Biocalorimetry 2: Applications of Calorimetry in the Biological Sciences

Biocalorimetry 2

Biocalorimetry 2 Applications of Calorimetry in the Biological Sciences Edited by

John E. Ladbury Department of Biochemistry and Molecular Biology, University College London, UK

Michael L. Doyle Department of Gene Expression and Protein Biochemistry, Bristol-Myers Squibb Pharmaceutical Research Institute, Princeton, NJ, USA

Copyright u 2004

John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England Telephone (+44) 1243 779777

E-mail (for orders and customer service enquiries): [email protected] Visit our Home Page on www.wileyeurope.com or www.wiley.com All Rights Reserved. No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, scanning or otherwise, except under the terms of the Copyright, Designs and Patents Act 1988 or under the terms of a licence issued by the Copyright Licensing Agency Ltd, 90 Tottenham Court Road, London W1T 4LP, UK, without the permission in writing of the Publisher. Requests to the Publisher should be addressed to the Permissions Department, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, or emailed to [email protected], or faxed to (+44) 1243 770620. This publication is designed to provide accurate and authoritative information with regard to the subject matter covered. It is sold on the understanding that the Publisher is not engaged in rendering professional services. If professional advice or other expert assistance is required, the services of a competent professional should be sought. Other Wiley Editorial Offices John Wiley & Sons Inc., 111 River Street, Hoboken, NJ 07030, USA Jossey-Bass, 989 Market Street, San Francisco, CA 94103-1741, USA Wiley-VCH Verlag GmbH, Boschstr. 12, D-69469 Weinheim, Germany John Wiley & Sons Australia Ltd, 33 Park Road, Milton, Queensland 4064, Australia John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01, Jin Xing Distripark, Singapore 129809 John Wiley & Sons Canada Ltd, 22 Worcester Road, Etobicoke, Ontario, Canada M9W 1L1 Wiley also publishes its books in a variety of electronic formats. Some content that appears in print may not be available in electronic books. Library of Congress Cataloging-in-Publication Data Biocalorimetry II : applications of calorimetry in the biological sciences / edited by John E. Ladbury, Michael L. Doyle. p. ; cm. Rev. ed. of: Biocalorimetry. Includes bibliographical references and index. ISBN 0-470-84968-1 (cloth : alk. paper) 1. Calorimetry. 2. Biology — Technique. [DNLM: 1. Biology — methods. 2. Calorimetry. ] I. Title: Biocalorimetry 2. II. Title: Biocalorimetry two. III. Ladbury, John E., 1960- IV. Doyle, Michael L., 1959- V. Biocalorimetry. VI. Title. QH324.9.C3B5482 2004 572’.43 — dc22 2004008487 British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN 0 470 84968 1 Typeset by Dobbie Typesetting Ltd, Tavistock, Devon Printed and bound in Great Britain by TJ International Ltd, Padstow, Cornwall This book is printed on acid-free paper responsibly manufactured from sustainable forestry in which at least two trees are planted for each one used for paper production.

Contents Preface List of Contributors

Part I General Introduction 1

Applications of Biocalorimetry: Binding, Stability and Enzyme Kinetics Ronan O’Brien and Ihtshamul Haq 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9

Introduction Principles of isothermal titration calorimetry (ITC) Applications of ITC in the life sciences Thermodynamic signatures of non-covalent interactions Thermodynamic discrimination (TD) ITC as a tool for studying drug–DNA interactions ITC as a tool for studying protein–DNA interactions The application of calorimetry for examining hydration effects The use of ITC for studying the kinetics and thermodynamics of enzyme catalysis 1.10 Principles of differential scanning calorimetry (DSC) 1.11 Applications of DSC in the life sciences 1.12 Thermodynamic stability 1.13 Shelf life versus thermodynamic stability 1.14 Specific and non-specific binding 1.15 Intrinsic and extrinsic macromolecular stability 1.16 Oligomerization 1.17 The use of DSC for examining nucleic acid helix?coil transitions 1.18 Summary Acknowledgements References

Part II Isothermal Titration Calorimetry 2

xi xiii 1

3 3 7 8 10 10 13 15 17 18 21 22 23 24 25 25 25 26 30 30 31

35

Isothermal Titration Calorimetry: A Tutorial James A. Thomson and John E. Ladbury

37

2.1 2.2 2.3 2.4

37 38 40 42

Introduction Thermodynamic characterization Instrumentation Raw data

vi

CONTENTS

2.5 Basic considerations for experimental set-up 2.6 Data analysis 2.7 Summary Application notes Acknowledgement References

3 The Application of Isothermal Titration Calorimetry to Drug Discovery Geoff Holdgate, Stewart Fisher and Walter Ward

43 50 53 53 57 57

59

3.1 Introduction 3.2 Overview of the drug discovery process 3.3 Experimental measurement of thermodynamic binding parameters 3.4 ITC in drug discovery 3.5 Summary

59 60

References

78

4 Dissecting the Thermodynamics of DNA–Protein Interactions Torleif Ha¨rd 4.1 Introduction 4.2 Model systems 4.3 Comparison with the hydrophobic effect 4.4 Protonation and charged–charged hydrogen bonds 4.5 Dissection of the binding entropy 4.6 Entropy contributions to the Sso7d–DNA interaction 4.7 Entropy contributions to the GCN4–DNA interaction 4.8 Discussion Acknowledgements References

5 Salt Effects in Ribonuclease–Ligand Interactions: Screening or Competitive Binding? Kenneth P. Murphy, Travis T. Waldron and Greta L. Schrift 5.1 Introduction 5.2 Anion binding to a protein–protein complex 5.3 Charge–charge interactions in ribonuclease binding 5.4 Conclusions Acknowledgement References

6 Thermodynamics–Structure Correlations of Sulfonamide Inhibitor Binding to Carbonic Anhydrase Daumantas Matulis and Matthew Todd 6.1 6.2 6.3 6.4 6.5

Introduction Identification of protonation reactions occurring upon binding Observed thermodynamics of inhibitor binding to CA Energetics of inhibitor protonation Sulfonamide ‘anion’ binding thermodynamics

63 67 78

81 81 82 83 84 86 87 88 90 90 90

93 93 94 96 103 104 104

107 107 108 111 119 123

CONTENTS

7

8

9

vii

6.6 Correlations between structures and the thermodynamics of sulfonamide binding to CA 6.7 Conclusions References

125 131 132

Energetics of the Interaction of Human Acidic Fibroblast Growth Factor with Heparin and the Functional Analogue Myo-Inositol Hexasulfate Mercedes Guzma´n-Casado, Marı´a M. Garcı´a-Mira, Pedro CanoSoldado, Guillermo Gime´nez-Gallego, Jose M. Sanchez-Ruiz and Antonio Parody-Morreale

133

7.1 Introduction 7.2 Thermodynamic parameter derived from ITC experiments 7.3 Discussion Acknowledgement References

133 137 142 148 148

Thermodynamics of SH2 Domain Binding Gabriel Waksman, Sangaralingam Kumaran and Olga Lubman

151

8.1 Introduction 8.2 SH2 domains and human pathologies 8.3 Structure and ligand binding 8.4 SH2-domain-binding inhibitors 8.5 Conclusions References

151 152 155 166 168 169

Titration Calorimetry as a Tool to Determine Thermodynamic and Kinetic Parameters of Enzymes M. Lucia Bianconi

175

9.1 Introduction 9.2 ICT of enzyme catalysed reactions 9.3 Summary Acknowledgements References

175 177 184 184 184

Part III Differential Scanning Calorimetry

187

10 Energetics of Site-Specific DNA Recognition by Integrase Tn916 Stoyan Milev, Hans Rudolf Bosshard and Ilian Jelesarov

189

10.1 Introduction 10.2 Conformational stability of INT-DBD and the target DNA duplex 10.3 Thermodynamics of complex formation measured by titration calorimetry

189 190 194

viii

CONTENTS

10.4 Thermal dissociation and unfolding of the protein–DNA complex 10.5 The heat capacity change of protein–DNA association 10.6 Discussion References

195 198 201 202

11 Linkage Between Temperature and Chemical Denaturant Effects on Protein Stability: the Interpretation of CalorimetricallyDetermined m Values Beatriz Ibarra-Molero, Raul Perez-Jimenez, Raquel Godoy-Ruiz 203 and Jose M. Sanchez-Ruiz 11.1 Introduction 11.2 Linkage between temperature and chemical denaturant effects on protein stability 11.3 Calorimetrically determined urea m values 11.4 Calorimetrically determined guanidine m values 11.5 Concluding remarks Acknowledgement References

12 Thermodynamic Indications of the Molten Globule State of Cytochrome c Induced by Hydrophobic Salts Ali A. Moosavi-Mohavedi and Jamshid Chamani 12.1 Introduction 12.2 Thermodynamic description of molten globule state 12.3 Molten globule structure detection by isothermal titration calorimetry (ITC) 12.4 Enthalpic relationship with electronic transfer of molten globule state 12.5 Indication of molten globule state by differential scanning calorimetry (DSC) Acknowledgement References

13 Microcalorimetry as Applied to Psychrophilic Enzymes Salvino D’Amico, Daphne´ Georlette, Tony Collins, Georges Feller and Charles Gerday 13.1 13.2 13.3 13.4 13.5

Introduction Cold adaptation Uniformly unstable enzymes Enzymes with local flexibility Thermal inactivation recorded by isothermal titration calorimetry 13.6 Microcalorimetric determination of enzyme kinetic parameters 13.7 Conclusion References

203 205 208 210 212 212 212

215 215 218 221 223 225 227 227

231 231 231 233 234 236 238 239 240

CONTENTS

14 An Autosampling Differential Scanning Calorimeter for Study of Biomolecular Interactions Valerian Plotnikov, Andrew Rochalski, Michael Brandts, John F. Brandts, Samuel Williston, Verna Frasca, and Lung-Nan Lin 14.1 Introduction to DSC 14.2 Description of instrument 14.3 Materials and methods 14.4 Results for ribonuclease binding to anionic inhibitors 14.5 Discussion References Appendix. Data analysis (reprinted from reference 10)

Index

ix

241 241 242 245 246 248 249 250

253

Preface A rigorous understanding of any biological system that undergoes some change requires quantification of its thermodynamic and kinetic properties. Combined with structural detail, these data can be assimilated to enable a picture of the mechanism of the change (for example, going from the free to bound state in a biomolecular interaction, or from the folded to the unfolded state of a biomolecule). Isothermal titration calorimetry (ITC) and differential scanning calorimetry (DSC) provide the thermodynamic data required for developing this understanding. The microcalorimetric instrumentation currently available can provide a highly accurate measurement of the heat (or molar change in enthalpy, DH) associated with the change in the system. This direct determination of one of the thermodynamic properties is unique to the calorimetric method. Using heat as the probe of the extent of change, the equilibrium constant K can be determined. After determination of DH and K a full characterization of the thermodynamic properties of the change in the system can be calculated. This provides a unique description of the event and allows for comparison with other similar data. This allows wholesale comparison of absolute affinities of interactions, the stability of biomolecules, changes in non-covalent bond formation etc. As a result calorimetry has become one of the primary tools in the arsenal of the biological scientist. The inaugural Applications of Bio-Calorimetry (ABCI) meeting took place at St Anne’s College, Oxford, UK, in September 1996. This meeting drew together a group of scientists from all over the world with a large variety of professional affiliations to discuss applications of the relatively recently commercially available microcalorimeters (both isothermal titration calorimeters and differential scanning calorimeters). The meeting provided a base for users of the techniques of different levels of experience to meet. It was clear at this first such meeting that it was necessary to provide for the dissemination of ideas, data and methodologies in this rapidly evolving area. As a result the impetus to establish meetings on a more regular basis was born. Since that meeting there have been additional meetings focusing on the application aspects of the microcalorimetric instrumentation (ABCII (1998), Halle, Germany; ABCIII (2002), Dublin, Ireland; ABCIV (to be held 2004),

Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

xii

PREFACE

Budapest, Hungary) as well as meetings in which trends and developments in the field have been highlighted (Trends in Calorimetry (2001), Philadelphia, PA, USA; Trends in Calorimetry (2003), Boston, MA, USA). As a result of these meetings a forum for the microcalorimetry field has been established. The book Biocalorimetry: Applications of Calorimetry in the Biological Sciences (Eds. J. E. Ladbury and B. Z. Chowdhry) was published as a tribute to the development of the calorimetric field and to summarize the state of the art after the inaugural meeting. Since there are now several hundred microcalorimeters in various laboratories throughout the world, and the techniques have become almost routine in areas as diverse as drug discovery to whole cell interactions, we felt that it would be appropriate to update the field on current applications and trends. This book is based on a selection of issues and applications that were presented by contributors to the meetings in 2002 and 2003. As such we are indebted to all the participants at both of these meetings. Clearly, this volume cannot hope to cover all possible applications for calorimetric instrumentation; however, we attempt to provide a flavour of the types of study possible and also an idea of how the data derived can be interpreted and utilized in a better understanding of biological science. This book pays particular attention to areas where calorimetry has made a significant impact in recent years and attempts to highlight the work of people we consider as making key contributions to the modern field. The editors have decided, despite attempts to unify calorimetric data under IUPAC directives, to allow individual authors to present their work using the system adopted in their country. This means that some data are reported in different units. It should thus be noted that 1 cal ¼ 4.18 J. Finally, both editors would like to thank the following people, who have had some impact on this work: Dr Matthew Cliff, Dr Mark A. Williams, Dr Aldo F. Gutierrez, and Professor E. Freire. J.E.L. would like to thank especially Professor J. M. Sturtevant and Professor B. Z. Chowdhry, without whom this work would not exist. John E. Ladbury Michael Doyle

List of Contributors M. Lucia Bianconi Departamento de Bioquı´ mica Me´dica, Universidade Federal do Rio de Janeiro, Brazil Hans Rudolf Bosshard Department of Biochemistry, University of Zurich, Switzerland John F. Brandts MicroCal, LLC, Northampton, MA, USA Michael Brandts MicroCal, LLC, Northampton, MA, USA Pedro Cano-Soldado Departamento de Quı´ mica-Fı´ sica e Instituto de Biotecnologı´ a, Universidad de Granada, Spain Jamshid Chamani Tehran, Iran

Institute of Biochemistry and Biophysics, University of

Tony Collins Laboratory of Biochemistry, Institute of Chemistry B6, University of Lie`ge, Belgium Salvino D’Amico Laboratory of Biochemistry, Institute of Chemistry B6, University of Lie`ge, Belgium Georges Feller Laboratory of Biochemistry, Institute of Chemistry B6, University of Lie`ge, Belgium Stewart Fisher AstraZeneca, R&D Boston, Waltham, MA, USA Verna Frasca MicroCal, LLC, Northampton, MA, USA Marı´ a M. Garcı´ a-Mira Departamento de Quı´ mica-Fı´ sica e Instituto de Biotecnologı´ a, Universidad de Granada, Spain Daphne´ Georlette Laboratory of Biochemistry, Institute of Chemistry B6, University of Lie`ge, Belgium Charles Gerday Laboratory of Biochemistry, Institute of Chemistry B6, University of Lie`ge, Belgium Guillermo Gime´nez-Gallego Centro de Investigaciones Biolo´gicas, CSIC, Ramiro de Maeztu 9, 28040 Madrid, Spain Raquel Godoy-Ruiz Departamento de Quı´ mica-Fı´ sica, Universidad de Granada, Spain

xiv

LIST OF CONTRIBUTORS

Mercedes Guzma´n-Casado Departamento de Quı´ mica-Fı´ sica e Instituto de Biotecnologı´ a, Universidad de Granada, Spain Ihtshamul Haq Centre for Chemical Biology, Department of Chemistry, University of Sheffield, UK Torleif Ha¨rd Swedish NMR Centre at Go¨teborg University and Institute of Medical Biochemistry, Go¨teborg, Sweden Geoff Holdgate Protein Function, AstraZeneca, Macclesfield, Cheshire, UK Beatriz Ibarra-Molero Departamento de Quı´ mica-Fı´ sica, Universidad de Granada, Spain Ilian Jelesarov Switzerland

Department of Biochemistry, University of Zurich,

Sangaralingam Kumaran Washington University School of Medicine, Department of Biochemistry and Molecular Biophysics, Saint Louis, MO, USA John E. Ladbury Department of Biochemistry and Molecular Biology, University College London, UK Lung-Nan Lin

MicroCal, LLC, Northampton, MA, USA

Olga Lubman Washington University School of Medicine, Department of Biochemistry and Molecular Biophysics, Saint Louis, MO, USA Daumantas Matulis Stoyan Milev Switzerland

3-Dimensional Pharmaceuticals, Exton, PA, USA

Department of Biochemistry, University of Zurich,

Ali A. Moosavi-Mohavedi Institute of Biochemistry and Biophysics, University of Tehran, Iran Kenneth P. Murphy City, IA, USA Ronan O’Brien

Department of Biochemistry, University of Iowa, Iowa

MicroCal LLC, Milton Keynes, UK

Antonio Parody-Morreale Departamento de Quı´ mica-Fı´ sica e Instituto de Biotecnologı´ a, Universidad de Granada, Spain Raul Perez-Jimenez Departamento de Quı´ mica-Fı´ sica, Universidad de Granada, Spain Valerian Plotnikov MicroCal, LLC, Northampton, MA, USA Andrew Rochalski MicroCal, LLC, Northampton, MA, USA

LIST OF CONTRIBUTORS

xv

Jose M. Sanchez-Ruiz Departamento de Quı´ mica-Fı´ sica e Instituto de Biotecnologı´ a, Universidad de Granada, Spain Greta L. Schrift Department of Biochemistry, University of Iowa, Iowa City, IA, USA James A. Thomson Biophysics Group, Pfizer Global Research and Development/Agouron Pharmaceuticals, San Diego, CA, USA Matthew Todd 3-Dimensional Pharmaceuticals, Exton, PA, USA Gabriel Waksman Institute of Structural Molecular Biology, Birkbeck College and Department of Biochemistry and Molecular Biology, University College London, London, UK Travis T. Waldron Department of Biochemistry, University of Iowa, Iowa City, IA, Walter Ward

AstraZeneca, R&D Alderley, Macclesfield, Cheshire, UK

Samuel Williston MicroCal, LLC, Northampton, MA, USA

Part I General Introduction

1 Applications of Biocalorimetry: Binding, Stability and Enzyme Kinetics Ronan O’Brien and Ihtshamul Haq

1.1 Introduction The past 50 years has witnessed the accumulation of a huge amount of highresolution structural data on biological molecules and their complexes.1 These NMR and X-ray crystallography studies have been crucial for our improved understanding of cellular processes. Structural biology has provided a threedimensional picture of nucleic acids and proteins and thus helped us to appreciate how structure is related to function. Nearly all the chemistry of life is mediated through non-covalent interactions between molecules, and some of the most important of these reactions involve proteins and nucleic acids. High-resolution structures of these complexes are important for elucidating the shape complementarity of interacting surfaces and the exact spacial orientation of functional groups on interacting molecules. Thus numbers of hydrogen bonds are enumerated and hydrophobic or electrostatic interactions are defined. However, examination of structural information alone tells us nothing about the energetic forces that drive complex formation or are responsible for maintaining the folded conformations of biological macromolecules in solution. Structural detail also fails to explain the folding pathway followed by biological molecules or the mechanism of bimolecular complex formation. Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

4

APPLICATIONS OF BIOCALORIMETRY

Therefore, in order to gain a fuller understanding of biological processes it is necessary to combine three-dimensional structure with an understanding of the underlying thermodynamics and the kinetics of the process. All three factors can then be used to explain and understand biological function. This approach is necessary because any effort to modulate cellular activity, for example in therapeutics, is predicated on a sufficiently detailed understanding of the molecular events that take place. In recent years there have been an ever increasing number of detailed and rigorous thermodynamic studies on a number of biologically important systems. This has been especially true for studies involving the stability and binding reactions of proteins and nucleic acids.2–6 In part these efforts have been aided by advances in the techniques of molecular biology and chemical synthesis. Cloning technology has meant that recombinant proteins can be over-expressed and purified in quantities that make them amenable to biophysical studies. Similarly, nucleic acid and peptide synthesis, whilst not inexpensive, is routine, and oligonucleotides or peptides of any given sequence can be prepared. Another factor that has underpinned the increasing number of thermodynamic studies of biomolecules has been the advent of commercially available high-sensitivity calorimetry.7–10 Calorimeters are instruments that directly and quantitatively measure the heat of a reaction. Whilst they have been used by chemists since the 18th century, it has only been in the last four decades that electronics, materials design and temperature sensing technology have advanced to the point where calorimeters have been sensitive enough to allow the meaningful study of biological molecules and their interactions. The field of biological calorimetry was pioneered by John Brandts,7,8 Julian Sturtevant,11,12 Stan Gill13 and Peter Privalov.14 By designing and building their own instruments in their respective laboratories these workers laid the foundations for the subsequent commercial development of biological calorimetry. They were also amongst the first to present directly measured thermodynamic data for protein and nucleic acid unfolding and binding interactions. Thermodynamic studies of biologically important processes are broadly concerned with specific and non-specific molecular recognition and macromolecular stability as summarized in Figure 1.1. These concepts are fundamental to molecular organization and cellular function. Understanding the thermodynamics of the equilibria shown in Figure 1.1 can reveal the nature of the energetic forces that drive complex formation or maintain threedimensional structure. In addition, the relative contributions of specific molecular interactions to the overall free energy of the process can be revealed, as well as the variation of these interactions with temperature, pH and ionic strength. Therefore thermodynamic studies provide a wealth of information that is not only of fundamental scientific interest but also of

INTRODUCTION

5

immense practical utility in terms of biotechnology, medicine and drug design. A corollary to these facts is that experimental approaches are required that will yield accurate, reliable and rapid methods for directly obtaining thermodynamic information. Calorimetry is a technology that fulfils all these criteria. The starting point of most thermodynamic studies is the experimental determination of the equilibrium association (or binding) constant (Kb ).

Figure 1.1 A schematic diagram to illustrate two basic phenomena that are often the focus of biophysical studies: biopolymer thermal stability and biopolymer–ligand interactions. All these equilibria can be studied using a combination of ITC and DSC. The blue spheres in the DNA complex represent Zn2+ ions

6

APPLICATIONS OF BIOCALORIMETRY

Hence the Gibbs free energy (DG) of the process under study, at a given temperature, can be evaluated using the standard thermodynamic relationship shown in Equation (1.1): DG ¼
RT ln Kb

ð1:1Þ

where T is the absolute temperature in Kelvin and R is the gas constant. It should be stated here that Kb is the reciprocal of the equilibrium dissociation constant (Kd ), which is a more commonly used and intuitive term (Kb ¼ 1=Kd ). For a binding interaction the equilibrium constant can be determined accurately in a number of ways, most of which rely upon measuring the concentrations of free and bound ligand. For example, optical spectroscopy or equilibrium dialysis are often convenient techniques. For a more detailed thermodynamic study it is necessary to examine the temperature dependence of the free energy change, which is reflected in the enthalpy change (DH) for the interaction. By measuring the enthalpy it is possible to quantify the exact enthalpic and entropic (DS) contributions to the overall observed free energy using Equation (1.2): DG ¼ DH
TDS

ð1:2Þ

This is important since two interactions with similar affinities and structures can have different enthalpic and entropic components to their free energies. By understanding these different energetic contributions to overall free energy and relating these to structure and other factors such as hydration, a detailed understanding of the interaction can be gained. It is of course possible to estimate the enthalpy change for an equilibrium by examining the temperature dependence of the equilibrium constant and then using the van’t Hoff relationship shown in Equation (1.3): ln Kb ¼


DH DS þ RT R

ð1:3Þ

By constructing a plot of ln Kb versus 1=T a straight line is produced that has a slope of
DHvH =R. However, it is now apparent that this experimental approach is often flawed since the enthalpy may not be temperature independent. The curvature expected in van’t Hoff plots as a result of a non-zero heat capacity change is often lost in experimental noise. Linear fits to such data yield systematically biased estimates of the enthalpy change, which vary considerably from directly measured calorimetric enthalpies. A number of studies have highlighted these discrepancies and demonstrated the inherent superiority of the calorimetric method for obtaining enthalpy values.15,16 The two different processes illustrated in Figure 1.1 can be studied using two different but complementary calorimetric techniques. Order–disorder

PRINCIPLES OF ISOTHERMAL TITRATION CALORIMETRY (ITC)

7

transitions, e.g. protein unfolding or nucleic acid melting, can be studied using differential scanning calorimetry (DSC). Ligand–macromolecule or macromolecule–macromolecule binding interactions can be examined using isothermal titration calorimetry (ITC). In this article we only present a brief overview of these two techniques, since the mechanics and operation of these instruments have been presented elsewhere in the literature.7,8,17–22 In addition several reviews have recently been published that describe detailed ITC and DSC methodology for conducting biophysical studies on biological molecules.23–26 There are also a number of other excellent articles that review the theory and applications of biological calorimetry.27–30 The principal aim of this chapter is to highlight the diversity of potential applications of calorimetry in the life sciences. We aim to demonstrate the versatility and power of these techniques for elucidating thermodynamic information on biologically significant systems. By using selected examples from the literature as well as work conducted in our own laboratories we will illustrate the utility and necessity of thermodynamic studies when examining macromolecular stability and molecular recognition. These data have direct application in rational drug design programmes as well as in biotechnology. The scope of calorimetric investigations covers factors such as hydration effects, dissection of observed free energy into component terms and establishing structure–thermodynamic relationships. Information of this type is proving to be very useful in helping us to understand key molecular events in greater detail. More recently, ITC has been used for examining the kinetics and thermodynamics of enzyme catalysed reactions. This relatively novel application of ITC, which has enormous potential for studying enzyme catalysis, will also be discussed here. By no means do we intend this to be an exhaustive list of possible applications of calorimetry. Rather, it is a survey of some of the uses of ITC and DSC that are particularly relevant in the life sciences, and for the most part these studies involve proteins and nucleic acids.

1.2 Principles of isothermal titration calorimetry (ITC) ITC directly and without the need of a predetermined model measures the enthalpy change for a bimolecular binding interaction at a constant temperature. This methodology relies upon a differential cell system within the calorimeter assembly. The reference cell contains only water or buffer, while the sample cell contains the macromolecule or ligand as well as a stirring device. Injection of the second component into the sample cell produces heat effects that are due to stirring and dilution of the ligand, dilution of the macromolecule and the heat of the interaction. The amount of power that must

8

APPLICATIONS OF BIOCALORIMETRY

be applied to actively compensate for the heat produced in the sample cell, after an injection of ligand, is measured directly. The applied thermal power as a function of time that is required to return the calorimeter to its steady state, following an injection, is directly proportional to the heat of reaction.

1.3 Applications of ITC in the life sciences ITC is a powerful technique for examining biological interactions because a well designed experiment will generate the binding enthalpy (DHb ), the equilibrium binding constant (Kb ) and the reaction stoichiometry (n) in a single two hour experiment. By performing these experiments over a range of temperature the change in heat capacity (DCp ) can be also be determined. ITC can be used to study almost any bimolecular complex formation with a defined stoichiometry since the absorption or evolution of heat is a universal property of all chemical reactions; it requires no ‘reporter’ groups and is performed free in solution. A major advantage of ITC is that there is no real assay development and as such affinities can be determined rapidly for a range of systems with very little prior knowledge other than sample concentration. The universal nature of the technique is reflected in the vast array of systems that have been interrogated using ITC to date. Examples of the types of biological system that can be studied by ITC include protein–protein and protein–peptide interactions involved in such diverse processes as cell signalling31–34 (e.g. kinases33,34); Alzheimer’s disease;35 transcription;36 chaperones37,38 and muscle contraction.39 Recent reviews have described the utility of the technique in the study of protein–drug,40 drug–DNA,2 protein–DNA6,41 and protein–carbohydrate interactions.42 ITC is not restricted in any way by the molecular weight of the ligand. For example, many studies have focused on examining the interaction of metal ions with biological molecules. Typically, this involves metal cofactor–protein interactions such as cation binding to calmodulin, parvalbumins and troponin C.43 ITC has also been used to determine critical micelle concentrations and other de/micellization parameters of bile salts and detergents.44,45 The experimental design of calorimetric titrations allows for an accurate determination of the reaction stoichiometry that is independent of the binding affinity. This characteristic has two major benefits: firstly as a quality control (QC) procedure and secondly as a method for determining absolute reaction stoichiometries. Many assays underestimate the affinity if some of the material is inactive. Accurate stoichiometry data are ideal for optimization of purification and manufacturing protocols and for testing protein constructs. A protein construct may have all the requirements for wild type affinity, but a higher propensity to unfold or aggregate. This could result in an

APPLICATIONS OF ITC IN THE LIFE SCIENCES

9

underestimate of the affinity by many techniques and yield incorrect structure–activity relationships (SARs). In biochemical studies it is often necessary to chemically modify a protein molecule. This may be required for down-stream assays or for examining the effects of active site modification. However, such chemical modifications can have two distinct effects: (a) alteration in the functional groups and therefore sterics of the active site and (b) increased propensity for the protein to denature during purification. Without being able to independently assess binding affinity and stoichiometry these two consequences of chemical modifications could not be distinguished. ITC is unique in its ability to separate the determination of these two parameters in a single experiment. Some examples of chemical modifications that are often used, especially in the pharmaceutical industry, include the addition of a reporter group for optical or radio-assays, the pegylation process employed to improve pharmacokinetic properties of biopharmaceuticals and the addition of linkers required in the preparation of microarray plates and chip technologies that are common in proteomics research. Binding between biomolecules and ligands that involve multiple interactions and hence multiple stoichiometries are conveniently studied using ITC. In systems that involve multiple binding events interactions can occur at two or more separate sites independently or through cooperative interactions. In either case binding at each individual site on the same molecule is described by a separate dissociation constant. Carefully designed ITC experiments allow each separate value of Kd to be quantified. Examples where this analysis has been applied include the interaction of the PI3 Kinase Zap 70 subunit with a bisphosphorylated peptide46 and the interaction of glycoconjugates to lectins.47 The increased use of ultrasensitive ITC has led to a growing body of thermodynamic data in the literature that details directly determined enthalpy estimates. This has allowed correlations between thermodynamics and functional/structural detail to be revealed. The dominant contributions to the energetics of binding are from hydrogen bonds, hydrophobic interactions, conformational changes and changes in molecular flexibility and electrostatics. Most of these have distinctive thermodynamic profiles, which can be used to distinguish them from one another. The change in entropy and enthalpy as a function of temperature is dictated by the change in the constant pressure heat capacity (DCp ). Empirical relationships, derived from small molecule heats of transfer data, have been developed that relate DCp to changes in solvent accessible surface area (SASA) upon binding.48 These relationships have been validated for a number of systems involving biopolymers, such as protein folding/unfolding, protein–ligand interactions and drug–DNA interactions. This is an important step change since for the first time it is possible to directly relate structural information to a bulk

10

APPLICATIONS OF BIOCALORIMETRY

thermodynamic parameter. This now makes it possible to undertake rational drug design using both a structural and thermodynamic basis.

1.4 Thermodynamic signatures of non-covalent interactions Observed enthalpies arise largely as a result of changes in interatomic interactions,49,50 the most important in biological systems being the hydrogen bond. The magnitude of the interaction enthalpy is dependent on bond lengths and bond angles. However, the sign indicates whether there is a net favourable (negative) or unfavourable (positive) redistribution of the hydrogen bond network between the reacting species (including solvent). Hydrophobic interactions are related to the relative degrees of disorder in the free and bound systems and therefore these interactions are reflected in the entropy change. The release of water molecules from a ‘wet’ surface to the bulk solvent is a common source of favourable entropy. This, coupled with the inability of non-polar groups to hydrogen bond with surrounding water molecules, is the main reason for the strong energetic influence of hydrophobicity in biology. The thermodynamic signature of this force is typically characterized by a small enthalpy, either positive or negative, and a favourable (positive) entropy. A large negative DCp is thought to arise from the accommodation of non-polar groups by water and is therefore another useful indicator of hydrophobic interactions. Conformational changes are entropically unfavourable.40,51 Large unfavourable entropies are often indicative of an ‘induced fit’ during the interaction. Such modes of binding often play an important physiological role. For example, large conformational changes may be required for allosteric processes common in signalling pathways or receptor binding.52,53 ‘Rigid body’ interactions also incur an entropic penalty, but to a lesser extent. This is because there are always losses in the rotational and translational degrees of freedom, or flexibility when two molecules are brought together to form a complex.48 All of these processes result from a delicate balance of enthalpic and entropic forces. However, trends emerging from the literature support the idea that one will be dominant for a specific type of interaction under approximately physiological conditions. It must also be made clear that other factors, such as electrostatics, which will be specifically discussed later, should not be ignored.

1.5 Thermodynamic discrimination (TD) Using these energetic signatures it is often possible to determine the energetic source of the interaction. Figure 1.2 is a schematic representation of a range of

THERMODYNAMIC DISCRIMINATION (TD)

11

thermodynamically distinct interactions all with a Kd of about 0:1 mM (equivalent to a binding free energy, DGb , of
40 kJ mol
1 ). These profiles can be used to thermodynamically discriminate between binding reactions, infer a function and be used in the drug discovery process in quantitative SAR (QSAR) and lead optimization programmes. A molecular interpretation is given for each. In Scheme (A) the dominant negative enthalpy suggests that there are a large number of favourable hydrogen bond contacts or van der Waals interactions between the biomolecule and ligand. The unfavourable entropy implies a conformational change in either or both of the molecules. An unopposed DH of
100 kJ mol
1 would result in a Kd of 3 attomolar (if TDS ¼ 0). For comparison, an unopposed energy of only 40 kJ mol
1 less (equivalent to two hydrogen bonds54) would result in a 10 million-fold reduction in the affinity to 30 picomolar. Thus these relatively small changes in entropy and enthalpy can potentially result in huge changes in the affinity of an interaction. The large positive entropy in Scheme (B) signifies that this reaction is dominated by solvent rearrangement and hydrophobic forces. This type of profile is often observed in interactions that involve nucleic acids due to the displacement of site specifically bound water molecules from the minor groove, hydrophobic transfer of ligand from bulk solvent to the DNA binding site and release of condensed counterions.2 In Scheme (C) there is a small favourable redistribution of the hydrogen bonding network and a modest overall hydrophobic contribution.

