrins that couple to the CSK through focal adhesions to activate this mechanochemical transduction response. The rapidity of the response also suggests that this mechanism likely occurs directly at the site of force application within the focal adhesion. If this is true, then how do integrins gate MS ion channels? One possibility is that integrins may bind directly to certain MS channels or associate within common macromolecular complexes on the surface membrane, and there is some evidence to support this possibility. For example, ENaC channels and voltage‐gated calcium channels (VGCC) coprecipitate in b1‐integrin immune complexes isolated from mouse chondrocytes (Mobasheri et al., 2002; Shakibaei and Mobasheri, 2003). Polycystin 1 (PC1), a component of the polycystin 2 (PC2) TRPP family channel complex that also associates with other members of the TRP family (Nilius and Voets, 2005), colocalizes with the collagen receptor, integrin a2b1, in renal epithelial cells (Wilson et al., 1999; Wilson, 2001). Integrin‐associated focal proteins, such as vinculin and FAK, co‐immunoprecipitate with PC1 as well (Wilson et al., 1999; Wilson, 2001). Interestingly, most of the TRP channels contain ankyrin domains (Minke and Cook, 2002; Nilius and Voets, 2005; Nilius et al., 2005) which can bind to cytoskeletal adaptor proteins (Hryniewicz‐Jankowska et al., 2002). Intriguingly, ankyrin domains in integrin‐linked kinase (ILK) appear to be important for its association with integrins in focal adhesions (Wu, 2004; Hannigan et al., 2005). The focal adhesion protein kinase FAK also has been shown to directly associate with C‐terminus of the hSlo a‐subunit of the large conductance Ca2þ‐activated potassium (BK) MS channel (Rezzonico et al., 2003). Thus, taken together, these observations support the possibility that integrins interact with MS channel complexes in focal adhesions. Hence, external forces may activate MS channels as a result of being channeled across the cell surface through transmembrane integrin receptors, rather than as a result of generalized lipid bilayer distortion. Given these observations, we need to readdress the gating of MS channels in the light of the previously proposed bilayer tension or tethered gate models (Fig. 1). TRAC1 channels that contain short ankyrin adapter domains can be directly activated by membrane stretch in reconstituted liposomes that are devoid of CSK (Maroto et al., 2005). Other MS channels can be activated in membrane ‘‘blebs’’ that are torn free from underlying cytoskeletal connections also supporting the idea that these channels may be activated by force transfer through the bilayer; however, these channels do not exhibit the normal regulated behavior of channels observed in intact cells (Hamill and McBride, 1997; Hamill and Martinac, 2001). Thus, while forces

76

Matthews et al.

may be able to be transmitted through the bilayer, this may not be what is happening under physiological conditions. In fact, TRAP1 activation appears to require multiple ankyrin domains that could mediate binding interactions with CSK proteins (Corey et al., 2004) or integrins (Wu, 2004; Hannigan et al., 2005). Also, while the newly cloned BK channels (Ca2þ‐activated Kþ channels) from chick ventricular myocytes can be activated by changes of membrane tension induced by amphipaths, deletion of the STREX domain in the channel abolishes this response (Tang et al., 2003; Qi et al., 2005). Thus, the observed changes in channel gating are likely mediated by interactions with adapter proteins, rather than resulting from direct eVects of amphipaths on the lipid bilayer alone (Tang et al., 2003; Qi et al., 2005). Thus, there is strong evidence that ECM–integrin–focal adhesion (CSK) linkages might be crucial for gating of certain MS channels; however, the mechanism underlying this regulation is presently unclear. However, here we propose potential mechanisms that may mediate this response in eukaryotic cells as shown in Fig. 3. As described above, eukaryotic cells generally contain excess membrane area relative to their volume such that many small membrane extensions are observed on the cell surface. This is possible because the submembranous CSK that provides most of the shape stability of the plasma membrane is highly flexible, and this is largely due to the great extensibility of the spectrin molecules that form the core lattice. If MS channels contain structural motifs that physically link them to this submembranous CSK, then they might experience deformation and change their gating activity when this lattice is mechanically stressed. However, because there is so much excess membrane and the submembranous CSK is so flexible, it is likely that the applied forces would dissipate through restructuring of the membrane before altering channel mechanics (Fig. 3A), except for high levels of mechanical strain or very rapid and highly focused perturbations, but these perturbations also may cause membrane tearing. In contrast, when integrins are bound and activated, they form focal adhesions that must in some way physically integrate with portions of the submembranous CSK that is present throughout the cell cortex (Fig. 3B). The high density of tightly bound focal adhesion molecules will stiVen the associated portion of the submembranous CSK and thus focus stresses that propagate from the ECM, or from within the cell, on associated MS channel proteins in an integrin‐dependent manner. These channels will experience a high local stress due to the increased rigidity of the lattice and enhance their gating activity. In contrast, application of the same force to other transmembrane proteins that do not associate with a rigidified focal adhesion (but still interact with the lipid bilayer) will not activate this response, again because


77

the stresses would dissipate within the flexible submembranous CSK lattice before they produced MS channel distortion. In summary, focal adhesion assembly might provide a way to rigidify structural linkages between integrins and MS channels, and thereby channel forces between these molecules. It also might recruit MS channels and thus increase the number of functional channels at these mechanosensing sites. The finding that disruption of the actin CSK with cytochalasin D increases (rather than decreases) stress‐dependent MS channel activation in many cells also needs to be readdressed in context of what we have learned about integrins and cellular tensegrity. First, it is important to clarify that cytochalasins do not cause F‐actin depolymerization in intact cells; rather they produce breakage of the central actin network (Schliwa, 1982). In fact, when cut, the tensed actin filaments usually retract back to the cell cortex which appears to remain intact in cytochalasin‐treated cells, and cells can adhere to ECM and form focal adhesions in the presence of high concentrations of cytochalasin (Ingber et al., 1995, unpublished observations). However, disruption of the integrity of the actin CSK will dissipate prestress in the cell. If MS channel activity is due to the level of isometric tension exerted on the protein, then lowering this prestress by disrupting the central actin CSK should decrease force sensitivity (i.e., increased stress would need to be applied to reach the same final state of tension that is necessary to produce channel activation). On the other hand, if MS channel activation depends on local mechanical strain in the rigidified focal adhesion/MS channel complex, then loss of prestress will increase channel activation in response to stress because greater local distortion of the focal adhesion will be produced when it is severed from the tensed internal CSK, much like when an untethered sail ‘‘luVs in the wind.’’ Thus, MS channels may sense local mechanical strain that will be greater when forces are transmitted over the stiVened focal adhesion relative to the flexible submembranous CSK, and even greater when stabilizing tethers that connect the entire stiVened focal adhesion to the underlying contractile CSK are severed.

VII. CONCLUSIONS AND FUTURE IMPLICATIONS Mechanotransduction is fundamental to many physiological responses and deregulation of this process leads to disease. MS ion channels constitute the first line of force transducers and regulate important functions such as hearing, touch, and vestibular function. However, both the identity of these channels and the precise gating mechanisms remain unknown in most cells. Force‐dependent distortion of the lipid bilayer represents one potential way to gate MS channels, and this might occur in certain cells (e.g., in

78

Matthews et al.

some bacteria). But in specialized organs, such as the inner ear, MS channels are directly connected to force‐bearing elements in the CSK, and their activity is sensitive to the level of prestress in the entire ECM–CSK lattice (Ingber, 2006). Integrin receptors, by connecting the ECM to the CSK and resisting cell‐ generated forces, are perfectly poised to modulate mechanical force transfer through living cells. Thus, they are excellent candidates for controlling mechanical gating of MS channels. Integrins connect to the CSK through a number of focal adhesion adapter proteins, which relay chemical and mechanical signals into the cells through change in their conformation and binding kinetics. Importantly, there is increasing evidence to suggest that some MS channels associate with focal adhesions, and thus, may form part of this nanoscale mechanochemical signaling complex. Moreover, recent work from our laboratory has confirmed that direct force application to integrins and associated focal adhesions activates Ca2þ entry through MS channels, whereas application of the same force to other transmembrane channels that do not form focal adhesions fails to produce this response in the same cells. Thus, we believe that, at least in the multiple mammalian cells we have studied, generalized cortical membrane distortion is not suYcient to activate these MS channels and, instead, that this is an integrin‐dependent mechanotransduction response. Focal adhesion formation may be viewed to enhance channel sensitivity by locally increasing the rigidity of this macromolecular complex and hence, more eYciently channeling stresses to these critical transduction molecules. Future studies on the mechanism of MS channel activation in eukaryotic cells will, therefore, need to explore the role of links between MS channels, integrins, and focal adhesion proteins, as well as how the deeper CSK influences channel sensitivity and adaptation responses. This will likely require a combination of biophysical, electrophysiological, genetic, biochemical, and cell biological techniques. But to fully understand this mechanism, it probably will be necessary to develop entirely new methods that will permit us to analyze single channel activities in the normal structural context of whole living cells, rather than within isolated membranes or regions of the intact membrane that are locally fixed to a rigid glass pipette (i.e., as is done with patch‐clamp approaches now). Only this way, will it be possible to tease out the critical local and global structural elements that govern stress‐dependent activation of MS ion channels in living cells. References Aikawa, R., Nagai, T., Kudoh, S., Zou, Y., Tanaka, M., Tamura, M., Akazawa, H., Takano, H., Nagai, R., and Komuro, I. (2002). Integrins play a critical role in mechanical stress‐induced p38 MAPK activation. Hypertension 39, 233–238.


79

Alenghat, F. J., and Ingber, D. E. (2002). Mechanotransduction: All signals point to cytoskeleton, matrix, and integrins. Sci. STKE 2002, PE6. Brangwynne, C. P., MacKintosh, F. C., Kumar, S., Geisse, N. A., Talbot, J., Mahadevan, L., Parker, K. K., Ingber, D. E., and Weitz, D. A. (2006). Microtubules can bear enhanced compressive loads in living cells because of lateral reinforcement. J. Cell Biol. 173, 733–741. Burridge, K., and Chrzanowska‐Wodnicka, M. (1996). Focal adhesions, contractility, and signaling. Annu. Rev. Cell Dev. Biol. 12, 463–518. Cai, S., Pestic‐Dragovich, L., O’Donnell, M. E., Wang, N., Ingber, D., Elson, E., and De Lanerolle, P. (1998). Regulation of cytoskeletal mechanics and cell growth by myosin light chain phosphorylation. Am. J. Physiol. 275, C1349–C1356. Caspar, D. L. (1980). Movement and self‐control in protein assemblies. Quasi‐equivalence revisited. Biophys. J. 32, 103–138. Chicurel, M. E., Singer, R. H., Meyer, C. J., and Ingber, D. E. (1998). Integrin binding and mechanical tension induce movement of mRNA and ribosomes to focal adhesions. Nature 392, 730–733. Cho, H., Shin, J., Shin, C. Y., Lee, S. Y., and Oh, U. (2002). Mechanosensitive ion channels in cultured sensory neurons of neonatal rats. J. Neurosci. 22, 1238–1247. Connelly, R., and Back, A. (1998). Mathematics and tensegrity. Am. Sci. 86, 142–151. Corey, D. P., and Hudspeth, A. J. (1983). Kinetics of the receptor current in bullfrog saccular hair cells. J. Neurosci. 3, 962–976. Corey, D. P., Garcia‐Anoveros, J., Holt, J. R., Kwan, K. Y., Lin, S. Y., Vollrath, M. A., Amalfitano, A., Cheung, E. L., Derfler, B. H., Duggan, A., Geleoc, G. S., Gray, P. A., et al. (2004). TRPA1 is a candidate for the mechanosensitive transduction channel of vertebrate hair cells. Nature 432, 723–730. Davies, P. F., Robotewskyj, A., and Griem, M. L. (1994). Quantitative studies of endothelial cell adhesion. Directional remodeling of focal adhesion sites in response to flow forces. J. Clin. Invest. 93, 2031–2038. Davies, P. F., Zilberberg, J., and Helmke, B. P. (2003). Spatial microstimuli in endothelial mechanosignaling. Circ. Res. 92, 359–370. Evans, E. (1992). Composite membranes and structured interfaces: From simple to complex designs in biology. In ‘‘Biomembranes Structure and Function—The State of the Art’’ (B. P. Gaber and K. R. M. Easwaran, eds.), pp. 81–101. Adenine, New York. Farrell, H. M., Jr., Qi, P. X., Brown, E. M., Cooke, P. H., Tunick, M. H., Wickham, E. D., and Unruh, J. J. (2002). Molten globule structures in milk proteins: Implications for potential new structure‐function relationships. J. Dairy Sci. 85, 459–471. Galbraith, C. G., Yamada, K. M., and Sheetz, M. P. (2002). The relationship between force and focal complex development. J. Cell Biol. 159, 695–705. Geiger, B., and Bershadsky, A. (2001). Assembly and mechanosensory function of focal contacts. Curr. Opin. Cell Biol. 13, 584–592. Geiger, B., Bershadsky, A., Pankov, R., and Yamada, K. M. (2001). Transmembrane crosstalk between the extracellular matrix—cytoskeleton crosstalk. Nat. Rev. Mol. Cell. Biol. 2, 793–805. Gillespie, P. G., and Walker, R. G. (2001). Molecular basis of mechanosensory transduction. Nature 413, 194–202. Glogauer, M., Ferrier, J., and McCulloch, C. A. (1995). Magnetic fields applied to collagen‐ coated ferric oxide beads induce stretch‐activated Ca2þ flux in fibroblasts. Am. J. Physiol. 269, C1093–C1104. Glogauer, M., Arora, P., Yao, G., Sokholov, I., Ferrier, J., and McCulloch, C. A. (1997a). Calcium ions and tyrosine phosphorylation interact coordinately with actin to regulate cytoprotective responses to stretching. J. Cell Sci. 110(Pt. 1), 11–21.

80

Matthews et al.

Glogauer, M., Arora, P., Yao, G., Sokholov, I., Ferrier, J., and McCulloch, C. A. and Mechanosensitive (1997b). Calcium ions and tyrosine phosphorylation interact coordinately with actin to regulate cytoprotective responses to stretching. J. Cell Sci. 110(Pt. 1), 11–21. Glogauer, M., Arora, P., Chou, D., Janmey, P. A., Downey, G. P., and McCulloch, C. A. (1998). The role of actin‐binding protein 280 in integrin‐dependent mechanoprotection. J. Biol. Chem. 273, 1689–1698. Guharay, F., and Sachs, F. (1984). Stretch‐activated single ion channel currents in tissue‐ cultured embryonic chick skeletal muscle. J. Physiol. 352, 685–701. Hamill, O. P. (2006). Twenty‐odd years of stretch‐sensitive channels. Pflugers Arch. 453(3), 333–351. Hamill, O. P., and Martinac, B. (2001). Molecular basis of mechanotransduction in living cells. Physiol. Rev. 81, 685–740. Hamill, O. P., and McBride, D. (1993). Molecular clues to mechanosensitivity. Biophys. J. 65, 17–18. Hamill, O. P., and McBride, D. W., Jr. (1994). The cloning of a mechano‐gated membrane ion channel. Trends Neurosci. 17, 439–443. Hamill, O. P., and McBride, D. W., Jr. (1996). A supramolecular complex underlying touch sensitivity. Trends Neurosci. 19, 258–261. Hamill, O. P., and McBride, D. W., Jr. (1997). Induced membrane hypo/hyper‐mechanosensitivity: A limitation of patch‐clamp recording. Annu. Rev. Physiol. 59, 621–631. Hannigan, G., Troussard, A. A., and Dedhar, S. (2005). Integrin‐linked kinase: A cancer therapeutic target unique among its ILK. Nat. Rev. Cancer 5, 51–63. Hase, C. C., Le Dain, A. C., and Martinac, B. (1995). Purification and functional reconstitution of the recombinant large mechanosensitive ion channel (MscL) of Escherichia coli. J. Biol. Chem. 270, 18329–18334. Helmke, B. P., Rosen, A. B., and Davies, P. F. (2003). Mapping mechanical strain of an endogenous cytoskeletal network in living endothelial cells. Biophys. J. 84, 2691–2699. Howard, J., Roberts, W. M., and Hudspeth, A. J. (1988). Mechanoelectrical transduction by hair cells. Annu. Rev. Biophys. Biophys. Chem. 17, 99–124. Hryniewicz‐Jankowska, A., Czogalla, A., Bok, E., and Sikorsk, A. F. (2002). Ankyrins, multifunctional proteins involved in many cellular pathways. Folia Histochem. Cytobiol. 40, 239–249. Hu, S., Chen, J., Fabry, B., Numaguchi, Y., Gouldstone, A., Ingber, D. E., Fredberg, J. J., Butler, J. P., and Wang, N. (2003). Intracellular stress tomography reveals stress focusing and structural anisotropy in cytoskeleton of living cells. Am. J. Physiol. Cell Physiol. 285, C1082–C1090. Hu, S., Eberhard, L., Chen, J., Love, J. C., Butler, J. P., Fredberg, J. J., Whitesides, G. M., and Wang, N. (2004). Mechanical anisotropy of adherent cells probed by a three‐dimensional magnetic twisting device. Am. J. Physiol. Cell Physiol. 287, C1184–C1191. Hu, S., Chen, J., Butler, J. P., and Wang, N. (2005). Prestress mediates force propagation into the nucleus. Biochem. Biophys. Res. Commun. 329, 423–428. Huang, S., and Ingber, D. E. (1999). The structural and mechanical complexity of cell‐growth control. Nat. Cell Biol. 1, E131–E138. Huang, M., Gu, G., Ferguson, E. L., and Chalfie, M. (1995). A stomatin‐like protein necessary for mechanosensation in C. elegans. Nature 378, 292–295. Hudspeth, A. J., and Gillespie, P. G. (1994). Pulling springs to tune transduction: Adaptation by hair cells. Neuron 12, 1–9. Hutchison, C. J. (2002). Lamins: Building blocks or regulators of gene expression? Nat. Rev. Mol. Cell. Biol. 3, 848–858.


81

Ingber, D. (1991). Integrins as mechanochemical transducers. Curr. Opin. Cell Biol. 3, 841–848. Ingber, D. E. (1993). Cellular tensegrity: Defining new rules of biological design that govern the cytoskeleton. J. Cell Sci. 104(Pt. 3), 613–627. Ingber, D. E. (1997a). Integrins, tensegrity, and mechanotransduction. Gravit. Space Biol. Bull. 10, 49–55. Ingber, D. E. (1997b). Tensegrity: The architectural basis of cellular mechanotransduction. Annu. Rev. Physiol. 59, 575–599. Ingber, D. E. (2003a). Mechanobiology and diseases of mechanotransduction. Ann. Med. 35, 564–577. Ingber, D. E. (2003b). Tensegrity I. Cell structure and hierarchical systems biology. J. Cell Sci. 116, 1157–1173. Ingber, D. E. (2006). Cellular mechanotransduction: Putting all the pieces together again. FASEB J. 20, 811–827. Ingber, D. E., and Jamieson, J. D. (1985). Cells as tensegrity structures: Architectural regulation of histodifferentiation by physical forces transduced over basement membrane. In ‘‘Gene Expression During Normal and Malignant Differentiation’’ (L. C. Andersson, C. G. Gahmberg, and P. Ekblom, eds.), pp. 13–32. Academic Press, Orlando. Ingber, D. E., Prusty, D., Sun, Z., Betensky, H., and Wang, N. (1995). Cell shape, cytoskeletal mechanics, and cell cycle control in angiogenesis. J. Biomech. 28, 1471–1484. Itano, N., Okamoto, S., Zhang, D., Lipton, S. A., and Ruoslahti, E. (2003). Cell spreading controls endoplasmic and nuclear calcium: A physical gene regulation pathway from the cell surface to the nucleus. Proc. Natl. Acad. Sci. USA 100, 5181–5186. Kanzaki, M., Nagasawa, M., Kojima, I., Sato, C., Naruse, K., Sokabe, M., and Iida, H. (1999). Molecular identification of a eukaryotic, stretch‐activated nonselective cation channel. Science 285, 882–886. Kloda, A., and Martinac, B. (2001a). Mechanosensitive channel of thermoplasma, the cell wall‐less archaea: Cloning and molecular characterization. Cell Biochem. Biophys. 34, 321–347. Kloda, A., and Martinac, B. (2001b). Molecular identification of a mechanosensitive channel in archaea. Biophys. J. 80, 229–240. Kloda, A., and Martinac, B. (2001c). Structural and functional diVerences between two homologous mechanosensitive channels of Methanococcus jannaschii. EMBO J. 20, 1888–1896. Knutton, S., Jackson, D., Graham, J. M., Micklem, K. J., and Pasternak, C. A. (1976). Microvilli and cell swelling. Nature 262, 52–54. Komulainen, J., Takala, T. E., Kuipers, H., and Hesselink, M. K. (1998). The disruption of myofibre structures in rat skeletal muscle after forced lengthening contractions. Pflugers Arch. 436, 735–741. Kudoh, S., Komuro, I., Hiroi, Y., Zou, Y., Harada, K., Sugaya, T., Takekoshi, N., Murakami, K., Kadowaki, T., and Yazaki, Y. (1998). Mechanical stretch induces hypertrophic responses in cardiac myocytes of angiotensin II type 1a receptor knockout mice. J. Biol. Chem. 273, 24037–24043. Kumar, S., Maxwell, I. Z., Heisterkamp, A., Polte, T. R., Lele, T. P., Salanga, M., Mazur, E., and Ingber, D. E. (2006). Viscoelastic retraction of single living stress fibers and its impact on cell shape, cytoskeletal organization, and extracellular matrix mechanics. Biophys. J. 90, 3762–3773. Kuster, J. E., French, A. S., and Sanders, E. J. (1983). The eVects of microtubule dissociating agents on the physiology and cytology of the sensory neuron in the femoral tactile spine of the cockroach, periplaneta americana L. Proc. R. Soc. Lond. B Biol. Sci. 219, 397–412. Lin, D. C., Yurke, B., and Langrana, N. A. (2004). Mechanical properties of a reversible, DNA‐ crosslinked polyacrylamide hydrogel. J. Biomech. Eng. 126, 104–110.

82

Matthews et al.

Liu, D., Wang, M., Deng, Z., Walulu, R., and Mao, C. (2004). Tensegrity: Construction of rigid DNA triangles with flexible four‐arm DNA junctions. J. Am. Chem. Soc. 126, 2324–2325. Mammoto, A., Huang, S., Moore, K., Oh, P., and Ingber, D. E. (2004). Role of RhoA, mDia, and ROCK in cell shape‐dependent control of the Skp2‐p27kip1 pathway and the G1/S transition. J. Biol. Chem. 279, 26323–26330. Maniotis, A. J., Chen, C. S., and Ingber, D. E. (1997). Demonstration of mechanical connections between integrins, cytoskeletal filaments, and nucleoplasm that stabilize nuclear structure. Proc. Natl. Acad. Sci. USA 94, 849–854. Maroto, R., Raso, A., Wood, T. G., Kurosky, A., Martinac, B., and Hamill, O. P. (2005). TRPC1 forms the stretch‐activated cation channel in vertebrate cells. Nat. Cell Biol. 7, 179–185. Martinac, B. (2001). Mechanosensitive channels in prokaryotes. Cell Physiol. Biochem. 11, 61–76. Martinac, B. (2004). Mechanosensitive ion channels: Molecules of mechanotransduction. J. Cell Sci. 117, 2449–2460. Martinac, B., Adler, J., and Kung, C. (1990). Mechanosensitive ion channels of E. coli activated by amphipaths. Nature 348, 261–263. Martinez‐Lemus, L. A., Wu, X., Wilson, E., Hill, M. A., Davis, G. E., Davis, M. J., and Meininger, G. A. (2003). Integrins as unique receptors for vascular control. J. Vasc. Res. 40, 211–233. Matthews, B. D., Overby, D. R., Alenghat, F. J., Karavitis, J., Numaguchi, Y., Allen, P. G., and Ingber, D. E. (2004). Mechanical properties of individual focal adhesions probed with a magnetic microneedle. Biochem. Biophys. Res. Commun. 313, 758–764. Matthews, B. D., Overby, D. R., Mannix, R., and Ingber, D. E. (2006). Cellular adaptation to mechanical stress: Role of integrins, Rho, cytoskeletal tension and mechanosensitive ion channels. J. Cell Sci. 119, 508–518. Mayer, C., Maaser, K., Daryab, N., Zanker, K. S., Brocker, E. B., and Friedl, P. (2004). Release of cell fragments by invading melanoma cells. Eur. J. Cell Biol. 83, 709–715. McNeil, P. L., and Ito, S. (1989). Gastrointestinal cell plasma membrane wounding and resealing in vivo. Gastroenterology 96, 1238–1248. Meyer, C. J., Alenghat, F. J., Rim, P., Fong, J. H., Fabry, B., and Ingber, D. E. (2000). Mechanical control of cyclic AMP signalling and gene transcription through integrins. Nat. Cell Biol. 2, 666–668. Minke, B., and Cook, B. (2002). TRP channel proteins and signal transduction. Physiol. Rev. 82, 429–472. Mobasheri, A., Carter, S. D., Martin‐Vasallo, P., and Shakibaei, M. (2002). Integrins and stretch activated ion channels: Putative components of functional cell surface mechanoreceptors in articular chondrocytes. Cell Biol. Int. 26, 1–18. Mohandas, N., and Evans, E. (1994). Mechanical properties of the red cell membrane in relation to molecular structure and genetic defects. Annu. Rev. Biophys. Biomol. Struct. 23, 787–818. Naruse, K., Sai, X., Yokoyama, N., and Sokabe, M. (1998a). Uni‐axial cyclic stretch induces c‐src activation and translocation in human endothelial cells via SA channel activation. FEBS Lett. 441, 111–115. Naruse, K., Yamada, T., Sai, X. R., Hamaguchi, M., and Sokabe, M. (1998b). Pp125FAK is required for stretch dependent morphological response of endothelial cells. Oncogene 17, 455–463. Naruse, K., Yamada, T., and Sokabe, M. (1998c). Involvement of SA channels in orienting response of cultured endothelial cells to cyclic stretch. Am. J. Physiol. 274, H1532–H1538. Nilius, B., and Voets, T. (2005). TRP channels: A TR(I)P through a world of multifunctional cation channels. Pflugers Arch. 451, 1–10.


83

Nilius, B., Voets, T., and Peters, J. (2005). TRP channels in disease. Sci. STKE 2005, re8. Opsahl, L. R., and Webb, W. W. (1994). Transduction of membrane tension by the ion channel alamethicin. Biophys. J. 66, 71–74. Perozo, E., and Rees, D. C. (2003). Structure and mechanism in prokaryotic mechanosensitive channels. Curr. Opin. Struct. Biol. 13, 432–442. Pickett‐Heaps, J. D., Forer, A., and Spurck, T. (1997). Traction fibre: Toward a ‘‘tensegral’’ model of the spindle. Cell Motil. Cytoskeleton 37, 1–6. Qi, Z., Chi, S., Su, X., Naruse, K., and Sokabe, M. (2005). Activation of a mechanosensitive BK channel by membrane stress created with amphipaths. Mol. Membr. Biol. 22, 519–527. Ralphs, J. R., Waggett, A. D., and Benjamin, M. (2002). Actin stress fibres and cell‐cell adhesion molecules in tendons: Organisation in vivo and response to mechanical loading of tendon cells in vitro. Matrix Biol. 21, 67–74. Raucher, D., and Sheetz, M. P. (1999). Characteristics of a membrane reservoir buVering membrane tension. Biophys. J. 77, 1992–2002. Rezzonico, R., Cayatte, C., Bourget‐Ponzio, I., Romey, G., Belhacene, N., Loubat, A., Rocchi, S., Van Obberghen, E., Girault, J. A., Rossi, B., and Schmid‐Antomarchi, H. (2003). Focal adhesion kinase pp125FAK interacts with the large conductance calcium‐activated hSlo potassium channel in human osteoblasts: Potential role in mechanotransduction. J. Bone Miner. Res. 18, 1863–1871. Riveline, D., Zamir, E., Balaban, N. Q., Schwarz, U. S., Ishizaki, T., Narumiya, S., Kam, Z., Geiger, B., and Bershadsky, A. D. (2001). Focal contacts as mechanosensors: Externally applied local mechanical force induces growth of focal contacts by an mDia1‐dependent and ROCK‐independent mechanism. J. Cell Biol. 153, 1175–1186. Sachs, F. (1992). Stretch‐sensitive ion channels: An update. Soc. Gen. Physiol. Ser. 47, 241–260. Salter, D. M., Robb, J. E., and Wright, M. O. (1997). Electrophysiological responses of human bone cells to mechanical stimulation: Evidence for specific integrin function in mechanotransduction. J. Bone Miner. Res. 12, 1133–1141. Sasamoto, A., Nagino, M., Kobayashi, S., Naruse, K., Nimura, Y., and Sokabe, M. (2005). Mechanotransduction by integrin is essential for IL‐6 secretion from endothelial cells in response to uniaxial continuous stretch. Am. J. Physiol. Cell Physiol. 288, C1012–C1022. Schliwa, M. (1982). Action of cytochalasin D on cytoskeletal networks. J. Cell Biol. 92, 79–91. Shakibaei, M., and Mobasheri, A. (2003). Beta1‐integrins co‐localize with Na, K‐ATPase, epithelial sodium channels (ENaC) and voltage activated calcium channels (VACC) in mechanoreceptor complexes of mouse limb‐bud chondrocytes. Histol. Histopathol. 18, 343–351. Sheetz, M. P., Sable, J. E., and Dobereiner, H. G. (2006). Continuous membrane‐cytoskeleton adhesion requires continuous accommodation to lipid and cytoskeleton dynamics. Annu. Rev. Biophys. Biomol. Struct. 35, 417–434. Shen, T., and Wolynes, P. G. (2005). Nonequilibrium statistical mechanical models for cytoskeletal assembly: Towards understanding tensegrity in cells. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 72, 041927. Shimaoka, M., and Springer, T. A. (2003). Therapeutic antagonists and conformational regulation of integrin function. Nat. Rev. Drug Discov. 2, 703–716. Sitharaman, B., Kissell, K. R., Hartman, K. B., Tran, L. A., Baikalov, A., Rusakova, I., Sun, Y., Khant, H. A., Ludtke, S. J., Chiu, W., Laus, S., Toth, E., et al. (2005). Superparamagnetic gadonanotubes are high‐performance MRI contrast agents. Chem. Commun. (Camb.) 31, 3915–3917. Small, D. L., and Morris, C. E. (1994). Delayed activation of single mechanosensitive channels in Lymnaea neurons. Am. J. Physiol. 267, C598–C606.

84

Matthews et al.

Sokabe, M., Sachs, F., and Jing, Z. Q. (1991). Quantitative video microscopy of patch clamped membranes stress, strain, capacitance, and stretch channel activation. Biophys. J. 59, 722–728. Solsona, C., Innocenti, B., and Fernandez, J. M. (1998). Regulation of exocytotic fusion by cell inflation. Biophys. J. 74, 1061–1073. Springer, T. A., and Wang, J. H. (2004). The three‐dimensional structure of integrins and their ligands, and conformational regulation of cell adhesion. Adv. Protein Chem. 68, 29–63. Stamenovic, D., and Coughlin, M. F. (1999). The role of prestress and architecture of the cytoskeleton and deformability of cytoskeletal filaments in mechanics of adherent cells: A quantitative analysis. J. Theor. Biol. 201, 63–74. Stamenovic, D., and Ingber, D. E. (2002). Models of cytoskeletal mechanics of adherent cells. Biomech. Model Mechanobiol. 1, 95–108. Stamenovic, D., Mijailovich, S. M., Tolic‐Norrelykke, I. M., and Wang, N. (2003). Experimental tests of the cellular tensegrity hypothesis. Biorheology 40, 221–225. Sukharev, S., and Anishkin, A. (2004). Mechanosensitive channels: What can we learn from ‘simple’ model systems? Trends Neurosci. 27, 345–351. Sukharev, S., and Corey, D. P. (2004). Mechanosensitive channels: Multiplicity of families and gating paradigms. Sci. STKE 2004, re4. Sukharev, S. I., Martinac, B., Arshavsky, V. Y., and Kung, C. (1993). Two types of mechanosensitive channels in the Escherichia coli cell envelope: Solubilization and functional reconstitution. Biophys. J. 65, 177–183. Sukharev, S. I., Blount, P., Martinac, B., Blattner, F. R., and Kung, C. (1994). A large‐ conductance mechanosensitive channel in E. coli encoded by mscL alone. Nature 368, 265–268. Sultan, C., Stamenovic, D., and Ingber, D. E. (2004). A computational tensegrity model predicts dynamic rheological behaviors in living cells. Ann. Biomed. Eng. 32, 520–530. Sung, L. A., and Vera, C. (2003). Protofilament and hexagon: A three‐dimensional mechanical model for the junctional complex in the erythrocyte membrane skeleton. Ann. Biomed. Eng. 31, 1314–1326. Suzuki, M., Naruse, K., Asano, Y., Okamoto, T., Nishikimi, N., Sakurai, T., Nimura, Y., and Sokabe, M. (1997). Up‐regulation of integrin beta 3 expression by cyclic stretch in human umbilical endothelial cells. Biochem. Biophys. Res. Commun. 239, 372–376. Takakuwa, Y., and Mohandas, N. (1988). Modulation of erythrocyte membrane material properties by Ca2þ and calmodulin: Implications for their role in regulation of skeletal protein interactions. J. Clin. Invest. 82, 394–400. Tang, Q. Y., Qi, Z., Naruse, K., and Sokabe, M. (2003). Characterization of a functionally expressed stretch‐activated BKca channel cloned from chick ventricular myocytes. J. Membr. Biol. 196, 185–200. Tavernarakis, N., and Driscoll, M. (1997). Molecular modeling of mechanotransduction in the nematode Caenorhabditis elegans. Annu. Rev. Physiol. 59, 659–689. Thurm, U. (1964). Mechanoreceptors in the cuticle of the honey bee: Fine structure and stimulus mechanism. Science 145, 1063–1065. Tzima, E. (2006). Role of small GTPases in endothelial cytoskeletal dynamics and the shear stress response. Circ. Res. 98, 176–185. Tzima, E., del Pozo, M. A., Shattil, S. J., Chien, S., and Schwartz, M. A. (2001). Activation of integrins in endothelial cells by fluid shear stress mediates Rho‐dependent cytoskeletal alignment. EMBO J. 20, 4639–4647. Tzima, E., Del Pozo, M. A., Kiosses, W. B., Mohamed, S. A., Li, S., Chien, S., and Schwartz, M. A. (2002). Activation of Rac1 by shear stress in endothelial cells mediates both cytoskeletal reorganization and eVects on gene expression. EMBO J. 21, 6791–6800.


85

Vera, C., Skelton, R., Bossens, F., and Sung, L. A. (2005). 3‐D nanomechanics of an erythrocyte junctional complex in equibiaxial and anisotropic deformations. Ann. Biomed. Eng. 33, 1387–1404. Vogel, V., and Sheetz, M. (2006). Local force and geometry sensing regulate cell functions. Nat. Rev. Mol. Cell. Biol. 7, 265–275. Walker, R. G., Willingham, A. T., and Zuker, C. S. (2000). A Drosophila mechanosensory transduction channel. Science 287, 2229–2234. Wang, N., Butler, J. P., and Ingber, D. E. (1993). Mechanotransduction across the cell surface and through the cytoskeleton. Science 260, 1124–1127. Wang, N., Naruse, K., Stamenovic, D., Fredberg, J. J., Mijailovich, S. M., Tolic‐Norrelykke, I. M., Polte, T., Mannix, R., and Ingber, D. E. (2001). Mechanical behavior in living cells consistent with the tensegrity model. Proc. Natl. Acad. Sci. USA 98, 7765–7770. Wendling, S., Canadas, P., and Chabrand, P. (2003). Toward a generalised tensegrity model describing the mechanical behaviour of the cytoskeleton structure. Comput. Methods Biomech. Biomed. Engin. 6, 45–52. Wilson, P. D. (2001). Polycystin: New aspects of structure, function, and regulation. J. Am. Soc. Nephrol. 12, 834–845. Wilson, P. D., Geng, L., Li, X., and Burrow, C. R. (1999). The PKD1 gene product, ‘‘polycystin‐1,’’ is a tyrosine‐phosphorylated protein that colocalizes with alpha2beta1‐ integrin in focal clusters in adherent renal epithelia. Lab. Invest. 79, 1311–1323. Wu, C. (2004). The PINCH‐ILK‐parvin complexes: Assembly, functions and regulation. Biochim. Biophys. Acta 1692, 55–62. Zanotti, G., and Guerra, C. (2003). Is tensegrity a unifying concept of protein folds? FEBS Lett. 534, 7–10. Zhang, Y., and Hamill, O. P. (2000). On the discrepancy between whole‐cell and membrane patch mechanosensitivity in Xenopus oocytes. J. Physiol. 523(Pt. 1), 101–115.

CHAPTER 4 Thermodynamics of Mechanosensitivity V. S. Markin* and F. Sachs{ *Department of Anesthesiology and Pain Management, UT Southwestern, Dallas, Texas 75235 { Physiology and Biophysical Sciences, SUNY BuValo, New York 14214

I. Overview II. Introduction A. General Equations III. Area Sensitivity A. Line Tension and Area Sensitivity B. Direct Observations of the EVect of Line Tension and Shape Transformation IV. Shape Sensitivity A. Experimental Observation of Shape Sensitivity V. Length Sensitivity and Switch Between Stretch‐Activation and Stretch‐Inactivation Modes A. Channel Activation by LPLs B. Other Parameters Regulating Switch Between Stretch‐Activation and Inactivation Modes VI. Thermodynamic Approach and Detailed Mechanical Models of MS Channels A. Detailed Mechanical Models VII. Conclusions References

I. OVERVIEW Mechanosensitivity of ion channels is conventionally interpreted as being driven by a change of their in‐plane cross‐sectional area AMSC. This, however, does not include factors relating to membrane stiVness, thickness, spontaneous curvature or changes in channel shape, length or stiVness. Because the open probability of a channel may be sensitive to multiple factors, we constructed a general thermodynamic formalism. These equations allow the Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58004-4

88

Markin and Sachs

analysis of mechanosensitive (MS) channels in lipids of diVerent geometric and chemical properties including hydrophobic mismatch at the boundary between the protein and lipid and the eVect of changes in the bilayer intrinsic curvature caused by the adsorption of amphipaths. The model predicts that the midpoint 1/2 and the slope1/2 of the gating curve are generally not independent, and on the basis of this relationship, we predicted the line tension at the channel/lipid border of MscL to be on the order of 10 pN suggesting that MscL channel is well matched to its lipid environment. For gramicidin, the theory predicts conversion from a stretch‐activated to a stretch‐inactivated gating as a function of bilayer thickness and composition.

II. INTRODUCTION Mechanosensitivity manifests itself in many physiological processes, and MS ion channels are prototype transducers that appear to be found in all species (Bass et al., 2002). As opposed to the prototypical family of homologous S4 voltage‐sensitive channels, these channels are a phenotypic family with no significant homology in sequence or structure, even for channels within Escherichia coli (Kloda and Martinac, 2001; Martinac, 2001; Bass et al., 2002; Perozo and Rees, 2003). Structural details and gating mechanisms proposed for these channels are extensively discussed in the literature (Hamill and Martinac, 2001; Betanzos et al., 2002; Anishkin and Sukharev, 2004; Chiang et al., 2004; Iscla et al., 2004; Sukharev and Anishkin, 2004), but here we concentrate on the general principles underlying mechanical transduction by ion channels (Markin and Sachs, 2004). There appear to be two general types of MS channels: those that receive stress from fibrillar proteins and those that receive stress from the lipid bilayer. The former are associated with the specialized receptors such as cochlear hair cells (Hackney and Furness, 1995), and touch receptors in Cenorhabditis elegans (Garcia‐Anoveros and Corey, 1996) and Drosophila (Walker, 2000). The channels stimulated by bilayer stress seem to be universally distributed (Sachs and Morris, 1998), although their physiological function is not generally known and do not seem to require an intact intra‐ or extracellular matrix in order to function (Hase et al., 1995; Suchyna and Sachs, 2004). It is this latter class of channels that we address in this chapter, although the approach can be generalized to include the specific receptor families. Mechanosensitivity probably appears in all membrane channels (Gu et al., 2001; Calabrese et al., 2002; Laitko and Morris, 2004; Morris, 2004) and transporters (Jutabha et al., 2003; Gradmann and Boyd, 2005) in the same way as their activity is modulated by membrane potential. The bacterial channels have nanosiemens conductances that do not significantly distinguish anions from cations, and are phenotypically and structurally

89

4. Thermodynamics of Mechanosensitivity

diVerent from the channels in eukaryotes such as the 2P (Patel et al., 2001; Honore and Patel, 2004) and TRP (Kim, 2004; Lin et al., 2005; Maroto et al., 2005) channels. The requirements for channels to become MS appear not to require specialized local structures such as utilized by the S4 voltage‐sensitive channel family but have probably evolved many times. Mechanosensitivity simply requires a significant change in channel dimensions between the closed and open states. The relevant stresses are global such as the far‐field tension and intrinsic curvature, or local such as amphipath modulation of boundary lipids (Martinac et al., 1990; Markin and Martinac, 1991; Hwang et al., 2003). These stresses interact with the changes in channel dimensions to change the relative energy of the closed and open states. There are three basic types of channel deformation that can change energies of the states (Fig. 1): change of in‐plane area, change of shape, or change of length normal to the membrane. If a channel increases its in‐plane A

B

C

Lipid

Channel

lc

Ω A L

L

FIGURE 1 Cartoon of three basic types of MS channel deformation during transition between closed and open states. (A) A change of area A occupied by the channel in plane of the membrane also changes the length L of the border between the channel complex and surrounding lipid where the line tension f resides; the ‘‘shape’’ of the complex does not change. (B) A change of shape of the MS expressed as a body angle O; average in‐plane area of the complex does not change. (C) Change of the length lc of the MS complex normal to the membrane; it can lead to changes in line tension f at the border with lipid.

90

Markin and Sachs

area A (case A) (Sachs and Lecar, 1991), the channel is stretch‐activated (SAC). If the area decreases, it is stretch‐inactivated (SIC) (Morris and Sigurdson, 1989). In another limiting case, an MS channel can change its shape, expressed as a body angle O, without changing its in‐plane area (case B). This type of mechanosensitivity is called shape sensitivity (Petrov and Usherwood, 1994; Petrov, 1999). The movement can be assisted (or resisted) by torque M produced by membrane curvature. The torque in the membrane can be produced either by global bending or by introduction of noncylindrical lipids such as lysolipids. The third limiting case of deformation (case C) is a change of the length lc of the channel without a change of in‐plane area or shape. This can result in hydrophobic mismatch between an MS channel and the surrounding lipid bilayer and is expressed as a line tension along the border with lipid. If the bilayer is stretched, or thinned with voltage, then its thickness decreases, changing the hydrophobic mismatch for open and closed states. If the energy of the closed state increases relative to the energy of open state, then the channel will tend to open under tension. This type of mechanosensitivity is called the length sensitivity. Natural MS channels may combine one or more of these basic deformations. A. General Equations Deformation of the channel is described by the Gibbs free energy that consists of three contributions: MS channel area, shape, and length 1 GA ¼ U0 þ BðAMSC A0 Þ2 þ fL gAMSC 2 G ¼ M

Gl ¼ fL

ð1Þ

Here B is the area stiVness (sometimes denoted KA), and A0 is the in‐plane area of the closed channel, and AMSC the in‐plane area of the open channel, f is the line tension at the border between the MS channel and surrounding lipids, L is the length of this border (Fig. 1). The shape of the MS channel can be described by the body angle O which changes in the transition and can be influenced by the total torque M acting on the MS channel. The area Gibbs energy GA describes the phenomenon of mechanosensitivity per se because it is directly related to membrane stretching. The body angle Gibbs energy GO is not necessarily related to membrane stretching and may occur without it. It describes ion channels that are not MS in conventional sense, but do respond to local mechanical stresses. Changing the

91

4. Thermodynamics of Mechanosensitivity Parameters of the MS channel Generalized forces

“Coordinates” Area: Arest, Aopen, Change: ∆A

Membrane tension: g

Perimeter: Lrest, Lopen, Change: ∆L

Line tension, f

Body angle: Ωrest, Ωopen, Change: ∆Ω

Torque, M

f L A

g

M

M Ω

FIGURE 2 Parameters of MS channels. The cartoon presents generalized coordinates (area, perimeter, and body angle) and generalized forces (membrane tension, line tension, and torque).

composition of the membrane can also alter gating of the channels. Therefore, there are three pairs of generalized coordinates and forces (Fig. 2): area–membrane tension, perimeter–line tension, and body angle–torque.

III. AREA SENSITIVITY In the absence of line tension, the gating of an MS channel is described as the transition between two energy levels corresponding to the closed and open states, 1 ¼ U0close þ Bclose ðAMSC Aclose Þ2 gAMSC ; Gclose 0 A 2 1 ¼ U0open þ Bopen ðAMSC Aopen Þ2 gAMSC Gopen 0 A 2

ð2Þ

Transitions occur between the two minima of these curves. Let us consider the case when the elasticity moduli of both states are equal to B (Fig. 3). In the closed state, the coordinates of the minimum of the energy curve close close are given by Aclose and Gclose ðg2 =2BÞ gAclose . 0 min ¼ ðg=BÞ þ A0 A;min ¼ U0 open and The same coordinates for the open state are Amin ¼ ðg=BÞ þ Aopen 0 open Gopen ðg2 =2BÞ gAopen . Therefore, in the transition from closed 0 A;min ¼ U0 to open, the area changes by,

92

Markin and Sachs G/kT 15

10

1

5 2 −2

2

4

6

8

A, nm2

3

−5

FIGURE 3 Plot of energy of the MS channel in closed (left part) and open (right part) states. Curve 1 is plotted in the absence of membrane tension, curve 2 corresponds to the midpoint of the transition, and curve 3 corresponds to the completely open state. open close Amin ¼ Aopen Aclose A0 0 min Amin ¼ A0 and the energy by, close GA;min ¼ Gopen A;min GA;min ¼ U0 gA0

ð3Þ ð4Þ

These equations and Fig. 3 show that with increasing tension, the minimum shifts down and to the right. Notice that the distance between the minima on the area axis A does not change, while on the energy axis it decreases, becomes zero, and then again increases in another direction. The open probability of the channel is defined by the Boltzmann function, popen ¼

1 1 þ expðG min =kTÞ

ð5Þ

and hence it increases with as shown in Fig. 4, changing from 0 to 1. The curves in Fig. 3 represent three characteristic states. Curve 1 is drawn in the absence of membrane tension, = 0. The energy diVerence between the two minima is large, and the channel is closed. On the Boltzmann curve (Fig. 4), this point is shifted far to the left. Curve 2 represents the situation when the two minima are at the same level, Gmin ¼ 0, and hence popen = 1/2. The corresponding membrane tension is designated 1/2. Curve 3 corresponds to high tension and is associated with the open channel. There are two parameters characterizing this function popen(): the midpoint of the transition 1/2 and the slope of the curve at this point (the slope sensitivity) equal to A0 =4kT.

93

4. Thermodynamics of Mechanosensitivity p open 1 0.8 0.6 0.4 0.2 9

10

11

12

13

14

15

g , mN/m

g 1/2 FIGURE 4

Open probability as a function of membrane tension.

The two‐state model with equal rigidity described the kinetics of MscL but runs into diYculties (Sukharev and Markin, 2001). The parameter of mechanosensitivity, A0 at the midpoint of the transition, was found to be 6 nm2. However, the X‐ray structure of the channel predicted that the diVerence of in‐plane area between the open and closed states should be about 20 nm2! This huge discrepancy has to be explained, and there are a few ways to do this. One is to assume diVerent elasticity moduli in the closed and open states. Sukharev and Markin (2001) assumed that Bclosed < Bopen. Then the transition parameter A is no longer independent of membrane tension: g 1 1 open open g ð6Þ Amin ¼ Amin ¼ þ A0 ¼ A0 B B close B open Figure 5 illustrates this situation. With increasing , the transition parameter Amin decreases. If in the beginning it was equal to 18.5 nm2, then at the middle point of transition it is about 6 nm2. We have to add that this explanation still utilizes the simple two‐state model, though more sophisticated alternatives were suggested (Sukharev and Anishkin, 2004). A. Line Tension and Area Sensitivity The area Gibbs free energy as a function of AMSC has a minimum at the point determined by the equation,

94

Markin and Sachs A G/kT 50 40 30 20 10

5 B

10

15

20

25

10

15

20

25

A, nm2

10 5 5

A, nm2

−5 −10 −15 −20 FIGURE 5 Energy plots of MS channel with diVerent elasticity in closed and open states. (A) No membrane tension; (B) midpoint of transition.

f BðAmin A0 Þ þ pffiffiffiffiffiffiffiffiffiffiffiffi g ¼ 0 pAmin

ð7Þ

If the energy associated with the line tension is small compared to that associated with change in area, then: Amin

g f þ A0 pffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffi B pB g þ A B

ð8Þ

0

Notice that Amin decreases with line tension f, as to be expected from compression of the channel at its periphery. The minimal value of G is given by,

95


Gmin ¼ U0 gA0

g2 2f þ pffiffiffi 2B p

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi g þ A0 B

ð9Þ

and the diVerence between two minima is: 0 1 g2 @ 1 1 A Gmin ¼ U0 gA0 2 Bopen Bclose 0 1 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2f @ g g A þ Aopen þ Aclose þ pffiffiffi 0 0 Bopen Bclose p

ð10Þ

Now, the open probability of the channel defined by the Boltzmann function is, popen ¼

1 1 þ expðGmin =kTÞ

ð11Þ

To further simplify the calculations, let us assume that the stiVness of the closed and open states are equal, that is, the Young’s moduli Bopen and Bclose are equal to B (but see Sukharev et al., 1999). Then the midpoint of the transition, when popen ¼ 1/2, is: sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ! U0 2f U0 U0 open g1=2 ¼ ð12Þ þ pffiffiffi þ A0 þ Aclose 0 A0 BA0 BA0 p A0 The slope of the curve at the midpoint is: S ¼ Slope1=2 ¼ 2 ¼

dp0 dg

3

g1=2 A0 U0 1 6 7 4A0 þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiqffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5 ð13Þ open 4kT B ðU =BA Þ þ A ðU =BA Þ þ Aclose 0

0

0

0

0

0

Notice that both the position of the midpoint (12) and the slope of the curve (13) depend on the line tension, and they both increase with increasing f. These equations contain a few characteristic parameters. The first equation can be transformed to, g1=2 ¼

U0 2f þ pffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi A0 p ðU =BA Þ þ Aopen þ ðU =BA Þ þ Aclose 0

0

0

0

0

0

ð14Þ

96

Markin and Sachs

The first term in this sum gives the characteristic membrane tension when the line tension f is zero: U0 g0 ¼ ð15Þ A0 By the analysis of the dimensions, one can establish that the denominator in the second term represents a characteristic length l, pffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi p ðU0 =BA0 Þ þ Aopen þ ðU0 =BA0 Þ þ Aclose ; ð16Þ l¼ 0 0 2 which converts line tension f to membrane tension . The physical meaning of this parameter is as follows. The ratio f =l gives the force compressing the MS channel. Therefore to open the channel, the membrane tension should be increased by this amount, and l is the eVective radius of a cylinder surrounding the channel, not necessarily a van der Waals enclosure. Finally, the tension at the midpoint of the transition can be presented as, g1=2 ¼ g0 þ

f l

ð17Þ

Analogously the slope of transition at the midpoint can be transformed to, A0 f ; ð18Þ S¼ 1þ 4kT lgs where the denominator contains a characteristic membrane tension, s ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi U0 U0 open close gs ¼ þ A0 B þ A0 B A0 A0

ð19Þ

Comparing two characteristic tensions (15) and (19), one can see that g s > g0

ð20Þ

These equations describe the role of line tension in the apparent area sensitivity. B. Direct Observations of the Effect of Line Tension and Shape Transformation MS channels can be reconstituted in diVerent lipid bilayers (Perozo et al., 2002; Moe and Blount, 2005) permitting the evaluation of local physical mechanisms for a role in MS channel gating: the hydrophobic mismatch at

97


the boundary between the protein and lipid and variations of bilayer intrinsic curvature. In addition, control of lipid composition permits altering f with heard group variation (Moe and Blount, 2005). The first mechanism can be attributed to the variation of the line tension around the molecule (area variation), and the second can be described by the body angle variation. Perozo et al. (2002) studied the bacterial wild‐type MscL in DPPC bilayers with monosaturated chains of 16, 18, and 20 carbons. They found that the midpoint of the gating transition and the slope of the transition in all three bilayers were diVerent. With increasing lipid chain length, there was a parallel increase of both the midpoint and the slope of the transition (Fig. 6), which is in correspondence with theoretical predictions from Eqs. (17) and (18) if we assume that the line tension increases with lipid thickness. The detailed analysis of the intrinsic parameters of the channel and their variation from one membrane to another was done in Markin and Sachs (2004). They used the normalized values of midpoint membrane tension rt = 1/2/ 1/2(16) and the slope of transition rs = s/s(16). Using Eq. (17), the first ratio of two tensions was presented as rt ¼

g1=2 ðnÞ 1 þ ½ f ðnÞ=lg0 1 þ gðnÞ ¼ ¼ g1=2 ð16Þ 1 þ ½ f ð16Þ=lg0 1 þ gð16Þ

ð21Þ

And the function g(n) was found as: gðnÞ ¼ 1 þ 0:5714ð0:0625n2 1:75n þ 13Þ

ð22Þ

p open

1 0.8 0.6 n=

16

18

20

0.4 0.2 0 0

20

40

60

80 100 q, mmHg

FIGURE 6 Open probability popen of MS channel as a function of the trans‐patch hydrostatic pressure q. Numbers on the curves indicate the length of the lipid chain.

98

Markin and Sachs

The experimental points are presented in Fig. 7A. To find the line tension at diVerent points, one needs to estimate the characteristic length l. Equation (16) can be simplified to: pffiffiffi qffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffi p l Aopen þ Aclose ð23Þ 0 0 2 If the cross section of the molecule were circular, then l could be expressed via the radii of the molecules in open and closed states: l ¼ pðropen þ rclose Þ=2 0 0 and is proportional to the average radius of the MS channel in the transition state. A g 1/2(n )/g 1/2(16) 3.5 3.0 2.5 2.0 1.5 1.0 15

16

17

18

19

20 n

21

16

17

18

19

20 n

21

B

n(n)/n(16) 2.5 2.25 2.0 1.75 1.5 1.25 1.0 15

FIGURE 7 Parameters of the MS channel as the functions of the length of the lipid chain. (A) Membrane tension corresponding to the midpoint of the transition. (B) Slope of the transition curve at the midpoint.


99

Parameter was l 8.8 nm, and the predicted line tension was f(18) 10.6 pN, and f(20) 25.6 pN. For comparison, the line tension of the hydrophobic edge of a phospholipid membrane against water is 120 pN, so the mismatch of the lipid at the MS molecule comprises only a fraction of the potential hydrophobic edge energy. The function f(n) also predicts the slope of the transition (Fig. 7B). The ratio of the slopes is given by: rs ¼

1 þ ½ f ðnÞ=lgs 1 þ gðnÞ=x ¼ 1 þ ½ f ð16Þ=lgs 1 þ gð16Þ=x

ð24Þ

with x = 1.5 and characteristic tension of s = 2.4 mN/m. In accordance with the theory, this value exceeds the resting tension 0 = 1.6 mN/m.

IV. SHAPE SENSITIVITY If the channel complex is shape sensitive, then its open probability will be aVected by torque M in the membrane: popen ¼ functionðMÞ. The torque is related to membrane curvature that can be aVected by applied pressure (Sokabe et al., 1991; Akinlaja and Sachs, 1998), for example, and intrinsic curvature. In particular, it can be generated by diVerent concentrations, cout and cin, of conical molecules in the two membrane leaflets (Fig. 8): M ¼ aðcout cin Þ

ð25Þ

If the shape of the molecule can be described by a single parameter—a certain body angle O, then its shape deformation energy can be presented in the same way as for area sensitivity: 1 G ¼ U0 þ K ð 0 Þ2 M ; 2

ð26Þ

where KO is the shape elasticity modulus. The energy of transition can be given by, G ;min ¼ U0 M 0

ð27Þ

The open probability, as before, is given by Eq. (11), which transforms to: popen ¼

1 1 þ expf½a þ bðcout cin Þ=kTg

ð28Þ

This is the simplest case when the shape can be eVectively described by a single parameter—body angle. This kind of channel is activated when the molecule is deformed in one direction and inactivated in another (Fig. 9A).

100

Markin and Sachs Torque in the membrane P

LP

Control PC18

Asymmetric LPC

Symmetric LPC, (compensation effect) FIGURE 8 Torque in the membrane. Torque is created by an asymmetric distribution of LPC in two membrane leaflets. Appearance of lysolipids creates torque even if the global curvature does not change. In other words, the eVect can be described as change of spontaneous curvature.

This is called one‐sided shape activation, and this phenomenon has been observed in vitro (Bowman and Lohr, 1996b). One can imagine another mechanism: the channel opens when the body angle is deformed in the both positive and negative directions (Fig. 9B). It is two‐sided shape activation, and the open probability in the first approximation can be presented as, popen ¼

1 1 þ expf½a þ bðcout cin Þ2 =kTg

ð29Þ

A. Experimental Observation of Shape Sensitivity It was reported in a number of papers that addition of charged amphiphiles, or lysophospholipids (LPL), to bilayers containing MS channels, dramatically lowered the activation threshold. With MscL, externally applied lysophosphatidylcholine (LPC) strongly favors the open state. The addition of LPC (1.5 mM), in the presence of a small transbilayer pressure, produces a pronounced increase in MscL activity. Moreover, once the pressure is released, a large fraction of the channels remain constitutively open. Even more

101

4. Thermodynamics of Mechanosensitivity A popen

Ω B popen

Ω FIGURE 9 Shape sensitivity of the MS channel. The MS molecule deforms in the positive direction (O > 0 or upper side is larger than lower side) and/or the negative direction (O < 0 or upper side is smaller than lower side). If the channel can open only when deformed in the positive direction, this is one‐sided shape sensitivity (panel A). Another possibility is that it opens in both directions; this is two‐sided shape sensitivity (panel B).

remarkably, at larger LPC concentrations (3 mM), MscL activity gradually increased with time in the absence of any applied pressure. It is important to remember, however, that there is always a resting tension present due to the energy of adhesion of the lipid for the glass (Opsahl and Webb, 1994). The LPC eVects suggest a second type of mechanosensitivity. The asymmetric addition of conically shaped LPC in the outer monolayer generates the torque that aVects the transition of MscL. At the same time it suggests that the gating of MscL is accompanied by a shape change. Remarkably, Perozo et al. (2002) found that if LPC is added symmetrically to both monolayers, the gating does not change. This is explained by the fact that LPC in two diVerent leaflets produces opposite eVect because min ¼ mout and hence M ¼ 0, and there is no eVect on the energy of the channel. Given that MscL is not symmetric along the membrane normal, the eVects of LPC may be asymmetric also. However, preliminary data suggest that if

102

Markin and Sachs

LPC was added to the inner monolayer only, the channel would also be activated (B. Martinac, private communication). If true, that would make the channel belong to the two‐sided shape activation family. An example of one‐sided shape activation channel was found by Maingret et al. (1999, 2000). They studied the MS Kþ channels TREK‐1 and TRAAK, and found that they too can be activated by LPLs and other amphipaths. LPC activation was a function of the size of the polar headgroup, and length of the acyl chain, but independent of the charge. These channels, which are found commonly in the central nervous system, are also opened by inhalation anesthetics such as chloroform, ether, halothane, and isoflurane at clinically relevant concentration (Patel et al., 1999). The authors proposed that activation of these Kþ channels may form the basis of general anesthesia. Perhaps MS channels evolved as amphipath detectors (Patel et al., 2001), and only later the far‐field mechanosensitivity became useful. While amphipaths can cause changes in global membrane curvature, that eVect appears to be a correlation rather than the cause of changes in MS channel gating. Amphipaths probably act locally (Suchyna et al., 2004b). We should not expect that global bilayer curvature (Sukharev et al., 1999; Moe and Blount, 2005) can change channel gating. For molecular sized objects like channels, the available energy input from changes in global curvature (radius of curvature > 1 mm) is well below kTB, and thus cannot have a significant eVect on gating. Indeed, if the in‐plane area of the MS molecule is A, bending rigidity is KB and the radius of curvature is Rp, then its bending energy is Ebend ¼ 12 KB ð2=Rp Þ2 A. Bending rigidity of MS channel is unknown so we substitute for it the bending rigidity of the lipid bilayer. For freely sliding monolayers KB can be estimated (Markin and Albanesi, 2002) as 0.8 1019J ¼ 20 kT. Taking A ¼ 30 nm2 and Rp ¼ 1 mm, one can find Ebend ¼ 1.2 103 kT. This is indeed a small amount. Even if to vary the parameter in this equation in reasonable range, one cannot approach 1 kT! There is no data to suggest that MS channels are, in fact, sensitive to global bilayer curvature (Lee, 2006), but there is evidence for membrane curvature sensitivity in biological membranes. The 2P Kþ channels, and some cationic MS channels, demonstrate pronounced curvature sensitivity, activating with curvature away from the cell (Maingret et al., 2000, 2002) or toward the cell (Bowman and Lohr, 1996a). Even the breaking strength of biological membranes is sensitive to the sign of the curvature, being stronger when bent toward the cytoplasm (Akinlaja and Sachs, 1998). This curvature sensitivity reminds us to be cautious applying simple homogenous physical models to the heterogeneous and anisotropic biological membranes.