Figure 1.2 A schematic diagram to show the different thermodynamic profiles that might be observed for distinct types of molecular interaction often observed in biology: (A) a high degree of hydrogen bonding in addition to conformational changes; (B) here binding is dominated by hydrophobic interactions; (C) a small favourable contribution from hydrogen bonding along with a hydrophobic contribution

12

APPLICATIONS OF BIOCALORIMETRY

Thermodynamic discrimination allows us to differentiate between systems and sometimes gain an overview of the dominant forces responsible for binding. It has been identified as a method for choosing HIV protease inhibitors as drug candidates.51 This involved designing second-generation inhibitors less reliant on entropic factors for tight binding so that they could be more conformationally flexible and adapt better when faced with naturally occurring mutations on the target protein. Other applications include the differentiation between intercalating and groove binding DNA targeted drugs24 and identifying agonist and antagonist properties in drug–receptor binding.52 The natural progression from this is to identify, dissect out and thermodynamically discriminate the energetics of each individual microscopic event involved in the binding process. This is done by two distinct but complementary methods. The first separates the binding based on structural criteria (QSAR) and the second is based solely on energetic considerations. The QSAR approach is particularly effective when coupled to high-resolution structural data. The method involves mutating a protein, changing the base sequence of a nucleic acid, modifying a functional group on a potential drug, altering the head group or aliphatic chain of a phospholipid and so on. Finally, the thermodynamic profiles (DDH, TDDS and DDG) are compared.40,55 If done in a methodical way this approach can be used to dissect out and identify individual microscopic events that have thermodynamically distinctive profiles and can be used to chart an energetic map of a binding interface. This allows for the identification of those features that are essential and non-essential for function. Such information cannot be gained using methods based solely on structural or affinity considerations.40 Scanning alanine mutagenesis is a commonly used approach to map binding sites in the absence of structural information. Thermodynamic studies have shown that such minor structural changes, even at interfacial sites, are often associated with compensatory changes in the enthalpy and entropy and therefore minimal changes in affinity are observed. These go unnoticed in conventional assays but are obvious using ITC since enthalpy and affinity are measured simultaneously. If enthalpy–entropy compensation is not observed then it is reasonable to assume that the mode of binding has changed.56,57 This approach, where enthalpy–entropy compensation is used as a probe of novel binding interactions, is of value to the pharmaceutical industry for use in identification of new drug classes.40 An alternative strategy for evaluating the thermodynamic contributions of defined microscopic events is to disturb the energetics of the system. This may be achieved by changing the pH and salt concentration, which affects electrostatic interactions. In addition, altering the temperature or adding cosolutes affects the hydrophobicity of the system. The effects of altering these solution conditions on overall free energy can provide considerable insight

ITC AS A TOOL FOR STUDYING DRUG–DNA INTERACTIONS

13

into the different thermodynamic contributions to binding. By carrying out this type of analysis it is possible to be more definitive about the molecular origins of the observed thermodynamic profile for any given interaction. For instance, a measured enthalpy may include contributions from protonation events, rather than hydrogen bonding; or a positive entropy may be derived from the release of counterions upon binding rather than solvent release. Both of these phenomena are discussed in the case studies below.

1.6 ITC as a tool for studying drug–DNA interactions It is probably true to say that, to date, rational drug design has only met with limited success in terms of developing new and effective drug treatments for various disease states. This is because rational drug design programmes have tended to be over-reliant on structural information without an adequate consideration of the thermodynamics of the interaction. Several years ago we embarked on a quest to try and quantify different energetic contributions to the binding free energy for drug–DNA interactions. In order to dissect the binding free energy it is necessary to define at least five component terms; these are (1) DGconf , a free energy contribution that arises from conformational changes in the drug or DNA during binding; (2) DGrþt , an unfavourable free energy effect due to losses of rotational and translational degrees of freedom during complex formation; (3) DGhyd comes from the hydrophobic transfer of the free drug to the DNA binding site; (4) DGpe , which is a favourable free energy contribution that arises from coupled polyelectrolyte effects during the binding of cationic ligands; (5) DGmol is the contribution of non-covalent molecular interactions such as hydrogen bond formation, van der Waals interactions etc. Being able to quantify each of these five terms would yield a detailed insight into a binding interaction and elucidate the important energetic factors that might best be modulated in order to achieve greater affinity and selectivity. In fact, all of these free energy terms can be estimated for any given drug–DNA interaction by using a combination of ITC data for the interaction and semi-empirical/theoretical approaches described in the literature. A comprehensive and detailed review of each of the free energy terms defined above as well as some literature values for each of them is contained within reference 58. The first attempt to carry out such a detailed partitioning exercise on any drug–DNA interaction was carried out in 1997. In this study we examined the binding of a model minor groove binding drug (Hoechst 33258) to a short DNA oligonucleotide containing one specific binding site.59 In the first instance we measured the equilibrium binding constant using fluorescence titrations and determined the overall DGobs , which was found to be
48:9ð2:5Þ kJ mol
1 . A proportion of this free energy is due to unfavourable

14

APPLICATIONS OF BIOCALORIMETRY

effects arising from losses of rotational and translational degrees of freedom when the Hoechst 33258 ligand binds to the DNA binding site (DGrþt ). This term is equal to
TDSrþt , and whilst there is some debate over the exact value of DSrþt Spolar and Record have empirically derived a value of þ209 ð42Þ J mol
1 K
1 .48 Using this value DGrþt can be estimated for the Hoechst 33258–DNA interaction as þ62:3 kJ mol
1 . From X-ray structures of the drug–DNA complex and the free oligonucleotide it is clear that there is little or no conformational change in either component upon complex formation. Therefore, in this case DGconf can be set as zero. It is therefore clear that in order for the observed binding to occur the remaining free energy terms must be favourable and large enough to overcome any unfavourable contributions. The hydrophobic contribution to binding free energy can be estimated based on semi-empirical relationships derived from heats of transfer of small molecules from the liquid to aqueous environments. Based on these data one possible way of quantifying the hydrophobic contribution is using the relationship DGhyd ¼ ð80+10ÞDCp .60 ITC allows DGhyd to be measured in a simple and direct way, since it can be used to determine the temperature dependence of DHb . By using calorimetry to measure the binding induced change in heat capacity we determined that DGhyd is
110 kJ mol
1 . A convenient way of obtaining the polyelectrolyte contribution to binding free energy is to examine the salt concentration dependence of Kb and then carry out analysis of the data using polyelectrolyte theory.61,62 For the Hoechst 33258–DNA interaction we found DGpe was
7:4 kJ mol
1 . Therefore, from the above analysis we can conclude that the remaining contribution to the overall observed binding free energy that arises from non-covalent molecular interactions is þ6:2 kJ mol
1 . This conclusion was somewhat surprising given the emphasis placed on hydrogen bonding and other molecular interactions in structural studies. In this particular case these interactions actually constitute a small unfavourable contribution to overall free energy. By far the largest driving force for binding is the hydrophobic transfer of the drug from bulk solvent. This is also consistent with the overall thermodynamic picture that emerges for this interaction, i.e. positive enthalpy, positive entropy and negative heat capacity change. The lesson from this study is that, whilst molecular interactions are undoubtedly important for recognition, they might not always provide the energetic driving force for the interaction. In this particular case it is likely that favourable free energy for hydrogen bond formation between the drug and DNA is at the expense of hydrogen bond breakage between site specifically bound water molecules in the DNA minor groove. Therefore, the affinity of DNA binding drugs might best be improved by modulating hydrophobicity. The general thermodynamic picture elucidated in this initial study has now been confirmed in a several other drug–DNA systems. These include the

ITC AS A TOOL FOR STUDYING PROTEIN–DNA INTERACTIONS

15

binding of cationic diphenylfuran compounds to the DNA minor groove63 as well as a number of DNA intercalators.64 In all of these studies the use of ITC to directly and accurately measure thermodynamic quantities was central in revealing the underlying energetic contributions to drug–DNA binding. This expanding database of thermodynamic data will be of benefit in rational drug design efforts.

1.7 ITC as a tool for studying protein–DNA interactions A major aim in biochemistry and molecular biology is to understand the precise nature of protein–DNA interactions. This is especially the case in terms of understanding the factors that dictate or influence sequence specificity. This is important not least because of the fact that gene regulation, transcription, translation and DNA replication are mediated through numerous protein–nucleic acid interactions. Malfunctions in this delicate molecular machinery can often lead to diseases such as cancer. Restriction endonucleases are very useful proteins for probing sequence specific protein– DNA recognition using biophysical and structural approaches. This is because this class of enzyme exhibits a large degree of specificity for its target DNA. This clearly has to be the case given the biological function of these enzymes. Recently we have undertaken thermodynamic studies of the interaction between the type II restriction endonuclease MunI and a DNA duplex possessing a single specific binding site.65 MunI is a very useful model protein since in the absence of Mg2+ it is able to reversibly and specifically bind to its recognition sequence. Previous studies have reported that at least two acidic amino acid residues at the active site are important in modulating sequence specificity.66 The presence of these ionizable groups at the active site means that the protein–DNA interaction will be thermodynamically linked to the release and uptake of protons. This phenomenon can be readily studied using ITC by examining the pH dependence of the interaction. Figure 1.3 shows an example of ITC binding data for this system. As expected the protein dimer binds with a 1:1 stoichiometry to the DNA duplex and in solutions containing 200 mM NaCl at pH 6.5 the binding constant was found to be  1  106 M
1 . The binding enthalpy under these conditions at 19.5 8C was found to be
80 kJ mol
1 . One of the principal aims in this study was to evaluate linked protonation effects. ITC is very useful in this endeavour since it can be used to examine the pH dependence of the binding constant. The correct way to analyse these data has been previously reviewed67 and so will not be reproduced here. We found that our Kobs versus pH data were best fitted when we used a model that assumed protonation of ionizable groups within two distinct pKa ranges.

16

APPLICATIONS OF BIOCALORIMETRY

Figure 1.3 ITC data for the binding of MunI to the DNA duplex d(GCCAATTGGC)2 at 19.58C in 200 mM NaCl and pH 6.5. The upper panel shows two sets of data: heats for the injection of DNA into buffer (top) and heats for the injection of an identical DNA solution into a MunI solution (bottom). These peaks were integrated and plotted as a binding isotherm in the lower panel. The solid line is a fit to the data using a single set of identical binding sites model

The fact that protons are taken up or released by the buffer upon complex formation means that the observed binding enthalpy measured using ITC must contain contributions from protonation. In fact the experimentally observed binding enthalpy can be defined as shown in Equation (1.4): DHb ¼ DH0 þ zðNÞDHion

ð1:4Þ

CALORIMETRY FOR EXAMINING HYDRATION EFFECTS

17

where DH0 is the intrinsic binding enthalpy that would be measured if DHion were zero, DHion is the ionization enthalpy of the buffer and zðNÞ is the change in the number of protons bound by the protein in going from free to bound state. The values of DH0 and zðNÞ can be easily determined using ITC by measuring DHb at constant pH but in a number of different buffers with different values for DHion . A plot of DHb versus DHion yields a straight line with the slope equal to zðNÞ and an intercept of DH0 .67 Experiments of this type conducted on the MunI system showed that at each of the pH values studied protons are taken up from the buffer during complex formation. Evaluation of the intrinsic binding enthalpy using this approach is useful since it allows different protein–DNA systems to be compared directly without the influence of protonation effects. This calorimetric study showed that protonation plays an important role in binding such that Kb increases about 20-fold when the pH is decreased from 9.0 to 5.5.

1.8 The application of calorimetry for examining hydration effects Water is an important component of DNA and protein structure and therefore it is necessary to account for solvation effects when examining macromolecular stability and biomolecule–ligand interactions. However, the interaction of water and ions with biological molecules is complicated and the thermodynamic consequences of hydration effects can be difficult to rationalize. Currently, there is a lack of detailed thermodynamic studies that are designed to quantify the role of water in binding interactions. Therefore it can be problematic to include a consideration of hydration effects in drug design strategies. One possibility for assessing the role of water in binding interactions as well as nucleic acid/protein stability is to combine the use of calorimetry with the osmotic stress technique. The application of osmotic stress to a solution using neutral solutes is useful for studying water uptake/release since there is a thermodynamic linkage between osmotic pressure dependence and functionally significant changes in hydration of a macromolecular system.68 By adding neutral solutes such as PEG or dextran to a solution containing the biological molecule osmotic stress is created and movement of water away from the substrate (i.e. into the higher osmotic pressure of the bulk solvent) is favoured. Therefore, if the macromolecule undergoes a conformational change where the molecules expand and water must be taken up to cover the additional surface area, then osmotic stress will inhibit the conformation change. Conversely, if there is a net loss of solvent exposed surface area, then water must be removed into bulk solvent and an increased osmotic stress will favour

18

APPLICATIONS OF BIOCALORIMETRY

the process. Most reactions of interest that involve changes in the hydration of the molecule can be monitored using the osmotic stress technique. ITC is a convenient method for obtaining this type of datum since the variation of the binding constant with osmotic pressure (concentration of osmolyte) can be evaluated. To date osmotic stress has been used to measure the changes in macromolecular hydration for several different reactions involving biological molecules. The effects of altering water activity by the addition of co-solutes on melting of duplex and triplex DNA has been investigated.69 Here changes in the number of bound waters as DNA unfolds were evaluated and the concomitant effect of the free energy of DNA melting was determined. The release/uptake of water and counterions has also been addressed using osmotic stress methods for drug–DNA interactions70,71 as well as TATA binding protein–DNA interactions.72,73 Here ITC was used to differentiate the effects of water release and cation binding. An important addition to the repertoire of techniques that can be used to interrogate protein and nucleic acid systems in order to gain insights into hydration effects is pressure perturbation calorimetry (PPC).74,75 This novel technique, developed at MicroCal, is designed to be used in conjunction with the ultra-sensitive VP-DSC.8 PPC works by evaluating the thermal coefficient of expansion of the partial volume of a solute (
a). This is achieved by changing the applied pressure above the solution of protein or DNA under study. The measured heat of the sample with respect to buffer can be used to calculate a
. These data are useful because they can be used to quantify accessible surfaces where biomolecules and biomolecule–ligand complexes interact with solvent. PPC directly measures volume changes that result from heat induced conformational changes. Therefore, the relationship between volumetric changes, structure and solvation can be conveniently evaluated. It has recently been successfully applied to lipid micellar systems.76

1.9 The use of ITC for studying the kinetics and thermodynamics of enzyme catalysis To understand any biochemical phenomenon it is necessary to study the enzyme reactions that make up that system. The aim is to learn how and why the enzyme recognizes its substrate and the mechanism it employs in catalysing the reaction to form the product. Modern techniques in molecular biology, sequencing and proteomics have led to an upsurge in the number of enzymes that can be routinely expressed as recombinant proteins. Substrates or substrate analogues can also be produced using the same technology or chemically synthesized. Once the two components have been obtained it is then necessary to develop an assay technique that allows the enzyme catalysed

KINETICS AND THERMODYNAMICS OF ENZYME CATALYSIS

19

reaction to be studied. When these data are combined with structural studies a picture of the enzyme mechanism can emerge that allows potential inhibitors to be developed based on a rational approach. However, development of suitable assays is often problematic, since many assay technologies rely upon spectroscopic or radiolabelled probes. The presence of these reporter groups can often produce unwanted effects on enzyme activity. In an effort to address these problems ITC has been successfully applied to quantitatively monitor enzymatic reactions in order to obtain the associated kinetic constants that describe the system. In order to measure enzyme kinetics it is necessary that an observable event accompany the transformation of substrate into product. ITC can be used to measure enzyme kinetic parameters because the thermal power generated as the reaction proceeds is a direct and sensitive observable event. The use of ITC to measure the heat generated during an enzymatic reaction is well established.77–82 The rate of a reaction is directly proportional to the thermal power, defined as the heat (Q) produced as a function of time (dt): power ¼ dQ=dt. The minimum response time of the MicroCal VP-ITC is about 15 seconds and therefore the kinetics of many processes are slow enough to be studied with this instrument; this means that the observable thermal power need not be corrected for the time constant of the calorimeter.82 By using the normal titration mode of the ITC, multiple injections of substrate can be made, providing multiple rate determinations under steady-state conditions within a single experiment. Furthermore, the high sensitivity of modern ITC instruments means the amounts of enzyme required are often similar to those used in spectrophotometric assays, but generally higher than the amounts required in radioactivity assays. The theoretical basis of using ITC for determining enzyme kinetics has been outlined in three excellent papers.79,80,82 These authors have shown that the amount of heat involved in converting n moles of substrate to product is given by the following expression: Q ¼ n DHapp ¼ ½Ptotal V DHapp

ð1:5Þ

where DHapp is the total molar enthalpy for the reaction, determined experimentally, P the concentration of product generated and V the volume of the reaction solution, i.e. the cell volume. It can be shown from Equation (1.5) that measuring the thermal power generated by the enzyme as it catalyses conversion of substrate gives a measure of the reaction rate since power ¼

dQ d½P ¼ V DHapp dt dt

ð1:6Þ

where d½P=dt is equal to the rate of product formation, i.e. the rate of reaction. Equation (1.6) can be rearranged to give

20

APPLICATIONS OF BIOCALORIMETRY

rate ¼

1 dQ V DHapp dt

ð1:7Þ

From Equation (1.7) it is clear that in order to obtain a Michaelis–Menten plot it is necessary to use the calorimeter to measure two parameters. (1) The total molar enthalpy must be determined in experiments where there is sufficient enzyme in the cell to convert all the injected substrate into product in a given time period. These experiments give rise to peaks where the baseline response returns to the same value after the substrate injection as before the injection. Integration of these peaks with respect to time yields the total heat produced by the reaction and dividing this total heat by the amount of substrate converted gives the total enthalpy change. (2) The power generated (dQ=dt) must be determined at different substrate concentrations. An alternative strategy for obtaining enzyme kinetic parameters involves continuous rate measurements after a single injection of substrate at concentrations higher than Km . In these experiments thermal power is monitored as the substrate is completely depleted. At any given time the reaction rate can be determined using Equation (1.7) and the concentration of substrate at any given time can be determined from the integral of the heat evolved. Therefore plots of rate versus substrate concentration can be plotted to give a continuous kinetic curve. We are currently employing ITC to examine a number of enzyme catalysed reactions. As a prerequisite we felt it was necessary to compare kinetic parameters for an enzyme catalysed reaction that had been measured using ITC and analogous spectrophotometric assays. This exercise can validate the use of ITC for assaying an enzyme using a known enzyme–substrate complex before analysing novel substrates. We have studied a serine/threonine phosphatase (PP1-g), which catalyses the direct attack of water during the hydrolytic removal of phosphate groups. PNPP is a model substrate that is often used in conventional spectroscopic assays for this enzyme. We have used this substrate to obtain kinetic parameters using both the spectroscopic assay and ITC. Figure 1.4 shows the Michaelis–Menten curves obtained using the two techniques and these data show that the techniques yield very closely similar kinetic parameters. ITC has the potential to become a ubiquitous tool for examining enzyme kinetics. A further advantage of the calorimetric method is that the experiments simultaneously provide thermodynamic information and kinetic data. Determining the molar enthalpy for the reaction as a function of buffer ionization enthalpy can yield important information on release or uptake of protons during catalysis. In many cases these linked effects are likely to be important in understanding enzyme mechanisms. In addition, the effects of inhibitors can be easily studied using ITC and values for Ki can be determined.

PRINCIPLES OF DIFFERENTIAL SCANNING CALORIMETRY (DSC)

21

Figure 1.4 Rate versus substrate concentration data for the hydrolysis of PNPP by PP1-g. Two sets of data are shown (normalized for enzyme concentration), first rate data from ITC experiments (solid squares) and second the analogous data from spectrophotometric assays (open squares). The line is a non-linear least-squares best fit to the spectroscopic data. The fitting was carried out in Origin 5.0 using the Michaelis–Menten equation

1.10 Principles of differential scanning calorimetry (DSC) A differential scanning calorimeter continuously measures the apparent specific heat of a system as a function of temperature. Therefore, DSC can be used to examine a heat induced phase transition or conformational change.8 A DSC instrument contains two cells suspended in an adiabatic jacket and connected by various heating and temperature/power sensing circuits. During a normal experimental set-up the reference cell is filled with buffer and the sample cell is filled with identical buffer plus macromolecule (e.g. protein, nucleic acid, lipid etc; cell volume ¼ 0:5–1 mL) and the temperature is increased in the range 0.1–1008C. During a heat induced endothermic transition the temperature of the sample cell falls behind that of the reference since some of the energy required is used to induce the transition rather than heat the solution. This lag is detected and additional electrical power is supplied to the sample cell in order to compensate. This additional energy is proportional to the energy associated with the thermally induced transition. Knowledge of the solute concentration permits the conversion of the observed

22

APPLICATIONS OF BIOCALORIMETRY

electrical energy against temperature profile to a curve corresponding to an excess heat capacity versus temperature plot. A single DSC experiment can provide a large amount of thermodynamic information, much of which cannot be obtained by any other technique. Equation (1.8) shows that integration of the experimental heat capacity curve yields the calorimetric transition enthalpy (DHcal ): ð ð1:8Þ Cp dT ¼ DH0 This calorimetrically determined enthalpy is model independent and therefore does not depend on the nature of the transition. The temperature at which excess heat capacity is at a maximum defines the transition temperature (Tm ). Differences in the initial and final baselines provide a measure of the heat capacity change that accompanies the transition. Equation (1.9) shows that by converting the experimental data into a Cp,xs =T versus T curve the entropy change for the transition can be determined from the area under such a curve. ð DS ¼ ðCp,xs =TÞdT ð1:9Þ Using these data the value of DG can be evaluated at any temperature. DSC experiments suitably analysed can also provide important information on the cooperativity of a transition. This can be achieved by comparing the model-dependent van’t Hoff enthalpy (obtained by shape analysis of the calorimetric data) and the calorimetric enthalpy. If DHvH ¼ DHcal then the transition proceeds in a two-state manner and meaningful thermodynamic data can be obtained by examining the temperature dependence of an equilibrium property. The ratio of DHvH =DHcal provides a quantitative insight into the nature of the transition; specifically, it provides a measure of the fraction of the structure that melts as a single thermodynamic entity, i.e. it defines the size of the cooperative unit. This is a unique advantage of DSC in the study of biological molecules.

1.11 Applications of DSC in the life sciences DSC is routinely used to study an entire range of biomolecular interactions, protein stability, lipid phase transitions, surfactant micellization, nucleic acid ‘melts’ and stability of liquid biopharmaceuticals as well as less defined cellular systems. Macromolecular structures stabilized by the cooperation of numerous weak forces undergo conformational or phase transitions upon heating; significant information about these structures can be derived from high-sensitivity DSC.83 DSC has an enormous range of applications and can be used semi-quantitatively for comparing stability of molecules and for

THERMODYNAMIC STABILITY

23

detecting both specific and non-specific binding. In addition DSC can be used to quantify affinities, enthalpies, entropies and heat capacities. In some cases it is possible to gain insights into the multimeric nature of the system as well.

1.12 Thermodynamic stability Figure 1.5 shows the reversible thermal unfolding of a representative monomeric protein. From this Tm , DHcal and DCp are all measured directly and hence DG and DS can be derived. The DG and DS values are therefore model dependent and require validation. One of the main advantages of DSC is that this validation can be achieved using the data from the DSC experiment itself. Specifically, this involves comparing the measured enthalpy (DHcal ) with the model dependent van’t Hoff enthalpy (DHvH ). Disagreement between the two enthalpies means that the thermodynamic model used to analyse the data is inappropriate and that other factors need to be considered. Noncalorimetric methods do not have this internal test for the correctness of the fitting model. Once the correct model has been determined the entire range of thermodynamic data can be extracted.

Figure 1.5 A DSC trace for the reversible unfolding of a monomeric protein. The maximum of the heat capacity curve is the Tm and is 608C. The dashed lines are the linear extrapolations of the pre- and post-transition baselines into the transition region. The difference between these values at Tm is the DCp . The thick black line is the theoretical ‘progress’ baseline computed by the software supplied by MiroCal with the VP-DSC instruments. The area between the DSC trace and the progress baseline is the calorimetric enthalpy, DHcal

24

APPLICATIONS OF BIOCALORIMETRY

1.13 Shelf life versus thermodynamic stability The term ‘stability’ means different things to different people. Some are predominantly concerned with the operational stability of a protein, i.e. its shelf-life, whereas others will think more in terms of the strength of interactions maintaining a native 3D structure. DSC can be used to study both. Quantification of the energetics responsible for maintaining the native structure involves equilibrium thermodynamic analysis whereas the operational stability is controlled, additionally by kinetic parameters. In 1954 Rufus Lumry and Henry Eyring proposed a feasible mechanism to account for protein denaturation;84 this is summarized in the following scheme: N

K

k2

! U
! D

(1:10)

where N is the native state, U is the unfolded state and D is the irreversibly denatured state. The transition between N and U is controlled by an equilibrium constant K and the rate of formation of D from U is dictated by a rate constant k2. The arrows in this scheme show that the first process is reversible and the second is a unidirectional, irreversible process. The longevity of the protein is dependent on K and k2 whereas the thermodynamic stability is only dependent on K. Conditions can be found for many small proteins (420 kD) in which k2 is so slow that the protein is operationally immortal, i.e. it will ultimately be destroyed by external factors such as bacterial proteases due to imperfect aseptic technique rather than intrinsic weaknesses. These not only have a long shelf-life but are also more useful as models for studying the thermodynamics of protein unfolding. Unfortunately many proteins will unfold irreversibly. This means that these proteins have a limited shelf-life and extracting thermodynamic data for their unfolding is difficult. There have been many discussions in the literature about the thermodynamic treatment of irreversible systems.85 Treatments to extract rates of protein inactivation from DSC have become established and have been successfully applied to a number of systems.86,87 The single most important finding is that, as with reversible systems under thermodynamic rather than kinetic control, the Tm remains an exceptionally good probe for comparing relative stabilities.88 The Tm is both conceptually straightforward and very informative. In its simplest form it is the temperature at which the DSC trace is at a maximum (Cp, max ). It represents an accessible experimental window into the stability of the molecule. An increase in the Tm of a transition demonstrates an increase in the stability of the system whereas a decrease infers a destabilization. This appears trivial, but the number of applications for such an assay is enormous.

OLIGOMERIZATION

25

Many of these applications can be categorized into three major groups: (a) specific and non-specific binding; (b) intrinsic and extrinsic macromolecular stability; (c) oligomerization.

1.14 Specific and non-specific binding Macromolecules and macromolecular structures can be stabilized by both specific and non-specific interactions. There are numerous articles in the literature that use Tm as a probe for specific binding. These include protein– small molecule,89 drug–DNA,90 DNA–PNA,91 peptide–lipid92,93 and intramolecular protein domain interactions.94 Non-specific binding measured in this way has been particularly useful in the study of micellar systems such as detergent lipid mixtures95 and DNA–lipid interactions.96 This DSC method is an indirect approach for measuring affinities but is particularly useful for systems with poorly defined stoichiometries and low ligand solubilities and where ultratight binding (4109 M) is evident.22,97 Its applicability has been broadened further by the recent development of a DSC integrated with an autosampler for high capacity protocols such as screening.22

1.15 Intrinsic and extrinsic macromolecular stability A major challenge for many protein biochemists is to find conditions in which a protein is soluble, stable and functional. There are normally two major approaches to this problem. The first is to optimize the extrinsic factors such as pH, ionic strength and stabilizing agents such as sugars, polyols or dithiothrietol (to maintain reduced thiol groups). The second involves protein engineering to increase the intrinsic stability of the molecule. By combining these approaches a protein can be produced that will remain functional long enough to study. DSC measured Tm shift experiments are an ideal method for screening and identifying the optimal conditions. Indeed, Tm has been shown to be predictive of the long-term stability of a protein and its propensity to aggregate.98 As such this assay is now widely used in industry for identifying optimal formulation conditions for liquid biopharmaceuticals.

1.16 Oligomerization In the first two categories the semi-empirical information obtained from Tm shift analysis is valid regardless of the reversible or irreversible nature of the transition. However, it should be stated that for useful comparisons to be

26

APPLICATIONS OF BIOCALORIMETRY

made protein concentrations should be held constant. To test for the presence of oligomers the concentration needs to be varied and therefore this method is only truly applicable to reversible systems, but it has been identified as valid for some irreversible systems as well. The Tm of a unimolecular process should be independent of sample concentration. If this is not so then either the native or the unfolded state must be multimeric. Any two-state protein unfolding process or phase transition can be represented by the following expression: AX !BY

ð1:11Þ

where A and B are the pre- and post-transition states, respectively and the subscripts X and Y refer to the oligomeric state of the species. If X is larger than Y, e.g. there is a dissociation of a dimer to a monomer, then Tm will increase with concentration, whereas a decrease indicates that state B is larger than A. These observations are explained by Le Chatalier’s principle and are in fact intuitive if one considers that a native dimer is more stable than a monomer at higher concentrations. Following this logic, if aggregation is the source of irreversible protein denaturation then the Tm should decrease with increased concentration. If aggregation is apparent (i.e. precipitation) and the Tm increases with concentration then dissociation must occur before aggregation. Therefore, thermodynamic principles apply and the protein can be confidently assigned as multimeric even though the unfolding process is irreversible. As mentioned above an extension of these conclusions has led to the use of DSC in formulations.98 This ability to assess the propensity of molecules to aggregate has been used to study gelation processes.99 In addition this application of DSC has enormous potential in the study of amyloid plaques that are central to understanding neurodegenerative disorders such as Alzheimer’s disease.

1.17 The use of DSC for examining nucleic acid helix!coil transitions Nucleic acids were first subjected to analysis by DSC in the early 1960s.100–102 This pioneering work was limited to examining a small number of synthetic polymers and DNA samples purified from biological material. This is because, at the time, obtaining highly pure nucleic acids of defined sequence was a major problem due to lack of appropriate synthetic methodologies. However, despite these limitations the thermodynamic data obtained by these workers have been subsequently confirmed and validated as DSC technology has advanced and the availability of purified nucleic acids has increased. During the last two decades major advances have been made in thermodynamically characterizing the DNA helix!coil transition using calorimetry (as well as

DSC FOR EXAMINING NUCLEIC ACID HELIX!COIL TRANSITIONS

27

UV melting) in a number of different laboratories.103 These data have been crucial in helping us understand the nature of the molecular interactions that are responsible for maintaining nucleic acid secondary structure. A major practical benefit of this work has been the development of a number of algorithms that are designed to use sequence information to predict the stability of various nucleic acid conformations under given solution conditions. The Benight laboratory has been especially prodigious in the number and quality of rigorous thermodynamic studies conducted on DNA substrates.104–106 Often the focus of these studies has been to examine sequence-dependent effects on the thermodynamic stability of DNA in the context of nearest-neighbour base pair interactions. This has led to significant refinements in analytical methods for calculating the sequence-dependent melting stability of duplex DNA.106 One of the most comprehensive calorimetric studies of DNA duplexes conducted to date was carried out in the Privalov laboratory.107 This study used both DSC and ITC to examine the thermal stability and energetics of formation of three DNA oligonucleotides that were 10, 12 and 16 base pairs long. The rationale for undertaking this work was that the duplexes represented specific targets for the HMG box from mouse Sox-5. Therefore, in order to gain a complete understanding of this protein–DNA interaction it is necessary to understand the thermal properties of all the interacting species in isolation as well as the energetics of complex formation. These workers noticed a phenomenon for which there has been considerable anecdotal evidence over the years. Specifically, the pre- and post-transitional baselines from DSC scans were extrapolated into the region of the transition and this indicated that strand dissociation occurs with little or no apparent heat capacity change. However, titration experiments conducted as a function of temperature using ITC were used to measure enthalpies of strand association. These data indicate that duplex formation is associated with a significant decrease in heat capacity. In addition the enthalpy for strand association determined using ITC was considerably smaller than the enthalpy of cooperative melting measured by DSC. The importance of this study is that, for the first time, this paradox was resolved by using DSC to examine the effects of heating and cooling the individual strands of the duplexes. Cooling experiments on single-stranded DNA showed that there is a significant release of heat that arises from both inter- and intramolecular interactions, for example base stacking. By correcting for enthalpies of residual structure in single strands, and for the temperature dependence of heat capacities of duplexes and their component strands both ITC and DSC results can be brought into agreement. This work convincingly demonstrates that the observed increase in heat capacity of duplex DNA, prior to cooperative unfolding, as temperature is increased is due to increasing fluctuations in structure. This effect hides the

28

APPLICATIONS OF BIOCALORIMETRY

þ200 ð40Þ J K
1 (mol bp)
1 heat capacity change that results from strand separation. The large number of accumulated calorimetrically derived data on the unfolding/folding transitions of duplex DNA and the resulting thermodynamic insights are now being used to guide studies of higher-order tripleand quadruple-stranded DNA, the latter of which has been of particular recent interest. Intramolecular quadruplexes are formed by the self-assembly of DNA from short G-rich repeat sequences, in the presence of monovalent cations, to form stacked guanine tetrad planes joined by short loop regions.108 One of the reasons G quadruplexes are important structures is because telomeres (protein–DNA complexes that occur at 3’ termini of eukaryotic chromosomes) are composed of repeating G-rich sequences.109 Telomeres function to maintain chromosome integrity and also to overcome the endreplication problem. After 30–50 rounds of mitosis somatic cells enter a state of senescence due to loss of telomeric DNA. Stem cells and the majority of tumour cells overcome this limitation since they over-express the ribonucleoprotein telomerase.110 This enzyme mediates the synthesis of telomeric DNA de novo using an RNA template and reverse transcriptase activity. Since telomerase has been associated with the longevity of cancer cells it has become an important and specific target for novel chemotherapeutics. One strategy for achieving telomerase inhibition is to target the telomeric DNA receptor for the protein. Thus interest in quadruplex DNA arose partly because of the observed telomerase inhibitory effects of Kþ and Naþ ions as well as quadruplex interactive ligands. In order to develop low molecular weight ligands capable of binding to quadruplex DNA in a sequence and structure specific manner it is necessary to understand the energetic forces that drive assembly of quadruplexes. With this information we will also be able to better assess the conditions under which single-stranded telomeric DNA might fold into the quadruplex conformation in vivo. Recently, we have undertaken some preliminary DSC studies to examine the quadruplex!random coil transition. We examined the melting behaviour of a model intramolecular quadruplex DNA with a sequence based on the human telomere tandem repeat. This DNA is referred to as G3 and has the sequence 5’-AGGG(TTAGGG)3. DSC melting experiments were carried out in either all potassium or all sodium conditions. Figure 1.6 shows a DSC endotherm obtained for the melting of 243 mM quadruplex in 200 mM KCl at a scan rate of 18C min
1 . The buffer baseline has been subtracted and the raw melting curve, which is clearly composed of multiple transitions, was deconvolved using Origin 5.0 software and fitted to a non-two-state dissociation model. This hitherto unobserved melting behaviour indicates that unfolding of G3 does not proceed in a simple two-state manner. In fact the overall experimental melting endotherm is best fitted to three component two-state transitions. The DSC melting data shows that there are three

DSC FOR EXAMINING NUCLEIC ACID HELIX!COIL TRANSITIONS

29

component transitions involved in melting of G3 quadruplex. The raw data were also fitted to two component transitions and a statistically worse fit was obtained, although the difference is small. We interpret the low-temperature transitions observed in DSC experiments as being due to unstacking or premelting of the loop residues prior to global unfolding of the G quartets. The melts in an all sodium environment were qualitatively similar to the data shown in Figure 1.6; however, there is only evidence of two component transitions. The main difference is observed in the Tm , which for Naþ -G3 is 518C; this means that G3 is 228C less stable in sodium than in potassium. The DSC data also allowed us to estimate an enthalpy change for quadruplex unfolding, and this was found to be þ69 kJ (mol base tetrads)
1 . These preliminary DSC experiments are important because they reveal complexity in the melting process. This clearly indicates that the composition and length of loop residues plays an important role in determining overall quadruplex thermodynamic stability. We have also conducted complementary ITC experiments in efforts to quantify the thermodynamics of quadruplex folding. These experiments are technically challenging but nevertheless we have prepared the G3 DNA completely free of any metal cations, so that the DNA is single stranded. Potassium ions were then titrated into the DNA and the heats associated with cation binding and consequent quadruplex folding were measured. These data yielded enthalpy values of about
25 kJ (mol base

Figure 1.6 DSC scan of G3 quadruplex in 200 mM K+. The solid line is the experimental data curve and the dotted line represents a fit to the data that yields three component twostate transitions shown as dashed lines

30

APPLICATIONS OF BIOCALORIMETRY

tetrad)
1 . It is interesting to note that this value does not agree well with the enthalpy data obtained using DSC. At the moment the exact source of this discrepancy is not clear since a number of molecular events may be giving rise to compensatory thermodynamic effects. However, there is a possibility and perhaps even a likelihood that the heat capacity effects observed for duplex melting in the Privalov study107 may also be apparent for unfolding of highorder nucleic acids. This possibility is currently under active investigation in our laboratories.

1.18 Summary In this review we have described a number of examples where calorimetry has given us information that is crucial to understanding the stability and interactions of biological molecules. This knowledge is fundamental to cellular function and successful rational, structure based drug design. It is only by having a clear understanding of the underlying forces that govern biomolecular interactions that we can hope to rationalize affinity, predict the consequences of naturally occurring mutations, design specific inhibitors and, in turn, develop efficient drugs. It is becoming increasingly apparent that high-throughput screening methods may not be the most cost effective way of producing a commercial drug. This fact combined with the increase in crystallographic and NMR structural data has led to the increasing use of rational, knowledge-rich, lower-throughput methods such as calorimetry. Thermodynamic data measured for the formation of nucleic acid/protein–ligand complexes where structural information is available has led to a deeper understanding of the forces responsible for biomolecular interactions. This approach is now routinely used both by biochemists interested in the machinery of binding and drug designers alike.

Acknowledgements We are very grateful to Professors Terry Jenkins and Babur Chowdhry for their help with DSC studies of G quadruplexes. We would also like to thank Drs Peter Lee-Robichaud, Nick Williams and Quaiser Sheikh for their help with purification of PP-1g and spectroscopic assays; we also thank the BBSRC for funding this project. We are also very grateful to Dr Matthew. J. Todd for supplying a very useful spreadsheet, which has helped us to analyse enzyme kinetic data from ITC. IH is an EPSRC Advanced Research Fellow.

REFERENCES

31

References 1. Orengo CA, Bray JE, Buchan DWA, Harrison A, Lee D, Pearl FMG, Sillitoe I, Todd AE and Thornton JM. (2002) Proteomics 2: 11–21. 2. Haq I. (2002) Arch. Biochem. Biophys. 403: 1–45. 3. Cooper A. (1999) Protein 2: 217–270. 4. Cooper A. (1999) Curr. Opin. Chem. Biol. 3: 557–563. 5. Chaires JB. (1998) Curr. Opin. Struct. Biol. 8: 314–320. 6. Ha¨rd T and Lundba¨ck T. (1996) Biophys. Chem. 62: 121–139. 7. Wiseman T, Williston S, Brandts JF and Lin L-N. (1989) Anal. Biochem. 179: 131–137. 8. Plotnikov VV, Brandts JM. Lin L-N and Brandts JF. (1997) Anal. Biochem. 250: 237– 244. 9. Freire E, Mayorga OL and Straume M. (1990) Anal. Chem. 62: 950A–959A. 10. Spokane RB and Gill SJ. (1981) Rev. Sci. Instrum. 52: 1728–1733. 11. Sturtevant JM. (1996) Prot. Sci. 5: 391–394. 12. Sturtevant JM. (1987) Annu. Rev. Phys. Chem. 38: 463–488. 13. McKinnon IR, Fall L, Parody-Morreale A and Gill SJ. (1984) Anal. Biochem. 139: 134–139. 14. Privalov PL and Plotnikov VV. (1989) Thermochim. Acta 139: 257–277. 15. Naghibi H, Tamura A and Sturtevant JM. (1995) Proc. Natl Acad. Sci. USA 92: 5597– 5599. 16. Chaires JB. (1997) Biophys. Chem. 64: 15–23. 17. Haq I. (1998) In Applications of Calorimetry in the Biological Sciences, Ladbury JE and Chowdhry BZ (eds.), Wiley, Chichester, pp. 41–61. 18. Breslauer KJ, Freire E and Straume M. (1992) Meth. Enzymol. 211: 533–567. 19. Sturtevant JM. (1974) Annu. Rev. Biophys. Bioeng. 3: 35–51. 20. Privalov GP and Privalov PL. (2000) Meth. Enzymol. 323: 31–62. 21. Livingstone JR. (1996) Nature 384: 491–492. 22. Plotnikov VV, Rochalski A, Brandts JM, Brandts JF, Williston S, Frasca V and Lin L-N. (2002) Assay Drug Devel. Techn. 1: 83–90. 23. Haq I, Chowdhry BZ and Jenkins TC. (2001) Meth. Enzymol. 340: 109–149. 24. Haq I, Jenkins TC, Chowdhry BZ, Ren J and Chaires JB. (2000) Meth. Enzymol. 323: 373–405. 25. Jenkins TC. (1997) In Methods in Molecular Biology, vol. 90, Fox KR (ed.), Humana, Totowa, NJ, pp. 195–218. 26. O’Brien R, Ladbury JE and Chowdhry BZ. (2001) In Protein–Ligand Interactions: Hydrodynamics and Calorimetry, Harding SE and Chowdhry BZ (eds.), Oxford University Press, Oxford, pp. 263–286. 27. Haq I and Ladbury JE. (2000) J. Mol. Recognit. 13: 188–197. 28. Jelesarov I and Bosshard HR. (1999) J. Mol. Recognit. 12: 3–18. 29. Ladbury JE and Chowdhry BZ. (1996) Chem. Biol. 3: 791–801. 30. Ladbury JE. (1995) Structure 3: 635–639. 31. Bradshaw JM, Grucza RA, Ladbury JE and Waksman G. (1998) Biochemistry 37: 9083–9090. 32. Mandiyan V, O’Brien R, Zhou M, Margolis B, Lemmon MA, Sturtevant JM and Schlessinger J. (1996) J. Biol. Chem. 271: 4770–4775. 33. Renzoni DA, Pugh DJR, Siligardi G, Das P, Morton CJ, Rossi C, Waterfield MD and Ladbury JE. (1996) Biochemistry 35: 15646–15653. 34. Rudolph MG, Linnemann T, Grunewald P, Wittinghofer A, Vetter IR and Herrmann C. (2001) J. Biol. Chem. 27: 23914–23921.

32

APPLICATIONS OF BIOCALORIMETRY

35. Dong J, Peters-Libeau CA, Weisgraber KH, Segelke BW, Rupp B, Capilla I, Hernaiz MJ, LeBrun LA and Linhardt RJ. (2001) Biochemistry 40: 2826–2834. 36. Lee L, Stollar E, Chang J-F, Grossman G, O’Brien R, Ladbury JE, Carpenter B, Roberts S and Luisi B. (2001) Biochemistry 40: 6580–6588. 37. Scheufler C, Brinker A, Bourenkov G, Pegoraro S, Moroder L, Bartunik H, Hartl FU and Moarefi I. (2000) Cell 101: 199–210. 38. Prodromou C, Siligardi G, O’Brien R, Woolfson DN, Regan L, Panaretou B, Ladbury JE, Piper PW and Pearl LH. (1999) EMBO J. 18: 754–762. 39. Joseph C, Stier G, O’Brien R, Politou AS, Atkinson RA, Bianco A, Ladbury JE, Martin SR and Pastore A. (2001) Biochemistry 40: 4957–4965. 40. Ward WH and Holdgate GA. (2001) Prog. Med. Chem. 38: 309–376. 41. Morton CJ and Ladbury JE. (1996) Protein Sci. 5: 2115–2118. 42. Dam TK and Brewer CF. (2002) Chem. Rev. 102: 387–429. 43. Yamada K. (1999) Mol. Cell Biochem. 190: 39–45. 44. Majhi PR and Blume A. (2001) Langmuir 17: 3844–3851. 45. Hildebrand A, Neubert R, Garidel P and Blume A. (2002) Langmuir 18: 2836–2847. 46. O’Brien R, Rugman P, Renzoni D, Layton M, Handa R, Hilyard K, Waterfield MD, Driscoll PC and Ladbury JE. (2000) Protein Sci. 9: 570–579. 47. Dam TK, Roy R, Page D and Brewer CF. (2002) Biochemistry 41: 1351–1358. 48. Spolar RS and Record MT Jr. (1994) Science 263: 777–783. 49. Velazquez-Campoy A, Kiso Y and Freire E. (2001) Arch. Biochem. Biophys. 390: 169– 175. 50. Holdgate GA. (2001) Biotechniques 31: 164–184. 51. Kwong P, Doyle ML, Casper DJ, Cicala C, Leavitt SA, Majeed S et al. (2002) Nature 420: 678–682. 52. Borea PA, Varani K, Gessi S, Gilli P and Gilli G. (1998) Biochem. Pharmacol. 55: 1189–1197. 53. Ayala YM, Vindigni A, Nayal M, Spolar RS, Record MT Jr and DiCera E. (1995) J. Mol. Biol. 253: 787–798. 54. Fersht AR. (1999) Structure and Mechanism in Protein Science, Freeman, New York. 55. Henriques DA and Ladbury JE. (2001) Arch. Biochem. Biophys. 390: 158–168. 56. Sarver SW, Yuan P, Marshall VP, Petzcold GL, Poorman RA, DeZwaan J and Stockman BJ. (1999) Biochim. Biophys. Acta 1434: 304–316. 57. Payne JW, Grail BM, Gupta S, Ladbury JE, Marshall NJ, O’Brien R and Payne GM. (2000) Arch. Biochem. Biophys. 384: 9–23. 58. Chaires JB. (1998) Biopolymers 44: 201–215. 59. Haq I, Ladbury JE, Chowdhry BZ, Jenkins TC and Chaires JB. (1997) J. Mol. Biol. 271: 244–257. 60. Ha JH, Spolar RS and Record MT Jr. (1989) J. Mol. Biol. 209: 801–806. 61. Record MT Jr, Anderson CF and Lohman TM. (1978) Q. Rev. Biophys. 11: 103–178. 62. Chaires JB. (1996) Anti Cancer Drug Des. 11: 569–580. 63. Mazur S, Tanious FA, Ding D, Kumar A, Boykin DW, Simpson IJ, Neidle S and Wilson WD. (2000) J. Mol. Biol. 300: 321–337. 64. Ren J, Jenkins TC and Chaires JB. (2000) Biochemistry 39: 8439–8447. 65. Haq I, O’Brien R, Lagunavicius A, Siksnys V and Ladbury JE. (2001) Biochemistry 40: 14 960–14 967. 66. Lagunavicius A and Siksnys V. (1997) Biochemistry 36: 11086–11092. 67. Baker BM and Murphy KP. (1996) Biophys. J. 71: 2049–2055. 68. Parsegian VA, Rand RP and Rau DC. (2000) Proc. Natl Acad. Sci. USA 97: 3987–3992. 69. Spink CH and Chaires JB. (1999) Biochemistry 38: 496–508.

REFERENCES

70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 86. 87. 88. 89.

90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. 104. 105. 106.

33

Qu X and Chaires JB. (1999) J. Am. Chem. Soc. 121: 2649–2650. Qu X and Chaires JB. (2001) J. Am. Chem. Soc. 123: 1–7. Bergqvist S, Williams MA, O’Brien R and Ladbury JE. (2002) Structure 10: 629–637. Bergqvist S, O’Brien R and Ladbury JE. (2001) Biochemistry 40: 2419–2425. Lin L-N, Brandts JF, Brandts JM and Plotnikov V. (2002) Anal. Chem. 302: 144–160. Cooper A, Johnson CM, Lakey JH and No¨llmann M. (2001) Biophys. Chem. 93: 215– 230. Heerklotz H and Seelig J. (2002) Biophys. J. 82: 1445–1452. Spink C and Wadso I. (1976) Meth. Biochem. Anal. 23: 1–159. Watt GD. (1990) Anal. Biochem. 187: 141–146. Morin PE and Freire E. (1991) Biochemistry 30: 8494–8500. Williams BA and Toone EJ. (1993) J. Org. Chem. 58: 3507–3510. Lonhienne T, Baise E, Feller G, Bouriotis V and Gerday C. (2000) Biochim. Biophys. Acta 1545: 349–356. Todd MJ and Gomez J. (2001) Anal. Biochem. 296: 179–187. Chowdhry BZ and Cole SC. (1989) TIBTECH 7: 11–16. Lumry R and Eyring H. (1954) J. Phys. Chem. 58: 110–120. Sanchez-Ruiz JM. (1992) Biophys. J. 61: 921–935. Galisteo ML, Mateo PL and Sanchez-Ruiz JM. (1991) Biochemistry 30: 2061–2066. Sanchez-Ruiz JM, Lopez-Lacomba JL, Cortijo M and Mateo PL. (1988) Biochemistry 27: 1648–1652. Remmele R and Gombotz WR. (2000) BioPharm. 13: 36–46. Cooper A, Nutley MA and Wadood A. (2001) In Protein–Ligand Interactions: Hydrodynamics and Calorimetry, Harding SE and Chowdhry BZ (eds.), Oxford University Press, Oxford, pp. 287–313. Hofr C and Brabec V. (2001) J. Biol. Chem. 276: 9655–9661. Chakrabarti MC and Schwarz FP. (1999) Nucleic Acid Res. 27: 4801–4806. Morein S, Killian JA and Sperotto MM. (2002) Biophys. J. 82: 1405–1417. Liu F, Lewis RNAH, Hodges RS, McElhaney RN. (2002) Biochemistry 41: 9197– 9207. Brandts JF, Hu CQ, Lin L-N and Mas MT. (1989) Biochemistry 28: 8588–8596. Gandhavadi M, Allende D, Vidal A, Simon SA and McIntosh TJ. (2002) Biophys. J. 82: 1469–1482. Tarahovsky YS, Rakhmanova VA, Epand RM and MacDonald RC. (2002) Biophys. J. 82: 264–273. Brandts JF and Lin L-N. (1990) Biochemistry 29: 6927–6940. Remmele R, Nightlinger NS Srinivasan S and Gombotz W. (1998) Pharm. Res. 15: 200–208. Vega-Warner V and Smith DM. (2001) J. Agric. Food Chem. 49: 906–912. Rawitscher MA, Ross PD and Sturtevant JM. (1963) J. Am. Chem. Soc. 85: 1915– 1918. Klump H and Ackermann T. (1971) Biopolymers 10: 513–522. Privalov PL, Ptitsyn OB and Birshtein TM. (1969) Biopolymers 8: 559–571. Breslauer KJ, Frank R, Blo¨cker H and Marky LA. (1986) Proc. Natl Acad. Sci. USA 83: 3746–3750. Riccelli PV, Vallone PM, Kashin I, Faldasz BD, Lane MJ and Benight AS. (1999) Biochemistry 38: 11 197–11 208. Vallone PM and Benight AS. (2000) Biochemistry 39: 7835–7846. Owczarzy R, Vallone PM, Gallo FJ, Paner TM, Lane MJ and Benight AS. (1998) Biopolymers 44: 217–239.

34 107. 108. 109. 110.

APPLICATIONS OF BIOCALORIMETRY

Jelesarov I, Crane-Robinson C and Privalov PL. (1999) J. Mol. Biol. 294: 981–995. Williamson JR, Raghuraman MK and Cech TR. (1989) Cell 59: 871–880. Blackburn EH. (2000) Nature 408: 53–56. Bryan TM and Cech TR. (1999) Curr. Opin. Cell Biol. 11: 318–324.

Part II Isothermal Titration Calorimetry

2 Isothermal Titration Calorimetry: a Tutorial James A. Thomson and John E. Ladbury

2.1 Introduction A decade ago, the isothermal titration calorimeter (ITC) was considered as a specialist apparatus; few laboratories had access to one and even fewer understood the implications of much of the data generated. Currently, ITC instruments can be found in hundreds of laboratories, whose specialities range from determining affinities for whole cellular systems to drug design. This rapid flourishing of the method and widespread availability of data has dramatically increased our understanding of many systems. It has also provided crucial data towards improving our ability to predict thermodynamic parameters for biomolecular processes. However, the desire to produce user-friendly instrumentation and make increased amounts of data more rapidly available has led to perplexity associated with experimental design and data interpretation. In this chapter we attempt to cover some of the potential pitfalls in these areas and make suggestions as to how the less experienced experimentalist can discern and overcome these. This tutorial aims to familiarize the reader with the ITC experimental process. It is aimed primarily at the enthusiastic beginner or the inquisitive dilettante in this field, and we recommend looking at other reviews on ITC (for example references 1–5 and references therein) and other chapters in this book for more advanced detail and description. To enhance this text and to introduce examples we have included a series of ‘Application Notes’ as an appendix. Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

38

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

2.2 Thermodynamic characterization Fundamental biological processes mediated by equilibrium biomolecular interactions are defined by the dependence of thermodynamic properties on structural detail. Thus, determination of the physico-chemical parameters, and relation of these to the structural consequences of changing the system from the free to the bound state, is essential to comprehending how interactions occur. The development of high-sensitivity microcalorimetric instrumentation makes the direct measurement of the thermodynamic parameters associated with biomolecular interactions accessible. Unlike any other method used to determine thermodynamic parameters, calorimetry, as the name suggests, uses the direct measurement of the heat of the interaction to probe the extent of binding between molecules. The exchange of heat with the environment is a ubiquitous property of all equilibrium interactions. A modern ITC can measure the heat of interaction (or molar enthalpy DH) with great accuracy (typically the minimum heat pulse is of the order of 1.0– 0.1 mcal). In the ITC experiment the heat of interaction is used as a probe of the extent of binding as one component of the interaction is titrated into the other. As a result, the concentration of the complex formed is known at any point in the titration, and thus the equilibrium binding constant, KB ð¼ 1=KD Þ can be determined (see below). If DH and KB are known then a full thermodynamic characterization of the interaction at the experimental temperature can be determined from the relationship in Equations (2.1) and (2.2). DG ¼
RT ln KB

ð2:1Þ

DS ¼ ðDH
DGÞ=T

ð2:2Þ

Thus, from KB we can determine the change in Gibbs free energy, DG, at the experimental temperature, T, using Equation (2.1), where R is the gas constant. If we know DG, DH and T then the change in entropy, DS, for the interaction can be determined from Equation (2.2). One further term that can be determined from the ITC experiment is the change in constant pressure heat capacity, DCP , on going from the free to the bound state. This term is the temperature derivative of DH. Therefore, performing ITC experiments over a range of temperatures will provide a value for this term (as shown in Equation (2.3)). DCP ¼ ðDHT2
DHT1 Þ=ðT2
T1 Þ

ð2:3Þ

where T1 and T2 are two different experimental temperatures. The value of the thermodynamic terms obtained from the ITC experiment is often not fully appreciated and sometimes over-interpreted. The DG term provides a concentration independent term that allows comparison of the

THERMODYNAMIC CHARACTERIZATION

39

affinities of interactions at a given temperature and set of conditions. As with all the thermodynamic terms, this term is based on a number of different contributions. These can be parsed into the following general, additive contributions to DG: intermolecular contacts (non-covalent bonds), conformational changes, hydrophobic transfer (e.g. of ligand to binding site), polyelectrolyte contribution (e.g. ionization effects) and changes in rotational/translational motion. The parsing of DG into its respective contributions has been attempted for some systems (see for example reference 6). The component of DG, DH, is the measure of heat energy associated with going from the free to the bound state at a given temperature. This can be thought of as the net heat associated with the making and breaking of noncovalent bonds in forming the biomolecular complex. It should be emphasized that this value includes not just the bonds associated with ligand binding in the appropriate site, but also those due to solvent rearrangement and also any conformational change experienced by the components of the interaction. As a result of the potentially large number of bonds made and broken in the formation of biomolecular interaction the DH measured can be complicated to interpret. This term can best be appreciated when high-resolution structural detail is available of the interacting molecules in the free and bound states. This is because the net change in bonds in the binding site can be visualized. Comparison of the DH terms for interactions with a given biomolecule and subtly modified ligands can give some insight as to the importance of specific, non-covalent bond formation. The DS term is far more difficult to interpret. Entropy is usually considered in simplistic terms as the thermodynamic state function that is a measure of the tendency of a system to disorder. This rather loose definition can be applied on the atomic/molecular level by asking whether the system becomes more ordered or disordered on going from free to bound state. Clearly the coming together of two (or more) biomolecules imposes some order on the system (with respect to the molecules involved (translational/cratic entropy) and the restriction of atomic degrees of freedom through bonding in the biomolecular interface (configurational entropy)), whereas the release of water molecules from a binding site into the bulk solvent results in disorder. The DS term is very ambiguous in the definition of an interaction. The most important correlation of this term with structural effects on complex formation is based on the burial of hydrophobic surface area.7–10 The organization of water molecules on a hydrophobic surface of a biomolecule can often extend to several layers of hydration. As a result, the burial of this surface in the complex interface results in the liberation of these waters. This can provide a major contribution in terms of DS to DG. (See Application Note 1. DH and DS provide a further level of characterization of an interaction.)

40

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Perhaps the most widely used term in relating the measured thermodynamics of an interaction to structure is DCP . This term has been shown to correlate with the burial of surface area. Several authors have optimized this correlation to enable some predictive capability of the structural or thermodynamic changes on binding of biomolecules.10–15 Nonetheless some discrepancies exist with this correlation16–20 and it should be used with caution. For example, it is not appropriate, for the most part, to attempt to make structural inferences on large conformational changes based on a DCP measurement.

2.3 Instrumentation In a typical ITC instrument the change in enthalpy on binding is measured in a small reaction cell (with an approximate volume of 0.5–1.5 ml). Two identical cells, one for the interaction (the sample cell) and one to act as a reference (see Figure 2.1), are located in a jacket (usually called the adiabatic jacket, although it is not strictly adiabatic), which is kept at a temperature below that at which the experiment is to be conducted (typically by about 5–108C). The sample cell has an opening into which a syringe is placed. Aliquots of one of the interacting solutions are injected into the other via the syringe. The reference cell contains buffer or water (or a solution of similar heat capacity as the interacting components in the sample cell) and is usually sealed. The two cells are heated to the experimental temperature. The power applied to the cell to maintain constant temperature is measured and provides the baseline value at thermal equilibrium. Their temperature is continually monitored by a thermocouple device. The sample cell and the reference cell are always maintained at thermal equilibrium at identical temperatures. The injection of one of the components (to avoid confusion this will be designated component 1 throughout this chapter) from the syringe into the other component (2) of the interaction in the calorimeter sample cell will result in a heat change. If this is exothermic (i.e. heat is given out) the sample cell will require less power input to maintain thermal equilibrium with the reference cell. This change in power requirement over a given time is measured. The data can be presented as a plot of power against time (see Figure 2.2). As the interaction returns to equilibrium the power has to return to its original value. As a result a peak is obtained on the plot. Subsequent injections will have the same effect. If the respective concentration regimes are set up accordingly, a series of injections of component 1 will gradually saturate the binding sites available on component 2 in the calorimeter cell. Once the binding sites on 2 are occupied the only heat that is observed is that derived from component 1 being dissolved into the solution in the cell. This heat is

INSTRUMENTATION

41

Figure 2.1 (a) A schematic diagram of a typical cell arrangement in an ITC. (b) A schematic diagram of a typical cell arrangement in an ITC on performing a titration. The raw data output is shown after the titration is complete

42

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Figure 2.2 Typical data derived from an ITC experiment. The upper panel shows the raw data output. The peaks are plotted as power against time. The lower panel shows the integrated raw data output plotted as molar change in enthalpy against mole ratio

known as the heat of dilution (see below). Since this heat does not relate to the binding interaction it has to be determined and subtracted from the binding isotherm. As a result a separate control experiment is necessary to ascertain this heat (see below).

2.4 Raw data A typical set of raw data is shown in Figure 2.2 (top). The figure shows a plot of the power (mcal s71) against time for the interaction of ribonuclease A and 2’CMP. This interaction has been used as a standard model system in calorimeters due to its suitably large enthalpy of interaction, simple stoichiometry and ready availability (components can be readily purchased). As is apparent from the plot the experiment was set up so the 13 injections of the protein (at a concentration of 200 mM) were made into the nucleotide (10 mM). The injections (10 ml) were made every four minutes (after an initial small injection

BASIC CONSIDERATIONS FOR EXPERIMENTAL SET-UP

43

of a smaller volume which is used to remove any air bubble from the end of the syringe and reduce any diffusion of nucleotide into the syringe during the equilibration period prior to initiating the titration). The change in molar enthalpy of the interaction can be determined by integrating the raw data peaks with respect to time. Since we know the total concentration of the components of the interaction at any given point in the titration we can plot DH against the molar ratio of 1 and 2. For a simple interaction with a stoichiometry of one independent binding site under appropriate conditions this gives a sigmoidal isotherm. This is shown for the ribonuclease A-2’CMP interaction in Figure 2.2 (bottom).

2.5 Basic considerations for experimental set-up Before beginning an ITC experiment there are several points that have to be considered. These are itemized below.

Detectable signal/heat of interaction As stated above heat is a ubiquitous signal derived from equilibrium interactions; however, the DH of interaction is temperature dependent, i.e. it varies according to the change in heat capacity (see Figure 2.3). Thus at some

Figure 2.3 Graph of DH against temperature showing typical linear relationship for three different protein–DNA interactions32

44

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

temperatures DH will be low or zero and thus undetectable even with the highly sensitive microcalorimeters commercially available. As described above DH depends on the temperature. As a result the temperature at which DH is zero is system dependent, thus the experimental temperature has to be chosen such that sufficient heat is obtained from the binding event. Typically, if in doubt, we would recommend trying the experiment at say 258C initially. If no heat is measurable then the temperature should be adjusted by, say, 108C either up or down. Caution has to be exercised in adjusting the temperature, since the interacting components can become denatured. (See Application Note 2. Effects of temperature on DH.) Clearly, since DH is temperature dependent, it is absolutely vital that the experimental temperature is reported. Furthermore, if two or more interactions are to be compared (for example, a protein interacting with a range of drug compounds) all the experiments should be carried out at the same temperature. The heat of an interaction can be enhanced by changing conditions other than the temperature. For example, if the interaction is accompanied by a protonation event the choice of a buffer system with a high DH of ionization will enhance the binding enthalpy. In Figure 2.4 effects on the DH of binding

Figure 2.4 Graph of raw data output from a protein–ligand interaction performed in two different buffer systems. The endothermic peaks are derived from an experiment performed in TRIS buffer. The exothermic peaks are derived from an experiment performed in HEPES (see also Figure 2.11)

BASIC CONSIDERATIONS FOR EXPERIMENTAL SET-UP

45

are shown for ITC experiments in two different buffer systems, TRIS and HEPES. The interaction performed in TRIS buffer gives an endothermic interaction (more power required), whereas that in HEPES is exothermic. The raw data are clearly very different and as a result the values for DH will also be different. As a result of this effect it is of utmost importance to report the buffer system used and be aware that comparison of DH values from different experiments should be done with caution. (See Application Note 3. DH of ionization allows determination of number of protonation events occurring on complex formation.)

Titration curve The ITC experiment can be used solely to determine highly accurate DH values for interactions alone; however, the method is also usually applied to the determination of KB . The binding constant is determined by fitting the ITC isotherm to obtain the change in free ligand concentration with respect to total ligand concentration (see below). As a basic consideration for the experimental set-up it is important to obtain an isotherm that provides maximum data points for the fitting process. In other words the shape of the binding isotherm will dictate how accurately KB can be determined, if at all. The shape of the isotherm is dependent on KB and the concentration of the interacting component in the calorimeter cell (the total binding site concentration). The product of these two terms gives a number called the C value,1 i.e. C ¼ n KB ½component 2

ð2:4Þ

where n is the stoichiometry of the interaction. C is a unitless parameter since KB is measured in M71. Analysis of the data requires an appropriate C value. As a rule of thumb, the C value should between 10 and 100. The reason for this is demonstrated in the data simulations shown in Figure 2.5. These binding isotherms were simulated using n ¼ 1, DH ¼ 710 kcal mol71, [component 1] ¼ 200 mM, [component 2] ¼ 10 mM and KB values ranging from 104 to 109 M71 (KD from 100 mM to 1 nM), resulting in C values ranging from 0.1 to 10 000. As can be seen from the simulated data in Figure 2.5 under the concentration regimes usually adopted for ITC experiments the data sets with low C values are not sigmoidal and tend towards featureless straight lines. In these experiments the complex formed, and hence DH, changes very little from injection to injection. Thus the determination of the free versus bound ligand concentrations becomes inherently inaccurate. In the isotherms with the very high C values the sigmoidal shape becomes more angular. The data corresponds to a situation where in the early injections all of the ligand is binding to form complex. At a given concentration of component 1 the

46

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Figure 2.5 Plot of the effects on the binding isotherm of different C values and different KD values

interaction is saturated and hence the final injections show no net heat. This data set also does not facilitate the fitting of the free versus bound concentrations as the titration comes to completion. It should be borne in mind that the useful range in terms of KB of modern instruments is approximately between 103 and 109 M71 (or KD between 1 mM and 1 nM). As a result Equation (2.4) would suggest that concentrations as low as 10 nM can be used for titrations. Unfortunately, for most biomolecular systems concentrations in the nanomolar range are not suitable since they do not provide sufficient heat change on binding. (See Application Note 4. Applications based on the stoichiometry of the interaction.)

Choice of buffer solution Most modern ITC instruments are robust and almost anything goes with respect to suitable solvents (always check the ITC manufacturer’s recommendations for the calorimeter cell).

BASIC CONSIDERATIONS FOR EXPERIMENTAL SET-UP

47

The heat of dilution of component 1 into the solution in the calorimeter cell should be kept to a minimum. The accuracy of the data will be affected adversely if a large heat of dilution has to be subtracted from a comparatively minute heat of binding. To ensure that this heat is minimized the buffers used in the cell and syringe must be carefully matched. This is accomplished most effectively by dialysing both component 1 and component 2 against the same buffer. If one of the components is too small for dialysis then dialyse the macromolecule and then dissolve the other component in the dialysate. This does not always work; for example, small molecules are usually synthesized and can often be accompanied by residues of the synthetic process or the buffers used in their separation/purification. Buffer solutions containing reactive ingredients such as reducing agents can be problematic in that these will often give a background heat effect. For example, the heat associated with the constant oxidation of a reducing agent will usually result in a drift in the raw data baseline. This can be a problem if it is severe. As a result, reducing agents with low heats of oxidation are recommended. If reducing agent is required use either TCEP or BME instead of DTT. The accompanying oxidation effect can be reduced by making sure that the interacting components are dialysed in solvents that have been treated so as to reduce the oxygen content. For example, dissolving nitrogen or argon gas will displace the oxygen in aqueous buffers. Also, keeping the calorimeter sample cell from exposure to atmospheric oxygen can help to reduce the background effects. Making the entrance to the calorimeter cells airtight using an appropriate film during equilibration can be helpful. In many applications of the ITC, one or both of the interacting components are insoluble in pure aqueous solvent. This can cause problems; for example, if a non-aqueous or organic buffer is used the heats of dilution can often be substantial. In most cases when performing titrations on biomolecular samples a compromise has to be reached in terms of the buffer solution adopted. For example, in experiments commonly encountered in the pharmaceutical industry many drug compounds need to be dissolved in an organic solvent; however, most proteins are unlikely to prevail in a non-aqueous buffer system. In these situations ideally the organic buffer should be diluted as low as possible with the buffer used to dialyse the protein. Serial dilutions can be used to reduce the organic solvent down to suitably low concentration to have limited effect on the heats of dilution. This incremental dilution of the stock solution should help to avoid precipitation of the biomolecule. If, for example, the ligand is provided as a stock solution (for instance DMSO), dilute the stock solution with the dialysate (ideally to below 1 per cent) and then add a corresponding amount of stock solution solvent to the macromolecule solution.

48

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Control experiments As mentioned above the experiment is set up to determine the heat of binding as component 1 is added to component 2. However, in a typical experiment there are a number of sources of additional heat that have to be accounted for by performing control experiments. These additional heat effects have to be determined and removed from the raw data for the interaction. The addition of component 1 to the solution in the cell will result in a heat of dilution of this component. As a result, an independent control experiment is required to obtain this heat per injection. The control experiment is performed where the identical solution of component 1 used for the binding experiment is added to the buffer solution that is used for component 2 (usually the same as that for component 1). This control experiment will also account for any effects resulting in mismatch of the buffers used (see above regarding impurities from synthesis of small molecules and the effects of organic solvents). Figure 2.6 shows the heat of dilution experiment when sucrose was used as component 1. As can be seen the heats per injection are rather low. In addition it can be seen that there is a gradual trend in the peak heights towards a lower heat per injection. This is due to the phenomenon of the dilution effect being progressively reduced as the concentration of sucrose in the cell gradually increases. Usually this gradual change in signal for the heat of dilution

Figure 2.6 Control experiment showing the titration of sucrose into buffer solution

BASIC CONSIDERATIONS FOR EXPERIMENTAL SET-UP

49

experiment is ignored since it is very small compared to the overall heats associated with binding (see Figure 2.6). The heat of dilution control experiment can also reveal complications in the experiment, which might not have previously been expected. For example, at the elevated concentrations required for the ITC experiments the component in the syringe could be in an associated state. If the injection of this component into the solution in the cell results in dissociation of component 1 from its associated state, a gradually declining heat of dilution will be observed. Since the dissociation effect is likely to be concentration dependent, to verify that this is occurring the dilution can be conducted at different concentrations. Significantly different heats of dilution will be obtained. These data can actually be used to determine the equilibrium dissociation constant in some cases (for example see reference 21). A further contribution to the overall heats measured in a binding experiment comes from the effect of the dilution of component 2 in the calorimeter cell. As a result a heat of dilution experiment should be performed to control for this. In this case the experiment requires that heats are determined as buffer is injected into the calorimeter cell. This effect is usually very small, because the actual dilution of component 2 is very small. For example an injection of 10 ml of buffer into a 1.5 ml cell will only effect a dilution of less than one-hundredth. Because heat measurement in modern ITC instrumentation is so highly sensitive a final control experiment, known as the machine blank, is also required. This experiment involves adding buffer into buffer and takes account of the (usually very small) heat associated with machine effects such as pressure changes encountered. Figure 2.7 shows the different heat effects for adding water into water using progressively shorter injection times. As can be seen the heats are very small. Nonetheless, the shorter injection time used in obtaining the peak on the far right gives a noticeably larger heat. The heats associated with this can thus be determined and subtracted from the raw interaction data. Figure 2.8 shows the titration of ribonuclease A into 2’CMP as shown in Figure 2.2 accompanied by all the heats of dilution control experiments. The final injections in the isotherm for the raw data for the interaction still give a heat even though the binding of the interacting components has come to an end. Attempting to fit the isotherm without correcting for the various additional heats would give erroneous values. As is clearly apparent the heats are small for all the control experiments; however, they must be taken into account. The final dataset should be fitted after the following: fheat of interactiong
fheat of dilution component 1g
fheat of dilution component 2g þ fmachine blankg

50

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Figure 2.7 Control experiment showing the machine blank in which 10 ml of water has been injected into water. The injections differ in their durations

The machine blank should be added since in effect it has been subtracted twice, since it is also incorporated in the heat of dilution experiments. In practice the heat of dilution of component 2 and the machine blank are often neglected since they are usually insignificant. If there is a discrepancy between the heat at the end of an experiment after saturation of the binding sites and the heat of dilution of component 1 it could be due to significant effects from these additional heats.