103


V. LENGTH SENSITIVITY AND SWITCH BETWEEN STRETCH‐ACTIVATION AND STRETCH‐INACTIVATION MODES Gramicidin is a wonderful example of a channel that does not change either in‐plane area or shape. From a conventional point of view, it should not be MS. Nevertheless, it can be activated by either membrane stretch or membrane torque. In addition, it can switch between activation and inactivation modes (Hamill and Martinac, 2001; Martinac and Hamill, 2002). As an example of the utility of the general thermodynamic approach, we will analyze gramicidin. Gramicidin A (gA) forms dimer channels, with the gA in one monolayer binding to a mate in the opposite monolayer (Andersen et al., 1996) (Fig. 10). Other than head–head dimer formation, gA does not change its conformation significantly between open and closed, and still it is strongly influenced by mechanical stresses in the bilayer. The allosteric parameter governing activation and inactivation is membrane thickness. The process of channel formation is described by the dimerization reaction between two monomers (m) from adjacent lipid leaflets (Lundbaek and Andersen, 1994). If the numbers of monomers in two leaflets are equal (Nm) and the number of dimers is Nd then in equilibrium, 2 k N ¼0 kþ1 Nm 1 d Nm þ Nd ¼ Ntot

ð30Þ

where Ntot is total amount of gramicidin in each leaflet and kþ1 and k1 are the reaction rates. Solution of these equations is

No tension A

Membrane tension B

h

lg g

C

g

D

g

g

FIGURE 10 Gramicidin A in the lipid bilayer: formation of dimers causes deformation of the bilayer.

104

Markin and Sachs

Nd ¼

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2KD Ntot þ 1 4KD N þ 1 ; 2KD

ð31Þ

where KD ¼ kþ1 =k1 is the dimerization constant. If KD Ntot 1, which is 2 . The dimerization constant includes the usually the case, then Nd KD Ntot free Gibbs energy of transition: G ; ð32Þ KD ¼ K exp kT where K* is the pre‐exponential coefficient not depending on temperature. The free Gibbs energy of transition can be found in the following way. If the length of the dimer does not coincide with the thickness of the hydrophobic core of the bilayer (Fig. 10A and D), then the hydrophobic mismatch with a formed channel will cause deformation of the bilayer with positive or negative local curvature (Hladky and Haydon, 1972). As demonstrated above, this deformation is equivalent to a line tension around the dimer, changing its energy relative to the two monomers. So the Gibbs free energy diVerence between these two states can be presented as G ¼ fL Gass ;

ð33Þ

where L is the length of the perimeter and Gass is the component of the energy of association that does not depend on membrane thickness. The degree of the deformation of the membrane depends on the relationship between the thickness of the monolayer h and the length of gramicidin lg. Their diVerence m ¼ lg h determines the value of the deformation energy and hence the line tension; in the Hookean approximation it can be presented as f ¼ km ðh lg Þ2 ;

ð34Þ

where km is the proportionality coeYcient. The applicability of the Hookean approximation to this case was discussed by Lundbaek and Andersen (1999) and Lundbaek et al. (1997) who demonstrated that the deformation energy can be quantified based on a linear spring description. When membrane tension is applied, the membrane area Am increases. Due to the volume incompressibility of lipids (Ah ¼ constant), the thickness of monolayers h decreases: h Am g ¼ ¼ h0 KA Am

ð35Þ

where h0 is the monolayer thickness in the absence of membrane tension, and KA is the elasticity modulus of the lipid bilayer. Then the line tension is


g 2 f ¼ km h0 lg h0 KA and the Gibbs energy can be presented as: lg g 2 Gass G ¼ Lkm h20 1 h0 KA

105 ð36Þ

ð37Þ

where L is the external perimeter of gramicidin. The number of dimers is Nd ¼

2 K Ntot exp

" # Lkm h20 lg G Gass g 2 2 exp 1 ¼ K Ntot exp kT kT h 0 KA kT ð38Þ

One can introduce the maximum number of dimers that can be formed by 2 membrane stretching Ndmax ¼ K Ntot expðGass =kTÞ and can define the ratio max popen ¼ Nd =Nd as the open probability: " 2 # Lkm h20 lg g 1 popen ¼ exp ð39Þ kT h0 KA The definition of this parameter as an open probability is not very rigorous because the channels do not exist a priori, but rather are formed in the process of stretching. It might be better to call it the degree of activation. Nevertheless, we shall use this definition in this section, because the number of open channels cannot exceed Ndmax . However, in the next section this parameter will be discussed in more details. The degree of activation or the open probability (39) is a function of two variables: popen ¼ popen (h0, ) (Goulian et al., 1997). With increasing membrane tension, the popen can decrease or increase, depending on the thickness of the monolayer, h0. The phase space [h0, ] presented in Fig. 11 is divided into two parts corresponding either to stretch‐ inactivation of the channels, or stretch‐activation. Interestingly enough, if h0 lg then application of membrane tension causes inactivation of the channels. However, if h0 > lg the behavior is more complicated. Small tensions cause activation of the channels, but after exceeding some critical value corresponding to a point at the curve in Fig. 11, the channels become SIC. Open probability, corresponding to these two cases is presented in Fig. 12A. The general case is presented in three‐dimensional plot in Fig. 12B.

106

Markin and Sachs h 0 /l 2 1.8

Stretch-activated

1.6 1.4 1.2 1.0

Stretch-inactivated 0.1

0.2

0.3

0.4

0.5

g/K A FIGURE 11 Phase space [h0, ] is divided into two parts corresponding either to stretch‐ inactivation of the channels or to stretch‐activation. The curve gives the maximum of the open probability presented by Eq. (39).

If membrane tension can influence formation of dimers, then there should be a force normal to the membrane that pulls the monomers inside the bilayer or pulls them apart disrupting the channel. This force is exerted by the deformed border area of lipid monolayers as presented in Fig. 10. The value of this force Fnormal can be found from the energy of deformation (37) and (34): dG g ¼ 2Lkm ðlg hÞ 2Lkm lg h 1 Fnormal ¼ ð40Þ dh Km The sign of the force is selected in such a way that positive force (lg h > 0) compresses the dimers inside the bilayer, as in Fig. 10D, while negative forces (when lg h < 0) pull them apart, as in Fig. 10A. Both positive and negative forces prevent formation of the channels. This is obvious in the thick bilayer where this force pulls the monomers apart. But it is also true in thin bilayers, where the force pulls the monomers inside, because the energy of the dimer protruding from the membrane is higher than fitting deeper side by side. This can also be viewed as the thin membrane preventing channel formation because the interior faces of the gA cannot interact since the gA monomers bump into each other side‐by‐side. Global membrane tension changes these forces by changing the thickness of the monolayer h; it decreases the magnitude of the negative force in Fig. 10A and eliminates its destructive influence on channel formation. This stretch‐activation situation happens in thick monolayers. In thin monolayers, we have the opposite case:

107

4. Thermodynamics of Mechanosensitivity A popen 1.0 0.8

hl = 1.15

0.6 0.4 0.2 hl = 1.0 0 0

0.05

0.1 g/K A

0.15

0.2

B

p open 1 0.75 0.5 0.25 0

0.2 0.15 0.1

0. 0.9

0.05

1 h0 /l

g/K A

1. 1.1 0

FIGURE 12 Open probability P0 as a function of two dimensionless variables h0 =lg and g=KA according to Eq. (39). For illustrative purposes, we arbitrarily selected Lkm h20 =kT ¼ 300. (Panel A) Two curves corresponding to h0 =lg ¼ 1 and h0 =lg ¼ 1:1. These are cross sections of the surface presented in panel B. (Panel B) Three‐dimensional plot of function (39).

The force does not decrease but rather increases and prevents formation of dimers. Therefore, the general rule is that in thin membranes, gA channels should be SIC, while in thick membranes they are stretch‐activated at small tensions, but switch polarity at high tensions. This diVerence of polarity in thin

108

Markin and Sachs

and thick bilayers was observed (Hamill and Martinac, 2001), but Eq. (39) predicts more than that: It says that the switch from stretch‐activation to stretch‐inactivation can occur in the same membrane at suYciently large tensions. However, channel opening usually occurs at relatively high membrane tensions. Therefore, to observe the curve marked 1.15 in Fig. 12A, one might need tensions greater than the lytic limit (10 mN/m). However, the lytic strength of membranes depends on duration of the stimulus, so that short duration stimuli can apply much higher tensions without lysis (Evans et al., 2003). We leave this prediction for the experimentalists. A. Channel Activation by LPLs Another interesting question is the role of membrane torque generated by LPLs. We demonstrated above that if the channel changes shape during the transition between two states, then the addition of LPLs can either facilitate or inhibit this transition. Gramicidin channel does not change shape so that torque should not have direct eVect. However, if there is a hydrophobic mismatch as in Fig. 12A, then the monolayer bends with a positive curvature. If LPLs are added, they can generate a positive torque, and facilitate monolayer bending in the same direction. They should facilitate stretch‐activation of the gramicidin channels and activate the channels in the absence of far‐field tension. Lundbaek and Andersen (1994) demonstrated that LPLs can increase the dimerization constant of membrane‐bound gramicidin up to 500‐fold (at the concentration of 2 mM). They found that the relative potency increases as a function of the size of the polar headgroup but does not depend on headgroup charge. It also depends on the channel length: as the channel length is decreased, the potency of the LPC increased. Membrane curvature is extensively employed for explanation of mechanical membrane transformations in the process of membrane fusion, fission, and poration (cf. Markin and Albanesi, 2002; Tamm et al., 2003). The key idea is that bending energy per unit area of a monolayer is determined by the diVerence between the actual geometric curvature of the monolayer C, and its spontaneous curvature C0 as 1 Ecurvature ¼ kC ðC C0 Þ2 ; 2

ð41Þ

where kC is the bending modulus or curvature elasticity. In the region of hydrophobic mismatch, where deformation occurs, this quantity should be compared with the energy of the initial, planar monolayer so that the elastic energy change will be:


109

1 1 1 Ecurvature ¼ kC ðC C0 Þ2 kC C02 ¼ kC CC0 þ kC C 2 ð42Þ 2 2 2 The mean geometrical curvature of the monolayers near a gramicidin channel is determined by the diVerence between monolayer hydrophobic thickness and the length of gA monomers. As a first approximation, the mean curvature can be presented as C ¼ aðh0 lg Þ. Spontaneous curvature is created by the lysolipids and should be proportional to their concentration, C0 ¼ bcLPL . The proportionality coeYcient can be positive, as for lysolipids (positive spontaneous curvature), or negative for other amphipaths. The change of elastic energy of each monolayer (42) near the channel gives the contribution to Gibbs free energy of transition between closed and open states: G ¼ Arim kC ½a2 ðh0 lg Þ2 2abðh0 lg ÞcLPL Gass ;

ð43Þ

where Arim is the area of the distorted lipid rim around gramicidin. According to Eq. (38), the number of dimers is given by " ( #) Arim a2 kC h20 2b cLPL Gass lg lg 2 2 Nd ¼ K Ntot exp exp 1 1 kT kT ah0 h0 h0 We shall define the characteristic amount of dimers " # 2 k ðh l Þ2 G A a ass C 0 g rim 2 Nd0 ¼ K Ntot exp kT

ð44Þ

ð45Þ

that can be formed in the absence of lysolipids (cLPL ¼ 0) and introduce the degree of activation of channels by lysolipids: lg lg Nd 2Arim abkC h0 cLPL cLPL exp 0 qact ¼ 0 ¼ exp 1 1 kT h0 h0 Nd cLPL ð46Þ This equation introduces a characteristic concentration of lysolipids c0LPL ¼

kT 2Arim abkC h0 cLPL

ð47Þ

that determines the potency of the given type of LPL to activate the channel: the lower c0LPL the higher is its potency. Lundbaek and Andersen (1994) observed that LPI and LPC are the most potent activators; at concentration of 2 mM they increased the number of open channels 630‐ and 450‐fold, respectively, while LPE and LPS at the same concentration produce 80‐ and

110

Markin and Sachs

40‐fold increase. This is related to the shape of lysolipid molecules: The shape is conical and the angle of this cone is determined by the size of their polar headgroups which go along with this series. The size of the headgroups is determined not only by the atoms constituting the polar heads but the hydration of these heads. PC headgroups are much more hydrated than PE and PS headgroups. So the shape of LPL molecule accounts for the potency of diVerent LPLs. The reason we call qact the degree of activation rather than open probability is that its value can exceed 1. Figure 13 presents function (46) for three diVerent values of monolayer thickness. In thick membranes (h0 > lg) lysolipids activate channels even without membrane tension and qact goes up. In thin membranes (h0 < lg) lysolipids inactivate channels and qact goes down, and if the length of the gramicidin monomer coincides with the hydrophobic thickness of the monolayer there is no eVect of LPL: qact remains constant and equal to 1. B. Other Parameters Regulating Switch Between Stretch‐Activation and Inactivation Modes The change of polarity of MS channels was described above in the example of gramicidin that does not preexist as a channel but rather forms qact

4 3

hl = 1.1

2 hl = 1.0

1

hl = 0.9 0

2.5

5.0

7.5

10.0

12.5

15.0

cLPL/ c 0LPL FIGURE 13 The degree of activation of gramicidin channels by LPLs. The curves are plotted according to Eq. (46). Channels are activated in thick membranes (h0 > lg) and inactivated in thin membranes (h0 < lg). The border line (h0 ¼ lg) goes horizontally showing no eVect in this case.


111

in process of mechanotransduction. But this phenomenon can also be observed in the preexisting channels where activation can be regulated by additional allosteric factors such as membrane potential. A well‐known example is Shaker‐IR, a voltage‐gated Kþ channel. It can exhibit a rich behavior including transition from stretch‐activation to stretch‐inactivation depending on the value of the membrane potential (Gu et al., 2001). The mechanism of this phenomenon is not known and Gu et al. (2001) discussed a multistate scheme where some states with diVerent time scale can play decisive MS role. Stretch‐activation occurs in the channels that have rather low popen at rest and stretch‐inactivation at high popen. This was demonstrated above with gramicidin channel. In Shaker, where electrical potential causes a shift of open probability, stretch tends to open closed channels and close open channels suggesting that an intermediate sate has significantly diVerent geometry than the end states (Tabarean and Morris, 2002). The shift from SAC to SIC behavior has been shown in 2P Kþ channels (Honore et al., 2006) and in MS channels from dystrophic muscle compared to normal muscle (Franco‐Obregon and Lansman, 1994). However, it has been shown that SIC behavior can be produced by SACs that are under tension at rest (Honore et al., 2006). Patches can be stressed by the cytoskeleton pulling the membrane toward the tip. Then suction can flatten the membrane reducing tension and closing the channels—SIC behavior. Thus, the diVerence between the dystrophic and normal muscle behavior may represent diVerences in cytoskeletal structure rather than channel structure (Suchyna and Sachs, 2007).

VI. THERMODYNAMIC APPROACH AND DETAILED MECHANICAL MODELS OF MS CHANNELS When describing the mechanosensitivity phenomenon, we use the general thermodynamic approach which is free of specific model assumptions. The model‐free thermodynamic approach is a powerful method permitting to establish the relationship between generalized forces and reactions of the system. A disadvantage is that it does not consider how mechanical forces gate the channel and leaves apart all parameters and specific properties of real systems. These properties can be found only in experimental studies and accounted for by mechanical models that need to include mechanics of the patch itself (Chiang et al., 2004; Suchyna and Sachs, 2004; Honore et al., 2006), and to include the heterogeneous distribution of mechanical stresses between the bilayer and the cytoskeleton.

112

Markin and Sachs

Hamill and Martinac (2001) provided a list of specific models used by diVerent authors to describe the behavior of MS channels, focusing on MscL. The first in this list is the multimerization model. It considered the tension‐sensitive recruitment of MscL monomers into a multimeric pore. The recruitment occurs because the energy of the complex under tension is lower than the energy of separated monomers. In the thermodynamic approach, the additional energy (that can be either positive or negative) of the complex was associated with interaction energy at the border between the channel complex and surrounding lipid. As is customary in two‐dimensional (2D) thermodynamics, this energy was described as a line tension. If the line tension decreases with area tension, then it favors formation (opening) of the channel, which will be stretch‐activated. The multimerization model probably does not apply to MscL. One serious objection is the very rapid opening transition (i.e., 1 kHz with a phase lead of 60 –120 , as expected of a displacement current. Tip displacement was outward with depolarization, meaning a positive sign of flexocoeYcient. From the estimations made in Mosbacher et al. (1998), one could infer a value of 1019 C for the flexocoeYcient of HEK293 membrane. This value is lower than the locust muscle one, which is not surprising in view of the marked mechanoelectrical behavior of the muscle membrane. 4. Converse FlexoeVect of Native Membranes at Pulsed Electric Excitation Further experimental results using pulsed excitation were obtained (Zhang et al., 2001). Experiments were performed with whole‐cell voltage‐clamped HEK293 cells. The cell membrane was indented using the AFM cantilever of a

137

5. Flexoelectricity and Mechanotransduction

modified Quesant Nomad AFM with a force of typically 0.5 nN. For each experiment, the authors averaged 20 repetitions of the cantilever displacement associated with hyperpolarizing or depolarizing pulses (from a holding potential of –60 mV). The displacement was taken as the average of 5 ms about the peak. Positive displacements represent the AFM moving into the cell. Rectangular pulses of linearly increasing amplitude produced membrane displacements in linear proportion to the voltage pulse amplitude ( 1 nm/ 100 mv) (Fig. 5). Hyperpolarizing the membrane would increase the local curvature around the tip as it is moving inward, that is, positive sign of the flexocoeYcient of HEK293 membrane as in Mosbacher et al. (1998) would be confirmed at normal ionic strength. Interestingly, a sign reversal was found at lower ionic strengths of the bath, below 10 mM. It is important to note that membrane movement and ionic current were uncorrelated, suggesting the motor mechanism is not electroosmotic (cf. Fig. 5B and C). Zhang et al. (2001) oVered an explanation of this novel finding in terms of Lippmann equation for membrane tension in the presence of electric field.

A

0 mV 30 mV 60 mV 90 mV

−80 −100 −120

1500 1000 500 0 −500

−1000 −1500

−140 −160 −20 0

−2000 −20 0

20 40 60 80 100 120 140 160

20 40 60 80 100 120 140 160

Time (ms)

C 8 7 6 5 4 3 2 1 0 −1 −2 −3 −4 −5 −6 −7 −8 −20

0 mV 30 mV 60 mV 90 mV

D 1

Displacement (A)

Displacement (A)

0 mV 30 mV 60 mV 90 mV

2000

Current (pA)

Voltage (mV)

−60

B

0

20

40

60

80 100 120 140 160

Time (ms)

0 −1 −2 −3 −4 −160

−140

−120

−100

−80

−60

Potential (mV)

FIGURE 5 EVect of voltage on membrane movement of a patch‐clamped HEK 293 cell. Steady‐state displacement vs command voltage. (A)–(C) show the membrane potential, ionic current, and membrane movement from the same cell (average of 20 sweeps). (D) illustrates that the movement is linear with voltage. From Snyder et al. (2002, Fig. 8, p. 448) with permission from the authors and the publisher.

138

Alexander G. Petrov

While in a symmetrically charged membrane Lippmann eVect predicts tension changes that are quadratic with respect to transmembrane voltage, Zhang et al. (2001) have shown for the first time that with asymmetrically charged membranes the Lippmann tension that is a sum of the two interfacial tensions displays a leading term that is linear with respect to the voltage, thus resembling flexoelectricity. Actually, flexoelectric torque by converse flexoeVect is roughly governed by the diVerence of the two interfacial tensions rather than by their sum. Such a diVerence will ultimately induce a surface torque that will curve the membrane. The Poisson–Boltzmann type of treatment (Zhang et al., 2001) shows that this diVerence will also be linear with voltage, thus providing a model of the monopole contribution to the flexoeVect that depends on surface charges alone. The actual symmetry of the AFM problem is such that both tension and torque variations of the distended membrane will move the AFM tip. 5. Flexoelectricity in Channel‐Containing Model and Native Membranes Locust muscle membrane contains two types of voltage‐activated channels, as found in the studies of Kþ‐selective channels in the membrane of adult locust muscle. The two types of Kþchannels of maximum conductance

150 pS (BK‐channel) and 35 pS (IK‐channel) described by Gorczynska et al. (1996) were found to be also stretch‐sensitive (Petrov et al., 1992; Mellor et al., 1999). The IK‐channel displayed a monotonic reversible increase of its open probability when the transmembrane pressure was raised (Fig. 6). In contrast, during recordings from BK‐channels (observed only under high Kþ concentration in the pipette), an increase of transmembrane pressure triggered the channel to its open state in an irreversible (or only slowly reversible), cumulative fashion (Fig.7). In confirmation to the earlier results (Petrov et al., 1989), we have repeatedly observed in locust, patches containing these Kþ channels that are also electrically gated, a dramatic amplification of direct flexoelectric response during channel opening (Petrov et al., 1993) (Fig. 8). Channel opening was aVected by applying transmembrane voltages larger than 20 mV, a property characteristic of IK channels. Increasing in a stepwise manner, the holding potential of the voltage clamp up to 50 mV we have observed, apart from the instant jumps of flexocurrent on a fast time scale (inset, 40 mV), a steady amplification of the first harmonic current amplitude of more than 50 times on the slow time scale of minutes. Initial very low flexocurrent rms value of 5 fA can be fully recovered with holding potential brought back to zero, and could be reamplified with larger negative holding potentials that also open the IK channels. This striking eVect could be comprehended as a transition


139

FIGURE 6 (A–D) EVect of pressure ramps on a cell‐attached patch containing a single IK channel. The pipette contained low‐Kþ saline and the patch was held at Vpip ¼ 70 mV. (A) Data from two pressure ramps of ‘‘0 torr to 50 torr to 0 torr.’’ The total duration of a ramp was 75 s, that is, the rate of change of pressure was 1.33 torr/s. There were no rest periods between the ramps. (B) Average patch current expressed as a function of pressure. Current data were sampled into 380 pressure bins (0.13 torr each) and averaged over seven rising pressure ramps (0 to 50 torr). The solid curve represents a fit by a Boltzmann function, with the P50% ¼ 23 torr. (C) A plot of ln [ p0/(lp0)] vs P. p0 was calculated by dividing the averaged current (Fig. 6) by the unitary channel current, Iunit ¼ 2.3 pA. The solid line represents a linear regression fit to the data (slope ¼ 0.235 0.005 torr1, correlation coeYcient ¼ 0.94). (D) Higher time resolution traces from the rising phase of the second ramp in (A) at pressures of –15, –30, and –50 torr. Figure reprinted with kind permission of Springer Science and Business Media from Mellor, I. R., Miller, B. A., Petrov, A. G., Tabarean, I., and Usherwood, P. N. R. (1999). Eur. Biophys. J. 28, 346, Fig. 2. Copyright (1999) by the European Biophysical Societies’ Association.

from detailed to global neutrality regime of the monopole flexopolarization during channel opening (Petrov, 1999): curvature‐induced transport of charges along open channels and across the whole membrane thickness leaves each one of the half‐spaces charged with respect to the other one and creates very large dipoles, situated at the same time across the low dielectric constant membrane core; therefore the enhancement of flexocoeYcient can reach two orders of magnitude depending on the electric coupling parameter H (Petrov, 1999).

140

Alexander G. Petrov A

Vpip = +45 mV −17.0 torr

0 torr

0 torr

8 pA

Vpip = +25 mV −14.8 torr

0 torr

B

0 torr

200 ms

5 Down

Vpip = +45 mV

Up

In (P0/1−P0)

0

Down

Down

−5

Vpip = +25 mV

Up

−10

Up

0

10

20 30 Suction pressure (torr)

40

50

FIGURE 7 Dependence of open probability po of a single BK‐channel on static pressure diVerence. Cell‐attached patch, high Kþ pipette. (A) Example of current traces at Vpip ¼ þ25 and þ45 mV and successive pressures of 0 torr, 14.8 (17.0) torr, and again 0 torr. (B) Log of open probability vs pressure for the two pipette potentials following initial increase and subsequent decrease of pressure. Pressure triggers the BK‐channel to activity levels, which are voltage‐dependent, but no longer pressure‐dependent. From Petrov (1999, Fig. 8.20) with the permission of the publisher.

Demonstration of the role of channels on the converse flexoeVect was performed by transfecting HEK cells with voltage‐gated Kþ (Shaker) channels (Beyder, 2005). The voltage‐induced membrane deviations’ (VD) shape of the Shaker‐transfected HEK cells (ShHEK) is notably diVerent than the VD of wtHEK and AchR HEKs (Fig. 9). The diVerence can be seen as a marked nonlinearity at the point of channel activation. For both on and oV steps, in the hyperpolarized portion of the VD, ShHEK membrane


141

FIGURE 8 Amplification of flexocurrent (If) during channel opening. Cell‐attached patch of locust muscle membrane. The rms amplitude of If (lower trace) was recorded at various values of Vhold (upper trace) on a slow time scale (note that time in this record runs from right to left). Direct flexoresponse was driven by pressure oscillations of 10 torr(pp), 20 Hz. Pipette resistance was 9 MO, seal resistance was 1 GO. Insets show pressure and flexocurrent in real time at indicated holding potentials. Opening of Kþ channels occurred mainly at positive holding potentials. Because of the relationship between flexopolarization and channel state, channel openings and closings could be resolved at 40 mV by sudden changes of the amplitude of flexocurrent. Figure reprinted with kind permission of Springer Science and Business Media from Petrov, A. G., Miller, B. A., Hristova, K., and Usherwood, P. N. R. (1993). Eur. Biophys. J. 22, 289. Copyright (1993) by the European Biophysical Societies’ Association.

displacement increases linearly with the voltage step size and the force clamp. However, from the voltage where Shaker channel activates (–40 to –20 mV), the VD traces for the on and oV steps take a strong turn and nearly saturate in displacement (Fig. 9B). Such nonlinearity was observed in 83% (19/23) of the experiments, and the experiments lacking nonlinear behavior were performed at lowest force clamp, where mechanical noise often

142

Alexander G. Petrov

A

B 1.0

ShIR (5) wtHEK (1) Fhold = 0.1 nN

Lever displacement (nm)

Lever displacement (nm)

0.8 0.6 0.4 0.2 0.0

−120

−90

−60

−30

Vm (mV)

0

30

0.8

wtHEK (2) ShIR (3) Fhold = 0.2 nN

0.6 0.4 0.2 0.0 −150

−100

−50

0

50

Vm (mV)

FIGURE 9 Voltage‐induced membrane movements of channel‐containing membranes: Shaker‐transfected HEK cell (ShHEK) vs wild‐type human embryonic kidney (wtHEK) cell. Movement of the cantilever engaged onto a voltage‐clamped ShHEK (upper trace) was larger than movement of the cantilever engaged onto a wtHEK membrane (lower trace) for force‐ clamp of 100 pN (A) and 200 pN (B). Values in parentheses are the number of traces averaged. From Beyder (2005, Fig. 14, p. 49) with permission from the author.

dominates the displacement recording. This nonlinear behavior was noted at all force clamp values, down to 50 pN (Fig. 9A). This suggests that the common linear increase in movement of the cantilever drastically diminishes after the activation of Shaker channels. The basic idea we oVer here for explanation of these results is that open channels switch the regime of flexoelectric polarization from detailed to global electric neutrality (detailed electric neutrality: each halfspace on both sides of the membrane is neutral in itself; global electric neutrality: both halfspaces are oppositely charged with respect to one another). In the case of converse flexoeVect, detailed and global neutrality regimes should be discussed with respect to membrane‐related charges only, that is, charges that are producing the electric field should not be considered. Open channels permit the transfer of charges from one side to the other, thus induced polarization could be quite large as opposite charges are separated by a large distance (membrane thickness) and by a low dielectric constant layer (hydrophobic membrane core). We have to admit in addition that a low area density of open channels will result in strict global neutrality conditions not over entire membrane area, but just over small portion of it, which will then be spread with time over the whole membrane area due to equipotentiality. However, this may be enough to explain the observed saturation (on the average) of the electric‐ induced curvature, in view of the fact that global flexocoeYcient is some 30 times larger than the detailed one, and of opposite sign (see Petrov, 1999, Sections 6.5.1 and 6.5.2).


143

As Eq. (10) demonstrates, the ratio of monopole contribution at global neutrality vs monopole contribution at detailed neutrality is: f MB ew d2 30 5 150 ¼ ¼ ¼ ¼ 31:25 CB f eL 2lD dð1 þ lD =dÞ 2 2 1 ð1 þ 1=5Þ 4:8 Closed channel: f closed ¼ f DB þ f CB Open channel: locally opposite and strong polarizations of the channels and their close surrounding area will be spread over the whole membrane area due to the equipotential condition for the electrolyte halfspace. Average flexocoeYcient: f open;av ¼ ð1 aÞð f DB þ f CB Þ aj f MB j; where a is the relative portion of the area around channels per unit area. Now, since f open, av ffi 0, it follows that a ffi 1/30.

V. FLEXOELECTRICITY AND MECHANOTRANSDUCTION As we have seen, flexoelectricity is a fundamental property of liquid crystals, relating their mechanical and electrical degrees of freedom. In a membrane system with just these two degrees of freedom, we can encounter flexoelectric eVects discussed above, direct and converse ones. In membranes (living and model), flexoelectricity provides a linear relationship between membrane curvature and membrane polarization, or transmembrane voltage and membrane‐bending stress. It is thus closely related to mechanosensitivity and mechanotransduction, basic features of all living systems. Currently, mechanosensitivity of membranes is supposedly related to the presence of mechanosensitive channels in them. In hair cells, although no specific mechanosensitive channel has been identified, much is known about the channel’s location and the cytoskeletal proteins involved in its gating and adaptation (Garcia‐Anõveros and Corey, 1997; Gillespie and Walker, 2001). In particular, the channels are located at the tips of a hair cell’s stereocilia, whose core filaments are composed of cross‐linked actin. Extracellularly, the channels are linked by an unidentified protein or protein complex known as the tip‐link, which is normally under tension; it is ‘‘prestressed,’’ much like a tuned violin string. When the stereocilia are deflected in response to sound vibrations, the tip‐link (or an associated protein at the base of the tip‐link) becomes stretched, and the resulting increase in tension opens the channel.

144

Alexander G. Petrov

Earlier, a diVerent model using direct flexoelectric eVect for transformation of mechanical into electrical energy in the hearing process in stereocilia was proposed (Petrov and Usherwood, 1994). Unlike mechanisms involving specialized structures like stress‐activated channels, flexoelectricity may well operate in channel‐free membrane regions, although its combination with ion channels brings about some interesting new possibilities (see above). We concentrated our attention on the stereocilia tips (Fig. 10). These are highly curved membrane regions: with a stereocilium diameter of 300 nm (after Passechnik, 1988) the tip curvature 2/R amounts to 13 106 m1. During oscillations of a stereocilium in an excitatory direction, this curvature would increase because of the pull by the tip‐link, respectively, decreasing in the inhibitory direction. Assuming a curvature variation of only 10% and a flexocoeYcient of only 1020 C, a 1.5 mV oscillation amplitude of the membrane potential may be calculated from Eq. (2) for a single stereocilium. The generation of such flexoelectric potential is concentrated in the tip region.

FIGURE 10 A model for applying stress to the membranes of stereocilia in hair cells. The left panel shows how tilting of the bundle of cilia to the left (excitatory excursion) leads to stretching of the membranes because of the pull by the tip‐links. This observation is supposedly related to the stress‐sensitive channel functioning. The right panel shows that the tip curvature under excitatory/inhibitory tilt is increased/decreased, while the cilia shaft membrane corrugation is decreased/increased. The implications of these observations in the flexoelectric sensing mechanism is discussed in the text. Figure reprinted with kind permission of Springer Science and Business Media from Petrov, A. G., and Usherwood, P. N. R. (1994). Eur. Biophys. J. 23, 1, Fig. 3. Copyright (1994) by the European Biophysical Societies’ Association.


145

There is experimental evidence that this region is the site of the mechanoelectric transducer (Hudspeth, 1982, 1983). Further, the shaft of the stereocilium membrane is corrugated at rest (Passechnik, 1988), but the folds would disappear because of the pull during excitatory excursions (Fig. 10, right). Thus, an additional source of flexoelectric potential could be recognized, with a comparable magnitude of that of the tip region, but with an opposite phase. The value of 1.5 mV favorably compares to the known values of the hair cell sensitivity of (2–4) 105 V/m (Howard et al., 1988), which yields a few millivolts change of the membrane potential at 10‐nm displacement. The flexoelectric generators of all stereocilia are in parallel, so their total e.m.f. would remain the same. However, the generated flexoelectric (displacement) current (being proportional to the membrane area) would increase in proportion to the number of stereocilia being activated in concert. With a tip area of 2pR ¼ 1.4 109 and a broadly accepted value of specific biomembrane capacity of 1 F/cm2, the displacement current equation [Eq. (12)] gives, at 1000‐Hz vibration frequency and a flexoelectric potential amplitude of 1.5 mV, a flexoelectric current amplitude of 13 fA per stereocilium, that is, 1.3 pA for a bunch of 100 stereocilia. This is a lower estimate in view of the possible additional flexoelectric current generated along the shaft of the stereocilium (see also below). The flexoelectric displacement current is of greatest interest with high frequency stimuli, when it is largest, and when its linear growth with frequency would overcome the frequency‐ dependent decrease in amplitude of the imposed displacement of the stereocilium. In the low frequency region, the conductive component of the current becomes important, its value being directly dependent on the conducting state of the ion channels of the stereocilium membrane (Fig. 8). Regarding converse flexoelectric eVect involvement in mechanotransduction, Petrov and Usherwood (1994) have predicted: ‘‘Thus, further exciting possibilities for the participation of flexoelectricity in the active process of mechanoamplification may be discussed, inspired in particular by the fact that the curvature‐generating flexoelectric mechanism may be fast enough. By pointing out these possibilities we hope closer attention will be paid to them in future studies of the functioning of hair cells.’’ The converse flexoeVect into question was employed by Raphael et al. (2000) for description of the electromotility of outer hair cell (OHC) membrane (Fig. 11). Electromotility plays a central role in the process of mechanoamplification, which is vital for the hearing of high frequency sounds (Passechnik, 1988; Brownell et al., 2001). The OHC displays a repetitive arc and pillar nanoarchitecture, containing sharp points at the confluence of any two adjacent arcs (Fig. 11). This architecture is inherently polar and serves well the flexoelectric mechanism (e.g., a sine wave membrane shape will not do much in flexoelectric respect: while

146

Alexander G. Petrov

A

B

Plasma membrane

Pillar

Spectrin

Actin +

−

FIGURE 11 Nanomechanical model for OHC converse flexoeVect. (A) A schematic of the OHC. These cells are cylindrically shaped with lengths ranging from 20 to 90 mm along the cochlea and with a radius of 4–5 mm. The hair bundle, composed of stereocilia, is located at the apex of the cell. The lateral wall is the source of electromotility and it appears smooth under a light microscope. When examined with electron microscopy, the lateral wall appears corrugated. The folds in the membrane appear to terminate at pillar proteins that extend to the cytoskeleton. The cytoskeleton is composed of actin filaments crosslinked by spectrin molecules. (B) Curvature changes in the elemental motile unit cause extension of the spectrin molecules attached to the pillar proteins. Three units are shown in the figure. A membrane depolarization (þ) leads to a decrease in the radius of curvature and a shortening of the cell while hyperpolarization () leads to an increase of the radius of curvature and cell lengthening. Figures reprinted with kind permission of the authors and the publisher from Raphael, R. M., Popel, A. S., and Brownell, W. E. (2000). Biophys. J. 78, 2844, Fig.1 and Fig. 2. Copyright (2000) by the Biophysical Society.

one of the halfwaves is reduced the opposite one will be extended, and vice versa). Such an arc motive is repeated a few thousand times along the OHC cell membrane; this is how a nanometer displacement of the end of any single arc is amplified by three orders of magnitude resulting in a micrometer displacement of the cell end. Membrane arcs are also found in several other electromotile cells, for example, Oscillatoria, Flexibacter BH3 (Brownell, 2001). This makes a membrane arc terminating on protein pillars, a basic nanoscale unit of a unique linear flexoelectric motor. Interestingly, on a microscopic scale this motor looks like a piezoelectric one, that is, a change of cell elongation due to a change of membrane potential. Indeed, the idea that OHC possesses piezoelectric properties has been advanced by Iwasa (2001) and Dong et al. (2002).


147

However, such an apparent piezomotor features an enormously eVective piezocoeYcient, c12 ¼ 20,000,000 1012 C/N, which has no analogue in organic or other materials (the best piezoelectric ceramic PZT has c12 ¼ 400 1012 C/N only). We have been able to explain the apparent OHC piezoelectricity in terms of the Raphael et al. (2000) model, thus showing that even a very weak membrane flexoelectricity on a nanolevel, of f ¼ 1.1021 C, combined with a thousand times repetition of the elemental motile unit, is capable to produce the huge apparent piezoelectricity of OHC on a microlevel (Petrov, 2003, 2006). Brownell (2006) claims: ‘‘The fundamental motor unit in the flexoelectric based model for OHC somatic electromotility are circumferential plasma membrane ripples organized by cortical lattice F‐actin. If evolution followed the suggested scenario, the ripples represent a morphed version of the stereocilium motor units of the bundle amplifier. In both cases a voltage‐ driven change in membrane curvature generates mechanical force. Depolarization leads to bending of the bundle toward the tallest row and to shortening of the soma. Both of which could be working in concert for the high frequency requirements of the mammalian cochlear amplifier.’’ One eventual consequence of this claim that would further enlarge the domain of bioflexoelectricity is to test whether ripples of muscle fiber membrane display the same pillar and arc nanoarchitecture like OHC.

VI. CONCLUSIONS The process of mechanotransduction in membranes seems to benefit not only from specialized and localized elements like stress‐activated channels, but also from the collective properties of the mechanosensitive membranes as a whole. These collective properties evolve from the liquid crystal character of membranes and are best understood in terms of liquid crystal physics (Petrov, 1999). Among them, curvature elasticity and flexoelectricity hold the first places. They ensure an eVective and direct way of transformation of mechanical energy of the whole membrane into electrical one and vice versa. In doing so, both the lipid bilayer part of the membrane and the cytoskeleton of a special architecture are vital. The localized protein structures like membrane channels are obviously able to interact with global membrane properties in a striking way that is not fully understood. By confirming the existing concepts of membrane flexoelectricity in other cell types and by the eventual discovering of new concepts our exciting flexoelectric journey in the world of mechanoperception can proceed further and can reveal new important facets of the problem.

148

Alexander G. Petrov

References Beyder, A. (2005). Electro‐Mechanics on the Cell Surface. Ph.D. Thesis, State University of New York at BuValo. Brehm, P., Kullberg, R., and Moody‐Corbett, F. (1984). Properties of non‐junctional acetylcholine receptor channel on innervated muscle of Xenopus laevis. J. Physiol. 350, 631–648. Brownell, W. E. (2001). Membrane based motor mechanisms. In ‘‘1st World Flexoelectric Congress.’’ SUNY‐BuValo. Brownell, W. E. (2006). The piezoelectric outer hair cell: Bidirectional energy conversion in membranes. In ‘‘Auditory Mechanisms: Processes and Models’’ (A. L. Nuttall, P. Gillespie, T. Ren, K. Grosh, and E. de Boer, eds.), pp. 176–186. World Scientific, Singapore. Brownell, W. E., Spector, A. A., Raphael, R. M., and Popel, A. S. (2001). Micro‐ and nanomechanics of the cochlear outer hair cell. Annu. Rev. Biomed. Eng. 3, 169–194. De Gennes, P. G. (1974). ‘‘The Physics of Liquid Crystals.’’ Clarendon Press, Oxford. Derzhanski, A. (1989). Curvature induced polarization of bilayer lipid membrane. Phys. Lett. 139A, 170–173. Dong, X.‐X., Ospeck, M., and Iwasa, K. H. (2002). Piezoelectric reciprocal relationship of the membrane motor in the cochlear outer hair cell. Biophys. J. 82, 1254–1259. Garcia‐Anõveros, J., and Corey, D. P. (1997). The molecules of mechanosensation. Annu. Rev. Neurosci. 20, 567–594. Gillespie, P. G., and Walker, R. G. (2001). Molecular basis of mechanosensory transduction. Nature 413, 194–202. Gorczynska, E., Huddie, P., Miller, B., Mellor, I., Vais, H., Ramsey, R., and Usherwood, P. N. R. (1996). Potassium channels of adult locust muscle. Pflugers Arch. 432, 597–606. Green, D. E., Ji, S., and Bru¨cker, R. F. (1973). Structure‐function unitization model of biological membranes. J. Bioenerg. 4, 253–284. Guharay, F., and Sachs, F. (1984). Stretch activated single ion channel currents in tissue‐ cultured embryonic chick skeletal muscle. J. Physiol. 352, 685–701. Hamill, O. P. (1983). Potassium and chloride channels in red blood cells. In ‘‘Single Channel Recording’’ (B. Sakmann and E. Neher, eds.), pp. 451–471. Plenum Press, New York London. Hamill, O. P., and Martinac, B. (2001). Molecular basis of mechanotransduction in living cells. Physiol. Rev. 81, 685–740. Howard, J., Roberts, W. M., and Hudspeth, A. J. (1988). Mechnaoelectrical transduction by hair cells. Annu. Rev. Biophys. Biophys. Chem. 17, 99–124. Hristova, K., Bivas, I., Petrov, A. G., and Derzhanski, A. (1991). Influence of the electric double layers of the membrane on the value of its flexoelectric coeYcient. Mol. Cryst. Liq. Cryst. 200, 71–77. Hristova, K., Bivas, I., and Derzhanski, A. (1992). Frequency dependence of the membrane flexoelectric voltage response. Adsorption of multivalent counterions on the surface of curved lipid bilayer. Mol. Cryst. Liq. Cryst. 215, 237–244. Hudspeth, A. J. (1982). Extracellular current flow and the site of transduction by vertebrate hair cells. J. Neurosci. 2, 1–10. Hudspeth, A. J. (1983). Mechanoelectrical transduction by hair cells in the acousticolateralis sensory system. Annu. Rev. Neurosci. 6, 187–215. Ingber, D. E. (1997). Tensegrity: The architectural basis of cellular mechanotransduction. Annu. Rev. Physiol. 59, 575–599. Iwasa, K. H. (2001). A two‐state piezoelectric model for outer hair cell motility. Biophys. J. 81, 2495–2506. Mellor, I. R., Miller, B. A., Petrov, A. G., Tabarean, I., and Usherwood, P. N. R. (1999). Mechanosensitive potassium channels in locust muscle membrane. Eur. Biophys. J. 28, 346–350.


149

Meyer, R. B. (1969). Piezoelectric eVects in liquid crystals. Phys. Rev. Lett. 22, 918–922. Morris, C. E. (1990). Mechanosensitive ion channels. J. Membr. Biol. 113, 93–107. Mosbacher, J., Langer, M., Horber, J. K. H., and Sachs, F. (1998). Voltage‐dependent membrane displacements measured by atomic force microscopy. J. Gen. Physiol. 111, 65–74. Orr, A. W., Helmke, B. P., Blackman, B. R., and Schwartz, M. A. (2006). Mechanisms of mechanotransduction. Dev. Cell 10, 11–20. Passechnik, V. I. (1988). Mekhanizmi ulitki organa slukha. In ‘‘Accounts’’ (P. G. Kostyuk, ed.), pp. 6–121, Sci. Techn. VINITI, Moscow. Hum. Anim. Physiol. Ser. 39. Petrov, A. G. (1975). Flexoelectric model of active transport. In ‘‘Physical and Chemical Bases of Biological Information Transfer’’ (J. Vassileva, ed.), pp. 111–125. Plenum Press, New York, London. Petrov, A. G. (1999). ‘‘The Lyotropic States of Matter. Molecular Physics and Living Matter Physics.’’ Gordon & Breach Publsisher, New York, London. Petrov, A. G. (2001). Flexoelectricity of model and living membranes. Biochim. Biophys. Acta 1561, 1–25. Petrov, A. G. (2003). Membrane flexoelectricity at nanoscale level. In ‘‘2nd World Flexoelectric Congress.’’ Rice‐Houston. Petrov, A. G. (2006). Electricity and mechanics of biomembrane systems: Flexoelectricity in living membranes. Anal. Chim. Acta 568, 70–83. Petrov, A. G., and Derzhanski, A. (1976). On some problems in the theory of elastic and flexoelectric eVects in bilayer lipid membranes and biomembranes. J. Phys. Suppl. 37, C3‐155–C3‐160. Petrov, A. G., and Sachs, F. (2002). Flexoelectricity and elasticity of asymmetric biomembranes. Phys. Rev. E 65, 021905–021910. Petrov, A. G., and Sokolov, V. S. (1986). Curvature‐electric eVect in black lipid membranes. Eur. Biophys. J. 13, 139–155. Petrov, A. G., and Usherwood, P. N. R. (1994). Mechanosensitivity of cell membranes. Ion channels, lipid matrix and cytoskeleton. Eur. Biophys. J. 23, 1–19. Petrov, A. G., Seleznev, S. A., and Derzhanski, A. (1979). Principles and methods of liquid crystal physics applied to the structure and functions of biological membranes. Acta Phys. Pol. A55, 385–405. Petrov, A. G., Ramsey, R. L., and Usherwood, P. N. R. (1989). Curvature‐electric eVects in artificial and natural membranes studied using patch‐clamp techniques. Eur. Biophys. J. 17, 13–17. Petrov, A. G., Miller, B. A., and Usherwood, P. N. R. (1992). Mechanoelectricity of guest‐host membrane systems: Lipid bilayer containing ion channels. Mol. Cryst. Liq. Cryst. 215, 109–119. Petrov, A. G., Miller, B. A., Hristova, K., and Usherwood, P. N. R. (1993). Flexoelectric eVects in model and native membranes containing ion channels. Eur. Biophys. J. 22, 289–300. Raphael, R. M., Popel, A. S., and Brownell, W. E. (2000). A membrane bending model of outer hair cell electromotility. Biophys. J. 78, 2844–2862. Sachs, F. (1990). Stress‐sensitive ion channels. The Neurosciences 2, 49–57. Snyder, K., Zhang, P. C., and Sachs, F. (2002). Dynamic AFM of patch clamped membranes. In ‘‘Ion Channel Localization Methods and Protocols’’ (A. Lopatin and C. G. Nichols, eds.), pp. 425–459. Humana Press Inc., Totowa, NJ. Suchyna, TM., and Sachs, F. (2004). Dynamic regulation of mechanosensitive channels; capacitance used to monitor patch tension in real time. Phys. Biol. 1, 1–18. Todorov, A. T., Petrov, A. G., Brandt, M. O., and Fendler, J. H. (1991). Electrical and real‐ time stroboscopic interferometric measurements of bilayer lipid membrane flexoelectricity. Langmuir 7, 3127–3137.

150

Alexander G. Petrov

Todorov, A. T., Petrov, A. G., and Fendler, J. H. (1994a). Flexoelectricity of charged and dipolar BLM studied by stroboscopic interferometry. Langmuir 10, 2344–2350. Todorov, A. T., Petrov, A. G., and Fendler, J. H. (1994b). First observation of the converse flexoelectric eVect in bilayer lipid membranes. J. Phys. Chem. 98, 3076–3079. Winterhalter, M., and Helfrich, W. (1992). Bending elasticity of electrically charged bilayers: Coupled monolayers, neutral surfaces, and balansing stresses. J. Phys. Chem. 96, 327–330. Zhang, P. C., Keleshian, A. M., and Sachs, F. (2001). Voltage‐induced membrane movement. Nature 413, 428–432.

CHAPTER 6 Lipid Effects on Mechanosensitive Channels Andrew M. Powl and Anthony G. Lee School of Biological Sciences, University of Southampton, Southampton SO16 7PX, United Kingdom

I. II. III. IV.

Overview Intrinsic Membrane Proteins EVects of Lipid Structure on Membrane Protein Function How to Explain EVects of Lipid Structure on Membrane Protein Function A. The Lipid Annulus B. The Fluidity of a Lipid Bilayer and Its Consequences C. The Importance of Hydrophobic Thickness D. Curvature Stress E. Elastic Strain and Pressure Profiles F. General Features of Lipid–Protein Interactions V. What Do These General Principles Tell Us About MscL? References

I. OVERVIEW The structure and function of a membrane protein depend on the properties of the lipid molecules, the annular lipid molecules, that surround it in a membrane. Fatty acyl chain length is important because lipid chain length determines the hydrophobic thickness of the lipid bilayer, and the eYciency of hydrophobic matching between a membrane protein and the surrounding lipid bilayer is high. Lipid headgroup structure can also be important because the lipid annulus can be heterogeneous with, for example, hot-spots where anionic lipids can bind with high aYnity. An important question is whether Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58006-8

152

Powl and Lee

the activity of a membrane protein depends only on the properties of the annular lipid molecules or is also dependent on the properties of the bulk lipid molecules in the membrane. Bulk properties that have been considered to be important include fluidity, curvature stress, elastic strain, and transmembrane pressure profiles. This chapter considers to what extent these general features of lipid–protein interactions help us to understand the properties of the bacterial mechanosensitive channel of large conductance (MscL).

II. INTRINSIC MEMBRANE PROTEINS All membrane proteins operate in the environment of a lipid bilayer and so are likely to be aVected by changes in the chemical composition or physical properties of the lipid bilayer. The bacterial mechanosensitive channels are unusual membrane proteins in that their dependence on the lipid bilayer is the key to their function: It is an increase in the tension in the lipid bilayer that leads to channel opening (Moe and Blount, 2005). However, insights into how bacterial mechanosensitive channels might function are likely to come from studies of other classes of membrane protein because lipid–protein interactions are important for all membrane proteins. Some membrane proteins require specific lipids for function, these lipids binding to specific sites on the membrane proteins (Lee, 2003, 2004, 2005). The eVects of such lipids clearly need to be understood in terms of the specific interactions between the bound lipid molecules and the membrane protein. However, the majority of the lipids around a membrane protein will not interact with the protein in this way, playing a role more like a conventional solvent molecule. The first question that needs to be answered is whether the function of a membrane protein changes when the structures of the ‘‘solvent’’ lipid molecules surrounding it in a membrane are changed. If protein function is found to be sensitive to the nature of the surrounding lipids, then the question arises as to whether this dependence is best explained at the molecular level, in terms of specific molecular interactions between the lipids and proteins, or at the macroscopic level, in terms of bulk properties of the lipid bilayer.

III. EFFECTS OF LIPID STRUCTURE ON MEMBRANE PROTEIN FUNCTION That membrane protein function is sensitive to lipid structure is most readily demonstrated for membrane-bound enzymes, since the techniques of presteady-state and steady-state kinetics allow detailed insights into mechanism (Michelangeli et al., 1991). Figure 1 shows the eVects of lipid

153

6. Lipid EVects on Mechanosensitive Channels 4 3 40 2 20 1

0 14

16

18 20 Chain length

22

24

Activity (IU/mg)

Activity (IU/mg)

60

0

FIGURE 1 The eVect of fatty acyl chain length on enzyme activity in bilayers of phosphatidylcholine in the liquid crystalline phase. Ca2þ-ATPase (□; right-hand axis) or diacylglycerol kinase (○; left-hand axis) were reconstituted into phosphatidylcholines containing monounsaturated fatty acyl chains of the given chain lengths. ATPase activities were determined at 25 C. For diacylglycerol kinase, the substrate was DHG present at 20 mol% in the bilayer. Data from Pilot et al. (2001a) and Lee (2003).

fatty acyl chain length on the activities of two membrane-bound enzymes, the Ca2þ-ATPase from skeletal muscle sarcoplasmic reticulum (Froud et al., 1986a,b; Starling et al., 1993) and the bacterial enzyme diacylglycerol kinase (DAGK) that uses ATP to phosphorylate a diacylglycerol to give the corresponding phosphatidic acid (Pilot et al., 2001a). In both cases, highest activity is seen in bilayers of phosphatidylcholines in the liquid crystalline phase when the fatty acyl chain length is C18, with lower activities in bilayers of lipids with shorter or longer fatty acyl chains. However, the reasons why short- and long-chain lipids give low activities for the Ca2þ-ATPase are diVerent from the reasons why they give low activities for DAGK, and the reasons for the low activities in short-chain lipids are diVerent from the reasons for the low activities in long-chain lipids. For example, the low rate of ATP hydrolysis observed for DAGK in di(C14:1)PC (Fig. 1) follows from a very high Km value for dihexanoylglycerol (DHG) in di(C14:1)PC, the value for vmax being the same as in di(C18:1)PC (Pilot et al., 2001a) (Fig. 2). In contrast, the low activity measured for DAGK in di(C24:1)PC follows from a low value for vmax, the value for Km being comparable to that in di(C18:1)PC (Pilot et al., 2001a) (Fig. 2). For the Ca2þ-ATPase, the low rate of ATP hydrolysis in di(C14:1)PC follows in large part from a slow rate of phosphorylation of the Ca2þ-ATPase by ATP (Starling et al., 1995a) (Fig. 3). In di(C24:1)PC, the rate of phosphorylation is the same as that in di(C18:1)PC (Fig. 3) and the low rate of ATP hydrolysis by the Ca2þATPase in di(C24:1)PC follows from a slow rate of dephosphorylation of

154

Powl and Lee 80 60

60

40 40 20 20

0 14

0 16

18 20 Chain length

22

Km (mol%) for DHG

νmax (IU/mg protein)

80

24

FIGURE 2 EVects of phosphatidylcholine chain lengths on Km and vmax values for DHG. vmax (○) and Km (□) values for DHG (expressed as mol% DHG in the membrane) at a fixed ATP concentration of 5 mM are plotted vs chain length. Data from Pilot et al. (2001a).

[EP] (nmol/mg protein)

3.0 2.5 2.0 1.5 1.0 0.5 0.0 0.00

0.05

0.10 Time (s)

0.15

0.20

FIGURE 3 Rate of phosphorylation of Ca2þ-ATPase as a function of fatty acyl chain length. The Ca2þ-ATPase was reconstituted in di(C18:1)PC (○), di(C24:1)PC (□), or di(C14:1)PC () in the presence of Ca2þ and mixed with ATP to give a final concentration of 50 mM, and the level of phosphorylation was determined at the given times. The lines show fits to single exponential rate processes. Data from Starling et al. (1995a).

the phosphorylated intermediate (Starling et al., 1995a,b). EVects of lipid chain length on the properties of the Ca2þ-ATPase are particularly complex, most, if not all, of the steps in the reaction sequence being aVected, the most surprising change being a change in the stoichiometry of Ca2þ binding to the Ca2þ-ATPase, from 2:1 in the native membrane or in di(C18:1)PC, to 1:1 in short- or long-chain lipid (Michelangeli et al., 1990; Starling et al., 1993).

6. Lipid EVects on Mechanosensitive Channels

155

Lipid headgroup structure also has important eVects on activity. The activity of the Ca2þ-ATPase is lower in a phosphatidylethanolamine than in the corresponding phosphatidylcholine, largely due to a decreased rate of dephosphorylation of the phosphorylated intermediate (Starling et al., 1996a). The Ca2þ-ATPase also has low activity in bilayers of anionic lipids (Dalton et al., 1998) and the activity of DAGK is also low in bilayers of phosphatidylethanolamines and anionic lipids (Pilot et al., 2001b). Studies of the Ca2þ-ATPase in mixtures of short-chain phosphatidylcholines and normal-chain anionic lipids suggest that the eVects of lipid headgroup and lipid fatty acyl chain on protein function can be considered to be separate (Dalton et al., 1998). It has been suggested that changes in lipid structure could aVect protein function by changing the aggregation state of the protein, but this has been shown not to be the case for the Ca2þ-ATPase (Starling et al., 1995b). A number of conclusions can be drawn from these results. The first is that there is unlikely to be any one, unique, explanation for the eVects of lipid structure on membrane protein function. The second is that changing lipid structure results in changes in protein structure, the changes in protein structure underlying the observed changes in function. Thus, membrane proteins are not rigid, but have a degree of plasticity, allowing them to deform to help provide optimum matching to the surrounding lipid bilayer.

IV. HOW TO EXPLAIN EFFECTS OF LIPID STRUCTURE ON MEMBRANE PROTEIN FUNCTION A. The Lipid Annulus To what extent can the structures and functions of membrane proteins be explained using the language and principles developed for water-soluble proteins? Explanations for the eVects of solvent water on the structure and function of a water-soluble protein are based largely on the hydrophobic eVect and on hydrogen bonding between water molecules and polar residues on the solvent-exposed surface of the protein; explanations rarely involve bulk properties of the water such as viscosity or the ‘‘pressure’’ exerted on the protein molecule as a result of collisions of the water molecules with the protein surface. Can the eVects of the lipid molecules that cover the transmembrane surface of a membrane protein also be understood in the same way, as simple solvent eVects? Certainly, crystal structures of membrane proteins containing resolved lipid molecules show the importance of hydrogen bonding and charge interactions between polar or charged residues on the protein and the lipid headgroup and backbone region (Lee, 2003).

156

Powl and Lee

Changing lipid headgroup structure, leading to changes in the interaction between the lipid headgroup and the protein, could well lead to changes in protein structure and thus to changes in protein function. The solvent lipid molecules covering the hydrophobic surface of a membrane protein are generally referred to as boundary or annular lipids, as they form an annular shell around the protein (Lee, 2003, 2004, 2005). The importance of this annular shell has been shown in experiments with the Ca2þ-ATPase; the number of lipid molecules required to form an annular shell around the Ca2þ-ATPase is about 30 (East et al., 1985), and the activity of the Ca2þ-ATPase remains constant as the number of lipid molecules per Ca2þ-ATPase molecule is reduced from 90, corresponding to the lipid:protein molar ratio in the native membrane, to about 30, below which activity declines (Warren et al., 1974). The fact that activity is maintained in a membrane with essentially no bulk lipid shows that the presence of bulk lipid is not essential for activity and thus that the properties of the bulk lipid are less important determinants of activity than the properties of the annular lipid: In other words, the annular lipid molecules largely buVer a membrane protein from the eVects of bulk lipid. B. The Fluidity of a Lipid Bilayer and Its Consequences The lipid bilayer component of a biological membrane is generally in a fluid state, commonly referred to as the liquid crystalline state. This is marked by considerable motional freedom of the lipid molecules. Figure 4 illustrates the types of motion that we might expect in a lipid bilayer in the

Flexing of headgroups

Rapid lateral diffusion in plane of bilayer

Rapid internal motion of chains

Slow flip-flop

Rapid rotation about long axis FIGURE 4

Classes of motion in a phospholipid bilayer.