2.6 Data Analysis The way in which the raw ITC data output is handled to determine the stoichiometry, n, DH and KB has been described in detail elsewhere (see for example references 1 and 22). The data analysis computer software from the different ITC manufacturers varies somewhat but can be broadly described by the following general overview. Let us call component 1 from the syringe X and component 2 in the cell Y. After making an injection, a number of moles of substance X has been added to a total amount of substance Y, [Ytot], in the calorimeter cell. As a consequence of binding the concentrations of free X, [X], and the complex, [XY], change. The interaction is accompanied by the production of heat. The

DATA ANALYSIS

51

Figure 2.8 ITC data for the interaction of ribonuclease A and 2’CMP. The figure shows the heats of interaction as well as all the control experiments

quantity of heat associated with the injection is a direct probe of the amount of binding that occurs. In the data fitting the problem is to relate the change in [X] to change in [XY]. This requires us to assume a model for the binding event. Taking the simplest model for an interaction, i.e. one where there is one independent binding site (i.e. we can neglect the stoichiometry term), the heat derived from an injection Q is related to the change in enthalpy by the following equation (2.5). Q ¼ ½Ytot V0 ðF2
F1 ÞDH

ð2:5Þ

where V0 is the volume of the cell and F is the fraction of Y bound. The numerical subscripts refer to the terms in going from injection 1 to injection 2. For the sake of simplicity we have ignored the change in volume associated with the injection of X. Since F ¼ ½XY=½Ytot 

ð2:6Þ

the above equation provides a relationship between heat and the amount of complex between X and Y being formed at each injection. For our thermodynamic parametrization we need to relate F to KB , where

52

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Figure 2.9 ITC data for the interaction of ribonuclease A and 2’CMP with the heats derived from the control experiments subtracted. The fit is based on a single set of identical binding sites model

KB ¼ ½XY=ð½X½YÞ:

ð2:7Þ

Knowing that ½X ¼ ½Xtot 
½XY and ½Y ¼ ½Ytot 
½XY, substituting into (2.7) we obtain the equation KB ¼ ½XY=ðð½Xtot 
½XYÞ ð½Ytot 
½XYÞÞ:

ð2:8Þ

Rearrangement of (2.8) gives the quadratic equation ½XY2
ð½Xtot  þ ½Ytot  þ 1=KB Þ½XY þ ½Xtot ½Ytot  ¼ 0

ð2:9Þ

The real root of this equation then gives a value for [XY] in terms of the known concentrations ½Xtot  and ½Ytot . Thus, fitting of a series of injections provides the value of the equilibrium constant. The fit of the binding data for the interaction of ribonuclease A and 2’CMP is shown in Figure 2.9. This fit is based on the simplest binding model based on a single set of independent and identical binding sites. This fit gives KB ¼ 8:04ð0:79Þ  105 M
1 and DH ¼
14:55ð0:37Þ kcal mol71.

APPLICATION NOTES

53

2.7 Summary In this tutorial we have attempted to emphasize the value of the methodology and highlight some of the areas in which the operator should exercise due care and attention. On the one hand it is important to realize that as ITC instrumentation becomes more widespread it is imperative that users become aware of some of the issues that can compromise their data. On the other hand it is also important that users learn that the range of experiments is large and the method provides a valuable approach to thermodynamic measurement that does not necessitate problems such as chemical modification, immobilization and potentially hazardous (e.g. radioactive) competing ligands that are required in other techniques.

Application notes 1. DH and DS provide a further level of characterization of an interaction. Using the example of a well studied system we can see that the determination of the thermodynamic parameters for an interaction provides a further level of characterization beyond the affinity. Table 2.1 includes ITC data for the interaction of the SH2 domain from the protein Src. SH2 domains are involved in processing intracellular signals and recognize ligands that have post-translationally phosphorylated tyrosine amino acids.23–26 In any given cell there can be more than a hundred different SH2 domains expressed. Since it is imperative that a given signalling pathway does not interfere with another, the individual SH2 domain interactions have been presumed to be highly specific. Since errors in these signalling pathways have been identified as giving rise to numerous disease states (e.g. cancer, immunodeficiency, diabetes) there has been significant interest in understanding how specific these interactions are (see chapter 8). This domain will bind peptide based ligands that include a phosphorylated tyrosine residue (pY) and that mimic physiological interactions. The domain makes interactions with the pY residue and amino acids proximal and C-terminal of this (i.e. the recognition motif is pYEEI). Table 2.1 shows data for the SH2 domain binding to a range of peptides that are only different by one amino acid sited three amino acids down from the pY residue. This residue was deemed important in the recognition process, making a fundamental interaction in the SH2 binding site. From Table 2.1 it can be seen that substitution of this residue has little effect on the overall KB values. Indeed, the DG terms for all the interactions only span a range of 0.8 kcal mol71. This suggests little in the way of specificity and represents a major conundrum in understanding cell signalling mediated by pY.27,28

54

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

Table 2.1 Binding data derived from ITC for the interaction of peptides with the SH2 domain from the Src protein. The data reflect a limited level of specificity of the peptides for the SH2 domain with respect to the change in free energy

Peptide sequence Ac-EPQpYEEIPIYL-NH2 Ac-EPQpYEEVPIYL-NH2 Ac-EPQpYEEEPIYL-NH2 Ac-EPQpYEEWPIYL-NH2 Ac-EPQpYEEDPIYL-NH2

KB
106 M71

DG kcal mol71

DH kcal mol71

TDS kcal mol71

10.8 6.24 4.87 3.18 2.64

79.6 79.3 79.1 78.9 78.8

79.3 76.8 77.8 77.7 76.6

0.3 2.5 1.3 1.2 2.2

The values for the DH and TDS terms do show some differences in the modes of interaction. For example, the interaction of the pYEEI motif has a significantly higher enthalpic contribution than the pYEEV peptide despite their similar aliphatic side chains. The pYEEV interaction on the other hand has a more favourable entropic term. We speculate (based on molecular dynamics simulations) that this is due to the greater mobility that the valine residue has in the binding site compared with the isoleucine. Quite unexpectedly (and by way of an example where it is difficult to predict the thermodynamic–structural correlation) the binding data for the aliphatic acidic glutamic acid residue (E) is almost identical to that of the large, aromatic, apolar tryptophan residue (W). 2. Effects of temperature on DH. The temperature dependence of DH is an important thermodynamic property of biomolecular interactions. As stated above the DCP term can be correlated with the burial of surface area on forming the complex. It is important to be aware, however, that increasing the temperature can also result in denaturation of the biomolecules under investigation. Thus assuming that the interaction occurs between the native, folded biomolecules at temperatures where denaturation becomes apparent, a heat associated with renaturation or folding of the components could be included in the binding. This is illustrated in Figure 2.10 where the binding of a short peptide to the subtilisin cleaved ribonuclease A fragment, ribonuclease S, is shown over the temperature range 10–408C.29 At lower temperatures (10– 208C) the relationship between DH and temperature is linear (as shown by the straight line in Figure 2.10). As the temperature increases the protein fragment begins to denature. As a result the heat associated with refolding the protein is added to the heat of binding (DHobs ¼ DHbind þ DHfold ). As the amount of unfolded protein increases with temperature the DHfold contribution to DHobs increases and results in curvature of the plot. 3. DH of ionization allows determination of number of protonation events occurring on complex formation. Since this DH of ionization is a unique

APPLICATION NOTES

55

Figure 2.10 The plot of DH against the temperature for a series of ITC experiments on the interaction of ribonuclease S polypeptide and S-peptide. The linearity of the DCP plot is lost at higher temperatures as the DH of refolding of the polypeptide is included with the DH of binding29

Figure 2.11 Plot of DHobs against DHionization for a protein–ligand interaction performed in two different buffer systems (see Figure 2.4). The slope of the line gives the number of protons released by the buffer

56

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

property of different buffer systems when protonation occurs on complex formation, it can be used to determine the number of protons interacting. The observed DH (DHobs ) from the titration experiment at a given pH is thus made up of the heat of binding and this protonation effect. These are related by Equation (2.10) DHobs ¼ DHbind þ ðDnÞDHionization

ð2:10Þ

Figure 2.11 shows DHobs against DHionization for two different buffers. The line described by Equation (2.10) will have a slope equal to Dn, the number of protons released/taken up by the buffer on binding.30–32 In Figure 2.11 the number of protons released is 1 (0.99). 4. Applications based on the stoichiometry of the interaction. In addition to the direct measurement of DH and the calculation of KB the ITC experiment provides a direct reading of the stoichiometry of an interaction. For example, if component 2 has two binding sites the equivalence point of the titration will be at a mole ratio of 2:1. This ability to directly, and independently, determine the stoichiometry of a system can be used in a number of different ways. For example, if the real

Figure 2.12 ITC data for a system from which the active concentration of the component in the calorimeter cell is only about one-third of that expected. The curve fit gives a stoichiometry of 0.33 (+0.02)

REFERENCES

57

stoichiometry of an interaction is known the ITC experiment can be used to assess the active concentration or purity of interacting compounds. This is of obvious benefit in a situation where one of the interacting components could be mixed with some impurity. Figure 2.12 shows such an experiment. In this case a pure synthetic inhibitor is added to a protein preparation that is known to interact with a 1:1 stoichiometry. The binding isotherm reveals that the stoichiometry is only 0.33. This suggests that either the inhibitor is more concentrated than expected, or the protein is less concentrated than expected. Investigation showed in this case that in fact the refolding process in the protein purification procedure was only approximately 30 per cent effective, resulting in a large portion of the material being incapable of binding the inhibitor.

Acknowledgement JEL is a Wellcome Trust Senior Research Fellow.

References 1. 2. 3. 4. 5. 6. 7 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21.

Wiseman T, Williston S, Brandts JF and Lin LN. (1989) Anal. Biochem. 179: 131–137. Freire E, Mayorga OL and Straume M. (1990) Anal. Chem. 62: 950–959. Ladbury JE. (1995) Structure 3: 635–639. Ladbury JE and Chowdhry BZ. (1996) Chem. Biol. 3: 791–801. Jelasarov I and Bosshard HR. (1999) J. Mol. Recognit. 12: 3–18. Haq I, Ladbury JE, Chowdhry BZ, Jenkins TC and Chaires JB. (1997) J. Mol. Biol. 271: 244–257. Sturtevant JM. (1977) Proc. Natl Acad. Sci. USA 74: 2236–2240. Searle MS and Williams DH. (1992) J. Am. Chem. Soc. 114: 10 690–10 697. Karplus M, Ichiye T and Pettitt BM. (1987) Biophys J. 52: 1083–1085. Spolar RS and Record MT Jr. (1994) Science 263: 777–783. Livingstone JR, Spolar RS and Record MT Jr. (1991) Biochemistry 30: 4237–4244. Spolar RS, Livingstone JR and Record MT Jr. (1992) Biochemistry 31: 3947–3955. Gomez J and Freire E. (1995) J. Mol. Biol. 252: 337–350. Luque I, Todd MJ, Gomez J, Semo N and Freire E. (1998) Biochemistry 37: 5791– 5797. Ren JS, Jenkins TC and Chaires JB. (2000) Biochemistry 39: 8439–8447. Ladbury JE, Wright JG, Sturtevant JM and Sigler PB. (1994) J. Mol. Biol. 238: 669– 681. Morton CJ and Ladbury JE. (1996) Prot. Sci. 5: 2115–2118. Guinto ER and DiCera E. (1996) Biochemistry 35: 8800–8804. Sleigh SH, Seavers PR, Wilkinson AJ, Ladbury JE and Tame JRH. (1999) J. Mol. Biol. 291: 393–415. Henriques DA, Ladbury JE and Jackson RM. (2000) Prot. Sci. 9: 1975–1985. Cooper A. (1998) In Biocalorimetry: Applications of Calorimetry in the Biological Sciences, Ladbury JE and Chowdhry BZ (eds), Wiley, Chicester, pp. 103–111.

58

ISOTHERMAL TITRATION CALORIMETRY: A TUTORIAL

22. Blandamer M. (1998) In Biocalorimetry: Applications of Calorimetry in the Biological Sciences, Ladbury JE and Chowdhry BZ (eds), Wiley, Chichester, pp. 5–25. 23. Ladbury JE, Lemmon MA, Zhou M, Green J, Botfield MC and Schlessinger J. (1995) Proc. Natl Acad. Sci. USA 92: 3199–3203. 24. Kuriyan J and Cowburn D. (1997) Annu. Rev. Biophys. Biomol. Struct. 26: 259–288. 25. Bradshaw JM, Grucza RA, Ladbury JE and Waksman G. (1998) Biochemistry 37: 9083–9090. 26. Hunter T. (2000) Cell 100: 113–127. 27. Ladbury JE and Arold S. (2000) Chem. Biol. 7: R3–R8. 28. O’Rourke L and Ladbury JE. (2003) Acc. Chem. Res. 36: 410–416. 29. Thomson J, Ratnaparkhi GS, Varadarajan R, Sturtevant JM and Richards FM. (1994) Biochemistry 33: 8587–8593. 30. Baker BM and Murphy KP. (1996) Biophys J. 71: 2049–2055. 31. Bradshaw JM and Waksman G. (1998) Biochemistry 37: 15 400–15 407. 32. Haq I, O’Brien R, Lagunavicius A, Siksnys V and Ladbury J. (2001) Biochemistry 40: 14 960–14 967.

3 The Application of Isothermal Titration Calorimetry to Drug Discovery Geoff Holdgate, Stewart Fisher and Walter Ward

3.1 Introduction Although technologies and methodologies have progressed significantly, the fundamental components of the drug discovery process have remained unchanged for many years, and have been responsible for the introduction of numerous modern medicines. The process begins with a hypothesis about the primary cause of a disease state, which leads to a potential molecular target for intervention. Compounds capable of binding to this target and eliciting the required response are sought, and a prototype lead structure identified. An optimization process then takes place, to produce the compound with the best activity profile. This candidate drug then enters the development phase of drug discovery, the progression of which will not be covered in this chapter. The research component of the drug discovery process is presented schematically in Figure 3.1. Isothermal titration calorimetry (ITC) monitors the heat change when a test compound binds to a target protein. The development of highly sensitive instruments has allowed this technique to be applied increasingly, to different extents, at several of the stages of drug discovery. It is the unique ability of ITC to measure binding enthalpies directly that gives the technique its broad application. An overview of the drug discovery process and of the Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

60

ITC IN DRUG DISCOVERY

measurement of thermodynamic binding parameters, along with the range of applications and potential of ITC in drug discovery, is given below.

3.2 Overview of the drug discovery process Target selection and validation The success of drug discovery depends upon the ability to identify novel chemical entities that have the ability to treat a particular disease state. Therefore, there is a need to identify biological targets that will enable competitive entry into novel therapeutic areas. A target may be defined in many ways, but one of the most useful is ‘a biomolecule against which chemical effort may be mounted to find small molecules that modulate that biomolecule’s activity’. Most targets are proteins that are particularly important for the structure and function of the living organism, and it is for this reason that substantial efforts are made to elucidate the structures and biological functions of proteins. The choice of target may be derived from traditional sources such as published literature or patents and other information available in the public domain. Targets may also be identified from novel biology, conducted in-house, or via commercial or academic alliances. Genetic studies can identify large numbers of gene products, which may be potential targets, which must be further analysed and prioritized into a workable number. In the context of new molecular targets, following initial identification, validation studies begin to verify that modulation of the target activity is relevant to alleviating the disease state. Important questions to answer during the validation process include whether the potential target is expressed in the appropriate cell type at the appropriate time during pathological conditions and whether the potential target plays a causative role in the biological effect or disease pathology. These activities of expression analysis and functional validation are crucial, since poor choice of targets

Figure 3.1 A scheme representing the process of drug discovery

OVERVIEW OF THE DRUG DISCOVERY PROCESS

61

ultimately leads to a poor discovery pipeline. Often, validation can only be achieved following the identification of selective and efficacious compounds, which can be used in the clinic for proof of principle studies. This is an expensive and unpredictable process, and represents a major risk. Target selection is based upon several criteria, including ongoing commercial evaluation of the opportunity, strength of the hypothesis, novelty of the mechanism, and feasibility of the high-throughput and secondary screens.

Hit identification Having selected a suitable target, the next stage in drug discovery is hit identification. Usually, the development of a high-throughput screen (HTS), through which several thousand compounds are tested each week, is the route for identification of ‘active’ compounds. These are compounds that show the desired activity against the target, but which may not necessarily meet the additional criteria needed for further progression. It should be noted that screening is a random process, even though it is focused on a defined molecular target. Its usefulness is limited by the selection of compounds to be screened, often based on existing knowledge of the target. Ideally, screening should be carried out on diverse structural libraries. The active compounds (which usually represent only 0.1–1 per cent of the total tested) would then be tested in a dose-dependent manner in the primary assay, to confirm activity. At this stage work is often initiated to determine the 3D structure of the target, as well as beginning preliminary in vitro drug metabolism and pharmacokinetic (DMPK) studies. The outcome from the hit identification phase should be a number of discrete chemical entities (hits), which are active in the primary and secondary screens (usually isolated protein assays), and which also possess an element of selectivity, as well as chemical characteristics amenable for optimization into a drug.

Lead identification The next stage of drug discovery often involves the synthesis of analogues of hits, which display the desired activity in cells, with the potential for activity in vivo. These compounds are usually known as ‘leads’. Significant amounts of chemical, biological and DMPK effort are often required to identify a proprietary lead series with the potential for optimization to produce a candidate drug. It is advantageous to have a stringent set of conditions by which to assess the quality of lead compounds. For example, does it have the required potency and selectivity in vitro, is it effective in vivo, is the structure suitable for chemical modification, does the compound contain groups likely

62

ITC IN DRUG DISCOVERY

to be toxic, is the compound of suitably low molecular weight, is the compound soluble, does the compound confer a strong intellectual property position? The lead compound should, overall, be an advance over previously studied compounds. As well as HTS there are several other processes, which can bring about the discovery of new chemical leads, including modification of old chemical leads, modifications of known drugs, new chemical ideas, exploitation of clinical and pharmacological side effects and intervention at a new point in the disease process, as well as the use of structural and mechanistic information in rational design. Each of these processes is associated with its own pitfalls and benefits, although an understanding of the 3D structure of the target and its interaction with ligands early in a project can give valuable information. However, it is the screening of chemical libraries in relevant bioassays that is still really the most important mechanism for lead generation. It is during lead identification that efforts are made to develop biological tests capable of reliably measuring some of the properties required in a development candidate. Nearly all aspects of novel chemical entity testing involve the concept of protein–ligand interaction, according to the law of mass action, derived from receptor theory1,2 (Figure 3.2). It is in the characterization of these interactions that ITC has its value in the drug discovery process. The hit to lead process is characterized by chemistry aimed at establishing structure–activity relationships (SARs), which are used in the optimization process. Several factors may influence the ability to find leads, including the type of assay and the profile of the compound set that is screened.

Figure 3.2 Binding interactions for receptors and enzymes. The symbols are as follows: R, receptor; L, ligand; E, enzyme; S, substrate; P, product. Various rate constants are shown, k

Lead optimization Once one or more suitable series of related lead compounds have been identified, the effort is then concentrated on enhancing activity, selectivity and bioavailability. This lead optimization phase typically lasts for around 2 years, and leads to three to four compounds with potential for concept testing in humans. Lead optimization is carried out using experience gained from other projects or series, substituting bioisosteric groups and making use of experimental structures and modelling.

EXPERIMENTAL MEASUREMENT

63

Bioisosteric replacement involves the substitution of atoms, or groups of atoms, in the original molecule to produce new compounds with improved biological properties. 3D structure information on compounds interacting at a particular target site, coupled with molecular modelling, may lead to alternative pharmacophores being proposed. These models can then be used to drive the iterative cycle of design and optimization of further chemical series. Obtaining functional efficacy is still a significant problem for drug discovery, as the mechanisms contributing to the functional response may not be well understood. Achieving bioavailability may be equally difficult, and often highly potent and selective compounds can lack activity in vivo. Metabolism and inability to reach the site of action (for example due to poor absorption) can limit bioavailability. The biological activity of any candidate drug is profoundly influenced by binding to a range of proteins: the target, side-effect proteins, plasma proteins, transport proteins and metabolizing proteins. ITC may be used to measure the strength of these diverse interactions.

3.3 Experimental measurement of thermodynamic binding parameters During a typical ITC experiment, a sample of the target protein is placed inside the thermally insulated calorimeter and heat changes are measured as aliquots of the test compound are introduced using a syringe. The experiment is usually designed so that most of the test compound becomes bound during the initial injections. Later in the titration, the magnitudes of the heat changes decrease as the sites on the target protein become saturated (Figure 3.3). Fitting an equation to these dose–response data can allow the estimation of the magnitudes of the association constant (Ka ¼ 1=Kd ), the standard enthalpy change and the stoichiometry of binding.3,4

Instrumentation The basic design of isothermal titration calorimeters has changed relatively little over the last 10 years, with most of the commercially available instruments operating in a differential or cell feedback mode. This involves the temperature difference between the sample and reference cells being continuously monitored, and a constant power being applied to the reference cell that activates the feedback circuit, providing a variable power to the sample cell. The voltage applied to the sample cell varies proportionally to the heat generated in or absorbed by the sample cell, and it is this applied voltage

64

ITC IN DRUG DISCOVERY

Figure 3.3 Typical thermogram for the binding of a triazine to the 24 kDa fragment from the gyrase B subunit

that is the experimentally measured quantity. Integration of this power with respect to time gives the heat associated with the reaction. However, significant progress has been made in increasing the sensitivity of ITC instruments, so that they may now measure heat effects as low as 0.1 mJ (VP-ITC from Microcal, www.microcalorimetry.com, and Nano ITC from CSC, www.calscorp.com). This corresponds to around a 10–20-fold improvement over instruments such as Microcal’s Omega, which marked the entry into calorimetric technology for many drug-hunting companies. This improvement, coupled with advances in recombinant DNA technology and protein purification, has allowed ITC to be used increasingly in drug discovery to monitor interactions of biological significance. Several publications give detailed descriptions of these older- and newer-generation instruments.5–7

Experimental data collection and analysis This topic has been covered in detail elsewhere 6,8–10 (see also chapter 2 of this book) and here will be covered only briefly. High-quality ITC data may be obtained relatively quickly as the detection method is unchanged from system to system, therefore reducing the need for specific optimization of the method. However, assay design is still an important consideration, and the ITC experimental conditions will be dictated by the objective of the experiment.

EXPERIMENTAL MEASUREMENT

65

Often this objective is to measure the apparent enthalpy of binding as well as the affinity and stoichiometry of the interaction. The choice of protein and ligand concentrations should be such that data are obtained that give information on the magnitude of DH (points at the beginning of the binding isotherm), on the affinity (related to the mid-point of the corrected total heat change) and on the heat of dilution and mixing (points at the end of the isotherm) (Figure 3.3). A protocol can often be designed that allows such determinations to be made. It is important to note that the accuracy of the measured thermodynamic parameters is dependent upon the accuracy of the determination of the concentration of the interacting reagents. Efforts should be made to ensure that the functional concentration of these reagents is known prior to conducting the ITC experiment. Heat effects observed during the titration are composed of several components, the heat of complex formation, the heat of dilution of the protein, the heat of dilution of the ligand, the heat of mixing and the heat associated with any further equilibria linked to the binding reaction. Control titrations are required in order to correct for the dilution and mixing effects. The presence of additional equilibria may be checked for in a variety of ways, including looking at the shape of the binding isotherm (symmetry), checking that the magnitude of the heats of dilution are similar to those points at the end of the titration, performing titrations at a range of concentrations and performing the titration with the opposite arrangement of reagents in the cell and syringe. Data fitting is normally carried out by non-linear regression using software packages such as ORIGIN. A single independent binding site is often the most appropriate model for data fitting, although more complex models can also be applied if appropriate. However, it is usually difficult to deconvolute heat signals from dissimilar or cooperative binding sites. As is usual in model discrimination, it is necessary to determine whether the models adequately describe the data, and whether one model is better than another. Some criteria that have been useful for evaluating goodness of fit and model discrimination are convergence in the regression analysis, meaningful parameter values with low standard errors, randomly distributed residuals and a low residual sum of squares.11 ITC is the only technique that measures Ka , DH and n (stoichiometry) in a single experiment. After correction of DH for any linked equilibria effects, to obtain DH8, the magnitudes of DG8 and DS8 are also obtained from the relationship DG8 ¼ DH8
TDS8 ¼
RT ln Ka

ð3:1Þ

where R is the molar gas constant, T is the experimental temperature and 8 signifies a parameter estimate relating to standard thermodynamic conditions.

66

ITC IN DRUG DISCOVERY

With relatively few extra experiments at different temperatures, ITC also allows measurement of an additional, central thermodynamic parameter, DCp (the change in heat capacity at constant pressure): DCp ¼ dðDH8Þ=dT

ð3:2Þ

Measurement of DCp in addition to DG8, DH8 and DS8 allows a full thermodynamic characterization of the binding interaction, as it is this parameter that controls the temperature variation of the other three parameters.10 Equation (3.1) is an integrated form of the van’t Hoff equation, and suggests that the variation in the measured values of the binding constant with temperature should lead to estimation of DH8. Comparison of enthalpies determined directly by ITC with those determined from van’t Hoff analysis of the binding constant, even where the binding constant is determined from the same calorimetric experiment, has led to discrepancies between enthalpy values.12 Analysis of van’t Hoff enthalpies using Equation (3.1) is a simplistic approach, and determination of binding enthalpies from affinity measurements alone requires slightly more complex analysis. Equation (3.1) assumes that DH8 is constant over the experimental temperature range. However, DCp is rarely zero, and so DH8 is found to vary with temperature, leading to the requirement for the DCp term to be included for the determination of van’t Hoff enthalpies: DHvHref
Tref DCp ln Ka ¼ R




1 1
Tref T

 þ

DCp T ln þ ln Karef R Tref

ð3:3Þ

where ln Ka is plotted against T and the variable parameters are DHvHref (the van’t Hoff enthalpy at a reference temperature, Tref ), ln Karef (Karef is the association constant at the reference temperature) and DCp (the temperatureindependent change in heat capacity). Further problems arise because enthalpy–entropy compensation (see below) causes DG8 to change relatively little with temperature, leading to a poor signal to noise ratio, the van’t Hoff relationship must derive two parameters (DH8 and DCp ) from changes in DG8, leading to correlation of the estimated parameters, and furthermore DCp may itself be temperature dependent. These factors, as well as other possibilities, such as cooperativity, non-ideal behaviour and assay artefacts, may have contributed to recent discrepancies between van’t Hoff and calorimetrically determined enthalpies.12–15 Both calorimetric and van’t Hoff enthalpies are representative of the same closed thermodynamic system, and should, in the absence of cooperative effects, and as long as all variables are allowed to change with temperature, be equivalent (DH8cal ¼ DH8vH ),16,17 even in the presence of linked equilibria.18,19

67

ITC IN DRUG DISCOVERY

Range and uncertainty in measured parameter values For any experimental technique, it is useful to know the potential range of the parameters and the uncertainty associated with measurements. In order to provide useful information, the variation of a measured parameter between compounds must be greater than the uncertainty in its magnitude. The range of DG8 values measured by ITC is limited by both the sensitivity of the instrument and the relatively narrow window of free energies that have a functional significance.20 This narrow window, corresponding to around
15 to
50 kJ mol
1 (Kd  1 mM to 1 nM), gives DG8 measurements the lowest signal to noise ratio for any of the thermodynamic parameters measured by ITC. There are no constraints on the values of DH8 and DS8, other than their combination producing a free energy lying in the functional range. The information content of DG8 is high, however, as this measures affinity, rather than just contributing to it, as do DH8 and DS8. DH8 has a larger signal to noise than DG8. The calculation of TDS8 as the difference between DH8 and DG8 leads to a less precise estimation of the entropy change. ITC measures extremely small heat changes, which can potentially relate to many sources of error, including evaporation, mixing and adsorption. However, these sources of error can usually be eliminated by the appropriate use of calibration, reference cells and controls. The result is that, compared with many other techniques, ITC provides a more precise measure of affinity. Typical ranges and standard errors of the thermodynamic parameters measured by ITC during drug discovery are given in Table 3.1.

3.4 ITC in drug discovery Characterization of target proteins The supply of sufficient quantities of purified protein is critical to drug discovery. Although high purity is a well recognized quality criterion for supply of protein for structure determination, functional criteria are often overlooked. It is extremely important to verify that protein preparations Table 3.1 Typical values of thermodynamic parameters for drug-like molecules binding to proteins Parameter DG8 DH8 TDS8 DCp

Typical range

Typical SE

30 to 50 kJ mol1 80 to þ20 kJ mol1 60 to þ40 kJ mol1 1700 to þ450 J K1 mol1

5 1 kJ mol1 5 1 kJ mol1 1:5 kJ mol1 10%

68

ITC IN DRUG DISCOVERY

function correctly prior to their use in assays, screening or structural studies. ITC is a powerful method for the quality assessment of protein preparations. This is not only due to the precision with which it measures affinity, but also because the stoichiometry and enthalpy of ligand binding are obtained. The stoichiometry may reveal the presence of non-functional protein within a preparation, whilst the affinity gives information on whether the active fraction is functioning correctly. The enthalpy of binding gives additional insight, because aberrant behaviour may not be detected as a change in affinity due to enthalpy–entropy compensation (see below). ITC has been used in our laboratory to assist in the choice between two different preparations of acyl carrier protein enoyl reductase (ACPER) for structural studies. Both samples showed similar, expected, affinities for the ligand, NADH, but with differing stoichiometries and enthalpies. The enthalpy and stoichiometry of the sample chosen for structural work were shown to be within 10 per cent of the parameters determined for a previous sample, which yielded crystals. ITC is particularly useful for the evaluation of protein constructs produced by recombinant DNA technology. Recombinant proteins are often produced as inactive fragments, autolysis incompetent mutants, or with tags in order to facilitate the purification process. The validity of this approach for producing proteins to be used within drug discovery depends upon the authentic and engineered protein following the same SAR. ITC may be informative in the validation of target proteins. Since ITC is a direct binding method, it can be used to avoid the problem of kinetic equivalence. Kinetic equivalence occurs in assays monitoring protein–protein interactions, since perturbing the interaction by binding to either protein will follow the same dose–response relationship. Performing ITC on each protein individually with the ligand of interest often allows definitive identification of the target.

Assay validation The ability of high-throughput and secondary screening assays to identify and characterize potent compounds depends upon the validity of the protocol. This is important in the drug discovery industry as the quest for higher throughput and lower reagent consumption may lead to use of less rigorous assays. The high-precision data generated by ITC and the direct nature of the technique favour it as a method for assay validation. Many kinetic and binding assays employ immobilized proteins (for example scintillation proximity assays, ELISA methods and BIAcore). ITC can be used as a benchmark to evaluate the effects of immobilization on potency and ranking of compounds. ITC is also useful for scrutinizing the reliability of model substrates. Many assays use model peptides in place of protein substrates, as

ITC IN DRUG DISCOVERY

69

they offer a wide range of benefits including ease of measurement, preparation, purification and modification. Verifying that these model systems are adequate surrogates for the full-length protein in terms of affinity and thermodynamic profile is well suited to ITC. We have used ITC to reveal artefacts within a secondary screen aimed at measuring potencies of compounds capable of inhibiting the interaction of the PDGFR receptor with SH2 domains of the p85 subunit of PI-3 kinase. The ELISA-type assay used a fusion protein containing glutathione transferase (GST) and the Cterminal SH2 domain of p85 (SHZC) immobilized on 96-well plates, and aimed to measure the ability of compounds to perturb the binding of an 11residue biotinylated phosphopeptide (Bio-P-Pep11) derived from the PDGFR, including the sequence flanking pTyr751. A representative quinoxaline, compound 1 (Figure 3.4) was found to have an IC50 in the assay of 0.7 mM, however, no heat change was observed when this compound was titrated with isolated p85 SH2C domain. This suggested a lack of binding was consistent with NMR data, which showed Kd 4 mM. The isolated p85 SH2C domain was shown to be functional, in that it bound a fiveresidue phosphopeptide (P-Pep5) derived from the PDGFR, when monitored by either ITC or NMR. ITC failed to detect binding of compounds or phosphopeptides to the GST–p85-SH2C fusion protein, suggesting that much of the fusion protein was not functional. A further preparation of GST–p85SH2C, prepared by elution from the glutathione affinity column in urea (rather than glutathione, as used previously), followed by renaturation, yields fully functional protein, as determined by calorimetry studies of P-Pep binding. ITC did not detect quinoxaline compound binding to the refolded fusion protein, either directly or in competition experiments with phosphopeptides. In the original ELISA-type assay, the IC50 for the quinoxaline compound with the refolded fusion protein was unchanged, with no change in the binding capacity for Bio-P-Pep11. The combination of ITC and NMR data suggested that there was a change in the specificity of the fusion protein conferred by the immobilization process, and highlighted the importance of careful functional checks on specificity, affinity and stoichiometry whenever possible for such systems. In a related example21 the thermodynamic profile of binding of a 14-residue peptide from the p85 subunit of PI-3 kinase to the Fyn SH3 domain was found to be significantly different from the full-length p85.

Figure 3.4 Structure of quinoxaline compound 1

70

ITC IN DRUG DISCOVERY

Affinity and mechanistic characterization of active compounds Once compounds have been identified as hits or confirmed as leads, it is important to characterize the affinity for the target protein, and the molecular mechanism of compound action. It is the direct nature of ITC that makes it exceptionally informative regarding the influence of another ligand on the activity of the compound of interest. The presence of another ligand may be required for binding of the test compound, may compete with the test compound, either directly or indirectly, or may have no effect. In another drug discovery project directed towards finding inhibitors of myristoyl CoA:protein N-myristoyltransferase (NMT), ITC was used to show that a dipeptidecontaining molecule, compound 2 (Figure 3.5), bound more strongly to NMT–substrate complex than to free NMT. NMT follows an ordered mechanism by which the myristoyl coenzyme A (MyrCoA) substrate binds before the peptide substrate. Thus inhibitors that are analogues of MyrCoA are expected to favour binding to free NMT, and compounds that mimic peptide substrate are expected to bind preferentially to NMT.MyrCoA complex. The concentration of MyrCoA in yeast or rat liver cells is approximately 10 mM, whilst the Km values for this substrate are around 1 mM. Thus, under physiological conditions the concentration of NMT.MyrCoA is much higher than the concentration of free NMT. This has implications for inhibitor design, since an inhibitor that functions by binding to free NMT is likely to require much greater affinity than a compound that binds to NMT.MyrCoA complex. ITC experiments produced only heats of dilution for titrations of compound in the absence of MyrCoA, whereas titration into NMT.MyrCoA complex yielded an enthalpy approaching
63 kJ mol
1 and affinity (Kd ) of 0.6 mM. Similarly, the inhibitor SG58272 (Figure 3.6) was found to bind to NMT.MyrCoA complex but not free enzyme.22 The results suggest that, for these peptide-containing compounds, crystal structures of enzyme–compound complex would give misleading information for structurebased drug design. ITC has also been useful in the analysis of acylCoA recognition in this system.22–24 A further example of the use of ITC in the elucidation of the intermolecular complex that gives biological activity is demonstrated by studies on the inhibition of enoyl (acyl carrier protein) reductase by triclosan. Our studies

Figure 3.5 Structure of dipeptide-containing compound 2

ITC IN DRUG DISCOVERY

71

Figure 3.6 Structure of SG58272

have shown that inhibition by this compound requires prior binding of NADþ .25 X-ray crystallography shows that the bound NADþ forms part of the binding site for triclosan. However, crystals are also obtained for enzyme–triclosan complex in the absence of coenzyme. The 3D structure determined from such crystals would not be appropriate to increasing the understanding of the biological activity of triclosan. Additionally, it would not be a relevant structure on which to base the rational design of improved analogues. Finally, in an anti-infective drug discovery program at AstraZeneca, based on inhibition of the bacterial ligase MurC, ITC was used to evaluate the inhibition properties of compounds from rational design. This enzyme utilizes ATP to catalyse the ligation of alanine to UDP-N-acetylmuramic acid. A prototype analogue was designed, based on the proposed tetrahedral transition state of the enzyme26 (Figure 3.7). This compound was found to be a potent (IC50 ¼ 42 nM), reversible, time-independent inhibitor of MurC. Kinetic studies revealed mixed type inhibition with respect to all three substrates for this compound, suggesting that it bound to multiple forms of

Figure 3.7 (a) Proposed MurC enzyme reaction transition state intermediate with a tetrahedral carbon centre and (b) phosphinate inhibitor

72

ITC IN DRUG DISCOVERY

Figure 3.8 Schematic representation of the observed binding to several enzyme forms, exhibited by phosphinate inhibitor (Kd values represent the mean values obtained from non-linear regression fitting to the calorimetric thermograms)

the enzyme.27 These findings were in contrast to predicted behaviour, which suggested a single dead-end E.ATP.inhibitor complex. Direct binding studies using ITC clearly demonstrated potent binding in the presence of ATP, as expected. However, unexpected binding was detected to two other enzyme forms. The inhibitor was shown to bind to the free enzyme and to a productbound form of the enzyme (E.ADP, Figure 3.8) with moderate affinities (Kd values from 0.5 to 2 mM). The results suggest that the compound acts as a transition state/multisubstrate analogue to produce potent inhibition, but that it can also serve as a product analogue in weaker, but kinetically relevant, complexes. These studies suggested that the design and testing of further compounds developed in this program should account for binding to several different enzyme forms to enable accurate interpretation of SAR.

Thermodynamics and structure-based design In certain circumstances, interpretation of binding thermodynamics may be relatively straightforward, for example the detection of proton movement on binding in buffers with different enthalpies of ionization. However, the detailed interpretation of binding thermodynamics may not always be trivial. It is important to consider that the measured parameters reflect differences

ITC IN DRUG DISCOVERY

73

between the reactants and the products. That is, DG8 and its constituent enthalpic and entropic components depend on differences between free and bound states for both of the interacting partners. Ideally 3D structural information on the free partners as well as the protein–ligand complex should be combined with the thermodynamic data in order to provide a detailed characterization of the binding process. The situation is complicated further when the structure of the ligand is modified during drug discovery. Sometimes the changes in affinity (DDG8) and in the other thermodynamic parameters are mistakenly assigned solely to modified interactions between the ligand and the protein. However, account must be taken of the reorganization terms reflecting the changes in thermodynamics on moving from the original to the modified ligand (Figure 3.9). This may be extremely important in medicinal chemistry and biomolecular SAR (e.g. alanine-scanning mutagenesis), as modified groups on a ligand that have been demonstrated to contribute increased interactions with the protein in a crystal structure may actually interact equally well, or better, with solvent water. This may mean that the modified group may not enhance, or could even decrease, affinity. Thus, the dependence of the thermodynamic parameters on differences between free and bound states, and their interactions with solvent and other solutes, highlights the fact that 3D structural information reveals only some of the factors affecting affinity. Biomolecular recognition is further complicated by the fact that each of the molecules and complexes is distributed between a plethora of conformations, which differ in energy. It is the combination of structural and thermodynamic data that has the greatest potential to support computational chemistry in drug discovery. The most reliable information comes from studies utilizing small changes in structure, where ITC data on the interaction are combined with structural data on both the free partners and the complex. Favourable systems capable of generating such detailed information are, unfortunately, rare.

Figure 3.9 Scheme for interpretation of the thermodynamic structure activity relationships (SARs) encountered during drug discovery. The parameter J may represent any of the thermodynamic parameters (DG8, DH8, DS8, DCp ). DDJ8 is given by (DJ80  DJ8Þ ¼ ðDJ8PL  DJ8L ), where DJ8L represents the reorganization term

74

ITC IN DRUG DISCOVERY

However, it is possible to use less comprehensive studies to aid the lead optimization phase of drug discovery. Such structural and thermodynamic information was used in our project working on inhibitors of DNA gyrase. In E. coli this enzyme, which catalyses the ATP-dependent supercoiling of DNA, is a tetramer of two subunit types, A and B, with molecular masses of 97 and 90 kDa respectively. The B subunit contains the ATP-binding site. It was not possible to generate preparations of the large, native enzyme for biomolecular studies. However, ITC was used to show that catalytically inactive 43 kDa and 24 kDa fragments of the N-terminal domain of the B subunit were valid models for the corresponding parts of intact gyrase. The binding of the antibiotic novobiocin,28 and of a series of inhibitors, the triazines10,29 (Figure 3.10), to the 24 kDa fragment, as well as a mutant form (Arg-136!His) of this fragment, was investigated by ITC. Compound 3 was the first triazine to yield a 3D structure by protein crystallography. Each of the anilino substituents is located in a subsite, which may be identified by the key residues in the gyrase protein: Asp-73, Arg-136 and Ile-94. Five different modifications (compounds 4–8) to any one of the aniline substituents produced only small changes in the energetics for binding to the 24 kDa fragment, compared with compound 3. DDH8 was found to range from
5:9 to þ6:7 kJ mol
1 , whilst DDG8 was smaller at
1:3 to þ4:2 kJ mol
1 . 3D structural data on four of these complexes (3–5 and 8) indicate that the charged substituent extends toward solvent via the Ile-94 subsite. These small energetic changes seem to reflect modified hydrogen bonding arrangements. A much larger change in enthalpy was found for compound 9, where DDH8 ¼
36:0 kJ mol
1 , although this compound bound with similar affinity (a less than two-fold change in Kd , DDG8 ¼
1:7 kJ mol
1 ). NMR suggests that the charged propanoate group reaches solvent via the Arg-136 subsite, so this could be considered to be a change in binding orientation, relative to the other triazines. The relatively large number of modifications to these series of compounds, especially between compound 3 and 9, may influence DH8, making it difficult to explain the large DDH8, although the altered binding mode almost certainly makes a large contribution. Relatively large changes in enthalpy have also been observed for inhibitors binding at different subsites in stromelysin.30 The results described above illustrate how differences in binding mode may be highlighted more readily by the measurement of DH8, rather than DG8. This is because compensation masks changes in DG8 more than changes in DH8 (see below). Measured enthalpy values can therefore emphasize discontinuities in SAR that are not apparent as shifts in binding affinity, and so help to reduce the errors associated with molecular design based on extrapolations from the 3D structures of related complexes. Large ( 4 10 kJ mol
1 ) changes in enthalpy with a related compound suggest that further iterative design should be based upon an experimentally determined

ITC IN DRUG DISCOVERY

75

Figure 3.10 Structures of the triazine compounds

structure of this protein–ligand complex, because predicted interactions may prove unreliable. The application of ITC may therefore be of value in projects for which a change in binding mode is likely, for example for proteins that crystallize in multiple conformations, or that have multiple binding sites, or when there is structural symmetry in the compounds of interest. A further example of the advantage of combining structural and thermodynamic data in the characterization of potent compounds is the work carried out on rosuvastatin (Crestor, a trademark of AstraZeneca). The

76

ITC IN DRUG DISCOVERY

statin inhibitors have high affinities for the enzyme 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMG-CoAR). The structure of rosuvastatin resembles that of an intermediate in the catalytic reaction, mevaldyl-CoA (Figure 3.11). The tighter binding of catalytic intermediates relative to substrates may explain the over 50 000-fold increase in affinity of rosuvastatin (DG8 ¼
59:8 kJ mol
1 ) compared with HMG-CoA (DG8 ¼
31:0 kJ mol
1 ). C5 is tetrahedral for both rosuvastatin and mevaldyl-CoA, and trigonal for HMG-CoA (Figure 3.11). A carbonyl oxygen on C5 of HMG-CoA is replaced by a hydroxyl group on mevaldyl-CoA and rosuvastatin, which appears to facilitate interaction with the carboxyl group of Glu-559, which is probably charged. Several groups on rosuvastatin form multiple hydrogen bonds with the enzyme. This may help to explain the strongly enthalpy-driven nature of rosuvastatin binding (DH8 ¼
69:0 kJ mol
1 ). It is possible that the entropic penalty normally experienced when hydrogen bonds are formed is smaller for the second and subsequent interactions when a group makes multiple hydrogen bonds, although overall there is still an unfavourable
1 entropy change for rosuvastatin binding (DS8 ¼
29:7 J K
1 mol ). In addition to the formation of multiple hydrogen bonds, rosuvastatin also makes more hydrophobic interactions (usually entropically favourable) with the enzyme than does the substrate, although the entropy change for binding
1 of HMG-CoA (DS8 ¼ þ76:9 J K
1 mol ) is more favourable than for rosuvastatin. A further entropic advantage would appear to be conferred to the binding of rosuvastatin, as the helix La11 (Gly-860 to C terminus) is disordered in the complex with rosuvastatin. These results highlight how ITC and structural data can be used to understand molecular interactions of potent compounds, and they also demonstrate that care must be exercised in interpreting the observed thermodynamics.

Figure 3.11 Structure of rosuvastatin, a reaction intermediate (melvaldyl-CoA) and the substrate (HMG-CoA)

ITC IN DRUG DISCOVERY

77

Enthalpy–entropy compensation Enthalpy–entropy compensation is the tendency for changed interactions producing a more negative DH8 to be associated with a more negative DS8 and vice versa. This is perhaps unsurprising, as increased bonding, producing more negative DH8 values, tends to occur at the expense of increased order, leading to a more negative DS8. Since both DH8 and DS8 are dependent upon DCp , this approximately linear correlation may be expected. Further, as mentioned above, the DG8 range measurable by ITC is around
15 to
50 kJ mol
1 . This limited window of free energy may help to explain the observation of enthalpy–entropy compensation for a wide range of biomolecules and different types of interaction.20 If DDH8 and/or TDDS8 are large compared with DDG8 for a series of compounds, enthalpy–entropy plots tend to appear linear with a slope of magnitude  T. Marked enthalpy– entropy compensation was observed for the binding of 11 triazines (including the seven compounds listed above) to the 24 kDa gyrase fragment (Figure 3.12). Enthalpy–entropy compensation is both a challenge and a benefit for the medicinal chemist, since the phenomenon must be overcome in the drive for affinity, but may allow substitution often with little affinity penalty, once sufficient affinity has been acquired. An important consequence is that changing the structure of an interacting group at the protein–ligand interface often has a small effect on affinity, and a larger effect on the enthalpy of binding. Thus, ITC can reveal changed interactions, which may be more difficult to detect using other functional assays in the absence of 3D structural data.

Figure 3.12 Enthalpy–entropy compensation for triazines binding to the 24 kDa fragment of DNA gyrase. The solid line represents the linear regression fit (correlation coefficient 0.997) with slope ¼ 314 K, close to the experimental temperature of 300 K

78

ITC IN DRUG DISCOVERY

3.5 Summary Drug discovery is an expensive and time-consuming process, directed towards the identification and optimization of compounds with the potential to become medicines. ITC is now an established technique within this process. Although currently used predominantly in the later stages (lead identification– lead optimization) of the drug discovery process, improvements in technology and understanding should help to widen the application of the method. Increased sensitivity, miniaturization and throughput, leading to instruments that could characterize biomolecular interactions in 96-well plates, could see the introduction of calorimetric methods into high-throughput screening laboratories. Novel calorimetric methods and integrated circuit calorimeters31 are already beginning to take calorimetry in this direction. Further increasing our understanding of biomolecular interactions may take longer. Although the combination of ITC data and 3D structural information has helped to improve our comprehension of the forces involved in molecular recognition, characterizing the individual components implicated in non-covalent interactions is still extremely difficult. A major advance would be the capability of the medicinal chemist to control the phenomenon of enthalpy–entropy compensation during compound optimization.

References 1. 2. 3. 4. 5. 6.

7.

8. 9. 10. 11. 12. 13. 14. 15.

Langley JN. (1905) J. Physiol. (Lond) 374–413. Ehrlich P. (1909) Ber. Deutsch. Chem. Ges. 42: 17–47. Wiseman T, Williston S, Brandts JF and Lin LN. (1989) Anal. Biochem. 179: 131–137. Freire E, Mayorga OL and Straume M. (1990) Anal. Chem. 62: 950A–959A. Blandamer MJ. (1998) In Biocalorimetry: Applications of Calorimetry in the Biological Sciences, Ladbury JE and Chowdry BZ (eds), Wiley, Chichester, pp. 5–25. Cooper A and Johnson CM. (1994) In Methods in Molecular Biology, Vol. 22: Microscopy, Optical Spectroscopy, and Macroscopic Techniques, Jones C, Mulloy B and Thomas AH (eds), Humana, Totowa, NJ, pp. 137–150. O’Brien R, Ladbury JE and Chowdry BZ. (2001) In Protein–Ligand Interactions: Hydrodynamics and Calorimetry, Harding SE and Chowdry BZ (eds), Oxford University Press, Oxford, pp. 263–286. Bundle DR and Sigurskjold BW. (1994) Methods Enzymol. 247: 288–305. Holdgate GA. (2001) BioTechniques 31: 164–184. Ward WHJ and Holdgate GA. (2001) Prog. Med. Chem. 38: 309–376. Mannervik B. (1982) Methods Enzymol. 87: 370–390. Naghibi H, Tamura A and Sturtevant JM. (1995) Proc. Natl Acad. Sci. USA 92: 5597– 5599. Liu Y and Sturtevant JM. (1995) Protein Sci. 4: 2559–2561. Liu Y and Sturtevant JM. (1997) Biophys. Chem. 64: 121–126. Thomson J, Liu Y, Sturtevant JM and Quiocho FA. (1998) Biophys. Chem. 70: 101–108.

REFERENCES

79

16. Cooper A and Johnson CM. (1994) In Methods in Molecular Biology, Vol. 22: Microscopy, Optical Spectroscopy, and Macroscopic Techniques, Jones C, Mulloy B and Thomas AH (eds), Humana, Totowa, NJ, pp. 109–124. 17. Cooper A. (1999) Curr. Opin. Chem. Biol. 3: 557–563. 18. Horn JR, Russell D, Lewis EA and Murphy KP. (2001) Biochemistry 40: 1774–1778. 19. Horn JR, Brandts JF and Murphy KP. (2002) Biochemistry 41: 7501–7507. 20. Cooper A, Johnson CM, Lakey JH and No¨llmann M. (2001) Biophys. Chem. 93: 215– 230. 21. Renzoni DA, Pugh DJR, Siligardi G, Das P, Morton CJ, Rossi C, Waterfield MD, Campbell ID and Ladbury JE. (1996) Biochemistry 35: 15 646–15 653. 22. Bhatnagar RS, Schall OF, Jackson-Machelski E, Sikorski JA, Devadas B, Gokel GW and Gordon JI. (1997) Biochemistry 36: 6700–6708. 23. Bhatnagar RS, Jackson-Machelski E, McWherter CA and Gordon JI. (1994) J. Biol. Chem. 269: 11 045–11 053. 24. Bhatnagar RS and Gordon JI. (1995) Methods Enzymol. 250: 467–486. 25. Ward WHJ, Holdgate GA, Rowsell S, McLean EG, Pauptit RA, Clayton E, Nichols WW, Colls JG, Minshull CA, Jude DA, Mistry A, Timms D, Camble R, Hales NJ, Britton CJ and Taylor IWF. (1999) Biochemistry 38: 12 514–12 525. 26. Reck F, Marmor S, Fisher SL and Wuonola MA. (2001) Bioorg. Med. Chem. Lett. 11: 1451–1454. 27. Marmor S, Petersen CP, Reck F, Yang W, Gao N and Fisher SL. (2001) Biochemistry 40: 12 207–12 214. 28. Holdgate GA, Tunnicliffe A, Ward WHJ, Weston SA, Rosenbrock G, Barth PT, Taylor IWF, Pauptit RA and Timms D. (1997) Biochemistry 36: 9663–9673. 29. Ward WHJ, Holdgate GA, Pauptit RA, Breeze AL, Timms D, Block MH, Poyser JP, Hales NJ, Tunnicliffe A and Weston SA. (1999) 2nd International Conference on Applications of Biocalorimetry, Halle, Germany. 30. Sarver SW, Yuan P, Marshall VP, Petzcold GL, Poorman RA, DeZwaan J and Stockman BJ. (1999) Biochim. Biophys. Acta 1434: 304–316. 31. Lerchner J, Wolf A and Wolf G. (1999) J. Therm. Anal. Calorim. 57: 241–251.

4 Dissecting the Thermodynamics of DNA–Protein Interactions Torleif Ha¨rd

4.1 Introduction A deeper understanding of the thermodynamics of biomolecular interactions requires that we come to terms with, and understand how to dissect, the effects of many coupled molecular events that collectively contribute to the interaction (binding) free energy under any given conditions. This is a very challenging task and requires a database of calorimetric measurements on several model systems and the structure determination and biophysical characterization of the free and uncomplexed components in solution. It also requires further studies and an advance in the theoretical treatment of biophysical data to allow parametrization of structural and dynamic effects. Significant progress is being made in many laboratories and it is put to practical use in for instance drug design; see for example reference 1. In this chapter we illustrate how the thermodynamic profiles for binding of proteins to DNA can be analysed in terms of three mechanisms that can dominate the observed binding thermodynamics: the hydrophobic effect, linked protonation equilibria and folding of a protein upon binding. We also suggest a simple procedure to dissect the binding entropy into contributions that can be verified separately and (in most cases) experimentally.

Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

82

THERMODYNAMICS OF DNA–PROTEIN INTERACTIONS

4.2 Model systems Figure 4.1 shows the structures of three protein–DNA complexes for which the binding thermodynamics has been thoroughly characterized, and for which also the free protein components have been characterized with regard to structure and dynamics. Thermodynamic profiles containing the calorimetrically measured binding free energy (DG8obs ), binding enthalpy (DHobs ), and

Figure 4.1 Structures of three protein–DNA complexes that serve as model systems for studies of binding thermodynamics. Top. The glucocorticoid receptor DNA-binding domain (GR DBD). The detail illustrates a hydrogen bond between the His451 side chain and the DNA phosphate backbone. Lower left. The archaeal Sso7d protein (nonsequencespecific binding). Lower right. The GCN4 transcription factor. The two figures of GCN4 illustrate the structural disorder of the DNA-binding helices in the uncomplexed state and that DNA binding in this case is coupled to folding. Reprinted with permission from Nature (Nature Struct. Biol. 7, pp. 11–13). Copyright 2000 Macmillan Magazines Limited

COMPARISON WITH THE HYDROPHOBIC EFFECT

83

binding entropy (DS8obs ) are shown in subsequent figures throughout this chapter. The three model systems discussed here are the following. (1) The glucocorticoid receptor DNA-binding domain (GR DBD). This zinccoordinating domain is present in all nuclear receptor transcription factors. The isolated GR DBD is a monomer in the free state, but binds to its cognate DNA sequence (the glucocorticoid response element, or GRE) as a dimer. The structures of the free proteins and protein–DNA complexes of this and many homologous systems have been determined by NMR and X-ray crystallography (respectively, for instance, references 2 and 3). (2) The Sso7d protein from Sulfolobus solfataricus. The function of this archaeal protein is not precisely known, but it is thought to play a role in DNA packing in the cell. Sso7d binds without DNA sequence specificity to the minor groove of double-stranded DNA. A number of structural and biophysical studies have been conducted on Sso7d4–7 and the closely related Sac7d protein. (3) The basic leucine zipper (bZIP) C-terminal region of the yeast GCN4 transcription factor. The bZIP domains form dimeric a-helical motifs that recognize specific DNA sequences. This protein has not been studied in our laboratory, but data on structures,8 dynamics9 and binding thermodynamics10 are available in the literature.

4.3 Comparison with the hydrophobic effect As a basis for further discussion it is worth illustrating the very large similarity of binding of well folded proteins to DNA on one hand, and the hydrophobic effect on the other. Here, we refer to the hydrophobic effect as it is defined by for instance reference 11; as the characteristic and very strong temperature dependence observed for the transfer of nonpolar substances from water to organic solvent; that is a large and negative heat capacity change (DCp ). Molecular interpretations of the effect include dehydration, cavity formation and/or change in supermolecular water ordering,12 but a detailed discussion of these is outside the scope of this chapter. A large number of studies have addressed and discussed the contribution of the hydrophobic effect to a range of biomolecular recognition processes, and Figure 4.2 provides an illustration: sequence-specific binding of the GR DBD (gene regulation in human cells) proceeds with a thermodynamic profile that is strikingly similar to that observed for transfer of cyclohexane from water solution into its own phase. (The scaling might be considered arbitrary, because it can for instance be

84

THERMODYNAMICS OF DNA–PROTEIN INTERACTIONS

adjusted by comparing dimeric GR DBD binding to transfer of two moles of cyclohexane.) Hence, it appears that (a) nature has evolved biomolecular recognition processes to make significant use of the hydrophobic effect and (b) the hydrophobic effect provides the background against which we have to evaluate other effects.

Figure 4.2 Role of the hydrophobic effect. Comparison of thermodynamic profiles for sequence specific DNA binding by the GR DBD (left; data from reference 5) and transfer of cyclohexane from water into its own phase (right). In the case of cyclohexane, the large temperature dependence in DH and DS and the typical convergence temperatures TH and TS (at which these quantities are equal to zero) are the thermodynamic fingerprint of the hydrophobic effect. Experimental data on GR DBD are shown with error bars, and the lines drawn outside the experimentally accessible temperature window are extrapolated assuming a temperature-independent DCp

4.4 Protonation and charged–charged hydrogen bonds Next, we would like to show an example of how specific interactions can be distinguished. If isothermal titration calorimetry (ITC) binding isotherms on the GR DBD–GRE complex are measured at lower salt concentrations than that of Figure 4.2, in buffers with different ionization enthalpies one finds that the measured binding enthalpy is a linear function of the buffer ionization enthalpy.13 The origin of the phenomenon is that complex formation results in a changed pKa of one or several residues or groups at the protein or DNA. The binding event is therefore coupled to uptake or release of protons from the buffer and the buffer ‘reports’ on this event via its heat of ionization. A thorough theoretical account of the effect of linked protonation events and how to analyse these has been published.14 The number of protons involved (and the on/off direction of the protonation) is given by

PROTONATION AND CHARGED–CHARGED HYDROGEN BONDS

DHcal ¼ DHbind þ DnHþ DHion

85

ð4:1Þ

where the indices ‘cal’, ‘bind’ and ‘ion’ indicate the measured, the intrinsic (no protonation effects) and buffer ionization enthalpies, respectively, and DnHþ is the number of protons involved in the reaction. For GR DBD binding at 100 mM NaCl in different buffers at pH ¼ 7:5 one can fit a value DnHþ ¼ þ 0:98, which is equivalent to stating that about one proton per GR DBD monomer is taken from the buffer upon binding to DNA.13 The site-specific location of the proton can be inferred by analysing binding enthalpies measured as a function of pH and verified by inspecting the structure of the complex. Figure 4.3 shows binding enthalpies and free energies measured in Tris buffer at moderate salt concentrations. It is possible to account for the pH dependences by the protonation of a histidine residue (His451) upon binding. The relevant parameters are in this case the ionization enthalpy of histidine (the literature value is DH His ¼ 7:1 kcal mol
1 , and a value in the range 6–7.3 can be deduced from the salt concentration dependence not shown here), the pKa of this residue in the free state (a pKa ¼ 5:9 was measured by NMR), the ionization enthalpy of Tris (DH Tris ¼ 11:4 kcal mol
1 ) and the value of DH0 ¼ DHbind þ DH His ¼ 3 kcal mol
1 , with the two adjustable parameters being the pKa in the bound state and the binding constant at high pH (where no protonation occurs). The dashed curves in Figure 4.3 correspond to a shift of the pKa of His451 from 5.9 to 7.9 for binding in buffers containing 150 mM NaCl and 2 mM MgCl2.13 There are two histidines in the GR DBD and one of these – His451 – is located very close to a DNA phosphate at an orientation that is consistent with a charged–charged hydrogen bond in the protonated state, as shown in

Figure 4.3 Protonation and charged–charged hydrogen bond. Enthalpies (left) and free energies (right) of binding of GR DBD to DNA in Tris buffer as function of pH. The dashed lines correspond to the expected behaviour if the pKa of a histidine side chain is shifted from 5.9 to 7.9 upon binding. The specific interaction is illustrated in Figure 4.1

86

THERMODYNAMICS OF DNA–PROTEIN INTERACTIONS

Figure 4.1 (top). Hence, one can conclude that the observed buffer effects on binding enthalpy arise due to this interaction. The interaction is probably also biologically significant because the binding constant is strongly dependent on pH in the physiologically relevant pH interval and salt concentration, and His451 is completely conserved throughout the family of nuclear receptors. The change in pKa , and the linked protonation, is not observed at salt concentrations higher than  300 mM NaCl. An analysis of the salt dependence (not shown here) shows that a competition between protonation (binding of Hþ ) and association of a Naþ ion to DNA qualitatively accounts for this observation.13

4.5 Dissection of the binding entropy We would now like to outline some principles that can be used to qualitatively reproduce binding entropies (DSbind ) without resorting to structural interpretations and surface area calculations. A detailed motivation and test of the procedures will be presented elsewhere. As a basis for the discussion we first write the entropy of binding as a sum of contributions from several physical effects: DSbind ðTÞ
DSHE ðTÞ þ DSPE þ DSrt þ DSconf

ð4:2Þ

where the subscripts indicate contributions to the total binding entropy (bind) from the hydrophobic effect (HE), polyelectrolyte effect (PE), loss of rotational and translational entropy (rt) and conformational entropy (conf). It is valid to separate these contributions as long as they do not affect one another via coupling terms. For instance, the polyelectrolyte effect is not expected to notably influence desolvation (water activity; hydrophobic effect) at low salt concentrations. It is reasonable to assume that the hydrophic effect provides the largest contribution to the heat capacity change and that the other terms are essentially temperature independent in the interval 10–40 8C. We therefore write ðT DSHE ðTÞ ¼

1 T




 dDHðT 0 Þ dT 0 dT 0

ð4:3Þ

TS

where the temperature dependence of the binding enthalpy is the binding heat capacity (DCp ). The integral in Equation (4.3) must be calibrated and for this we make use of the fact that the hydrophobic effect is characterized by convergence to zero entropy and enthalpy changes at common temperatures:

ENTROPY CONTRIBUTIONS TO THE SSO7D–DNA INTERACTION

87

TS and TH .15 Convergence temperatures for entropy changes fall in a narrow interval around 386 K. We therefore require that DSHE ðTS Þ ¼ 0 for TS ¼ 386 K

ð4:4Þ

The standard procedure to account for desolvation has been to first calibrate heat capacity in terms of buried surface area and then back-calculate the hydrophobic effect from optimized coefficients (proportionalities). Here, we instead obtain the contribution to binding entropy directly from the temperature dependence in DH8obs . Entropy changes arising from polyelectrolyte (salt) effects can be obtained experimentally, either from direct measurements of binding entropy as function of salt concentrations7 or based on the salt concentration dependence of the binding free energy, in which case it is assumed that this dependence resides in the entropy term.16 Entropy changes due to loss of rotational and translational freedom may be estimated based on theoretical arguments, although there is an ongoing discussion about how these should be formulated, which results in an uncertainty in this term. Here, we use a value of DSrt ¼
20  5 cal mol
1 K
1 for the formation of binary complexes and
40 cal mol
1 K
1 for ternary complexes at 10
4 M concentrations. These values, which encompass the results of several recent estimates,17 are larger in magnitude than those expected based on changes in cratic entropy18 but smaller than values calculated from the Sackur–Tetrode equation which applies to gases. The magnitude and uncertainty in DSrt is on the order of the accuracy that we aim for in this qualitative analysis, i.e. an accuracy in the calculated TDS around 5 kcal mol
1 at 293 K. The most difficult term to evaluate in Equation (4.2) is the change in conformational entropy upon binding. NMR relaxation data provide a powerful measure of the disorder which is manifested in dynamics19 but there are still some uncertainties associated with the interpretation as discussed below.

4.6 Entropy contributions to the Sso7d–DNA interaction To test the division of binding entropy we first compare calculations with experimental findings with the Sso7d protein. We have previously collected calorimetric data on the DNA binding of this protein as a function of temperature, buffer, salt concentration, and pH.7,20 Given these data and the value of DSrt mentioned above we obtain the binding entropy components shown in Figure 4.4. We find that the treatment (Equations (4.2)–(4.4)) reproduces the experimentally observed binding entropy, which is encouraging. The

88

THERMODYNAMICS OF DNA–PROTEIN INTERACTIONS

Figure 4.4 Thermodynamics of Sso7d binding to DNA. Left. Calculated contributions to binding entropy. Right. Experimental thermodynamic profile and the sum of the calculated binding entropy components. See the text for details

contributions from DSPE and DSrt are of opposite sign and cancel each other, whereas DSHE is larger in magnitude. The remaining term DSconf can be expected to be unfavourable for binding, but comparatively small, because Sso7d is well folded in the free state. We have recently measured NMR order parameters for backbone dynamics in the free and DNA-bound states and found that this assumption is correct (Berglund et al., manuscript in preparation). The difference between the calculated and observed entropies in Figure 4.4 might reflect changes in side-chain entropy, but the uncertainties of the calculations are too large for a meaningful discussion of this possibility.

4.7 Entropy contributions to the GCN4–DNA interaction To get a perspective on the results with Sso7d and to get an impression of the expected effects of coupled protein folding we must repeat the analysis with a DNA-binding protein that folds upon binding. GCN4 is such a protein and the thermodynamic profile for binding10 is shown in Figure 4.5. One can immediately note that the GCN4–DNA interation is enthalpy driven and strongly opposed by the entropy term at all temperatures. Previous analyses have led to the conclusion that this is indeed due to coupled folding.9,10 Here, we calculate DSHE and DSrt as described above, but detailed experimental data on the polyelectrolyte effect (salt concentration dependence) are not available. In this case we therefore estimate the polyelectrolyte effect contribution to DSPE ¼ 23  10 cal mol
1 K
1 . This estimate is based on the number of salt bridges observed in the structure of the GCN4–DNA complex8 and the

ENTROPY CONTRIBUTIONS TO THE GCN4–DNA INTERACTION

89

further assumption that this number reflects the valency from which the salt concentration dependence can be calculated.16 The entropy dissection is shown in Figure 4.5. Despite the uncertainties in DSPE and DSrt it is obvious that there is now a difference between the calculated
TðDSHE þ DSPE þ DSrt ) and observed
TðDS8obs ) of more than 60 kcal mol
1 , reflecting a substantial contribution from DSconf due to coupled folding and binding of GCN4. The contribution to DSconf that can be attributed to ordering of the protein backbone can be calculated from NMR relaxation data (S 2 order parameters) measured in the free and complexed states.9 The effect of the corresponding entropy change (DSbb  140 cal mol
1 K
1 ) is shown in Figure 4.5. The remaining binding entropy to be accounted for can be attributed to changes in sidechain entropy upon binding.9 However, it is important here to note three difficulties with calculations of entropy from NMR order parameters: (a) the order parameter only represents the dynamic averaging of the orientation of the amide HN vector in the protein backbone and might therefore not be an accurate measure of the backbone entropy as a whole; (b) the order parameter only reflects disorder that results in dynamics on the pico- to nanosecond timescales, but disorder can also establish itself on other timescales, and (c) the analysis assumes that individual motions are independent, whereas order parameters measured at different sites in the protein in fact could represent the same protein motion, which then will be ‘counted twice’. These uncertainties must be addressed in further studies before NMR order parameters can be considered as a quantitative, rather than qualitative, tool to calculate conformational entropies.

Figure 4.5 Thermodynamic profile for the binding of GCN4 to DNA and calculated contributions to binding entropy: (i) TDSHE (T ), (ii) T(DSHE (T) þ DSPE þ DSrt ) and (iii) T(DSHE (T ) þ DSPE þ DSrt þ DSconf,backbone ). Calorimetric data were taken from reference 10 and data on dynamics used to calculate the protein backbone contribution to DSconf are from reference 9

90

THERMODYNAMICS OF DNA–PROTEIN INTERACTIONS

4.8 Discussion We have illustrated how some effects on the thermodynamics of protein–DNA interactions can be dissected and understood. The starting point of the discussion is the hydrophobic effect, which can be completely dominant. The origin of the large contribution from the hydrophobic effect is that protein– DNA interaction interfaces are large and highly complementary and exclude most of the water that solvate these surfaces in the free DNA and protein components. Next, we describe how different buffers and variation of the reaction pH can reveal protonation events and the specific interactions that cause shifts in pKa values. This type of analysis is straightforward and informative, albeit sometimes laborious. Finally, we show that the assumption that the temperature dependence in DH is due to the hydrophobic effect allows for a dissection of this and other contributions to the binding entropy, DS. A test on a protein (Sso7d) for which an extensive biophysical characterization has been carried out shows that one can account for the entropy of binding with a precision in TDS of about 5 kcal mol
1 . Application on GCN4 then reveals the imprint on binding thermodynamics left by the folding of this protein upon binding. The method is in a preliminary stage of development and needs to be further elaborated and verified. Weak points that need to be addressed are for instance differences in the contributions from desolvation of polar and charged groups, the origin of weak temperature dependences also in DCp , and the relevance in the use and choice of convergence temperatures. There is also a need for further theoretical work on the interpretation of NMR relaxation data as discussed above. A more thorough discussion of these issues and an application on the GR DBD and the related estrogen receptor (ER) DBD will be presented elsewhere. We also note that a similar treatment should allow for an account of the enthalpy component using the TH convergence to calibrate the contribution from the hydrophobic effect. However, for this enthalpy data must be corrected for buffer effects and this problem is also currently being addressed.

Acknowledgements This research is supported by grants from the Swedish Research Council (VR) and the Knut and Alice Wallenberg Foundation (KAW) for NMR instrumentation.

References 1. Luque I and Freire E. (2002) Proteins Struct. Funct. Genet. 49: 181–190.

REFERENCES

2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20.

Luisi BF, Xu WX, et al. (1991) Nature 352: 497–505. Baumann H, Paulsen K, et al. (1993) Biochemistry 32: 13 463–13 471. Baumann H, Knapp S, et al. (1994) Nature Struct. Biol. 1: 808–819. Lundba¨ck T and Ha¨rd T. (1996) Proc. Natl Acad. Sci. USA 93: 4754–4759. Agback P, Baumann H, et al. (1998) Nature Struct. Biol. 5: 579–584. Lundba¨ck T, Hansson H, et al. (1998) J. Mol. Biol. 276: 775–786. Keller W, Koenig P, et al. (1995) J. Mol. Biol. 254: 657–667. Bracken C, Carr PA, et al. (1999) J. Mol. Biol. 285: 2133–2146. Berger C, Jelesarov I, et al. (1996) Biochemistry 35: 14 984–14 991. Dill KA. (1990) Science 250: 297. Chandler D. (2002) Nature 417: 491. Lundba¨ck T, van den Berg S, et al. (2000) Biochemistry 39: 8906–8916. Baker BM and Murphy KP. (1996) Biophys. J. 71: 2049–2055. Baldwin RL. (1986) Proc. Natl Acad. Sci. USA 83: 8069–8073. Record M, Ha J-H, et al. (1991) Meth. Enzymol. 208: 291–343. Brady G and Sharp K. (1997) Curr. Opin. Struct. Biol. 7: 215–221. Murphy KP, Xie D, et al. (1994) Proteins Struct. Funct. Genet. 18: 63–67. Yang D, Mok Y-K, et al. (1997). J. Mol. Biol. 272: 790–804 Lundba¨ck T and Ha¨rd T. (1996) J. Phys. Chem. 100: 17 690–17 695.