157

liquid crystalline phase. The most important of the intramolecular motions is rotation about C---C bonds in the fatty acyl chains and in the headgroup region; it is this that makes the lipids ‘‘liquid-like.’’ Motions of the whole lipid molecule include a fast lateral diVusion in the plane of the membrane and a fast axial rotation of the lipid about its long axis. Flip-flop motion in which a lipid moves from one side of the membrane to the other is, however, slow, since it involves moving the polar headgroup of the lipid through the hydrocarbon core of the membrane; it is the slowness of this motion that allows an asymmetric distribution of lipids between the two halves of a biological membrane. In some ways the interior of a lipid bilayer is like a simple hydrocarbon. For example, an analysis of NMR spin-lattice relaxation times has suggested that trans-gauche isomerization rates (109–1010 s1) in fatty acyl chains in a lipid bilayer are very similar to those in a free chain and that the eVective viscosity for the bilayer is ca. 0.01 P (Poastor et al., 1988). This can be compared with the viscosity of neat hexadecane at 50 C, which is 0.019 P (Small, 1986). Similarly, molecular packing of the chains in the liquid crystalline bilayer is equivalent to that of a liquid alkane, with average ˚ 3, methylene and methyl group volumes in the bilayer interior of 28 and 54 A respectively (Petrache et al., 1997), comparable to the methylene and methyl ˚ 3, respectively (Nagle and group volumes in a liquid alkane of 27 and 57 A Wiener, 1988). Nevertheless, there are important diVerences between a lipid bilayer and a normal liquid. The chemical nature of the lipid molecule, with fatty acyl chains ‘‘anchored’’ at the top of the chain to a relatively immobile backbone, most commonly a glycerol group, results in a gradient of motion along the lipid fatty acyl chains. A proper description of this chain motion requires a clear distinction to be made between the rates and the amplitudes of motion. The term ‘‘fluidity’’ and its inverse, viscosity, refers strictly, only to the rate of motion but in the biological literature the word fluidity is used very informally to include both factors. This is largely a reflection of the fact that most methods used to measure rates of motion (NMR, ESR, fluorescence depolarization) are, in fact, sensitive to both rates and extents of motion, and separating the two can be diYcult. The extent or range of motion is described in terms of an order parameter that describes the timeaveraged disposition in space of each group of atoms in the fatty acyl chain. The rate of motion can be described in terms of a correlation time, a measure of the rate of movement of a group of atoms between its various possible positions in space. The idea of an order parameter can be introduced by considering the lipid molecules in a bilayer to be long, thin cylinders (Fig. 5). In a bilayer, the long axes of the cylinders will tend to align about a direction normal to the surface of the bilayer. At finite temperatures, the thermal motions of the

158

Powl and Lee A

B

Ψ

z

n

q x

y

f FIGURE 5 Orientation of lipid molecules (represented as cylinders) in one leaflet of a bilayer (A) and the angles required to describe the orientation of a molecule in the bilayer (B).

molecules will prevent the alignment from being perfect; the cylinders will occupy a range of angles about the direction of the bilayer normal. In Fig. 5, the bilayer normal is aligned along the z-axis of a fixed rectangular coordinate system. The orientation of the cylinder can then be described by the three angles shown. The cylindrical symmetry of the molecules means that there can be no order in the system in the angle c (describing rotation of the cylinder about its long axis) or in the angle ’ (describing rotation of the cylinder about the z direction): If any particular angle c or ’ had been preferred in some way, then the shape of the molecule could not have been cylindrical. However, a degree of order can exist for the angle y. The most likely value of y for the bilayer structure will be 0 , with the long axis of the cylinder parallel to the bilayer normal. However, since order is not perfect, a range of values of y will be observed, centered around this most probable value of 0 . In contrast, in a normal isotropic liquid, where no direction in space is preferred, all values of the angle y will be equally likely. The angle y itself, although a measure of order, is not a very convenient one. Rather, it is normal to introduce an order parameter, a function of cos2y. Actually, it is not the instantaneous value of cos2y for one given molecule that is important, but rather the averaged value of cos2y over the timescale of the measurement for all the molecules in the sample; this average is written as . Finally, it is traditional to write an order parameter in such a way that it has the value of 1 for a perfectly ordered


159

sample and a value of 0 for a completely disordered isotropic liquid. For a fully ordered system, y ¼ 00 for all molecules and so the value of will be 1, but for a totally disordered system when all values for y are equally likely, the value of is 1/3. Thus, the order parameter S that is used is: S¼

3 < cos2 y > 1 2

ð1Þ

This has the required properties: S ¼ 1 for a completely ordered system and S ¼ 0 for the isotropic phase. The formalism described above is that used to describe the order of a C---H bond in a fatty acyl chain of a lipid molecule in terms of the average angle between the C ---H bond and the bilayer normal. Anchoring the fatty acyl chains of a phospholipid molecule at one end to the lipid backbone results in a gradient of motion along the chain, the extent or range of motion increasing from the backbone to the terminal methyl end of the chain. Each CH2 group in the chain will, therefore, have its own characteristic range of motion and thus its own order parameter. The most powerful technique for measuring these order parameters is 2H NMR, studying the motion of C---D groups introduced at specific positions in the chains (Seelig and Seelig, 1980; Bloom et al., 1991). The order parameter of the C---D bond, SCD, is defined as SCD ¼

3 < cos2 j > 1 ; 2

ð2Þ

where j is the angle between the bilayer normal and the C---H vectors at carbon atom j. SCD describes the extent of the time-averaged excursions experienced by the C---H group. For an all-trans configuration of a saturated fatty acyl chain rotating about the bilayer normal, the value of SCD will be 0.5. At higher temperatures, trans-gauche isomerizations about the C---C bond leads to a mixture of trans and gauche configurations along the chain, which will reduce the absolute value (magnitude) of SCD. The interpretation of the experimental order parameter SCD in molecular terms may not always be as simple as suggested above since measured value of SCD can also depend on the geometry of the deuterated molecule. This can be important for chains containing C-----C double bonds, since the orientation of a C---D bond for a C atom in a double bond will necessarily be diVerent to that in a normal CH2 group. Order parameter profiles for all phospholipid bilayers in the liquid crystalline phase are very similar. That for the palmitoyl chain in 1-palmitoyl-2oleoylphosphatidylcholine [(C16:0, C18:1)PC] is shown in Fig. 6 (Seelig and

160

Powl and Lee 0.25

A

0.20 0.15

Order parameter −SCD

0.10 0.05 0.00 0.25

B

0.20 0.15 0.10 0.05 0.00

0

2

4

6 8 10 12 Labeled carbon atom

14

16

FIGURE 6 The experimental order parameters (SCD) for the palmitoyl (○) and oleoyl (□) chains of (A) (C16:0, C18:1)PC; (B) lipids of E. coli labeled in the palmitoyl (○) and oleoyl (□) chains. Data from Seelig and Seelig (1980).

Seelig, 1980). The measured order parameters are observed to lie between the values expected for an all-trans chain rotating about its long axis (SCD ¼ 0.5) and for complete orientational disorder, as found in an isotropic liquid (SCD ¼ 0). Thus, some order persists in the fatty acyl chain region despite the liquid-like state of the chains. The degree of order varies along the chain, an initial plateau region of constant order being followed by a region of rapidly decreasing order toward the center of the bilayer. The plateau region has its origin in intermolecular restrictions on chain motion. Excluded volume eVects are very important in the upper part of the chain since rotation about a single C---C bond results in a large bend in the chain. Lower down the chain, lateral displacements resulting from rotations about single C---C bonds are very much smaller, and, therefore, steric restrictions on motion become less important. The order parameter profile for the unsaturated chain in (C16:0, C18:1)PC appears to be very diVerent to that for the saturated chain with the


161

experimental order parameters for carbon atoms 10 and 11 being low in the oleoyl chain (Fig. 6). This does not, however, indicate a high degree of motional disorder for these carbons, but rather follows from eVects of the cis double bond on the orientation of C---D bonds with respect to the bilayer normal. Nevertheless, molecular dynamics simulations do show an increased motion for the C-----C double bond and the methylene groups next to it, particularly for that on the terminal methyl side (Heller et al., 1993; Huang et al., 1994). Increased disorder in the region of the double bond is the result of shallow energy barriers for rotation about C---C bonds adjacent to a double bond (Li et al., 1994). The eVects of polyunsaturation have also been studied in a series of phosphatidylcholines with a deuterated stearoyl chain at the sn-1 position and an unsaturated chain at the sn-2 position. EVects are rather small, the sn-1 chain becoming slightly more disordered as the unsaturation of the sn-2 chain is increased, but with the eVect of unsaturation reaching a maximum at three double bonds (Holte et al., 1995). There has been special interest in the properties of lipids containing polyunsaturated docosahexaenoic acid (DHA) chains because of its high concentration in retinal rod membranes. The DHA chain shows considerable flexibility because of the large number of cis double bonds that it contains (Feller et al., 2002), and it has been suggested that the extreme flexibility for the DHA chain could be important for interaction with membrane proteins (Grossfield et al., 2006). A molecular dynamics simulation of rhodopsin in a bilayer of 1-stearoyl-2-docosohexaenoyl-phosphatidylcholine showed that the DHA chains penetrate deeper into the protein interface than do the stearic acid chains (Grossfield et al., 2006). It was suggested that the extreme flexibility of the DHA chain could allow it to adapt better to the rugged surface of the protein (Grossfield et al., 2006). Profiles of chain order are largely unaVected by lipid headgroup, although absolute values of order parameters can be aVected. Thus, order parameters for phosphatidylethanolamines in the liquid crystalline phase are almost constant for the first part of the chain, but decrease rapidly toward the terminal methyl group, as for the phosphatidylcholines, but the order parameters are higher for phosphatidylethanolamines than for phosphatidylcholines, at all positions of the chain (Perly et al., 1985; Lafleur et al., 1990). The higher order parameters in phosphatidylethanolamines can be attributed to the smaller headgroup of the phosphatidylethanolamine and to strong intermolecular hydrogen bonding between the headgroups, both factors leading to a greater packing density throughout the bilayer. However, diVerences in packing density between phosphatidylcholines and phosphatidylethanolamines in the chain region must be quite small since the thicknesses of bilayers of di(C18:1)PC and di(C18:1)PE are equal (Fenske et al., 1990).

162

Powl and Lee

An important observation is that order parameter profiles for intact biological membranes are the same as for those for simple lipid bilayers. The profiles for Escherichia coli membranes labeled in the palmitoyl and oleoyl chains are shown in Fig. 6; the values of the order parameters and their variation along the chain are the same as in bilayers of (C16:0, C18:1) PC (Seelig and Seelig, 1980). Thus, the presence of membrane proteins has no significant eVect on the extent of motion of the average lipid fatty acyl chain in the membrane. A second important point also follows from these results that the lipids in the E. coli inner membrane must adopt a bilayer structure since the order parameter profile is that characteristic of a bilayer, despite the fact that the predominant lipid in the E. coli inner membrane is phosphatidylethanolamine, a lipid that, on its own, would prefer the hexagonal HII phase. This is consistent with model system studies which show that even a small proportion of lipids favoring a bilayer structure will stabilize nonbilayer favoring lipids in a bilayer structure (Boni and Hui, 1983). There is also considerable motion in the lipid headgroup region in the liquid crystalline phase. Average orientations of the headgroups in phosphatidylcholines and phosphatidylethanolamines are roughly parallel to the bilayer surface in the liquid crystalline phase (Buldt and Wohlgemuth, 1981), but molecular dynamics simulations for phosphatidylcholines show that the orientations of the P–N vectors in individual molecules can vary from an angle of zero with respect to the bilayer normal, so that the NMe3þ group is pointing out into the solvent, to values greater than 90 , so that the NMe3þ group is pointing into the hydrocarbon core of the bilayer (Heller et al., 1993; Stouch et al., 1994; Hyvonen et al., 1997). Proper function of a membrane protein generally seems to require that the lipid bilayer be in a liquid crystalline phase but there is, however, no compelling evidence to suggest that the exact fluidity in the fluid crystalline phase is important for function (East et al., 1984; Lee, 1991). The importance of the liquid crystalline phase is that it gives the bilayer the important property of plasticity; the relatively weak interactions between neighboring lipid molecules means that the bilayer can distort around a foreign body, without any large-scale breakdown of bilayer structure. For example, Fig. 7 shows the results of a molecular dynamics simulation of the hemagglutinin fusion peptide bound to a bilayer of (C16:0, C18:1)PC (Lague et al., 2005). EVects are largely limited to the immediate neighbors of the peptide; the presence of the peptide has no eVect on the properties of the bulk of the lipids in the bilayer. On the peptide-containing side of the bilayer, the chains immediately adjacent to the peptide become more disordered (SCD values are low) and thus shorter, as the chains curl to fill the space under the peptide. In contrast, chains immediately below the peptide in the other monolayer become more ordered and thus longer, and extend into the upper monolayer,


163

FIGURE 7 Packing of (C16:0, C18:1)PC around the hemagglutinin fusion peptide. Modified from Lague et al. (2005).

helping to fill the space under the peptide (Lague et al., 2005). Similar results were obtained in a molecular dynamics simulation of a simple tripeptide binding to a bilayer surface, where again the bilayer was able to accommodate the peptide without significant change to the properties of the bulk phospholipids (Damodaran et al., 1995). This high degree of plasticity means that the bilayer will provide little resistance to a change in shape for a membrane protein, allowing the protein to undergo any conformational changes required for function. C. The Importance of Hydrophobic Thickness An important way in which a lipid bilayer diVers from a bulk solvent like water is in possessing a distinct thickness. Given that the cost of exposing hydrophobic regions of either a lipid bilayer or a protein to water is high, it can be expected that the hydrophobic thickness of a membrane protein will

164

Powl and Lee

match closely that of the surrounding lipid bilayer. This has been shown in dramatic fashion for the potassium channel KcsA where Trp residues at the ends of the transmembrane a-helices have been shown to maintain their positions close to the glycerol backbone region in bilayers of phosphatidylcholines over a chain length range C12-C24, representing a more than twofold change in bilayer hydrophobic thickness (Williamson et al., 2002). The two most likely ways in which this hydrophobic matching could be achieved are illustrated in Fig. 8. The first involves compression or stretching of the lipid fatty acyl chains around the membrane protein to achieve matching. However, compressing or stretching the lipid bilayer requires work and, if the required work is suYciently high, it may become more favorable energetically to distort the protein; if a membrane protein can exist in states with diVerent hydrophobic thicknesses, then the equilibrium between these states will be shifted toward the state whose hydrophobic thickness best matches the hydrophobic thickness of the unperturbed surrounding lipid bilayer, so minimizing the requirement to distort the lipid bilayer. For example, gramicidin is a small peptide that dimerizes to form channels across the membrane; the hydrophobic length of the gramicidin dimer is relatively short and so thick lipid bilayers shift the monomer-dimer equilibrium toward monomer (Lundbaek and Andersen, 1999; Lundbaek et al., 2004). For an intrinsic membrane protein, the most obvious distortion to achieve hydrophobic matching is tilting of the transmembrane a-helices (Fig. 8). Such tilting was

Optimal matching

Bilayer too thick

Bilayer too thin

A

B

FIGURE 8 Possible responses to hydrophobic mismatch between a membrane protein and its surrounding lipid bilayer. In (A) matching to a too thick or a too thin bilayer results from compressing or stretching of the lipid fatty acyl chains, respectively. In (B) matching to a too thick or a too thin bilayer results from decreasing or increasing the tilt angle of a transmembrane a-helix, respectively.

165


suggested to explain the behavior of simple model transmembrane helices in lipid bilayers (Webb et al., 1998), in agreement with molecular dynamics simulations (Kandasamy and Larson, 2006). Tilting of transmembrane a-helices to minimize hydrophobic mismatch has been detected experimentally for the pore region of the M2 proton channel (Duong-Ly et al., 2005) and for the protein Vpu from HIV-1 (Park and Opella, 2005). An important observation is that eVects of bilayer thickness on membrane protein function are cooperative. This is shown in Fig. 9 for the Ca2þATPase in mixtures of di(C18:1)PC and di(C14:1)PC; the changes in Ca2þbinding stoichiometry and ATPase activity characteristic of short-chain lipid only occur when the bilayer contains more than 50 mol% of the short-chain lipid (Starling et al., 1993). Since the binding aYnity of the Ca2þ-ATPase for lipid varies little with chain length (East and Lee, 1982), the results illustrated in Fig. 9 show that the structural changes in the Ca2þ-ATPase leading to low activity only occur when more than about 15 of the 30 lipids in the annular shell of the Ca2þ-ATPase (East et al., 1985) are short-chain lipids; the energy required to distort one lipid molecule might be less than that required to distort the Ca2þ-ATPase, but the energy required to distort 15 lipid molecules is presumably comparable to that required to distort the Ca2þ-ATPase.

5 12

ATPase activity

4

10 3

8 6

2

4 1 0

2

Ca2+ bound (nmoles/mg)

14

0 0.0

0.2 0.4 0.6 0.8 Mole fraction di(C14:1)PC

1.0

FIGURE 9 EVects of mixtures of di(C14:1)PC and di(C18:1)PC on Ca2þ binding and activity of the Ca2þ-ATPase. The Ca2þ-ATPase was reconstituted with mixtures of di(C14:1)PC and di(C18:1)PC containing the given mole fraction of di(C14:1)PC. The bars show the levels of Ca2þ binding (nmol Ca2þ bound/mg protein). Also shown are the ATPase activities measured at 25 C (○). Data from Starling et al. (1993).

166

Powl and Lee

FIGURE 10 Loops between transmembrane helices on the luminal side of the Ca2þ-ATPase in its Ca2þ-bound form. The view is end-on, from the luminal side of the membrane. The locations of the loops between transmembrane helices M1-M2, M3-M4, M5-M6, M7-M8, and M9-M10 are marked. The two bound Ca2þ ions are shown in space fill format (PDB file 1EUL).

It is possible that the structure of the Ca2þ-ATPase makes distortion of the structure relatively easy (Lee, 2002). Loops connecting the transmembrane a-helices on the luminal side of the Ca2þ-ATPase are short, the loops making little contact with each other (Fig. 10), so that it will be largely the relative strengths of the helix–helix and helix–lipid interactions that keep the transmembrane helical bundle of the Ca2þ-ATPase intact. Changing the lipid composition around the Ca2þ-ATPase could, therefore, lead to changes in the packing of the transmembrane helical bundle, likely to result in changes in function since this is a critical region for the Ca2þ-ATPase, containing the two Ca2þ-binding sites on the Ca2þ-ATPase. D. Curvature Stress Lipids with small polar headgroups, such as the phosphatidylethanolamines, are said to have a ‘‘conical’’ shape in contrast to a lipid such as a phosphatidylcholine with a larger headgroup, which is said to have a ‘‘cylindrical’’ shape. Whereas lipids with a cylindrical shape will pack into a planar bilayer structure, a conically shaped lipid has a tendency to form curved, hexagonal HII phases (Cullis and de KruijV, 1979). In a biological membrane, the presence of both the intrinsic membrane proteins and bilayer-preferring lipids will, however, force lipids such as the phosphatidylethanolamines to adopt a bilayer structure (Lee, 2004) and the lipid will, therefore, be in a state


167

of curvature stress. However, although the concept of a ‘‘conical’’ shape for a phosphatidylethanolamine is helpful in considering the phase preferences of the isolated lipid, it would not be accurate to say that a phosphatidylethanolamine adopted a ‘‘conical’’ shape when in a lipid bilayer; if it did, the presence of a phosphatidylethanolamine in a bilayer of phosphatidylcholine would create a greater packing density toward the center of the bilayer and a smaller packing density near the glycerol backbone region and thus increase order parameters for phosphatidylcholine chains at the terminal methyl ends of the chains and decrease order parameters at the carboxyl end. Such eVects are not seen, addition of a phosphatidylethanolamine to a bilayer of a phosphatidylcholine increasing order parameters at all positions in the chains of the phosphatidylcholine (Fenske et al., 1990). Further, it has been shown that, in mixtures of (C16:0, C18:1)PE and (C16:0, C18:1)PC, the order parameters for the palmitoyl chains in (C16:0, C18:1)PE and (C16:0, C18:1)PC are the same (Lafleur et al., 1990). This is another example of the important plasticity of the lipid bilayer. It has been suggested that the curvature elastic energy stored in a membrane could shift the equilibrium between conformational states of an intrinsic membrane protein to that with the greatest hydrophobic thickness (Fig. 11) (Botelho et al., 2002). The presence of a lipid such as phosphatidylethanolamine favoring negative curvature would favor the conformational state with largest hydrophobic thickness (Fig. 11). Extensive studies of the eVects of hydrophobic additives on the function of a sodium channel were shown to be consistent with such a model (Lundbaek et al., 2004). However, it is not yet clear how common will be large diVerences in hydrophobic thickness between the diVerent conformational states of a membrane protein. For the Ca2þ-ATPase where crystal structures are available for a number of conformational states (Moller et al., 2005; Obara et al., 2005),

FIGURE 11 Possible eVects of stored curvature elastic energy on the function of membrane proteins. The presence of lipids favoring the hexagonal HII phase shifts the conformational equilibrium of an intrinsic membrane protein toward the conformation with the greatest hydrophobic thickness.

168

Powl and Lee

it is clear that the diVerent conformations all have very similar hydrophobic thicknesses and yet the activity of the Ca2þ-ATPase is aVected by the presence of phosphatidylethanolamine (Starling et al., 1996a). The presence of phosphatidylethanolamine has been shown to aVect channel opening of MscL (Moe and Blount, 2005) and, as described later, the hydrophobic thickness of MscL probably decreases on channel opening. However, even for MscL molecular dynamics simulations show diVerent patterns of hydrogen bonding between MscL and phosphatidylethanolamines and phosphatidylcholines, associated with significant conformational changes on MscL (Elmore and Dougherty, 2003) that could be the underlying reason for the eVects of phosphatidylethanolamine on MscL function. E. Elastic Strain and Pressure Profiles At equilibrium in a bilayer, hydrophobic forces exactly balance the repulsive lateral pressures present in the bilayer, and there is no net tension in the membrane (Seddon, 1990; Marsh, 1996). These forces are illustrated in Fig. 12. At about the position of the glycerol backbone region, just below the lipid headgroups, an attractive force Fg arises from the unfavorable contact of the hydrocarbon chains with water (the hydrophobic eVect). Tight packing in this region ensures the minimum exposure of the hydrocarbon interior of the membrane to water, leading to a negative lateral pressure (a positive membrane tension), tending to contract the bilayer. In the headgroup region of the bilayer, a positive lateral pressure Fh arises because of steric, hydrational, and electrostatic eVects; these will normally be repulsive but may contain attractive contributions from, for example, hydrogen bonding interactions. Similarly, in the hydrocarbon interior of the membrane, attractive van der Waals interactions between the chains will be opposed by the repulsive interactions due to the thermal motions of the chains, the net eVect being a positive lateral pressure Fc tending to expand the membrane. For a membrane to stay flat, the forces illustrated in Fig. 12 must be in balance across the monolayer. We can say that a negative pressure in the glycerol backbone region, arising from the hydrophobic eVect and serving to contract the membrane, balances a positive pressure in the chain region, serving to expand the membrane. Since the positive pressure in the chain region arises from collisions between the chains, and since the extent of chain motion varies down the chain as shown by the order parameter profiles shown in Fig. 6, diVerent positions in the chains will make diVerent contributions to this positive pressure. This has given rise to the idea of a pressure profile within the lipid bilayer. Pressure is an inherently macroscopic property, and so the concept of a local pressure is not absolutely straightforward (Lindahl and Edholm, 2000).

169

6. Lipid EVects on Mechanosensitive Channels Fh

Interfacial

A

Fg

g

tension

Fc

B Pressure

FIGURE 12 Pressure profile in a lipid bilayer. At the top is shown the distribution of lateral pressures and tensions across a lipid monolayer. The repulsive lateral pressure Fc in the chain region is due to thermally activated bond rotational motion. The interfacial tension g, tending to minimize the interfacial area, arises from the hydrophobic eVect (unfavorable hydrocarbonwater contacts). Finally, the lateral pressure Fh in the headgroup region arises from steric, hydrational, and electrostatic eVects; it is normally repulsive, but may contain attractive contributions from, for example, hydrogen bonding interactions. After Seddon (1990). Below is the pressure profile in a lipid bilayer that will aVect the conformational change A Ð B for a membrane protein that involves a change in shape for the protein.

Nevertheless, pressure profiles can be calculated from molecular dynamics simulations by dividing the membrane up into a series of slices and calculating the pressures in each slice from the interactions between the atoms in that slice (Lindahl and Edholm, 2000; Gullingsrud and Schulten, 2004). As expected and as shown in Fig. 12, the distribution of lateral pressures within the fatty acyl chain region of a bilayer is not uniform (Cantor, 1999; Lindahl and Edholm, 2000; Gullingsrud and Schulten, 2003, 2004). The largest negative pressure occurs close to the backbone region of the bilayer, with the pressure of lowest magnitude in the center of the bilayer, corresponding to the region of lowest chain order parameter and lowest packing density where the ends of the chains from the two monolayers meet. The exact shape of the pressure profile was found to vary markedly with lipid structure (Cantor, 1999; Gullingsrud and Schulten, 2004). Cantor (1997) has suggested that this pressure profile could be important for membrane protein function. In classical physical chemistry, to increase the volume of an object under constant pressure requires work, the work term being PV where P is the applied pressure and V is the increase in volume. In a lipid bilayer, because

170

Powl and Lee

there is no net tension (all the pressures in the membrane cancel out), simple expansion of a membrane protein with no change in shape requires no work. However, if a conformational change A>B for a membrane protein involves a change in shape in the transmembrane region of the protein, with the crosssectional area of the protein at one depth in the membrane changing by more than the cross-sectional area at another depth, as illustrated in Fig. 12, then for the conformational change to occur, work will have to be done against the pressure profile since the PV terms across the bilayer will not cancel. These energy terms could be large because the pressures involved have been estimated to be hundreds of atmospheres, that near the backbone region, for example, being estimated to be ca. 1000 atm (Cantor, 1999; Gullingsrud and Schulten, 2004). However, the calculations assume that the pressure profile of the lipid bilayer does not change as a result of the presence of the membrane protein or as a result of the change in shape of the membrane protein, and this seems unlikely since a key feature of the lipid bilayer is its plasticity, as illustrated in Fig. 7. A high degree of plasticity in the lipid bilayer means that the lipid bilayer will be able easily to distort around a membrane protein and so will provide little in the way of an energy barrier to changes in local volume of the type illustrated in Fig. 12. Given that pressure profiles across lipid bilayers are predicted to change markedly with changes in the degree of unsaturation of the fatty acyl chains (Cantor, 1999), an experimental test for the importance of the pressure profile is to look for a marked dependence of protein function on the pattern of fatty acyl chain unsaturation. In fact, experiments in which Ca2þ-ATPase was reconstituted into a series of phosphatidylcholines containing fatty acyl chains of the same length but containing diVerent numbers of cis double bonds showed no dependence of activity on chain structure, as long as the chain length remained constant (East et al., 1984). F. General Features of Lipid–Protein Interactions A key feature of the lipid bilayer is its plasticity, allowing it to accommodate to the rough surface of a membrane protein and to accommodate readily those changes in membrane protein shape that are involved in function. The adaptations of the lipid bilayer required to accommodate a membrane protein are largely confined to the annular shell of lipid molecules surrounding the membrane protein (Lee, 1977) and, at least for some membrane proteins, activity is normal at low molar ratios of lipid to protein when only annular lipid is present (Warren et al., 1974). Exchange between annular and bulk lipids is generally fast (East et al., 1985), and selectivity in interactions between annular lipid and the transmembrane surface of a


171

membrane protein is correspondingly generally low (Lee, 2003). However, regions of high positive charge on a membrane protein close to the lipid headgroup region are likely to give rise to ‘‘hot-spots’’ where anionic lipids could bind with higher aYnity than zwitterionic lipids (Marius et al., 2005; Powl et al., 2005a; Lee, 2006).

V. WHAT DO THESE GENERAL PRINCIPLES TELL US ABOUT MscL? We would expect hydrophobic matching between MscL and the surrounding lipid bilayer to be highly eYcient. This appears to be the case; Trp residues introduced at the ends of the second transmembrane a-helix of MscL from Mycobacterium tuberculosis maintain their positions close to the glycerol backbone region of the lipid bilayer when the lipid fatty acyl chains are changed in length from C12 to C24 (Powl et al., 2005b). The hydrophobic thickness of MscL in the closed state has been estimated from ˚ (Powl et al., 2005b), in good Trp scanning fluorescence studies to be 26 A agreement with theoretical calculations based on the energetics of solvation of the protein in a model lipid bilayer (Lomize et al., 2006). Relative lipid-binding constants for MscL have been determined using a fluorescence quenching method, measuring the level of fluorescence quenching of Trp-containing mutants of MscL in mixtures of phospholipids with bromine-containing fatty acyl chains and non-bromine-containing fatty acyl chains (Powl et al., 2003). As shown in Fig. 13, the phosphatidylcholine that bound most strongly to MscL was di(C16:1)PC, a lipid which forms a ˚ (Powl et al., 2003), in good bilayer with a hydrophobic thickness of ca. 24 A ˚ (Powl et al., agreement with a hydrophobic thickness for MscL of 26 A 2005b). The hydrophobic thickness of the native M. tuberculosis membrane ˚ since the predominant lipid fatty acyl chains in the will be close to 24 A membrane are C18:1 and C16:0 (Coren, 1984). As shown in Fig. 13, the chain length dependence of lipid binding to MscL in the closed state is less marked than that for the b-barrel protein OmpF (O’KeeVe et al., 2000). b-Barrel proteins are relatively rigid so that, at least for moderate degrees of hydrophobic mismatch, hydrophobic matching can be expected to follow from distortion of the lipid bilayer around a relatively undistorted b-barrel (Lee, 2004). The fact that the variation in lipid-binding constant with chain length is less for MscL than for OmpF, therefore, suggests that hydrophobic matching for MscL is achieved by distorting both the protein and the lipid bilayer because it is less costly to distort an a-helical membrane protein than a b-barrel membrane protein (Powl et al., 2003). Distortion of an a-helical membrane protein is likely to involve tilting of the transmembrane a-helices, as illustrated in Fig. 8.

172

Powl and Lee

Relative binding constant

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 10

12

14 16 18 20 22 Fatty acyl chain length

24

26

FIGURE 13 Relative lipid-binding constants for MscL. Binding constants for phosphatidylcholines relative to that for di(C18:1)PC are plotted as a function of acyl chain length for the closed channel (○) and for a gain of function mutant (□). Data are compared with relative lipid-binding constants for OmpF (). Data from O’KeeVe et al. (2000), Powl et al. (2003), and Powl and Lee (2006).

To investigate lipid binding to the open state of the MscL channel, a gainof-function mutant was prepared (V21K) in which Val21, at the narrowest region of the channel, was mutated to Lys, where charge repulsion is expected to keep the channel in an open state (Powl et al., 2005a). The lipid binding most strongly to the V21K mutant is di(C14:1)PC (Fig. 13), a lipid ˚ , suggesting that giving a bilayer of hydrophobic thickness of ca. 20 A ˚ channel opening is associated with a ca. 4-A thinning of the lipid bilayer. This is consistent with the observation that decreasing bilayer thickness leads to a decrease in the tension required to open the MscL channel (Perozo et al., ˚ would be consistent with a ca. 16 increase in tilt 2002b). A thinning of 4 A angle for the transmembrane a-helices, in good agreement with the estimate made by Perozo et al. (2002a) for the E. coli protein. The eVect of lipid headgroup structure on the strength of lipid binding to MscL was investigated separately on the two sides of the membrane by introducing Trp residues on either the periplasmic or cytoplasmic sides of MscL (Powl et al., 2005a). These studies showed that there was no selectivity in binding lipid of diVerent classes on the periplasmic side of the membrane whereas on the cytoplasmic side, although phosphatidylcholines bound with equal aYnity to phosphatidylethanolamines, anionic lipids, particularly phosphatidic acid, bound with high aYnity to a site consisting of the three positively charged residues Arg98, Lys99, and Lys100 (Powl et al., 2005a). These positively charged residues, together with adjacent negatively charged residues Glu102 and Glu104, were suggested to act as a molecular ‘‘Velcro,’’

173


B

K100

K99

A

E102 R98

E

E102 E104

E104

E A B

FIGURE 14 The location of the positively charged cluster Arg98, Lys99, and Lys100 in MscL at the C-terminal end of TM2 of the MscL monomer. The locations of Arg98, Lys99, and Lys100 in the central monomer (A) of the homopentameric structure are shown surrounded by the anionic residues Glu102 and Glu104 in the neighboring subunits B and E. The side chain of Lys100 is not resolved in the crystal structure.

holding together the ends of the TM helices in the closed form of the channel (Fig. 14). Opening the channel by helix tilting would move the ends of the helices further apart, only possible if the charge interactions observed in the closed channel were broken up. Indeed, simultaneous mutation of the three positively charged residues led to formation of the open channel (Powl et al., 2005a). In the V21K open state mutant, the specific binding of anionic lipid was lost, showing that in the open state of the channel the positively charged cluster must have been repositioned so that high aYnity interaction with anionic lipid headgroups is no longer possible (Powl et al., 2005a). The presence of a cluster of positively and negatively charged residues appears to be a characteristic of all MscL channels, although, given that the positively and negatively charged residues are located in a loop region of the structure (Fig. 14), it would not be expected that the positions of the charged residues would be absolutely conserved in the MscL sequences. Deletion of the charge cluster in the C-terminal region of E. coli MscL leads to loss of function, providing additional evidence that this region of MscL has an important role in channel function (Blount et al., 1996). On the basis of

174

Powl and Lee

results obtained with a set of charge reversal mutants, Kloda et al. (2006) suggested that the charge cluster could act as a pH sensor in E. coli MscL. Interaction of anionic lipid with the positively charged cluster on MscL seems not to be functionally important since Moe and Blount (2005) found that the tension required to open the MscL channel from E. coli was unaVected by the presence of anionic lipid. In contrast, Moe and Blount (2005) found that the presence of phosphatidylethanolamine increased the tension required to open the MscL channel, even though we found no selectivity for binding phosphatidylethanolamine over phosphatidylcholine, on either side of the membrane (Powl et al., 2005a). Molecular dynamics simulations of MscL in bilayers of phosphatidylcholine and phosphatidylethanolamine suggest that the diVerent patterns of hydrogen bonding to the phosphatidylcholine and phosphatidylethanolamine headgroups led to changes in protein structure (Elmore and Dougherty, 2003), and it is possible that these were responsible for the observed changes in tension for channel opening. Trp fluorescence spectroscopy has also been used to investigate the role of lipid in maintaining MscL in the closed state. Wiggins and Phillips (2004) proposed that the closed state of the channel is stabilized by hydrophobic mismatch between the channel and the surrounding lipid bilayer and that energy diVerences between the open and closed states of the channel are dominated by bilayer deformation energies rather than by energy diVerences between the diVerent conformational states of the protein itself. A Trp residue introduced into position 80 in the mutant F80W shows fluorescence emission centered at ca. 321 nm in the closed channel, but emission shifts to ca. 332 nm in the open mutant F80W:V21K (Powl et al., 2005a). In a wide variety of detergents, fluorescence emission for F80W remained at ca. 321 nm, suggesting that a lipid bilayer is not required to maintain the MscL channel in the closed state and that the closed state of the channel is the most stable state for the wild-type protein (Powl and Lee, 2006). The driving force for channel opening would then be the less tight packing of the lipid molecules in the stretched membrane, leading to potential exposure of hydrophobic residues in the transmembrane a-helices to water, an exposure that can be minimized by expansion of the membrane protein to maintain good contact with the surrounding lipid bilayer. References Bloom, M., Evans, E., and Mouritsen, O. G. (1991). Physical properties of the fluid lipid-bilayer component of cell membranes: A perspective. Q. Rev. Biophys. 24, 293–397. Blount, P., Sukharev, S. I., Schroeder, M. J., Nagle, S. K., and Kung, C. (1996). Single residue substitutions that change the gating properties of a mechanosensitive channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 93, 11652–11657.


175

Boni, L. T., and Hui, S. W. (1983). Polymorphic phase behaviour of dilinoleoylphosphatidylethanolamine and palmitoyloleoylphosphatidylcholine mixtures. Structural changes between hexagonal, cubic and bilayer phases. Biochim. Biophys. Acta 731, 177–185. Botelho, A. V., Gibson, N. J., Thurmond, R. L., Wand, Y., and Brown, M. F. (2002). Conformational energetics of rhodopsin modulated by nonlamellar-forming lipids. Biochemistry 41, 6354–6368. Buldt, G., and Wohlgemuth, R. (1981). The headgroup conformation of phospholipids in membranes. J. Membr. Biol. 58, 81–100. Cantor, R. S. (1997). Lateral pressures in cell membranes: A mechanism for modulation of protein function. J. Phys. Chem. B 101, 1723–1725. Cantor, R. S. (1999). Lipid composition and the lateral pressure profile in bilayers. Biophys. J. 76, 2625–2639. Coren, M. B. (1984). ‘‘The Mycobacteria. Biosynthesis and Structures of Phospholipids and Sulfatides’’ (G. P. Kubica and L. G. Wayne, eds.), pp. 379–415. Marcel Dekker, New York. Cullis, P. R., and de KruijV, B. (1979). Lipid polymorphism and the functional roles of lipids in biological membranes. Biochim. Biophys. Acta 559, 399–420. Dalton, K. A., East, J. M., Mall, S., Oliver, S., Starling, A. P., and Lee, A. G. (1998). Interaction of phosphatidic acid and phosphatidylserine with the Ca2þ-ATPase of sarcoplasmic reticulum and the mechanism of inhibition. Biochem. J. 329, 637–646. Damodaran, K. V., Merz, K. M., and Gaber, B. P. (1995). Interaction of small peptides with lipid bilayers. Biophys. J. 69, 1299–1308. Duong-Ly, K. C., Nanda, V., DeGrado, W. F., and Howard, K. P. (2005). The conformation of the pore region of the M2 proton channel depends on lipid bilayer environment. Protein Sci. 14, 856–861. East, J. M., and Lee, A. G. (1982). Lipid selectivity of the calcium and magnesium ion dependent adenosinetriphosphatase, studied with fluorescence quenching by a brominated phospholipid. Biochemistry 21, 4144–4151. East, J. M., Jones, O. T., Simmonds, A. C., and Lee, A. G. (1984). Membrane fluidity is not an important physiological regulator of the (Ca2þ-Mg2þ)-dependent ATPase of sarcoplasmic reticulum. J. Biol. Chem. 259, 8070–8071. East, J. M., Melville, D., and Lee, A. G. (1985). Exchange rates and numbers of annular lipids for the calcium and magnesium ion dependent adenosinetriphosphatase. Biochemistry 24, 2615–2623. Elmore, D. E., and Dougherty, D. A. (2003). Investigating lipid composition eVects on the mechanosensitive channel of large conductance (MscL) using molecular dynamics simulations. Biophys. J. 85, 1512–1524. Feller, S. E., Gawrisch, K., and MacKerell, A. D. (2002). Polyunsaturated fatty acids in lipid bilayers: Intrinsic and environmental contributions to their unique physical properties. J. Am. Chem. Soc. 124, 318–326. Fenske, D. B., Jarrell, H. C., Guo, Y., and Hui, S. W. (1990). EVect of unsaturated phosphatidylethanolamine on the chain order profile of bilayers at the onset of the hexagonal phase transition. A2H NMR study. Biochemistry 29, 11222–11229. Froud, R. J., Earl, C. R. A., East, J. M., and Lee, A. G. (1986a). EVects of lipid fatty acyl chain structure on the activity of the (Ca2þ-Mg2þ)-ATPase. Biochim. Biophys. Acta 860, 354–360. Froud, R. J., East, J. M., Jones, O. T., and Lee, A. G. (1986b). EVects of lipids and long-chain alkyl derivatives on the activity of (Ca2þ-Mg2þ)-ATPase. Biochemistry 25, 7544–7552.

176

Powl and Lee

Grossfield, A., Feller, S. E., and Pitman, M. C. (2006). A role for direct interactions in the modulation of rhodopsin by o-3 polyunsaturated lipids. Proc. Natl. Acad. Sci. USA 103, 4888–4893. Gullingsrud, J., and Schulten, K. (2003). Gating of MscL studied by steered molecular dynamics. Biophys. J. 85, 2087–2099. Gullingsrud, J., and Schulten, K. (2004). Lipid bilayer pressure profiles and mechanosensitive channel gating. Biophys. J. 86, 3496–3509. Heller, H., Schaefer, M., and Schulten, K. (1993). Molecular dynamics simulation of a bilayer of 200 lipids in the gel and in the liquid-crystal phases. J. Phys. Chem. 97, 8343–8360. Holte, L. L., Peter, S. A., Sinnwell, T. M., and Gawrisch, K. (1995). 2H Nuclear magnetic resonance order parameter profiles suggest a change of molecular shape for phosphatidylcholines containing a polyunsaturated acyl chain. Biophys. J. 68, 2396–2403. Huang, P., Perez, J. J., and Loew, G. H. (1994). Molecular dynamics simulations of phospholipid bilayers. J. Biomolec. Struct. Dynamics 5, 927–956. Hyvonen, M. T., Rantala, T. T., and Ala-Korpela, M. (1997). Structure and dynamic properties of diunsaturated 1- palmitoyl-2- linoleoyl- sn-glycero-3-phosphatidylcholine lipid bilayer from molecular dynamics simulation. Biophys. J. 73, 2907–2923. Kandasamy, S. K., and Larson, R. G. (2006). Molecular dynamics simulations of model transmembrane peptides in lipid bilayers: A systematic investigation of hydrophobic mismatch. Biophys. J. 90, 2326–2343. Kloda, A., Ghazi, A., and Martinac, B. (2006). C-terminal charged cluster of MscL, RKKEE, functions as a pH sensor. Biophys. J. 90, 1992–1998. Lafleur, M., Cullis, P. R., and Bloom, M. (1990). Modulation of the orientational order profile of the lipid acyl chain in the La phase. Eur. Biophys. J. 19, 55–62. Lague, P., Roux, B., and Pastor, R. W. (2005). Molecular dynamics simulations of the influenza hemagglutinin fusion peptide in micelles and bilayers: Conformational analysis of peptide and lipids. J. Mol. Biol. 354, 1129–1141. Lee, A. G. (1977). Annular events: Lipid-protein interactions. Trends Biochem. Sci. 2, 231–233. Lee, A. G. (1991). Lipids and their eVects on membrane proteins: Evidence against a role for fluidity. Prog. Lipid Res. 30, 323–348. Lee, A. G. (2002). Ca2þ-ATPase structure in the E1 and E2 conformations: Mechanism, helixhelix and helix-lipid interactions. Biochim. Biophys. Acta 1565, 246–266. Lee, A. G. (2003). Lipid-protein interactions in biological membranes: A structural perspective. Biochim. Biophys. Acta 1612, 1–40. Lee, A. G. (2004). How lipids aVect the activities of integral membrane proteins. Biochim. Biophys. Acta 1666, 62–87. Lee, A. G. (2005). How lipids and proteins interact in a membrane: A molecular approach. Mol. BioSyst. 1, 1–11. Lee, A. G. (2006). Lipid interactions with ion channels. Future Lipidology 1, 103–113. Li, S., Lin, H. N., Wang, Z. Q., and Huang, C. (1994). Identification and characterization of kink motifs in 1-palmitoyl-2-oleoyl-phosphatidylcholines: A molecular mechanics study. Biophys. J. 66, 2005–2018. Lindahl, E., and Edholm, O. (2000). Spatial and energetic-entropic decomposition of surface tension in lipid bilayers from molecular dynamics simulations. J. Chem. Phys. 113, 3882–3893. Lomize, M. A., Lomize, A. L., Pogozheva, I. D., and Mosberg, H. I. (2006). OPM: Orientations of proteins in membranes database. Bioinformatics 22, 623–625. Lundbaek, J. A., and Andersen, O. S. (1999). Spring constants for channel-induced lipid bilayer deformations estimated using gramicidin channels. Biophys. J. 76, 889–895.


177

Lundbaek, J. A., Birn, P., Hansen, A. J., Sogaard, R., Nielsen, C., Girshman, J., Bruno, M. J., Tape, S. E., Egebjerg, J., Greathouse, D. V., Mattice, G. L., Koeppe, R. E., et al. (2004). Regulation of sodium channel function by bilayer elasticity: The importance of hydrophobic coupling. EVects of micelle-forming amphiphiles and cholesterol. J. Gen. Physiol. 123, 599–621. Marius, P., Alvis, S. J., East, J. M., and Lee, A. G. (2005). The interfacial lipid binding site on the potassium channel KcsA is specific for anionic phospholipids. Biophys. J. 89, 4081–4089. Marsh, D. (1996). Lateral pressure in membranes. Biochim. Biophys. Acta 1286, 183–223. Michelangeli, F., Orlowski, S., Champeil, P., Grimes, E. A., East, J. M., and Lee, A. G. (1990). EVects of phospholipids on binding of calcium to (Ca2þ-Mg2þ)-ATPase. Biochemistry 29, 8307–8312. Michelangeli, F., Grimes, E. A., East, J. M., and Lee, A. G. (1991). EVects of phospholipids on the function of the (Ca2þ-Mg2þ)-ATPase. Biochemistry 30, 342–351. Moe, P., and Blount, P. (2005). Assessment of potential stimuli for mechano-dependent gating of MscL: EVects of pressure, tension, and lipid headgroups. Biochemistry 44, 12239–12244. Moller, J. V., Nissen, P., Sorensen, T. L. M., and le Maire, M. (2005). Transport mechanism of the sarcoplasmic reticulum Ca2þ-ATPase pump. Curr. Opin. Struct. Biol. 15, 387–393. Nagle, J. F., and Wiener, M. C. (1988). Structure of fully hydrated bilayer dispersions. Biochim. Biophys. Acta 942, 1–10. O’KeeVe, A. H., East, J. M., and Lee, A. G. (2000). Selectivity in lipid binding to the bacterial outer membrane protein OmpF. Biophys. J. 79, 2066–2074. Obara, K., Miyashita, N., Xu, C., Toyoshima, L., Sugita, Y., Inesi, G., and Toyoshima, C. (2005). Structural role of countertransport revealed in Ca2þpump crystal structure in the absence of Ca2þ. Proc. Natl. Acad. Sci. USA 102, 14489–14496. Park, S. H., and Opella, S. J. (2005). Tilt angle of a trans-membrane helix is determined by hydrophobic mismatch. J. Mol. Biol. 350, 310–318. Perly, B., Smith, I. C. P., and Jarrell, H. C. (1985). EVects of the replacement of a double bond by a cyclopropane ring in phosphatidylethanolamines: A deuterium NMR study of phase transitions and molecular organization. Biochemistry 24, 1055–1063. Perozo, E., Cortes, D. M., Sompornpisut, P., and Martinac, B. (2002a). Open channel structure of MscL and the gating mechanism of mechanosensitive channels. Nature 418, 942–948. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2002b). Physical principles underlying the transduction of bilayer deformation forces during mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703. Petrache, H. I., Feller, S. E., and Nagle, J. F. (1997). Determination of component volumes of lipid bilayers from simulations. Biophys. J. 70, 2237–2242. Pilot, J. D., East, J. M., and Lee, A. G. (2001a). EVects of bilayer thickness on the activity of diacylglycerol kinase of Escherichia coli. Biochemistry 40, 8188–8195. Pilot, J. D., East, J. M., and Lee, A. G. (2001b). EVects of phospholipid headgroup and phase on the activity of diacylglycerol kinase of Escherichia coli. Biochemistry 40, 14891–14897. Poastor, R. W., Venable, R. M., Karplus, M., and Szabo, A. (1988). A simulation based model of NMR T1 relaxation in lipid bilayer vesicles. J. Chem. Phys. 89, 1128–1140. Powl, A. M., and Lee, A. G. (2006). Unpublished observations. Powl, A. M., East, J. M., and Lee, A. G. (2003). Lipid-protein interactions studied by introduction of a tryptophan residue: The mechanosensitive channel MscL. Biochemistry 42, 14306–14317. Powl, A. M., East, J. M., and Lee, A. G. (2005a). Heterogeneity in the binding of lipid molecules to the surface of a membrane protein: Hot-spots for anionic lipids on the

178

Powl and Lee

mechanosensitive channel of large conductance MscL and eVects on conformation. Biochemistry 44, 5873–5883. Powl, A. M., Wright, J. N., East, J. M., and Lee, A. G. (2005b). Identification of the hydrophobic thickness of a membrane protein using fluorescence spectroscopy: Studies with the mechanosensitive channel MscL. Biochemistry 44, 5713–5721. Seddon, J. M. (1990). Structure of the inverted hexagonal (HII) phase, and non-lamellar phase transitions of lipids. Biochim. Biophys. Acta 1031, 1–69. Seelig, J., and Seelig, A. (1980). Lipid conformation in model membranes. Q. Rev. Biophys. 13, 19–61. Small, D. M. (1986). ‘‘The Physical Chemistry of Lipids. Handbook of Lipid Research Volume 4.’’ Plenum Press, New York. Starling, A. P., East, J. M., and Lee, A. G. (1993). EVects of phosphatidylcholine fatty acyl chain length on calcium binding and other functions of the (Ca2þ-Mg2þ)-ATPase. Biochemistry 32, 1593–1600. Starling, A. P., East, J. M., and Lee, A. G. (1995a). EVects of phospholipid fatty acyl chain length on phosphorylation and dephosphorylation of the Ca2þ-ATPase. Biochem. J. 310, 875–879. Starling, A. P., East, J. M., and Lee, A. G. (1995b). Evidence that the eVects of phospholipids on the activity of the Ca2þ-ATPase do not involve aggregation. Biochem. J. 308, 343–346. Starling, A. P., Dalton, K. A., East, J. M., Oliver, S., and Lee, A. G. (1996a). EVects of phosphatidylethanolamines on the activity of the Ca2þ-ATPase of sarcoplasmic reticulum. Biochem. J. 320, 309–314. Starling, A. P., East, J. M., and Lee, A. G. (1996b). Separate eVects of long-chain phosphatidylcholines on dephosphorylation of the Ca2þ-ATPase and on Ca2þbinding. Biochem. J. 318, 785–788. Stouch, T. R., Alper, H. E., and Bassolino-Klimas, D. (1994). Supercomputing studies of biomembranes. Supercomputer Appl. High Performance Comput. 8, 6–23. Warren, G. B., Toon, P. A., Birdsall, N. J. M., Lee, A. G., and Metcalfe, J. C. (1974). Reversible lipid titrations of the activity of pure adenosine triphosphatase-lipid complexes. Biochemistry 13, 5501–5507. Webb, R. J., East, J. M., Sharma, R. P., and Lee, A. G. (1998). Hydrophobic mismatch and the incorporation of peptides into lipid bilayers: A possible mechanism for Golgi retention. Biochemistry 37, 673–679. Wiggins, P., and Phillips, R. (2004). Analytic models for mechanotransduction: Gating a mechanosensitive channel. Proc. Natl. Acad. Sci. USA 101, 4071–4076. Williamson, I. M., Alvis, S. J., East, J. M., and Lee, A. G. (2002). Interactions of phospholipids with the potassium channel KcsA. Biophys. J. 83, 2026–2038.

CHAPTER 7 Functional Interactions of the Extracellular Matrix with Mechanosensitive Channels Anita Sengupta* and Christopher A. McCulloch{ *Department of Anatomy, University of Bristol, School of Medical Sciences, University Walk, Clifton, BS8 1TD Bristol, United Kingdom { CIHR Group in Matrix Dynamics, University of Toronto, Toronto, Ontario, Canada M5S 3E2

I. II. III. IV. V.

Overview Mechanotransduction Mechanosensitive Channels in Connective Tissue Cells The Extracellular Environment of Cells Force Transmission from Matrix to Cytoskeleton A. Focal Adhesions B. Selectins VI. Experimental Models of Force Application to Connective Tissue Cells VII. EVects of Force on Cell Surface Structures VIII. Future Approaches References

I. OVERVIEW Mechanical stimuli generate responses in many types of mammalian cells that interact with the extracellular matrix. As the extracellular matrix is a potential force conductor, in many tissues mechanical loads can be directed through fibrillar matrix proteins to matrix receptors such as integrins. These transmembrane receptors direct forces into the cytoskeleton and can influence the gating properties of mechanosensitive channels in the adjacent plasma membrane. Stretch-sensitive channels colocalize with integrins and Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58007-X

180

Sengupta and McCulloch

functional studies indicate that stable matrix-integrin attachments are required for stretch activation. We review here the mechanical properties of prominent matrix proteins, the nature of cell attachments to the matrix, how cell-matrix attachments interact with the cytoskeleton to regulate stretch sensitivity of mechanosensitive channels and contemplate how matrix proteins may interact with mechanosensors to eVect mechanotransduction. These findings point to conserved regulatory mechanisms by which cells in vertebrates respond to external forces and convert these forces into signals that mediate alterations in the structure and function of connective tissues.

II. MECHANOTRANSDUCTION Cellular mechanics can be considered as a three-step process comprising mechanosensation, mechanotransduction, and mechanoresponse (Vogel and Sheetz, 2006). Mechanosensation involves the ability of cells to detect forces and to explore the topography of the extracellular matrix. This information about the physical environment is then translated into intracellular biochemical signals, a process known as mechanotransduction that leads to the generation of cellular responses (mechanoresponses) which may impact the structure and function of the extracellular matrix itself. Target cell adaptation and desensitization of mechanosensitive channels to constant or repeated mechanical stimulation in connective tissue fibroblasts has been reported (Glogauer et al., 1997) and may prevent ‘‘information overload’’ including overt Ca2þ induced cell death. Consequently, cellular adaptations to applied mechanical forces, including the regulation of mechanosensitive channel function, is important for optimizing cellular responses to fluctuations in the physical environment of cells (Alberts et al., 2002). Mechanosensing, mechanotransduction, and mechanoresponses comprise tightly integrated processes involving extracellular matrix, cytoskeletal, and signaling proteins that ultimately provide cells with the ability to respond and adapt to applied physical forces. Direct mechanical influences that evoke cellular responses may impact the cell membrane and include stress, strain, fluid shear, hydrostatic pressure, volume change, osmolarity, and geometry sensing (Orr et al., 2006). In particular, mechanoresponses are dependent on the magnitude and rate of loading or complexity of surface geometry as well as the age and diVerentiation state of the cell (Clements et al., 2001). Some of the earliest mechanoresponses following mechanical stimulation of connective tissue cells are transient increases in intracellular calcium or reorganization of the cortical cytoskeleton. Longer term responses include the secretion of cell mediators into the matrix such as prostaglandin E2 and interleukin-1, signals that may increase cell proliferation and upregulate

7. Matrix and Mechanosensitive Channels

181

the synthesis of matrix components, matrix degrading enzymes, and growth factors (Chiquet, 1999). Mechanoresponses are complex and diverse but the pathways can be broadly characterized as associated with physical force or those involved in geometry sensing and cell motility (Vogel and Sheetz, 2006). Cytoprotection is a group of individual mechanoresponses that serve to protect the cell from potentially harmful mechanical forces, possibly leading to cell death and loss of tissue homeostasis. These responses can include reinforcement of the cell membrane to prevent excessive mechanical deformation (e.g., assembly of actin filaments in orthogonal arrays; Glogauer et al., 1998) or the operation of stretch-inactivated ion channels (Gu et al., 2001). Some of these adaptations include the ability of cells to repair damaged cell membranes and to survive high amplitude forces without undergoing cell death (Kainulainen et al., 2002; McNeil and Kirchhausen, 2005) and the ability to remodel the extracellular matrix in response to applied forces (Ozaki et al., 2005). Cytoprotection manifests in the response to sustained loading which in fibroblasts results in an upregulation in the expression of actin cross-linking proteins that enhance local membrane rigidity via subcortical actin filaments (D’Addario et al., 2001). It seems likely that at least some of the proteins involved in mechanotransduction can form aggregates that optimize the transmission and processing of mechanically induced signals (Tzima et al., 2005). Mechanosensitive channels are functionally linked to transmembrane proteins that provide anchorage for cells to the extracellular matrix (e.g., integrins; Glogauer et al., 1997) or for intercellular adhesions (e.g., cadherins; Ko et al., 2001). These attachment proteins are linked by actin-binding proteins to the subcortical actin network and form complexes with proteins that regulate tension in the cell membrane and the conductance of mechanosensitive channels. Indeed, one of the central notions of the regulation of ion channel conductance by attachment proteins and the cytoskeleton is that cell membranes are important targets for delivery of exogenous mechanical loads (Martinac, 2004) that can then be translated into membrane tension. Conceivably, increased membrane tension can lead to increased ion channel conductance, and ultimately, altered gene expression. In cells of connective tissues, force transmission through fibrillar extracellular matrix proteins (Provenzano and Vanderby, 2006) can be focused on to discrete patches of the cell membrane (Fig. 1) through the organization of matrix receptors into clusters such as focal complexes or the more mature and extensively cross-linked focal adhesions (Galbraith et al., 2002). As the formation of focal adhesions is dependent on cell-generated tension by myosin motors bound to actin, why do cells anchored to the matrix not spontaneously activate mechanosensitive channels as has been reported for

182


FIGURE 1 Diagram to illustrate potential functional relationship between tensile forces applied through collagen fibrils to integrins and how these tensile forces may impact adjacent mechanosensitive ion channels.

migrating cells (Lee et al., 1999)? Presumably, if the matrix to which cells are anchored is suYciently stiV, cells resist the deformations generated by internal forces (Pelham and Wang, 1997) and the stretch channels will not be activated. Evidently, there are very precise balances in cells between matrixapplied forces, cell deformation, membrane tension, and cytoskeletal elements that resist deformation (Coughlin et al., 2006). The impact of these balances on alterations of cell metabolism likely depends on the activity of the cell at any given moment (migrating, quiescent, dividing, undergoing diVerentiation, subjected to tension), its relationship with adjacent cells (via intercellular adhesions), and its attachment to the substrate (extracellular matrix). We explore below how elements of the extracellular matrix not only provide force transmission to cells but also aid in force sensing and in determining cellular responses to force. The generation of mechanically induced responses may facilitate the migration of cells through the matrix and the organization of tissues, particularly as regulated by matrix adhesion receptors that are deeply involved in mechanotransduction processes (Katsum et al., 2004).

III. MECHANOSENSITIVE CHANNELS IN CONNECTIVE TISSUE CELLS Mechanical loading of bone leads to tissue remodeling (Lanyon, 1984) as does force application to soft connective tissues such as the periodontal ligament (Bumann et al., 1997). Connective tissues are evidently impacted by applied forces and are an attractive system for determining how force leads to alteration of gene expression and matrix remodeling. Bone remodeling in particular has highlighted the characterization of mechanosensitive


183

ion channels in cultured osteoblastic cells and osteoclasts. By patch clamping, Davidson et al. (1990) identified three classes of channels in osteoblastic cells that were characterized on the basis of conductance, ionic selectivity, and sensitivity to membrane tension. The coexistence of mechanosensitive nonselective cation and Kþ-selective channels in these cells indicated that applied forces could promote either membrane hyperpolarization, depolarization or a multiphasic response, depending on the density of the channels in the deformed region of the membrane. One of the diYculties with these types of patch-clamp studies is to relate the magnitude of the membrane deformations to physiological bone matrix turnover since the stresses induced by the patch pipette might be very diVerent than those encountered in vivo (Sachs, 1988). Ypey et al. (1992) also found stretch-activated Kþ-selective channels in chick osteoclasts, thereby providing the basis for a cellular system by which both synthetic and resorptive cells of the bone matrix could be integrated into a stretch-sensitive tissue remodeling system. In studies that were related to physiological regulatory systems of osteoblasts, Duncan and colleagues (Ryder and Duncan, 2001) used shear forces (1 dyne/cm2) applied to cultured murine osteoblastic cells to examine interactions between parathyroid hormone and fluid shear-induced Ca2þ signals. Notably, parathyroid hormone is an osteotropic hormone that, when combined with mechanical stimulation, increases bone mass. Their data indicated that parathyroid hormone enhanced the Ca2þ response to shear force by protein kinase C modulation of mechanosensitive cation channels and voltagesensitive Ca2þ channels. While the identity of these channels has not been determined, reconstitution experiments using expression of the a-subunit of the epithelial Naþ channel from osteoblasts demonstrated a Ca2þ-permeable, cation-selective, stretch-sensitive channels with some of the expected properties of mechanosensitive channel in intact osteoblasts. RT-PCR, Western blotting, and immunohistochemistry have been used to confirm that human-derived osteoblasts and MG63 cells express TREK-1 mRNA and protein. In human osteoblasts, functional expression of TREK-1 indicates that these channels may contribute to the resting membrane potential of human osteoblast cells (Hughes et al., 2006). Recent data indicate that the focal adhesion kinase pp125FAK interacts with the large conductance Ca2þactivated hSlo Kþ channel in human osteoblasts, thereby suggesting a potential role in mechanotransduction (Rezzonico et al., 2003). Evidently, there is an important role for mechanosensitive channels in the physiological regulation of connective tissues but currently, the identity and molecular regulation of these channels, as well as their functional interactions with matrix elements are not well defined. Below, we consider the molecular determinants of connective tissues that may be important for the regulation and function of mechanosensitive channels.

184


IV. THE EXTRACELLULAR ENVIRONMENT OF CELLS Cells of soft and mineralized connective tissues reside within an extracellular matrix that they may elaborate themselves (e.g., dermal fibroblasts in the lamina propria of the skin or osteoblasts in the bone matrix) or into which they have migrated (e.g., the myofibroblasts in the provisional matrix of healing wounds) or in which they become embedded during tissue formation (e.g., chondrocytes of cartilage; osteocytes of mineralized bone). In other types of tissues, epithelial cells (e.g., gastrointestinal-lining cells) and endothelial cells (blood vessel-lining cells) attach to underlying connective tissue matrices that provide anchorage, metabolic exchange, and the molecular factors required for growth and diVerentiation. Thus, in many tissues that are subjected to either gravitational, muscle, or endogenous cellgenerated forces, the extracellular matrix, in addition to its roles in anchorage, metabolism, and chemical signaling, also provides a medium through which mechanical signals can be directed to impact cell function. Notably, many types of connective tissue cells such as fibroblasts exhibit extensive intercellular junctions (Beertsen and Everts, 1980; El-Sayegh et al., 2005), enabling them to function as part of a syncitium and thereby facilitate transmission of mechanical forces between cells and throughout a tissue (Xia and Ferrier, 1992; Boitano et al., 1994). Consequently, in connective tissues, mechanosensing and mechanotransduction processes may involve force transmission through fibrillar elements of the extracellular matrix and/or by direct force transmission through intercellular adhesions. Sensing mechanical signals from the immediate matrix environment is important not only for appropriate cellular adaptation and mechanoprotective phenomena (McNeil and Kirchhausen, 2005) but for linkage of mechanotransduction processes to degradation (e.g., phagocytosis; Beningo and Wang, 2002) and tissue synthesis and remodeling (Guo et al., 2006). A combination of synthetic and degradative processes is involved in extracellular matrix remodeling and for mediating adaptations of the biomechanical properties of the matrix in response to metabolic and functional demands (Goodship et al., 1979). Collectively, these adaptive processes act to protect the tissue as a whole, particularly if subjected to high amplitude mechanical loading. This is an important determinant of tissue homeostasis since during physiological ageing there is an overall decrease in cell density and global cellular responsiveness to external stimuli. Accordingly, tissue repair responses may be outpaced by accumulating fatigue failure in fibrillar matrix proteins, thereby illustrating the vital role of cells in maintaining the biomechanical health of tissues in both developing and aging mammals (Clements et al., 2001, 2004). Mechanical forces that act on cells vary in frequency, amplitude, and duration and are in turn further diversified by the composition and biomechanical


185

properties of the extracellular matrices that ‘‘deliver’’ forces to resident cells. Connective tissues exhibit, in general, three types of matrices: (1) mainly fibrillar (e.g., tendons and ligaments); (2) a mixture of fibers and specialized ground substance (e.g., bone); and (3) fluids (e.g., blood) (Nordin and Frankel, 1980). The relative abundance of fibers and, in particular, the mineralized components of connective tissues have very large eVects on the mechanical properties of specific connective tissues and on the delivery of mechanical forces to cells. While shear forces in the arterial blood stream may be transmitted directly to circulating blood cells, compressive, bending, torsion, and tensile forces are largely dissipated after delivery to osteocytes embedded in the bone matrix. Indeed, the ability of osteocytes to function as tissue mechanosensors for bone relies in part on their exquisite sensitivity to minute changes in fluid flow through the surrounding osteocyte lacuna (Han et al., 2004). In health, there is an optimal distribution of forces between matrix and cell. The extracellular matrix provides stress shielding to cells but, in situations of high amplitude mechanical force loading, forces can exceed cell- and matrixdependent protective systems to induced cell loss by apoptosis (Chen et al., 2003; Goga et al., 2006). Extracellular matrix composition is governed by functional demands which in turn determine the physical nature of the matrix. Reciprocal relationships between function and form, and the physical properties of connective tissues, impact not only cells that reside within connective tissues such as fibroblasts and osteocytes, but also regulate the forces applied to cells which are attached to connective tissues such as endothelial or epithelial lining cells. Bone, cartilage, enamel, and dentine are subjected to very high compressive forces and deformation of these tissues is resisted by their high content of hydroxyapatite, calcium phosphate lattices that are extremely rigid. Many load-bearing tissues also have a high concentration of collagen (Oloyede and Broom, 1996). For fibrillar collagens, the tight, triple helical arrangement of tropocollagen endows collagen fibers with an extraordinarily high tensile strength and resistance to shear deformation (Li et al., 2005). Collagens also exhibit critical molecular domains that are required for cell attachment, diVerentiation and have been shown to associate with mechanosensitive channels (Liu et al., 1996). Collagen fibers are, however, very narrow and are subject to buckling under compression; when subjected to excessive mechanical loading, collagen fibers exhibit thickening and deformation (Kaab et al., 2000). In cartilage, tissue resistance to compressive deformation is dependent on the function of a gel phase comprising proteoglycans whose long and highly charged side chains can bind high molar concentrations of water. The combination of collagen fibrils and the gel matrix in articular collagen and intervertebral discs provides these tissues with enhanced viscoelastic properties, which are seen as creep and stress relaxation during

186


complete recovery from deforming loads (Adams et al., 1996; Kerin et al., 1998). In addition to the extracellular matrix that surrounds connective tissue cells, many cell types exhibit a surface coating of matrix molecules such as glycoproteins (e.g., fibronectin), proteoglycans (e.g., heparan sulfate), glycolipids, and other cell adhesive molecules that comprise the glycocalyx, a structure that varies in thickness in diVerent cell types and is modulated by the level of cell diVerentiation (Sengupta et al., 2000) (Fig. 2). In endothelium, shear forces from blood flow are dampened by the glycocalyx before they are delivered to the cell membrane, underlining the importance of the glycocalyx in regulating mechanical signaling. The glycocalyx is composed of proteins with highly glycosylated extracellular domains, and some proteins with membrane-spanning domains and short cytoplasmic tails. Notably, in endothelial cells, heparan sulfate proteoglycan may play a critical role as a mechanosensor in mechanical signaling events related to fluid flow (Florian et al., 2003). The glycocalyx overlying the endothelial surface layer can extend up to several microns peripheral to the plasma membrane; when the glycocalyx is intact, it can dissipate fluid shear at the plasma membrane of endothelial cells to near zero levels (Tarbell and Pahakis, 2006). The total bulk of charged carbohydrate moieties oVers resistance to cellular deformation, while specific end sugars may serve as mechanosensors. For cells to detect forces within the extracellular matrix, it would seem that mechanosensors must be located at the glycocalyx surface or protrude beyond.

FIGURE 2 Diagram to show cell surface glycocalyx and possible interactions with membrane-associated cell adhesion receptors.


187

V. FORCE TRANSMISSION FROM MATRIX TO CYTOSKELETON An eVective transducer of mechanical forces that can ultimately alter cellular metabolism would be expected to exhibit membrane-spanning domains and a functional association with the cytoskeleton (Watson, 1991). The integrin superfamily of matrix adhesion receptors seems well-suited for this purpose and these molecules have been considered as critical elements of the mechanosensory machinery (Katsum et al., 2004). Integrins provide a critical link for transmembrane communication between the extracellular matrix and subcortical actin filaments via specialized adhesions that comprise a large array of actin-binding proteins, signaling proteins, and the termini of actin filaments (Galbraith et al., 2002). Integrins and other cell surface matrix receptors can provide a continuum from the extracellular matrix to the actin cytoskeleton (and other cytoskeletal systems) that link the cell exterior to the cell interior as a mechanically coherent unit. Integrins are heterodimers formed from at least 18 distinct a-subunits and 8 distinct -units which dimerize in various combinations (Miranti and Brugge, 2002). Integrins are involved in intercellular signaling and exhibit cross talk with other receptors; they also play critical roles in cell survival and regulation of cell proliferation (Miranti and Brugge, 2002). Each integrin exhibits a measure of specificity for binding extracellular matrix ligands. For example, a11, a21, a31, a101, and a111 integrins bind type 1 collagen; some of these receptors play a role in regulation of phagocytosis by fibroblasts (Lee et al., 1996) and in the transmission of mechanical signals from collagen to cells. Fibronectin is a ubiquitous extracellular matrix protein, which is secreted as subunits and assembled extracellularly into a fibrillar network. This assembly process is integrin dependent and is particularly important for assembly of the glycocalyx. Notably, av1 and a51 integrins bind to fibronectin (Mao and Schwarzbauer, 2005). Integrins have a single membrane-spanning domain and short cytoplasmic tails which interact with the cytoskeleton via the actinbinding protein talin. It is not known if all integrins linking the extracellular matrix to the actin cytoskeleton participate in mechanotransduction nor have direct integrin-mechanosensitive channel interactions been identified to date. However, 1 integrins have been colocalized with epithelial Naþ channels and voltage-activated Ca2þ channels in putative mechanoreceptor complexes in mouse chondrocytes (Shakibaei and Mobasheri, 2003), suggesting the possibility that mechanical interactions between these two groups of molecules may be involved in mechanosensitivity. A. Focal Adhesions Studies of extracellular matrix ligand binding in cultured connective tissue cells have shown that integrin clustering acts as a nucleation site, recruiting to the adhesion complexes a large array of signaling molecules such as the

188


focal adhesion kinase, Src kinases, as well as phosphatases such as SHP-2 (MacGillivray et al., 2000; Herrera Abreu et al., 2006). The nascent focal complexes exhibit assembly of actin filaments and the recruitment of a large array of actin-binding proteins such as vinculin, talin, paxillin, and a-actinin (Zamir and Geiger, 2001). As the adhesions mature into focal adhesions and then super-mature focal adhesions (Dugina et al., 2001), more complex aggregates of proteins including a-actinin are assembled into these tightly adherent protein arrays. On the basis of matrix adhesion receptors such as integrins, specialized adhesive plaques, like focal adhesions in vitro and the fibronexus complex in vivo, permit cells to interact with a broad array of extracellular matrix molecules such as laminin, vibronectin, fibronectin, and collagen. When immunostained in vitro for vinculin or paxillin, focal complexes appear as tiny spear tips that are distributed on the ventral surface of cells adhering to the underlying substrate. The use of green fluorescent protein tagging has enabled an improved understanding of the structural and temporal aspects of the diVerent molecular components of focal adhesions (Wehrle-Haller and Imhof, 2002), including microstructural analyses and the development of a more sophisticated classification of focal adhesions. This is relevant to the function of mechanosensitive channels in connective tissues since focal adhesions are enriched with signaling molecules and they provide critical force transfer sites into the cellular interior. Further, focal adhesions appear to be crucial for our understanding of cell-mediated matrix remodeling. Notably, actin stress fibers insert into focal adhesions and these organelles enhance the development of tension that is involved in matrix contraction. In healing wounds and morphogenesis, contractile forces are crucial for remodeling of the nascent extracellular matrix and for the generation of strong propulsive forces which are reduced as the adhesions grown in size (Beningo et al., 2001). The converse seems to be the case for focal adhesions involved in contraction of the matrix. In myofibroblasts, they develop into ‘‘supermature’’ focal adhesions, which are considerably larger and are able to exert two- to fourfold greater forces on the extracellular matrix than earlier stage focal adhesions (Hinz, 2006). B. Selectins Circulating leukocytes marginate to walls of the blood vessels at sites of inflammation. In order to gain access to the inflamed site, leukocytes adhere to specific regions of the endothelial cell wall and then migrate through intercellular contacts, a process known as extravasation. Leukocytes roll along the endothelial wall and can form more permanent tethering types of


189

adhesions to the endothelium which is the first stage in their eventual migration across the blood vessel wall and into connective tissues. The a41 integrin binds both to fibronectin and to vascular cell adhesion molecule-1 expressed on endothelial cell surfaces. The integrin-vascular cell adhesion molecule bond strength is of the order of >50 pN and thus provides a firm attachment for cells in spite of shear forces from flowing blood (Zhang et al., 2004). Other binding ligands for endothelial cell adhesion molecules are transmembrane glycoproteins known as selectins that are also expressed on the cell surface of leukocytes. In order to promote leukocyte adherence, selectins require exposure to a threshold level of shear force. Slip bonds are weak noncovalent bonds that break under low shear stress, but allow cells to adhere briefly, ‘‘survey’’ the local endothelium, and then roll on. Catch bonds are much stronger adhesions that form under high shear force and are responsible for tethering. Depending on the local shear forces experienced by adherent cells, L-selectins can rapidly switch between expression of slip bonds and catch bonds, and thus may function as highly eVective mechanosensors. In addition, they can activate shape change within the cells by modulating the organization of the actin cytoskeleton. This can be mediated directly by the cross-linking actin-binding protein a-actinin, or indirectly through ezrin-radixin-moesin proteins (Ivetic and Ridley, 2004; Yago et al., 2004). This putative role of selectins as mechanosensors is highlighted by studies of adhesion of eosinophils to interleukin 4-stimulated endothelial cells. In this model, adhesion of eosinophils to endothelial cells induced shear-dependent increases of endothelial cell intracellular calcium and increased phosphorylation of extracellular signal-regulated kinase. Further, ligation of either vascular cell adhesion molecule-1 or E-selectin induced a shear-dependent increase in ERK2 phosphorylation in cytokine-stimulated endothelial cells (Cuvelier et al., 2005). Collectively, these data indicate that selectins can regulate the activity of mechanosensitive Ca2þ-permeable channels in endothelial cells in a collaborative manner requiring inputs from multiple molecular sources.

VI. EXPERIMENTAL MODELS OF FORCE APPLICATION TO CONNECTIVE TISSUE CELLS Although a great deal has been learned by examining regulators of mechanical signaling in intact organisms (e.g., stretch activation of transcription factors in Drosophila; Somogyi and Rorth, 2004), improved understanding of the mechanisms of cell biomechanics has largely involved the development of in vitro model systems. A wide variety of experimental

190


force application systems have been designed and several complex surfaces have been created to study geometry sensing (Liu et al., 1999; Rovensky et al., 2001). While cells in vivo may experience several types of deforming forces simultaneously, several laboratories have designed experimental devices that resolve force application into unidirectional mechanical stretch (static or cyclic) or shear stress. Several systems can eVect repetitive as well as static loading and can localize loading to discrete regions of the cell membrane by ‘‘prodding’’ with micropipettes. This technique can be eVected, for example, with micropipettes mounted on micromanipulators and can involve simple touching and analysis of cell responses by, for example, microscopic fluorimeter measurements of fluorescent dyes that report [Ca2þ]i (Xia and Ferrier, 1992). These techniques can be used mainly for studies of single cells and as such are largely restricted to measurement of ion fluxes. Biochemical analyses require typically >5 105 cells per analysis. High-resolution electrophysiological measurements such as patch clamping provided the initial data on mechanosensitive channels in connective tissue cells (Guharay and Sachs, 1984) but as this technique also damages the cell membrane, inferences about cell-matrix interactions are not feasible (Sokabe et al., 1991). The most commonly employed method for studies of mechanical signaling in connective tissue cells is the use of devices that subject cells to substrate elongation (MacKenna et al., 1998). Cells are plated on matrix-coated flexible substrates and are mechanically stretched by application of vacuum or air pressure, or by shaped inserts (Pender and McCulloch, 1991). The strain gradients created in substrate elongation systems are not uniform across the diameter of the dish, so if constancy of force levels is an important issue in experimental design and interpretation, these types of devices limit observations on cells to relatively small zones at a fixed radius from the center of the dish. Further, observing and measuring mechanoresponses in substrate elongation systems restricts the type of measurements that are made. This is particularly critical when the mechanoresponse may rapidly follow the force stimulus, such as an increase of [Ca2þ]i. This limitation can be partly overcome for fluorescence dye measurements of ions if the force application system is physically integrated with the spectrofluorimetric measurement system (Arora et al., 1994). Global mechanical stretching of the cell membrane can also be eVected by inducing regulatory volume increase with hyposmotic buVers (Star et al., 1992; Bibby and McCulloch, 1994) but the influence of the extracellular matrix on stretch-induced mechanical responses are not easy to quantify in these systems since forces are not applied through matrix adhesion sites but rather across the whole cell membrane.


191

On the basis of early measurements of the rheological properties of cytoplasm using internalized magnetite beads, a system of mechanical loading was developed (Glogauer et al., 1995) in which cells that bound collagencoated magnetite beads were placed within a magnetic field (Fig. 3). With this system the magnitude of loading can be controlled by the flux density of the external magnetic field and by the bead-loading density. Notably, coating of the magnetite beads with collagen permits binding of the beads to collagen receptors (mainly the a21 integrin) and consequently, integrindependent mechanotransduction processes can be analyzed (Katsum et al., 2004). Analysis by single-cell ratio fluorimetry of fura 2-loaded cells demonstrated large Ca2þ transients (50–300 nM above baseline) in response to magnetic force applications through collagen beads. Experiments using either the stretch-activated channel blocker gadolinium chloride or EGTA to chelate external Ca2þ ions, or addition of extracellular manganese ions, indicated that there was a Ca2þ influx through putative stretch-activated channels. The probability of a Ca2þ influx in single cells was increased by higher surface bead loading and the degree of cell spreading. Depolymerization of actin filaments by cytochalasin D increased the amplitude of Ca2þ response twofold. The regulation of Ca2þ flux by actin filament content indicated a possible modulatory role for the cytoskeleton in channel sensitivity. With this system, suYciently large membrane distortions are induced to activate stretch-sensitive Ca2þ-permeable channels. Notably, the development and use of rotational bead models also provides excellent opportunities for study of mechanotransduction through matrix adhesion receptors and the cytoskeleton (Wang et al., 1993), although this particular model has not been frequently applied to studies of mechanosensitive channel activation.

FIGURE 3 Simplified diagram of magnetic system for applying force to collagen-coated beads attached via integrins to the dorsal surfaces of cultured cells.

192


As noted above, cells in many connective tissues act in an integrated and collective manner for transducing mechanical forces into biochemical signals. Thus, forces delivered through both extracellular matrices and intercellular junctions can impact mechanotransduction processes (Ko and McCulloch, 2001). Cells in mechanically active, soft connective tissue environments such as the periodontal ligament (Beertsen and Everts, 1980) form extensive, cadherin-mediated intercellular junctions that are important in tissue remodeling and cell diVerentiation. For examination of cadherinmediated force transmission in connective tissue cells, human gingival fibroblasts in suspension were plated on established homotypic monolayer cultures. The cells formed intercellular adherens junctions. Controlled mechanical forces were applied to intercellular junctions by electromagnets acting on cells containing internalized magnetite beads. At early but not later stages of intercellular attachment, force application visibly displaced magnetite bead-loaded cells and induced Ca2þ transients (65 9 nm above baseline). Similar Ca2þ transients were induced by force application to anti-N-cadherin antibody-coated magnetite beads. Ca2þ responses depended on influx of extracellular Ca2þ through mechanosensitive channels because both Ca2þ chelation and gadolinium chloride abolished the response and manganese chloride quenched fura-2 fluorescence after force application. Force application induced accumulation of microinjected rhodamine-actin at intercellular contacts; actin assembly was inhibited by buVering intracellular Ca2þ fluxes. These results indicated that mechanical forces applied to intercellular junctions activate stretch-sensitive Ca2þ-permeable channels, increase actin polymerization and that N-cadherins in fibroblasts are evidently intercellular mechanotransducers (Ko et al., 2001). Elucidation of the nature of mechanosensors at a subcellular level has been advanced by tools developed through nanotechnologies. These tools include the use of innovative substrates with highly precise determination of the placement of size and locales of adherent proteins (Balaban et al., 2001). Notably, recently developed tissue scaVolds have been produced which enable improved understanding of the mechanosensing, mechanotransduction, and response elements of force signaling (Vogel and Sheetz, 2006). Further advances include substrates with precisely determined rheological properties and the application of laser tweezers and atomic force microscopy by which operators can estimate the magnitude of the forces required to disrupt, for example, cell-matrix interactions (Jiang et al., 2003; Zhang et al., 2004). In combination with the development of technologies for isolation of mechanosensors associated with extracellular matrix adhesion receptors (Fig. 4), there is now hope for identifying what molecular components of adhesive cellular systems are involved in regulating


193

FIGURE 4 Method for purifying focal adhesion associated proteins following application of tensile forces to cultured cells. Panel on left shows removal of proteins from beads and their subsequent analysis by tandem mass spectrometry.

and determining the activation of mechanosensitive channels in connective tissues.

VII. EFFECTS OF FORCE ON CELL SURFACE STRUCTURES If applied forces are of suYciently high amplitude, alterations of protein folding and conformational changes at quaternary and tertiary structural levels can occur (Alberts et al., 2002). As a result of force application, the properties of proteins can be considerably altered as new binding sites are uncovered and new folds are created at other sites in the molecule (Vogel and Sheetz, 2006). These changes in structure have important implications for information-rich molecules such as extracellular matrix proteins including collagen that are heavily endowed with cell attachment motifs, degradation initiation sites, and domains that impact the diVerentiation of cells that attach to these proteins. Conceivably, alterations of protein configuration

194


may be very early events at the initiation of mechanotransduction processes. In this context, we consider that mechanotransduction is a fundamental process by which mechanical signals are converted into biochemical outcomes. Nanotools have the capacity to reveal with considerable precision the sites where initial contacts between cells and matrix molecules take place and, possibly, where the mechanosensors are located. Nanotools may also be able to measure to within a few piconewtons, the amplitudes of force required to generate signals within cells (Jiang et al., 2003). At cell-matrix interfaces, forces may be able to expose cryptic peptide sequences, permitting new receptor-ligand binding interactions, or strengthening preexisting interactions. Although not examined in any detail, changes in protein folding could result in altered conformation of mechanosensitive channels, thereby leading to activation or inactivation of channels. Notably, in addition to the exposure of cells to exogenous forces, cell-generated forces also exert eVects on the surrounding matrix that have profound eVects on tissue formation and maintenance and that are regulated by the rigidity of the substrate (Guo et al., 2006).

VIII. FUTURE APPROACHES This chapter explores the progress that has been made on the structure, function, and regulation of mechanosensitive channels. While connective tissues and their extracellular matrices are strongly impacted by mechanical forces as shown by the tight interdependence of mechanical loading and alteration of matrix structure/function, it is evident that our knowledge of mechanosensitive channels in connective tissues is rather meager. Further, the structural relationships between mechanosensitive channels and matrix adhesion receptors are only starting to be explored, in spite of the large amount of data showing functional connections. The advent of proteomic approaches that examine protein–protein interactions and particularly in those restricted to subcellular fractions suggests novel approaches to characterize channel-adhesion receptor and channel-matrix protein interactions. One possible approach for defining these interactions is shown in Fig. 4 in which focal adhesion proteins are isolated from cells with collagen-coated beads. The bead-associated proteins, with presumptive interacting channel proteins, can then be examined by tandem mass spectrometry and confirmations of potential interactions between adhesion receptors and channel proteins then examined by coimmunoprecipitations. The connective tissueextracellular matrix-mechanoreceptor field is a subset of the global mechanotransduction arena which holds considerable promise for future studies. What is now needed are structural and functional definitions of how forces


195

directed along extracellular matrix proteins interact with and regulate mechanosensitive channels. References Adams, M. A., McNally, D. S., and Dolan, P. (1996). ‘Stress’ distributions inside intervertebral discs. The eVects of age and degeneration. J. Bone Joint Surg. Br. 78, 965–972. Alberts, B., Johnson, A., Lewis, J., RaV, M., Roberts, K., and Walter, P. (2002). ‘‘Molecular Biology of the Cell,’’ 4th edn. Garland Publishing, New York. Arora, P. D., Bibby, K. J., and McCulloch, C. A. G. (1994). Slow oscillations of free intracellular calcium ion concentration in human fibroblasts responding to mechanical stretch. J. Cell. Physiol. 161, 187–200. Balaban, N. Q., Schwarz, A. Z., Riveline, D., Goichberg, P., Tzur, G., Sabanay, I., Mahalu, D., Safran, S., Bershadsky, A., Addadi, L., and Geiger, B. (2001). Force and focal adhesion assembly: A close relationship studied using elastic micropatterned substrates. Nat. Cell Biol. 3, 466–472. Beertsen, W., and Everts, V. (1980). Junctions between fibroblasts in mouse periodontal ligament. J. Periodontal Res. 15, 655–668. Beningo, K. A., and Wang, Y. L. (2002). Fc-receptor-mediated phagocytosis is regulated by mechanical properties of the target. J. Cell Sci. 115(Pt. 4), 849–856. Beningo, K. A., Demobo, M., Kaverina, I., Small, J. V., and Wang, Y. (2001). Nascent focal adhesions are responsible for the generation of strong propulsive forces in migrating fibroblasts. J. Cell Biol. 153, 881–887. Bibby, K. J., and McCulloch, C. A. (1994). Regulation of cell volume and [Ca2]i in attached human fibroblasts responding to anisosmotic buVers. Am. J. Physiol. 266, C1639–C1649. Boitano, S., Sanderson, M. J., and Dirksen, E. R. (1994). A role for Ca(2þ)-conducting ion channels in mechanically-induced signal transduction of airway epithelial cells. J. Cell Sci. 107(Pt. 11), 3037–3044. Bumann, A., Carvalho, R. S., Schwarzer, C. L., and Yen, E. H. (1997). Collagen synthesis from human PDL cells following orthodontic tooth movement. Eur. J. Orthod. 19, 29–37. Chen, C. T., Bhargava, M., Lin, P. M., and Torzilli, P. A. (2003). Time, stress, and location dependent chondrocyte death and collagen damage in cyclically loaded articular cartilage. J. Orthop. Res. 21, 888–898. Clements, K. M., Bee, Z. C., Crossingham, G. V., Adams, M. A., and Sharif, M. (2001). How severe must repetitive loading be to kill chondrocytes in articular cartilage? Osteoarthr. Cartil. 9, 499–507. Clements, K. M., Burton-Wurster, N., and Lust, G. (2004). The spread of cell death from impact damaged cartilage: Lack of evidence for the role of nitric oxide and caspases. Osteoarthr. Cartil. 12, 577–585. Chiquet, M. (1999). Regulation of extracellular matrix gene expression by mechanical stress. Matrix Biol. 18, 417–426. Coughlin, M. F., Puig-de-Morales, M., Bursac, P., Mellema, M., Millet, E., and Fredberg, J. J. (2006). Filamin-A and rheological properties of cultured melanoma cells. Biophys J. 90, 2199–2205. Cuvelier, S. L., Paul, S., Shariat, N., Colarusso, P., and Patel, K. D. (2005). Eosinophil adhesion under flow conditions activates mechanosensitive signaling pathways in human endothelial cells. J. Exp. Med. 202, 865–876. D’Addario, M., Arora, P. D., Fan, J., Ganss, B., Ellen, R. P., and McCulloch, C. A. (2001). Cytoprotection against mechanical forces delivered through 1 integrins requires induction of filamin A. J. Biol. Chem. 276, 31969–31977.

196


Davidson, R. M., Tatakis, D. W., and Auerbach, A. L. (1990). Multiple forms of mechanosensitive ion channels in osteoblast-like cells. Pflugers Arch. 416, 646–651. Dugina, V., Fontao, L., Chaponnier, C., Vasiliev, J., and Gabbiani, G. (2001). Focal adhesion features during myofibroblastic diVerentiation are controlled by intracellular and extracellular factors. J. Cell Sci. 114(Pt. 18), 3285–3296. El-Sayegh, T. Y., Arora, P. D., Fan, L., Laschinger, C. A., Greer, P. A., McCulloch, C. A., and Kapus, A. (2005). Phosphorylation of N-cadherin-associated cortactin by Fer kinase regulates N-cadherin mobility and intercellular adhesion strength. Mol. Biol. Cell 16, 5514–5527. Florian, J. A., Kosky, J. R., Ainslie, K., Pang, Z., Dull, R. O., and Tarbell, J. M. (2003). Heparan sulfate proteoglycan is a mechanosensor on endothelial cells. Circ. Res. 93, 136–142. Galbraith, C. G., Yamada, K. M., and Sheetz, M. P. (2002). The relationship between force and focal complex development. J. Cell Biol. 159, 695–705. Glogauer, M., Ferrier, J., and McCulloch, C. A. (1995). Magnetic fields applied to collagencoated ferric oxide beads induce stretch-activated Ca2þ flux in fibroblasts. Am. J. Physiol. 269(5 Pt. 1), C1093–C1104. Glogauer, M., Arora, P., Yao, G., Sokholov, I., Ferrier, J., and McCulloch, C. A. (1997). Calcium ions and tyrosine phosphorylation interact coordinately with actin to regulate cytoprotective responses to stretching. J. Cell Sci. 110(Pt. 1), 11–21. Glogauer, M., Arora, P., Chou, D., Janmey, P. A., Downey, G. P., and McCulloch, C. A. (1998). The role of actin-binding protein 280 in integrin-dependent mechanoprotection. J. Biol. Chem. 273, 1689–1698. Goga, Y., Chiba, M., Shimizu, Y., and Mitani, H. (2006). Compressive force induces osteoblast apoptosis via caspase-8. J. Dent. Res. 85, 240–244. Goodship, A. E., Lanyon, L. E., and McFie, H. (1979). Functional adaptation of bone to increased stress. An experimental study. J. Bone Joint Surg. Am. 61, 539–546. Gu, C. X., Juranka, P. F., and Morris, C. E. (2001). Stretch-activation and stretch-inactivation of Shaker-IR, a voltage-gated Kþ channel. Biophys. J. 80, 2678–2693. Guharay, F., and Sachs, F. (1984). Stretch-activated single ion channel currents in tissuecultured embryonic chick skeletal muscle. J. Physiol. 352, 685–701. Guo, W. H., Frey, M. T., Burnham, N. A., and Wang, Y. L. (2006). Substrate rigidity regulates the formation and maintenance of tissues. Biophys. J. 90, 2213–2220. Han, Y., Cowin, S. C., SchaZer, M. B., and Weinbaum, S. (2004). Mechanotransduction and strain amplification in osteocyte cell processes. Proc. Natl. Acad. Sci. USA 101, 16689–16694. Herrera Abreu, M. T., Wang, Q., Vachon, E., Suzuki, T., Chow, C. W., Wang, Y., Hong, O., Villar, J., McCulloch, C. A., and Downey, G. P. (2006). Tyrosine phosphatase SHP-2 regulates IL-1 signalling in fibroblasts through focal adhesions. J. Cell. Physiol. 207, 132–143. Hinz, B. (2006). Masters and servants of the force: The role of matrix adhesions in myofibroblast force perception and transmission. Eur. J. Cell Biol. 85, 175–181. Hughes, S., Magnay, J., Foreman, M., Publicover, S. J., Dobson, J. P., and El Haj, A. J. (2006). Expression of the mechanosensitive 2PK þ channel TREK-1 in human osteoblasts. J. Cell. Physiol. 206, 738–748. Ivetic, A, and Ridley, A. J. (2004). The telling tail of L-selectin. Biochem. Soc. Trans. 32, 1118–1121. Jiang, G., Giannone, G., Critchley, D. R., Fukomoto, E., and Sheetz, M. P. (2003). Twopiconewton slip bond between fibronectin and the cytoskeleton depends on talin. Nature 424, 334–337.


197

Kaab, M. J., Ito, K., Rahn, B., Clark, J. M., and Notzli, H. P. (2000). EVect of mechanical load on articular cartilage collagen structure: A scanning electron-microscopic study. Cells Tissues Organs 167, 106–120. Kainulainen, T., Pender, A., D’Addario, M., Feng, Y., Lekic, P., and McCulloch, C. A. (2002). Cell death and mechanoprotection by filamin A in connective tissues after challenge by applied tensile forces. J. Biol. Chem. 277, 21998–22009. Katsum, A., Orr, A. W., Tzima, E., and Schwartz, M. A. (2004). Integrins in mechanotransduction. J. Biol. Chem. 279, 12001–12004. Kerin, A. J., Wisnom, M. R., and Adams, M. A. (1998). The compressive strength of articular cartilage. Proc. Inst. Mech. Eng. [H] 212, 273–280. Ko, K. S., and McCulloch, C. A. (2001). Intercellular mechanotransduction: Cellular circuits that coordinate tissue responses to mechanical loading. Biochem. Biophys. Res. Commun. 285, 1077–1083. Ko, K. S., Arora, P. D., and McCulloch, C. A. (2001). Cadherins mediate intercellular mechanical signalling in fibroblasts by activation of stretch-sensitive calcium-permeable channels. J. Biol. Chem. 276, 35967–35977. Lanyon, L. E. (1984). Functional strain as a determinant for bone remodeling. Calcif. Tissue Int. 36(Suppl. 1), S56–S61. Lee, J., Ishihara, A., Oxford, G., Johnson, B., and Jacobson, K. (1999). Regulation of cell movement is mediated by stretch-activated calcium channels. Nature 400, 382–386. Lee, W., Sodek, J., and McCulloch, C. A. (1996). Role of integrins in regulation of collagen phagocytosis by human fibroblasts. J. Cell. Physiol. 168, 695–704. Li, L. P., Herzog, W., Korhonen, R. K., and Jurvelin, J. S. (2005). The role of viscoelasticity of collagen fibers in articular cartilage: Axial tension versus compression. Med. Eng. Phys. 27, 51–57. Liu, J., Schrank, B., and Waterston, R. H. (1996). Interaction between a putative mechanosensory membrane channel and a collagen. Science 273, 361–364. Liu, M., Tanswell, K., and Post, M. (1999). Mechanical force-induced signal transduction in lung cells. Am. J. Physiol. 277, L667–L683. MacGillivray, M. K., Cruz, T. F., and McCulloch, C. A. (2000). The recruitment of the interleukin-1 (IL-1) receptor associated kinase (IRAK) into focal adhesion complexes is required for IL-1-induced ERK activation. J. Biol. Chem. 275, 23509–23515. MacKenna, D. A., Dolfi, F., Vuori, K., and Ruoslahti, E. (1998). Extracellular signalregulated kinase and c-Jun NH2-terminal kinase activation by mechanical stretch is integrin-dependent and matrix-specific in rat cardiac fibroblasts. J. Clin. Invest. 101, 301–310. Mao, Y., and Schwarzbauer, J. (2005). Fibronectin fibrillogenesis, a cell-mediated matrix assembly process. Matrix Biol. 24, 389–399. Martinac, B. (2004). Mechanosensitive ion channels: Molecules of mechanotransduction. J. Cell Sci. 117(Pt. 12), 2449–2460. McNeil, P. L., and Kirchhausen, T. (2005). An emergency response team for membrane repair. Nat. Rev. Mol. Cell. Biol. 6, 499–505. Miranti, C. K., and Brugge, J. S. (2002). Sensing the environment: A historical perspective on integrin signal transduction. Nat. Cell Biol. 4, E83–E90. Nordin, M., and Frankel, V. H. (1980). ‘‘Basic Biomechanics of the Musculoskeletal System,’’ 2nd edn. Williams and Wilkins, Philadelphia, USA. Oloyede, A., and Broom, N. (1996). The biomechanics of cartilage load-carriage. Connect. Tissue Res. 34, 119–143. Orr, A. W., Helmke, B. P., Blackman, B. R., and Schwartz, M. A. (2006). Mechanisms of mechanotransduction. Dev. Cell 10, 11–20.

198


Ozaki, S., Kaneko, S., Podyma-Inoue, K. A., Yanagishita, M., and Soma, K. (2005). Modulation of extracellular matrix synthesis and alkaline phosphatase activity of periodontal ligament cells by mechanical stress. J. Periodont. Res. 40, 110–117. Pelham, R. J., Jr., and Wang, Y. (1997). Cell locomotion and focal adhesions are regulated by substrate flexibility. Proc. Natl. Acad. Sci. USA 94, 13661–13665. Pender, N., and McCulloch, C. A. (1991). Quantitation of actin polymerization in two human fibroblast sub-types responding to mechanical stretching. J. Cell Sci. 100(Pt. 1), 187–193. Provenzano, P. P., and Vanderby, R., Jr. (2006). Collagen fibril morphology and organization: Implications for force transmission in ligament and tendon. Matrix Biol. 25, 71–84. Rezzonico, R., Cayatte, C., Bourget-Ponzio, I., Romey, G., Belhacene, N., Loubat, A., Rocchi, S., Van Obberghen, E., Girault, J. A., Rossi, B., and Schmid-Antomarchi, H. (2003). Focal adhesion kinase pp125FAK interacts with the large conductance calcium-activated hSlo potassium channel in human osteoblasts: Potential role in mechanotransduction. J. Bone Miner. Res. 18, 1863–1871. Rovensky, Y. A., Domnina, L. V., Ivanova, O. Y., and Vasiliev, J. M. (2001). Responses of epithelial and fibroblast-like cells to discontinuous configuration of the culture substrate. Membr. Cell Biol. 14, 617–627. Ryder, K. D., and Duncan, R. L. (2001). Parathyroid hormone enhances fluid shear-induced [Ca2þ]i signaling in osteoblastic cells through activation of mechanosensitive and voltagesensitive Ca2þ channels. J. Bone Miner. Res. 16, 240–248. Sachs, F. (1988). Mechanical transduction in biological systems. CRC Crit. Rev. Biomed. Eng. 16, 141–169. Sengupta, A., Valdramidou, D., Huntley, S., Hicks, S. J., Carrington, S. D., and Corfield, A. P. (2000). Distribution of MUC1 in the normal human oral cavity is localized to the ducts of minor salivary glands. Arch. Oral Biol. 46, 529–538. Shakibaei, M., and Mobasheri, A. (2003). Beta1-integrins co-localize with Na, K-ATPase, epithelial sodium channels (ENaC) and voltage activated calcium channels (VACC) in mechanoreceptor complexes of mouse limb-bud chondrocytes. Histol. Histopathol. 18, 343–351. Sokabe, M., Sachs, F., and Jing, Z. Q. (1991). Quantitative video microscopy of patch clamped membranes stress, strain, capacitance, and stretch channel activation. Biophys. J. 59, 722–728. Somogyi, K., and Rorth, P. (2004). Evidence for tension-based regulation of Drosophila MAL and SRF during invasive cell migration. Dev. Cell 7, 85–93. Star, R. A., Zhang, B. X., Loessberg, P. A., and Maullem, S. (1992). Regulatory volume decrease in the presence of HSCO 3 by single osteosarcoma cells UMR‐106‐01. J. Biol. Chem. 267, 17665–17679. Tarbell, J. M., and Pahakis, M. Y. (2006). Mechanotransduction and the glycocalyx. J. Int. Med. 259, 339–350. Tzima, E, Irani-Tehrani, M., Kiosses, W. B., Dejana, E., Schultz, D. A., Engelhardt, B., Cao, G., DeLisser, H., and Schwartz, M. A. (2005). A mechanosensory complex that mediates the endothelial cell response to fluid shear stress. Nature 437, 426–431. Wang, N., Butler, J. P., and Ingber, D. E. (1993). Mechanotransduction across the cell surface and through the cytoskeleton. Science 260, 1124–1127. Watson, P. A. (1991). Function follows form: Generation of intracellular signals by cell deformation. FASEB J. 5, 2013–2019. Wehrle-Haller, B., and Imhof, B. A. (2002). The inner lives of focal adhesions. Trends Cell Biol. 12, 382–389.


199

Vogel, V., and Sheetz, M. P. (2006). Local force and geometry sensing regulate cell functions. Nat. Rev. Mol. Cell. Biol. 7, 265–275. Xia, S. L., and Ferrier, J. (1992). Propagation of a calcium pulse between osteoblastic cells. Biochem. Biophys. Res. Commun. 186, 1212–1219. Yago, T., Wu, J., Wey, D., Klopochki, A. G., Cheng, Z., and McEver, R. P. (2004). Catch bonds govern adhesion through L-selectin at threshold shear. J. Cell Biol. 166, 913–923. Ypey, D. L., Weidema, A. F., Hold, K. M., Van der Laarse, A., Ravesloot, J. H., Van Der Plas, A., and Nijweide, P. J. (1992). Voltage, calcium, and stretch-activated ionic channels and intracellular calcium in bone cells. J. Bone Miner. Res. 7(Suppl. 2), S377–S387. Zamir, E., and Geiger, B. (2001). Components of cell-matrix adhesions. J. Cell Sci. 114, 3577–3579. Zhang, X., Craig, S. E., Kirby, H., Humphries, M. J., and Moy, V. T. (2004). Molecular basis for the dynamic strength of the integrin 41/VCAM-1 interaction. Biophys. J. 87, 3470–3478.

CHAPTER 8 MscL: The Bacterial Mechanosensitive Channel of Large Conductance Paul Blount, Irene Iscla, Paul C. Moe, and Yuezhou Li Department of Physiology, University of Texas-Southwestern Medical Center, Texas 75390

I. Overview II. Introduction and Historical Perspective A. The Discovery of MS Channels in Bacteria B. Proposing a Function C. The Identification of Multiple MS Channel Activities in E. coli D. Identification of the E. coli mscL Gene E. Early Mutagenesis Studies III. A Detailed Structural Model: An X-Ray Crystallographic Structure from an E. coli MscL Orthologue A. The Crystal Structure B. Fitting the Structure with the Findings from Mutagenesis Studies C. Comparing Tb-MscL with Eco-MscL IV. Proposed Models for How the MscL Channel Opens A. Opening the Channel: Twist and Turn B. Molecular Dynamic Simulations V. Physical Cues for MscL Channel Gating: Protein–Lipid Interactions A. Studies of the Energetic and Spatial Parameters for MscL Gating B. Does MscL Sense the Pressure Across the Membrane or the Tension Within It? C. Sensing the Biophysical Properties of the Membrane D. Specific Protein–Lipid Interactions VI. MscL as a Possible Nanosensor VII. Conclusions References


1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58008-1

202

Blount et al.

I. OVERVIEW MscL is mechanosensitive (MS) channel of large conductance found in the bacterial cytoplasmic membrane. It, with other MS channels, serves as a biological ‘‘emergency release valve’’ that protects the cell from lysis resulting from acute downward shifts in the osmotic environment. To date, more is known of MscL and its mechanism of action than perhaps any other MS channel. Many genetic and other functional studies have been performed on the Escherichia coli channel and a crystal structure exists for a closed or ‘‘nearly closed’’ structure of one of its orthologues from Mycobacterium tuberculosis. Models for the mechanisms of channel gating and the open structure have been generated and tested by several diverse approaches. Finally, several studies have begun to determine the precise stimuli that are sensed by this channel. Together, the data and projected models are giving a glimpse to the molecular mechanisms underlying an MS channel activity.

II. INTRODUCTION AND HISTORICAL PERSPECTIVE Bacterial MS channels are not only a means to study aspects of bacterial physiology but are also currently the most advanced model system for studying MS channel function. They have thus emerged as a paradigm for studying how a protein can sense and respond to changes in its lipid environment. Among the bacterial MS channels, MscL has been the most tractable and is currently the best studied. Identification of the gene that encodes the MscL activity gave us a first glimpse and chance for the genetic study of a channel that senses and responds to mechanical force. A. The Discovery of MS Channels in Bacteria On genesis of the patch-clamp technique (Hamill et al., 1981), it was just a matter of time before membranes that were not associated with electrical or other conductances, such as those belonging to cellular internal structures and organelles, were studied. Perhaps one of the more creative candidates was the cytoplasmic membrane of microbes, including bacteria. Ching Kung’s group, the pioneer in the study of the physiology of microbial channels, was the first to patch the bacterial envelope of E. coli (Martinac et al., 1987). To provide enough membrane area to be patched, the authors used cephalexin, a drug that inhibits septation; hence, the cells grew but could not divide. The resulting elongated cells were then collapsed by treatment with lysozyme and EDTA to form giant cells, on the order of a few microns in diameter,

8. MscL: A Bacterial Mechanosensitive Channel

203

which were then amenable to the patch-clamp technique. On examination, what were discovered were not voltage-gated channels, but rather those that responded to mechanical force, specifically suction or pressure within the electrode. These channels are most frequently referred to as MS channels, and as discussed in Section V.B have been shown to be gated by membrane tension. Remembering that E. coli is a Gram-negative bacterium, it was important to know which membrane, inner or outer, the channels resided in; subsequent studies demonstrated that the activities are not porins or other outer membrane channels but instead are restricted to the inner membrane (Berrier et al., 1989). B. Proposing a Function A curious feature of these channels was that they had a relatively large conductance, on the order of 1 nanoSiemen (nS), whereas normal eukaryotic channels are at best measured in a few tens of picoSiemens. This observation begged the question of what function such a large MS channel activity could have in such a small cell. Previous studies had demonstrated that bacterial cells, when exposed to a high osmotic environment, accumulated compatible solutes such as Kþ, proline, and glycine betaine to very high levels, presumably as osmoprotectants that would allow the cells to maintain turgor and continue to divide (Britten and McClure, 1962). When a culture so treated was subsequently subjected to an acute decrease in the osmotic environment, an osmotic downshock, the cells jettisoned the accumulated solutes to the medium without a substantial decrease in viability (Britten and McClure, 1962; Schleyer et al., 1993). It seemed logical, therefore, to propose that this large-conductance channel is the conduit through which accumulated compatible solutes are expelled on osmotic downshock. However, as discussed in Section II.C, the demonstration that bacterial MS channels played a role as a ‘‘biological emergency-release-valve’’ required the cloning and characterization of multiple bacterial MS channels. C. The Identification of Multiple MS Channel Activities in E. coli The initial studies of bacterial MS channels (Martinac et al., 1987, 1990) examined what was thought to be a single activity that was 1 nS in conductance. In a subsequent study, it was demonstrated that when a larger stimulus was given, a larger conductance channel activity was observed; the two activities could be separated by biochemical fractionation of solubilized membrane proteins over a gel filtration column and reconstitution of the

204

Blount et al.

subsequent fractions into azolectin lipids, which were subsequently subjected to patch clamp (Sukharev et al., 1993), thus demonstrating that the MscS and MscL channel activities were not only produced by separate proteinatious entities but that they survived solubilization, biochemical enrichment, and reconstitution. An additional study showed that with smaller stimuli a smaller conductance was also observed (Berrier et al., 1996). The gradation of channel activities with increasing conductance correlating with an increase in stimuli to gate them thus led to the hypothesis that as osmotic downshock increased, more and larger cytoplasmic components would be jettisoned to the medium to prevent cell lysis (Berrier et al., 1996). From these studies, a new nomenclature emerged, with the largest conducting channel (3.6 nS) named MscL for MS channel of large conductance, MscS for the smaller conducting channel (1 nS), and MscM for the even smaller or mini conducting channel (0.3 nS). Although the molecular identity of MscM has remained elusive, the other channels have been cloned and sequenced (see Section II.D for a discussion of the cloning of the MscL channel). The final added twist of the story came when it was realized that what had been deemed ‘‘MscS activity’’ was probably composed of two similar channels, now referred to as MscS and MscK; ‘‘K’’ for Kþ regulated (Levina et al., 1999; Li et al., 2002). From what we now know of the abundance and conductance of the diVerent channel activities, it appears that the early studies characterizing MS channels in E. coli were characterizing primarily the MscS, and perhaps a small amount of the MscK activity. Please note that the E. coli MscS channel is reviewed in Chapter 9 and MscS/MscK related putative channels found within plants reviewed in Chapter 13 of this volume. As discussed in Section II.B, the role of these channels was suspected to be as biological emergency-release-valves that allow the cell to rapidly adapt to an acute osmotic downshock. However, the discovery of multiple channel activities left open the possibility that these channels would be redundant in function. Indeed, it required the cloning of both the mscS and the mscL genes, and the generation of a double null to definitively demonstrate that these channels played such a role (Levina et al., 1999). However, making a triple-null strain that, in addition, is null for mscK does not amplify the ‘‘osmotic-lysis’’ phenotype observed for double-null mutant. MscK is one of a handful of MscS homologues predicted to be expressed in E. coli but is the sole homologue to be observed in patch clamp. MscK is only observed under specific ionic environments, thus leading to the hypothesis that some MscS homologues may only function under as yet unidentified environmental conditions (Li et al., 2002). Of all of the bacterial channels, the gene encoding MscL was the first to be identified. Although initially the isolation of this single gene could not confirm the physiological function of the bacterial channels because of the redundancy of function with the, at the


205

time, unidentified MscS, it did reflect an important step in the field of bacterial mechanosensors thus allowing for the study of the molecular basis of mechanosensation in a well-defined system. D. Identification of the E. coli mscL Gene The identification of the E. coli mscL gene was performed by a laborious eVort to first biochemically enrich the protein responsible for the MscL activity (Sukharev et al., 1994). To our knowledge, this approach had not been used previously, or since, for the molecular identification of a channel. Briefly, membrane fractions were solubilized, fractionated using columns that separated the proteins according to biochemical properties, then a portion of each fraction reconstituted into azolectin lipids and assayed by patch clamp for channel activity. Fractions containing a significant number of channels were pooled, assayed for protein content by SDS-PAGE, and then chromatographically fractionated over a second, independent column. This was performed until SDS-PAGE showed only one primary protein band remained, thus suggesting that the channel was generated from a single gene product. This 17-kDa band was sequenced, and 37 amino acids at the N-terminal identified. At the time, the E. coli genomic sequence was just under way and the sequence did not match any of the genome-project sequences. But, the authors were lucky because researchers studying another gene, trkA, had sequenced slightly further than necessary, thus generating a sequence that predicted the first 38 residues of the next gene (Hamann et al., 1987). This was a perfect match with the putative MscL microsequence. Thus, this information identified the region of the genome that held the proposed mscL gene and allowed for its cloning and sequencing (Sukharev et al., 1994). Surprisingly, the gene was very small, encoding a protein of 136 amino acids in length and only two -helical transmembrane domains. The discrepancy between the size of the protein on SDS-PAGE and in a nondenaturing gel filtration column strongly suggested a homomultimer. The authors of this first report of the identification of mscL went to great lengths to demonstrate that they had indeed cloned the gene responsible for the MscL activity (Sukharev et al., 1994). An mscL-null mutant was generated by insertional disruption; to demonstrate that the channel activity was missing from this null mutant, 50 spheroplast membrane patches that survived a high level of stimulus were shown not to contain activity. Expression of mscL in trans in the null mutant reconstituted the activity. The SDSPAGE protein band that initially correlated with MscL activity was missing from the mscL-null mutant on an identical enrichment scheme used to identify it, and no activity could be reconstituted from the final fraction

206

Blount et al.

when assayed by patch clamp. Finally, when the gene was translated using a cell-free system and the product reconstituted and assayed by patch clamp, MscL channel activity was observed. At the time, little was known of the molecular basis of mechanosensation. Perhaps the best-studied mechanosensory system at the time was in Caenorhabditis elegans where several genes were found to correlate with touch sensation and the few that predicted transmembrane proteins were candidates for channel subunits (Tavernarakis and Driscoll, 1997); however, there was no gene, or combination of genes, shown to encode MS channel activity. Even today, the precise functional role of many of the candidates for eukaryotic MS channels is still being debated. Hence, mscL underwent a rare set of rigorous tests to demonstrate that it truly encoded the MscL activity; as a result, it was the first gene shown to encode any MS channel activity. E. Early Mutagenesis Studies As discussed in Section II.B, the proposed function of bacterial MS channels was as biological emergency-release-valves that allowed the cell to rapidly adapt to an acute osmotic downshock. However, after the cloning of mscL, it soon became clear that the MscL-null mutant did not have an obvious phenotype; it did not lyse or show any distress on osmotic downshock. Hence, the early mutagenic studies of MscL served two purposes: first to determine the functional significance of domains and residues within the protein, and second, to strengthen the correlation between osmotic and ‘‘compatible solute-flux’’ phenotypes and the functional properties of the channel activities. The earliest mutagenic studies were largely directed at determining functional regions of the protein. Deletion analysis demonstrated that much of the C-terminal region was not necessary for normal channel function, while deletion or alteration of the N-terminal led to a disruption of channel function (Blount et al., 1996a,c; Ha¨se et al., 1997a). In addition, site-directed mutations within the channel often led to changes in channel sensitivity to pressure or open dwell times (Blount et al., 1996c). While site-directed mutagenesis can, and did, give a gross resolution of some regions of the protein that did or did not play a role in MS channel activity, it was not overlooked that one of the real advantages of working with a bacterial system is the ability to randomly mutate a gene of interest and select or screen for rare phenotype-eVecting mutations. However, as stated above, the mscL-null mutant did not show an obvious loss-of-function (LOF) or null phenotype, thus limiting the possibility of isolating mutated genes that led to total or partial LOF phenotype. Ching Kung’s group therefore isolated randomly mutated mscL genes that, when expressed, led


207

to a slowed- or no-growth gain-of-function (GOF) phenotype (Ou et al., 1998). The approach was to place the randomly mutated mscL genes under the transcriptional control of an inducible promoter (lacUV5) and use a replica-plate strategy; colonies that grew when the gene was not induced, but failed to grow well when the channel expression was induced, were selected for further characterization. The hope was that a subset of mutations would lead to channels that mis-gate in vivo; this ‘‘loose cannon’’ would cause the cells to inappropriately lose valuable cytoplasmic components and, therefore, play ‘‘metabolic catch‐up’’ and either slow their growth or even decrease their viability. This study led to two major findings. First, it identified a ‘‘hot-spot’’ within the protein where specific mutations led to extremely severe GOF phenotypes; this region was predicted to be the N-terminal half of the first transmembrane domain (TMD1). Second, further analysis of the functional properties of the mutated channels demonstrated a correlation between the severity of the phenotype, the flux of the osmoprotectant Kþ, and an increase in sensitivity of the channel activity to stimuli. Most of the mutations in this hot-spot were hydrophilic residues, and many were to charged amino acids. Furthermore, in addition to an increase in channel sensitivity, the authors noted that the GOF mutants also showed a decrease in the open dwell time, suggesting that the transition barrier between closed and open states was decreased. However, a more complete appreciation of these findings required a more detailed structural model.

III. A DETAILED STRUCTURAL MODEL: AN X-RAY CRYSTALLOGRAPHIC STRUCTURE FROM AN E. COLI MscL ORTHOLOGUE Subsequent to the identification of mscL, the topology of the MscL protein had been well defined (Blount et al., 1996b) using an approach common for bacterial systems, PhoA fusion (Manoil and Beckwith, 1986). This study supported the hypothesis that the protein had two transmembrane domains, and strongly suggested that the both N- and C-termini were cytoplasmic and the loop between the transmembrane domains periplasmic (Blount et al., 1996b). By contrast, other data were misleading and misinterpreted to suggest that the complex was composed of six identical subunits (Blount et al., 1996b; Saint et al., 1998), rather than the five we now know it contains. Finally, one study using circular dichroism (CD) spectral analysis demonstrated that the protein was highly -helical and relatively resistant to denaturation (Arkin et al., 1998). This was essentially the extent of the knowledge of the structural features of the channel when an X-ray crystallographic structure of an MscL from M. tuberculosis was obtained and revolutionized the field (Chang et al., 1998). Because most experimental data has

Blount et al.

208

been derived from the MscL from E. coli, researchers have been forced to make structural and functional comparisons between these orthologues. A. The Crystal Structure Douglas Rees’s group revolutionized the field of bacterial MS channels when they solved a structure of an MscL by X-ray crystallography to a ˚ (Chang et al., 1998) (Fig. 1). At the time, because of the resolution of 3.5 A numerous microbial genomic sequencing projects, it was becoming obvious that MscL is almost ubiquitous among the prokaryotic kingdom and that large portions of the protein are highly conserved (Moe et al., 1998; Maurer et al., 2000). The authors, therefore, apparently used the strategy of attempting to generate refracting crystals from homologues derived from various organisms. Hence, the solved structure was derived from the organism M. tuberculosis. (For those readers that would like a more detailed discussion of the structural properties of MscL and MscS than that presented within this chapter, below, please refer to Chapter 1 of this volume.)

Side view

Periplasmic view

FIGURE 1 Schematic representation of MscL based on the M. tuberculosis crystal structure, which some evidence now suggests is in a ‘‘nearly closed’’ state. In the view from the side, the approximate position of the membrane is shown (left). The disposition of the TMD1 lining the pore of the channel and the TMD2 surrounding them is better shown in the periplasmic view of the protein (right). A single subunit is highlighted for clarity.


209

The solved structure confirmed some predictions, challenged others, and shed new light on structural features for which no data existed. As had been anticipated from previous studies (Blount et al., 1996b; Arkin et al., 1998), the structure is highly -helical and contains what appear to be two transmembrane domains. However, the structure revealed a radially symmetrical homopentamer in contrast to the hexamer (Blount et al., 1996b; Saint et al., 1998) or even monomer (Ha¨se et al., 1997b) that had been predicted from previous studies. Within the crystallographic study, a cross-linking experiment supported the pentameric structure for both the M. tuberculosis (Tb-MscL) and E. coli MscL (Eco-MscL). A subsequent study that utilized multiple cross-linking reagents as well as equilibrium ultracentrifugation further supported the pentameric design of the MscL channel (Sukharev et al., 1999a), which is now generally accepted. The C-terminal region of the channel appeared to form a cytoplasmic -helical bundle, which the authors noted may be an artifact of the low pH in which the channel was crystallized. Although deletion experiments suggested that the N-terminus was important in forming a functional channel, no clues came from the structure because the first nine residues were not resolved. The second transmembrane domain (TMD2) faced what would normally be the lipid bilayer, while the TMD1 ˚ in diameter; so, formed the pore. The constriction point was just over 4 A given that the conductance of the channel and molecular sieving experiments ˚ , the authors predicted the open channel to form a pore greater than 30 A speculated that the channel was in a closed or ‘‘nearly closed’’ state. There was nothing obvious from the structure to reveal the functional working of an MS channel. Hence, mechanistic models required a combination of the solved structure with data that had been derived from the earlier mutagenesis studies. B. Fitting the Structure with the Findings from Mutagenesis Studies The crystallographic structure of Tb-MscL showed the constriction of the pore of the channel being formed by TMD1, with the periplasmic half of the domain forming a vestibule, while the cytoplasmic half appeared more tightly packed with a true constriction point at a valine at position 21 (V21; the analogous residue is V23 in E. coli) (Chang et al., 1998). The authors noted this ‘‘hydrophobic barrier or gate’’ at the cytoplasmic half of the TMD1, and discussed the consistency with the random mutagenesis study (Ou et al., 1998) that found this region to be a hot-spot, with many mutations in this region leading to GOF phenotypes (Section II.E). A closer look at the mutagenesis data showed that 14 of the 18 GOF phenotype-eVecting mutations isolated were within this subdomain, all but 3 of the 14 were

210

Blount et al.

to residues that were significantly more hydrophilic, 7 of which were to charged residues. In addition, it was noted that there was a correlation between an increase in severity of the GOF phenotype in vivo with both, an increase in sensitivity of the channel to membrane tension and a decrease in channel open dwell times as assayed in patch clamp (Ou et al., 1998). This correlation was subsequently further substantiated by two studies: one in which multiple residue substitutions at a single position, G22 (Yoshimura et al., 1999), and another in which it was shown that adding a charge to the channel posttranslationally by reacting sulfhydryl reagents with a cysteine mutant (Yoshimura et al., 2001). Combining these data with the structure led to what was coined the ‘‘hydrophobic lock’’ hypothesis (Blount and Moe, 1999; Yoshimura et al., 1999; Moe et al., 2000). Briefly, according to this theory, it is the transit of this hydrophobic constriction point of the pore through an aqueous environment, presumably the channel lumen, during a normal transition state that is the primary energy barrier to channel opening. The substitution of a residue in this region to a more hydrophilic amino acid decreases this energy barrier, thus allowing the channel to more easily transit from the closed to the open state, and back again. Thus, the lowering of the transition barrier leads to the increased sensitivity of the channel as well as the observed decrease in open dwell time. C. Comparing Tb-MscL with Eco-MscL Given that the best structural information is for Tb-MscL, but that the bulk of the experimental data has been obtained from the Eco-MscL, it is tempting to shoehorn all of the data obtained for Eco-MscL into the obtained structure. Aspects of the Tb-MscL structure have been confirmed for the Eco-MscL channel by electron paramagnetic resonance (EPR) spectroscopy in combination with site-directed spin labeling (SDSL), albeit to a far less precise resolution (Perozo et al., 2001). However, one must be careful to take into account any functional diVerences between these potential orthologues, and to realize that assumptions that details of the structure are conserved, or that the structure reflects a fully closed state of the channel, may not be correct. 1. Functional Comparisons One report had assayed seven homologues from both Gram-positive and Gram-negative organisms, when expressed in an MscL-null E. coli strain, and found that all of the putative channel genes encoded detectable channel activity (Moe et al., 1998); hence, the channels all appeared to be orthologues. However, functional properties were not always identical. For example,


211

the channel isolated from Synechocystis was found to be significantly less sensitive to membrane tension and that isolated from Staphylococcus aureus was found to have significantly shorter open dwell times when compared to Eco-MscL. The report had two important implications. First, that the function was conserved; a phenomenon that would seem unlikely unless the channel activity that could be measured was a critical part of the protein’s normal in vivo function. Second, that the most conserved parts of the protein among these and other homologues, for example TMD1, must have exceptional functional relevance. Unfortunately, the Tb-MscL was not among these initially characterized MscL channels. Subsequent to the report of the crystallization of the Tb-MscL channel, a study was made of the functionality of this potential orthologue when expressed in an MscL-null E. coli strain (Moe et al., 2000). Although channel activity was observed, the channel was among the least sensitive of all homologues assayed to date; the Synechocystis MscL is the only other channel that is this diYcult to open. Consistent with this finding, the Tb-MscL, when expressed in trans, was unable to suppress the osmotic-lysis phenotype observed for the mscS/mscL double-null E. coli mutant. On the other hand, Eco-MscL mutations that had been shown to eVect a GOF phenotype, when transposed into Tb-MscL, were also eVective at making this channel more sensitive, suggesting a mechanistic correlation between the two channels (Moe et al., 2000). Hence, the inability of the Tb-MscL channel to gate at physiological membrane tensions when expressed within E. coli may simply be due to environmental factors such as native membrane composition; this functional diVerence should be considered when trying to extrapolate data derived from the Eco-MscL to the structure of Tb-MscL. 2. Does the Structure Reflect a Fully Closed State? The authors of the crystallization paper were extremely careful to state that the channel appeared to be in a closed or nearly closed state (Chang et al., 1998). Perhaps because it was suspected that the closed state would be of the lowest energy, many researchers dogmatically believed that the TbMscL structure reflected a fully closed structure. However, relatively recent and independent lines of evidence have suggested that for the Eco-MscL channel G26, rather than V23, is the true constriction point of the fully closed channel. In a study in which the two transmembrane domains were sequentially scanned by cysteine mutagenesis, G26C, not V23C, appeared to eYciently make disulfide bridges; channel activity was only eYciently observed for the G26C mutant in patch clamp when DTT was added to the patch buVer (Levin and Blount, 2004). Furthermore, in an independent study, a number of single histidine mutations were generated in the pore region of Eco-MscL, and the ability of these residues to coordinate metals,

212

Blount et al.

including Ni2þ, Cd2þ, and Zn2þ, was assayed as the ability of these metals to keep the channel closed or inhibit it from gating (Iscla et al., 2004). Again, G26H, not V23H, was the most eYcient at coordinating these metals, suggesting that this is indeed the constriction point. If true, this could simply be species specific. On the other hand, the Tb-MscL structure shows a ˚ pore. It, therefore, seems possible that for channel with a greater than 4-A both channels the analogous and conserved glycine is closer to the constriction point. This would predict that the constriction point would be more periplasmic and the vestibule smaller. It would also require a counterclockwise rotation of TMD1 when viewed from the periplasm; since, as discussed in Section IV.A.2, a clockwise rotation of TMD1 has been predicted to occur on channel opening, such a counterclockwise rotation may be necessary to achieve a fully closed channel. Hence, the Tb-MscL structure may not reflect a truly fully closed but merely a nearly closed state of the channel.