91

5 Salt Effects in Ribonuclease– Ligand Interactions: Screening or Competitive Binding? Kenneth P. Murphy, Travis T. Waldron and Greta L. Schrift

5.1 Introduction Numerous interactions are involved in stabilizing the complex of a protein and its ligand. These include the hydrophobic effect, hydrogen bonding and charge–charge interactions. A considerable body of data is available on the contributions of the hydrophobic effect and hydrogen bonding to both the enthalpy and entropy of protein–ligand interactions (e.g. reference 1). Fewer experimental data exist on the enthalpic and entropic contributions of charge– charge interactions. The interaction between two charges has been investigated for centuries. The energy of interaction between charges increases in magnitude with increasing charge, as given by Coulomb’s law. The energy of interaction between charges decreases with increasing salt concentration, as given by the Poisson–Boltzmann equation. However, other factors can contribute to the interaction between charged groups. These include changes in solvation as charged groups interact, and the ability of these groups to participate as hydrogen bond donors or acceptors. The effects of changes in charge and salt concentration on the enthalpy and entropy of interaction can be studied using isothermal titration calorimetry Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

94

SALT EFFECTS IN RIBONUCLEASE–LIGAND INTERACTIONS

(ITC) experiments.2 We have studied the effect of charge and added salt on binding energetics in two systems. The first system is the binding of phosphate or sulfate to a complex between a serine protease and an inhibitor protein. The second system is the binding of nucleotides to ribonucleases. In the first system the charge on the anion is varied between 71 and 72 by varying pH or ion type. The second system is used to investigate the effect of salt concentration.

5.2. Anion binding to a protein–protein complex Kazal-type protease inhibitors are small proteins that bind tightly to serine proteases.3 The thermodynamics of binding between porcine pancreatic elastase (PPE), a serine protease, and turkey ovomucoid third domain (OMTKY3), a protease inhibitor, have been studied in detail using ITC.4 The binding enthalpy in these studies showed a dependence on phosphate concentration, suggesting that phosphate binds to the complex between the proteins.4 This assertion was supported by the observation that phosphate binds to the complex between a different serine protease, Streptomyces griseus protease B (SGPB), and OMTKY35 as shown in Figure 5.1. Because the binding of phosphate to the protein complex is weak, it cannot be measured directly. Heats of dilution of the highly concentrated phosphate solution required for titration would be too high to detect binding heats. However, because of the linkage between complex formation and phosphate binding, the binding can be measured. The linkage scheme is illustrated in Figure 5.2.

Figure 5.1 Ribbon diagram of the complex between Streptomyces griseus protease B (light grey to left) and turkey ovomucoid third domain (dark grey to right) showing the binding of phosphate to the interface between the two proteins. Coordinates are from the structure by James and coworkers6

ANION BINDING TO A PROTEIN–PROTEIN COMPLEX

95

Figure 5.2 Scheme for the linkage between protein–protein binding (horizontal) and phosphate binding (vertical) used in determining the energetics of phosphate binding to the protein–protein complex (adapted from reference 16)

The enthalpy of binding OMTKY3 to PPE will have two components: the DH of the protein–protein interaction, and the DH for the binding of phosphate to the complex. The second component will be equal to the fraction of complex which binds phosphate multiplied by the molar enthalpy of phosphate binding. Thus the observed binding enthalpy, DHobs, is given as DHobs ¼ DHpro þ

KPi ½Pi  DHphos 1 þ KPi ½Pi 

ð5:1Þ

where KPi is the equilibrium constant for binding phosphate to the proteaseinhibitor complex, ½Pi  is the phosphate concentration and DHphos is the molar enthalpy of phosphate binding.7 A plot of DHobs versus [Pi] will describe a Langmuir binding isotherm from which KPi and DHphos can be determined. Figure 5.3 illustrates the dependence of DHobs on [Pi] at pH 6 and pH 7 and 258C. The solid lines are the best fit of the data to Equation (5.1). These data

Figure 5.3 Dependence on phosphate concentration of the observed enthalpy of binding OMTKY3 to PPE as determined from ITC experiments at 258C and the indicated pH values. Data are from reference 16

96

SALT EFFECTS IN RIBONUCLEASE–LIGAND INTERACTIONS

Table 5.1 Thermodynamics of binding anions to the complex between PPE and OMTKY3 at 258C (adapted from reference 16) Conditions KH2PO4, pH 7 NaH2PO4, pH 7 NaH2PO4, pH 6 Na2SO4, pH 6

Ka

DG8/kJ mol71

DH8/kJ mol71

TDS8/J mol71

7+3 9+5 10+5 17+7

75+1 75+1 76+1 77+1

18+2 17+4 17+3 18+2

23+2 22+4 23+3 25+2

indicate that phosphate binds quite weakly with an unfavourable enthalpy. More surprisingly, they also indicate no dependence of the binding energetics on pH as shown in Table 5.1. The displacement of one curve from the other is the result of the difference in the DH of the protein–protein interaction at the two pH values.4 As the average charge on the phosphate varies from 71 to 71.7 in going from pH 6 to pH 7, there is no indication that the binding of phosphate to this protein complex is dependent on charge. The lack of a charge dependence was verified by examining the binding of sulfate to the complex at pH 6. Sulfate has a full 72 charge at this pH, in contrast to the 71 charge on the phosphate. The binding energetics of sulfate are also given in Table 5.1. It can be seen that there are no significant differences in either Ka or DH in binding sulfate at pH 6 or phosphate at pH 6 or 7. Finally, an experiment was performed at 100 mM phosphate at pH 6 with 1 M added NaCl. The DH of binding was observed to be the same as for 100 mM phosphate at pH 6 in the absence of added salt, suggesting that salt has no effect on the binding of phosphate to the complex. The results on the binding of phosphate and sulfate to this protein–protein complex are surprising. There is little or no observed dependence of the binding energetics on either charge or salt, even though the interaction is between a charged ligand and the protein. The results suggest that the binding is mediated by hydrogen bonding and desolvation, rather than by charge– charge interactions per se.

5.3. Charge–charge interactions in ribonuclease binding Although the energetics of binding of anions to the complex between OMTKY3 and PPE does not show a dependence on either charge or salt concentration, this is not necessarily the case for other protein–ligand interactions. Indeed, the binding of anionic nucleotide ligands to ribonucleases is known to be salt dependent.8 This salt dependence is illustrated in Figure 5.4. Here the binding of 3’guanosine monophosphate (3’-GMP) to ribonuclease Sa (RNase Sa) is studied

CHARGE–CHARGE INTERACTIONS IN RIBONUCLEASE BINDING

97

Figure 5.4 Results of ITC experiment of the binding of 3’-GMP to RNase Sa at pH 5.5 and 258C. Solutions were dialysed overnight against 15 mM acetate buffer. The results indicate that the DH8 of binding is less exothermic and the affinity is weaker in 100 mM added NaCl

by ITC at pH 5.5 and 258C. The figure shows the integrated heat from each injection as a function of the number of injections. The heats in the absence of added salt are shown in light grey triangles, and those in the presence of 100 mM added NaCl are shown in dark grey squares. The initial heats in the 100 mM NaCl experiment have a decreased magnitude, indicating that the DH of binding is decreased. The curvature is also seen to be less steep, indicating that the binding constant has decreased. Does this salt dependence indicate that charge–charge interactions are more important in this system? To answer this question, we must determine the source of the salt dependence of the binding energetics. In general, the salt dependence of the energetics of binding of nucleotides to ribonucleases can result from either a salt-dependent destabilization of the complex, or a saltdependent stabilization of the free protein and ligand. In order to test the former possibility, Poisson–Boltzmann calculations were performed using the UHBD program suite.9 The electrostatic potential of the complex in the presence and absence of 100 mM NaCl was determined. The calculations indicate that the phosphate interacts with an area of negative potential, which is reduced with added salt. Thus the expectation would be that the complex becomes more, not less, stable with increasing salt concentration. As the calculations are in conflict with the complex being destabilized by salt, it appears that the free protein and/or ligand are stabilized by the addition of salt. This stabilization can arise in two ways: (1) it can be the result of screening between charges, and (2) it can be the result of a direct interaction between salt ions and either the protein or the ligand, i.e. competitive binding. The former possibility can be examined using Poisson–Boltzmann calculations while the latter can be examined from several approaches.

98

SALT EFFECTS IN RIBONUCLEASE–LIGAND INTERACTIONS

The hypothesis that the salt dependence of ribonuclease–nucleotide interactions arises from competitive binding can be tested in a variety of ways: structural methods, thermodynamic measurements, stability studies and comparison with computation. Each of these will be examined.

Structural studies The crystal structure of the ribonuclease, RNase A, has been determined at 2 M NaCl by Almo and co-workers.10 This structure shows a chloride ion in the ligand binding site, which would preclude binding of a nucleotide. In order to determine whether Cl7 also interacts with the binding site of RNase Sa, we have determined the HSQC NMR spectrum, which shows cross peaks between amino nitrogens and the protons to which they are bonded, in the presence and absence of high salt. The HSQC spectrum is shown in Figure 5.5. The addition of NaCl results in perturbations of chemical shifts of some of the peaks, either in the 15N or 1H shift, or both. Figure 5.6 shows the changes in chemical shifts, normalized to the largest shift, as a function of residue number. Asterisks mark those residues of RNase Sa that are seen in the crystal structure to interact with the phosphate of 3’-GMP.11 It can be seen that the largest changes in chemical shift upon the addition of salt are for those residues that interact with the

Figure 5.5 HSQC spectrum of 15N-labelled RNase Sa. Data were collected in a 500 MHz Varian NMR at the University of Iowa College of Medicine NMR Facility

CHARGE–CHARGE INTERACTIONS IN RIBONUCLEASE BINDING

99

Figure 5.6 Fractional change in the chemical shift of residues upon addition of 100 mM NaCl. The change in chemical shift, relative to the spectrum with no added salt, was normalized to the maximum change. Positions marked with an asterisk are those residues that are seen to interact with the phosphate of 3’-GMP in the crystal structure

phosphate of 3’-GMP. This suggests that Cl7 binds to the site on RNase Sa where the phosphate of the ligand binds and therefore would compete with ligand binding.

Binding thermodynamics The thermodynamics of ligand binding will behave in defined ways if competitive binding is occurring. The binding constant is observed to decrease with increasing salt concentration. The observed binding constant, Kobs (assuming the competitive binding of Cl7), is given as Kobs ¼ Kint

1 1 þ KCl ½Cl


ð5:2Þ

where Kint is the binding constant in the absence of Cl7, KCl is the binding constant for Cl7 and [Cl7] is the molar concentration of chloride ion. Equation (5.2) indicates that the competitive binding of Cl7 to the ligand binding site will result in a decrease in the observed binding constant with added salt. Furthermore, the observed binding constant will continue to decrease with added salt, even after the binding site is completely saturated with Cl7. The decrease in the binding with salt is not saturable. Competitive binding also has a well defined effect on the observed enthalpy of binding. The observed binding enthalpy, DHobs, is given as DHobs ¼ DHint
DHCl

KCl ½Cl
 1 þ KCl ½Cl


ð5:3Þ

where DHint is the binding enthalpy in the absence of added salt, and DHCl is the enthalpy of chloride binding.

100

SALT EFFECTS IN RIBONUCLEASE–LIGAND INTERACTIONS

The observed DH will either increase or decrease with the addition of salt, depending on the sign of the chloride binding enthalpy. In contrast to the effect on Kobs, the effect of salt on DHobs is saturable and should show a constant value at salt concentrations high enough to saturate the binding sites. The effect of salt concentration on the binding constant is shown in Figure 5.7. It can be seen that the decrease in the binding constant does not saturate up to a salt concentration of 500 mM. It is difficult to determine the binding constant at higher salt concentrations because Kobs becomes quite weak and very high protein concentrations would be required for an ITC experiment. It is not possible to ascertain whether DHobs saturates with increasing salt because of the inability to measure binding at high salt concentrations. However, the competitive binding model indicates that the salt dependence of both Kobs and DHobs will depend on chloride concentration through KCl. Consequently, if the competitive binding model is correct, fits of the salt dependence of both Kobs and DHobs should yield the same value for KCl. This can be tested by globally fitting the ITC data taken at different salt concentration to the competitive binding model and determining the quality of the fit. The results of a global fit of the data are shown in Figure 5.8. The results of the global fit in Figure 5.8 indicate that both DHobs and Kobs exhibit a salt dependence that is consistent the same value of KCl. The fit yields a value of KCl of 14 with a DHCl of 79 kJ mol71. The results of the thermodynamics of binding are consistent with the competitive binding model.

Stability studies In the absence of binding to the denatured state of the protein, binding of ligand to the native state stabilizes the protein against unfolding.12 Consequently, if the salt dependence of the binding results from the competitive binding of chloride ions to the protein, then there should be a salt dependence to the stability as well. This salt dependence will have the form Kunf ¼ Kint

1 1 þ KCl ½Cl


ð5:4Þ

where Kunf is the unfolding constant (i.e. the ratio of denatured to native protein) and Kint is the unfolding constant in the absence of added Cl7. As with the observed binding constant, the effect of salt on the stability constant does not saturate at high salt concentrations. The stability of RNase Sa was studied as a function of salt concentration using differential scanning calorimetry (DSC). In a DSC experiment the excess

CHARGE–CHARGE INTERACTIONS IN RIBONUCLEASE BINDING

101

Figure 5.7 The effect of added salt on the observed binding constant for binding 3’-GMP to RNase Sa at pH 5.5 and 258C. The binding constants were determined from fits of ITC experiments. It can be seen that the observed binding constant continues to decrease even at 0.5 M added NaCl

heat capacity of the protein solution is measured as a function of temperature. The maximum in the heat capacity curve corresponds roughly to the melting temperature, Tm, where there are equal populations of native and denatured protein. The area under the curve yields the unfolding enthalpy, DHm, so that the unfolding entropy can be determined as the ratio of DHm and the melting temperature (in Kelvin). The results of a series of DSC experiments at pH 5.5 and various salt concentrations are shown in Figure 5.9. The DSC scans show that the stability of RNase Sa increases with increasing salt from 0 to 1 M NaCl. It is also clear that the stabilization does

Figure 5.8 Global fits of ITC data for the binding of 3’-GMP to RNase Sa at different salt concentrations. The solid lines represent the best fit of the data to the competitive binding model as described in the text

102

SALT EFFECTS IN RIBONUCLEASE–LIGAND INTERACTIONS

Figure 5.9 Differential scanning calorimetry studies of the thermal stability of RNase Sa at 258C. Curves show the effect of added NaCl on the stability with the concentration of added NaCl varying from zero (leftmost curve) to 1 M (rightmost curve). The melting temperature continues to increase with increasing salt concentration

not saturate at high salt, in agreement with the prediction from the competitive binding model. While the DSC data are consistent with the binding of one or more chloride ions to the native RNase Sa, it is also possible that the increasing stability with added salt is due to a screening of unfavourable electrostatic interactions on the surface of the protein. In order to examine this possibility, we have performed Tanford–Kirkwood calculations13 using the programs developed by Makhatadze and coworkers.14,15 The electrostatic contribution to the free energy of folding is shown in Figure 5.10 as a function of pH and salt concentration. At pH 5.5 it can be seen that the electrostatic free energy is stabilizing, so the addition of salt would be expected to decrease the stability of the protein, not increase it. These results support the conclusion that chloride binds to the native state of the protein and that competitive binding is the origin of the salt dependence of nucleotide binding.

Computational studies Finally, we have performed Poisson–Boltzmann (PB) calculations using the UHBD program9 to calculate the expected salt dependence of the DG8 of binding 3’-GMP to RNase Sa. As there are many factors besides electrostatics that contribute to the affinity, it is not expected that the PB calculations will predict the absolute DG8 accurately; however, the salt dependence of the electrostatic contribution to DG8 should be well predicted. The calculated and experimental values of DG8obs as a function of salt concentration are shown in Figure 5.11. The salt dependence of DG8 is given as the slope. The experimental slope is about three times greater than that predicted by the PB calculations. This result indicates that other factors, in addition to screening of charge–charge interactions, give rise to the observed

CONCLUSIONS

103

Figure 5.10 Contribution of surface electrostatic interactions to the DG8 of folding of RNase Sa as calculated using Tanford–Kirkwood theory. The electrostatics contributes favourably between about pH 3.5 and pH 6 so that the addition of salt would be expected to decrease stability at pH 5.5

Figure 5.11 Comparison of the salt dependence of the DG8 of binding 3-GMP to RNase Sa as determined experimentally (lower curve) and as determined from Poisson– Boltzmann calculations (upper curve). The salt dependence (i.e. slope) is much greater experimentally than is predicted from the calculations

salt dependence. The competitive binding of chloride is likely to be the source of this additional salt dependence. Numerous lines of evidence, including structural data, binding thermodynamics, stability data and computational results, support the idea that most, if not all, of the salt dependence of nucleotide binding to ribonucleases arises from competitive binding of anions and not from screening of electrostatic interactions. Thus a salt dependence to binding, in and of itself, cannot be used as evidence of significant contributions from charge–charge interactions.

5.4. Conclusions Two of the hallmarks of electrostatic interactions are a dependence on charge and a dependence on salt concentration. We have investigated two systems in

104

SALT EFFECTS IN RIBONUCLEASE–LIGAND INTERACTIONS

which charged ligands interact with proteins. In the first system, the binding of phosphate (or sulfate) to a protein–protein complex, we see no evidence of the energetics of binding depending on either charge or salt concentration.16 This indicates that the presence of a charge on a ligand does not necessitate that the charge itself makes a critical contribution to the binding. Rather, the ability of phosphate and sulfate to act as hydrogen bond acceptors is probably the basis for their interaction with this complex. In the second system, the binding of nucleotides to ribonucleases, a salt dependence of the binding energetics is observed. However, several lines of evidence suggest that the origin of the salt dependence is not screening of electrostatic interactions, but the competitive binding of anions to the ligand binding site. Thus, the presence of a salt dependence of the binding energetics does not, in itself, indicate that charge–charge interactions are important to binding. While more work is clearly needed to understand the contribution of charges in molecular recognition in proteins, the present work suggests that the presence of charged groups in a protein–ligand interface does not necessarily mean that charges are necessary to the interaction. It is possible that these charged groups could be replaced with appropriate hydrogen bond acceptors or donors and still have high affinity and specificity. This could be advantageous to drug design as polar, but uncharged, compounds are better able to cross biological membranes than anionic compounds, thus improving bioavailability.

Acknowledgement Financial support for this work was provided in part by a grant from the National Science Foundation and from the Department of Biochemistry at the University of Iowa.

References 1. 2. 3. 4. 5.

Myers JK and Pace CN. (1996) Biophys. J. 71: 2033–2039. Wiseman T, Williston S, Brandts JF and Lin L-N. (1989) Anal. Biochem. 179: 131–137. Laskowski MJ and Ikunoshin K. (1980) Annu. Rev. Biochem. 49: 593–626. Baker BM and Murphy KP. (1997) J. Mol. Biol. 268: 557–569. Read RJ, Fujinaga M, Sielecki AR and James MNG. (1983) Biochemistry 22: 4420– 4433. 6. Huang K, Lu W, Anderson S, Laskowski, M Jr. and James MNG. (1995) Protein Sci. 4: 1985–1997. 7. Edgcomb, SP, Baker BM and Murphy KP. (2000) Protein Sci. 9: 927–933. 8. Flogel M, Albert A and Biltonen R. (1975) Biochemistry 14: 2616–2621.

REFERENCES

105

9. Davis ME, Madura JD, Luty BA and McCammon JA. (1991) Comp. Phys. Comm. 62: 187–197. 10. Fedorov AA, Diane J-M, Fedorov E, Sirakova D, Graf I and Almo SC. (1996) Biochemistry 35: 15 962–15 979. 11. Sevcik J, Dodson EJ and Dodson GG. (1991) Acta Crystallogr. B 47: 240–253. 12. Schellman JA. (1975) Biopolymers 14: 999–1018. 13. Tanford C and Kirkwood JG. (1957) J. Am. Chem. Soc. 79: 5333–5339. 14. Loladze VV, Ibarra-Molero B, Sanchez-Ruiz JM and Makhatadze GI. (1999) Biochemistry 38: 16 419–16 423. 15. Sanchez-Ruiz JM and Makhatadze GI. (2001) Trends Biotech. 19: 132–135. 16. Waldron TT, Modestou MA and Murphy KP. (2003) Protein Sci. 12: 871–874.

6 Thermodynamics–Structure Correlations of Sulfonamide Inhibitor Binding to Carbonic Anhydrase Daumantas Matulis and Matthew Todd 6.1 Introduction Microcalorimetry is an excellent method of determining the thermodynamics of ligand–protein binding in solution, because the signal being measured is the observed binding enthalpy (Db-obs H [Editors’ note: this term can also be written DHobs or Db Hobs as seen elsewhere in this book]), which is free of any model assumptions. The affinity (Db G) can be computed from the shape of a binding curve, and the binding entropy (Db S) is obtained from the difference between the two. The calorimetrically observed binding enthalpy, however, can be the sum of several partial reactions, including changes in the protonation of ligand, protein, and buffer, each of which may have its own contribution to the observed enthalpy of binding. Which enthalpy is the ‘true’ enthalpy of a reaction? In order to obtain an enthalpy of binding that can be correlated with structure, one must elucidate the thermodynamics for all reactions that occur simultaneously upon ligand binding. It is fairly straightforward to distinguish net protonation effects by measuring Db-obs H in buffers with varying ionization enthalpies,1 and extrapolating the value to zero Dbuffer-protonation H. This process simultaneously gives the net protonation effect for all remaining linked reactions. However, a thorough deconvolution of the energetics of the remaining linked reactions remains complex. Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

108

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

Carbonic anhydrase is a clinically relevant and biochemically well characterized protein.2 It catalyses hydration–dehydration of carbon dioxide to carbonic acid3,4 and is involved in vital physiological processes such as pH and CO2 homeostasis, transport of bicarbonate and CO2, biosynthetic reactions, bone resorption, calcification, tumorigenicity, and many other physiological or pathological processes.5 Therefore, the enzyme is an important target for inhibitors with clinical applications:6 inhibitors are primarily used as antiglaucoma agents but also for the therapy of various pathologies such as epilepsy and Parkinson’s disease.7 Carbonic anhydrase has a single active site containing a catalytically active zinc coordinated by three histidine residues and a hydroxide ion (or water molecule) in the resting state. Thermodynamics of metal ion binding to apocarbonic anhydrase has been measured.8–11 The catalytic mechanism of the enzyme and the mechanism of inhibition has been studied extensively.12–18 Carbonic anhydrase has been used as a model protein to study molten globules and their refolding (19, 20 and references therein). However, there have been few thermodynamic measurements of sulfonamide inhibitor binding to carbonic anhydrases.21,22

6.2 Identification of protonation reactions occurring upon binding The secondary structure of human carbonic anhydrase I is shown in Figure 6.1(a). Beta sheets form a core of carbonic anhydrase, and there are several

IDENTIFICATION OF PROTONATION REACTIONS

109

Figure 6.1 (a) A ribbon view of carbonic anhydrase I showing the secondary structure of the protein with acetazolamide (ball–stick model) bound to the active site zinc atom (large sphere). (b) An electrostatic surface of carbonic anhydrase I drawn using identical orientation and scale as in (a). (c) Chemical structures and abbreviations used in the text of the seven carbonic anhydrase inhibitors reviewed

110

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

alpha helices at the edges of the enzyme. Coordinates are from Protein Data Bank 1AZM.pdb. The surface structure is shown in Figure 6.1(b). The large sphere of the zinc atom is positioned deep in the active site cleft. Acetazolamide inhibitor is shown bound in the cleft as a ball–stick model. Crystallographic studies of carbonic anhydrases have recently been reviewed.23 Chemical structures of the several carbonic anhydrase inhibitors are shown in Figure 1(c). Every inhibitor bears a sulfonamide group, which binds to the zinc atom of the enzyme active site ((a), (b)),24–26 and the hydrophobic group, which affects the pKa of the sulfonamide group and interacts with the hydrophobic pocket of the enzyme. Only the ionized (deprotonated, anionic RNH
) sulfonamide form is thought to bind to the carbonic anhydrase;22,27 however, most sulfonamide pKa values are above seven. Thus, upon binding of most of these inhibitors to carbonic anhydrase, a pH-dependent deprotonation reaction must occur. Likewise, the active site Zn-coordinated hydroxide must protonate prior to being replaced by the amino group of the sulfonamide. Other linked protonation– deprotonation reactions associated with protein side-chains may also occur upon inhibitor binding.22,27 Such reactions may be summarized as follows (R – inhibitor chemical groups other than sulfonamide, CA – carbonic anhydrase enzyme): RSO2 NH
þ CA--Zn--H2 O $ CA--Zn--NH
SO2 R þ H2 O þ

CAH $ CA þ H


þ

ðR1Þ ðR2Þ

þ

CA--Zn--OH þ H $ CA--Zn--H2 O


ðR3Þ þ

RSO2 NH2 $ RSO2 NH þ H buffer þ Hþ $ bufferHþ

RSO2 NH2 þ CA--Zn--OH
þ buffer þ CAHþ $ CA--Zn--NH
SO2 R þ bufferHþ þ CA þ H2 O

ðR4Þ ðR5Þ

ðR6Þ

(R1) is a binding reaction, free of protonation–deprotonation contributions (a displacement of water molecule by a deprotonated inhibitor molecule), (R2) is a linked deprotonation or protonation of any ionizable group of the enzyme, (R3) is the protonation of the hydroxide ion coordinated to the active site zinc atom, (R4) is the inhibitor deprotonation, (R5) is buffer protonation, and (R6) is the sum of all the above events. Observed thermodynamic parameters that are measured by an isothermal titration calorimetry (ITC) experiment represent the sum of the linked events (R6). This review describes experiments designed to dissect individual reactions’ contributions to the thermodynamic parameters of binding and to correlate them with the structural features of protein–ligand interaction surface.

OBSERVED THERMODYNAMICS OF INHIBITOR BINDING TO CA

111

6.3 Observed thermodynamics of inhibitor binding to CA Figure 6.2 shows typical raw and integrated ITC data for inhibitor binding to carbonic anhydrase. Integrated ITC curves of five inhibitors binding to carbonic anhydrase isozyme II from bovine erythrocytes (b-CAII) are shown in Figure 6.3. Both the cell and the syringe contained 25 mM phosphate buffer, pH 7.0, 50 mM NaCl, and 2 per cent DMSO. Datapoints are experimental ITC data and lines are fitted curves. Inhibitors were & – ACTAZ, ^ – METHZ, * – TFMSA, X – DCHPA, and ~ – SULFA. Both the observed

Figure 6.2 ITC curve (raw data, upper panel) of carbonic anhydrase I (h-CAI) (initial concentration in the cell 3.0 mM) titrated with TFMSA (concentration in the syringe 40 mM, injection size 12.5 ml, injected at 3 minute intervals) at 258C, where both the cell and syringe contained 20 mM Na phosphate buffer of pH 7.5, 100 mM NaCl, and 0.4 per cent dimethylsulfoxide (DMSO). The lower panel shows the integrated ITC curve of the same data obtained after adding 1.2 kcal mol1 for the heat of dilution

112

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

Figure 6.3 Integrated ITC curves of five inhibitors binding to bovine carbonic anhydrase II at 258C

enthalpies (initial plateaus of the curves) and the observed binding constants varied significantly depending on the inhibitor. The observed binding constants of the seven inhibitors span a large range, from 103 to 108 M
1 (Table 6.1) and are specific for each isozyme. Figure 6.4 shows integrated ITC curves of acetolazamide binding to three carbonic anhydrase isozymes. Both the cell and the syringe contained 20 mM phosphate buffer, pH 7.0, 100 mM NaCl, and 1 per cent DMSO. Datapoints are experimental ITC data and lines are curves regressed against the data. Proteins were & – h-CAI, * – b-CAII, and ~ – h-CAII. The observed binding constants varied significantly depending on the protein – h-CAI bound ACTAZ significantly weaker than any CAII, but the observed enthalpies (plateaus of the curves) were quite similar, with less than 1 kcal mol
1 difference for all three proteins. Because there is a limit of binding constants that can be determined by ITC (about 108–109 m71 depending on protein concentration), the binding constants of six inhibitors were also measured by a high-throughput thermal shift assay (ThermoFluor1) that measures the increase in protein melting temperature in the presence of the inhibitor and has no upper limit to measured affinity.28–30 A good agreement between the ITC and ThermoFluor1 data confirmed the accuracy of binding constant determination by ITC, as shown in Figure 6.5. A solid line shows the trend of exact match between the two methods. Open small symbols show binding constants for h-CAI, closed symbols b-CAII, and open large symbols h-CAII. Phosphate buffer of pH 7.0

107  10 79  20 35  10 155  20 52  20 50  40 NA

Carbonic anhydrase II (bovine, b-CAII ) 10:43  0:07 11:5  1:2 1:0  1:2 10:13  0:07 10:0  1:6 0:1  1:6 10:80  0:02 9:3  1:0 1:5  1:0 11:34  0:23 7:1  1:1 4:2  1:1 7:02  0:18 5:0  3:0 2:0  3:0 6:22  0:19 3:0  2:8 3:2  2:8 8:22  0:13 10:8  1:5 2:6  1:3

2:23
107  2:63
106 1:37
107  1:53
106 4:10
107  1:41
106 9:83
107  4:55
107 8:90
104  2:97
104 2:40
104  8:49
103 1:06
106  2:4
105

8:23
107 2:33
107 4:81
107 4:81
107

ACTAZ METHZ TFMSA DCHPA SULFA PAMBS CARBSg

ACTAZ METHZ TFMSA DCHPA

Carbonic anhydrase II (human, h-CAII ) 12:5 1:3 11:3 0:9 12:6 1:7 8:0 2:9

0:23 0:22 0:23 0:40

0:29 0:30 0:17 0:58 NA NA NA

0:30 0:61 0:29 0:29 NA NA

nR1R4 f, mol mol1

a Protein–sulfonamide observed binding constant at T0 ¼ 378C, obtained from ITC experiments. bObserved Gibbs free energy of protein–inhibitor binding at T0 ¼ 378C, Db Gobs,T0 ¼
RT lnðKb-obs Þ. cObserved enthalpy of protein–inhibitor binding at T0 ¼ 378C, obtained from ITC experiments and extrapolated to zero enthalpy of buffer ionization. dObserved entropy of protein–inhibitor binding at T0 ¼ 378C (T0 Db Sobs,T0 ¼ Db Hobs,T0
Db Gobs,T0 ). eObserved heat capacity of protein–inhibitor binding, assumed to be constant in the temperature range of 10–508C, equal to the slope of the observed enthalpy dependence on temperature. f Slopes of the observed enthalpy dependences on the enthalpy of buffer protonation, equal to the net number of linked protonation events. gValues for CARBS binding to b-CAII (258C, phosphate buffer of pH 7.0) were taken from the study by M. Doyle to determine the repeatability of ITC measurements.

11:23 10:45 10:90 10:90

64 73 48 148

58  15 65  20 72  30 130  30 50  35 50  0

2:2  0:3 0:2  1:8 2:0  0:6 0:1  1:6 2:4  4:0 3:3  5:0

10:6  0:3 9:6  1:8 12:6  0:6 8:3  1:6 3:0  4:0 1:0  5:0

Carbonic anhydrase I (humans, h-CAI )

Db Cp,obs e, cal mol1 K1

8:42  0:13 9:77  0:21 10:47  0:10 8:38  0:25 5:42  0:42 4:26  0:42

T0 Db Sobs,T0 d, kcal mol1

8:63
105  1:97
105 7:61
106  3:13
106 2:40
107  4:27
106 8:07
105  4:06
105 6:61
103  6:43
103 1:00
103  9:70
102

Db Hobs,T0 c, kcal mol1

ACTAZ METHZ TFMSA DCHPA SULFA PAMBS

Db Gobs,T0 b, kcal mol1

Kb-obsT0 a, M1

Inhibitor

Table 6.1 Observed thermodynamic parameters of inhibitor binding to carbonic anhydrases h-CAI, b-CAII, and h-CAII at 378C representing reactions (R1)–(R4). The enthalpies are extrapolated to zero buffer protonation enthalpy; binding constants were independent of buffer within the error of the experiment OBSERVED THERMODYNAMICS OF INHIBITOR BINDING TO CA

113

114

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

Figure 6.4 Integrated ITC curves of acetazolamide binding to three carbonic anhydrase isozymes at 258C

Figure 6.5 Correlation of the binding constants obtained by ITC and ThermoFluor1

OBSERVED THERMODYNAMICS OF INHIBITOR BINDING TO CA

115

Figure 6.6 (a) Dependence of observed molar enthalpies of inhibitor binding to h-CAI on temperature (in phosphate buffer, pH 7.0). (b) Temperature dependence of the molar enthalpy of inhibitor binding to CAII (bovine – solid lines, human – dashed lines) as determined by ITC

116

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

was used in all cases. Approximate values for ligand binding enthalpy and heat capacity (Db HT0 ¼378C ¼
5:0 kcal mol
1 , Db Cp ¼
300 cal mol
1 K
1 ) were used to extrapolate estimated Kb values by ThermoFluor1 (accurate at protein melting temperature) to 378C. Data for h-CAII are from single determinations and have no error bars. When an ITC experiment is carried out at several temperatures the heat capacity of the reaction can be obtained from the enthalpy dependence on temperature. Figure 6.6(a) shows the dependence of the observed Db-obs H on temperature for isozyme I (Table 6.1). Datapoints are enthalpies of binding as derived from ITC experiments and solid lines are linear fits of the datapoints. Slopes of the linear fits are equal to the observed heat capacity of inhibitor binding to the protein. The heat capacity is assumed to be constant and temperature independent in the range of interest. Inhibitors were: * – DCHPA, ~ – METHZ, & – ACTAZ, ~ – SULFA, and * – TFMSA. Figure 6.6(b) shows the dependence for isozyme II. Datapoints are enthalpies derived from ITC experiment and solid lines are linear fits of the datapoints. Buffers were PIPES for b-CAII buffer (except TFMSA in HEPES), phosphate for h-CAII. For DCHPA, inhibitor data is shown for h-CAII for both phosphate (open diamonds) and Tris (filled diamonds with dotted line). All buffers were of pH 7.0. Inhibitors were X – DCHPA, ^ – METHZ, & – ACTAZ, ~ – SULFA, * – TFMSA, and + – PAMBS. This Db-obs Cp , however, is also an ‘observed’ value that may be affected by linked protonation events. Thus, attempts to correlate these values to structure and the buried hydrophilic and hydrophobic surface area become dependent on identifying these values. There are often linked protonation–deprotonation reactions accompanying binding. In order to determine whether such protonation reactions are linked to the binding reaction, the ITC experiment can be carried out in several buffers of different protonation enthalpies. If a net change in protonation occurs, a corresponding change in buffer protonation will add to the observed enthalpy. We have carried out the binding experiments in the following buffers at pH 7.0 (enthalpy of deprotonation at 378C is in brackets): phosphate (0.70 kcal mol
1 ), PIPES (2.79 kcal mol
1 ), MES (3.76 kcal mol
1 ), HEPES (5.16 kcal mol
1 ), and Tris (11.13 kcal mol
1 ). The observed binding constants were found to be buffer independent (within the error of the experiment). However, Db-obs H varied greatly depending on the buffer, protein, and the inhibitor. Figure 6.7(a) shows Db-obs H for isozyme I as a function of Dbuffer
deprotonation H. Datapoints are experimental ITC data and solid lines are linear fits of the datapoints. Inhibitors were X – DCHPA, ^ – METHZ, & – ACTAZ, and * – TFMSA. Buffers (25 mM) and the enthalpies of their deprotonation at 378C were phosphate 0.70 kcal mol
1 , PIPES 2.79 kcal mol
1 , MES 3.76 kcal mol
1 , HEPES 5.16 kcal mol
1 , and Tris 11.13 kcal mol
1 . Figure 6.7(b) shows Db-obs H for isozyme II. Datapoints are

OBSERVED THERMODYNAMICS OF INHIBITOR BINDING TO CA

117

Figure 6.7 (a) Dependence of molar enthalpies of inhibitor binding to h-CAI on the enthalpy of buffer deprotonation as determined by ITC at 378C and pH 7.0. (b) Molar enthalpy of inhibitor binding to CAII depends on the enthalpy of buffer deprotonation

118

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

enthalpy obtained from ITC experiment and solid lines are linear fits. Solid lines with filled symbols are for b-CAII at 258C; dashed lines with unfilled symbols are for h-CAII at 258C. Inhibitors were X – DCHPA, ^ – METHZ, & – ACTAZ, and * – TFMSA. Only TFMSA binding was coupled with the proton binding; others were linked to proton release. Buffers (25 mM) and the enthalpies of their deprotonation at 258C were phosphate 1.22 kcal mol
1 , PIPES 2.74 kcal mol
1 , MES 3.71 kcal mol
1 , HEPES 5.02 kcal mol
1 , and Tris 11.34 kcal mol
1 . A zero slope in Figure 6.7 indicates no net protonation change on binding, while a non-zero slope indicates a linked protonation reaction. A negative slope indicates a net release of protons upon ligand binding; a positive slope indicates a net uptake of protons.1 Data in Figure 6.7 gives two important parameters: (1) the slope gives net change in protonation (Table 6.1); (2) the intercept at zero enthalpy of buffer protonation gives buffer-independent binding enthalpies from a collection of Db-obs H values. Thus, if Db-obs H represents the sum of (R1)–(R5), (R6), then the intercept gives Db H for the sum of (R1)–(R4). Table 6.1 summarizes observed buffer-independent thermodynamic parameters of seven inhibitors binding to three CA isozymes. When comparing the inhibitors, the observed binding constants spanned a wide range of strengths. However, the underlying mechanism of the large differences between inhibitors is not immediately clear without detailed analysis of all underlying reactions. Carbonic anhydrase I bound sulfonamide inhibitors consistently more weakly than both isozymes of carbonic anhydrase II. Converted to binding Gibbs free energy, Db G, inhibitor affinity varied by 5 kcal mol
1 , whereas Db H varied by more than 8 kcal mol
1 . When similar experiments were performed with other isozymes (Figure 6.4), an equally large difference in Db G and Db H was observed (Table 6.1) for a given protein. The buffer-independent binding enthalpies ((R1)–(R4), Table 6.1) of the seven inhibitors spanned the range from
1 to
12:6 kcal mol
1 . Typically the carbonic anhydrase, isozyme I, from human erythrocytes (h-CAI) Db G was 1– 3 kcal mol
1 less favourable than for the other enzymes. The only deviation from this ranking is trifluoromethanesulfonamide (TFMSA), which bound equally tightly, but with more favourable enthalpy, to h-CAI. Differences between the three proteins were small with the largest difference being for TFMSA:
12:6 kcal mol
1 for carbonic anhydrase isozyme II from human erythrocytes (h-CAII) and
9:3 kcal mol
1 for b-CAII. The entropy contributions were smaller than the enthalpies. Entropies of most inhibitors binding were usually positive, especially for b-CAII. All inhibitors bound with a negative change in heat capacity. The values spanned the range of
70  35 cal (mol K)
1 with the exception of DCHPA binding, which gave an approximately double value. DCHPA is the only ligand with two sulfonamide groups and two chlorine atoms. Since Db Cp

ENERGETICS OF INHIBITOR PROTONATION

119

scales with buried hydrophobic surface areas, the greater value possibly arises from the bulkiness of the inhibitor. The numbers of linked protonation–deprotonation events obtained from the slopes in Figure 6.7 varied from
0:61 to þ0:29. DCHPA, ACTAZ, and METHZ binding was linked to a net release of protons, while the binding of TFMSA was linked to a net uptake of protons. Linked protonation numbers were nearly identical for the two isozymes of CAII.