IV. PROPOSED MODELS FOR HOW THE MscL CHANNEL OPENS Given a crystal structure for a (nearly) closed state of a channel, it is tempting for researchers to speculate on what the open structure would look like. In the original crystallization paper, the authors included a crude model in which the TMD1s separated (Chang et al., 1998). To form a pore of ˚ with near-vertical transmembrane helices, all 10 transgreater than 30 A membrane domains must contribute to the channel lumen. More recently derived evidence strongly suggests that this simple model of the pore being generated by transmembrane domains positioned in the membrane like staves of a barrel is incorrect. Although the concrete evidence of a crystal structure for the open structure of the MscL channel is still lacking and much of the details are still in debate, from several independent studies an image of the open structure of MscL channel is beginning to emerge. A. Opening the Channel: Twist and Turn The first detailed model proposed for the open and transition states for MscL opening was both revolutionary and bold (Sukharev et al., 2001a,b). It was revolutionary because it proposed a mechanism by which the lumen of the pore could be generated almost solely by the TMD1. This could be achieved by increasing the angle in which these domains sit within the membrane; the opening would then not be much diVerent from the opening of the iris of a camera. The model was bold because it not only proposed a critical function as a second gate for the N-terminal end of the protein


213

(Sukharev et al., 2001a), which was not resolved in the crystal structure (Chang et al., 1998), but it also predicted the precise locations for each individual residue in the open and several transition states (Sukharev et al., 2001b). The notion of the iris-like opening of the channel and TMD1 lining the pore can now be seen as a great insight that is generally agreed on by researchers in the field, and perhaps therefore often underappreciated. On the other hand, it is perhaps not surprising that many of the other details of the model are either highly controversial or largely discredited by the numerous subsequent studies. 1. Tilting the Transmembrane Domains In proposing the detailed model, the authors chose to work with the EcoMscL channel, presumably because of the wealth of mutagenic information and the ability to test the models on a channel that had a practical functionality. Therefore, before trying to derive the open structure, the authors generated a model for the closed Eco-MscL channel largely by imposing the sequence onto the Tb-MscL structure and making a best-guess for any discrepancies (Sukharev et al., 2001b). From this model, predictions were made on how the Eco-MscL channel opens. One of the major predictions was that the transmembrane domains would tilt within the thinning membrane and then TMD1 would separate to open the channel. This would allow the lumen of the pore to be formed largely by TMD1. The resulting models for the structures for the closed and open states of the channel are shown in Fig. 2. Several studies, using various approaches, have now provided strong support for the idea that the transmembrane domains tilt and TMD1 lines the lumen of the pore. In one study, lysophospholipids were used to gate the Eco‐ MscL channel in vitro and structural aspects of the open channel were determined by SDSL and EPR (Perozo et al., 2002a). The findings were consistent with the tilting and expansion of the transmembrane domains, as predicted (Sukharev et al., 2001b). In another study, Eco-MscL was divided in half so TMD1 and TMD2 were expressed independently (Park et al., 2004). By patch-clamp analysis, the half-containing TMD1 formed spontaneously gating channels, albeit of varying conductance, while TMD2 segments were completely silent as assayed by patch clamp. The coexpression of the two domains formed an MS channel with a similar conductance to the wild-type channel. Hence, the data from this study suggested that TMD1 plays the large part in forming the channel lumen, while TMD2 is more important for assembly and sensing the lipid environment. In yet another study, tryptophan or tyrosine residues were substituted into strategic positions in the transmembrane domains in an attempt to decrease the ability of the domains to tilt (Chiang et al., 2005). The results are consistent with gating of the channel

214

Blount et al.

FIGURE 2 One proposed model for the structure of the E. coli MscL protein (Sukharev et al., 2001b). (A) In the upper part of the panel, the lateral views of the homopentameric complex are shown in its open (right) and closed (left) states. The approximate position of the membrane is shown. Note the flattening of the open structure in comparison with the closed conformation due to the tilting of the transmembrane domains. Although the S3 domain is shown to separate, a more recent proposal suggests that the bundle may remain intact (Anishkin et al., 2003). In the lower panels, the respective periplasmic views are shown. (B) The diVerent domains of the protein are shown in a single subunit.

being linked with TMD tilt; however, specificity of the phenotypes to the substitutions was not demonstrated [e.g., a previous study (Levin and Blount, 2004) demonstrated that a C at position F93 yields a similar decreasedsensitivity phenotype as the F or W substitutions studied]. Finally, one study used an approach coined an in vivo SCAM (Bartlett et al., 2004); SCAM for substituted cysteine accessibility method (Akabas and Karlin, 1999). Here, a characterized cysteine Eco-MscL library in which each residue was independently mutated in both transmembrane elements (Levin and Blount, 2004) was used to determine the accessibility of each residue in the closed and open states of the channel. The authors, taking advantage of the fact that there are no endogenous cysteines in MscL, extended previous findings that the addition of hydrophilic residues in the pore often led to a channel that gave


215

a GOF phenotype (Ou et al., 1998; Yoshimura et al., 1999, 2001), that a cysteine substitution allowed for posttranslational modification of the residue by using sulfhydryl reagents (Akabas and Karlin, 1999), and that this approach had been successfully used for a single substitution previously (Batiza et al., 2002). The positively charged sulfhydryl reagent methanethiosulfonate bromide (MTSET) was used to modify the cysteine mutations in vivo, and the assay was performed in the presence or absence of an osmotic downshock suYcient to gate the channel to assay accessibility in the open and closed states of the channel, respectively. A viability cell count was used to determine if the residue was accessible and modified the channel such that it inappropriately gated and thus eVected a GOF phenotype. Patch-clamp analysis of these mutants in the presence and absence of MTSET has substantiated and extended these results (Bartlett et al., 2006). Although a negative result in this assay may simply mean that the modification does not lead to a misgating channel, it is still of interest to note that all of the 11 residues that gave a positive result in this assay are in the TMD1 (Bartlett et al., 2004). While none of the experiments above are as definitive as a crystal structure of the open channel, together they form a strong support for the original postulate that the transmembrane domains tilt and that TMD1 alone forms the bulk of the lumen of the pore. 2. Rotating TMD1 The original model for the open channel of Eco-MscL simply separated the TMD1 domains and predicted only a very small counterclockwise rotation of this domain of the channel when viewed from the periplasm (Sukharev et al., 2001b). The model predicted several interactions, including unique interactions in the open state between TMD1 and TMD2, that were subsequently tested by trying to trap the channel in specific states by the generation of disulfide bonds in double-cysteine mutants (Betanzos et al., 2002); this approach is often called disulfide trapping. However, it is important to note that the cysteine mutations were highly targeted so only very limited combinations of cysteine substitutions were tested. In addition, when disulfide trapping was reported to stabilize the open state of the channel, as determined by an increase in open dwell time of the fully open state in three patches, membrane tension was still required to open the channel (Betanzos et al., 2002), and the observation that other mutations at independent locations within the channel have been reported to lead to such increases in open dwell times (Blount et al., 1996c) makes the interpretation of the disulfide trapping slightly more diYcult to interpret. Finally, a reversibility of the stabilization of the open state with reducing reagents was not shown. In contrast to the above model, the model derived from SDSL and EPR suggested a 110 clockwise rotation of the TMD1 domain on gating (Perozo

216

Blount et al.

FIGURE 3 The residues predicted by diVerent studies to line the pore region of E. coli MscL in its open state. (A) Helical net representation of residues R13 to I41 from the TMD1 (left). Residues predicted to be exposed to the pore in the open state of the channel by the original (Sukharev et al., 2001b) model are shown in dark gray circles, while those residues derived from EPR studies (Perozo et al., 2002a) are shown in light gray circles. Residue G26, the only one that is included by the both models, is shown in gray. A table listing the diVerent set of residues lining the pore of the channel in the two models is shown (right). (B) Helical net representation of residues D18 to A38 where residues predicted to be exposed to the pore of the channel on activation are shown in light gray circles, as indicated by the results of an ‘‘in vivo’’ SCAM study (Bartlett et al., 2004) (see text). Residues in dark gray circles were predicted to be exposed to the lumen in the closed state.

et al., 2002a). Hence, the discrepancy with the previously proposed model was close to a full 180 . A summary of the residues predicted to line the lumen of the open pore for the two models is presented in Fig. 3; note that G26 is the only common residue. The results of the in vivo SCAM (Bartlett et al., 2004), which identified residues exposed to the aqueous environment when the channel was in the closed and opening states as discussed in Section IV.A.1, appear to resolve which residues are exposed to the channel lumen (Fig. 3). Note that this approach, in contrast to those used previously, has the advantage that movements of the channel are assayed in vivo, when the channel is in its natural environment. Consistent with the hypothesis that G26 is the constriction point of the closed channel (Section III.C.2), G26C appeared to be highly reactive with MTSET in the closed state. The residue clockwise to this position, A27, however, required gating to observe maximal eVects even though it is more periplasmic. Furthermore, the sequential residues G22,


217

V23, and I24 all required gating for maximal eVects, and absolutely no eVect was observed for I24 unless gating occurred. The only way I24 could eYciently be exposed to the lumen of the open pore would be if there was a significant clockwise rotation of the TMD1 domain. Hence, these data are consistent with the EPR results. In another study, mentioned in Section III.C.2, a number of single histidine mutations were generated in the pore region, and the capacity of these residues to coordinate metals, including Ni2þ, Cd2þ, and Zn2þ, was tested by their ability to keep the channel closed or inhibit it from gating (Iscla et al., 2004). Again, the I24 residue was one of the positions assayed. Each of the metals assayed was able to inhibit gating of the I24H mutant. These data not only support the hypothesis of a relatively large clockwise rotation of the TMD1 domain but also suggest that this rotation often occurs quite early in the gating process, even preceding ion permeation. As discussed in Sections II.E, a random mutagenesis study identified a number of GOF-eVecting mutations. In one study, two of these mutated channels, V23A and G26S, were further randomly mutated and suppressors of the GOF phenotype isolated (Li et al., 2004). Partial suppressors were preferentially isolated to avoid nonfunctional channels. All of the suppressors isolated were found to be ‘‘general suppressors’’ that suppressed both GOF-eVecting mutations, with one exception, I92V, which suppressed exclusively the G26S mutation. These data suggest a direct interaction between G26 and I92 on gating. When these positions are imposed on the original open-state model, they fall in proximity but do not face each other (Fig. 4). However, a simple rotation of close to 180 would do the trick. Here again, the data are consistent with a clockwise rotation of the TMD1 domain. In sum, the original model for the open channel structure predicted a slight counterclockwise rotation of the TMD1 domain. Although some disulfide-trapping experiments have supported this model, an EPR study, the in vivo SCAM, metal binding to histidine substitutions, and suppressor mutagenesis all support a model in which there is a significant clockwise rotation of the TMD1 domain. 3. A Closed-Expanded State An early study defining the energetic and spatial parameters of the gating of the Eco-MscL channel suggested an expansion of the system prior to channel gating (Sukharev et al., 1999b). Assuming the expansion does not occur in the lipid membrane itself, this could be accounted for if the channel expanded to two-thirds its fully open size prior to gating. This begged the question: could there be expansion of the pore, but permeation barred by a second gate? One of the boldest proposals in the original gating models for the Eco-MscL channel was that the N-terminal region of the protein, which

218

Blount et al.

FIGURE 4 The MscL G26 and I92 residues may be in proximity. One study using a ‘‘suppressor mutagenesis’’ approach, as described in text, predicted an interaction between G26 and I92 (Li et al., 2004). As seen in a current model for the open structure of the homopentameric E. coli MscL channel (Sukharev et al., 2001b), G26, shown in CPK style in blue on the cyan subunit, and I92, in green on the adjacent orange subunit, are in relatively proximity. Cytoplasmic views (top) and side views (bottom) containing all five (left) or just the two noted adjacent subunits (right) are presented. Note that all is necessary for interaction of these residues is the rotation of TM1 (shown by the arrows in the right panels).

was not resolved in the Tb-MscL crystal structure, served as this second gate (Sukharev et al., 2001a). The authors proposed that this region of the protein formed a helical bundle just cytoplasmic to the TMD1 pore. These proposed -helical domains at the extreme N-terminal end were referred to as ‘‘S1 domains.’’ The model predicts that the TMD1 pore could expand without ion permeation and stabilize in what was called the closed-expanded (CE) state. The model further predicted that it was the separation of the S1 bundles that actually allowed ions and other molecules to permeate. This model was appealing because it could be the sequential separation of the S1 segments that accounted for the substates of the channel that were often observed. In addition, previous studies had demonstrated the importance of this region


219

of the protein for channel function; deletion of the first 11 residues abolished all channel activity (Blount et al., 1996a,c; Ha¨se et al., 1997a). Here again, targeted disulfide-trapping experiments were performed to support the model (Sukharev et al., 2001a). While appealing, the model cannot account for all of the data. First, previous studies had demonstrated that a substate that is four-fifths normal conductance is the one that is most often stabilized (Sukharev et al., 1999b), which seems unlikely with three bundled cytoplasmic S1 helices as proposed. Second, as discussed in Section II.E, substitutions within TMD1 invariably lead to changes in the kinetic properties of the channel as well as increasing its sensitivity; these observations led to the hydrophobic lock hypothesis, a proposal that separation of the TMD1 domains is the primary energy barrier for gating (Blount and Moe, 1999; Yoshimura et al., 1999; Moe et al., 2000) (Section III.B). This change in kinetics strongly suggests that the separation of TMD1, not S1 domains, is coupled to ion permeation. If the separation of S1 is also required, it must be coupled to TMD1 separation, and thus a stable CE state would be impossible. Third, if the S1 helix was truly important for channel gating, one would expect to have isolated numerous GOF- and LOFeVecting mutated Eco-mscL genes within the first 11 amino acids of the protein, where the significant interactions of the S1 helices were predicted to occur. However, in the original random mutagenesis study (Ou et al., 1998), none of the 19 mutations isolated that eVected a GOF phenotype were in this region. In a subsequent random mutagenesis study in which a rapid screening was used to isolate both GOF- and LOF-eVecting mutants (Maurer and Dougherty, 2003), none of the 52 phenotype-eVecting mutations were in this region; instead, the 26 mutations that were isolated in this domain gave no phenotype. Hence, while the proposal of the CE state and S1 domains as a second gate is attractive given the importance of this region and the disulfide-trapping evidence that supports it, several other lines of evidence are more consistent with the proposal that this region serves a more structural role for the channel, such as being critical for correct folding or assembly of the complex. Given this apparent discrepancy, a more rigorous and detailed study of this region and its role in channel formation and/or function is needed. 4. Does the C-Terminal a-Helical Bundle Open? In the original gating model for the Eco-MscL, the C-terminal bundle, now called S3, was predicted to separate on channel opening (Sukharev et al., 2001b) (also see Fig. 2). However, the authors noted the sequence of the S3 of MscL was similar to that of the oligomerization domain of the cartilage protein COMP, which has a fivefold coiled-coil structure stabilized by apolar interactions inside the bundle and by salt bridges on the periphery

220

Blount et al.

(Efimov et al., 1996; Malashkevich et al., 1996). The authors, therefore, tested if the S3 domain would associate in a similar structure, and if this structure is stable during channel gating (Anishkin et al., 2003). Computer modeling suggested that this MscL structure could be quite stable, and disulfide trapping-experiments supported the predicted structure, and seemed to suggest that it remained so on gating. Using disulfide-trapping and biochemical analysis, the authors demonstrated that the predicted interactions between the S3 helices occurred quite eYciently under ambient conditions, strongly suggesting that their assumptions of the helical bundle are correct. Because of technical diYculties using oxidizing reagents in patch clamp, the authors made the assumption that ‘‘ambient’’ oxidative conditions for Eco-MscL channels undergoing biochemical analysis (which includes French press cellular disruption, solubilization, purification and reconstitution) would be the same for channels in spheroplasts subjected to patch-clamp buVer. The results were largely negative or subtle, but all were consistent with a stable S3 bundle during the gating process. The proposed function of a stabilized S3 bundle is as a filtration device (Anishkin et al., 2003). However, an early molecular sieving study suggested ˚ (Cruickshank that the constriction site of the pore must be at least 30–40 A et al., 1997)! In addition, studies comparing mscL-null strains with their parental have suggested that moderately sized proteins including thioredoxin, elongation factor Tu, and DnaK pass through the MscL channel on osmotic downshock (Ajouz et al., 1998; Berrier et al., 2000). At the time the study on the S3 bundle emerged, the ability of MscL to pass such large proteins had been put into question by another study that did not reproduce the results (Vazquez-Laslop et al., 2001); however, it has been demonstrated that the discrepancy between the two studies is simply due to subtle diVerences in experimental approach (Ewis and Lu, 2005); these proteins are truly transported through the MscL channel on osmotic downshock. Hence, future studies will have to address the question: if the S3 bundle truly is stable on gating, how can such large molecules pass through the MscL channel? 5. Role for the Periplasmic Loop Several studies have implied a function for the periplasmic loop located between TMD1 and TMD2. First, early mutagenesis demonstrated that substitutions in this region could influence the kinetics (Blount et al., 1996c) and sensitivity (Ou et al., 1998; Maurer et al., 2000; Tsai et al., 2005) of the channel. An additional study showed that when the channel is subjected to proteases, the channel sensitivity to membrane tension dramatically increases (Ajouz et al., 2000). Consistent with this latter finding, as described in Section IV.A.1, when the Eco-MscL was divided in half so TMD1 and TMD2 were expressed independently (Park et al., 2004), the channels that


221

were formed that had a similar conductance to the wild-type channel but had an increased sensitivity to membrane tension. Together, these data are consistent with the periplasmic loop playing the role of a torsional spring, inhibiting the channel from gating. 6. Asymmetric Movements For simplicity, and given the radial symmetry of the channel, the model for gating the MscL channel assumes a smooth and coordinated movement of all of the subunits simultaneously; radial symmetry is largely maintained in the predicted transition states (with the exception of the S1 domains, which sequentially separate, as discussed in Section IV.A.2) (Sukharev et al., 2001b). However, recent data has challenged the assumption that much of the channel remains radially symmetrical as the channel opens. Few functional studies have been performed with the Tb-MscL channel because of its insensitivity to membrane tension (Section III.C.1), which leads to technical diYculties detecting channel activity prior to rupture of the patch (Moe et al., 2000). However, a GOF-eVecting mutated Tb-MscL channel, V15C, displayed the interesting property of eYcient disulfide bridge formation as assayed by SDS-PAGE. Because this mutated channel contains only a single cysteine, these interactions presumably occurred between subunits within the complex. The resulting activity could easily be studied in patch clamp because the V15C-mutated channel showed a dramatic increase in sensitivity to membrane tension (Shapovalov et al., 2003). The authors found that the channel often times did not go through a normal closure, but instead locked into an open state in response to pressure; the channel then slowly and irreversibly stabilized into partially open ‘‘signature events’’; such events occurred even after the cessation of stimuli. The signature events become smaller with time, and subsequent opening of the channel is impossible. This phenomenon is not observed in the presence of reducing reagents. These data suggest that once a disulfide bridge is formed, the channel, on trying to close, locks into an unstable open structure, and presumably attains unnatural and irreversible conformational changes under the strain. Because the Tb-MscL crystal structure does not predict that the V15 residues can easily interact with one another within the channel complex, the findings suggest that asymmetric movements, in which one V15 approaches its neighbor, often occur on gating. A similar and more recent study has been performed using the Eco-MscL (Iscla et al., 2006). The analogous residue to the Tb-MscL V15 residue in the Eco-MscL channel, V17, did not demonstrate the same properties when mutated to cysteine, presumably because of a species diVerence. The authors, however, had scanned the area with cysteines and found a nearby mutation, N15C, which had interesting properties. Here, disulfide

222

Blount et al.

trapping led to a channel that was more sensitive to membrane tension. The eYciency and rapid kinetics of this reaction suggested that the channel is trapped into a normal transition state that retains the capacity to attain both closed and fully open states. Furthermore, it is reversible by adding reducing agents. Similar to the Tb-MscL V15 residue, the position of the Eco-MscL N15 residue in structural models did not predict that this residue would interact unless asymmetric movements normally occur in the closed to open transition. These data predict that although the MscL channel appears to have radial symmetry, it appears that gating is initiated with the movement of one region of one subunit of the protein. This region is just cytoplasmic to TMD1 and presumably is involved in or initiates the cascade of movements that leads to full channel opening. B. Molecular Dynamic Simulations Molecular dynamic simulation (MDS) is an approach in which detailed atomic interactions and structural mobility are assessed under diVerent conditions by computer modeling. This is a relatively new approach that can help support or direct experimental studies. One of the limitations of this technique is that, because of the computer-processing requirements, only short stretches of time, normally a few tens of nanoseconds at best, can be followed, so often extreme or nonphysiological stimuli must be applied. The approach may vary by the presence or absence of solvent molecules or membrane, as well as the form and extent of stimulation to gate the channel. Several aspects of the structural changes that occur on MscL channel gating have been studied. For example, using either the Tb-MscL (Elmore and Dougherty, 2001; Gullingsrud et al., 2001; Colombo et al., 2003; Valadie et al., 2003) or Eco-MscL (Kong et al., 2002; Gullingsrud and Schulten, 2003; Valadie et al., 2003) protein structural models (Figs. 1 and 2), researchers have obtained data that support the hypothesis that the transmembrane helices tilt within the lipid bilayer on gating and that the lumen of the open pore is generated primarily by TMD1, as discussed in Section IV.A.1. In addition, one MDS study also showed the clockwise rotation of the TMD1 and exposure of the I24 residue (Colombo et al., 2003), as discussed in Section IV.A.2. In another experiment (Elmore and Dougherty, 2001), the C-terminal region of the protein was assayed under diVerent pH conditions and it was found that the bundle was stable only at the low pH in which the Tb-MscL was crystallized; these data would be consistent with an alternative structure for this region, as has been proposed (Anishkin et al., 2003) and is discussed in Section IV.A.4. There are, however, inconsistencies between


223

studies for the presence or absence of the stabilized CE structure and putative S1 helices forming a second gate of the Eco-MscL channel, as discussed in Section IV.A.3. For example, starting from the proposed EcoMscL-closed structure (Fig. 2), one study found a total expansion of the pore could occur without seeing a disruption of the S1 bundle (Gullingsrud and Schulten, 2003), while another group found that the S1 and TM1 do not behave as two independent gates, but the separation of these domains is coupled (Kong et al., 2002). In one experiment, the membrane was curved to stimulate the channel; this study found significant movements within the periplasmic loop structure, which would be consistent for the ‘‘torsional spring’’ function discussed in Section IV.A.5 (Meyer et al., 2006). Many of the studies found asymmetric movements of the complex, as opposed to maintaining radial symmetry, as the channel opens (Bilston and Mylvaganam, 2002; Kong et al., 2002; Colombo et al., 2003), thus supporting the possibility of the asymmetric movements discussed in Section IV.A.

V. PHYSICAL CUES FOR MscL CHANNEL GATING: PROTEIN–LIPID INTERACTIONS Perhaps one of the more intriguing questions concerning mechanosensors is: what exactly are they sensing? Clues have been derived from several studies using a variety of approaches. The data thus far suggest that the MscL channel is able to directly sense biophysical changes in its membrane environment. A. Studies of the Energetic and Spatial Parameters for MscL Gating When the MscL channel is stimulated in a membrane, the amount of stimulus, in pressure, can be measured by a pressure transducer. In addition, with the proper equipment, the patched membrane can be imaged and the radius of curvature also measured. Using these parameters, tension within the membrane can be calculated using Laplace’s law: tension in the membrane equals the pressure across it times the radius of curvature divided by 2. A Boltzmann model for the relationship of the probability of the channel opening vs the membrane tension can then be used to plot and study the energetic and spatial parameters for the gating of the Eco-MscL channel. This analysis was performed (Sukharev et al., 1999b), updated (Chiang et al., 2004), and has been used for some of the GOF-mutated channels (Anishkin et al., 2005). From these analyses, the current-derived values for opening the wild-type channel hold that it takes 7–13 dynes/cm2 of tension in the

224

Blount et al.

membrane to achieve a 50% probability of channel gating, the energy required to gate the channel at this level (E) is 51 13kT, and the change in area (A) is 20 5 nm2. Note that the latter parameter is consistent with current models for the approximate pore size of the open channel. In addition, when the rate constants for achieving specific substates of the channel are plotted against membrane tension, it appears that it is only the closed to first substate conversion that is responsive to tension; once the channel has begun ion permeation, the progression to the diVerent substates occurs independent of external stimulus. B. Does MscL Sense the Pressure Across the Membrane or the Tension Within It? As mentioned in Section V.A, pressure across the membrane and tension within it are related, but not the same. According to Laplace’s law, tension in the membrane equals the pressure across it times the radius of curvature divided by 2. The first study to determine if an MS channel was sensing pressure across the membrane or tension within it was performed on a yeast channel (Gustin et al., 1988). The authors performed the whole-cell patch technique with positive pressure in the patch to observe the channels. Three cells of very diVerent diameter, which would thus have a diVerent radius of curvature, were assayed. When the probability of opening, Po, was plotted against the pressure, three distinct curves were found. However, when the Po was plotted against the calculated tension, the data collapsed to form a single curve. Thus, this yeast channel appears to sense membrane tension. When the spatial and energetic parameters were measured for Eco-MscL (Section V.A), the data were consistent with MscL sensing membrane tension and the authors assumed this was the case, but it was not formally demonstrated (Sukharev et al., 1999b; Chiang et al., 2004). However, using appropriate imaging equipment so that the curvature of the patch can be measured, an experiment similar to that performed in the yeast system was performed, where patches with quite diVerent radius of curvature were compared (Moe and Blount, 2005). These experiments demonstrated that the MscL channel does indeed sense tension within the membrane, not the pressure across it. C. Sensing the Biophysical Properties of the Membrane Soon after the identification of MS channel activities in bacterial membranes, it was realized that changes in the biophysical properties of the membrane are important for channel stimulation. Amphipaths, having both


225

hydrophilic and hydrophobic character, will intercalate into membranes and thus change the lateral pressure profile within the membrane. A combination of a charge on the amphipath and the potential across the membrane can lead to an amphipath preferentially partitioning into one leaflet or the other. An early study determined the eVect of addition of amphipaths on bacterial MS channels (Martinac et al., 1990). This study found that addition of either positively charged, negatively charged, or uncharged amphipaths could increase the sensitivity of the channel to stimuli. After the activation of the channel by an amphipath of one charge, its replacement with an amphipath of the opposite charge would partially reverse the eVects prior to activation, presumably because the amphipaths would partition within opposing sides of the membrane, thus canceling the eVect. While this original study was performed prior to the distinction between the MscS and MscL channel activity, and thus was probably studying primarily MscS, a similar study has been performed for the purified Eco-MscL where channel activity and structure were monitored by patch clamp and EPR, respectively (Perozo et al., 2002b). The findings were consistent, demonstrating that lysophospholipids, which are strongly amphipathic molecules, could gate the Eco-MscL channel. The same study to determine the influence of the possible thinning of the membrane, or hydrophobic mismatch, on channel gating also reconstituted the Eco-MscL channel into lipids with varying chain length. While the sensitivity of the channel increased with decreasing chain length, the channel did not spontaneously gate. Hence, it appears that while hydrophobic mismatch may play some role in channel sensitivity, it appears that the true stimuli for the channel are changes in the biophysical properties of the membrane and perhaps modification of the lateral pressure profile. D. Specific Protein–Lipid Interactions Given that the MscL channel detects tension in the membrane, is modified by hydrophobic mismatch, and appears to detect biophysical changes in the membrane, it is tempting to speculate that there are specific protein–lipid interactions involved in the ability of the channel to sense stimuli. Several studies have suggested this could be the case. For example, it was noted in one of the random mutagenesis studies that LOF-eVecting mutations often occurred near the proposed site of interaction with membrane lipid headgroups, suggesting such specific interactions may occur (Maurer and Dougherty, 2003). In another study looking for intragenic suppressors for a GOF mutant, I41N, the authors noted that some of the suppressing mutations were clustered on the periplasmic side of the transmembrane domains, again near the predicted headgroups (Yoshimura et al., 2004). The authors, therefore,

226

Blount et al.

performed an asparagine scan of the region and found that mutation of many of the residues predicted to face the lipid led to channels that were less functional (Fig. 5). While the study is well done and the interpretation attractive, some predictions have not yet been fulfilled by other studies, as one might have expected. For instance, a similar clustering was not noted in another intragenic suppressors study (Li et al., 2004); but this may simply be because in the latter study partial suppressors were selected, and neither study saturated their screen. In addition, a similar scanning of the region with cysteines did not find the same results (Levin and Blount, 2004). In another set of studies, tryptophan mutagenesis and tryptophan fluorescence spectroscopy revealed a relatively nonspecific association between the protein and uncharged lipids, but a highly specific binding with anionic lipids. Three positively charged residues in a cluster near the cytoplasmic end of TMD2

FIGURE 5 Protein–lipid interactions in the MscL channel. Residues implicated in lipid binding are in a side view of the channel (left) with individual subunits indicated by color and, for clarity, on a single subunit (right). Mutation of residues lining the periplasmic rim of the channel (A) yielded an LOF phenotype, presumably through disruption of lipid binding (Yoshimura et al., 2004). These residues form the periplasmic end of TMD1 (gray) and TMD2 (black). Tryptophan fluorescence spectroscopy revealed heterogeneity in lipid binding to the channel (Powl et al., 2005). No binding preference was detected for uncharged lipids, however, anionic lipids were found to associate strongly with a pocket of charged residues at the cytoplasmic end of TMD2 (B). This charged pocket lies at the beginning of a charge cluster, RKKEE, postulated to form a pH-sensing domain that regulates the channel sensitivity (C) (Kloda et al., 2006).


227

were implicated (Powl et al., 2005) (Fig. 5). Disruption of one charge in this cluster significantly reduced this anionic lipid binding. Further manipulation of this pocket was found to perturb lipid association and yield a GOF phenotype, possibly reflecting a conformational disruption of the protein. Interestingly, this charge cluster is proximal to, and included in, a charged domain that has been postulated to function as a pH sensor. In this latter study, this and other charges just cytoplasmic to TMD2 were shown to lead to a pH modification of channel sensitivity in patch clamp. The authors suggested that this region may interact with charted lipids in a pH-sensitive manner (Kloda et al., 2006). A previous study was consistent with the findings shown for the wild-type channel (Iscla et al., 2004); however, a functional role for this potential pH modulation has yet to be demonstrated in vivo. Together the data discussed above are beginning to give a glimpse of possible interactions between the MscL protein and the headgroups of the lipids. If the interactions are specific, as some of the studies imply, then one would anticipate that changing lipid headgroups would have influences at least as profound as mutation of the protein. To test this hypothesis, one group assessed the influence of changing lipid composition on channel function (Moe and Blount, 2005). Lipids with phosphatidylcholine headgroups are not generated by bacteria and were, therefore, used as a standard. The addition of phosphatidylserine, which contains a negative charge, did not influence the tension needed to gate the channel. The two major lipid headgroups in the E. coli cytoplasmic membrane were also tested. The major negatively charged E. coli lipid headgroup, phosphatidylglycerol also had no eVect. The other major headgroup, phosphatidylethanolamine, actually led to a channel activity with at a lower rather than higher sensitivity to membrane tension, suggesting that this lipid eVects altered activity through changes in the biophysical properties of the membrane rather than through MscL-lipid specific interactions. This study cannot rule out the possibilities that there are nonfunctional interactions or that some minor lipid headgroup plays a positive role in MscL channel function. However, it does appear that none of the major E. coli lipid headgroups specifically interact functionally with residues of the MscL channel.

VI. MscL AS A POSSIBLE NANOSENSOR As nanotechnology increases in promise and scope, researchers are beginning to realize that biosensors may fill the role of nanosensors in many systems, allowing electrical current, or the release of small chemicals or drugs on stimulation. MscL, with its large pore size and streamlined structure, is a prime candidate for such a nanosensor.

228

Blount et al.

In several experiments, the modality of the channel has been measured or changed. One study, using whole cells from wild-type and an MscL-null mutant, implied that the MscL channel responds to increases in temperature (Jones et al., 2000); however, recent experiments have found the opposite influence of temperature on reconstituted channels (Li et al., unpublished results), suggesting the cellular responses may be indirect and not an inherent property of the simplest lipid-protein system. In other studies, either the lipid composition (Folgering et al., 2004) or protein structure (Koçer et al., 2005) has been modified to generate a system that is sensitive to light of specific wavelengths. Finally, mutagenesis (Iscla et al., 2004) and protein modification (Koçer et al., 2006) have made channels with an increased sensitivity or spontaneous activity at a specific pH. The channel has been shown to function while reconstituted in lipid vesicles (Koçer et al., 2005, 2006); unfortunately, the one published attempt at reconstituting the channel into a nonlipid matrix led to channels that were frozen in specific conformations, depending on the pressures used to generate the matrix (Ornatska et al., 2003). Researchers have also investigated whether the MscL channel can be synthesized in vitro rather than producing it in a biological system; the obvious advantage would be that large quantities could be generated. One study demonstrated that the channel can be synthesized in a cell-free system in the presence of detergent rather than phospholipid membranes; a functional channel is produced as assayed by patch clamp after reconstituted into membranes (Berrier et al., 2004). In other studies, the channel was generated by total chemical synthesis (Clayton et al., 2004); in a later work, such a synthesized channel was shown to be able to form functional channels when reconstituted into lipid membranes (Becker et al., 2004).

VII. CONCLUSIONS MscL has given us our first and deepest glimpse into the workings of a mechanosensory channel. In less than 13 years, we have progressed from identifying the gene that encodes the MscL activity to detailed structural and mechanistic models. However, there are still many details to be worked out and questions to be addressed. It remains the challenge of the future to determine exactly how the channel detects membrane tension and the precise structural rearrangements the channel undergoes on stimulation. Acknowledgments Our work is supported by Grant I-1420 of the Welch Foundation, Grant FA9550-05-1-0073 of the Air Force OYce of Scientific Review, and Grant GM61028 from the National Institutes of Health.


229

References Ajouz, B., Berrier, C., Garrigues, A., Besnard, M., and Ghazi, A. (1998). Release of thioredoxin via the mechanosensitive channel MscL during osmotic downshock of Escherichia coli cells. J. Biol. Chem. 273, 26670–26674. Ajouz, B., Berrier, C., Besnard, M., Martinac, B., and Ghazi, A. (2000). Contributions of the diVerent extramembranous domains of the mechanosensitive ion channel MscL to its response to membrane tension. J. Biol. Chem. 275, 1015–1022. Akabas, M. H., and Karlin, A. (1999). Substituted-cysteine accessibility method. In ‘‘Methods Enzymology,’’ Vol. 294, pp. 123–144. Academic Press, New York. Anishkin, A., Gendel, V., Sharifi, N. A., Chiang, C. S., Shirinian, L., Guy, H. R., Sukharev, S., and Department of Biology, U. o. M. C. P. U. S. A. (2003). On the conformation of the COOH-terminal domain of the large mechanosensitive channel MscL. J. Gen. Physiol. 121(3), 227–244. Anishkin, A., Chiang, C. S., and Sukharev, S. (2005). Gain-of-function mutations reveal expanded intermediate states and a sequential action of two gates in MscL. J. Gen. Physiol. 125, 155–170. Arkin, I. T., Sukharev, S. I., Blount, P., Kung, C., and Brunger, A. T. (1998). Helicity, membrane incorporation, orientation and thermal stability of the large conductance mechanosensitive ion channel from E. coli. Biochim. Biophys. Acta 1369, 131–140. Bartlett, J. L., Levin, G., and Blount, P. (2004). An in vivo assay identifies changes in residue accessibility on mechanosensitive channel gating. Proc. Natl. Acad. Sci. USA 101, 10161–10165. Bartlett, J. L., Li, Y., and Blount, P. (2006). Mechanosensitive channel gating transitions resolved by functional changes upon more modification. Biophys. J. 91, 3684–3691. Batiza, A. F., Kuo, M. M., Yoshimura, K., and Kung, C. (2002). Gating the bacterial mechanosensitive channel MscL in vivo. Proc. Natl. Acad. Sci. USA 99, 5643–5648. Becker, C. F., Clayton, D., Shapovalov, G., Lester, H. A., and Kochendoerfer, G. G. (2004). On-resin assembly of a linkerless lanthanide(III)-based luminescence label and its application to the total synthesis of site-specifically labeled mechanosensitive channels. Bioconjug. Chem. 15, 1118–1124. Berrier, C., Coulombe, A., Houssin, C., and Ghazi, A. (1989). A patch-clamp study of ion channels of inner and outer membranes and of contact zones of E. coli, fused into giant liposomes. Pressure-activated channels are localized in the inner membrane. FEBS Lett. 259, 27–32. Berrier, C., Besnard, M., Ajouz, B., Coulombe, A., and Ghazi, A. (1996). Multiple mechanosensitive ion channels from Escherichia coli, activated at diVerent thresholds of applied pressure. J. Membr. Biol. 151, 175–187. Berrier, C., Garrigues, A., Richarme, G., and Ghazi, A. (2000). Elongation factor Tu and DnaK are transferred from the cytoplasm to the periplasm of Escherichia coli during osmotic downshock presumably via the mechanosensitive channel mscL. J. Bacteriol. 182, 248–251. Berrier, C., Park, K. H., Abes, S., Bibonne, A., Betton, J. M., and Ghazi, A. (2004). Cell-free synthesis of a functional ion channel in the absence of a membrane and in the presence of detergent. Biochemistry 43, 12585–12591. Betanzos, M., Chiang, C. S., Guy, H. R., and Sukharev, S. (2002). A large iris-like expansion of a mechanosensitive channel protein induced by membrane tension. Nat. Struct. Biol. 9, 704–710. Bilston, L. E., and Mylvaganam, K. (2002). Molecular simulations of the large conductance mechanosensitive (MscL) channel under mechanical loading. FEBS Lett. 512, 185–190. Blount, P., and Moe, P. (1999). Bacterial mechanosensitive channels: Integrating physiology, structure and function. Trends Microbiol. 7, 420–424.

230

Blount et al.

Blount, P., Sukharev, S. I., Moe, P. C., Nagle, S. K., and Kung, C. (1996a). Towards an understanding of the structural and functional properties of MscL, a mechanosensitive channel in bacteria. Biol. Cell 87, 1–8. Blount, P., Sukharev, S. I., Moe, P. C., Schroeder, M. J., Guy, H. R., and Kung, C. (1996b). Membrane topology and multimeric structure of a mechanosensitive channel protein of Escherichia coli. EMBO J. 15, 4798–4805. Blount, P., Sukharev, S. I., Schroeder, M. J., Nagle, S. K., and Kung, C. (1996c). Single residue substitutions that change the gating properties of a mechanosensitive channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 93, 11652–11657. Britten, R. J., and McClure, F. T. (1962). The amino acid pool in Escherichia coli. Bacteriol. Rev. 26, 292–335. Chang, G., Spencer, R. H., Lee, A. T., Barclay, M. T., and Rees, D. C. (1998). Structure of the MscL homolog from Mycobacterium tuberculosis: A gated mechanosensitive ion channel. Science 282, 2220–2226. Chiang, C. S., Anishkin, A., and Sukharev, S. (2004). Gating of the large mechanosensitive channel in situ: Estimation of the spatial scale of the transition from channel population responses. Biophys. J. 86, 2846–2861. Chiang, C. S., Shirinian, L., and Sukharev, S. (2005). Capping transmembrane helices of MscL with aromatic residues changes channel response to membrane stretch. Biochemistry 44, 12589–12597. Clayton, D., Shapovalov, G., Maurer, J. A., Dougherty, D. A., Lester, H. A., and Kochendoerfer, G. G. (2004). Total chemical synthesis and electrophysiological characterization of mechanosensitive channels from Escherichia coli and Mycobacterium tuberculosis. Proc. Natl. Acad. Sci. USA 101, 4764–4769. Colombo, G., Marrink, S. J., and Mark, A. (2003). Simulation of MscL gating in a bilayer under stress. Biophys. J. 84, 2331–2337. Cruickshank, C. C., Minchin, R. F., Le Dain, A. C., and Martinac, B. (1997). Estimation of the pore size of the large-conductance mechanosensitive ion channel of Escherichia coli. Biophys. J. 73, 1925–1931. Efimov, V. P., Engel, J., and Malashkevich, V. N. (1996). Crystallization and preliminary crystallographic study of the pentamerizing domain from cartilage oligomeric matrix protein: A five-stranded alpha-helical bundle. Proteins 24, 259–262. Elmore, D. E., and Dougherty, D. A. (2001). Molecular dynamics simulations of wild-type and mutant forms of the Mycobacterium tuberculosis MscL channel. Biophys. J. 81, 1345–1359. Ewis, H. E., and Lu, C. D. (2005). Osmotic shock: A mechanosensitive channel blocker can prevent release of cytoplasmic but not periplasmic proteins. FEMS Microbiol. Lett. 253, 295–301. Folgering, J. H., Kuiper, J. M., de Vries, A. H., Engberts, J. B., and Poolman, B. (2004). Lipidmediated light activation of a mechanosensitive channel of large conductance. Langmuir 20, 6985–6987. Gullingsrud, J., and Schulten, K. (2003). Gating of MscL studied by steered molecular dynamics. Biophys. J. 85, 2087–2099. Gullingsrud, J., Kosztin, D., and Schulten, K. (2001). Structural determinants of MscL gating studied by molecular dynamics simulations. Biophys. J. 80, 2074–2081. Gustin, M. C., Zhou, X. L., Martinac, B., and Kung, C. (1988). A mechanosensitive ion channel in the yeast plasma membrane. Science 242, 762–765. Hamann, A., Bossemeyer, D., and Bakker, E. P. (1987). Physical mapping of the Kþ transport trkA gene of Escherichia coli and overproduction of the TrkA protein. J. Bacteriol. 169, 3138–3145.


231

Hamill, O. P., Marty, A., Neher, E., Sakmann, B., and Sigworth, F. J. (1981). Improved patchclamp techniques for high-resolution current recording from cells and cell-free membrane patches. Pflugers. Arch. 391, 85–100. Ha¨se, C. C., Ledain, A. C., and Martinac, B. (1997a). Molecular dissection of the large mechanosensitive ion channel (Mscl) of E. coli-mutants with altered channel gating and pressure sensitivity. J. Membr. Biol. 157, 17–25. Ha¨se, C. C., Minchin, R. F., Kloda, A., and Martinac, B. (1997b). Cross-linking studies and membrane localization and assembly of radiolabelled large mechanosensitive ion channel (MscL) of Escherichia coli. Biochem. Biophys. Res. Commun. 232, 777–782. Iscla, I., Levin, G., Wray, R., Reynolds, R., and Blount, P. (2004). Defining the physical gate of a mechanosensitive channel, MscL, by engineering metal-binding sites. Biophys. J. 87(5), 3172–3180. Iscla, I., Levin, G., Wray, R., and Blount, P. (2006). Disulfide trapping the mechanosensitive channel MscL into a gating‐transition state. Biophys. J. [Eprint ahead of print 0006‐3495]. Jones, S. E., Naik, R. R., and Stone, M. O. (2000). Use of small fluorescent molecules to monitor channel activity. Biochem. Biophys. Res. Commun. 279, 208–212. Kloda, A., Ghazi, A., and Martinac, B. (2006). C-terminal charged cluster of MscL, RKKEE, functions as a pH sensor. Biophys. J. 90, 1992–1998. Koçer, A., Walko, M., Meijberg, W., and Feringa, B. L. (2005). A light-actuated nanovalve derived from a channel protein. Science 309, 755–758. Koçer, A., Walko, M., Bulten, E., Halza, E., Feringa, B., and Meijberg, W. (2006). Rationally designed chemical modulators convert a bacterial channel protein into a pH-sensory valve. Angew. Chem. Int. Ed. Engl. 45, 3126–3130. Kong, Y., Shen, Y., Warth, T. E., and Ma, J. (2002). Conformational pathways in the gating of Escherichia coli mechanosensitive channel. Proc. Natl. Acad. Sci. USA 99, 5999–6004. Levin, G., and Blount, P. (2004). Cysteine scanning of MscL transmembrane domains reveals residues critical for mechanosensitive channel gating. Biophys. J. 86, 2862–2870. Levina, N., Totemeyer, S., Stokes, N. R., Louis, P., Jones, M. A., and Booth, I. R. (1999). Protection of Escherichia coli cells against extreme turgor by activation of MscS and MscL mechanosensitive channels: Identification of genes required for MscS activity. EMBO J. 18, 1730–1737. Li, Y., Moe, P. C., Chandrasekaran, S., Booth, I. R., and Blount, P. (2002). Ionic regulation of MscK, a mechanosensitive channel from Escherichia coli. EMBO J. 21, 5323–5330. Li, Y., Wray, R., and Blount, P. (2004). Intragenic suppression of gain-of-function mutations in the Escherichia coli mechanosensitive channel, MscL. Mol. Microbiol. 53, 485–495. Malashkevich, V. N., Kammerer, R. A., Efimov, V. P., Schulthess, T., and Engel, J. (1996). The crystal structure of a five-stranded coiled coil in COMP: A prototype ion channel? Science 274, 761–765. Manoil, C., and Beckwith, J. (1986). A genetic approach to analyzing membrane protein topology. Science 233, 1403–1408. Martinac, B., Buechner, M., Delcour, A. H., Adler, J., and Kung, C. (1987). Pressure-sensitive ion channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 84, 2297–2301. Martinac, B., Adler, J., and Kung, C. (1990). Mechanosensitive ion channels of E. coli activated by amphipaths. Nature 348, 261–263. Maurer, J. A., and Dougherty, D. A. (2003). Generation and evaluation of a large mutational library from the Escherichia coli mechanosensitive channel of large conductance, MscL‐Implications for channel gating and evolutionary design. J. Biol. Chem. 278, 21076–21082. Maurer, J. A., Elmore, D., Lester, H., and Dougherty, D. (2000). Comparing and contrasting Escherichia coli and Mycobacterium tuberculosis mechanosensitive channels (MscL). New gain of function mutations in the loop region. J. Biol. Chem. 275, 22238–22244.

232

Blount et al.

Meyer, G. R., Gullingsrud, J., Schulten, K., and Martinac, B. (2006). Molecular dynamics study of MscL interactions with a curved lipid bilayer. Biophys. J. 91, 1630–1637. Moe, P., and Blount, P. (2005). Assessment of potential stimuli for mechano-dependent gating of MscL: EVects of pressure, tension, and lipid headgroups. Biochemistry 44, 12239–12244. Moe, P. C., Blount, P., and Kung, C. (1998). Functional and structural conservation in the mechanosensitive channel MscL implicates elements crucial for mechanosensation. Mol. Microbiol. 28, 583–592. Moe, P. C., Levin, G., and Blount, P. (2000). Correlating a protein structure with function of a bacterial mechanosensitive channel. J. Biol. Chem. 275, 31121–31127. Ornatska, M., Jones, S. E., Naik, R. R., Stone, M. O., and Tsukruk, V. V. (2003). Biomolecular stress-sensitive gauges: Surface-mediated immobilization of mechanosensitive membrane protein. J. Am. Chem. Soc. 125, 12722–12723. Ou, X., Blount, P., HoVman, R. J., and Kung, C. (1998). One face of a transmembrane helix is crucial in mechanosensitive channel gating. Proc. Natl. Acad. Sci. USA 95, 11471–11475. Park, K. H., Berrier, C., Martinac, B., and Ghazi, A. (2004). Purification and functional reconstitution of N- and C-halves of the MscL channel. Biophys. J. 86, 2129–2136. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2001). Site-directed spin-labeling analysis of reconstituted Mscl in the closed state. J. Gen. Physiol. 118, 193–206. Perozo, E., Cortes, D. M., Sompornpisut, P., Kloda, A., and Martinac, B. (2002a). Open channel structure of MscL and the gating mechanism of mechanosensitive channels. Nature 418, 942–948. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2002b). Physical principles underlying the transduction of bilayer deformation forces during mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703. Powl, A. M., East, J. M., and Lee, A. G. (2005). Heterogeneity in the binding of lipid molecules to the surface of a membrane protein: Hot spots for anionic lipids on the mechanosensitive channel of large conductance MscL and eVects on conformation. Biochemistry 44, 5873–5883. Saint, N., Lacapere, J. J., Gu, L. Q., Ghazi, A., Martinac, B., and Rigaud, J. L. (1998). A hexameric transmembrane pore revealed by two-dimensional crystallization of the large mechanosensitive ion channel (MscL) of Escherichia coli. J. Biol. Chem. 273, 14667–14670. Schleyer, M., Schmid, R., and Bakker, E. P. (1993). Transient, specific and extremely rapid release of osmolytes from growing cells of Escherichia coli K-12 exposed to hypoosmotic shock. Arch. Microbiol. 160, 424–431. Shapovalov, G., Bass, R., Rees, D. C., and Lester, H. A. (2003). Open-state disulfide crosslinking between Mycobacterium tuberculosis mechanosensitive channel subunits. Biophys. J. 84, 2357–2365. Sukharev, S., Betanzos, M., Chiang, C., and Guy, H. (2001a). The gating mechanism of the large mechanosensitive channel MscL. Nature 409, 720–724. Sukharev, S., Durell, S., and Guy, H. (2001b). Structural models of the MscL gating mechanism. Biophys. J. 81, 917–936. Sukharev, S. I., Martinac, B., Arshavsky, V. Y., and Kung, C. (1993). Two types of mechanosensitive channels in the Escherichia coli cell envelope: Solubilization and functional reconstitution. Biophys. J. 65, 177–183. Sukharev, S. I., Blount, P., Martinac, B., Blattner, F. R., and Kung, C. (1994). A largeconductance mechanosensitive channel in E. coli encoded by mscL alone. Nature 368, 265–268. Sukharev, S. I., Schroeder, M. J., and McCaslin, D. R. (1999a). Stoichiometry of the large conductance bacterial mechanosensitive channel of E. coli. A biochemical study. J. Membr. Biol. 171, 183–193.


233

Sukharev, S. I., Sigurdson, W. J., Kung, C., and Sachs, F. (1999b). Energetic and spatial parameters for gating of the bacterial large conductance mechanosensitive channel, MscL. J. Gen. Physiol. 113, 525–540. Tavernarakis, N., and Driscoll, M. (1997). Molecular modeling of mechanotransduction in the nematode Caenorhabditis elegans. Annu. Rev. Physiol. 59, 659–689. Tsai, I. J., Liu, Z. W., Rayment, J., Norman, C., McKinley, A., and Martinac, B. (2005). The role of the periplasmic loop residue glutamine 65 for MscL mechanosensitivity. Eur. Biophys. J. 34, 403–412. Valadie, H., Lacapcre, J. J., Sanejouand, Y. H., and Etchebest, C. (2003). Dynamical properties of the MscL of Escherichia coli: A normal mode analysis. J. Mol. Biol. 332, 657–674. Vazquez-Laslop, N., Lee, H., Hu, R., and Neyfakh, A. A. (2001). Molecular sieve mechanism of selective release of cytoplasmic proteins by osmotically shocked Escherichia coli. J. Bacteriol. 183, 2399–2404. Yoshimura, K., Batiza, A., Schroeder, M., Blount, P., and Kung, C. (1999). Hydrophilicity of a single residue within MscL correlates with increased channel mechanosensitivity. Biophys. J. 77, 1960–1972. Yoshimura, K., Batiza, A., and Kung, C. (2001). Chemically charging the pore constriction opens the mechanosensitive channel MscL. Biophys. J. 80, 2198–2206. Yoshimura, K., Nomura, T., and Sokabe, M. (2004). Loss-of-function mutations at the rim of the funnel of mechanosensitive channel MscL. Biophys. J. 86, 2113–2120.

CHAPTER 9 The Bacterial Mechanosensitive Channel MscS: Emerging Principles of Gating and Modulation Sergei Sukharev, Bradley Akitake, and Andriy Anishkin Department of Biology, University of Maryland, College Park, Maryland 20742

I. Overview II. Introduction III. MscS and Its Relatives A. A Brief Account of Bacterial Osmoregulation and the Discovery of MscS B. MscS Vs MscK: How to Interpret Early Functional Data? C. Purification and Reconstitution of MscS Showed Homo‐Multimeric Channels Activated by Tension in the Lipid Bilayer IV. Structural and Computational Studies A. Structure of MscS and First Hypotheses About Its Gating Mechanism B. Computational Studies of MscS V. Functional Properties of MscS A. MscS Conduction and Selectivity B. Gating Characteristics of MscS In Situ C. Mutations That AVect MscS Activity D. MscS Inactivation VI. What Do the Closed, Open, and Inactivated States of MscS Look Like? A. Is the Crystal Structure a Native State? B. Closed State C. Open State VII. Emerging Principles of MscS Gating and Regulation and the New Directions References


1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58009-3

236

Sukharev et al.

I. OVERVIEW The mechanosensitive channel of small conductance (MscS) is a tension‐ driven osmolyte release valve, which plays a critical role in bacterial adaptation to low osmolarity. Homologues of MscS have been identified in yeast and higher plants with at least two molecular species being involved in the regulation of chloroplast volume and fission. In bacteria, MscS and MscL channels are the major components of a turgor‐driven emergency osmolyte eZux system. Of the two, MscS is considerably more complex, as it shows time‐dependent patterns of gating and adaptation. This chapter will attempt to present a current interpretation of the main functional traits of MscS in light of its crystal structure. Responding directly to membrane tension, MscS generates either sustained or transient permeability responses cycling through at least three functional states. The response amplitude (Po) depends on the rate of tension application, whereas the duration, defined by inactivation kinetics, depends on the magnitude of the mechanical stimulus, voltage, and presence of specific solutes. Molecular dynamics (MD) simulations data from several groups show that the pore of MscS, in the crystal conformation, is largely dehydrated and likely nonconductive. In addition, instability of the splayed crystal conformation in the lipid bilayer during simulations suggests that the crystal structure may not represent a native state for the protein. Models of the opening process, based on experimental data, depict a substantial expansion of the hydrophobic pore accompanied by a critical wetting event required for the onset of conduction. Simulations also suggest that the characteristic kink, breaking the pore‐lining TM3 helix at G113, may straighten in the open conformation. Several hypotheses will be presented, based on the phenomenology and results of computational analysis, about the character of conformational transitions in MscS between its resting, open, and inactivated states, the significance of pore dehydration and the putative function of the cage‐like cytoplasmic domain as an internal, solute‐specific osmosensor.

II. INTRODUCTION Osmoregulation is fundamental problem for all organisms because even relatively small osmotic imbalances can produce large and damaging volume or intracellular solute concentration changes. It is for this reason that physicians often scrutinize a patient’s fluid and electrolyte balance for diagnostic clues. Although the cell physiology of osmoregulation has been studied for decades, the existing molecular information is fragmental. Even in bacteria, which are the smallest and therefore ‘‘simplest’’ of all free‐living

9. The Bacterial Mechanosensitive Channel MscS

237

organisms, the primary mechanisms of osmosensing and osmoregulation are not fully understood. Solving this puzzle of how a cell senses water activity inside and outside and then triggers multiple adaptive pathways to counteract perturbation would provide an important link between the fundamental thermodynamics of macromolecular (or membrane) solvation and organismal physiology (Wood, 1999). The study of these systems in bacteria is important for several reasons. First, understanding the mechanisms of osmosensing and the accumulation/release or exchange of compatible osmolytes in bacteria would provide insight into both the physical and evolutionary biochemistry of these systems. Second, osmoregulatory mechanisms play key roles in parasite–host or symbiont–host interactions (Stewart et al., 2005) and determine the environmental stability of pathogens/symbionts inside and outside of the host. Study of these mechanisms will potentially aid in fight to stop the spread of infectious diseases. It is, therefore, fortunate that these bacterial systems provide extremely convenient tools to gain a basic understanding of macromolecular function in particular for the biophysical studies of membrane proteins. The MscS is a bacterial osmolyte release valve that limits turgor pressure during osmotic downshock, thus rescuing cells from lysis (Levina et al., 1999). Together with three other mechanosensitive channels, MscK (Kþ dependent), MscL (large conductance), and the yet unidentified mini‐ channel MscM, MscS constitutes a partially redundant system, which provides graded permeability response to osmotic stress (Berrier et al., 1996; Blount et al., 1999; Martinac, 2001; Booth et al., 2003). Like the large conductance channel MscL, MscS gates directly by tension developed in the surrounding lipid bilayer. In contrast to MscL, which is characterized by steady, tension‐dependent, activities observed near the membrane’s lytic limit (8–12 dyne/cm), MscS responds to moderate tensions (4–6 dyne/cm) usually with a transient (decaying) response implying activation followed by transition to an inactivated or desensitized state(s) (Koprowski and Kubalski, 1998; Levina et al., 1999; Akitake et al., 2005). The crystal structures of MscL (homologue from Mycobacterium tuberculosis) and Escherichia coli MscS have been solved by Rees and coworkers (Chang et al., 1998; Bass et al., 2002) and provide critical information and a strong impetus to the field. The more elaborate structure of MscS currently guides eVorts of several experimental and computational groups toward unraveling the molecular mechanism of its gating and intricate adaptive behaviors. While MscL is primarily a prokaryotic molecule, with several distant homologues in fungi, MscS‐type channels appear to be more widespread (Pivetti et al., 2003). Multiple homologues have recently been identified in higher plants. Three such homologues in Arabidopsis thaliana were found to regulate the volume and division of chloroplasts (Haswell and Meyerowitz, 2006). This chapter

238

Sukharev et al.

will provide an update focused primarily on the functional studies of E. coli MscS and discuss the emerging paradigms of MscS‐gating mechanism in the frameworks of structural information and known principles of bacterial osmoregulation.

III. MscS AND ITS RELATIVES A. A Brief Account of Bacterial Osmoregulation and the Discovery of MscS Bacteria and other walled cells maintain their volumes and shape by creating positive turgor pressure in the cytoplasm against the confines of their cell wall. This is achieved through an osmotic network of solute uptake systems accompanied by de novo synthesis of substances that maintain normal osmotic pressure gradients across the cytoplasmic membrane. It is generally accepted that an intracellular pressure of 3–4 atm is required for normal proliferation of enteric bacteria such as E. coli and much higher pressures (15–25 atm) are typical for Gram‐positive species (Csonka, 1989; Wood, 1999). Under steady growth conditions, turgor pressure is collectively created by small molecular constituents such as free amino acids, polyols, nucleotides, and two major intracellular ions, Kþ and glutamate. If the medium suddenly becomes hypertonic due to increased salinity, bacterial cells initially loose turgor but quickly regain it by accumulating extra Kþ and glutamate (Epstein, 2003) through independent transport systems (McLaggan et al., 1994). Kþ, however, is not the most optimal intracellular osmolyte as its elevation changes the ionic strength inside the cell. Thus, having restored normal turgor, the cells gradually exchange Kþ for more inert substances such as proline, betaine, and to some extent trehalose (Csonka and Hanson, 1991). These solutes are called ‘‘compatible’’ as they minimally aVect cellular functions, though still aid in the retention of water. The growth conditions for enteric and soil bacteria are not always steady and periods of proliferation in high‐osmolarity environments are often interrupted by low‐osmolarity conditions (rain/drainage water). In these instances, previously accumulated solutes put the bacterium at risk for osmotic lysis as the peptidoglycan cell wall can restrain the volume change but only to a certain extent. It has been shown that peptidoglycan is distensible, and cell swelling results in substantial increases of internal volume (Koch and Woeste, 1992). As the excess slack of the inner membrane is used up, tension in the bilayer starts to build. Amelioration of this tension was found to occur through a drastic permeability increase of the cell membrane followed by the release small solutes to reduce the osmotic gradient (Britten and McClure, 1962; Tsapis and Kepes, 1977). This reversible permeability change, demonstrated


239

to take 1–2 s with little eVect on cell survival, was initially interpreted as transient membrane ‘‘crack’’ formation or some sort of reversible membrane breakdown. Later studies (Schleyer et al., 1993) refined the repertoire of released osmolytes and linked the reversible permeability pathways to the activities of bacterial stretch‐activated channels first reported by the Kung laboratory (Martinac et al., 1987). By applying the standard patch‐clamp technique (Hamill et al., 1981) to giant spheroplast preparations (Ruthe and Adler, 1985), Martinac et al. (1987) first documented the presence of 1‐nS pressure‐activated channels in the E. coli cell envelope. This channel was more active under depolarizing voltages (cytoplasm positive). Its activity was also found to be dependent on the presence of Kþ as the channel was completely inactive in Naþ buVers. Further studies involving membrane solubilization and reconstitution experiments (Sukharev et al., 1993) revealed that E. coli possesses two distinct mechanically activated channel species, one of smaller (1 nS) conductance, similar to those previously observed by Martinac and designated as MscS, and a channel with three times larger (3.2 nS) conductance and spiky activities named MscL. The MscL protein was subsequently isolated and its gene cloned (Sukharev et al., 1994). Surprisingly, it was found that the mscL‐ null mutant did not exhibit any growth/survival phenotype under a variety of osmotic conditions. Identification and cloning of MscS and MscK by the Booth laboratory came in 1999 through studies of mutations which altered Kþ exchange with betaine and genomic analysis of related sequences (Levina et al., 1999). In the event of osmotic upshock, primary Kþ uptake is mediated by the Trk, Ktr, and Kdp pumps activated by low turgor (Epstein, 2003). The subsequent accumulation of proline and betaine is accomplished by the ProU, ProP, and BetT active transporters, which also fulfill the roles of osmosensors (Wood et al., 2001). While searching for the Kþ eZux and accompanying proline– betaine uptake pathways, Booth and collaborators studied the kefA, kefB, and kefC loci. The two latter systems were found to be active primarily under toxic or oxidative stresses (Ferguson et al., 2000). A kefA‐null mutation did not change the course of betaine‐for‐Kþ exchange under high osmolarity either, but a point mutation in kefA (called RQ2) exhibited a growth‐ suppressing phenotype when the cells were grown in a high‐osmotic medium in the presence of Kþ and betaine. The RQ2 bacteria accumulated betaine normally, but were unable to extrude Kþ eYciently and became extremely sensitive to turgor created by the two osmolytes (McLaggan et al., 2002). From this phenotype, it seemed possible that the product of kefA was a stretch‐sensitive membrane protein. The kefA open reading frame coded for a large (1120 amino acid) multidomain membrane protein. The search for E. coli sequences similar to kefA revealed two proteins of the same length

240

Sukharev et al.

(AefA and YjeP) and a smaller protein resembling the C‐terminal part of KefA, named YggB. Generation of single and multiple null mutants followed by reexpression of individual genes ‘‘in trans’’ revealed that kefA and yggB both code for mechanosensitive channels of similar 1‐nS conductance. KefA, characterized by sustained activities under constant pressure gradients, was shown to be dependent on the presence of Kþ (Levina et al., 1999), thus it was renamed MscK. The product of yggB, which was more abundant and generated transient responses to pressure steps, was renamed MscS. None of the other related open reading frames (ORFs) produced measurable channel activities in situ. It was discovered that the generated triple mscK‐, mscS‐, mscL‐knockout strain (MJF465) had increased sensitivity to moderate downshocks (400–500 mOsm), and reexpression of either MscL or MscS fully restored the osmotic downshift tolerance to wild‐type (WT) levels. Expression of MscK alone in the triple knockout strain did not ameliorate its osmotic fragility (Levina et al., 1999). This ground‐breaking work not only identified two new genes for mechanosensitive channels of very similar conductance (previously though to be one), but also unequivocally demonstrated the partially redundant physiological roles for MscS and MscL as turgor‐limiting release valves. In addition, MJF 465 has become a highly useful ‘‘clean’’ background system for mechanosensitive channel expression, electrophysiology recording, and testing of the osmotically driven rescuing ability of multiple MscS and MscL mutants. B. MscS Vs MscK: How to Interpret Early Functional Data? The earliest functional studies of MscS‐like mechanosensitive channels in E. coli spheroplasts conducted by Martinac et al. reported on a population of pressure‐sensitive ion channels of 1‐nS conductance. Single channel recordings revealed pressure dependence, a strong and sharp increase of channel open probability observed with increasing suction, and voltage dependence. Increasingly depolarizing voltages made the channel population easier to open and shifted the activation curves progressively to the left. Subsequent study of these channels in the whole‐protoplast recording mode by Cui et al. (1995) revealed similar phenomenology. Data from these early studies were obviously collected on mixed populations of channels, and it is now known that a WT E. coli background contains both MscS and MscK with similar activities. Of the two, MscK is a much larger multidomain protein. From an alignment of the primary sequences, the 286‐ amino acid‐long MscS shares 23% identity and 53% similarity to the C‐terminal part of MscK. Although these numbers are not quite high enough to conclude structural identity, analysis of the residue


241

patterns, especially those in the transmembrane (TM) region, suggested very similar organization, and provides some rationale for their similar gating characteristics (Levina et al., 1999; Miller et al., 2003a). PhoA‐fusion experiments concluded that the common parts of MscK and MscS each contain three TM domains, with the N‐terminal ends of the protein being periplasmic and the larger soluble C‐terminal domains residing in the cytoplasm (Miller et al., 2003a). MscK also has a large N‐terminal domain that was predicted to form up to seven additional TM spans. In spite of the structural diVerences, both channels were characterized with similar 1‐nS conductances, a slight anionic preference, and weak inward rectification (Li et al., 2002). Studies of MscS expressed in a clean genetic background (Levina et al., 1999; Vasquez and Perozo, 2004; Akitake et al., 2005) revealed functional characteristics that substantially diVered from those originally reported by Martinac and Cui (Martinac et al., 1987; Cui et al., 1995). A detailed comparison revealed that MscK is activated at slightly lower tensions than MscS (Li et al., 2002); however the primary distinction between the two channels lies in their kinetic behavior and regulation. With symmetrical KCl on both sides, MscK exhibits sustained activity in the entire range of activating pressures, whereas under mechanical stimulation below saturating levels, MscS response was transient with characteristic inactivation (Levina et al., 1999; Akitake et al., 2005). Both Martinac and Cui used constant pressure stimuli held for extended durations, conditions at which MscS populations inactivate quickly. MscK was found to be critically dependent on the monovalent cation species present in the recording solution showing activity only when Kþ, NH4þ, Rbþ, or Csþ was present in the periplasmic medium and inactive when replaced with Naþ or Liþ (Li et al., 2002). In contrast, MscS activity was demonstrated to be the same in both Naþ and Kþ buVers (Li et al., 2002; Akitake et al., 2005). The channel population studied by Martinac showed marked changes in activity becoming harder to open with faster kinetics in Naþ buVers. Cui et al. also studied the RQ2 strain of E. coli and noted that in this mutant MscS‐like channel activity displayed a lowered opening threshold. RQ2 was later found to harbor a missense gain‐of‐ function (GOF) mutation in the mscK (formerly kefA) gene (McLaggan et al., 2002). The position of the RQ2 GOF mutation (G922S) was mapped to the gate region in the most C‐terminal TM span of MscK (equivalent to the A106 position in the TM3 helix of MscS). Finally, it has been shown that the opening transition in MscS is not aVected by voltage (Vasquez and Perozo, 2004; Akitake et al., 2005; Nomura et al., 2006). From these comparisons, it is now evident that the early phenomenology of MscS‐like channels in E. coli as reported by Martinac and Cui belongs primarily to MscK, not MscS.

242

Sukharev et al.

C. Purification and Reconstitution of MscS Showed Homo‐Multimeric Channels Activated by Tension in the Lipid Bilayer After the sequence became available, the mscS gene was amplified by PCR and the coding sequence was modified with a 6‐His tag on its C‐terminus (Sukharev, 2002; Nomura et al., 2006). The position of this tag was shown to have minimal eVects on channel function under normal ionic conditions (Koprowski and Kubalski, 2003). The protein expressed in E. coli was purified on an Ni‐NTA column and appeared homogeneous. When subjected to size‐exclusion chromatography in the presence of ‐octylglucoside and lipids, the protein emerged as a 200‐kDa particle, apparently representing a stable complex. In the presence of detergent only, the complexes degraded to 30‐kDa monomers. Cross‐linking experiments with the bifunctional reagent DSS suggested that at least six identical subunits form the active complex. Reconstitution of pure MscS into liposomes revealed fully active channels characterized with a weak anionic preference, slight inward rectification, and essentially nonsaturable conductance (Sukharev, 2002). Imaging of large liposome patches, along channel recording at diVerent pressures, permitted determination of the radii of patch curvature and in such the magnitude of tension acting on channel population from the law of Laplace. The midpoint tension for MscS activation was found near 5.5 dyne/cm, in good correspondence with previous measurements by Cui and Adler (1996) performed in the whole‐protoplast recording mode. In the same way that the slope of the conductance vs voltage (G–V) curve for a voltage‐gated channel gives an estimation of the gating charge, the slope of Po/Pc on tension estimates the in‐plane area change of a mechanosensitive channel. Treatment of dose‐response curves (Fig. 2C) in a two‐state Boltzmann approximation (Sachs and Morris, 1998) predicted a protein expansion of A ¼ 8.4 nm2 and an energy diVerence between the closed and open conformations, in the absence of tension, of E ¼ 11.4 kT. Independently conducted reconstitution experiments by Okada et al. (2002) confirmed that MscS is gated directly by tension in the lipid bilayer.

IV. STRUCTURAL AND COMPUTATIONAL STUDIES A. Structure of MscS and First Hypotheses About Its Gating Mechanism Successful crystallographic work by the Rees group solved the structure of ˚ resolution (Bass et al., 2002) revealing a homoheptameric MscS to 3.9‐A complex with three TM helices (TM1–TM3) per subunit and a large hollow


243

C‐terminal domain in the cytoplasm (Fig. 1). TM1 and TM2 are bundled together and splayed at about 30 relative to the axis of the pore. Due to the splay, a characteristic crevice is seen on the cytoplasmic side, between the TM1–TM2 pair and the pore‐forming TM3. The TM1–TM2 assembly resembles the ‘‘paddle’’ of KvAP’s voltage sensor (Strop et al., 2003) and, due to the presence of several arginines in its structure, was proposed to function similarly as the voltage‐sensitive domain for MscS. The third TM helices, TM3s, line a relatively wide but very hydrophobic pore, which after a visual inspection was deemed to be in the open conformation (Bass et al., 2002). The adjacent TM3s are tightly packed through characteristic juxtaposition of conserved glycines (101, 104, and 108) and alanines (98, 102, and 106). Two rings of leucines, 105 and 109, form the narrow constriction assigned as the gate. A number of positively charged residues are situated in the pore

FIGURE 1 The crystal structure of MscS (left) (1MXM.pdb) and a vertical slice through the TM domain shown in a space‐filled representation (right). The three membrane‐spanning helices from each subunit comprise the TM domain; approximate positions of membrane boundaries are shown as horizontal lines. A large hollow domain is formed by the C‐terminal domains of each subunit in the cytoplasm. On the magnified image of the TM domains, the solvent‐ accessible surfaces are colored according to the specific desolvation energy (from Wesson and Eisenberg, 1992). The most hydrophobic areas are yellow and most hydrophilic are blue. Averaging of solvation parameters and mapping was performed using HISTAN (www.life. umd.edu/biology/sukharevlab/download.htm#HISTAN). The pore with the hydrophobic gate (G) is formed by seven TM3 helices contributed from each subunit. A deep hydrophobic crevice separates the TM3 barrel from the peripheral TM1–TM2 helices. This crevice, likely filled with detergent in the crystals, may not exist in the native state.

244

Sukharev et al.

vestibules (R88, K169) and were identified as candidate residues to confer the slight anionic selectivity of the channel as observed in experiments (Sukharev et al., 1993; Li et al., 2002; Sukharev, 2002). An unusual feature of the TM3 helices is that they are quite long and break at L111, near residue G113. The stretch of TM3 beyond the kink lies almost parallel to the membrane plane. After TM3, the polypeptide continues with the middle ‐folded domain shaping the upper hemisphere of the cytoplasmic ‘‘cage.’’ This is followed by the C‐terminal domain, which forms the ‘‘bottom’’ of the cage. The cage is perforated by seven equatorial portals and one axially positioned ‘‘crown’’ hole. The structure of 26 N‐terminal residues residing on the extracellular side was unresolved. Rees and coauthors proposed that the crystal conformation of MscS represented the open state of the channel. A rough estimation of conductance using the HOLE program (Smart et al., 1996) deemed that the structure could satisfy the observed 1‐nS conductance. However, later estimations pre˚ pore (surface‐to‐surface diameter) to dicted the need for at least an 16‐A maintain such a high current (Anishkin and Sukharev, 2004). In reference to early work by Martinac (Martinac et al., 1987), the authors proposed that tension and depolarization worked synergistically to cause the upward splay of the TM1‐TM2 bundle, a motion that was associated with channel opening. The closed state, correspondingly, had to be more compact structure and have a narrower pore. The authors did not exclude the possibilities of kinking or asymmetric packing of the TM3 helices to achieve complete closure (Bass et al., 2002). More recent studies have identified hypothetical conformations for the TM domains that are more compact than the crystal structure. Cross‐linking experiments have identified pairs of cysteines which were shown to form disulfide bridges in the resting state, while in the crystal structure they appear to be too distant to interact (Miller et al., 2003b; Edwards et al., 2004). On the basis of the idea of a more compact closed state, Booth and coworkers proposed a tighter packing of the TM3 helices in the resting state; these ideas will be discussed below (Edwards et al., 2005). B. Computational Studies of MscS 1. How Compact Is the Resting State of MscS? The solved crystal structure provided a solid starting point for the modeling of gating transitions in MscS. The first systematic attempt to envision MscS gating was a collaborative eVort between the Booth, Blount, and Bowie groups which combined experimental analysis of pore mutants with Monte‐Carlo (MC) optimization of the pore domain (Edwards et al., 2005).


245

Relying on the initial assumption that the crystal structure of the TM3 bundle represents the open state (Bass et al., 2002), the authors presented a model of a narrower conformation held together by tighter interhelical interactions. These low‐energy conformations were obtained by searching for homodimers of the 16‐residue TM3 segments. The best pairs were then duplicated around the sevenfold symmetry axis to create a heptamer. The resultant ‘‘closed’’ conformation, when compared with the crystal structure, suggested that tightening of the bundle could be achieved by decreasing the tilts of the TM3 helices and by slight rotation, permitting more optimal packing of alanine side chains in the complimentary groves between the conserved ˚ glycines. The new structure deviated from the crystal conformation by 1.8‐A root mean square deviation (RMSD), however the energy gain and gradient for the transition was not presented, thus making estimation of the distribution in the ensemble of similar conformations around this minimum diYcult. In addition, this new ‘‘closed’’ structure, although very plausible, has not been tested for stability in the context of the entire TM assembly. The ‘‘closing’’ of the pore reduced the diameter of the hydrophobic constriction to about ˚ making the pore certainly nonconductive. The expansion of the pore 2 A ˚ in diameter) gaining about 0.3 nm2 in toward the crystal structure (6.5 A cross‐sectional area was deemed suYcient to achieve the observed 1‐nS open‐ state conductance (Edwards et al., 2005). It must be reiterated, however, that this hypothetical transition rests on the assumption that crystal structure is the open conformation. More recent studies suggest that even if the crystal conformation represents a native state it would likely be nonconducting. 2. The Pore as Predicted by the Crystal Structure Is Largely Dehydrated Analysis of the pore lining indicated that L105 and L109 form not only the narrowest but also the most hydrophobic region of the MscS pore (Anishkin and Sukharev, 2004). These two ‘‘gate‐keeping’’ residues also happen to be the most conserved residues in the larger family of MscS‐type channels. The ˚ in diameter and outer chamber (periplasmic side) adjacent to the gate is 12 A was also found to be predominantly hydrophobic. The diameter between opposite solvent‐accessible surfaces in the constriction ˚ . Estimation of the pore of the crystal structure was measured to be about 6.5 A conductance, in continuum approximation with Hall’s equation (Hall, 1975; Hille, 1992), gave a value of 70 pS, one order of magnitude less than the observed open‐state conductance for MscS. In order to satisfy the observed 1‐nS conductance, it was necessary to increase the eVective diameter of con˚ . These results, which resembled dewetting transitions in striction to about 16 A hydrophobic capillaries (Beckstein et al., 2001; Beckstein and Sansom, 2003) and nanotubes (Hummer et al., 2001), prompted further investigation of

246

Sukharev et al.

water’s behavior in the suspiciously hydrophobic MscS pore. Analysis of the pore lining using the GETAREA protocol (Fraczkiewicz and Braun, 1998) and atomic solvation energies of Wesson and Eisenberg (1992) allowed for the calculation of the hydration energy profile along the pore axis (Fig. 2) and an assessment of the thermodynamic possibility of dehydration. A dewetting event was represented as an imaginary break in the water column filling the channel, thus creating a small ‘‘vapor plug’’ in the constriction. The movements of the upper and lower water surfaces decrease the contact with the protein surface, but at the same time increase the size of the water–vapor interface. Maintaining these surfaces is more energetically costly than water’s contact with hydrophobic protein (surface tension of air–water interface is 72 dyne/cm ¼ 17.7 kT/nm2). Calculation of the energy diVerence for the water–vapor (wwv) and water– protein (wwp) boundaries produces the interfacial free energy for dewetting: wdw ¼ wwv wwp. By finding the minima for wdw in each half of the pore (Fig. 2), the positions of the water–vapor boundaries were predicted.

FIGURE 2 Capillary dewetting and hydration properties of MscS pore. (A) The cross section ˚ ) colored according to of the central pore region with its solvent‐accessible surface (probe r ¼ 1.4 A the residue type: acidic (red ), basic (blue), polar (green), and nonpolar (white). Positions of residues defining the polarity of pore lining are indicated. Water molecules (cyan) do not occupy the constriction displaying characteristic dewetting. The coordinates are from the final frame of 6‐ns molecular dynamics simulation (Anishkin, Sukharev, 2004). (B) EVective pore radius R (gray curve) and free energy of dewetting (blue curve; computed with surface tension parameter ¼ 20 mJ/m2 suggested for confined water) as functions of z coordinate. The initiation of the vapor phase formation was inferred to occur in the narrowest part of the pore (dashed line).


247

Equilibrium MD simulations of the MscS pore filled with flexible TIP3P water confirmed that irrespective to the initial configuration (water–filled or empty pore), the hydrophobic constriction of the crystal structure stays dehydrated most of the time and that the boundaries reflect the predicted minimums of wdw (Fig. 2, blue line). The wider outer chamber of the pore also displayed a tendency to dewet when the electrostatics calculation protocol in simulations was changed from particle mesh Ewald (PME) to cutoV. Statistical treatment of trajectories showed not only lower density but also lower hydrogen bonding of water during infrequent visits into the constriction. Steered passages of Cl through the pore consistently produced partial dehydration of the ion and required a force of 200–400 pN to overcome an estimated barrier of 10–20 kcal/mole, implying negligibly low conductance. Importantly, MD simulations showed that the pore of the L109S GOF mutant (shown to have a low activation threshold and fast kinetics in experiments) was always fully hydrated. From these data it was concluded that the crystal structure of MscS does not represent an open state (Anishkin and Sukharev, 2004). 3. Expansion Is Required to Promote Pore Hydration and Conduction Dewetting transitions in the MscS pore were independently observed by Sotomayor and Schulten (2004). In their study, not a segment but the complete crystal structure was embedded in the fully hydrated lipid bilayer and simulated with and without applied membrane tension. The conformation with the protein backbone harmonically restrained near the crystal coordinates displayed a dehydrated constriction most of the time with intermittent water permeation events. Without restraints, the TM domain quickly collapsed in an asymmetric occluded conformation without water inside. Application of tension (20 dyne/cm) and release of restraints resulted in a gradual expansion of the pore accompanied with increased water occupancy. The eVective radius of the pore constriction during these simulations increased only minimally and events of ion permeation through the channel have not been observed. Specific interactions were noticed between charged residues residing at the ends of the TM helices and those on the cytoplasmic ‘‘cage’’ during evolution of the system. Not satisfied with the conductance estimations obtained in their first rounds of simulations, Sotomayor et al. (2006) turned to a Boltzmann Transport Monte‐Carlo method. This technique, utilizing coarse‐grained representation of molecules, is capable of simulating considerably longer trajectories up to 5 ms (van der Straaten et al., 2005). The restrained crystal structure again produced very low conductance. However, new expanded conformations, obtained through steered all‐atom simulations (with pore ˚ ), gave conductance estimations close to the constrictions as wide as 16 A

248

Sukharev et al.

experimentally observed 1 nS. The only inconsistencies in the simulations when compared to experiments were high anionic selectivity (PKþ =PCl 0:1 vs experimentally observed PKþ =PCl 0:7) and an opposite direction of rectification. One possible explanation is that in the simulations pore expansion was achieved by applying not only lateral tension to the entire simulation cell but also outward forces to the TM helices, the result was that only the constriction showed substantial movement. The charge‐carrying vestibules remained in the crystal‐like conformation, which would confer high Cl selectivity. This property was illustrated in analysis of the patterns of ion distributions and detailed maps of electrostatic potential across the entire protein, including the cage domain. The largest impediment to cation permeation appeared to be positive potential from the ring of Arginines 88 in the outer vestibule and from Lysines 169 inside the cage. The potential in the entire pore was very positive. The absence of the unresolved N‐terminal domains may have also contributed to the electrostatic imbalance since there are three additional acidic residues (E2, D3, and D8) that, in combination with the protonated N‐terminus, would result in two extra negative charges per subunit. Despite getting the constriction diameter right, the authors apparently did not achieve a true open‐state conformation for the more peripheral parts of the protein. Nevertheless, one interesting change in the inner pore vestibule was observed: in the course of steered pore expansion, the characteristic kinks at G113 showed a tendency to straighten, hinting at the possible conformation of the TM3 helices in the open state (Sotomayor et al., 2006). In the very latest assessment of the conducting properties of the MscS crystal structure, Vora et al. (2006) performed Brownian Dynamics simulations on both the whole MscS complex and the isolated TM region. The crystal conformation of the pore produced at best 30‐pS conductance. When ˚ , the authors observed an increase in the constriction was expanded by 2.5 A current corresponding to 0.2 nS. From this data, it was concluded once again that the crystal structure does not represent the open state of the channel. It is known that high‐electric fields promote formation of aqueous pores in low‐dielectric membranes (Chernomordik et al., 1987). Previous MD simulations indicated that high voltage applied across the membrane causes filling of preexisting hydrophobic pores with water (Dzubiella et al., 2004) and readily creates electropores in the lipid bilayer (Tieleman et al., 2003). An attempt to probe the crystal structure not by tension, but by high voltage, was published by the Dougherty group (Spronk et al., 2006). Performed with the Gromos force field and SPC water, these simulations well reproduced both the dewetting events and the collapse of an unrestrained TM domain into an asymmetric occluded state as previously observed in CHARMM simulations with a TIP3P water model (Sotomayor and Schulten, 2004). Application of TM voltages


249

ranging from 220 to 1100 mV produced a moderate expansion of the pore and higher water occupancy. Importantly, it was observed that either sharpening or straightening of the kink near G113 accompanied the both asymmetric collapse and sporadic outward movements of the pore‐forming TM3 helices. The authors also found that the distribution of charges in the peripheral TM domains (TM1 and TM2) strongly influence the occupancy of pore with water apparently through long‐distance electrostatic interactions. At low and moderate potentials, the pore was found impermeable to ions. Under extreme TM potentials (1100 mV), when the pore area expanded approximately two times comparing to the crystal conformation, ion permeation events were observed, with the frequency corresponding to 0.75‐nS conductance. An outward rectification manifested as higher rates of inward Cl permeation were observed under depolarizing conditions. In addition, the frequency of permeation events, defining the net current, was practically independent on the bulk concentration of carrier ions, which the authors’ interpreted as a diVusion‐limited regime. It was proposed that because the ‘‘hydrophobic lock’’ in the pore ‘‘disappears’’ with voltage, the crystal structure must resemble more the open state than the closed. These computational data, although interesting, predict a sharply nonlinear I–V curve with essentially zero conductance at low potentials, saturation in high salt, high anionic selectivity, and strong inward rectification. All of these findings occur in stark contrast to the entire experimental phenomenology known for the conductive state of MscS described below (Li et al., 2002; Sukharev, 2002; Akitake et al., 2005). It appears that none of the existing models of the open state exactly predict the conductive properties of MscS or account for the physical expansion of the protein in the plane of the membrane, as defined by the slope of MscS activation with tension, observed in experiments. In order to be ‘‘accurate,’’ any structural hypothesis on the functional states of the channel and/or the character of its transitions must take into account the body of experimental data regarding the conduction, activation and kinetic behavior of MscS which will be detailed below.

V. FUNCTIONAL PROPERTIES OF MscS A. MscS Conduction and Selectivity Despite its high conductance, which signifies substantial conformational changes in the multimeric protein, the onset of MscS conductance is extremely rapid and cooperative. An eVort by the Lester group that employed high‐bandwidth recording to resolve the time course of conductance change

250

Sukharev et al.

in MscS from the closed to the fully open level revealed that most of the opening and closing events are straight transitions faster than 3 ms (Shapovalov and Lester, 2004). The recording system detected occasional substates at 2/3 of the full‐open conductance. In the open state, the current–voltage (I–V) relationship of MscS is almost ideally linear near 0 mV with the conductance of the channel remaining constant up to 80 mV. At positive membrane potentials, MscS current decreases due to mild inward rectification. Beyond þ40 mV the presence of subconducting states were found to further reduce the eVective channel conductance. Independent measurements showed that under a threefold concentration gradient of KCl, the I–V curve of MscS shifts by 5–8 mV toward the reversal potential for Cl (Li et al., 2002; Sukharev, 2002). According to the Goldman equation, a 5‐mV shift under such conditions corresponds to PKþ =PCl of 0.68. Channel conductance was found nonsaturable up to 1.5‐M KCl. The strictly linear dependence of the unitary conductance on the specific conductivity of bathing electrolyte, with a space constant of 3.74 108 cm, strongly suggested bulk‐like conditions for ion movement in the water‐filled pore (Sukharev, 2002). MscS conduction in NaCl‐based buVers was found to be very similar to that in KCl (Akitake et al., 2005) and the specific permeability characteristics of MscS to other ions have not been studied. The conducting properties and selectivity of MscS in situ (in its native setting) are practically indistinguishable from those in reconstituted patches (Sukharev et al., 1993; Li et al., 2002; Sukharev, 2002). B. Gating Characteristics of MscS In Situ The study by Akitake et al. (2005) was the first to employ a high‐speed pressure clamp apparatus to deliver reproducible pressure stimuli to WT MscS populations. Expression of MscS in a background free of other mechanosensitive channels [MJF465; triple mscS‐, mscK‐, mscL‐knockout strain (Levina et al., 1999)] gave robust responses with up to 100 channels in an excised patch. WT MscS channels steeply activate with pressure and remain stably open in cases of constant saturating pressure. However, at subsaturating pressures MscS inactivates and is unable to respond to subsequent stimuli immediately. It takes about 3 min at zero tension and voltage for the population to completely return from the inactivated, back to the resting state. The open dwell time for WT MscS depends on tension, but near the activation threshold, the distribution peaks around 100–150 ms (Edwards et al., 2005; Akitake, unpublished data).


251

An analysis of MscS population responses to linear ramps of negative pressure, after rescaling pressure to tension [the midpoint pressure p1/2 corresponds to a membrane tension of 5.5 dyne/cm (Sukharev, 2002)], yielded the thermodynamic parameters of E ¼ 24 kT, A ¼ 18 nm2 as estimates of the energy and in‐plane area changes associated with the opening transition. These values, extracted from MscS activities in native membranes, were higher than those reported from studies in liposomes (E ¼ 11.4 kT and A ¼ 8.4 nm2). The larger A and E in the former case may be a result of the diVerent stimulus protocols applied. In the liposome‐ reconstitution experiments, dose‐response curves were measured using a series of descending steps (under more reversible conditions). It now appears that the ascending pressure ramp regime, even with slow ramps, may be diVerent from the descending regime. A comparison of WT MscS traces recorded with symmetric linear ascending and descending ramps of pressure showed clear hysteresis, indicating that the opening and closing transitions are characterized with diVerent slopes on tension and are likely to proceed through diVerent transition states (Akitake et al., unpublished data). Responses to just the descending ramps produced estimates for A between 9 and 12 nm2, more consistent with the previous data obtained in liposomes (Sukharev, 2002). Another factor that may aVect the apparent slope of Po() is the uniformity of channel population (Chiang et al., 2004). The higher slope and respectively larger apparent A suggest that the channel population in native patches is more uniform (in terms of E and/or local tension) compared to liposome patches. The nonhomogeneity of MscS channels in liposomes could potentially arise from lateral clustering. It also cannot be excluded that liposome and spheroplasts patches, when stressed with steady pressure gradients, exhibit diVerent TM distributions of lateral tension. Indeed, in liposome patches, tension may be more concentrated in the pipette‐attached monolayer due to unrestrained slippage and relaxation of the opposite monolayer in the low‐protein membrane. If MscS expansion is nonuniform along the z‐axis, redistribution of tension along the same axis would lead to changes in the apparent expansion area in the x–y plane. Currently there is no reason to believe that MscS in native membranes expands more, or gates very diVerently compared to liposomes. In sharp contrast to the early phenomenology (Martinac et al., 1987), it was found that MscS exhibits essentially voltage‐independent activation by tension. The responses of channel populations to linear pressure ramps at diVerent voltages (from þ100 to 100 mV) showed no significant shift in the pressure midpoint of activation. The slope of E plotted as a function of voltage, produced the gating charge of q ¼ þ0.8 e per channel complex or about þ0.1 e per subunit signifying that activation is not associated with any significant transfer of charge across the membrane.

252

Sukharev et al.

The process of MscS activation was found to be influenced by high‐ molecular‐weight compounds such as ficoll or polyethylene glycols (PEGs) present on the cytoplasmic side of the membrane (Grajkowski et al., 2005). The right‐shift of activation (Po—pressure) curves was more pronounced (30%) in the presence of a high‐molecular‐weight PEG (6.0 kDa), apparently impermeable through the fenestrations in the cytoplasmic cage, and a much smaller shift was observed with a smaller PEG (0.2 kDa). Both small and large PEGs were found to be inhibitory for MscS channel populations, though the exact mechanisms of their action appear to be diVerent.

C. Mutations That Affect MscS Activity The V40D/G41S mutation was first isolated in a random mutagenesis screen (Okada et al., 2002) as it resulted in a GOF phenotype inhibiting bacterial growth in liquid culture on induction. Further analysis revealed that V40D alone produced the GOF phenotype and since this hydrophobic‐ to‐charged substitution mapped to the first TM domain, it was inferred that TM1 might line the conducting pore of MscS (Okada et al., 2002). In later experiments, three growth‐suppressing mutations, previously identified in the mscK locus of Salmonella typhimurium, were replicated into MscS and found to confer visible GOF phenotypes (Miller et al., 2003a). Two of these mutants T93R and L109S were severe GOFs, whereas the third mutation, G108S, displayed a milder phenotype. Unlike the V40D mutation, these substitutions mapped to TM3, suggesting that TM3, not TM1, is the pore lining helix. Solution of the crystal structure in 2002 confirmed that TM3 lines the pore, with the T93 residue mapped to the short loop connecting TM2 with TM3 and the residues G108 and L109 right in the gate. The substitution L109S in the most severe GOF mutant was predicted to remove part of the hydrophobic lock that keeps MscS closed. Guided by the crystal structure, Booth and coworkers generated a number of substitutions, which hypothetically aVected the packing of TM3 helices in the pore of MscS. It was found that complimentary TM3–TM3 interfaces play a pivotal role in setting the threshold for MscS activation. Replacement of conserved glycines with alanines in this region (G101A, G104A, and G108A) produced MscS channels with higher activation thresholds, closer to that for MscL. These ‘‘stiVening’’ eVects were partially reversed by making reciprocal substitutions on the neighboring surface (A106G/G108A and A106G/G104A). The authors concluded that substitution of the glycines with alanines created artificial ‘‘knobs’’ on the surface of the mutated TM3s. These ‘‘knobs’’ sterically clash with the existing alanine ‘‘knobs’’ on


253

the opposite interhelical face, thus creating a higher barrier for the opening transition. While this mechanism is plausible, the authors did not consider the relationship between the polarity of substitutions and the activation thresholds, limiting discussion to the steric aspect of the interhelical interactions (Edwards et al., 2005). Their own analysis of G101S and G108S, hydrophilic mutations that should also produce steric conflicts, revealed channels that were easily opened. Conversely, hydrophobic and bulkier substitutions A98L, A106V, A106L produced extremely ‘‘stiV’’ channels. It now appears that hydration eVects have a stronger contribution to the activation threshold than the steric eVects (Anishkin and Sukharev, 2004; Akitake et al., unpublished data). D. MscS Inactivation 1. Parameters AVecting the Rate of Inactivation Inactivation is one of the most intriguing features of MscS gating. The eVects of MscS inactivation appear to be reversible in native membranes and largely irreversible in liposomes (Sukharev, 2002). Typical MscS activity in response to a fast pressure stimulus, which is then held at a constant nonsaturating level, is a spike of conductance followed by decay due to inactivation (Levina et al., 1999; Akitake et al., 2005). Koprowski and Kubalski (1998) were the first to characterize time‐dependent inactivation/ adaptation and recovery of MscS‐like channel populations. It was found that the inactivation/adaptation time constant was dependent on the magnitude of pressure application. It was also noted that adaptation was insensitive to voltage at moderate (30 mV) potentials (Koprowski and Kubalski, 1998). In a following study of the inactivation process, populations of MscS were stimulated with reproducible pulses of pressures under a wider (100 mV) range of voltages (Akitake et al., 2005). The data shows that low‐activating pressures promoted visible inactivation at hyperpolarizing (positive pipette) voltages, a process that was sped up by more than an order of magnitude under depolarizations beyond 40 mV. Higher depolarizing voltages were able to drive the MscS population to an inactivated state even when stimulated by previously saturating pressures. Furthermore, it was noticed that subjecting the channel population to depolarizing voltages increased the frequency of channel transitions to subconducting states (Akitake et al., 2005). High‐resolution recordings of MscS at high voltage identified several subconducting states (Shapovalov and Lester, 2004). These observations are consistent with the notion that substates are intermediates between the open and inactivated states and entering a substate predisposes MscS channels to inactivation.

254

Sukharev et al.

The involvement of the C‐terminal domain of MscS in the inactivation process is an area of active study. The crystal structure of MscS identified several charged residues along TM1, TM2, and TM3, which were designated as putative voltage sensors (Bass et al., 2002; Spronk et al., 2006). Nomura et al. (2005) mutated each of these residues and discovered that the voltage‐ sensitive characteristics of MscS inactivation were unchanged. Deletion studies have revealed that almost the entire C‐terminal domain of MscS is required for proper channel assembly (Miller et al., 2003a). Study was conducted on a mild C‐terminal truncation mutant (266–286) in which the sevenfold ‐barrel forming the base of the cytoplasmic cage was removed. This mutant was able to incorporate into the membrane but displayed greatly reduced activity and a striking inability to recover from inactivation (Schumann et al., 2004). The involvement of the C‐terminal domain in MscS inactivation and recovery was initially demonstrated by Koprowski and Kubalski (2003) in cross‐linking experiments targeting multiple lysines concentrated in cage region. Later, the same group (Grajkowski et al., 2005) noticed that large‐molecular‐weight cosolvents, when added from the cytoplasmic face, had a marked eVect on MscS inactivation. Data collected with PEGs of various sizes showed that smaller molecular weight PEGs, those that presumably penetrate the C‐terminal cage and make their way to the gate, tended to reduce channel conduction. However, larger PEGs did not appear to aVect conduction, rather their presence increased the rate of MscS inactivation. Interpreted as a ‘‘cosolvent’’ eVect, the inactivation‐promoting action of PEGs could be linked to pressures across the wall of the cage and structural changes in the C‐terminal domain of MscS (Grajkowski et al., 2005). These findings implicate a critical role of the C‐terminal in the process of inactivation. 2. MscS Activity Depends on the Rate of Pressure Application Unlike MscL, the activity of MscS is strongly dependent on the rate of stimulus application. MscS populations were stimulated with ramps of pressure that linearly increased from zero to a saturating value over varying time span (i.e., with rates from 2.7 to 240 mm Hg/s). It was found that the midpoint activation pressure remained fixed; however, it was obvious that the activity (maximal conductance) at the end of the ramp was higher with steeper ramps (Akitake et al., 2005). With the slowest ramp tested, only 30% of the MscS population was shown to activate. Inactivation may play a critical role in this eVect as reduction of maximal current was observed to be much stronger at depolarizing voltages (beyond 40 mV). To obtain further insight into this phenomenon, a two‐step protocol was employed in which a longer prepulse was delivered to trigger activation (with concomitant inactivation), and a subsequent shorter step of saturating pressure


255

applied to test the availability of the active population. Kinetics of the recorded currents indicated that inactivation was faster at intermediate pressures where only a portion of the MscS population activates (Akitake et al., 2005). This was consistent with earlier inactivation studies that showed a tension dependence of the rate of inactivation (Koprowski and Kubalski, 1998). It was found that the rate of inactivation was greatest in a narrow range of pressures just above the opening threshold and below the activation midpoint of the MscS population. When MscS is subjected to a slow ramp of pressure, the channels spend more time in this window of pressures that promote channel inactivation resulting in lower final conductance of the population. This interplay of opening and tension‐dependent inactivation permits MscS to respond in full to an abrupt stimulus and ignore stimuli applied slowly as if the gate of the channel is connected to the tension‐transmitting element eVectively via a velocity‐sensitive ‘‘dashpot’’ (Akitake et al., 2005). This functional design allows the channel to adapt to mild or slow‐onset osmotic shocks, when collapse of membrane potential and indiscriminate release of ions and small osmolytes may not be desirable. 3. Small Amphipathic Compounds Promote MscS Inactivation and Subunit Separation Early studies revealed strong eVects of small amphipathic molecules on the gating of mechanosensitive channels (Martinac et al., 1990). Cationic and anionic amphipathic compounds, such as chlorpromazine (CPZ) and trinitrophenol (TNP), when added to the bath solution in excised patch‐clamp recordings did not appear to bind to the channels directly, but lowered the activation threshold pressure for the population presumably due to curvature and/or area stress in the surrounding lipid bilayer (Sheetz and Singer, 1974; Markin and Martinac, 1991). Since then membrane‐active amphipathic substances, such as lysophosphatidylcholines (LPCs), have been used to shift the state equilibrium in MscL, especially when applied asymmetrically (Perozo et al., 2002). Short‐chain alcohols are diVerent from typical amphipaths as they are smaller and act on membranes in higher concentrations, as cosolvents (Barry and Gawrisch, 1994; Koenig and Gawrisch, 2005). 2,2,2‐Trifluoroethanol (TFE) is a low‐dielectric solvent that has known, since the 1960s, to strongly aVect protein structure. At high concentrations, TFE tends to stabilize the secondary structure of soluble proteins, whereas in membrane proteins the eVects are mainly destabilizing. The ‘‘inverted’’ design of integral membrane proteins (hydrophobic on periphery) when compared to soluble proteins (hydrophobic inside) appears to make them more susceptible to disruption by TFE. MscS was identified as one of the inner membrane proteins that

256

Sukharev et al.

changed its oligomerization state in two‐dimensional gels on exposure to TFE (Spelbrink et al., 2005). Isolated MscS oligomeric complexes were shown to remain stably associated as homoheptamers and homotetramers in cold lithium dodecyl sulfate polyacrylamide gel electrophoresis (LDS‐PAGE) (Akitake et al., 2007). The addition of 10 vol% TFE resulted in complete dissociation of MscS into monomeric subunits. Patch‐clamp analysis of MscS channels asymmetrically exposed to TFE from either the pipette or the bath showed shifts in the activation dose response curves to the left and right, respectively. These shifts are consistent with a buildup of lateral pressure due to intercalation of TFE into one leaflet of the membrane. In addition, it was found that TFE added to the bath causes reversible inactivation of the MscS population. The amount of inactivation was found to increase with greater concentrations of TFE. Fifty percent MscS inactivation was observed with a TFE concentration near 0.7 vol%. The shift in equilibrium toward the inactivated state correlated with the increased rate of MscS inactivation and slower recovery. In the presence of 0.5 vol% TFE, application of subsaturating stimulus revealed 2.6 times quicker inactivation and 4.2 times longer recovery compared to TFE‐free control preformed in the same patch under identical conditions (Akitake et al., 2007). It appears that both, the change in state distribution at low TFE concentrations and subunit separation at higher TFE, can be explained by the same thermodynamic property of TFE to separate helices by preferential partitioning into hydrophobic protein crevices. The membrane in this case may also act as a hydrophobic reservoir for TFE increasing its concentration several fold relative to the bulk. At these concentrations, TFE would prefer to partition into the easily formed crevices between TM2 and TM3 (as depicted in the crystal structure, Fig. 1). Crevice formation may shift the channel equilibrium toward the inactivated state consistent with the idea that separation of these helices is the mechanism of inactivation (Akitake et al., 2005, 2007).

VI. WHAT DO THE CLOSED, OPEN, AND INACTIVATED STATES OF MscS LOOK LIKE? Like any other protein crystal structure, the structure of MscS represents only one of the many physical conformations for the channel. In the absence of other direct structural data, visualization of MscS’s functional states (closed, open, and inactivated) can be obtained only through extensive modeling and simulations, judicious analysis of permitted conformations and interpolation of smooth transitional paths between them, and by critical


257

comparison of experimentally observed parameters. A ‘‘good’’ model must maintain consistency between each of these diVerent types of parameters. A. Is the Crystal Structure a Native State? The first model of MscS gating by Bass and colleagues postulated that the crystal conformation represents the conducting state of the channel with the characteristically tilted TM1–TM2 helices signifying the voltage‐driven transfer of charges previously, and erroneously, associated with MscS activation. It was also implied that the resting (nonconductive) state of the channel has less tilted peripheral helices and that packing of the central TM3 helices should be tighter. Computational data discussed above provides compelling evidence that the crystal structure of MscS is actually nonconductive under physiological voltages and thus cannot represent an open conformation. The crystal structure does not appear to represent a closed (resting) state either, as the tilted lipid‐facing TM1 and TM2 helices are splayed out and make no physical contacts with the gate‐forming TM3s. The absence of TM2–TM3 interactions makes it unclear how membrane stretch, received by the peripheral helices, could be transmitted to the gate. In this regard, the crystal conformation would appear to be irresponsive to membrane stretch, an attribute of the inactivated state. Likewise the diameter of the crystal pore does not satisfy the observed 1‐nS conductance for MscS (Anishkin and Sukharev, 2004; Sotomayor and Schulten, 2004; Sotomayor et al., 2006; Spronk et al., 2006; Vora et al., 2006), and the splaying motion of TM1 and TM2 would produce only small in‐plane protein expansion (2–3 nm2), far short of 8–12 nm2 deduced from the slope of MscS activation curves. It, therefore, seems plausible that the crystal conformation, with kinked TM3 helices, may resemble the inactivated MscS channel. One major problem with the crystal conformation is that the peripheral helices TM1 and TM2 do not form a continuous lipid‐facing wall, but were found to protrude outward, forming deep hydrophobic crevices, which must be filled with the detergent in the crystals. Tilting of individual TM helices in the bilayer is energetically unfavorable (de Planque et al., 2003; Strandberg et al., 2004) and the complete absence of tilt‐stabilizing helical contacts between the TM1–TM2 pairs suggests that the unusual splay could be an artifact of delipidation. As previously described, independent MD simulations revealed that when the channel is embedded into lipids in its crystal conformation, the pore quickly collapses (Sotomayor and Schulten, 2004; Spronk et al., 2006). In this regard, as a starting conformation, the crystal structure appears to be incompatible with a typical bilayer environment. Packing of the peripheral

258

Sukharev et al.

helices to be more parallel as proposed by Booth and coworkers (Miller et al., 2003b; Edwards et al., 2004) may alleviate the problem of instability; however, the sharp kink at G113 would still impose some gap between TM2 and TM3. This gap, generally hydrophobic (see Fig. 1 and legend) can be filled with small apolar substances such as TFE, which was shown to stabilize the inactivated state (Akitake et al., 2007). Therefore, a conformation similar to the crystal state, but with more parallel packing of helices and smaller gaps, is proposed for the conformation of the native inactivated state. B. Closed State To achieve the resting (closed) conformation, an even tighter packing of the TM helices from the crystal conformation would be required (Miller et al., 2003b; Edwards et al., 2004). This does not necessarily mean that the TM3 barrel should be packed much tighter as well. The already tight packing and limited mobility (low ‐factor) of the TM3 helices in the crystal structure is consistent with this conformation being close to an energetic minimum. If the crystal structure pore were truly nonconducting, additional packing of TM3 as proposed by Edwards et al., although possible, would not be required to close MscS. The model shown in Fig. 2 (middle) represents the results of steered simulation, in which the crystal structure (left) was subjected to uniform constricting forces. This compaction restored the TM2–TM3 contacts, but also caused straightening of the G113 kink to resolve steric conflict between the distal part of TM3 and the TM1–TM2 linker. An alternative kink was created at the highly conserved G121. Preliminary 6‐ns simulations have ˚ in diameter) remains shown that the constriction in this conformation (5 A dehydrated and thus nonconductive. The backbone of the pore in this ˚ ‘‘relaxed’’ model (residues 98–110) had RMSD deviation of only 0.73 A ˚ when compared to the crystal structure and 1.79 A compared to the model of Edwards et al. (2005). C. Open State What does it take to arrive at the open conformation of MscS? It is thought that tension applied to the peripheral helices (TM1 and TM2) is transmitted to the central pore‐lining helices (TM3) straightening them completely and producing a substantial outward expansion of the pore. This transition (Fig. 3) was achieved through a series of expansion‐minimization


259

FIGURE 3 The crystal conformation of the TM domain and models of the closed and open states of MscS. Molecular structures are in the ‘‘ribbon’’ representation, while L105 and L109 residues forming the constriction are shown in white VDW spheres, and R88 are presented as black sticks. The two front subunits are removed for a better view of the pore interior. The first 26 N‐terminal residues were unresolved in the crystal structure, are not shown in the models for consistency. Models were derived from the crystal conformation using steered molecular dynamics and targeted energy minimization in vacuum and in a fully hydrated POPC bilayer. To produce the closed state model, centripetal harmonic forces were applied to the transmembrane helices TM1 and TM2 of the crystal structure. The amplitude of the forces was gradually increased up to the minimal value when the helices moved and restored the contact with the TM3 barrel, while the secondary structure remained intact. The transition from the closed to the open state was initiated by the centrifugal forces applied to the whole transmembrane region. ˚ displacement, the expansion of the protein was continued by intermittent After the initial 1 A ˚ ) extrapolated displacements, and energy cycles of unrestrained relaxations, small (0.2 A minimizations. An open conformation satisfying the experimentally measured conductance and in‐plane expansion area was selected (shown as the open state model), embedded into a fully hydrated POPC bilayer and simulated without restraints for 4 ns. The simulation confirmed the stability of the model. (A. Anishkin et al., in preparation).

cycles, followed by a 4‐ns equilibrium simulation in the fully hydrated POPC bilayer (A. Anishkin, in preparation). The structure resolved in a stable tilted ˚ in conformation of TM3s with an aqueous central lumen of about 20 A diameter. It was observed that a slight rotation of TM3s resulted in the gate‐keeping leucines 105 and 109 swinging to the side, thus opening

260

Sukharev et al.

conducting pathway. At the same time, rotation of TM3 fully exposed the conserved glycines 101, 104, and 108 to the lumen, a feature that dramatically increased the polarity (and wettability) of the pore surface. Exposure of these glycine residues, which are firmly buried in interhelical contacts in the resting conformation, appears to stabilize the hydrated state of the open pore. This glycine‐mediated stabilization of the open conformation is highly consistent with the data presented by Edwards et al. (2005) who demonstrated that substitutions of these glycines to alanines (covering the polar backbone with apolar methyl) substantially increases the tension required for activation, whereas substitution with larger but more polar serines facilitates the opening transition. In this hypothetical open conformation, the conductance of the entire channel satisfied the observed 1 nS and the entire opening transition provided 12 nm2 of in‐plane protein expansion consistent with that derived from MscS activation curves. The outward movement of the TM helices, which displace the charged R88 residues farther away from the pore axis, is expected to produce a moderately anion‐selective channel. Transition from the open to the inactivated state can be envisioned as a breakage of the TM3 helices at G113. Kink formation at G113 closes the pore by bringing the TM3 segments almost parallel to each other. In this conformation, the now closely associated leucine 105 and 109 residues act to expel water and cease conduction. This transition may be accompanied by a partial disengagement of TM2 from TM3 retuning the system to a crystal‐ like conformation. Interestingly, the kink‐supporting cluster of residues, equivalent to G113‐S114 in MscS, is absent in MscK, which does not inactivate under similar stimuli (Levina et al., 1999).

VII. EMERGING PRINCIPLES OF MscS GATING AND REGULATION AND THE NEW DIRECTIONS On the basis of the experimental and computational data discussed above, it appears that the gating of MscS involves cooperative entry and retraction of water from the pore. In its closed state, the hydrophobic nature of MscS’s conduction path leads to a complete dewetting of the constriction (L105/ L109) and possibly the outer chamber lined by A102 and A98. The formation of this vapor plug would create a tight seal that is impermeable not only to all typical inorganic ions but to protons as well. This is a critical feature of a large conductance channel because any residual water bridge would invariably create a proton leak in the cytoplasmic membrane and decouple the bacterial energetics. The eVects of severe GOF mutations (L109S) are highly consistent with the permanently hydrated state of such pores making them leak.


261

A complete retraction of water from the pore cavity would also create meniscus‐like interfaces above and below the constriction. These high‐ tension interfaces would stabilize the tightly packed closed conformation. One challenging future project would be direct detection of the vapor plug in MscS possibly through the use of spectroscopic or particle scattering techniques. When tension is applied to MscS, the TM domain of the channel undergoes substantial expansion in the plane of the membrane (8–12 nm2), moving the protein–lipid boundary outward. Simultaneous widening of the pore increases the number of polar atoms exposed to the lumen and eVectively changes the character of pore lining from hydrophobic to hydrophilic. This change promotes hydration, which must precede the onset of conduction. It is important to remember that the changes of conduction from zero to the fully open level and then back to zero are sharp processes occurring in an all‐ or‐none fashion within 3 ms (Shapovalov and Lester, 2004). These opening and closing events are significantly shorter than what would be expected for the entire transition of the bulky MscS protein involving a lateral displacement of lipids and possible cytoplasmic cage transformation (Miller et al., 2003b; Edwards et al., 2004). Therefore, the character of conduction change in MscS may signify critical wetting–dewetting events involving primarily the dynamics of water (not the entire protein), which in simulations occur in 1–3 ns for the preexpanded protein. An important next step in clarifying the opening and closing conformational pathways would be to carefully study the hysteresis between the two events and determine the positions of the transition states. The characteristic kink found in the crystal structure near G113 appears to be a signature of the inactivated MscS. In contrast, the noninactivating channel MscK shows strong helical propensity at this location. Dependence of the inactivation/recovery processes in MscS to the relative helical propensity of TM3 is currently under investigation. The MscS channel displays the unique ability to undergo time‐ and pressure‐dependent desensitization/ inactivation, with the channel population responding in full to an abrupt stimulus and not to those applied slowly (Akitake et al., 2005). This kinetic property of ‘‘dumping’’ a slowly applied force may be important in diVerent environmental situations where dilution of the external medium occurs gradually and thus fast and indiscriminate release of osmolytes is not desirable. Turning oV MscS activity under such conditions provides an opportunity for more specific transporters/exchangers to balance osmotic forces in a more selective way. One important experiment will be to compare the cellular responses of bacteria to acute and gradual osmotic changes in strains where MscS is the sole osmolyte release valve. It is possible that the presence or absence of aquaporins, which regulate the rate of water transport and

262

Sukharev et al.

turgor pressure onset (Booth and Louis, 1999), may influence the activity of MscS. The structural design of MscS is such that the gate‐forming TM3 helix is directly connected to a large C‐terminal cage domain. Deciphering the interplay between the opening and inactivation processes with the volume and/or conformation of this hollow domain is an attractive idea. It appears that the cage may function as an internal osmosensor that discriminates between the osmotic pressure created by small (permeable) osmolytes from the oncotic pressure created by large macromolecular constituents. If a substance easily penetrates the fenestrations in the cage domain, it will not contribute to a pressure gradient across the cage wall. One example is when the turgor pressure inside the cell is created primarily by inorganic ions such as Kþ during the first stage of the cellular response to hyperosmotic stress. The cage will not experience any extra pressure and the channel will open to release these extra ionic constituents and balance the osmotic gradient. However, larger macromolecules that are cage‐impermeable are also too large to be expelled by MscS. If these molecules are responsible for the cellular turgor pressure, their exclusion from the cage will exert a secondary osmotic imbalance on the C‐terminal domain, causing its collapse. This mechanism appears to prevent futile openings that would only destroy the membrane potential and normal ionic gradients. The cage of MscS senses the ‘‘squeezing’’ action of these high‐molecular‐weight osmolytes and renders the channel insensitive to membrane tension by driving the gate into an inactivated conformation. Data presented by the Kubalski group (Grajkowski et al., 2005) is highly consistent with this hypothesis. Further testing of MscS responses in the presence of natural cytoplasmic constituents, other high‐molecular‐weight compounds and compatible osmolytes is being conducted. In addition to being an internal osmosensor, the cage domain of MscS may also act as a voltage or ionic flux sensor. The report by Nomura et al. (2006) has demonstrated that the voltage dependence of inactivation remains even when most of the charges in the TM domain are neutralized by mutations. It is important to note that voltage‐dependent inactivation is observed to develop only after the channel opens. It appears that high depolarizing potentials are unable to drive the closed channel into the inactivated state. These observations suggest that the ionic flux itself, and/or the accompanying elecroosmotic water flux, drives MscS into the inactivated state. Careful testing of this hypothesis is certainly required; however, preliminary data show a strong dependence of inactivation by voltage on the type of permeant ions (Akitake, unpublished data). This is consistent with the idea that the voltage sensor in MscS is not located in the membrane where it would perceive static voltage, but is more likely located intracellularly where it perceives


263

unbalanced ion fluxes causing coupled water fluxes and integrates this perturbation with the osmotic contributions of diVerent solutes. It will be very interesting to see if channel activity is influenced by proline, betaine, and glutamate and ultimately whether MscS is part of the system that exchanges Kþ for compatible osmolytes during the upshock response (Wood et al., 2001; Booth et al., 2003). Recent studies of MscS homologues have revealed interesting clues about the role of the small mechanosensitive channel in higher organisms. Do all MscS‐type channels function in the same way? The answer is apparently not. Even the related bacterial channel MscK, which features activities similar to MscS in patch‐clamp experiments, cannot by itself rescue MJF465 cells from osmotic shock. On the other hand, the recently characterized MscS homologue from A. thaliana, MSL3, when expressed in E. coli, is capable of restoring osmoprotective function (Haswell and Meyerowitz, 2006). There are yet no reports of any electrophysiological data for MSL3, and it is possible that MSL3 has a smaller conductance that is not readily seen in patch clamp. MSL3 was demonstrated to participate in chloroplast shape formation and division. Deletion of MSL2 and MSL3 resulted in chloroplasts that were much larger and swollen. A failure of this type of regulation suggests that MSLs indeed act as tension‐activated release or exchange valves; however, to accomplish this function the rate of transport through these channels need not be as high as that of MscS (109 ions per second). From analysis of the sequence, A. thaliana MSLs have longer hydrophobic constrictions (three rings of bulky aliphatic residues instead of two in MscS), and this may critically change the rate of ion transport. The presence of these bacterial‐type channels in A. thaliana and their role in chloroplasts volume regulation is a vivid illustration of how a function of initially bacterial channels was preserved and adapted in the course of endosymbiotic evolution of plastids (Haswell and Meyerowitz, 2006). MscS is undoubtedly a high‐value system for experimentation and molecular computation that will aid in the discovery of new biophysical principles of ion channel regulation. It is hopeful that the continued study of MscS will eventually link its unique and complex phenomenology to a greater understanding of the mechanisms of cellular/organellar growth, development, and flexible adaptation to various environments. References Akitake, B., Anishkin, A., and Sukharev, S. (2005). The ‘‘dashpot’’ mechanism of stretch‐ dependent gating in MscS. J. Gen. Physiol. 125, 143–154. Akitake, B., Spelbrink, R. E., Anishkin, A., Killian, J. A., de Kruijff, B., and Sukharev, S. (2007). 2,2,2‐Trifluoroethanol changes the transition kinetics and subunit interactions in the small bacterial mechanosensitive channel MscS. Biophys. J. (in press).

264

Sukharev et al.

Anishkin, A., and Sukharev, S. (2004). Water dynamics and dewetting transitions in the small mechanosensitive channel MscS. Biophysics 86, 2883–2895. Barry, J. A., and Gawrisch, K. (1994). Direct NMR evidence for ethanol binding to the lipid‐ water interface of phospholipid bilayers. Biochemistry 33, 8082–8088. Bass, R. B., Strop, P., Barclay, M., and Rees, D. C. (2002). Crystal structure of Escherichia coli MscS, a voltage‐modulated and mechanosensitive channel. Science 298, 1582–1587. Beckstein, O., and Sansom, M. S. (2003). Liquid‐vapor oscillations of water in hydrophobic nanopores. Proc. Natl. Acad. Sci. USA 100, 7063–7068. Beckstein, O., Biggin, P. C., and Sansom, M. S. P. (2001). A hydrophobic gating mechanism for nanopores. J. Phys. Chem. B 105, 12902–12905. Berrier, C., Besnard, M., Ajouz, B., Coulombe, A., and Ghazi, A. (1996). Multiple mechanosensitive ion channels from Escherichia coli, activated at diVerent thresholds of applied pressure. J. Membr. Biol. 151, 175–187. Blount, P., Sukharev, S. I., Moe, P. C., Martinac, B., and Kung, C. (1999). Mechanosensitive channels of bacteria. Methods Enzymol. 294, 458–482. Booth, I. R., and Louis, P. (1999). Managing hypoosmotic stress: Aquaporins and mechanosensitive channels in Escherichia coli. Curr. Opin. Microbiol. 2, 166–169. Booth, I. R., Edwards, M. D., and Miller, S. (2003). Bacterial ion channels. Biochemistry 42, 10045–10053. Britten, R. J., and McClure, F. T. (1962). The amino acid pool in Escherichia coli. Bacteriol. Rev. 26, 292–335. Chang, G., Spencer, R. H., Lee, A. T., Barclay, M. T., and Rees, D. C. (1998). Structure of the MscL homolog from Mycobacterium tuberculosis: A gated mechanosensitive ion channel. Science 282, 2220–2226. Chernomordik, L. V., Sukharev, S. I., Popov, S. V., Pastushenko, V. F., Sokirko, A. V., Abidor, I. G., and Chizmadzhev, Y. A. (1987). The electrical breakdown of cell and lipid‐ membranes—the similarity of phenomenologies. Biochim. Biophys. Acta 902, 360–373. Chiang, C. S., Anishkin, A., and Sukharev, S. (2004). Gating of the large mechanosensitive channel in situ: Estimation of the spatial scale of the transition from channel population responses. Biophysics 86, 2846–2861. Csonka, L. N. (1989). Physiological and genetic responses of bacteria to osmotic stress. Microbiol. Rev. 53, 121–147. Csonka, L. N., and Hanson, A. D. (1991). Prokaryotic osmoregulation: Genetics and physiology. Annu. Rev. Microbiol. 45, 569–606. Cui, C., and Adler, J. (1996). EVect of mutation of potassium‐eZux system, KefA, on mechanosensitive channels in the cytoplasmic membrane of Escherichia coli. J. Membr. Biol. 150, 143–152. Cui, C., Smith, D. O., and Adler, J. (1995). Characterization of mechanosensitive channels in Escherichia coli cytoplasmic membrane by whole‐cell patch clamp recording. J. Membr. Biol. 144, 31–42. de Planque, M. R., Bonev, B. B., Demmers, J. A., Greathouse, D. V., Koeppe, R. E., Separovic, F., Watts, A., and Killian, J. A. (2003). Interfacial anchor properties of tryptophan residues in transmembrane peptides can dominate over hydrophobic matching eVects in peptide‐lipid interactions. Biochemistry 42, 5341–5348. Dzubiella, J., Allen, R. J., and Hansen, J. P. (2004). Electric field‐controlled water permeation coupled to ion transport through a nanopore. J. Chem. Phys. 120, 5001–5004. Edwards, M. D., Booth, I. R., and Miller, S. (2004). Gating the bacterial mechanosensitive channels: MscS a new paradigm? Curr. Opin. Microbiol. 7, 163–167.


265

Edwards, M. D., Li, Y., Kim, S., Miller, S., Bartlett, W., Black, S., Dennison, S., Iscla, I., Blount, P., Bowie, J. U., and Booth, I. R. (2005). Pivotal role of the glycine‐rich TM3 helix in gating the MscS mechanosensitive channel. Nat. Struct. Mol. Biol. 12, 113–119. Epstein, W. (2003). The roles and regulation of potassium in bacteria. Prog. Nucleic Acid Res. Mol. Biol. 75, 293–320. Ferguson, G. P., Battista, J. R., Lee, A. T., and Booth, I. R. (2000). Protection of the DNA during the exposure of Escherichia coli cells to a toxic metabolite: The role of the KefB and KefC potassium channels. Mol. Microbiol. 35, 113–122. Fraczkiewicz, R., and Braun, W. (1998). Exact and eYcient analytical calculation of the accessible surface areas and their gradients for macromolecules. J. Comput. Chem. 19, 319–333. Grajkowski, W., Kubalski, A., and Koprowski, P. (2005). Surface changes of the mechanosensitive channel MscS upon its activation, inactivation, and closing. Biophysics 88, 3050–3059. Hall, J. E. (1975). Access resistance of a small circular pore. J. Gen. Physiol. 66, 531–532. Hamill, O. P., Marty, A., Neher, E., Sakmann, B., and Sigworth, F. J. (1981). Improved patch‐ clamp techniques for high‐resolution current recording from cells and cell‐free membrane patches. Pflugers Arch. 391, 85–100. Haswell, E. S., and Meyerowitz, E. M. (2006). MscS‐like proteins control plastid size and shape in Arabidopsis thaliana. Curr. Biol. 16, 1–11. Hille, B. (1992). ‘‘Ionic Channels of Excitable Membranes.’’ Sinauer Asssociates Inc., Sunderland, MA. Hummer, G., Rasaiah, J. C., and Noworyta, J. P. (2001). Water conduction through the hydrophobic channel of a carbon nanotube. Nature 414, 188–190. Koch, A. L., and Woeste, S. (1992). Elasticity of the sacculus of Escherichia coli. J. Bacteriol. 174, 4811–4819. Koenig, B. W., and Gawrisch, K. (2005). Lipid‐ethanol interaction studied by NMR on bicelles. J. Phys. Chem. B 109, 7540–7547. Koprowski, P., and Kubalski, A. (1998). Voltage‐independent adaptation of mechanosensitive channels in Escherichia coli protoplasts. J. Membr. Biol. 164, 253–262. Koprowski, P., and Kubalski, A. (2003). C termini of the Escherichia coli mechanosensitive ion channel (MscS) move apart upon the channel opening. J. Biol. Chem. 278, 11237–11245. Levina, N., Totemeyer, S., Stokes, N. R., Louis, P., Jones, M. A., and Booth, I. R. (1999). Protection of Escherichia coli cells against extreme turgor by activation of MscS and MscL mechanosensitive channels: Identification of genes required for MscS activity. EMBO J. 18, 1730–1737. Li, Y., Moe, P. C., Chandrasekaran, S., Booth, I. R., and Blount, P. (2002). Ionic regulation of MscK, a mechanosensitive channel from Escherichia coli. EMBO J. 21, 5323–5330. Markin, V. S., and Martinac, B. (1991). Mechanosensitive ion channels as reporters of bilayer expansion. A theoretical model. Biophysics 60, 1120–1127. Martinac, B. (2001). Mechanosensitive channels in prokaryotes. Cell Physiol. Biochem. 11, 61–76. Martinac, B., Buechner, M., Delcour, A. H., Adler, J., and Kung, C. (1987). Pressure‐sensitive ion channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 84, 2297–2301. Martinac, B., Adler, J., and Kung, C. (1990). Mechanosensitive ion channels of, E. coli activated by amphipaths. Nature 348, 261–263. McLaggan, D., Naprstek, J., Buurman, E. T., and Epstein, W. (1994). Interdependence of K þ and glutamate accumulation during osmotic adaptation of Escherichia coli. J. Biol. Chem. 269, 1911–1917.