6.4 Energetics of inhibitor protonation The sulfonamide group of all studied inhibitors may be electrically neutral (protonated RNH2) or negatively charged (deprotonated RNH
) as represented in (R4) above. The predominant ionization form of the inhibitor depends on the pH of solution and its pKa . Most sulfonamide inhibitors that do not have strong electron withdrawing groups have pKa values above 7 and thus are usually protonated at physiological pH, thus binding affinity and enthalpy will be pH dependent. However, only the negatively charged (deprotonated) sulfonamide group is thought to bind to the zinc atom in the active site of carbonic anhydrase.22 Therefore, inhibitors that exist in a protonated form at pH 7 must undergo a linked deprotonation reaction upon binding to the protein. In order to understand the effects of inhibitor deprotonation on binding, the energetics of inhibitor protonation were evaluated independently. All seven inhibitors were titrated calorimetrically and potentiometrically with NaOH and HCl to determine the thermodynamics of protonation. Enthalpies and heat capacities were determined by titrating the deprotonated (alkaline) sulfonamides with HCl in the calorimeter, and the Gibbs free energies (pKa values) were determined by potentiometric titration with the pH-meter. The pKa values could not be determined by ITC because the binding constants are out of the range where ITC experiments can determine them accurately. It was chosen to titrate deprotonated (alkaline) inhibitors with HCl because the reaction of protonation is exothermic and could be determined with greater precision than the reaction of deprotonation with NaOH. Calorimetric titration with NaOH is complicated by a non-trivial behaviour of NaOH dilution. To obtain negatively charged inhibitors, 1.5 molar equivalents of NaOH were added. Titration curves of inhibitors with HCl exhibit two transitions: (1) an event with stoichiometry 0.5 and Dobs H 
13 kcal mol
1 , occurring due to neutralization of excess NaOH; (2) an event with stoichiometry 1.0 and inhibitor-dependent enthalpy, occurring due to protonation of the inhibitor sulfonamide group. Figure 6.8(a) shows a titration curve of METHZ with HCl. Concentrations and volumes were identical to pH titration experiments

120

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

shown in (c). The first portion of the titration curve represents the reaction between Hþ and excess OH
. The second portion of the titration curve with the stoichiometry of 1.0 represents inhibitor protonation. Figure 6.8(b) shows titration curves of TFMSA and METHZ with HCl. The enthalpy of the first portion of the titration (stoichiometry  0:5, DH 
13 kcal mol
1 ) is a value close to that expected for the reaction of Hþ þ OH
! H2 O. The enthalpy of the second portion of the titration curve (stoichiometry  1:0) represents the enthalpy of sulfonamide protonation. Heat capacities of sulfonamide protonation were determined by carrying out the titration at several temperatures. The slope of the enthalpy dependence on temperature is equal to the heat capacity of inhibitor protonation. The exact same solutions and volumes as in the calorimetric experiments were used for potentiometric titration by using a pH-meter. The injection syringe of the VP-ITC calorimeter was used to deliver injections of several microlitres. Such methodology gave high precision titration curves using small volumes of solution. The stoichiometry obtained from potentiometric titration curves precisely matched that obtained from the calorimetric curves. A function describing two transitions was derived according to reference 31 and curves were simulated to match the experimental datapoints. Several representative titration curves are shown in Figure 6.8(c). Concentrations and volumes are identical to those in ITC experiments (panels (a) and (b)). Two transitions are seen: 0.5 equivalent of Hþ reaction with OH
, and one

ENERGETICS OF INHIBITOR PROTONATION

121

Figure 6.8 (a) Raw ITC data of METHZ titration with HCl at 258C is representative of inhibitor protonation enthalpy determinations by ITC, where 1.5 equivalents of NaOH were added to the neutral inhibitor. (b) Examples of integrated ITC curves of TFMSA (*, 138C) and METHZ (^, 378C) titration with HCl. (c) pH titration curves of SULFA (~), METHZ (^), and TFMSA (*) at 248C

equivalent of sulfonamide protonation. The pKa is approximately equal to the pH at the midpoint of the second stage of the titration. The midpoint of the second transition closely matches the pKa of the inhibitor. A summary of all seven inhibitor protonation energetics is shown in Table 6.2. Sulfanilamide (SULFA) has the highest pKa of all studied inhibitors. Therefore, the fraction of deprotonated sulfonamide at pH 7 is quite small (0.0019) and results in weak affinity (last column in Table 6.2). On the other hand, TFMSA has a pKa of 6.02. Therefore, at pH 7 almost all of the inhibitor exists in a negatively charged form (fraction ¼ 0:91). The enthalpies of inhibitor protonation were similar (
4:5  0:5 kcal mol
1 ), except for DCHPA, which was
6 kcal mol
1 , perhaps because of two chlorine atoms – electron withdrawing groups – on the phenyl ring. Entropies of protonation were positive and equal to 16  2 cal (mol K)
1 for ACTAZ, METHZ, TFMSA, and DCHPA. The remaining three inhibitors have even greater positive entropy of

7.30 7.12 6.25 8.20 10.10 8.35 9.60

ACTAZ METHZ TFMSA DCHPA SULFA PAMBS CARBS

7.03 6.86 6.02 7.89 9.72 8.04 9.24

pKa,T0 b

Dprot HT0 ,d kcal mol1 4:5  0:3 4:9  0:2 4:0  0:3 6:2  0:4 4:7  2h 4:7  2h 4:1  0:7

Dprot GT0 ,c kcal mol1 9:98  0:2 9:73  0:2 8:55  0:2 11:2  0:2 13:8  0:5 11:4  0:5 13:1  0:2

5:4  0:5 4:8  0:5 4:5  0:5 5:0  0:5 9:1  2:5 6:7  2:5 9:0  0:5

T0 Dprot ST0 ,e kcal mol1 75  15 70  20 86  10 58  30 70  50h 70  50h 58  20

Dprot Cp ,f cal mol1 K1

0.48 0.58 0.91 0.11 0.084 0:084 0.0057

fdeprot,T0 at pH 7.0g

c

b

pKa measured experimentally by pH titration of the inhibitor as described in the ‘Methods’ section. pKa of protonation at 378C calculated from pKa at 258C, enthalpy, and heat capacity. Gibbs free energy of protonation at 378C calculated from pKa at 378C. d The enthalpy of protonation at 378C determined by ITC as described in the ‘Methods’ section. e T0 Dprot ST0 at 378C calculated by subtracting Gibbs free energy from enthalpy. f Heat capacity of protonation, obtained from the slope of linear regression of ITC enthalpy dependence on temperature obtained at 13, 25, and 378C. g Fraction of deprotonated sulfonamide at 378C and at pH 7.0 calculated by Henderson–Hasselbach equation. h These values were not determined; they were assumed to be an average of other studied inhibitors, with large uncertainty.

a

pKa 258C,a 0:15

Thermodynamic parameters of inhibitor sulfonamide group protonation (reaction (R4))

Inhibitor

Table 6.2

122 SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

SULFONAMIDE ‘ANION’ BINDING THERMODYNAMICS

123

protonation. However, the enthalpies for p-aminomethylbenzenesulfonamide (PAMBS) and SULFA were not determined. Enthalpies of methazolamide protonation have also been determined by measuring the temperature dependence of the pKa . The pKa of methazolamide protonation was earlier determined to be 7.26,22 close to our determined value of 7.12 at 258C. Our measured enthalpy of methazolamide protonation of
5:8 kcal mol
1 was close to the value of
7:0 kcal mol
1 at 258C.22 The pKa of acetazolamide has been previously determined to be 7.2,12 and we obtained a value of 7.30.

6.5 Sulfonamide ‘anion’ binding thermodynamics The process of inhibitor protonation (R4) can now be subtracted from the buffer-independent values ((R1)–(R4), Table 6.1), giving the thermodynamic parameters for reactions (R1)–(R3). To obtain the binding constant independent of inhibitor protonation (Kb,T0 ), the observed binding constant (Kb-obs,T0 ) was divided by the fraction of deprotonated sulfonamide ( fdeprot , Table 6.2), present at pH 7.0: Kb,T0 ¼

Kb-obs,T0 fdeprot

ð1Þ

The enthalpies of deprotonated sulfonamide binding to carbonic anhydrases were calculated from: Db HT0 ¼ Db-obs HT0 þ Dprot HT0 ð1
fdeprot Þ

ð2Þ

where Dprot HT0 is the enthalpy of sulfonamide protonation. The observed enthalpies of ligand binding (Db-obs HT0 ) are the ones obtained after extrapolating to zero buffer protonation enthalpy. Similarly, the ligand protonation-independent heat capacity of binding was estimated from: Db Cp,T0 ¼ Db-obs Cp,T0 þ Dprot Cp,T0 ð1
fdeprot Þ

ð3Þ

A summary of the inhibitor binding data after accounting for inhibitor deprotonation (R1)–(R3) is listed in Table 6.3. The data is most precise for the four strongly binding ligands as seen from the standard deviation and uncertainty of the data. Unexpected conclusions were drawn from the Table 6.3. After consideration of fractions of deprotonated inhibitor, the inhibitors bound to each carbonic anhydrase with similar affinities, differing by less than one order of magnitude. Such a large reduction in the difference of observed binding constants was unexpected when considering that the observed binding constants varied by four orders of magnitude (from 7  103 to 2  107 ) for h-CAI and three orders of magnitude (from 9  104 to 1  108 ) for b-CAII

146  18 108  28 43  14 206  36 122  54 114  64 NA

Carbonic anhydrase II (bovine, b-CAII) 10:88  0:21 13:8  1:2 2:9  1:2 10:46  0:21 11:9  1:6 1:5  1:6 10:87  0:20 9:7  1:0 1:1  1:0 12:68  0:31 11:1  1:1 1:5  1:2 10:89  0:27 9:5  3:0 1:4  3:1 7:74  0:27 7:2  2:8 0:6  2:9 11:23 15:6 4:4

4:61
107  1:89
107 2:35
107  9:55
106 4:53
107  1:75
107 8:66
108  5:62
108 4:69
107  2:55
107 2:86
105  1:60
105 1:69
108

1:71
108 3:99
107 5:31
107 4:23
108

ACTAZ METHZ TFMSA DCHPA SULFA PAMBS CARBS

ACTAZ METHZ TFMSA DCHPA

Carbonic anhydrase II (human, h-CAII) 14:9 3:2 13:4 2:6 12:9 2:0 13:5 1:2

0:77 0:64 0:14 1:29

0:81 0:72 0:08 1:47 NA NA NA

0:41 1:03 0:20 1:18 NA NA

nR1R3 f, mol mol1

a Binding constant obtained by dividing the observed binding constant at T0 ¼ 378C, from Table 6.1, by the fraction of deprotonated sulfonamide at pH 7.0 (from Table 6.2), Kb,T0 ¼ Kb-obs,T0 =fdeprot . bGibbs free energy of deprotonated inhibitor binding to protein at T0 ¼ 378C, Db GT0 ¼
RT lnðKb Þ. cEnthalpy of protein– inhibitor binding at T0 ¼ 378C, obtained from ITC experiments and accounting for the enthalpy of inhibitor deprotonation according to Formula (6.2). dEntropy of protein–inhibitor binding at T0 ¼ 378C (T0 Db ST0 ¼ Db HT0
Db GT0 ). eHeat capacity of protein–inhibitor binding, constant in the temperature range of 10–508C, calculated according to Formula (6.3). fSlopes of the observed enthalpy dependences on the enthalpy of buffer protonation equal to the net number of linked protonation events for reactions (R1)–(R3) calculated by subtracting observed n (Table 6.1) from the fraction of protonated sulfonamide (Table 6.2 lists fractions of deprotonated sulfonamide).

11:68 10:79 10:96 12:24

114 113 61 203

97  21 94  28 80  32 181  42 120  61 64  50

4:1  0:4 1:4  1:8 2:4  0:7 2:6  1:6 1:8  4:1 0:6  5:0

12:9  0:4 11:5  1:8 12:9  0:6 12:3  1:6 7:5  4:0 5:2  5:0

Carbonic anhydrase I (humans, h-CAI )

Db Cp e, cal mol1 K1

8:87  0:24 10:10  0:29 10:54  0:22 9:72  0:32 9:28  0:42 5:79  0:42

T0 Db ST0 d, kcal mol1

1:79
106  8:37
105 1:31
107  7:91
106 2:66
107  1:16
107 7:11
106  4:86
106 3:48
106  3:39
106 1:19
104  1:16
104

Db HT0 c, kcal mol1

ACTAZ METHZ TFMSA DCHPA SULFA PAMBS

Db GT0 b, kcal mol1

Kb,T0 a, M1

Inhibitor

Table 6.3 Thermodynamic parameters of inhibitor binding to carbonic anhydrases h-CAI, b-CAII, and h-CAII at 378C calculated from data in Tables 6.1 and 6.2 to represent reactions (R1)–(R3). This data represents the thermodynamic parameters of ‘anionic’ sulfonamide binding

124 SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

125

CORRELATIONS BETWEEN STRUCTURES

(excluding PAMBS, where binding constants were not sufficiently accurately determined). The largest portion of the observed difference in affinity lies with the difference of sulfonamide protonation thermodynamics and not with the differences of surface interactions between the inhibitors and protein. For all ligands, the binding remained enthalpy driven after accounting for inhibitor protonation. The entropy contribution was even closer to zero after accounting for inhibitor deprotonation. The heat capacities became significantly more negative than the buffer-independent values in Table 6.1. DCHPA remained an inhibitor with about twice the negative heat capacity of other inhibitors.

6.6 Correlations between structures and the thermodynamics of sulfonamide binding to CA Co-crystal structures of several sulfonamide inhibitors bound to human carbonic anhydrase I and II isozymes have been solved.23,32,33 Sequence alignments of the amino acid residues participating in the direct contact with the site where sulfonamide inhibitors bind is shown in Table 6.4. The numbering is that of h-CAI. Amino acids were selected if they are located within  4:5 A˚ of any ligand in the structures shown in Figure 6.9. Sequences are taken from the Swiss-Prot database and reference 2. The first three enzymes are where there are crystal structures and our experimental data are separated from other sequences by a horizontal line. Sequences where no crystal structures are available were aligned through homology. Histidines 94, Table 6.4 Amino acid sequence structural alignment of selected carbonic anhydrase enzyme active site region 91

94

96

119

121

131

143

198

199

200

209

h-CAI h-CAII b-CAII

F I V

H H H

H H H

H H H

A V L

L F F

V V V

L L L

T T T

H T T

W W W

h-CA3 h-CA4 h-CA5 h-CA6 h-CA7 h-CA9 h-CA12 h-CA13 h-CA14

R K K Q K L T R A

H H H H H H H H H

H H H H H H H H H

H H H H H H H H H

V V V V V V V V V

F V Y Y F D A F F

V V V V V L V V V

F L L L L L L L L

T T T T T T T T T

T T T T T T T V V

W W W W W W W W W

126

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

96, and 119 are conserved among all members because they are bound to the catalytic zinc atom. Also conserved are Thr199 and Trp209. Threonine 199 is thought to be crucial for the alignment of bound carbon dioxide, because the mutation of threonine into any amino acid (except serine) diminished catalytic activity by more than two orders of magnitude.34 Thermodynamic parameters listed in Table 6.3 represent the sum of reactions (R1), (R2), and (R3). They are not the ‘true’ parameters of reaction (R1). However, reactions (R2) and (R3) cannot be subtracted from (R1) without their independent experimental measurement. It has been hypothesized that the pKa of the water molecule bound to zinc–CA is the same as that bound to cobalt–CA, and that the enthalpies of protonation are the same.22 Here we do not make such assumptions. Figure 6.9 shows the following structures of the active site region. Portions of protein amino acids directly interacting with the ligand or otherwise important for visualization of the active site are shown in sticks. Atoms that have relatively close distances between the ligand and protein are shown in not-up-to-scale CPK mode. The active site zinc atom is shown in up-to-scale CPK mode. The models are drawn with Web Lab Viewer Lite 4.2 software. Amino acid numbers are the same as in Table 6.4, sequence alignment. (a) shows h-CAI complexed with hydrocarbonate (1HCB). Bicarbonate is bound by two hydrogen bonds to Thr199, similar to how sulfonamides bind. This structure is included to show that the overall positions of all amino acids involved in contacts with ligands are in essentially the same orientation as in structures with bound sulfonamide ligands. (b) shows h-CAI with acetazolamide (1AZM). The relatively bulky ACTAZ ligand is within contact to all amino acid groups shown in ball and stick. Two hydrogen bonds are formed to Thr 199; the rest are essentially hydrophobic van der Waals contacts. (c) shows h-CAI with methazolamide (1BZM). The structure is very similar to the structure of h-CAI with bound ACTAZ ligand. The only significant difference is that His200 is in steric contact with the extra methyl group of the METHZ ligand and, therefore, the His200 side chain is slightly rotated as compared to the ACTAZ structure. (d) shows h-CAI with sulfanilamide (in the presence of mercury, 1CZM). SULFA is a smaller ligand than ACTAZ or METHZ; therefore, it does not make contact with Leu131. The rest of the structure is similar to ACTAZ and METHZ. (e) shows h-CAII complexed with bromide (1RAZ). This structure is included as a reference to the structure of h-CAII with TFMSA. All amino acids are in nearly identical positions as when TFMSA is bound instead of bromide. (f) shows h-CAII with trifluoromethanesulfonamide (1BCD). TFMSA is bound by two hydrogen bonds to Thr 199, and the zinc atom by an ionic bond. TFMSA is a relatively small ligand and it fits into a small space between Leu198 and Trp209, where no other sulfonate ligands of this study could fit. Therefore, TFMSA is in a different orientation than other sulfonamides.

CORRELATIONS BETWEEN STRUCTURES

127

All bound sulfonamide inhibitors form an ionic bond with zinc atom and two hydrogen bonds to threonine 199: between the sulfonamide N and the backbone carbonyl, and between an O and Thr 200 backbone amide. METHZ, ACTAZ, and SULFA are oriented outwards, pointing towards the solvent with their hydrophobic moiety. TFMSA, however, is oriented differently, with fluorine atoms pointing towards the tryptophan 209. Only the relatively small molecule of TFMSA could fit in such an orientation. Therefore, TFMSA, as the smallest inhibitor in this study, is the only one that does not reach Ile91 or Phe91 in any of the isozymes. A co-crystal structure with TFMSA is available only for h-CAII. The exact mode of TFMSA binding to the other two enzymes is assumed to be in the same orientation as with h-CAII. Trifluoromethane sulfonamide bound with similar binding constants to all three carbonic anhydrases (Table 6.3). However, the enthalpies of binding, which were identical for h-CAI and h-CAII (
12:9 kcal mol)
1 , were less favourable for b-CAII (
9:7 kcal mol
1 ). This difference may be associated with either His200, which is Thr in b-CAII, amino acid 121 (Ala in h-CAI, Val in b-CAII, and Leu in h-CAII), or some group external to the active site. Most likely, the difference is due to His200, which is in direct contact with TFMSA. Other studied inhibitors bound significantly more strongly to CAII than to CAI. Active sites of the two carbonic anhydrases are structurally very similar. The following amino acids differ (Table 6.4) among h-CAI, h-CAII, and b-CAII near the active site based on sequence homology:23 F91(h-CAI)– I91(h-CAII)–V91(b-CAII), A121–V121–L121, L131–F131–F131, H200–T200– T200. These differences may account for the differences in thermodynamics of inhibitor binding. When comparing ACTAZ with METHZ, which differs by only one methyl group, we see that ACTAZ binds more strongly than METHZ to both CAII, but METHZ binds stronger to CAI than to both CAII. This difference cannot be explained by the difference of the amino acid in position 200. Most likely the difference lies with the amino acids in positions 91, 121, and 131. When comparing the enthalpies of ACTAZ and METHZ binding to the three enzymes, the enthalpy is most negative for h-CAII and least negative for hCAI. Another difference between ACTAZ and METHZ is in the linked deprotonation reaction of h-CAI. There is no such difference, however, for hCAII and b-CAII. The first explanation of such a difference is that His200 in h-CAI changes its pKa upon METHZ binding, but not on ACTAZ binding. The methyl group present in METHZ is located so close to His200 in h-CAI (Figure 9(c)) that the His side chain is significantly rotated. Experimentally, we observe that h-CAI exhibits such a linked protonation reaction, whereas both h-CAII and b-CAII, which have Thr instead of His, do not exhibit such a linked protonation reaction. PAMBS is structurally very similar to SULFA. They differ by only one methyl group, which causes the PAMBS amino group to be positively charged

128

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

Figure 6.9 The structures of the active site regions of human carbonic anhydrase I (panels a–d) and II (human, panels e, f) with bound inhibitors. (a) h-CAI with bound bicarbonate (PDB code 1HCB). (b) h-CAI with bound acetazolamide (ACTAZ, 1AZM). (c) h-CAI with bound methazolamide (METHZ, 1BZM). (d) h-CAI with bound sulfanilamide in the presence of Hg (SULFA, 1CZM). (e) h-CAII with bound bromide anion (1RAZ). (f) hCAII with bound trifluoromethanesulfonamide (TFMSA, 1BCD)

CORRELATIONS BETWEEN STRUCTURES

129

Figure 6.9 (continues)

130

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

Figure 6.9 (continued)

CONCLUSIONS

131

at pH 7.0. This difference is entirely enthalpic and causes over 100-fold reduction in binding constant from SULFA to PAMBS. The positive charge reduces the sulfonamide pKa from 10.10 (SULFA) to 8.35 (PAMBS), making the binding tighter. Therefore, the presence of the positive charge on the amino group of PAMBS reduces the affinity to carbonic anhydrase by about four orders of magnitude. The binding of 4-carboxybenzene sulfonamide (CARBS) to bovine carbonic anhydrase II has recently been studied by a number of laboratories worldwide with the goal of determining statistical uncertainty of titration calorimetry and other data.35 At pH 7.0 this inhibitor has additional negative charge on the carboxylic group, making it a significantly stronger binder than the structurally similar SULFA or PAMBS. Interestingly, the enthalpy obtained after accounting for the inhibitor deprotonation is the most negative of all inhibitors in this study, consistent with additional specific interactions.

6.7 Conclusions 1. Dissection of all contributing reactions to the observed thermodynamics of inhibitor binding to a protein is necessary in order to correlate the thermodynamic parameters with the structure of the protein–ligand complex. 2. Sulfonamide inhibitors bound more strongly to both CAII than to CAI. An especially large difference of over two orders of magnitude was observed for dichlorophenamide. The smallest difference among the three enzymes was observed for TFMSA and METHZ binding. 3. The major difference in affinities of the inhibitors lies within the differences of sulfonamide group pKa values. After accounting for the deprotonation of the ligand, the range of binding constants for the inhibitors was reduced from four to one order of magnitude. 4. For all ligands, the binding was enthalpy driven, especially for the most tightly binding ligands. After accounting for the enthalpic contribution to sulfonamide deprotonation and the sulfonamide pKa , the binding enthalpy was much more similar among inhibitors. 5. The change in heat capacity of binding was negative, consistent with the burial of hydrophobic surfaces upon binding. Heat capacity of protonation was positive, thus making the heat capacity of binding even more negative after accounting for inhibitor deprotonation.

132

SULFONAMIDE INHIBITOR BINDING TO CARBONIC ANHYDRASE

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35.

Baker BM and Murphy KP. (1996) Biophys. J. 71(4): 2049–2055. Sly WS and Hu PY. (1995) Annu. Rev. Biochem. 64: 375–401. Lindskog S. (1997) Pharmacol. Ther. 74: 1–20. Thoms S. (2002) J. Theor. Biol. 215: 399–404. Casini A, Scozzafava A, Mastrolorenzo A and Supuran LT. (2002) Curr. Cancer Drug Targets 2: 55–75. Mattioni BE and Jurs PC. (2002) J. Chem. Inf. Comput. Sci. 42: 94–102. Masereel B, Rolin S, Abbate F, Scozzafava A and Supuran CT. (2002) J. Med. Chem. 45: 312–320. Henkens RW, Watt GD and Sturtevant JM. (1969) Biochemistry 8: 1874–1878. DiTusa CA, Christensen T, McCall KA, Fierke CA and Toone EJ. (2001) Biochemistry 40: 5338–5344. DiTusa CA, McCall KA, Christensen T, Mahapatro M, Fierke CA and Toone EJ. (2001) Biochemistry 40: 5345–5351. Hunt JA, Ahmed M and Fierke CA. (1999) Biochemistry, 38: 9054–9062. Mansoor UF, Zhang XR and Blackburn GM. (2000) Exs: 437–459. Nair SK, Krebs JF, Christianson DW and Fierke CA. (1995) Biochemistry 34: 3981–3989. Abbate F, Supuran CT, Scozzafava A, Orioli P, Stubbs MT and Klebe G. (2002) J. Med. Chem. 45: 3583–3587. Baird TT Jr, Waheed A, Okuyama T, Sly WS and Fierke CA. (1997) Biochemistry 36: 2669–2678. Doyon JB, Hansen EA, Kim CY, Chang JS, Christianson DW, Madder RD, Voet JG, Baird TA Jr, Fierke CA and Jain A. (2000) Org. Lett. 2: 2557–2558. Madder RD, Kim CY, Chandra PP, Doyon JB, Baird TA Jr, Fierke CA, Christianson DW, Voet JG and Jain A (2002) J. Org. Chem. 67: 582–584. Tripp BC, Smith K and Ferry JG (2001) J. Biol. Chem. 276: 48615–48618. Carlsson U and Jonsson BH. (2000) Exs: 241–259. Semisotnov GV, Rodionova NA, Razgulyaev OI, Uversky VN, Gripas AF and Gilmanshin RI. (1991) Biopolymers 31: 119–128. Binford JS, Lindskog S and Wadso I. (1974) Biochim. Biophys. Acta 341: 345–356. Khalifah RG, Zhang F, Parr JS and Rowe ES. (1993) Biochemistry 32: 3058–3066. Stams T and Christianson DW. (2000) Exs: 159–174. Coleman JE. (1975) Annu. Rev. Pharmacol. 15: 221–242. Maren TH and Sanyal G. (1983) Annu. Rev. Pharmacol. Toxicol. 23: 439–459. Supuran CT, Briganti F, Tilli S, Chegwidden WR and Scozzafava A. (2001) Bioorg. Med. Chem. 9: 703–714. Matulis D, Salemme R and Todd M. (2004) in preparation. Pantoliano MW, Petrella EC, Kwasnoski JD, Lobanov VS, Myslik J, Graf E, Carver T, Asel E, Springer BA, Lane P and Salemme FR. (2001) J. Biomol. Screen. 6: 429–440. Matulis D and Todd M. (2004) Biochemistry in preparation. Todd MJ and Salemme FR. (2003) Genetic Eng. News 23. Butler JN and Cogley DR. (1998) Ionic Equilibrium. Solubility and pH Calculations, Wiley, New York. Chakravarty S and Kannan KK. (1994) J. Mol. Biol. 243: 298–309. Liljas A, Hakansson K, Jonsson BH and Xue Y. (1994) Eur. J. Biochem. 219: 1–10. Krebs JF and Fierke CA. (1993) J. Biol. Chem. 268: 948–954. Myszka DG, Abdiche Y, Arisaka F, Byron O, Eisenstein E, Hensley P, Lombardo C, Schwarz FP, Thommson J, Stafford W and Doyle ML. (2003) J. Biomol. Tech. 14: 247–269.

7 Energetics of the Interaction of Human Acidic Fibroblast Growth Factor with Heparin and the Functional Analogue Myo-Inositol Hexasulfate Mercedes Guzma´n-Casado, Marı´ a M. Garcı´ a-Mira, Pedro Cano-Soldado, Guillermo Gime´nez-Gallego, Jose M. Sanchez-Ruiz and Antonio Parody-Morreale

7.1 Introduction Fibroblast growth factors (FGFs) form a protein family of 23 members,1 which have a broad range of biological activities and are found throughout most life forms from nematodes to humans.2 Acidic fibroblast growth factor (aFGF or FGF-1) and basic fibroblast growth factor (bFGF of FGF-2) are the best characterized members of the family. These two proteins are 154 amino acids long and present 55 per cent homology.3 All FGFs have a high affinity for glycosaminoglycan heparin and for cellsurface heparan sulfate proteoglycans and their different biological activities appear to be mediated by the binding to these molecules.4 Heparin is a linear sulfated polysaccharide chain with alternating L-iduronic and D-glucosamino sugars containing a large number of 6-O-sulfate groups.5 The interaction of Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

134

INTERACTION OF aFGF WITH HEPARIN AND MHS

FGFs with heparin is the subject of intense research and controversy from both functional and molecular points of view (cf. reviews in References 6 and 7). Two different patterns of assemblage of the FGF:heparin complex seem possible. In a cys-type binding FGF molecules would bind as a dimer to the same side of the heparin molecule whilst in a trans-type binding FGF molecules would bind to opposite sites of the heparin molecule. Whatever the case, the concept of heparin-induced FGF oligomerization8 is widely accepted by researchers in the field. Experiments with newt aFGF have demonstrated, however, that oligomerization is not necessary for its cell proliferation activity.9 Myo-inositol hexasulfate (MHS) is a fully sulfated cyclic polyalcohol with some of the structural features of the heparin molecule. It was reported to stabilize in aqueous solution an otherwise unstable 141-amino-acid form of aFGF10,11 and it has been shown to mimic heparin, as an activator of aFGF in mitogenic assays.12 Most of the amino acids in the aFGF molecule that interact with MHS (Lys126, Lys127, Arg133, Lys142)12 also interact with heparin (Lys126, Lys127, Lys142).13 Figure 7.1 shows the structures of a common repeating disaccharide unit of heparin and MHS. We present here results obtained in calorimetric measurements of the interaction with low average molecular weight heparins (3000 or 6000 g mol71)14 or MHS15 of two aFGF forms (139 amino acids (139aFGF)16 or 132 amino acids (132aFGF), both of them truncated at the amino terminal end). The acronym aFGF will be used interchangeably for both forms. These studies were mainly addressed at gaining insight, from the thermodynamic point of view, into the interactions, with a special consideration of elucidating the energetics of the potential heparin-induced oligomerization.