266

Sukharev et al.

McLaggan, D., Jones, M. A., Gouesbet, G., Levina, N., Lindey, S., Epstein, W., and Booth, I. R. (2002). Analysis of the kefA2 mutation suggests that KefA is a cation‐specific channel involved in osmotic adaptation in Escherichia coli. Mol. Microbiol. 43, 521–536. Miller, S., Bartlett, W., Chandrasekaran, S., Simpson, S., Edwards, M., and Booth, I. R. (2003a). Domain organization of the MscS mechanosensitive channel of Escherichia coli. EMBO J. 22, 36–46. Miller, S., Edwards, M. D., Ozdemir, C., and Booth, I. R. (2003b). The closed structure of the MscS mechanosensitive channel. Cross‐linking of single cysteine mutants. J. Biol. Chem. 278, 32246–32250. Nomura, T., Yoshimura, K., and Sokabe, M. (2005). Exploring the lipid‐protein interface essential to the mechanosensitivity of MscS, a mechanosensitive channel from, E. coli. Biophys. J. 88, 290A. Nomura, T., Yoshimura, K., and Sokabe, M. (2006). Voltage dependence of the adaptation in mscS occurs independent of the charged residues in the transmembrane domain. Biophys. J. 90, XXA. Okada, K., Moe, P. C., and Blount, P. (2002). Functional design of bacterial mechanosensitive channels. Comparisons and contrasts illuminated by random mutagenesis. J. Biol. Chem. 277, 27682–27688. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2002). Physical principles underlying the transduction of bilayer deformation forces during mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703. Pivetti, C. D., Yen, M. R., Miller, S., Busch, W., Tseng, Y. H., Booth, I. R., and Saier, M. H., Jr. (2003). Two families of mechanosensitive channel proteins. Microbiol. Mol. Biol. Rev. 67, 66–85, table. Ruthe, H. J., and Adler, J. (1985). Fusion of bacterial spheroplasts by electric fields. Biochim. Biophys. Acta 819, 105–113. Sachs, F., and Morris, C. E. (1998). Mechanosensitive ion channels in nonspecialized cells. Rev. Physiol. Biochem. Pharmacol. 132, 1–77. Schleyer, M., Schmid, R., and Bakker, E. P. (1993). Transient, specific and extremely rapid release of osmolytes from growing cells of Escherichia coli K‐12 exposed to hypoosmotic shock. Arch. Microbiol. 160, 424–431. Schumann, U., Edwards, M. D., Li, C., and Booth, I. R. (2004). The conserved carboxy‐ terminus of the MscS mechanosensitive channel is not essential but increases stability and activity. FEBS Lett. 572, 233–237. Shapovalov, G., and Lester, H. A. (2004). Gating transitions in bacterial ion channels measured at 3 micros resolution. J. Gen. Physiol. 124, 151–161. Sheetz, M. P., and Singer, S. J. (1974). Biological membranes as bilayer couples. A molecular mechanism of drug‐erythrocyte interactions. Proc. Natl. Acad. Sci. USA 71, 4457–4461. Smart, O. S., Neduvelil, J. G., Wang, X., Wallace, B. A., and Sansom, M. S. (1996). HOLE: A program for the analysis of the pore dimensions of ion channel structural models. J. Mol. Graph. 14, 354–360, 376. Sotomayor, M., and Schulten, K. (2004). Molecular dynamics study of gating in the mechanosensitive channel of small conductance MscS. Biophysics 87, 3050–3065. Sotomayor, M., van der Straaten, T. A., Ravaioli, U., and Schulten, K. (2006). Electrostatic properties of the mechanosensitive channel of small conductance MscS. Biophysics 90, 3496–3510. Spelbrink, R. E., Kolkman, A., Slijper, M., Killian, J. A., and de, K. B. (2005). Detection and identification of stable oligomeric protein complexes in Escherichi coli inner membranes: A proteomics approach. J. Biol. Chem. 280, 28742–28748.


267

Spronk, S. A., Elmore, D. E., and Dougherty, D. A. (2006). Voltage‐dependent hydration and conduction properties of the hydrophobic pore of the mechanosensitive channel of small conductance. Biophysics 90, 3555–3569. Stewart, G. R., Patel, J., Robertson, B. D., Rae, A., and Young, D. B. (2005). Mycobacterial mutants with defective control of phagosomal acidification. PLoS Pathog. 1, 269–278. Strandberg, E., Ozdirekcan, S., Rijkers, D. T., van der Wel, P. C., Koeppe, R. E., Liskamp, R. M., and Killian, J. A. (2004). Tilt angles of transmembrane model peptides in oriented and non‐oriented lipid bilayers as determined by 2H solid‐state NMR. Biophysics 86, 3709–3721. Strop, P., Bass, R., and Rees, D. C. (2003). Prokaryotic mechanosensitive channels. Adv. Protein Chem. 63, 177–209. Sukharev, S. (2002). Purification of the small mechanosensitive channel of Escherichia coli (MscS): The subunit structure, conduction, and gating characteristics in liposomes. Biophysics 83, 290–298. Sukharev, S. I., Martinac, B., Arshavsky, V. Y., and Kung, C. (1993). Two types of mechanosensitive channels in the Escherichia coli cell envelope: Solubilization and functional reconstitution. Biophysics 65, 177–183. Sukharev, S. I., Blount, P., Martinac, B., Blattner, F. R., and Kung, C. (1994). A large‐conductance mechanosensitive channel in, E. coli encoded by mscL alone. Nature 368, 265–268. Tieleman, D. P., Leontiadou, H., Mark, A. E., and Marrink, S. J. (2003). Simulation of pore formation in lipid bilayers by mechanical stress and electric fields. J. Am. Chem. Soc. 125, 6382–6383. Tsapis, A., and Kepes, A. (1977). Transient breakdown of the permeability barrier of the membrane of Escherichia coli upon hypoosmotic shock. Biochim. Biophys. Acta 469, 1–12. van der Straaten, T. A., Kathawala, G., Trellakis, A., Eisenberg, R. S., and Ravaioli, U. (2005). BioMOCA—a Boltzmann transport Monte Carlo model for ion channel simulation. Mol. Simul. 31, 151–171. Vasquez, V., and Perozo, E. (2004). Voltage dependent gating in MscS. Biophys. J. 86, 545A. Vora, T., Corry, B., and Chung, S. H. (2006). Brownian dynamics investigation into the conductance state of the MscS channel crystal structure. Biochim. Biophys. Acta. 1758, 730–737. Wesson, L., and Eisenberg, D. (1992). Atomic solvation parameters applied to molecular dynamics of proteins in solution. Protein Sci. 1, 227–235. Wood, J. M. (1999). Osmosensing by bacteria: Signals and membrane‐based sensors. Microbiol. Mol. Biol. Rev. 63, 230–262. Wood, J. M., Bremer, E., Csonka, L. N., Kraemer, R., Poolman, B., van der, H. T., and Smith, L. T. (2001). Osmosensing and osmoregulatory compatible solute accumulation by bacteria. Comp Biochem. Physiol. A Mol. Integr. Physiol. 130, 437–460.

CHAPTER 10 Structure–Function Relations of MscS Ian R. Booth, Michelle D. Edwards, Samantha Miller, Chan Li, Susan Black, Wendy Bartlett, and Ulrike Schumann School of Medical Sciences, Institute of Medical Sciences, University of Aberdeen, Aberdeen AB25 2ZD, United Kingdom

I. Overview II. Introduction A. Functional Overview III. The Structure of MscS A. The Membrane Domain B. The Cytoplasmic Domain C. Variations in Structure D. Twisting MscS Around the Pore E. MscS Is Small but Beautifully Formed IV. MscS Mutational Analysis V. Structural Transitions in MscS A. The Need for the Closed State B. The Crystal State C. The TM3 Pore D. The Closed‐to‐Open Transition VI. Conclusions and Future Perspective References

I. OVERVIEW The mechanosensitive (MS) channel MscS is the more widespread of two major MS channels that have been characterized. MscS‐like proteins have been discovered in bacteria, archaea, yeasts and fungi, and in plants. In most organisms, multiple homologues have been found, although few have been characterized in detail. In Escherichia coli, where most work has been carried out, there are three small MscS homologues of 286 (MscS, YggB), 343 (YnaI), and 415 (YbdG) amino acids. In addition, there are three members Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58010-X

270

Booth et al.

of an MscS subfamily, typified by MscK (KefA) in E. coli, which are restricted to Gram‐negative bacteria and which feature a more complex organization in the membrane. These proteins are between 741 (YbiO) and 1120 (MscK) residues in length and possess both a large periplasmic domain and additional membrane domains N‐terminal to their ‘‘MscS channel’’ domain. Of the six E. coli proteins of the MscS family discovered to date, most is known about MscS and MscK. In this chapter, we will present current views on the function, expression, structure, and mechanism of the MscS proteins, making reference to MscK where appropriate.

II. INTRODUCTION MS channels in bacterial cells fall into two major categories defined by their core structures: MscL and MscS (Chang et al., 1998; Bass et al., 2002). Both have now been studied extensively using molecular genetics allied to electrophysiology and protein biochemistry. Both channel classes are widespread among bacteria and archaea, and there are also examples found in fungi and plants (Pivetti et al., 2003; Haswell and Meyerowitz, 2006). Their role in cell physiology is generally agreed, viz., to facilitate the rapid release of solutes in a nondiscriminating manner, such that cytoplasmic turgor is diminished (Levina et al., 1999). It is frequently observed that bacteria possess examples of both MscS and MscL types and that they are generally functionally redundant (Pivetti et al., 2003). However, the number of organisms in which the roles of the channels has been rigorously tested is limited. The analysis is complicated by the presence of multiple homologues, usually of the MscS class, but occasionally also MscL. Even in E. coli their role has only been investigated in laboratory isolates of E. coli K‐12, where the wall has been weakened by the loss of lipopolysaccharide (LPS). For both channels, there is an emerging consensus on the structural transitions that they undergo during the opening process. It is given that the channels are closed in the growing bacterial cell and that they undergo ˚ rapid structural transitions that create transient large pores 8‐ to 30‐A diameter. Bacterial cells rely on a selectively permeable membrane to maintain cytoplasmic homeostasis and to interconvert energy via ion gradients (Booth, 1985). The opening of MS channels subverts both of these processes by dispelling ion gradients, lowering the membrane potential, and allowing the nonselective movement of solutes. The lack of selectivity is a function of the large pore diameter that readily allows the passage of solutes probably without the loss of their hydration shell. In the case of the well‐characterized Naþ, Kþ, and Cl channels, ionic specificity is largely attained through the

271

10. MscS Mechanosensitive Channel

dehydration of the ion and the replacement of the coordinating water molecules by O and N atoms of the peptide backbone of the channel (Doyle et al., 1998; Zhou et al., 2001; Dutzler et al., 2002). The loss of specificity associated with MS channels is not simply a function of their pore size, since at least MscS homologues exhibit ion selectivity, even if at a more modest scale than observed for classical ion‐selective channels (Martinac et al., 1987; Li et al., 2002). The loss of selectivity may be oVset by the potentially higher rates of ion permeation, since although some Kþ channels exhibit ion conduction at rates close to the rate of free diVusion, there is a huge dynamic range in the observed conductances. MS channels in E. coli range in conductance from 0.3 to 3 nS and in patch‐clamp analyses they are activated by applied transmembrane (TM) pressure in order of their conductance: MscM, first, followed by MscK, MscS, and finally MscL (Table I; Fig. 1) (Sukharev et al., 1993; Batiza et al., 2002; Kung and Blount, 2004). The structural genes for MscS, MscK, and MscL are known, whereas the gene for MscM remains to be discovered. MscS and MscK are structurally related (see below) but diVer in their properties (Levina et al., 1999). In E. coli, MscS is an abundant channel activity, whereas MscK is less readily observed in membrane patches.

TABLE I MscS Homologues in E. coli Activitya

Expressionb

Protein

Gene

Size (amino acids)

MscS

yggB

286

Yes (M) (P)

F343

ynaI

343

No (P)

ND

YbdG

ybdG

415

Some (P)

Yes

MscK

kefA

1120c

Yes (Kþ)d (M) (P)

LeuO

yjeP

c

1101

NDe

ND

ybiO

c

ND

s38

YjeP F786

741

s70, s38

a Activity can be defined either in terms of activity in membrane patches determined by electrophysiology (M) or by protection (P) aVorded against hypoosmotic shock in the triple channel‐deficient mutant, MJF465. b Regulation of expression of the gene is reported: s70 refers to the vegetative sigma factor; s38 refers to the stationary phase sigma factor that is also expressed in response to osmotic stress in E. coli; LeuO is a general regulatory protein for which the mechanism is not fully understood. c The size includes the signal sequence required for export of the N‐terminal domain to the periplasm. d The MscK channel is dependent on high Kþ concentrations on the periplasmic side for activity. e Although not detected in E. coli, the growth of YjeP insertion mutants (i.e., null mutants) in Erwinia chrysanthemi is inhibited in media containing high Kþ and betaine or proline. In addition to MscS, five MscS‐related proteins are found in E. coli. This table summarizes the limited information available for each.

Booth et al.

272

YbdG Ynal 4.5 0

MscS

−4.5 YbiO YjeP SS

Periplasmic domain

Inner membrane

CTD

4.5

0

MscK

−4.5 0

100

200

300

400

500

600

700

800

900

1000 1100

FIGURE 1 The MscS/MscK family in E. coli. Hydrophobicity plots for MscS (upper) and MscK (lower) are depicted using a window of 19 residues to define average hydrophibicity at each position using the Protean program (DNAstar). The domain organization of MscK is indicated; note that there are ‘‘two’’ membrane domains—eight helices that lie N‐terminal to the ‘‘MscS’’ domain (gray) and the three helices of the ‘‘MscS’’ domain itself. Above each hydrophobicity plot a bar is presented depicting the length of the homologues. For YbiO a gap representing the in‐frame deletion in the periplasmic domain that has led to the shorter version of this protein is indicated by a broken line. For YbdG, the insertion that has arisen at the junction between the b‐ and the ab‐domains is indicated by an open bar connected to the main bar (filled) at the position of the insertion.

MscK opens at pressures just below those needed to activate MscS. MscK also requires Kþ at the periplasmic face for activation and is relatively nonselective for ions (Li et al., 2002), contrasting with MscS, which is variably ion selective (anions in E. coli and cations in Methanococcus jannaschii) (Martinac et al., 1987; Kloda and Martinac, 2001a,b). Correspondingly, the conductance of MscK is 0.9 nS and that of MscS is 1.2 nS (in 0.2‐M KCl in the bath and pipette). The two channels diVer in that MscS, but not MscK, inactivates under sustained pressure and can only be recovered by resting the membrane patch (Levina et al., 1999). However, in cells MscK appears to have only a minor role in relief of excessive turgor, whereas MscS plays a dominant role in this relief from stress.


273

A. Functional Overview MS channels are maintained in the closed state by balanced lateral pressure within the lipid bilayer that prevents the channel protein from expanding to the open state (Perozo et al., 2002a,b). Artificial activation can be achieved by diVerential intercalation of small amphipaths into one leaflet of the bilayer such that a pressure diVerential exists between the two halves (Martinac et al., 1990). However, it is generally accepted that in cells MS channels gate in response to pressure diVerentials across the membrane that cause distortion of the bilayer such that cell damage that would result from excessive turgor is avoided. Their function is to release solutes from the cytoplasm and thereby to diminish the TM pressure associated with water influx down the osmotic gradient (Berrier et al., 1992, 1996). Generally, bacteria accumulate solutes to concentrations much higher than the environment leading to water influx and this generates an outwardly directed turgor pressure (Booth et al., 1988). Although precise measurements of turgor pressure are still lacking, the earliest estimates for E. coli and Staphylococcus aureus of 4 and 20 atm, respectively, remain valid working assumptions (Booth et al., 1988). Analysis of MS channels usually takes place by electrophysiology in isolated membrane patches not protected by the peptidoglycan cell wall (Martinac et al., 1987). Here the most sensitive E. coli MS channels, MscM, MscK, and MscS, are generally active at 0.1‐atm pressure applied across the membrane. Clearly the outward turgor pressure in cells is much greater than this value and must be balanced by the resistance of the cell wall (peptidoglycan and LPS, in the case of Gram‐negative bacteria; peptidoglycan and lipoteichoic acids in Gram‐positive organisms) to maintain the channels closed. Under physiological conditions, MS channel activation occurs in response to a decrease in external osmolarity, which results in an immediate large increase in turgor pressure. Transfer from high osmolarity to low can generate an increase in turgor of up to 10 atm in a few milliseconds, as water rushes into the cell. In these circumstances, the change in the contact between the inner membrane and the peptidoglycan results in membrane distortions suYcient to activate the channels. As indicated above, channels can also be activated by amphipaths that intercalate into the membrane with a slow transfer time between the outer leaflet and the inner leaflet of the bilayer (Martinac et al., 1987; Perozo et al., 2002b). Such molecular properties could generate transient diVerences in lateral pressure in the two leaflets leading to distortion‐led channel activation. Conceivably, a wide range of molecules could have transient eVects on channel gating, for example fermentation products, antibiotics, fatty acids, and so on. Food preservatives, for example parabens (Nguyen et al., 2005) and weak organic acids, and drugs, for example local anaesthetics (Martinac

274

Booth et al.

et al., 1990) that act via the membrane, may be a particular source of transient channel activation. Of perhaps greater significance, periods of cell wall remodeling, particularly in Gram‐negative bacteria where a unimolecular layer of peptidoglycan is found, may result in channel activation due to localized changes in tension. 1. Other Functions for MS Channels To date, the primary function of MS channels has been seen to be relief from the stress associated with hypoosmotic shock—rapid transitions from high‐osmolarity environments to low. However, there are many MscS/MscK homologues in E. coli (and in many of other organisms) that cannot readily be assayed either as cellular functions or as electrical activities in patch clamp. Conceivably, these proteins have evolved diVerent functions; however, equally, the possession of a gene by an organism does not immediately imply a function within the physiology of that organism. For example, the YjeP, MscK homologue, can be deleted from E. coli without any apparent physiological consequence; however, a null mutant in Erwinia chrysanthemi leads to sensitivity to the osmoprotectant betaine when cells are grown at high osmolarity in the presence of KCl (Touze et al., 2001). Clearly, in two related organisms an important function for one homologue has been taken over by other proteins. In Arabidopsis MscS homologues are implicated in shape regulation and division of chloroplasts (Haswell and Meyerowitz, 2006). When expressed in E. coli, at least one of these MS channels is a functional channel and protects mutants lacking MS channels against hypoosmotic shock. While one cannot ascribe functions to all MS channel proteins, it remains possible that some are involved in more subtle processes than simple stress relief. 2. Expression of MS Channels Small‐scale, but significant, changes in MS channel gene expression have been observed in E. coli (Stokes et al., 2003). The increase in expression is generally two‐ to three‐fold and is in response to either increases in osmolarity or entry into stationary phase. Both the mscL and mscS genes are transcribed from promoters recognized by both s70 and s38, leading to low levels of transcription during vegetative growth. Enhanced expression takes place during osmotic stress and in stationary phase, two conditions where s38 (RpoS) protein abundance increases and forms an RNA polymerase with modified specificity (Hengge-Aronis et al., 1993; Hengge-Aronis, 1996). One of the smaller MscK homologues in E. coli, F783 (ybiO), is also known to be regulated by s38 (Schellhorn et al., 1998). Mutants lacking s38 exhibit reduced expression both of the MscS and MscL channel genes (and F783) and mutants in which s38 is stabilized express higher levels of the


275

channels (Stokes et al., 2003). RpoS mutants exhibit extreme hypoosmotic shock‐sensitivity after entry into stationary phase (Stokes et al., 2003). However, since s38 also regulates some enzymes involved in stationary phase cell wall remodeling, the mutant phenotype may arise from either the deficiency in channel genes, or altered cell wall. Not all of the expression pattern seen with MscS can be accounted for by the change of the sigma factor from s70 to s38. It seems highly probable that other protein factors are important for regulation; changes in DNA topology associated with higher osmolarity may also play a role in regulating expression. In the case of MscK, it has been shown that its expression can be lowered by inactivating the LeuO protein, which is considered to be a DNA‐binding protein of low specificity (Klauck et al., 1997). In addition to transcriptional regulation, both mscL and mscS mRNA molecules have relatively weak ribosome‐binding sites with the eVect that their translation can be diminished even when transcription takes place. In E. coli, and possibly other bacteria, the rate of translation of the mRNA for any protein is determined to a significant extent by the availability of ribosomes and the strength of the ribosome‐binding site. The presence of highly abundant mRNA molecules that have strong ribosome‐binding sites can cause a significant reduction in the translation of less abundant mRNA molecules with weak ribosome‐binding sites. One consequence of this for the MS channels may be to cause a strong degree of heterogeneity in the bacterial population with regard to the abundance of assembled channels. A stochastic distribution of channel subunits would lead to some cells with very few channels since a single channel requires five (MscL) or seven (MscS) subunits. The possession of two independently expressed channels may be a prerequisite for survival of hypoosmotic stress in cells subject to stochastic distribution of the number of subunits, since redundancy reduces the chance of any one cell having no channels. 3. MS Channel Function in Other Bacteria Lactococcus lactis is one of the few Gram‐positive organisms in which the functional role of MS channels has been investigated (Folgering et al., 2005). The genes encoding the MscL (MscL‐Ll) and MscS (MscS‐Ll) proteins were cloned and expressed in E. coli MJF465, which lacks the three major E. coli MS channels, MscL, MscS, and MscK. Expression of the L. lactis channels protected cells against hypoosmotic shock, indicating retention of function when expressed in E. coli. As expected, both channels gave electrophysiological signatures similar to their respective E. coli homologues. However, MscS‐Ll activity was not detected after fusion of L. lactis membranes with liposomes, whereas MscL‐Ll activity was readily observed. RT‐PCR experiments verified that both mscL‐Ll and mscS‐Ll genes were transcribed,

276

Booth et al.

suggesting possible posttranscriptional regulation of MscS protein production. Inactivation of MscL‐Ll led to a reduced rate of betaine eZux in response to hypoosmotic shock, but only small changes in survival of the mutant were observed. Since L. lactis grows in chains, the authors suggested that this growth morphology would lead to an overestimate of the number of survivors since only one cell in a chain was required to survive to allow a colony to form. An equally plausible explanation is that the incubation conditions predispose the cells to survive despite the inactivation of MS channels. Thus, in E. coli a variable fraction of cells of a triple channel mutant (lacking MscL, MscS, and MscK) survive depending on the precise growth conditions. Most significantly, the inclusion of betaine in the growth medium enhances survival (N. R. Stokes, W. Bartlett, and I. R. Booth, unpublished data) and this osmoprotectant was used in the L. lactis experiments. Similar influences of betaine may account for the failure of MscS, MscL double mutants of Corynebacterium glutamicum to exhibit significant changes in phenotype (RuVert et al., 1999; Nottebrock et al., 2003).

III. THE STRUCTURE OF MscS ˚ by Doug The crystal structure of the MscS protein was determined at 3.9 A Rees’s group in 2002 using the E. coli protein (Fig. 2) (Bass et al., 2002). Possessing the structure of the E. coli protein has generated a considerable advantage for the study of this channel since almost all of the genetics, molecular biology, and electrophysiology had already been conducted with this species. E. coli MscS is a 286‐amino acid protein and is one of the smallest homologues in this family of proteins. Almost the whole protein is visible in the crystal structure, with only the first 26 and the last 6 residues not resolved. Thus, the crystal structure represents an almost complete image of the protein (Bass et al., 2002). The channel is a homoheptamer with a central pore. The protein falls into two quite distinct domains, a three helix membrane domain and a large cytoplasmic domain. Although there are three helices, only two are genuinely TM, TM1 and TM2. The third helix lines the pore, but in fact only spans the region that is approximately equivalent to the inner leaflet of the lipid bilayer, the rest of the pore is made up from an extended linker that joins TM2 to TM3 and the outer mouth is formed from the N‐terminal region of TM1 (Fig. 2). The pore‐lining TM3 helix is considered to be in two halves: TM3A and TM3B, with the former lining the pore and the latter being an amphipathic helix that lies along the membrane surface at the junction between the membrane domain and the cytoplasmic domain (Fig. 2B).

277

10. MscS Mechanosensitive Channel A Membrane domain

TM1–TM2

TM3A

TM3B

b-Domain

Lateral portal ab-Domain

b-Barrel

B TM2–TM3 linker

TM3A

TM3B

FIGURE 2 The crystal structure of MscS. (A) The heptameric structure is depicted showing only the backbone of each subunit. One subunit is indicated in black to show the path followed by a single subunit. Specific domains referred to in the text are labeled. (B) The structure of the pore region of MscS showing only the path of the TM2–TM3 extended linker and the TM3A and TM3B helices. Again a single strand has been depicted. The images were created using Protein Explorer (Martz, 2002).

278

Booth et al.

A. The Membrane Domain The membrane domain constitutes only 40% of the total protein, but can be much more in some homologues that have between 4 and 11 TM spans in all (Fig. 1) (Levina et al., 1999; Pivetti et al., 2003). The E. coli MscS protein has three helices with an overall orientation NOUT –CIN, where OUT refers to the periplasmic face of the membrane and IN to the cytoplasmic face (Miller et al., 2003a). TM1 and TM2 pack against each other, but in the crystal state these two helices lie well‐separated from the TM3 pore‐lining helices and are slightly twisted relative to the axis of the pore (Bass et al., 2002). This conformation is unlikely to reflect the natural state of the channel in the membrane where the lipids will exert pressure to compact the helices such that they make direct contact with the outside of TM3. The removal of the lipid by detergent to facilitate crystallization is principally responsible for the observed conformation, but the formation of stable protein‐protein contacts in the crystal may also aVect the observed organization. In the crystal structure, the densities for TM1 and TM2 are less well‐ defined than for TM3, suggesting that they retain either some mobility in the crystal or a number of slightly diVerent alternative packing arrangements (Bass et al., 2002). The strongest conservation in MscS is TM3, but even here there is considerable diversity among the 19 subfamilies of MscS homologues (Bass et al., 2002; Pivetti et al., 2003). In the crystal structure, this is also the best region of well‐defined density. TM3 is considered to consist of two domains: TM3A, residues 96–112 line the pore and TM3B, residues 114–127 form a helix that lies along the surface of the inner leaflet of the membrane and oriented so that it is tangential to the axis of the pore (Fig. 2B) (Bass et al., 2002). Residues 112 and 113, asparagine (Asn) and glycine (Gly) in E. coli MscS, act as a hinge allowing the helix to bend. It is notable that while the Asn residue is moderately highly conserved, a number of diVerent residues replace the MscS Gly113 in other homologues. TM3A and TM3B residues define key attributes of the channels, particularly gating pressure, open dwell time, and inactivation kinetics. B. The Cytoplasmic Domain The C‐terminal domain hangs below the membrane domain, resembling a Chinese lantern—there is a large vestibule created from the seven subunits that is perforated by lateral portals at the subunit interfaces and an axial portal. The domain is suspended from TM3B and this structure may be critical to transmitting conformational changes to the C‐terminal domain


279

(Fig. 2). Each C‐terminal domain is an 17‐kDa unit that consists of three subdomains: b (132–177), ab (188–265), and the b‐barrel (271–280) (Fig. 1). In the channel, the oligomer of the seven C‐terminal domains creates a large ˚ and which is 70‐A ˚ long (Bass vestibule that has an external diameter 80 A et al., 2002). The enclosed chamber of the vestibule varies in diameter ˚ ) to the neck (27 A ˚ ). Access to narrowing from the portal region (50 A ˚ wide, created by the the vestibule is via seven lateral portals, each 14‐A junctions between the subunits at the interface between the upper b‐domain and the lower ab‐domain. In essence, the b‐domain lies immediately below the TM3B segment of the pore‐lining helix and narrows the upper part of the vestibule such that there is in eVect a wide neck to the cytoplasmic entrance to the pore. The b‐domains are themselves a recognized structural fold, the sm‐fold, associated with some classes of nucleoproteins, where they form rings around DNA. Usually, these proteins are heptameric but other oligomeric states are possible (Toro et al., 2002). The ab‐domains combine to form the bottom of the vestibule. The seven‐ strand b‐barrel created by residues 271–280 represents a potential eighth ˚ and its interior is lined with axial portal. However, the diameter is only 8 A hydrophobic residues and this may prevent easy passage of hydrated solutes. We have shown that the b‐barrel may be required for stable assembly of the channel—moreover, the b‐barrel is an important structural element that is required for some of the transitions undergone by MscS channels (see below) (Schumann et al., 2004). Small proteins (e.g., GFP and alkaline phosphatase) can be fused to the C‐terminus of E. coli MscS without severely impairing assembly or gating of the channel (unpublished data). Some homologues naturally have large domains attached to the C‐terminus of their MscS protein sequence. C. Variations in Structure Despite some specific variations at the C‐terminus, this end of the protein tends to be relatively conserved for length. In contrast, MscS homologues with large extensions at the N‐terminus are common, with MscK representing a particularly extreme case (Fig. 1) (Levina et al., 1999). MscK is 1120 amino acids (120 kDa) and has an MscS‐like domain at the C‐terminus. Immediately N‐terminal to this channel‐forming domain is a membrane region that has been proposed to form a further eight TM spans (i.e., making 11 in all). At the N‐terminus of the predicted protein is a signal sequence that is processed when the protein is exported to the periplasm. This signal sequence ensures that a large (45 kDa) domain is located to the periplasm. Little or nothing is known about this domain. Constructs that try to recreate

280

Booth et al.

‘‘MscS’’ from MscK domains are at best poorly active and require addition of a signal sequence for them to be correctly assembled in the membrane (C. Li and I. R. Booth, unpublished data). Thus, whereas E. coli MscS can readily achieve an NOUT –CIN organization, the equivalent MscK domain requires either a signal sequence or the rest of the protein to achieve the correct organization in the membrane. D. Twisting MscS Around the Pore One of the most important characteristics of MscS is the path followed by the individual subunits relative to the pore (Bass et al., 2002). Each subunit has its N‐terminus in the periplasm. The TM3A helices cross the membrane at 27 and pack tightly against each other with crossing angles of 22o. On leaving the membrane, the TM3B helix takes a path tangential to the axis of the pore (Fig. 2), and subsequent packing of the b‐ and ab‐domains causes the path of the subunits to twist around the axis of the channel, a process that is completed by the packing of the protein into the seven‐stranded b‐barrel. As a consequence of this packing arrangement, the C‐terminus of a strand exiting the b‐barrel is located 250 –270 relative to the N‐terminus of TM1. The eVect of twisting each subunit around the axis of the pore is probably critical for stability, but even more significant for structural transitions during gating (Edwards et al., 2004, 2005). Indeed it is one of the major properties of MscS that the protein spontaneously oligomerizes when freed from the membrane with detergents (R. Bass, personal communication; S. Miller and I. R. Booth, unpublished data). Given that the protein is stable as a heptamer of free monomers in the membrane, it is inferred from these observations that removal of the lateral pressure generated by the lipid bilayer, allows MscS to adopt alternative packing arrangements from those found in the closed state in the membrane. This process can be accelerated by cross‐linking introduced cysteine (Cys) residues. Uniquely, an S267C mutant forms SDS‐stable oligomers, up to and including the heptamer, when cross‐linked with the fixed‐length reagent o‐phenylenedimaleimide (o‐PDM) (Miller et al., 2003b). In rapid succession, dimers are supplemented by trimers through to the heptamer. Other Cys residues inserted close to the position of S267 in the crystal structure do not generate these stable oligomers, despite forming the initial dimer. Placing a Cys residue in MscK at the equivalent position to S267C also generates SDS‐ stable oligomers (C. Li, S. Miller, and I. R. Booth, unpublished data). This suggests that the property displayed when cross‐linked is intrinsic to the structural organization of the proteins. Analysis of the cross‐linked proteins


281

revealed no further cross‐links and the multiple forms persisted during purification with gentle detergents. The most remarkable observation is that the trimer, which can only covalently link two of the subunits, carries the third subunit into the SDS‐stable state. E. MscS Is Small but Beautifully Formed MscS in E. coli is a small protein and appears to require all of the elements evident in the sequence and structure. Trimming the E. coli MscS protein by making structured deletions aVects the stability of the assembled complex (Miller et al., 2003a; Schumann et al., 2004). Removal of the nonconserved stretch from residue 8 to 12 at the N‐terminus causes reduced accumulation of the MscS protein in the membrane. Larger deletions (removing residues 8–21) destabilize the protein resulting in the accumulation in the membrane of a truncated protein of 17 kDa, which may be the C‐terminal domain (Miller et al., 2003a). Extending the deletions into TM1 causes almost complete loss of the protein from the membrane. Similarly, the protein does not readily tolerate deletions from the C‐terminal end. Proteins that have the base of the vestibule and the b‐barrel deleted are less stable than the parent, but larger deletions to the boundaries of the ab‐ and b‐domains do not result in any active protein and no accumulation of protein in the membrane. The mutants that have the base of the vestibule, including the b‐barrel, deleted (266–286) are particularly interesting for function analysis. The assembled 266–286 channels were functional as indicated by their ability to protect a channel‐less mutant E. coli strain against hypoosmotic shock (Schumann et al., 2004). Further, the channels could be gated by pressure in isolated membrane patches, but the pressure required to gate the channel was slightly higher than that observed for the wild‐type channel. However, the significant change was observed after the channels were allowed to undergo desensitization (inactivation). MscS channels, uniquely among MS channels analyzed to date, exhibit the desensitization property. After being maintained open at high, subsaturating pressure (i.e., the pressure required to open multiple channels in a patch and assumed to open all MscS channels present but not great enough to open MscL channels) the channels close. Channel closure follows essentially first order kinetics and the rate is inversely proportional to the pressure on the patch (Akitake et al., 2005). However, this inactive state can be readily distinguished from the closed state. The latter is observed at lower pressures when channels undergo frequent openings followed, a few hundred milliseconds later, by spontaneous closure. Many cycles of opening and closure can be sustained without any apparent loss in channel function. The desensitized state is characterized

282

Booth et al.

by the fact that the patch must be rested at zero pressure for several minutes before channels will again respond to a change in pressure by opening (Koprowski and Kubalski, 1998; Akitake et al., 2005; Grajkowski et al., 2005). We observed that removal of the b‐barrel and the base of the vestibule did not aVect desensitization per se (Schumann et al., 2004). However, desensitized channels were impaired in their recovery of the active state even after rest for several minutes. The 266–286 protein was most aVected. The ability to recover could partially be restored by introduction of a sequence of eight amino acids that derive from the histidine (His)‐tag. Thus, one can argue that the base of the vestibule, including the b‐barrel, is critical for recovery from the desensitized state, but is not otherwise essential. Consistent with the known stability of b‐barrel structures, the presence of this feature in MscS may aid correct assembly and facilitate some structural transitions. Finally, the introduction of Cys residues into MscS must be undertaken with great care to avoid disruption of the structure. In creating Cys‐ containing proteins, we have observed that despite the cytoplasmic location of the substituted residues, oxidation to form cross‐linked proteins frequently occurs. This is not true for all residues (e.g., S267C and S196C are both exempted from this observation), but similar observations have been made for Cys residues located, respectively, on the surface and on the inside of the vestibule. Commonly the oxidized proteins are observed to locate poorly to the membrane and in some cases can only be observed to accumulate if cells are grown in the presence of a reducing agent during the period of induction of expression of the mutant proteins. It seems possible that the Cys residues may oxidize during assembly of the channel protein and that this leads to aberrant conformations that are then subject to degradation. The severity of this eVect is position specific and does not generally debar making X to C mutations, merely requiring greater caution than is possibly the case with other membrane proteins.

IV. MscS MUTATIONAL ANALYSIS The discovery of the yggB gene that encodes MscS was a consequence of the analysis of a gain‐of‐function (GOF) mutation in the kefA gene that was subsequently shown to encode MscK (Levina et al., 1999; McLaggan et al., 2002). This mutation aVected the ability of cells to grow at high osmolarity in the presence of 0.6‐M Kþ and betaine or proline, as osmoprotectants. The mutation was subsequently shown to reside in the TM helix equivalent to TM3A of MscS and altered the gating of the channel, rendering it inappropriately active. It is believed that the mutant channel activates only in the


283

presence of betaine or proline and high Kþ, because under these conditions the cell needs to release Kþ and the normal Kþ eZux systems are blocked. The rise in turgor associated with betaine accumulation precipitates premature channel activation. However, the phenotype of this mutant is critically dependent on the expression level, since placing the mutation (G922S) in the cloned kefA gene, which elevates expression 20‐ to 30‐fold, causes growth inhibition even at low osmolarity (C. Li and I. R. Booth, unpublished data), whereas similar expression of the wild‐type protein is tolerated. The discovery of the structural gene for MscS precipitated a flurry of analysis, leading to demonstration that yggB gene alone was suYcient to generate channels with the known properties of MscS (Okada et al., 2002; Sukharev, 2002), the crystal structure of the protein, and the search for mutants. The equivalent search for mscL gating mutants had been highly successful, simply by screening for growth defects associated with expression of mscL genes that had been mutagenized (Blount et al., 1996a,b, 1997). An equivalent analysis of MscS mutants yielded only a single mutant allele, V40D, expression of which blocked growth, accelerated Kþ loss, and rendered cells sensitive to hypoosmotic acid shock (Okada et al., 2002). The failure to find other MscS GOF mutants may arise from the abundance of this channel since many GOF alleles were isolated in MscK, which is expressed at lower levels, and could be constructed by site‐directed mutagenesis in MscS when suitable precautions are taken (Miller et al., 2003a; Edwards et al., 2005). Interesting mutations that modify the gating of MscK were a product of a screen in Salmonella typhimurium for mutations that would allow a nadB mutant strain to grow on 0.1‐mM quinolinic acid (QA). Normally nadB mutants require 10‐mM QA for growth, a phenotype that is believed to arise either from poor entry of the acid or rapid expulsion. Among the mutants allowing growth at 0.1‐mM QA were five kefA (mscK) alleles: R792P, L866Q, W909R, A918P, and G924S. The first two mutations are outside the TM3A pore‐forming helix, but the other three are either in the sequence equivalent to TM3A or in the extended loop connecting TM2 to TM3A. Transfer of the mutations to their equivalent positions in MscS (A918 and G924 are conserved residues) generated GOF phenotypes indicating functional equivalence between the pore structures in MscS and MscK. Similarly, creating T93R in MscS, the equivalent of W909R in MscK, also generated a GOF phenotype which was the first indication of the importance of this sequence in the gating transition (Miller et al., 2003a). The other two alleles are also interesting since both are less severe GOF alleles in MscK and R792 is not represented in the MscS structure since it forms part of the linker that connects the ‘‘MscS domain’’ to the rest of the MscK protein. L866Q, when recreated in MscS, does not have a strong phenotype and this reveals

284

Booth et al.

potentially significant diVerences in the two structures. In MscK, L866Q is a mild GOF allele, whereas a double mutation in MscS, close to the equivalent position (I48D, S49P; single mutations have no observable phenotype) blocks gating. Subsequent studies have created many mutations in both MscS and MscK that have facilitated model building for the gating transition. However, it is frequently observed that introducing mutations into MscS destabilizes the protein and thus the absence of observable mutants in the more generic growth‐inhibition screens may arise from the significant structural perturbation such mutations generate.

V. STRUCTURAL TRANSITIONS IN MscS A. The Need for the Closed State Bacterial cytoplasmic membranes are simultaneously the site of energy transduction and the location for MS channels. The opening of the latter will depolarize the membrane and will perturb cytoplasmic ion pools leading to a loss of homeostasis and diminished energy production. Consequently, MS channels must remain in the closed state for much of the time and after opening they must revert to the closed state quickly to avoid impairing the growth (and survival) of the cell. This has formed the basis for the selection of gating mutants in MscL and in a more limited sense for MscS mutants (Blount et al., 1996b, 1997; Okada et al., 2002). In both cases, growth inhibition results from expression of channels that gate more readily at lower pressure than the wild type. However, the correlation is not straightforward. Growth inhibition is the product of the expression level of the protein, its stability in the membrane and the eVect of the actual amino acid change on both the threshold pressure for channel activation and the open dwell time. A mutant that gates at lower pressure, but which also either aVects channel assembly or the open dwell time of the channels, may not inhibit growth of the bacterial cell. This is exemplified by the N15D MscL mutant (Buurman et al., 2004). At the low levels achieved by expression from the chromosome, the mutant channels facilitate growth at low Kþ concentrations of a mutant strain lacking the normal Kþ uptake systems, that is, N15D activity improves cell physiology because of the particular problems of this E. coli mutant strain. In contrast, expression of the same mutation from a high copy plasmid has a very severe eVect on growth (Ou et al., 1998). Similarly, we have often observed compensation arising from simultaneous changes in open dwell time and threshold pressure for activation, with one eVect oVsetting the other. In both MscL and MscS, the ion impermeability of the closed state of the channel is maintained by rings of hydrophobic residues. In the case of MscS,


285

these are two rings of leucine (Leu) residue, L105 and L109 (Bass et al., 2002). L109 lies immediately adjacent to the cytoplasmic neck of the pore, with L105 residues in a ring immediately above this (i.e., toward the periplasmic face)—thus the seal is not symmetrically located at the middle of the membrane, but lies closer to the cytoplasmic face. This feature may be critical to the gating transition (see below). Substitution of small residues or hydrophilic residues at positions 105 and 109 creates channels that gate at lower pressures. The greatest eVects are seen at position 105. However, the channels are closed until pressure is applied, contrary to speculation that such channels would be open pores (Edwards et al., 2005). Insertion of larger hydrophobic residues in place of Leu creates stable channels, but these have a tendency to require higher pressure for gating. B. The Crystal State In the crystal form, MscS is a homoheptamer that has been trapped in an open state. A central pore is seen down the long axis of the protein. The ˚ (depending of the method of assessobserved diameter of the pore at 8–11A ˚ ) from conductance measurements for ment) is smaller than predicted (14 A the fully open channel (Sukharev, 2002). This has led to a degree of controversy (Anishkin and Sukharev, 2004) concerning whether this protein is the open state, the closed state, or ‘‘an open state.’’ The latter represents a compromise between the two extreme states that the channel could occupy. Whatever state the structure represents, it is clear that a hydrated ion or low‐ molecular‐weight solute could pass through the pore as displayed in the crystal form. Other biochemical evidence suggests that the closed form of the channel is more compact than that depicted in the crystal structure (Miller et al., 2003b). Thus, we have shown that single Cys residues substituted for serine (Ser) (there are no endogenous Cys residues in E. coli ˚ apart in the crystal structure can readily be MscS) that are greater than 10 A oxidized by Cu/phenanthroline reagent. The most significant of these data were derived using studies of two mutants: S58C and S267C residues, which are located at the base of TM2 in the membrane domain and at the bottom of the vestibule in the C‐terminal domain, respectively. In adjacent MscS subunits, these residues are separated in the crystal structure by 19 (S267C) ˚ (S58C). The residues were rapidly cross‐linked. Bringing S58C and 33 A residues together could be achieved by packing the TM1–TM2 helices closer together against TM3; however, significant mobility would still be required to facilitate the formation of the cross‐link since TM1 helices would be placed between TM2 in such a packed structure. Perhaps more significantly for the compact closed structure model, S267C residues are buried in the

Booth et al.

286

crystal state, such that neither residue should easily react with the next one (Fig. 3). However, in the closed state, that is, the channel embedded in the membrane, the residues readily react and are cross‐linked by Cu/phenanthroline and MTS‐1‐MTS. Iodine cannot be used with any confidence as an oxidizing agent to study MscS, as incubation with this compound has been found to cause the rapid degradation of the protein, even in the native protein where Cys residues are absent (S. Miller, unpublished data). These data, plus other recent unpublished studies, point to the closed channel being in a compact state relative to that depicted in the crystal structure. Other data support a more compact form for the closed state of the channel (Koprowski and Kubalski, 2003; Grajkowski et al., 2005). Cross‐ linking lysine (Lys) residues by 1‐min exposure of membrane patches to bis (sulfosuccinimidyl)‐suberate caused loss of channel activity that could not be reversed by washing out the cross‐linking reagent. Since all except one Lys residue are situated in the cytoplasmic domain, these data were interpreted to indicate that preventing C‐terminal domain movement blocked channel gating (Koprowski and Kubalski, 2003). In the same study, Kubalski and colleagues demonstrated that Ni2þ could block the transition from the closed to the open channel when added to membrane patches from cells expressing MscS protein with a C‐terminal His6‐tag. The eVect could be

A

B

19 A

FIGURE 3 The position of the S267 residue. The position of the S267 residue, which when modified to S267C and cross‐linked with o‐PDM leads to SDS‐stable oligomers, is indicated. (A) The position of the residue relative to the whole channel protein is indicated by a space‐filled residue (yellow) against the backbone of the subunits. (B) A space‐filled model viewed from the base, in which the base of the vestibule is dark gray, the b‐barrel is pale gray, and the S267 residues are yellow. Images were created using Protein Explorer (Martz, 2002).


287

reversed by washout of the Ni2þ and was not seen when the channel lacked the inserted C‐terminal His6‐tag. Moreover, this group also demonstrated that addition of high‐molecular‐weight ficoll [400 kDa; 1–10% (w/v), final concentration] to the bath increased the rate of inactivation of the channel and diminished the total number of active channels (Grajkowski et al., 2005). The presence of ficoll increased the pressure required to activate the channels. Addition of ficoll to the periplasmic side of the patch slowed channel inactivation but was without eVect on the number of active channels. Clearly, these data indicate a greater eVect of ficoll from the cytoplasmic side of the patch, that is, action via the C‐terminal domain. The high molecular mass of the ficoll would probably prevent it entering the vestibule and consequently the eVects have been interpreted as arising from inhibition of structural transitions in the C‐terminal domain that are required for closed‐to‐open transitions in MscS. These data are consistent with the MscS protein making a large conformational change during the transition from the closed to the open state, which is consistent with the crystal structure representing one open state that the protein can achieve. C. The TM3 Pore One of the most obvious features from the structure of E. coli MscS is the very tight packing of the TM3A helices, which are in the closest proximity possible (Bass et al., 2002; Edwards et al., 2005). This is due to the conservation of Gly and alanine (Ala) residues such that the former creates a surface against which the Ala residues are packed. The E. coli MscS family of homologues carries the sequence A98hhG101A102hG104hA106hG108hA110hyG113, where A and G have their normal single letter code meaning, h ¼ hydrophobic and hy ¼ hydrophilic residues. In the crystal structure, A106 and G108 are within van der Waals radii of each other, but A98–G101 and A102–G104 are packed somewhat more loosely. The helices cross each other at an angle of 22o (Bass et al., 2002). We proposed that the Ala residues formed knobs that slid across the grooves created by the Gly residues in the adjacent helix and that the closed‐to‐open transition involved rotation of the TM3A helix such that new contacts were established between diVerent Gly and Ala pairs (Edwards et al., 2005). This hypothesis was tested by changing Ala to valine and Gly to Ala to create proteins in which bulkier residues replaced the simple knobs and grooves. All of these mutants proved more diYcult to gate. Conversely, the substitution of Ala by Gly removed the knobs and created channels that opened more easily (A106G). A similar mutation further up the pore (A102G) was aVected in that its open dwell time was much reduced, suggesting that in this case removing the knob removed the stabilizing factor

288

Booth et al.

for the open state. A double mutant that recreated Ala‐Gly packing but on opposite helices (i.e., A106G/G108A) exhibited a return to normal gating pressure. These data support the idea that TM3A helices rotate during the closed‐to‐open transition such that the Ala residues cross over the Gly surfaces. It seems likely that the limit of this structural transition is imposed by bulky hydrophobic residues both in the lumen of the pore (L105 and L109) and in the interfaces packed against the TM1–TM2 pair, but this hypothesis has not been fully tested. During the course of analyzing the importance of the conserved Gly‐Ala packing in MscS, we observed that the eVects of mutations was of increasing severity when the changes were made close to the seal of the channel compared with mutations created higher up TM3A (Edwards et al., 2005). For example, A106V displays two open states—an unstable wild‐type conductance that is seen at pressures intermediate between those that open MscS and MscL and then a low conductance state that is the dominant form of activity at high pressure equal to those needed to gate MscL. Similarly, G108A, A110V, A106L, A106S, and A106G channels exhibit lowered conductance. Ser residues are strongly perturbing where the path of the helix in the wild‐type state is constrained. Ser residues have the capacity to form intrahelical and interhelical H‐bonds that can perturb helix path and/or helix packing. Gly to Ser mutations at positions 101, 104, and 108 progressively lower the duration of the open channel from 250 to 1 ms. This open dwell time analysis points to greater constraints on the helix packing around the seal than was the case higher up the channel pore. Consequently, the model envisaged that to achieve the fully open state the helices would tilt outward to a greater extent at their periplasmic ends than at the pore region. Modeling of the TM3A helices also predicted this structural transition to account for the creation of the pore (Edwards et al., 2005). Given that the known state is an open one (though not necessarily the fully open state) in the crystal form, by reference to this structure one must note that to achieve the closed state the TM3A helices must attain a more vertical state, pack more closely (by moving to Ala98‐Gly102 tight packing), and be rotated such that their Leu residues point toward the center of the channel pore. D. The Closed‐to‐Open Transition High‐resolution recordings of MscS channel activity have indicated that the channel may open via an ion conducting substate that is short‐ lived (20 ms) and which has a conductance 2/3 of the fully open conductance (Shapovalov and Lester, 2004). This contrasts with MscL where many substates are seen and some of the GOF mutants lead to higher occupancy


289

of some of the subconducting states (Anishkin et al., 2005). No comparable analysis has been completed for MscS GOF mutants, although as referred to above, we have observed some GOF and loss‐of‐function mutants to exhibit lowered conductance. MscS displays the potential to be voltage‐gated and the capacity for inactivation. These two issues frame the discussion of the closed‐to‐open transition. Several charged residues reside in the TM1 and TM2 helices (R46, R74, and R88) and the structure resembles a voltage sensor (Bass et al., 2002). This led to the proposal that the channel was voltage and pressure sensitive, in line with earlier observations of channel activity in membrane patches (Martinac et al., 1987). Subsequent analysis has indicated that MscS activation is essentially voltage independent but that the inactivation process may be voltage sensitive (Akitake et al., 2005). At high negative holding potentials (negative patch pipette voltage), the rate of inactivation of channels was enhanced. If, as expected, the TM1–TM2 sensor paddle is able to move in the electrical field, then application of a TM voltage (pipette negative) could cause a significant displacement of this part of the channel. Given the importance of the link between TM2 and TM3 for channel activation, it is reasonable to expect that these conformational changes would be transmitted to the pore. Clearly, these phenomena are important aspects of the structural transitions that MscS can undergo in patches, but are they relevant in the context of the cell? Bacterial cells have membrane potentials varying between 60 and 240 mV (inside negative), depending on the organism and the environmental conditions. The polarity of the field is, however, more significant than the dimensions with respect to the activation/inactivation of MscS. In the normal state, the membrane potential is negative inside (i.e., positive outside) which is the opposite of the polarity applied in patch clamp to eVect changes in the inactivation rate. Perhaps more significantly, MscS is almost certainly the third MS channel in E. coli to open, since the sequence observed in patches is MscM, followed by MscK and then MscS and finally MscL (Batiza et al., 2002). The current‐carrying capacity of either MscM or MscK should be suYcient to depolarize the membrane such that when MscS is open there is no significant potential to aVect the kinetics of the channel. Conditions used to measure MscS activity may, therefore, lead to properties that do not have a corresponding cellular dimension. For example, inactivation of MscS channels is seen when high pressure is sustained on the patch for an extended period up to several seconds. However, in cells the action of opening the channels dissipates the pressure gradient and the expected duration of the open state should be in the order of milliseconds rather than seconds. Thus, inactivation may be a measurement artifact rather than an important functional attribute of MscS.

Booth et al.

290

We have proposed a model for the closed‐to‐open transition based on our cross‐linking data and the observations made by other groups (Fig. 4) (Edwards et al., 2004, 2005). In the closed state, we envisage MscS TM3A helices to be closer to the perpendicular than in the crystal structure and consequently to exhibit altered packing between the potential Gly‐Ala pairs formed between adjacent TM3A helices. TM1–TM2 pairs are held against TM3A by the lateral pressure within the lipid bilayer. The conformational change in TM3A is transmitted to TM3B such that the packing of the b‐domains is modified and this may aid maintenance of the closed state. The overall eVect is that both the membrane and the cytoplasmic domains are in a more compact conformation. Distortion of the membrane bilayer allows unpacking of the TM1–TM2 paddle from TM3A and this change is suYcient to allow the pore‐lining helices to rotate and tilt such that the pore

TM2–TM3 stretched loop

G101 G104 Seal: L105 L109

A98

A102

G108 A106

FIGURE 4 The closed‐to‐open transition in MscS. The backbone of two TM3A helices is depicted with the Gly (dark gray) and Ala (light gray) pairs indicated relative to the position of the Leu seal. The Gly residues provide surfaces over which the Ala residues slide to provide the smooth transition to the open state. In the closed state, A98‐G101 and possibly A102 and G104 are proposed to approach each other as the helices turn and straighten. The open state must be stabilized and this may require that the fully open state involves the crossing of bulkier residues to form a resistance to prevent the collapse back to the closed state. In support of such a model it has been observed that Ala to Gly mutations cause channels to become unable to sustain an open state, but that this can be suppressed by mutagenizing Gly to Ala at other positions in TM3A (unpublished data). Images were constructed using Protein Explorer (Martz, 2002).


291

enlarges. Such molecular motion must be accompanied by increased separation of the helices to create a pore of suYcient size for hydrated ions and small solutes to pass through rapidly. In molecular dynamics simulations of MscS, much has been made of the changes in hydration and the potential for the crystal structure to represent the closed state blocked by a vapor lock (Anishkin and Sukharev, 2004; Spronk et al., 2006). There can be no doubt that opening the pore must be accompanied by changes in water structure close to the surface of the pore but whether the vapor lock is real is unclear (Spronk et al., 2006). TM3A rotation will alter the conformation of TM3B with consequences for the packing of the C‐terminal domain, such that the expanded structure seen in the crystal form is generated. However, critical datasets that might allow verification of this model for the gating transition are lacking at present.

VI. CONCLUSIONS AND FUTURE PERSPECTIVE Like all good models the crystal structure has generated speculation, experimentation, and structured simulations. Not knowing the precise state represented by the crystal structure is a disadvantage, but simultaneously it has narrowed the options on the structural transitions undertaken by MscS. Further structures will be welcome additions to this canon, as will the publication of data from Perozo’s laboratory that have used site‐directed spin labeling to examine the movements of TM1–TM3 during gating (E. Perozo, personal communication). The analysis of MscL by this method was critical to building a model that is generally accepted for the gating transition (Perozo et al., 2001, 2002a,b). Finally, understanding the structural transitions in MscS lies at one end of the spectrum of our knowledge of this system. Equally important is to go back and place the channel in the context of cell physiology to increase the understanding of the cellular function of this channel and its homologues. Acknowledgments The authors are indebted to the Wellcome Trust, the BBSRC, The University of Aberdeen, and Unilever plc for their support of our research program on bacterial ion channels. At diVerent stages and over a long period, a number of members of the group have contributed to the analysis of these channels: Sabine To¨temeyer, Neil Stokes, Natasha Levina, Debbie McLaggan, Petra Louis, and Sally Dennison—we oVer them our profound thanks for their input. Ching Kung, Sergei Sukharev, Paul Blount, Sanguk Kim, Hochterl Jeong, John Roth, Eduardo Perozo, Boris Martinac, Jim Bowie, and Jim Naismith have made major contributions to our work through either experimental support or detailed discussions. They have no responsibility for the errors that we make, but greatly reduce their number!

292

Booth et al.

References Akitake, B., Anishkin, A., and Sukharev, S. (2005). The ‘‘dashpot’’ mechanism of stretch‐ dependent gating in MscS. J. Gen. Physiol. 125, 143–154. Anishkin, A., and Sukharev, S. (2004). Water dynamics and dewetting transition in the small mechanosensitive channel MscS. Biophys. J. 86, 2883–2895. Anishkin, A., Chiang, C. S., and Sukharev, S. (2005). Gain‐of‐function mutations reveal expanded intermediate states and a sequential action of two gates in MscL. J. Gen. Physiol. 125, 155–170. Bass, R. B., Strop, P., Barclay, M., and Rees, D. C. (2002). Crystal structure of Escherichia coli MscS, a voltage‐modulated and mechanosensitive channel. Science 298, 1582–1587. Batiza, A. F., Kuo, M. M. C., Yoshimura, K., and Kung, C. (2002). Gating the bacterial mechanosensitive channel MscL in vivo. Proc. Natl. Acad. Sci. USA 99, 5643–5648. Berrier, C., Coulombe, A., Szabo, I., Zoratti, M., and Ghazi, A. (1992). Gadolinium ion inhibits loss of metabolites induced by osmotic shock and large stretch‐activated channels in bacteria. Eur. J. Biochem. 206, 559–565. Berrier, C., Besnard, M., Ajouz, B., Coulombe, A., and Ghazi, A. (1996). Multiple mechanosensitive ion channels from Escherichia coli, activated at diVerent thresholds of applied pressure. J. Membr. Biol. 151, 175–187. Blount, P., Sukharev, S. I., Moe, P. C., Nagle, S. K., and Kung, C. (1996a). Towards an understanding of the structural and functional properties of MscL, a mechanosensitive channel in bacteria. Biol. Cell. 87, 1–8. Blount, P., Sukharev, S. I., Schroeder, M. J., Nagle, S. K., and Kung, C. (1996b). Single residue substitutions that change the gating properties of a mechanosensitive channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 93, 11652–11657. Blount, P., Schroeder, M. J., and Kung, C. (1997). Mutations in a bacterial mechanosensitive channel change the cellular response to osmotic stress. J. Biol. Chem. 272, 32150–32157. Booth, I. R. (1985). Regulation of cytoplasmic Ph in bacteria. Microbiol. Rev. 49, 359–378. Booth, I. R., Cairney, J., Sutherland, L., and Higgin, C. F. (1988). Enteric bacteria and osmotic stress: An integrated homeostatic system. Soc. Appl. Bacteriol. Symp. Ser. 17, 35S–49S. Buurman, E. T., McLaggan, D., Naprstek, J., and Epstein, W. (2004). Multiple paths for nonphysiological transport of Kþ in Escherichia coli. J. Bacteriol. 186, 4238–4245. Chang, G., Spencer, R. H., Lee, A. T., Barclay, M. T., and Rees, D. C. (1998). Structure of the MscL homolog from Mycobacterium tuberculosis: A gated mechanosensitive ion channel. Science 282, 2220–2226. Doyle, D. A., Cabral, J. M., Pfuetzner, R. A., Kuo, A. L., Gulbis, J. M., Cohen, S. L., Chait, B. T., and MacKinnon, R. (1998). The structure of the potassium channel: Molecular basis of Kþ conduction and selectivity. Science 280, 69–77. Dutzler, R., Campbell, E. B., Cadene, M., Chait, B. T., and MacKinnon, R. (2002). X‐ray structure of a CIC chloride channel at 3. 0 angstrom reveals the molecular basis of anion selectivity. Nature 415, 287–294. Edwards, M. D., Booth, I. R., and Miller, S. (2004). Gating the mechanosensitive channels: MscS a new paradigm? Curr. Opin. Microbiol. 7(2), 163–167. Edwards, M. D., Li, Y., Kim, S., Miller, S., Bartlett, W., Black, S., Dennison, S., Iscla, I., Blount, P., Bowie, J. U., and Booth, I. R. (2005). Pivotal role of the glycine‐rich TM3 helix in gating the MscS mechanosensitive channel. Nat. Struct. Mol. Biol. 12, 113–119. Folgering, J. H., Moe, P. C., Schuurman‐Wolters, G. K., Blount, P., and Poolman, B. (2005). Lactococcus lactis uses MscL as its principal mechanosensitive channel. J. Biol. Chem. 280, 8784–8792.


293

Grajkowski, W., Kubalski, A., and Koprowski, P. (2005). Surface changes of the mechanosensitive channel MscS upon its activation, inactivation, and closing. Biophys. J. 88, 3050–3059. Haswell, E. S., and Meyerowitz, E. M. (2006). MscS‐like proteins control plastid size and shape in Arabidopsis thaliana. Curr. Biol. 16, 1–11. Hengge‐Aronis, R. (1996). Back to log phase: Sigma S as a global regulator in the osmotic control of gene expression in Escherichia coli. Mol. Microbiol. 21, 887–893. Hengge‐Aronis, R., Lange, R., Henneberg, N., and Fischer, D. (1993). Osmotic regulation of rpoS‐dependent genes in Escherichia coli. J. Bacteriol. 175, 259–265. Klauck, E., Bohringer, J., and Hengge‐Aronis, R. (1997). The LysR‐like regulator LeuO in Escherichia coli is involved in the translational regulation of rpoS by aVecting the expression of the small regulatory DsrA‐RNA. Mol. Microbiol. 25, 559–569. Kloda, A., and Martinac, B. (2001a). Mechanosensitive channels in Archaea. Cell Biochem. Biophys. 34, 349–381. Kloda, A., and Martinac, B. (2001b). Structural and functional diVerences between two homologous mechanosensitive channels of Methanococcus jannaschii. EMBO J. 20, 1888–1896. Koprowski, P., and Kubalski, A. (1998). Voltage‐independent adaptation of mechanosensitive channels in Escherichia coli protoplasts. J. Membr. Biol. 164, 253–262. Koprowski, P., and Kubalski, A. (2003). C‐termini of the Escherichia coli mechanosensitive ion channel (MscS) move apart upon the channel opening. J. Biol. Chem. 278, 11237–11245. Kung, C., and Blount, P. (2004). Channels in microbes: So many holes to fill. Mol. Microbiol. 53, 373–380. Levina, N., Totemeyer, S., Stokes, N. R., Louis, P., Jones, M. A., and Booth, I. R. (1999). Protection of Escherichia coli cells against extreme turgor by activation of MscS and MscL mechanosensitive channels: Identification of genes required for MscS activity. EMBO J. 18, 1730–1737. Li, Y., Moe, P. C., Chandrasekaran, S., Booth, I. R., and Blount, P. (2002). Ionic regulation of MscK, a mechanosensitive channel from Escherichia coli. EMBO J. 21, 5323–5330. Martinac, B., Buehner, M., Delcour, A. H., Adler, J., and Kung, C. (1987). Pressure‐sensitive ion channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 84, 2297–2301. Martinac, B., Adler, J., and Kung, C. (1990). Mechanosensitive ion channels of E. coli activated by amphipaths. Nature 348, 261–263. Martz, E. (2002). Protein explorer: Easy yet powerful macromolecular visualization. Trends Biochem. Sci. 27, 107–109. McLaggan, D., Jones, M. A., Gouesbet, G., Levina, N., Lindey, S., Epstein, W., and Booth, I. R. (2002). Analysis of the kefA2 mutation suggests that KefA is a cation‐specific channel involved in osmotic adaptation in Escherichia coli. Mol. Microbiol. 43, 521–536. Miller, S., Bartlett, W., Chandrasekaran, S., Simpson, S., Edwards, M., and Booth, I. R. (2003a). Domain organization of the MscS mechanosensitive channel of Escherichia coli. EMBO J. 22, 36–46. Miller, S., Edwards, M. D., Ozdemir, C., and Booth, I. R. (2003b). The closed structure of the MscS mechanosensitive channel—Cross‐linking of single cysteine mutants. J. Biol. Chem. 278, 32246–32250. Nguyen, T., Clare, B., Guo, W., and Martinac, B. (2005). The eVects of parabens on the mechanosensitive channels of E. coli. Eur. Biophys. J. 34, 389–395. Nottebrock, D., Meyer, U., Kramer, R., and Morbach, S. (2003). Molecular and biochemical characterization of mechanosensitive channels in Corynebacterium glutamicum. FEMS Microbiol. Lett. 218, 305–309.

294

Booth et al.

Okada, K., Moe, P. C., and Blount, P. (2002). Functional design of bacterial mechanosensitive channels. Comparisons and contrasts illuminated by random mutagenesis. J. Biol. Chem. 277, 27682–27688. Ou, X., Blount, P., HoVman, R. J., and Kung, C. (1998). One face of a transmembrane helix is crucial in mechanosensitive channel gating. Proc. Natl. Acad. Sci. USA 95, 11471–11475. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2001). Site‐directed spin‐labeling analysis of reconstituted MscL in the closed state. J. Gen. Physiol. 118, 193–206. Perozo, E., Cortes, D. M., Sompornpisut, P., Kloda, A., and Martinac, B. (2002a). Open channel structure of MscL and the gating mechanism of mechanosensitive channels. Nature 418, 942–948. Perozo, E., Kloda, A., Cortes, D. M., and Martinac, B. (2002b). Physical principles underlying the transduction of bilayer deformation forces during mechanosensitive channel gating. Nat. Struct. Biol. 9, 696–703. Pivetti, C. D., Yen, M. R., Miller, S., Busch, W., Tseng, Y. H., Booth, I. R., and Saier, M. H. (2003). Two families of mechanosensitive channel proteins. Microbiol. Mol. Biol. Rev. 67, 66–85. RuVert, S., Berrier, C., Kramer, R., and Ghazi, A. (1999). Identification of mechanosensitive ion channels in the cytoplasmic membrane of Corynebacterium glutamicum. J. Bacteriol. 181, 1673–1676. Schellhorn, H. E., Audia, J. P., Wei, L. I. C., and Chang, L. (1998). Identification of conserved, RpoS‐dependent stationary‐phase genes of Escherichia coli. J. Bacteriol. 180, 6283–6291. Schumann, U., Edwards, M. D., Li, C., and Booth, I. R. (2004). The conserved carboxy‐ terminus of the MscS mechanosensitive channel is not essential but increases stability and activity. FEBS Lett. 572, 233–237. Shapovalov, G., and Lester, H. A. (2004). Gating transitions in bacterial ion channels measured at 3 microns resolution. J. Gen. Physiol. 124, 151–161. Spronk, S. A., Elmore, D. E., and Dougherty, D. A. (2006). Voltage‐dependent hydration and conduction properties of the hydrophobic pore of the mechanosensitive channel of small conductance. Biophys. J. 90, 3555–3569. Stokes, N. R., Murray, H. D., Subramaniam, C., Gourse, R. L., Louis, P., Bartlett, W., Miller, S., and Booth, I. R. (2003). A role for mechanosensitive channels in survival of stationary phase: Regulation of channel expression by RpoS. Proc. Natl. Acad. Sci. USA 100, 15959–15964. Sukharev, S. (2002). Purification of the small mechanosensitive channel of Escherichia coli (MscS): The subunit structure, conduction, and gating characteristics in liposomes. Biophys. J. 83, 290–298. Sukharev, S. I., Martinac, B., Arshavsky, V. Y., and Kung, C. (1993). Two types of mechanosensitive channels in the Escherichia coli cell envelope: Solubilization and functional reconstitution. Biophys. J. 65, 177–183. Toro, I., Basquin, J., Teo‐Dreher, H., and Suck, D. (2002). Archaeal Sm proteins form heptameric and hexameric complexes: Crystal structures of the Sm1 and Sm2 proteins from the hyperthermophile Archaeoglobus fulgidus. J. Mol. Biol. 320, 129–142. Touze, T., Gouesbet, G., Bolangiu, C., Jebbar, M., Bonnassie, S., and Blanco, C. (2001). Glycine betaine loses its osmoprotective activity in a bspA strain of Erwinia chrysanthemi. Mol. Microbiol. 42, 87–99. Zhou, Y. F., Morais‐Cabral, J. H., Kaufman, A., and MacKinnon, R. (2001). Chemistry of ion coordination and hydration revealed by a Kþ channel‐Fab complex at 2. 0 angstrom resolution. Nature 414, 43–48.

CHAPTER 11 The MscS Cytoplasmic Domain and Its Conformational Changes on the Channel Gating Piotr Koprowski, Wojciech Grajkowski, and Andrzej Kubalski Department of Cell Biology, Nencki Institute of Experimental Biology, 02‐093 Warsaw, Poland


Overview MscL and MscS: Primary Gates and Similarities in Activation The MscL Cytoplasmic Regions and Functioning of the Channel The MscS C‐Terminal Chamber: The Cage‐Like Structure and Kinetics Structural Alterations of the MscS Cytoplasmic Chamber on Gating Conclusions and Perspectives References

I. OVERVIEW The cytoplasmic domain of the bacterial mechanosensitive (MS) channel of small conductance (MscS) is shaped by its C‐termini forming a large chamber filled with water. Several independent studies indicate that the chamber is a dynamic structure that undergoes severe conformational changes on the channel gating. Various electrophysiological and biochemical methods combined with molecular biology have been used to investigate this phenomenon and the results are presented in the chapter. The size of the chamber and its shape resemble cytoplasmic domains from eukaryotic non‐MS channels whose function in stabilization of the channel closed state is established. Analogous role of the MscS cytoplasmic chamber is discussed. Bacterial MS channels protect these cells against hypoosmotic shock. Two types of MS channels from the cytoplasmic membrane of Escherichia coli, Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58011-1

296

Koprowski et al.

MscL and MscS (the large and small conductance MS channel, respectively, see also other chapters of the book), play an essential role in the physiology of this bacterium, allowing eZux of solutes from the cytoplasm when osmolarity of the external medium decreases (Ajouz et al., 1998; Levina et al., 1999; Batiza et al., 2002). Homologues of these channels have been found widely in other bacteria (Moe et al., 1998; Levina et al., 1999) and archea (Kloda and Martinac, 2002). Few eukaryotic homologues of both channels have also been identified and they include: structurally related MscL protein from Neurospora crassa and putative membrane proteins from Arabidopsis thaliana, Saccharomyces pombe, and Drosophila melanogaster showing homology to MscS (Koprowski and Kubalski, 2001; Pivetti et al., 2003). Two MscS‐like proteins from Arabidopsis thaliana have been shown to function as channels and to control plastid size, shape, and perhaps division of plant cells during normal development (Haswell and Meyerowitz, 2006). It is not known, at present, if analogous functions can be attributed to the bacterial MS channels.

II. MscL AND MscS: PRIMARY GATES AND SIMILARITIES IN ACTIVATION The activities of MscL and MscS can be recorded after reconstitution of purified proteins in planar lipid bilayers (Ha¨se et al., 1995; Blount et al., 1996; Okada et al., 2002; Sukharev, 2002), indicating that no auxiliary proteins are necessary for the MS conduction of ions and both channels sense membrane stress directly. Their single channel conductances are large if compared to the conductances of eukaryotic ion channels and are 1 and 3.5 nS for MscS and MscL, respectively. In a typical patch‐clamp experiment, MscS is activated at membrane tension of about 5.5 dynes/cm (Sukharev, 2002), whereas for the opening of MscL considerably higher tension must be applied (around 10 dynes/cm) (Chiang et al., 2004). The activities of both channels recorded directly from the E. coli membrane are kinetically distinct: MscS opens for hundreds of milliseconds (Martinac et al., 1987), whereas the MscL open‐dwell times are in the range of tens of milliseconds (Sukharev et al., 1993). The MscS activation is dependent on the rate of applied stimulus and channels fully respond to the abrupt pressure changes but do not open at those applied slowly (Akitake et al., 2005). MscS shows inactivation during sustained pressure (Koprowski and Kubalski, 1998; Akitake et al., 2005), while MscL does not. Except of being an MS channel, MscS is modulated by voltage (Martinac et al., 1987), and it has been demonstrated that at lower depolarizing voltages (below 40 mV) the channels inactivate easily (Akitake et al., 2005). Both channels are

11. The MscS Cytoplasmic Domain

297

regulated by pH but in a diVerent way. MscS is completely blocked by pH below 5.5 (Cui et al., 1995), while low pH shifts the MscL activation curve toward higher pressures (Kloda et al., 2006). Comparing gating characteristics of both channels clearly indicated that the basic conformational rearrangements in MscS should be more complex than those in MscL. A release of amino acid sequences first and then solving crystal structures of both channels confirmed this assumption. The MscL functional channel is a pentamer (Fig. 1B), and each 136‐amino acid subunit consists of two ‐helical membrane‐spanning domains TM1 and TM2 with both C‐ and N‐termini located in the cytoplasm (Sukharev et al., 1994; Blount et al., 1996; Chang et al., 1998). TM1s line the pore and their hydrophobic residues form the transmembrane (TM) gate (Chang et al., 1998; Batiza et al., 1999). The MscL quaternary structure reveals its closed conformation. On the basis of this structure and the analysis of the channel gating, the open conformation has been predicted (Sukharev et al., 2001a,b) and experimentally confirmed (Betanzos et al., 2002; Perozo et al., 2002). MscS is a 286‐amino acid protein and its crystal structure (Fig. 1A) reveals that the channel is a heptamer (Bass et al., 2002). Each of seven subunits is composed of three TM helices TM1, TM2, and TM3. The highly conserved TM3 helices rich in glycines and alanines line the channel pore. The cytoplasmic domains of the channel are composed mostly of ‐sheets ˚ . Each domain and surround a large water‐filled chamber of diameter 40 A consists of a middle ‐domain and a lower / ‐domain (Fig. 1A), and all seven subunits are linked together by a ‐barrel composed of seven strands that are located at the very ends of C‐termini. The chamber has seven pores, ˚ in diameter each, located at the subunit interfaces and the additional 14 A ˚ diameter formed by a ‐barrel at the bottom of the chamber. opening of 8‐A It has been suggested that the structure reveals an open channel conformation (Bass et al., 2002) but molecular dynamics studies of water inside the MscS channel implicated that the structure may represent an inactive state of the channel (Anishkin and Sukharev, 2004; Akitake et al., 2005). On the basis of crystallographic data, a closed nonconducting conformation of the channel has been proposed (Bass et al., 2002; Edwards et al., 2005). Although structurally diVerent both channels, according to the existing models, are activated in a very similar fashion. The gates in both channels in a nonconducting state are formed by a tight constriction rings of hydrophobic residues within the TM domains lining the pore and the opening of the channel involves a tilt and rotation of these domains (Yoshimura et al., 2001; Sukharev et al., 2001b; Betanzos et al., 2002; Perozo et al., 2002; Barlett et al., 2004; Edwards et al., 2005). The rearrangements within MscS are, however, of smaller magnitude than in MscL (see also other chapters of the book). If the structural rearrangements within TM domains lining the

FIGURE 1 Crystal structures of bacterial MS channels of small MscS (A), and large MscL (B) conductance. One subunit of each channel was colored with conservation scores (magenta, conserved; cyan, nonconserved) by the program Consurf (Glaser et al., 2003; Landau et al., 2005) available online at: http://consurf.tau.ac.il/. The gray bar in the middle of the figure represents an approximate thickness of the membrane.


299

pores in both channels are of similar character, what makes the channels kinetically distinct and which regions of MscS underpin its inactivation? According to the ‘‘dashpot’’ model of the MscS gating, the open conformation of the channel is accomplished by a concerted movement of all three TM helices aside from the vertical axis of the channel (Akitake et al., 2005). A transition from the open to the inactivated state, whose conformation is represented by the crystal structure, involves detachment of the TM3 helices from the TM1–TM2 assembly and their collapse to the closed, nonconductive conformation. The detachment of TM3 is stimulated by depolarizing voltage acting on positive charges located in TM1–TM2 and pushing them toward a position that is perpendicular to the plane of the membrane (Akitake et al., 2005).

III. THE MscL CYTOPLASMIC REGIONS AND FUNCTIONING OF THE CHANNEL The modeling of the conformational transitions on gating is much more advanced in MscL than in MscS and it includes also, except rearrangements of the TM helices, structural changes and a role of the channel cytoplasmic N‐ and C‐termini. It is postulated that there are two gates involved in the opening of the channel. Except the main TM gate, there is a second cytoplasmic gate (Sukharev et al., 2001a,b) composed of five ‐helical S1 segments of the cytoplasmic N‐termini (not resolved in the crystal structure and not indicated in Fig. 1B) acting in accordance. The TM gate is proposed to act as a pressure sensor and on application of pressure; this gate permits initial expansion of the channel without its full opening (Betanzos et al., 2002; Sukharev et al., 2005). The cytoplasmic gate that allows full activation of the channel is being connected with TM1s via flexible linkers. According to the model, the applied pressure is transmitted to the S1 segments through the flexible linkers and pulls them apart. The channel may fully open when the interactions between five S1 segments of the cytoplasmic gate break down. The cytoplasmic C‐terminal domain of MscL is formed by a bundle of five helices (Fig. 1B) connected to the TM2 by linkers containing clusters of charged residues RKKEE. According to the present model (Anishkin et al., 2003), the bundle remains stably associated on transition from the closed to open conformation of the channel, serving as a size‐exclusion filter. The investigation of the role of charged cluster revealed, however, that it functions as a proton sensor adjusting the channel sensitivity to membrane tension in a pH‐dependent manner and, therefore, having an influence on the channel gating (Kloda et al., 2006).

300

Koprowski et al.

IV. THE MscS C‐TERMINAL CHAMBER: THE CAGE‐LIKE STRUCTURE AND KINETICS On the basis of available experimental data, the current models of the MscS gating focus mainly on the function of the TM domains, and at present the contribution of the large cage‐like structure formed by its C‐termini is not clearly assessed. The size of the structure and its complexity create main obstacle in its investigation and a subsequent modeling. There is, however, a quite large body of evidence that the MscS large cytoplasmic chamber, unlike that of MscL, is a dynamic structure changing its shape on the channel gating and essential for structural transitions undergone by the channel. The cytoplasmic domain is composed mostly of ‐sheets and surrounds ˚ . Each of seven subunits of the large water‐filled chamber of diameter 40 A the assembly consists of a middle ‐domain and a lower 3‐domain and all of them are linked together by a ‐barrel composed of seven strands that are located at the very ends of C‐termini. The MscS cytoplasmic chamber resembles ‘‘hanging gondola‐like’’ structure of the cytoplasmic T1 domain of the eukaryotic voltage‐dependent potassium channels (Sokolova et al., 2001; Kobertz et al., 2002) or that described for the acetylcholine receptor (Miyazawa et al., 1999). It is well documented now that the cytoplasmic domains of eukaryotic potassium channels including inward rectifiers and those activated by voltage, Ca2þ, or cyclic nucleotides aVect the activity of the associated pores by controlling the ion flow and ultimately providing an additional gate, or by coupling intracellular signals to the channel primary gate (Yi et al., 2001; Roosild et al., 2004). It is of great interest, therefore, to provide an evidence for a possible function of the large cytoplasmic chamber of MscS by investigating its structural rearrangements, link them to the channel gating, and eventually couple to the intracellular signaling pathways. Results obtained independently by diVerent research groups indicate that the MscS cytoplasmic chamber is a flexible structure and it may undergo significant structural rearrangements occurring on the channel transitions from one functional state to another (Koprowski and Kubalski, 2003; Miller et al., 2003; Schumann et al., 2004; Grajkowski et al., 2005). An observation that in the amino acid sequence of MscS, all lysines but one (K60) are situated in the C‐terminus led to device a series of experiments in which the lysines were cross‐linked. It was expected that a cross‐link of lysines from diVerent C‐termini of the channel would hamper or prevent the channel opening, providing its C‐termini being pulled apart during opening. A series of patch‐clamp experiments was performed in which lysine‐specific reagents 1‐ethyl‐3‐(3‐dimethylaminopropyl) carbodiimide hydrochloride—EDC and highly lysine‐specific bis(sulfosuccinimidyl)suberate—BS3 were applied to


301

the cytoplasmic side of MscS in its closed configuration (Koprowski and Kubalski, 2003). Indeed, the cross‐link of the MscS C‐termini yielded inactive channels and the eVect was irreversible. In the same study, it has also been demonstrated that Ni2þ coordination in the MscS‐His6 channels (His‐tag added at the very end of C‐terminus) prevented the channels from opening. The Ni2þ coordination leading to the reversible inhibition of activities was observed in the channel closed state but not in the channel open conformation, suggesting that the closed state of the channel is the only configuration in which the intersubunit coordination of Ni2þ may occur. It has been suggested that the lack of the eVect of Ni2þ application to the open MscS‐His6 channels is due to binding of metal ions to histidines within individual poly‐histidine tag. This observation provided the first experimental evidence that C‐termini may move apart on the channel opening and may be involved in the process of the channel gating. These data were supported shortly by a biochemical study (Miller et al., 2003) that demonstrated in situ—and, accordingly in the closed state of the channel—a cross‐link of cysteines substituting serines 267 (indicated in pink in Fig. 2) located at the bottom of 3‐domain of the cytoplasmic chamber. The cysteines were cross‐linked by o‐phenylene‐1,2‐dimaleimide (o‐PDM) ˚ that cross‐links residues that are capable of adopting positions within 10 A of each other. Strikingly in the crystal structure of MscS that represents open ˚ apart. or inactivated channel conformation, the serines 267 are 20 A On the basis of these data, the authors suggested large flexibility of the MscS cytoplasmic chamber and proposed a model of the closed state of the channel with an assumption that the crystal structure shows the MscS open configuration. According to the model, the closed, nonconductive state would be represented by a more compact configuration of the cytoplasmic chamber in which an eventual collapse of the entire 3‐domain may lead ultimately to an impermeable conformation of the cytoplasmic structure (Edwards et al., 2004). It is, therefore, suggested that the chamber may represent an additional, flexible permeability filter. The model has been referred to as the ‘‘Chinese‐ lantern’’ representation by analogy with a lantern, whose light intensity is related to the extent of its expansion. However, this hypothesis needs experimental exploration. The molecular dynamics simulations that tested transitions from the channel conformation revealed by crystallography to the closed and then to the open configuration have shown that the side openings of the C‐terminal chamber did not assume a completely closed state in none of the simulations. The bottom opening, however, remained closed (did not conduct water molecules) even after the open state of the channel was imposed by application of surface tension (Sotomayor and Schulten, 2004). Another study, in which small cosolvents exerted a diVerent eVect on the

302

Koprowski et al.

FIGURE 2 The structure of MscS showing positions of mutations within the cytoplasmic chamber that aVect channel gating. Double mutations G160R and I162E (indicated in yellow) and single mutation N177C (indicated in orange) in ‐domain of the chamber yielded nonfunctional channels. Truncated channels with deletions below serine 267 were not found in the membrane, suggesting that the channel assembly was impaired (Schumann et al., 2004). The ‐barrel structure (indicated in green) could be deleted but the truncated channel shows altered gating.

MscS gating than the large ones, suggests that the openings of the cytoplasmic chamber remain accessible to them in various conformational states of the channel (Grajkowski et al., 2005). The relevance of the C‐terminal domain in the channel assembly and/or its ability to function has been tested in a series of mutants with deletions within their C‐terminal domains. Deletions at the MscS C‐terminal that were longer than the last 20 residues yielded proteins that did not incorporate into the membrane (Schumann et al., 2004). Removal of the last 20 amino acids was tolerated; the protein was found in the membrane and it formed a functional channel. However, the properties of the truncated channel were altered: the channel activated and inactivated, but the recovery from the


303

inactive state was impaired. After the channel inactivated, it lost its ability to make a transition from the inactivated to the closed state, from which it could reopen again. Interestingly, addition of poly‐histidine tag to the end of the truncated C‐termini promoted a transition back to the closed state and the mutated channel functioned similarly to the wild‐type one (Schumann et al., 2004). This result is reminiscent to the finding that replacement of highly conserved cytoplasmic T1 domain in voltage K (Kv) channels by an artificial tetramerization module restored those channel properties that were missing in the channel lacking T1 (Zerangue et al., 2000). The assembly of the Kv channels lacking the T1 domain was greatly improved in the presence of the artificial tetramerization domain. Since artificial heptamers are not available as yet, it is not possible to investigate if potential diYculties in assembly of the MscS with truncated C‐termini are accountable for the lack of functional channels. Except the channels with truncated C‐termini, some mutants with single or double substitutions within C‐terminal chamber have been tested electrophysiologically. It has been found that the double substitution of highly conserved residues in the middle ‐domain G160R and I162E (marked in yellow in Fig. 2) or the single substitution N177C in the same region (marked in orange in Fig. 2) both yielded nonfunctional channels (Koprowski et al., unpublished data).

V. STRUCTURAL ALTERATIONS OF THE MscS CYTOPLASMIC CHAMBER ON GATING The data mentioned above indicate that the cytoplasmic chamber may undergo large conformational changes on transition from the closed to open state of the channel. By analogy, similar changes in the opposite direction may be expected on transition from the open to inactivated and then back to the closed conformation. In eVort to predict these structural rearrangements, a series of experimental conditions was set, under which changes in the surface of the MscS cytoplasmic chamber and/or in its entire ion‐conducting pathway could be expected, and resulting alterations of the channel kinetics would be observed. It is well known that large molecules (cosolvents) present in the solution surrounding the protein of interest interact with it (TimasheV, 1998, 2002). The interaction can be positive (preferential binding) or negative (preferential exclusion resulting in preferential hydration). Cosolvents that preferentially bind to proteins (urea, guanidine hydrochloride, or propyleneglycol—solubilizers) are known to solubilize and denaturate proteins and promote an expanded, unfolded state. On the other hand, cosolvents that are preferentially excluded from proteins [polyethylenoglycans (PEGs), dextrans, ficoll, and sucrose—stabilizers] fix them in the native, compact state. In an

304

Koprowski et al.

FIGURE 3 Model of MscS conformational changes on activation, inactivation, and closing based on available experimental data and existing models. On the channel activation, the pore opens and both the inner and the outer surfaces of the cytoplasmic chamber increase. The inactivation is associated with a detachment of TM3 helices from TM1 and TM2, which results in a nonconductive pore conformation. The outer surface of the cytoplasmic chamber decreases and the chamber lowers its volume. The crystal structure is interpreted here as a representation of the inactivated conformation. On transition from the inactivated to the closed state, TM3 attaches again to TM1 and TM2 and the inner surface of the cytoplasmic chamber increases. The protein is represented as a schematic cross section so that the individual subunits twisted around MscS symmetry axis are undistinguishable.

approach combining patch‐clamp analysis with a use of various‐size stabilizers, it has been expected that the MscS kinetic states are linked to the conformational changes of the channel in the presence of cosolvents (Grajkowski et al., 2005). It has been found that large cosolvents that cannot enter the channelwater‐ filled cytoplasmic chamber impair channel activation and accelerate its inactivation when applied from the cytoplasmic side but they slow down inactivation when applied from the extracellular side. It has also been found that small cosolvents that can enter the channel, cytoplasmic chamber prevents the channel from opening much stronger than the large ones, having almost no eVect on the inactivation rate. On the basis of crystal structure, possible conformational changes of the channel molecule on transitions between its functional states have been suggested and they can be summarized as follows:


305

1. Activation is associated with an increase of the area of the channel inner surface (the chamber and the TM gate) accessible to small cosolvents, and possibly with an increase of the volume of this entire part of the channel. Large cosolvents that interacted with the outer side of the channel cytoplasmic domains also aVected activation but to a lesser degree. 2. Inactivation is associated with a decrease of the external surface of the cytoplasmic chamber of the channel and probably with a decrease of its volume. The periplasm‐exposed parts of the channel enlarge their surface on inactivation. 3. On closure (a transition from inactivated to closed state) the channel increases its inner surface area. These results are in an agreement with the previous suggestion that in the closed state cytoplasmic domains are in a much more compact conformation then in the crystal structure (Miller et al., 2003). A study utilizing steered molecular dynamics simulation revealed a widening of the channel, when restrains imposed on the channel to keep it in the crystal structure conformation were abolished, and the surface tension was applied (Sotomayor and Schulten, 2004). These data suggest that the channel may become larger than revealed by the crystal structure, and since the pore radius increases in the expanded conformation it may, indeed, represent the open channel state. Figure 3 shows possible rearrangements of MscS on gating and the presented model makes use of all available experimental data and existed models.

VI. CONCLUSIONS AND PERSPECTIVES Since the moment of crystallization of MscS, it has been noticed that the organization as well as the size of its cytoplasmic part is reminiscent of the cytoplasmic domains of eukaryotic channels including the best characterized T1 domain of the potassium channels. It is now well established that the T1 domain from potassium KV channels is involved not only in the channel assembly but also in gating by stabilization of the closed conformation of the channel, and thereby plays a key role in the conformational alterations leading to the channel opening (Cushman et al., 2000; Minor et al., 2000; Jiang et al., 2002). The data presented in this chapter, particularly those showing that certain mutations within the MscS cytoplasmic domains prevent the channel from opening, may in fact indicate the stabilizing role of the chamber in the closed state of the channel. It is not known at present if the MscS assembly is under control of its cytoplasmic domains, however, the channels with the truncated C‐termini are not found in the membrane. Understanding the structural changes in MscS, particularly in the region of its cytoplasmic chamber, may be of great importance in understanding

306

Koprowski et al.

how the MS channels are integrated with cell physiology. Since MscS crystallization, it has been speculated on the possibility that the cytoplasmic chamber may be a docking site for the cytoplasmic regulatory proteins (Bass et al., 2002). It is well documented that cytoplasmic domains of many eukaryotic channels of various types are sensitive to the binding of cytosolic molecules that aVect channel activities (Roosild et al., 2004; Pegan et al., 2005; Lu et al., 2006). It would be of interest to explore this possibility in MscS taking into consideration a complexity, flexibility, and a potential involvement in gating of its cytoplasmic chamber. This is a sound and not new idea, an MS channel as a signal transducer has been proposed in osmotaxis (Martinac et al., 1987; Li et al., 1988). Bacteria respond to the abrupt changes in osmolarity by changing its motile behavior trying to avoid high or low concentrations of solutes. One substance can be either an attractant or a repellent depending on a concentration used and, therefore, it is suggested that the system sensing osmolarity responds de facto to the changes in concentration of water. There are two types of E. coli behavior to osmotic upshifts: one utilizing the chemotactic machinery and the other for which the chemotaxis system is not required. It has been demonstrated that bacterial chemoreceptors that mediate the motile behavior change their spatial organization within the lipid bilayer in response to osmotic stress (Vaknin and Berg, 2005). The other system detecting changes in solute concentration and not dependent on chemotaxis system was not resolved as yet and the involvement of MS channels in this bacterial response was proposed (Li and Adler, 1993). The MscS‐like proteins are found in other species than bacteria, and the MscS‐like protein family is much diversified (Blount et al., 2005). The MscS cytoplasmic chamber is a conserved structure (more than the C‐terminal domain in MscL; Fig. 2), and, therefore, it can be assumed that its role might be of crucial importance for the potential mechanosensory activity of the MscS homologues. Investigation of structural rearrangements of the cytoplasmic chamber of the bacterial channel may indeed lead to better understanding of the principles of mechanotransduction on the level of single molecule and eventually integrate the detailed conformational changes of the chamber with the cell physiology. Acknowledgments Our work is supported by a grant from the Polish Ministry of Education (2 P04C 01627). P.K. is a fellow of Human Frontiers Science Program.

References Ajouz, B., Berrier, C., Garrigues, A., Besnard, M., and Ghazi, A. (1998). Release of thioredoxin via the mechanosensitive channel MscL during osmotic downshock of Escherichia coli cells. J. Biol. Chem. 273, 26670–26674.


307

Akitake, B., Anishkin, A., and Sukharev, S. (2005). The ‘‘dashpot’’ mechanism of stretch‐ dependent gating in MscS. J. Gen. Physiol. 125, 143–154. Anishkin, A., and Sukharev, S. (2004). Water dynamics and dewetting transitions in the small mechanosensitive channel MscS. Biophys. J. 86, 2883–2895. Anishkin, A., Gendel, V., Sharifi, N. A., Chiang, C. S., Shirinian, L., Guy, H. R., and Sukharev, S. (2003). On the conformation of the COOH‐terminal domain of the large mechanosensitive channel MscL. J. Gen. Physiol. 121, 227–244. Barlett, J. L., Lewin, G., and Blount, P. (2004). An in vivo assay identifies changes in residue accessibility on mechanosensitive channel gating. Proc. Natl. Acad. Sci. USA 101, 10161–10165. Bass, R. B., Strop, P., Barclay, M., and Rees, D. C. (2002). Crystal structure of Escherichia coli MscS, a voltage‐modulated and mechanosensitive channel. Science 5598, 1582–1587. Batiza, A. F., Kuo, M. M., Yoshimura, K., and Kung, C. (2002). Gating the bacterial mechanosensitive channel MscL in vivo. Proc. Natl. Acad. Sci. USA 99, 5643–5648. Batiza, A. N., Rayment, I., and Kung, C. (1999). Channel gate! Tension, leak and disclosure. Struct. Fold. Des. 7, R99–R103. Betanzos, M., Chiang, C. S., Guy, H. R., and Sukharev, S. (2002). A large iris‐like expansion of a mechanosensitive channel protein induced by membrane tension. Nat. Struct. Biol. 9, 704–710. Blount, P., Sukharev, S. I., Moe, P. C., Schroeder, M. J., Guy, H. R., and Kung, C. (1996). Membrane topology and multimeric structure of a mechanosensitive channel protein of Escherichia coli. EMBO J. 18, 4798–4805. Blount, P., Iscla, I., Li, Y., and Moe, P. C. (2005). The bacterial mechanosensitive channel MscS and its extended family. In ‘‘Bacterial Ion Channels and Their Eukaryotic Homologs’’ (A. Kubalski and B. Martinac, eds.), pp. 247–258. ASM Press, Washigton DC. Chang, G., Spencer, R. H., Lee, A. T., Barclay, M. T., and Rees, D. C. (1998). Structure of the MscL homolog from Mycobacterium tuberculosis: A gated mechanosensitive ion channel. Science 5397, 2220–2226. Chiang, C. S., Anishkin, A., and Sukharev, S. (2004). Gating of the large mechanosensitive channel in situ: Estimation of the spatial scale of the transition from channel population responses. Biophys. J. 86, 2846–2861. Cui, C., Smith, D. O., and Adler, J. (1995). Characterization of mechanosensitive channels in Escherichia coli cytoplasmic membrane by whole‐cell patch clamp recording. J. Membr. Biol. 144, 31–42. Cushman, S. J., Nanao, M. H., Jahng, A. W., DeRubeis, D., Choe, S., and PfaYnger, P. J. (2000). Voltage dependent activation of potassium channels is coupled to T1 domain structure. Nat. Struct. Biol. 7, 403–407. Edwards, M. D., Booth, I. R., and Miller, S. (2004). Gating the bacterial mechanosensitive channels: MscS a new paradigm? Curr. Opin. Microbiol. 7, 163–167. Edwards, M. D., Li, Y., Kim, S., Miller, S., Bartlett, W., Black, S., Dennison, S., Iscla, I., Blount, P., Bowie, J. U., and Booth, I. R. (2005). Pivotal role of the glycine‐rich TM3 helix in gating the MscS mechanosensitive channel. Nat. Struct. Mol. Biol. 12, 113–119. Glaser, F., Pupko, T., Paz, I., Bell, R. E., Bechor, D., Martz, E., and Ben‐Tal, N. (2003). ConSurf: Identification of functional regions in proteins by surface‐mapping of phylogenetic information. Bioinformatics 19, 163–164. Grajkowski, W., Kubalski, A., and Koprowski, P. (2005). Surface changes of the mechanosensitive channel MscS upon its activation, inactivation, and closing. Biophys. J. 88, 3050–3059. Ha¨se, C. C., Le Dain, A. C., and Martinac, B. (1995). Purification and functional reconstitution of the recombinant large mechanosensitive ion channel (MscL) of Escherichia coli. J. Biol. Chem. 270, 18329–18334.

308

Koprowski et al.

Haswell, E. S., and Meyerowitz, E. M. (2006). MscS‐like proteins control plastid size and shape in Arabidopsis thaliana. Curr. Biol. 16, 1–11. Jiang, Y., Lee, A., Chen, J., Cadene, M., Chait, B. T., and MacKinnon, R. (2002). Crystal structure and mechanism of a calcium‐gated potassium channel. Nature 417, 515–522. Kloda, A., and Martinac, B. (2002). Common evolutionary origins of mechanosensitive ion channels in Archaea, Bacteria and cell‐walled Eukarya. Archaea 1, 35–44. Kloda, A., Ghazi, A., and Martinac, B. (2006). C‐terminal charged cluster of MscL, RKKEE, functions as a pH sensor. Biophys. J. 90, 1992–1998. Kobertz, W. R., Williams, C., and Miller, C. (2002). Hanging gondola structure of the T1 domain in a voltage‐gated K(þ) channel. Biochemistry 39, 10347–10352. Koprowski, P., and Kubalski, A. (1998). Voltage‐independent adaptation of mechanosensitive channels in Escherichia coli protoplasts. J. Membr. Biol. 164, 253–262. Koprowski, P., and Kubalski, A. (2001). Bacterial ion channels and their eukaryotic homologues. Bioessays 23, 1148–1158. Koprowski, P., and Kubalski, A. (2003). C termini of the Escherichia coli mechanosensitive ion channel (MscS) move apart upon the channel opening. J. Biol. Chem. 278, 11237–11245. Landau, M., Mayrose, I., Rosenberg, Y., Glaser, F., Martz, E., Pupko, T., and Ben‐Tal, N. (2005). ConSurf 2005: The projection of evolutionary conservation scores of residues on protein structures. Nucleic Acids Res. 33, W299–W302. Levina, N., Totemeyer, S., Stokes, N. R., Louis, P., Jones, M. A., and Booth, I. R. (1999). Protection of Escherichia coli cells against extreme turgor by activation of MscS and MscL mechanosensitive channels: Identification of genes required for MscS activity. EMBO J. 18, 1730–1737. Li, C., and Adler, J. (1993). Escherichia coli shows two types of behavioral responses to osmotic upshift. J. Bacteriol. 175, 2564–2567. Li, C., Boileau, A. J., Kung, C., and Adler, J. (1988). Osmotaxis in Escherichia coli. Proc. Natl. Acad. Sci. USA 85, 9451–9455. Lu, R., Alioua, A., Kumar, Y., Eghbali, M., Stefani, E., and Toro, L. (2006). MaxiK chanel partners: Physiological impact. J. Physiol. 570, 65–72. Martinac, B., Buechner, M., Delcour, A. H., Adler, J., and Kung, C. (1987). Pressure‐sensitive ion channel in Escherichia coli. Proc. Natl. Acad. Sci. USA 84, 2297–2301. Miller, S., Edwards, M. D., Ozdemir, C., and Booth, I. R. (2003). The closed structure of the MscS mechanosensitive channel. Cross‐linking of single cysteine mutants. J. Biol. Chem. 278, 32246–32250. Minor, D. L., Jr., Lin, Y.‐F., Mobley, B. C., Avelar, A., Jan, Y. N., Jan, L. Y., and Berger, J. M. (2000). The polar T1 interface is linked to conformational changes that open the voltage‐gated potassium channel. Cell 102, 657–670. Miyazawa, A., Fujiyoshi, Y., Stowell, M., and Unwin, N. (1999). Nicotinic acetylcholine receptor at 4.6 A resolution: Transverse tunnels in the channel wall. J. Mol. Biol. 288, 765–786. Moe, P. C., Blount, P., and Kung, C. (1998). Functional and structural conservation in the mechanosensitive channel MscL implicates elements crucial for mechanosensation. Mol. Microbiol. 28, 583–592. Okada, K., Moe, P. C., and Blount, P. (2002). Functional design of bacterial mechanosensitive channels. Comparisons and contrasts illuminated by random mutagenesis. J. Biol. Chem. 277, 27682–27688. Pegan, S., Arrabit, C., Zhou, W., Kwiatkowski, W., Collins, A., Slesinger, P. A., and Choe, S. (2005). Cytoplasmic domain structures of Kir2.1 and Kir3.1 show sites for modulating gating and rectification. Nat. Neurosci. 8, 279–287.


309

Perozo, E., Cortes, D. M., Sompornpisut, P., Kloda, A., and Martinac, B. (2002). Open channel structure of MscL and the gating mechanism of mechanosensitive channels. Nature 6901, 942–948. Pivetti, C. D., Yen, M. R., Miller, S., Busch, W., Tseng, Y. H., Booth, I. R., and Saier, M. H., Jr. (2003). Two families of mechanosensitive channel proteins. Microbiol. Mol. Biol. Rev. 67, 66–68. Roosild, T. P., Le, K. T., and Choe, S. (2004). Cytoplasmic gatekeepers of Kþ‐channel flux: A structural perspective. Trends Biochem. Sci. 29, 39–45. Schumann, U., Edwards, M. D., Li, C., and Booth, I. R. (2004). The conserved carboxy‐ terminus of the MscS mechanosensitive channel is not essential but increases stability and activity. FEBS Lett. 572, 233–237. Sokolova, O., Kolmakova‐Partensky, L., and GrigorieV, N. (2001). Three‐dimensional structure of a voltage‐gated potassium channel at 2.5 nm resolution. Structure 9, 215–220. Sotomayor, M., and Schulten, K. (2004). Molecular dynamics study of gating in the mechanosensitive channel of small conductance MscS. Biophys. J. 87, 3050–3065. Sukharev, S. (2002). Purification of the small mechanosensitive channel of Escherichia coli (MscS): The subunit structure, conduction, and gating characteristics in liposomes. Biophys. J. 83, 290–298. Sukharev, S. I., Martinac, B., Arshavsky, V. Y., and Kung, C. (1993). Two types of mechanosensitive channels in the Escherichia coli cell envelope: Solubilization and functional reconstitution. Biophys. J. 65, 177–183. Sukharev, S. I., Blount, P., Martinac, B., Blattner, F. R., and Kung, C. (1994). A large‐ conductance mechanosensitive channel in E. coli encoded by mscL alone. Nature 6468, 265–268. Sukharev, S., Betanzos, M., Chiang, C. S., and Guy, H. R. (2001a). The gating mechanism of the large mechanosensitive channel MscL. Nature 6821, 720–724. Sukharev, S., Durell, S. R., and Guy, H. R. (2001b). Structural models of the MscL gating mechanism. Biophys. J. 81, 917–936. Sukharev, S., Anishkin, A., Chiang, Ch.‐S., Betanzos, M., and Guy, H. R. (2005). MscL, a bacterial mechanosensitive channel. In ‘‘Bacterial Ion Channels and Their Eukaryotic Homologs’’ (A. Kubalski and B. Martinac, eds.), pp. 259–290. ASM Press, Washigton DC. TimasheV, S. N. (1998). Control of protein stability and reactions by weakly interacting cosolvents: The simplicity of the complicated. Adv. Protein Chem. 51, 355–432. TimasheV, S. N. (2002). Protein‐solvent preferential interactions, protein hydration, and the modulation of biochemical reactions by solvent components. Proc. Natl. Acad. Sci. USA 99, 9721–9726. Vaknin, A., and Berg, C. (2005). Osmotic stress mechanically perturbs chemoreceptors in Escherichia coli. Proc. Natl. Acad. Sci. USA 103, 592–596. Yi, B. A., Minor, D. L., Jr., Lin, Y. F., Jan, Y. N., and Jan, L. Y. (2001). Controlling potassium channel activities: Interplay between the membrane and intracellular factors. Proc. Natl. Acad. Sci. USA 98, 11016–11023. Yoshimura, K., Batiza, A., and Kung, C. (2001). Chemically charging the pore constriction opens the mechanosensitive channel MscL. Biophys. J. 80, 2198–2206. Zerangue, N., Jan, Y. N., and Jan, L. Y. (2000). An artificial tetramerization domain restores eYcient assembly of functional Shaker channels lacking T1. Proc. Natl. Acad. Sci. USA 97, 3591–3595.

CHAPTER 12 Microbial TRP Channels and Their Mechanosensitivity Yoshiro Saimi, Xinliang Zhou, Stephen H. Loukin, W. John Haynes, and Ching Kung Laboratory of Molecular Biology, Department of Genetics, University of Wisconsin, Madison, Wisconsin 53706

I. II. III. IV. V. VI. VII. VIII.

Overview A History TRP‐Channel Research The Mechanosensitivity of Animal TRP Channels Distribution and the Unknown Origin of TRPs TRPY1: The TRP Channel of Budding Yeast Other Fungal TRP Homologues Sequence Information Does Not Explain TRP Mechanosensitivity Conclusions References

I. OVERVIEW The pioneering phenotype‐to‐gene approach that dissects biological pathways has repeatedly led to TRP channels. Follow‐up examinations of these TRPs and their homologues constitute the bulk of current research on this superfamily of fascinating channels, which are central to many aspects of animal physiology. Beyond animals, TRP genes can be found in Paramecium, Tetrahymena, Dictyostelium, Trypanosoma, Leishmania, and other protists, as well as most species of fungi. Thus, these channels and their basic mechanisms have apparently evolved long before the appearance of multicellular animals. Experiments showed that the TRPY1 channel of the budding yeast (Saccharomyces cerevisiae) releases Ca2þ from the vacuole into the cytoplasm when the yeast cell is subjected to a sudden osmotic upshock. Current Topics in Membranes, Volume 58 Copyright 2007, Elsevier Inc. All right reserved.

1063-5823/07 $35.00 DOI: 10.1016/S1063-5823(06)58012-3

312

Saimi et al.

Under patch clamp, TRPY1 displays a 300‐pS cation conductance that rectifies inwardly. Stretch force on the vacuolar membrane, on the order milliNewton per meter, activates TRPY1. TRPY1’s mechanosensitivity examined in vitro explains the osmotically induced Ca2þ release in vivo. TRPY2 and TRPY3, homologues of TRPY1 from other yeasts, have similar properties. Unlike most animal TRPs, the yeast TRP channels can be examined directly with patch clamp for their mechanosensitivity. The cloned TPRY genes can also be manipulated with ease, using yeast molecular genetics. The study of microbial TRPs should be of value in further analysis on the molecular basis of mechanosensitivity.

II. A HISTORY TRP‐CHANNEL RESEARCH There must be more than 50 review papers on TRP channels in the last three years, in English alone. One author wrote 19. The great enthusiasm for these channels stems from their novelty (associated with pepper, menthol, garlic, marijuana, and so on), physiological roles (hot, cold, pain, touch, hearing, osmotic senses, and so on), diseases (retinal degeneration, mucolipidosis, polycystic kidney disease, and glomerulosclerosis), and promises (therapy, drugs, and research grants). The following summary on the historical origins of the current bloom of TRP research should be instructive. The bulk of current research, mostly on mammalian TRPs, is the derivative of some 10 diVerent prospecting adventures. With no known sequence targets to start with, all of these projects independently arrived at diVerent TRP channels. The term TRP, ‘‘transient receptor potential,’’ describes the electroretinographic phenotype of a near‐blind mutant Drosophila isolated in the Pak laboratory in 1975 (Minke et al., 1975), the corresponding gene of which was cloned by Montell et al. in 1985 in the Rubin laboratory. This TRP channel is the founding member of TRPC (C for canonical) and is the crux of phototransduction in insects, although how it is activated in vivo remains obscure despite intensive research (Minke, 2006). In 1997, TRP channels were ‘‘rediscovered’’ in two diVerent contexts. In one, Julius and coworkers used expression cloning to search and found a heat/pain receptor, by using a heat surrogate (the pepper essence capsaicin) as a probe. This vanilloid receptor, VR1, turned out to have clear homology to Drosophila TRP and is now called TRPV1 (Caterina et al., 1997). In the other, the Bargmann Laboratory isolated and analyzed mutant worms (Caenorhabditis elegans) defective in their avoidance of 4‐M fructose. Position cloning of the mutation in one such mutants led to a gene, osm‐9, which is homologous to TRPV1 (Colbert et al., 1997). Expression cloning using a surrogate of cold (menthol) revealed the receptor CMR1, now a TRPM (McKemy et al., 2002).

12. Microbial TRP Channels and Mechanosensitivity

313

In the fly, mutations that cause defects in balance and touch response led to NOMPC, now a TRPN (Walker et al., 2000); those insensitive to pain to PAINLESS, now a TRPA (Tracey et al., 2003). In the worm, mutations causing defect in the males’ ability to locate the vulvas of hermaphrodites during mating were traced to LOV‐1, a homologue of PKD1 that forms channels by associating with PKD2, now TRPP (Barr and Sternberg, 1999). Cloning gene of heritable diseases is the medical equivalent of the phenotype‐ to‐gene forward genetics. Thus polycystic kidney disease was traced to PKD1 and PKD2, now TRPPs (Hughes et al., 1995; Mochizuki et al., 1996). IV was traced to MCOLN1, now TRPML (Sun et al., 2000). In short, each founding member of the TRP subfamily, TRPC, TRPV, TRPN, TRPP, TRPM, or TRPML, was independently discovered by a piece of prospecting research and not through alignment of known TRP sequence. [TRPA was founded by the identification of PAINLESS (Tracey et al., 2003) and through candidate sequence homology (Story et al., 2003) almost simultaneously.] The convergence of these multiple original studies onto the same superfamily of ion channels gives us strong confidence that TRPs are indeed central to many aspects of biology. Note that expression cloning of genes through probe recognition or position cloning of mutations with biological or clinical phenotypes are both ‘‘fishing expeditions,’’ not driven by known sequence targets or concrete preconceived hypotheses. The findings of such ‘‘forward‐genetic’’ studies are therefore more objective and original. Once these molecular targets are found, their sequence homologues can be recognized and used in further research. Commonly, mammalian homologues are heterologously expressed in oocytes or cultured cells and examined biophysically or biochemically. Knockout mice are also generated to examine possible phenotypes. These studies are generically referred to as ‘‘reverse genetics,’’ and constitute the bulk of the current research, much of which is summarized in the 50 reviews mentioned above.

III. THE MECHANOSENSITIVITY OF ANIMAL TRP CHANNELS Each subfamily of animal TRP channels (TRPC, TRPV, TRPN, TRPP, TRPM, or TRPML) has been associated with mechanosensitivity. The evidence for this association varies greatly from case to case. At the organismic level, evidence comes from mutant behavioral phenotypes such as deafness (Kim et al., 2003; Gong et al., 2004), touch‐blind (Walker et al., 2000), osmotactic failure (Colbert et al., 1997), drinking behavior (Liedtke and Friedman, 2003), and so on or from knockout animals’ physiological defects such as bladder malfunction (Birder et al., 2002), and so on. At the cellular and tissue level, circumstantial evidences include the presence of the TRP

314

Saimi et al.

proteins or their mRNAs being in the expected places or at the expected developmental time (Corey et al., 2004). At the molecular level, evidence most commonly comes from experimentation through heterologous expression. Here, mechanosensitivity is commonly indicated by osmotic downshock‐ induced entry of Ca2þ (monitored with a dye) into cultured cells expressing a foreign TRP transgene (Liedtke et al., 2000; Strotmann et al., 2000; Kim et al., 2003; Gong et al., 2004). More direct electrophysiological examinations of animal TRP channels are rare. Under a current clamp, TRPV1 in isolated magnocellular neurosecretory cells has been found to correlate with hypotonically induced spikes (Naeini et al., 2006). Under patch clamp, in a cell‐attached mode, heterologously expressed TRPV4 [previously OTRPC4 (Strotmann et al., 2000) and VR‐OAC (Liedtke et al., 2000)] and TRPV2 (Muraki et al., 2003) were shown to be activated by hypotonicity. Maroto et al. (2005) provided the most direct and convincing evidence for animal TRP‐channel mechanosensitivity. They showed that, when reconstituted into liposome, a TRPC1‐rich detergent‐ solubilized fraction of frog‐oocyte membrane correlates with unitary conductances that are activated by direct suction exerted on the bilayer patch. They also showed that human TRPC1, expressed in oocytes, correlates with a tenfold increase in stretch‐activated current (Maroto et al., 2005). A part of the diYculty in the analysis of animal TRP channels is that they are often located in specialized cells and strategically located even within those cells. For example, the transducing channels for hearing are located near the tips of stereocilia of vertebrate hair cells (Corey et al., 2004) or the sensory cilia of insect chordotonal organ (Kim et al., 2003; Gong et al., 2004). TRPP is located in the primary cilia of renal epithelial cells (Nauli et al., 2003); TRPML in intracellular endosomes and lysosomes (Di Palma et al., 2002); and others in the compound eyes, taste buds, and Merkel cells, Meissner corpuscle, and so on. These locations are currently nearly inaccessible to the patch‐clamp pipette. Since these TRP channels cannot be studied in situ, they are therefore expressed heterologously in arenas such as oocytes or culture cells and examined therein. Results from heterologous experiments may include artifacts such as the contributions (or the lack of such contribution) from host subunits or host enzyme modifications. By contrast, the yeast TRPY1 channels can be directly patch clamped and examined for its mechanosensitivity in its natural location (the vacuolar membrane) with relative ease. See in later section.

IV. DISTRIBUTION AND THE UNKNOWN ORIGIN OF TRPs Classification of TRPs, iterated in most review papers, is by primary‐ sequence comparison and not by biophysical characteristics or biological functions (see below). Primary sequence cannot predict confidently any


315

tertiary or quaternary structures of proteins using current bioinformatics. Thus, without a crystal structure, the commonly cited model of TRP as a tetramer with a funnellike center fitted with a filter is only by analogy to the known crystal structure of Kþ channels (Doyle et al., 1998; Jiang et al., 2002), based on the belief that these cation channels are distantly related (Yu and Catterall, 2004) and will have similar structures. However, primary sequence can predict secondary structures and general topology with some confidence. The sequence of TRP‐gene product is predicted to have six transmembrane (TM) ‐helices with extensive N‐ and C‐terminal domains in the cytoplasm. These cytoplasmic domains contain recognizable regions with proposed functions (e.g., ankyrin repeats, calmodulin‐binding sites) or unknown functions (‘‘TRP box,’’ ‘‘TRPM homology,’’ and so on) that sort the members found in animals into seven subfamilies (TRPC, TRPV, TRPA, TRPN, TRPM, TRPP, and TRPML). The resemblances between these subfamilies are limited. Most similarities are found in the sequence from the predicted TM5 to slightly beyond the C‐terminus of the predicted TM6, a region that comprises the presumed filter and gate (Fig. 1A). Though the visible animals and plants loom large in our mind, they are in fact a small part of the eukaryotic diversity (Embley and Martin, 2006). Currently, taxonomists divide Eukarya into six clusters (Adl et al., 2005), one of which comprises both animals and fungi. Thus, the very diverse kinds of unicells are generally lumped under the nondescript term ‘‘protists.’’ Using the above key sequence (TM5 through TM6) as the criterion, searches in the existing databases recognize TRP‐channel genes without ambiguity in the genomes of Paramecium, Tetrahymena (both ciliates), Dictyostelium (cellular slime mold), Trypanosoma (an agent of African sleeping sickness), and Leishmania (leishmaniasis) (Fig. 1A) (Haynes, unpublished results). Fragments of similarities can also be found in the genomes of Chlamydomonas, (a green flagellate), Plasmodium (malaria), and Thalassiosira (diatom), and so on, though full‐length TRP‐channel genes have not been recognized or assembled from these genomes due to technical diYculties. No experimental work has been reported on these putative TRP homologues in protists at this writing. The same search criterion revealed a TRP‐channel gene in the genome of the budding yeast S. cerevisiae, which has been experimentally studied at length (see below) (Palmer et al., 2001; Zhou et al., 2003). This channel, TRPY1, has homologues in some 30 diVerent fungal genomes of fungi. See below for the relatedness of 18 fungal TRPs (Zhou et al., 2005) (Fig. 5). An additional TRPP‐like channel gene in the Schizosaccharomyces pombe has similarity with the Drosophila TRPP (Palmer et al., 2005). The putative TRP channels in fungal and protist genomes usually do not bear the cytoplasmic features (ankyrin, ‘‘TRP box,’’ and so on) used to


317

distinguish the animal TRP subtypes. This makes it diYcult to fit these channels into the oYcial but animal‐centric classification system (Montell et al., 2002; Clapham et al., 2003). Blast searches using the strictly defined S5‐to‐S6 sequences from all vertebrate‐defined TRP subfamilies have protist TRPs locally aligned with a greater bit score to TRPML, while fungal sequences align with other TRP subtypes (Haynes, unpublished data). The no‐gap bootstrap cladogram (Fig. 1B) drawn from the ClustalW global alignment (Fig. 1A) also shows this same tendency for clustering with diVerent TRP subtypes. Whether this clustering is evolutionarily meaningful cannot be asserted at the moment since available microbial genome sequences are limited and include many reduced genomes of parasites. The criteria we used have so far failed to identify TRP candidates among bacteria and archaea. Since the relationship among the three domains of life (Bacteria, Archaea, and Eukarya) remains unclear at the moment (Embley and Martin, 2006), it would be presumptuous to speculate that TRPs of ‘‘higher’’ animals must have evolved from some prototypes found in bacteria or archaea. One cannot rule out the scenario that the commingled gene pool of primordial cell communities (Embley and Martin, 2006) encoded the first detectors of force and heat, from which the first TRPs were derived.

V. TRPY1: THE TRP CHANNEL OF BUDDING YEAST Long before the molecular biology of TRPY1, Wada et al. (1987) first described an 300‐pS conductance observed with a planar lipid bilayer into which a vacuolar‐membrane fraction of yeast had been reconstituted. Others have observed a similar conductance by patch clamping the vacuolar membrane, after releasing the vacuoles from yeast spheroplasts (Minorsky et al., 1989;

FIGURE 1 An alignment and unrooted cladogram of the major family members of TRP channels. (A) A clustalW (Gonnet 250) alignment made using the program ClustalX. Several representative TRP genes found in various protists were aligned with a member of each major family of TRP channel along with a Ca2þ channel (conserved in multicellular organisms) for comparison. The protist channels were found by Blast searching genomic sequences currently available for each organism listed. The majority of these protist sequences were predicted by automated annotation procedures at the respective sequencing centers. The Leishmania sequence was described (Chennik et al., 2005). (B) A single unrooted bootstrapped cladogram using neighbor joining method (Saitou and Nei) drawn from the clustalW (Gonnet 250) alignment (with all gapped sequence removed) shown above made using the program ClustalX. 1000 possible trees were compared and the numbers represent the number of trees in which the branches shown were present. The Ca2þ channel was selected as the outgroup for the purpose of drawing this tree.

Saimi et al.

318

Bertl and Slayman, 1990; Saimi et al., 1992) ( Fig. 2A). It rectifies inwardly, that is, from the vacuole into the cytoplasm (Fig. 2B). It is cation selective, PNaþ ¼ PKþ PCl . It also passes divalent cations, PCa2þ ¼ PBa2þ > PMg2þ (Zhou, unpublished result), and it passes the physiologically important Ca2þ even when it is the sole cation (Palmer et al., 2001). Vacuolar Ca2þ (mM) or low pH (193 mm behind the growing tip. Addition of 100‐ and 500‐mM Gd3þ decreased open time and the number of events of the MS channels but did not aVect the spontaneous channels. Similar concentrations of Gd3þ only transiently decreased elongation of hyphae for less than 5 min, before growth returned to pretreatment rates and did not change the tip‐high Ca2þ gradient after growth resumed to pretreatment rates (Levina et al., 1995). Rinsing oV the Gd3þ at the end of the experiment did nothing to change the average growth rate. Similar to the conclusion discussed above with Silvetia rhizoids, the two plasma membrane Ca2þ‐permeable MS channels in N. crassa hyphae are not being used to generate the tip‐high Ca2þ necessary for growth (Levina et al., 1995). In fact an intracellular IP3‐activated channel is thought to generate the intracellular tip‐high Ca2þ gradient (Silverman‐Gavrila and Lew, 2001, 2002). The Ca2þ‐permeable MS channels in N. crassa are thought to be involved with turgor regulation or sensing of mechanical stress (Levina et al., 1995).

406

Messerli and Robinson

VII. IS TURGOR NECESSARY FOR ACTIVATION OF MS CHANNELS? Expansion and growth of these cell wall‐enclosed, turgor maintaining systems is dependent on the interplay between the turgor and the restrictive mechanical barriers keeping turgor under control. The majority of the turgor measured under normal growth conditions could simply be oVset by these mechanical barriers. Slight yielding of these barriers either by increasing turgor to overcome the barriers or by directly weakening the mechanical barriers is all that is necessary to open MS channels assuming that there is no slack in the plasma membrane. However, the idea of turgor generated expansion of tip‐growing organisms has been brought into question by the inability to measure turgor in growing hyphae of two water molds, S. ferax and Achlya bisexualis, in increased external osmoticum (Money and Harold, 1993; Harold et al., 1996). We only include discussion of S. ferax as they have been shown to possess stretch‐activated MS channels, while no evidence for MS channels in A. bisexualis exists yet. It was reported that 20 kPa is the smallest increment in oil pressure that could be controlled with confidence and indicating that ‘‘no measurable turgor’’ means that the turgor is between 0 and 20 kPa (Money and Harold, 1993; Harold et al., 1996). Is a reduction in turgor from 400 kPa to

Mechanosensitive Ion Channels, Part A, Volume 58 (Current Topics in Membranes)

Mechanosensitive Ion Channels, Part B, Volume 59 (Current Topics in Membranes)

Mechanosensitive Ion Channels

Chloride Channels, Volume 42 (Current Topics in Membranes)

Current Topics in Membranes & Transport, Volume 21: Ion Channels: Molecular and Physiological Aspects

Current Topics in Membranes & Transport, Volume 21: Ion Channels: Molecular and Physiological Aspects

Aquaporins (Current Topics in Membranes, Volume 51)

Claudins (Current Topics in Membranes, Volume 65)

Ion Channels (Topics in Medicinal Chemistry, 3)

CURR TOPICS IN MEMBRANES & TRANSPORT V15, Volume 15 (Current Topics in Membranes and Transport)

Current Topics in Membranes and Transport, Volume 05

Current Topics in Membranes and Transport, Volume 10

Current Topics in Membranes and Transport, Volume 9

Current Topics in Developmental Biology, Volume 58 (Current Topics in Developmental Biology)

Leucocyte Adhesion (Current Topics in Membranes, Volume 64)

Current Topics in Membranes and Transport, Volume 3

Membrane Protein-Cytoskeleton Interactions, Volume 43 (Current Topics in Membranes)

The Nociceptive Membrane, Volume 57 (Current Topics in Membranes)

Current Topics in Membranes and Transport, Volume 20

Current Topics in Membranes and Transport, Volume 04

Current Topics in Membranes and Transport, Volume 33: Molecular Biology of Ionic Channels

Current Topics in Membranes and Transport, Volume 33: Molecular Biology of Ionic Channels

Structure and Function of Calcium Release Channels (Current Topics in Membranes, Volume 66)

Current Topics in Membranes and Transport, Volume 04

Current Topics in Membranes and Transport, Volume 14

Current Topics in Membranes and Transport, Volume 08

Current Topics in Membranes and Transport, Volume 07

Current Topics in Membranes and Transport, Volume 10

Current Topics in Membranes and Transport, Volume 07

Current Topics in Membranes and Transport, Volume 16: Electrogenic Ion Pumps

Mechanosensitive Ion Channels, Part A, Volume 58 (Current Topics in Membranes)

Mechanosensitive Ion Channels, Part B, Volume 59 (Current Topics in Membranes)