Figure 7.1 Haworth representation of the structure of the main repeating disaccharide of heparin and tridimensional structure of myo-inositol hexasulfate

INTRODUCTION

135

Our calorimetric measurements were performed using a Gill titration calorimeter described in previous papers.19,20 The instrument is of the heatconduction type, the protein being placed in a 203 ml reaction bulb into which successive 5–10 ml aliquots of heparin ligand are injected into protein. The upper panel in Figure 7.2 shows the raw calorimetric signals for a couple of titrations. The area under the peaks accounts for 5 per cent of the heat effect after an injection, the remaining 95 per cent being measured via power compensation through a heater wrapped around the reaction cell. We calculated heats of interaction and corrected for dilution effects and complete titration curves were obtained (see the lower panel of Figure 7.2), which could be analysed to obtain the enthalpy change and binding constant values21 when the protein–ligand affinities lay in the 102–107 M71 range. Details of the data analysis are given below. The corrected heats (qi), measured after each injection step i, can be related to the enthalpy per mole of macromolecule, ðH

H
0 Þi 22 at the free-ligand concentration at that step i, by the expression23 qi ¼ ðH

H
0 Þi ½MT,i V
ðH

H
0 Þi
1 ½MT,i
1 ðV
nÞ

ð7:1Þ

V ¼ 203 ml is the effective reaction volume, which remains constant throughout the experiment, since for each volume injected, v, an equal volume is ejected. [M]T,i is the total protein concentration in the cell after an injection step i and is related to the initial concentration in the cell, [M]T,0, by [M]T,i ¼ [M]T,0(17v/V)i . The enthalpy per mole of macromolecule at step i, ðH

H
0 Þi , is given by22 @lnPi ðH

H
0 Þi ¼
R @1=T

ð7:2Þ

where R is the gas constant, T the temperature and Pi the binding polynomial for the interaction in question. In our case, assuming that there are n identical, independent sites for the binding of ligand, L, to the aFGF molecule, with a constant K, the binding polynomial can be written as Pi ¼ ð1 þ K½Li Þn

ð7:3Þ

Incorporating Equation (7.3) into (7.2) and taking into account van’t Hoff’s equation relating the equilibrium constant to the change in enthalpy, DH, we obtain ðH

H
0 Þi ¼

nDHK½Li 1 þ K½Li

ð7:4Þ

where [L]i is the free-ligand concentration and is calculated from the conservation equation [L]T,i ¼ [L]i+[L]b,i, where [L]T,i and [L]b,i are the total and bound-ligand concentrations respectively at step i. The former varies in

136

INTERACTION OF aFGF WITH HEPARIN AND MHS

Figure 7.2 Upper panel: raw calorimeter signals in the titrations of human 132aFGF with heparin (upper left corner) and myo-inositol hexasulfate at 258C in 20 mM NaPi, 0.3 M NaCl, pH 7.0. Lower panel: human 139aFGF titrations with heparin (open circles) and myo-inositol hexasulfate (filled circles) in 20 mM NaPi, 0.15 M NaCl, 1 mM DTT (dithiothreitol), pH 7, 258C. The lines represent the best fit based on a model of identical and independent binding sites. The protein in the reaction bulb (203 ml) of the calorimeter, at 0.24 mM concentration in both cases, was titrated with 10 ml injections of 0.59 mM heparin or 0.59 mM myo-inositol hexasulfate

the cell after each injection according to [L]T,i ¼ [L]T,s[1
(17v/V)i], where [L]T,s is the constant ligand concentration in the syringe. In the case of identical and independent binding, [L]b,i would be given by [M]t,i [L]i/ (1+K[L]i) and [L]i is the unknown of a second-grade equation. The binding parameters DH, K and n were determined for a titration experiment by a non-linear, least-squares fitting of Equation (7.1) to the corrected heats qi, where ðH

H
0 Þi is given by Equation (7.4) and [L]i by the

THERMODYNAMIC PARAMETER DERIVED FROM ITC EXPERIMENTS 137

analytical solution of the second-grade equation. The fitting was carried out with a program developed in our laboratory24 based on the Simplex algorithm.25 The standard errors given in the parameters are the errors in the fitting procedure of a single titration experiment and were calculated by the Bevington method.26 The performance of the calorimeter in determining the thermodynamic parameters and their errors has been reported elsewhere,21 where it is shown that the errors obtained in the fittings are similar to the errors obtained after measuring the system several times. In the case of heparin, a single site molecular weight of 1500 g mol71 was considered in the calculation of its concentration. This involves the assumption that a 3000 g mol71 heparin molecule has two binding sites for aFGFs. This assumption is based on the fact that the 3000 g mol71 heparin must be composed of between eight and 10 monosaccharides and thus of two tetra- or pentasaccharides, which, according to various authors,27–29 constitute the minimum unit to permit interaction between heparin and aFGFs. A 6000 g mol71 heparin molecule would have four binding sites for the protein molecules. Thus, the heparin concentration used in the analysis is either double or four times the nominal concentrations that appear in the descriptions of the experiments. An analysis based on Equations (7.1)–(7.4) yielding a value of n ¼ 1 imply that we are looking at the interaction between the protein and the heparin binding site. Calorimetric measurements in the last experiments shown in this paper, to study the aFGF concentration dependence of heparin binding, were made with an MCS titration calorimeter (Microcal Inc, MA, USA).30 The data were analysed using software provided by Microcal. We calculated accessible surface area (ASA) using a program written by us and based on a modification of the Shrake–Rupley algorithm31 in which 2000 points are uniformly and randomly distributed on the surface of the solvated van der Waals sphere corresponding to each non-hydrogen protein atom. A radius of 1.4 A˚ for the solvent probe and the Chothia set for the protein atoms were always used.

7.2 Thermodynamic parameter derived from ITC experiments Our work with FGFs started with the measurement of heparin and MHS binding at neutral pH and the presence of 0.15 M NaCl. The lower panel of Figure 7.2 shows titrations with both ligands of 139aFGF. The qi corrected heat values measured are shown as circles in the figure. The lines of the plots represent the best fit of the model of n identical, independent sites stated in the preceding section to the data. The parameters DH, K, and n obtained from the fitting of the titrations for both proteins are shown in Table 7.1. The values of the number of sites are close to unity. This was to be expected for the

138

INTERACTION OF aFGF WITH HEPARIN AND MHS

interaction between aFGF and MHS; as Pineda-Lucena et al.12 showed in their NMR studies, it has a 1:1 stoichiometry. Besides DH, calculated thermodynamic parameters DG8 and DS for the reaction are also given in Table 7.1. All the thermodynamic parameters in this paper are given per mole of protein. Titration experiments (not shown) at different pH values in the 6–8 range at the same 0.15 M NaCl concentration showed that there is a negligible proton balance in the binding of any of the ligands to either one of the proteins. This result is confirmed as well by experiments in the presence of Pipes (piperazineN,N’-bis[2-ethanesulfonic acid]) buffer, which has a different enthalpy of protonation than phosphate: the same values for the enthalpy changes were obtained with both buffers. We also report in Table 7.1 the values for the heat capacity changes associated with ligand binding. These were derived from DH measurements at different temperatures, for example those for heparin binding to 139aFGF shown in the upper panel of Figure 7.4. The heat capacity change is calculated from the slope. An overall examination of Table 7.1 shows that the thermodynamic parameters for the binding of the same ligand to each protein are quite similar. In the case of MHS binding, the interaction is both enthalpically and entropically driven. In the case of heparin, it is the enthalpy change that drives the interaction, the entropy change being unfavourable. The affinity for MHS of any of the proteins is slightly higher than that of heparin. Experiments described so far were carried out with both forms of the protein at 0.15 M NaCl. At this point it is important to note that at 0.15 M NaCl the protein collected after gel filtration for the calorimetric experiment was slightly cloudy. This cloudiness was eliminated by centrifuging at 15 000 g for 10 min. The supernatant was then completely transparent with a protein loss of less than 2–3 per cent. We may fairly deduce that at the aFGF concentrations used in our calorimetric experiments at the ionic strength of the 0.15 M NaCl medium we are working close to the solubility threshold of the protein. We analysed our results accepting that the kinetics of any Table 7.1 Thermodynamic parameters for heparin and myo-inositol hexasulfate (MHS) binding to 139aFGF and 132 aFGF in 20 mM NaPi, 0.15 M NaCl, pH 7.0 DH/ kJ mol71 heparin + 139aFGF* 132aFGF MHS+ 139aFGF* 132aFGF

K/M71

n

DG8/ kJ mol71

DS/ DCp/ J K71 mol71 J K71 mol71

737.6+1.3 738.5+4.6

(6.5+0.6)6104 1.01+0.01 727.2+0.4 (5.1+1.7)6104 0.78+0.05 726.8+1.7

733+4 742+18

7720+160 7790+420

721.8+0.9 720.1+0.4

(3.8+2.2)6105 0.93+0.01 731.8+2.5 (5.9+1.3)6105 1.11+0.01 733.1+0.4

33+8 42+4

7360+90 7380+50

*Plus 1 mM DTT in the reaction medium.

THERMODYNAMIC PARAMETER DERIVED FROM ITC EXPERIMENTS 139

Table 7.2 Thermodynamic parameters for heparin binding to 132aFGF in 20 mM NaPi, pH 7.0, 258C and different NaCl concentrations [NaCl]/M

DH/kJ mol71

K/M71

n

0.20 0.30 0.40 0.60

728.4+3.3 725.9+2.1 720.9+0.8 716.7+2.1

(2.6+0.7)6105 (1.8+0.2)6105 (2.8+0.9)6104 (9.3+4.6)6103

1.08+0.02 1.02+0.01 1.05+0.04 1.08+0.1

DG8/kJ mol71 DS/J K71 mol71 726.7+2.1 725.1+0.8 725.5+0.4 723.0+0.8

75.4+12.5 72.9+7.1 15.5+2.9 20.9+7.1

aggregation process affecting the protein would be so slow as to be ignored. This point will need to be confirmed and the conclusions arrived at from an analysis of these results must be taken with all due caution. In view of the above, and taking into account a low yield in the purification of 139aFGF, we decided to continue our work with 132aFGF and at NaCl concentrations of 0.2 M and above, at which no solubility problems were detected. Ionic strength dependence of the interaction with 132aFGF was studied both for heparin and MHS, at neutral pH and 258C, in a NaCl concentration range from 0.2 to 0.6 M NaCl. Results obtained in the calorimetric titrations with heparin and the thermodynamic parameters calculated are shown in Table 7.2. For the case of MHS we show the thermodynamic parameters for the interaction in the upper panel of Figure 7.3. We see a decrease in the affinity for MHS of the protein with increasing ionic strength, an effect that is entirely enthalpic, since the binding entropy remains constant in the whole NaCl range. This is not the case in the interaction with heparin, in which the change in entropy makes a net contribution to the salt-concentration dependence of the affinity as can be seen in Table 7.2. In the lower panel of Figure 7.3 the dependence of the logarithm of the dissociation constant of the heparin 132aFGF complex is represented versus the logarithm of the salt concentration. A straight line is obtained with a slope of 1.3+0.6 and a y intercept of 73.8 at log([Na+] ¼ 1 M) ¼ 0. The heat capacity changes for the interaction of heparin and MHS with 132aFGF at neutral pH and 0.4 M NaCl concentration were determined from experiments at three different temperatures. To correct for any possible contribution from protonation changes to the heat measurements either in the protein or the buffer we performed our experiments in phosphate, Pipes and Bes (N,N-bis[2-hydroxyethyl]-2-aminoethanesulfonic acid) buffers, which have different protonation enthalpies. In this situation, the experimentally measured binding enthalpy DH at any temperature y can be expressed as14,15 DH ¼ DHb,25 þ DCp ðy
25Þ þ DnðDhhis
Dhbuffer Þ

ð7:5Þ

where Dhhis is the enthalpy of protonation of the histidine groups (under the assumption that, at pH 7, only histidine residues are involved in any ionization in the protein), and Dhbuffer the enthalpy of protonation of the buffer. DHb,25 is

140

INTERACTION OF aFGF WITH HEPARIN AND MHS

Figure 7.3 Variation with the logarithm of NaCl concentration of the thermodynamic parameters for the human 132aFGF reaction with myo-inositol hexasulfate (upper panel) and of the dissociation constant of the human 132aFGF and heparin complex (lower panel). 20 mM NaPi, pH 7.0, buffer at 258C in both cases

the intrinsic change in enthalpy due to the binding of heparin or MHS at the reference temperature of 258C, DCp the heat capacity change in the binding (presumed to be constant) and Dn the number of protons taken up by the protein from the medium upon binding of the ligand. These last three parameters were determined after the bidimensional fitting of Equation (7.5) to the experimental DH data, that is to say, taking DH to be a function of the two independent variables (y and Dhbuffer) and DHb,25, DCp and Dn the three adjustable parameters. The fits were excellent, as shown by comparison between experimental and predicted DH values (lower panel in Figure 7.4). The values for the fitted parameters are collected in Table 7.3.

THERMODYNAMIC PARAMETER DERIVED FROM ITC EXPERIMENTS 141

Table 7.3 Enthalpy, heat capacity change, proton uptake and changes in accessible surface area in the binding of heparin and myo-inositol hexasulfate (MHS) to human 132aFGF in 20 mM NaPi, 0.4 M NaCl, pH 7.0 Ligand Heparin MHS

DHb,25/kJ mol71 DCp/J K71 mol71 720.8+1.2 712.6+0.33

796+75 7142+13

Dn

DASAnp/A˚2

DASAp/A˚2

70.10+0.07 0.09+0.02

7190 7180

7230 7180

Figure 7.4 Temperature dependence of human aFGF reaction with heparin and myoinositol hexasulfate. Upper panel: temperature dependence of the enthalpy change for the binding to human 139aFGF of heparin in 20 mM NaPi, 0.15 M NaCl, 1 mM DTT, pH 7.0. Lower panel: measured enthalpy changes in the reaction of human 132aFGF with myoinositol hexasulfate in NaCl 0.4 M, pH 7.0 medium at three different temperatures and buffers (phosphate, Pipes and Bes) versus values predicted by Equation (7.5) with the fitted parameters shown in Table 7.3. Different symbols are used for the three temperatures of the experiments: *, 98C; &, 178C; !, 258C

142

INTERACTION OF aFGF WITH HEPARIN AND MHS

Work carried out during the last decade supports the view that the changes in thermodynamic quantities associated with folding and ligand-binding processes can be parametrized in terms of the corresponding changes in nonpolar (ASAnp) and polar (ASAp) areas exposed to the solvent. Thus, Freire and co-workers have suggested32,33 the following equations: (1) for the heat capacity capacity change, DCp ¼ 0:45DASAnp
0:26DASAp

ð7:6Þ

and (2) for the enthalpy change (at the reference temperature of 608C), DH60 ¼
8:44DASAnp þ 31:4DASAp

ð7:7Þ

(DCp, DH60 and DASA in cal K71 mol71, cal mol71 and A˚2 units respectively). From the measured enthalpy change at 258C, the corresponding value at 608C is calculated using DH60 ¼ DH25+DCp(60725). If we admit the expression for the enthalpy change to be valid for binding processes, we would have a system of two equations, (7.6) and (7.7), with two unknowns, from which we can calculate changes in accessible surface areas. Using the values of the first two data columns of Table 7.3 we arrived at the changes in DASAnp and DASAp in the binding of heparin or MHS to 132aFGF reported in the last two columns of the same table.

7.3 Discussion Almost identical results for the thermodynamic parameters were obtained with both aFGF forms in our studies at the lower ionic strength (0.15 M [NaCl]). This is consistent with the fact that the seven amino acids lacking in the amino-terminal end of 132aFGF (Asn–Leu–Pro–Pro–Gly–Asn–Tyr) are not involved in the binding to heparin13 or MHS.12 Although no measurements were made with 139aFGF at higher ionic strengths, one might expect that no significant differences between the two forms of the protein in the binding of either one of the ligands would arise. Energetic investigation of aFGF interactions is scarce. Spivak-Kroizman et al.8 have reported isothermal titration calorimetry studies of the interaction of aFGF with heparins of 4.8 and 16 kDa molecular weight. Their data were analysed assuming identical and independent binding of the protein to the polysaccharide, giving values (around 26106 M71) for the equilibrium constant higher than ours (66104 M71), whilst the change in enthalpy per site reported was lower (around 720 kJ mol71 versus 738 kJ mol71). As their experimental conditions were quite similar to ours (50 mM Hepes, 0.15 M NaCl, 1 mM EDTA, pH 7.4, 258C) our only conjecture at this moment is that these discrepancies might be due to the lack of chemical reproducibility between different heparin preparations.

DISCUSSION

143

The effects of ionic concentration upon the energetics of the binding between ligands and biological macromolecules have been mainly studied with nucleic acids.34 The effects observed in these cases are put down to the fact that the nucleic acid is a linear charged polyelectrolyte in which counterion condensation occurs on the molecule to reduce its charge density35 and that during the binding of ligands cationic counterions are released. It is important to note, however, that the effect of the salinity in these cases is purely entropic, associated with the dilution of the ions released as a result of the binding process. In the case of MHS binding to 132aFGF, the saline effect (upper panel of Figure 7.3) is fundamentally enthalpic and not entropic. We might perhaps imagine that the ions released as a consequence of the protein–ligand interaction were previously bound in a specific way. In fact, the structural studies carried out by Pineda-Lucena et al.12 suggest that four positively charged amino acid residues (Lys 126, Lys 127, Lys 142 and Arg 133) interact directly with MHS when bound to aFGF. It would seem reasonable to presume that, in the presence of relatively high concentrations of NaCl, these positive charges will be either partially or even wholly neutralized by chloride ions in the ligand-free protein and that these chloride ions are released upon the ligand’s binding to produce the observed saline effect. We thus postulated15 a simple model for MHS binding that encompasses these ideas and gave us an adequate framework for an analysis of our experimental results. The model implies a competitive binding of chloride ions and MHS to the same site in the aFGF molecule as depicted in Figure 7.5. The mathematical formulation of the model led us to the following equation for the dependence of the interaction Gibbs energy on the chloride concentration


@DG8 @ln½Cl


 ¼ RThii

ð7:8Þ

T

where hii is the mean number of chloride ions bound to 132aFGF and thus released upon binding of MHS. The slope of DG8 versus ln[Cl7] (upper panel of Figure 7.3) leads to a value of 3.3+0.8 for hii, a value consistent with our previous supposition of n ¼ 4. We may say then that the process that has been studied experimentally may imply the replacement of *3 bound chloride ions by MHS. This would mean that the relatively low enthalpy values for the binding process might well be simply the result of a partial cancelling out of the two large enthalpic effects. Since heparin is a linear charged polyelectrolyte, heparin binding to 132aFGF may be similar to the binding of charged molecules to linear nucleic acids. The interaction of a protein with heparin would be accompanied by counterionic (vgr.: Na+) release into the medium. As it happens with nucleic

144

INTERACTION OF aFGF WITH HEPARIN AND MHS

Figure 7.5 A model to interpret the effect of NaCl concentration upon the energetics of myo-inositol hexasulfate binding to 132aFGF

acids, this effect implies a favourable entropic contribution to the free energy of binding, which we in fact observed with NaCl concentrations slightly over 0.2 M NaCl, and increasing concomitantly with the salt concentration (see Table 7.2). A dependence upon ionic strength of the interaction between heparin and other heparin binding proteins such as thrombin,36 antithrombin III,37 mucus proteinase inhibitor38 and the basic form of FGF29 has been reported in the literature. When the log of the dissociation constant of the complex is represented versus the log of the salt concentration a straight line should be obtained, the slope of which gives the number of counterions released (in our case Na+) multiplied by a factor of C.29 This factor gives the fraction of Na+ ions bound to heparin, depends upon the axial charge density of the polymer and has a value of 0.8.39 Thompson et al.29 have calculated such slopes for the above-mentioned proteins. The lowest value (1.9) was obtained for bFGF and was similar to that of antithrombin. That value implied that two or three Na+ ions are displaced when bFGF binds to heparin. In the lower part of Figure 7.3 we represent our data for heparin binding to 132aFGF in the same way. We obtain a straight line with a lower slope of 1.3+0.6, which corresponds to one or two Na+ ions released during the binding of aFGF. Structural characterization13 of a 140-amino-acid form of aFGF binding to a heparin decasaccharide has shown the participation of three (Lys 126, Lys 127 and Lys 142) out of the four amino acids that interact with MHS. This implies that both ligands, heparin and MHS, bind to the same site in the protein, to which (according to the model to explain the ionic strength dependence of MHS binding) *3 chloride ions bind. We might speculate then that if one less amino acid participates in the interaction with heparin, one less chloride ion (i.e. *2) will be released from the site in the protein. Thus we could say that the global counterionic release in the aFGF binding to heparin

DISCUSSION

145

would be one of one or two Na+ ions from the heparin molecule and *2 Cl7 from the aFGF molecule. According to several authors (Thomson et al.,29 and references therein) the graph of the log of the dissociation constant versus the log of NaCl concentration also allows us to estimate the ionic versus non-ionic contribution during binding. It is assumed that at high ionic strength, say 1 M NaCl, the interaction is all non-ionic, as all charges are then neutralized. By extrapolating the dissociation constant values in the lower part of Figure 7.3 to that salt concentration we measure a change in free energy during the binding of aFGF to heparin of 722 kJ mol71. If afterwards we go to a low ionic strength, –27 kJ mol71 at 0.15 M NaCl for example, the difference would then be the ionic contribution to the binding in this case, 75 kJ mol71 out of 727 kJ mol71, which is a small (520 per cent) ionic contribution. Consequent to this line of argument, we might then assign a value of 75 kJ mol71 to the formation of one or two pure ionic interaction(s) between net negative charge(s) on the heparin molecule and net positive charge(s) on the aFGF molecule. To fully understand the molecular basis of the major non-ionic specificity between aFGF and heparin, an accurate knowledge of the chemical structure of the heparin molecule used in the experiments would be needed. Using Freire’s parametric equations we calculated changes in total wateraccessible surface areas, both non-polar and polar, upon binding to 132aFGF of heparin or MHS in a neutral, 0.4 M NaCl medium (Table 7.3). Very similar values, around 7200 A˚,2 were obtained for both ligands. The negative sign implies a contraction of the protein when the ligands binds, and the absolute value shows that the conformational change is very small. Because structural knowledge of the interaction of 132aFGF with MHS is available, we have calculated15 changes in accessible surface areas, both for free and MHS-bound aFGF, on the basis of 25 molecular dynamics NMR structures.12,40 Figure 7.6 shows the results obtained for each amino acid in the protein. Values determined for the total surface areas were 7961+233 A˚2 (4121+151 A˚2 nonpolar and 3840+124 A˚2 polar) for free aFGF and 7645+165 A˚2 (3928+120 A˚2 non-polar and 3717+114 A˚2 polar) for MHS-bound aFGF, from which values upon binding of 7193+193 A˚2 for the change in non-polar surface area and 7123+168 A˚2 for the change in polar surface area were calculated. These values are in very good agreement with the values obtained from thermodynamic data. Figure 7.6 also shows that, upon ligand binding, ASA changes occur significantly in amino acids Ile56, Gln57, Leu58 and Leu149, in the high-affinity receptor-binding region of the protein. Because of the similarities in the accessible surface area changes and the functional analogy between MHS and heparin binding to 132aFGF, it is not unreasonable to assume that heparin binding to aFGF causes similar changes in the same amino-acid cluster. We suggest that the main difference between the energetics of MHS and heparin bindings to aFGF is related to different

146

INTERACTION OF aFGF WITH HEPARIN AND MHS

Figure 7.6 Change in accessible surface area for 132aFGF amino acids upon myo-inositol hexasulfate binding. Amino-acid residues for which changes are significant are highlighted. NMR structures used for the calculations were obtained in a 10 mM NaPi, 0.1 M NaCl, 50 mM Na2SO4, 1 mM b-mercaptoethanol, pH 6.0 medium for the free protein and 10 mM NaPi, 1.5 mM NaCl, 2 mM myo-inositol hexasulfate, pH 6.0 for the bound protein

counterionic behaviour, which is reflected in the salt dependence of the binding entropy change. Two basic models7 have been proposed for the role of heparin in the aFGF (or any other FGF) signal transduction pathway. In one of the models FGF molecules bind in a cys-type mode to the same side of the heparin molecule. In the other, a trans-binding model, FGF molecules bind to opposite sides of the heparin molecule. Both binding modes could be significant from the physiological point of view and it may be that each of them is related to a different biological activity. The cys-type binding, for example, would favour the dimerization of the FGF receptor on the cell surface. In the case of basic FGF, a higher order model composed of four protein molecules complexed to heparin through both cis- and trans-interactions has been proposed as the minimum biologically active unit.41 In any case, the concept of heparin-linked FGF oligomerization (or just dimerization) seems to be widely accepted by

DISCUSSION

147

most researchers in the field. It is not at all clear, however, whether a true protein oligomerization occurs in the binding to heparin or we are merely dealing with the binding of two or more protein molecules to a linear lattice. From our point of view, a true protein oligomerization would imply some protein–protein specificity at a protein–protein interface. In the previously mentioned work,13 although a trans-dimeric structure is assigned to aFGF complexed to heparin, it is also stated that no protein–protein interface appears to exist. In fact, under the light of previous structure–energetics studies in the literature, our results support the hypothesis that no protein– protein interface is formed when aFGF binds to heparin. Thus, Spolar and Record42 have collected values for the heat-capacity change and ASA changes in protein oligomerization processes These values range from 72900 J K71 mol71 for the heat-capacity change and 72200 A˚2 for the total ASA change in the case of a-chymotrypsin dimerization to 76400 J K71 mol71, and 710 000 A˚2 for the same parameters in l cro repressor dimerization. We have obtained in the case of aFGF binding to heparin values far outside these ranges: around 7100 J K71 mol71 for the heat capacity change and 7400 A˚2 for the total ASA change. They suggest that under our experimental conditions (0.4 M NaCl neutral medium) no protein– protein interface is formed and hence there is no aFGF dimerization linked to heparin binding. As we have already pointed out, the small change in total ASA would be correlated to a small increase in protein compactness upon binding, as demonstrated with MHS. In the case of the binding of heparin to 132aFGF at the lower ionic strength of 0.15 M NaCl, we measured higher values for the heat capacity change (7700 to 7800 J K71 mol71; see Table 7.1) and calculated a total ASA change of around 71500 A˚2. These values are still outside the lower end of the range of values reported by Spolar and Record for dimerization processes. Therefore, we feel that aFGF dimerization can still be excluded in these experimental conditions. The conclusion that no dimerization occurs when aFGF binds to heparin is supported by other considerations. In Figure 7.7 we show experiments performed with 6000 g mol71 heparin, a molecule twice as long as the one talked about so far. Titrations were performed in a concentration range down to 0.03 mM aFGF, using an MCS titration calorimeter, an instrument that allows experiments to be performed at lower protein concentrations. The data were analysed using the same independent site model and the same molecular weight of 1500 g mol71 for the minimum protein-binding unit of heparin, which implies the existence of four independent binding sites for the protein in the heparin molecule. The fittings were good and results obtained per binding site were similar to those obtained with the shorter 3000 g mol71 heparin molecule (also shown in Figure 7.7), which strongly supports the idea that we are dealing with independent binding to a lattice.

148

INTERACTION OF aFGF WITH HEPARIN AND MHS

Figure 7.7 Dependence of energetic parameters upon heparin length and 132aFGF concentration. Squares: enthalpy change, DH. Circles: Gibbs energy change, DG8. Filled symbols: binding to aFGF of 3 kDa heparin. Open symbols: binding to aFGF of 6 kDa heparin. Experiments performed with an MCS titration calorimeter at 0.4 M NaCl, 20 mM NaPi, pH 7.0, 258C. Mean values with 3 kDa heparin in experiments at 0.3 mM aFGF concentration: DH ¼ 719.6 kJ mol71; DG8 ¼ 726.2 kJ mol71

Although no direct structural information can be obtained from our thermodynamic measurements, we might infer, both from the quality of the fittings of the simplest binding to a lattice model to our data and from the fact that the heat-capacity changes seem to be incompatible with the formation of a protein–protein interface, a non-cooperative binding to a lattice model for human aFGF binding to heparin. Thus, the concept of heparin-induced oligomerization we believe should be treated with caution. In fact, recent experiments with newt aFGF have demonstrated that oligomerization is not necessary for its cell proliferation activity.9

Acknowledgement This work was supported by Grants BIO96-0411, PB96-1439 and BIO20001437 from the Spanish Government.

References 1. Yamashita T, Yoshioka M and Itoh N. (2000) Biochem. Biophys. Res. Commun. 277: 494–498. 2. Ornitz DM and Itoh N. (2001) Genome. Biol. 2: 3005.1–3005.12.

REFERENCES

149

3. Thomas KA. (1987) FASEB J. 1: 434–440. 4. Yayon A, Zimmer Y, Shen GH, Avivi A, Yarden Y and Givol D. (1992) EMBO J. 11: 1885–1890. 5. Kjellen L and Lindahl U. (1991) Annu. Rev. Biochem. 60: 443–475. 6. Waksman G and Herr AB. (1998) Nat. Struct. Biol. 5: 527–530. 7. Faham S, Linhardt RJ and Rees DC. (1998) Curr. Opin. Struct. Biol. 8: 578–586. 8. Spivak-Kroizman T, Lemmon MA, Dikic I, Ladbury JE, Pinchasi D, Huang J, Jaye M, Crumley G, Schlessinger J and Lax I. (1994) Cell 79: 1015–1024. 9. Arunkumar AI, Kumar TK, Kathir KM, Srisailam S, Wang HM, Leena PS, Chi YH, Chen HC, Wu CH, Wu RT, Chang GG, Chiu IM and Yu C. (2002) Protein Sci. 11: 1050–1061. 10. Dabora JM, Sanyal G and Middaugh CR. (1991) J. Biol. Chem. 266: 23 637–23 640. 11. Volkin DB, Tsai PK, Dabora JM, Gress JO, Burke CJ, Linhardt RJ and Middaugh CR. (1993) Arch. Biochem. Biophys. 300: 30–41. 12. Pineda-Lucena A, Jimenez MA, Nieto JL, Santoro J, Rico M and Gimenez-Gallego G. (1994) J. Mol. Biol. 242: 81–98. 13. DiGabriele AD, Lax I, Chen DI, Svahn CM, Jaye M, Schlessinger J and Hendrickson WA. (1998) Nature 393: 812–817. 14. Guzma´n-Casado M, Garcı´ a-Mira MM, Sanchez-Ruiz JM, Gime´nez-Gallego G and Parody-Morreale A. (2002) Int. J. Biol. Macromol. 31: 45–54. 15. Guzma´n-Casado M, Sanchez-Ruiz JM, El Harrous M, Gime´nez-Gallego G and Parody-Morreale A. (2000) Eur. J. Biochem. 267: 3477–3486. 16. Guzma´n-Casado M, Cardenete A, Gime´nez-Gallego G and Parody-Morreale A. (2001) Int. J. Biol. Macromol. 28: 305–313. 17. Copeland RA, Ji H, Halfpenny AJ, Williams RW, Thompson KC, Herber WK, Thomas KA, Bruner MW, Ryan JA, Marquis-Omer D, et al. (1991) Arch. Biochem. Biophys. 289: 53–61. 18. Zazo M, Lozano RM, Ortega S, Varela J, Dı´ az-Orejas R, Ramı´ rez JM and Gime´nezGallego G. (1992) Gene 113: 231–238. 19. El Harrous M, Gill SJ and Parody-Morreale A. (1994) Meas. Sci. Technol. 5: 1065– 1070. 20. El Harrous M, Mayorga OL and Parody-Morreale A. (1994) Meas. Sci. Technol. 5: 1071–1077. 21. El Harrous M and Parody-Morreale A. (1997) Anal. Biochem. 254: 96–108. 22. Gill SJ, Richey B, Bishop G and Wyman J. (1985) Biophys. Chem. 21: 1–14. 23. Parody-Morreale A, Robert CH, Bishop GA and Gill SJ. (1987) J. Biol. Chem. 26: 10 994–10 999. 24. El Harrous M. (1994) Ph.D. Thesis, Universidad de Granada. 25. Nedler JA and Mead R. (1965) Comput. J. 7: 308–313. 26. Bevington PR. (1969) Data Reduction and Error Analysis for the Physical Sciences, McGraw-Hill, New York. 27. Ornitz DM, Yayon A, Flanagan JG, Svahn CM, Levi E and Leder P. (1992) Mol. Cell. Biol. 12: 240–247. 28. Mach H, Volkin DB, Burke CJ, Middaugh CR, Linhardt RJ, Fromm JR, Loganathan D and Mattsson L. (1993) Biochemistry 32: 5480–5489. 29. Thompson LD, Pantoliano MW and Springer BA. (1994) Biochemistry 33: 3831–3840. 30. Wiseman T, Williston S, Brandts JF and Lin LN. (1989) Anal. Biochem. 179: 131–137. 31. Shrake A and Rupley JA. (1973) J. Mol. Biol. 79: 351–371. 32. Go´mez J, Hilser VJ, Xie D and Freire E. (1995) Proteins 22: 404–412. 33. Hilser VJ, Go´mez J and Freire E. (1996) Proteins 26: 123–133.

150 34. 35. 36. 37. 38. 39. 40.

INTERACTION OF aFGF WITH HEPARIN AND MHS

Lohman TM and Mascotti DP. (1992) Methods Enzymol. 212: 400–424. Manning GS. (1978) Q. Rev. Biophys. 11: 179–246. Olson ST, Halvorson HR, and Bjork I. (1991) J. Biol. Chem. 266: 6342–6352. Olson ST and Bjork I. (1991) J. Biol. Chem. 266: 6353–6364. Faller B, Mely Y, Gerard D and Bieth JG. (1992) Biochemistry 31: 8285–8290. Mattai J and Kwak JC. (1988) Biophys. Chem. 31: 295–299. Pineda-Lucena A, Jime´nez MA, Lozano RM, Nieto JL, Santoro J, Rico M and Gime´nez-Gallego G. (1996) J. Mol. Biol. 264: 162–178. 41. Moy FJ, Safran M, Seddon AP, Kitchen D, Bohlen P, Aviezer D, Yayon A and Powers R. (1997) Biochemistry 36: 4782–4791. 42. Spolar RS and Record MT, Jr. (1994) Science 263: 777–784.

8 Thermodynamics of SH2 Domain Binding Gabriel Waksman, Sangaralingam Kumaran and Olga Lubman

8.1 Introduction Signal transduction is a key event in the response of the cell to changes in its extracellular environment. Cells respond to extracellular signals by transducing through their membranes the molecular message that these signals contain. A well documented signal transduction pathway is that induced by platelet-derived growth factor (PDGF). PDGF receptors play important roles during embryogenesis, particularly in the development of the kidneys, blood vessels, lungs, and the central nervous system.1–4 PDGF receptors also have a role in wound healing in adults.5 The PDGF receptor belongs to the family of receptor tyrosine kinases. Each contains a large extracellular region involved in growth factor binding and an intracellular region that contains a kinase domain responsible for propagation of the signal.6–8 Ligand binding to the PDGF receptor induces autophosphorylation of the kinase domain, which leads to increased catalytic kinase activity and to the addition of phosphate groups on specific tyrosine residues. Thus, the extracellular ‘signal’ that PDGF constitutes is ‘translated’ on the other side of the cell membrane by the specific phosphorylation of tyrosine residues. What happens next is the transmission of the phosphate signal from one protein to another through a succession of protein–protein interactions that ultimately leads to a cellular response such as gene activation or enzyme activation. Such a mechanism Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

152

THERMODYNAMICS OF SH2 DOMAIN BINDING

implies that tyrosines that are phosphorylated upon receptor activation are in some way recognized and serve as docking sites for downstream signal transduction molecules. Such a recognition process is made possible by small protein modules contained in signal transduction molecules, which bind to phosphorylated tyrosine residues in a sequence specific manner. These domains are called Src Homology 2 (SH2) domains.9 SH2 domains were initially discovered as a  100 amino-acid block that was conserved in a number of cytoplasmic tyrosine kinases, including Src, and that, when mutated, appeared to be involved either in regulating kinase activity or in promoting interactions of kinases with other components of the cell.10 Subsequent breakthroughs identified SH2 domains as binding specifically to tyrosine phosphorylated sites (reviewed in reference 11). Since this pioneering work, SH2 domains have been identified in numerous proteins of various origins and with various activities, which include protein tyrosine kinases, transcription factors, or molecular adaptors. SH2-domain-containing proteins are involved in pathways regulating many essential cellular processes such as cytoskeletal re-arrangement, homeostasis, the immune response, or development. Although mutations in SH2 domains are rarely the cause of pathological conditions (see the section ‘SH2 domains and human pathologies’), the involvement of these domains in such a variety of pathways make them ideal candidates as drug targets (see the section ‘SH2 domains as drug targets’): indeed, disease states often result from alterations in signal transduction proteins; thus the hope that SH2-domain-binding inhibitors could be designed to inhibit a faulty signal transduction pathway has in the last 15 years justified enormous investments by the pharmaceutical industry.12.13

8.2 SH2 domains and human pathologies Only a few SH2 domains have been directly implicated in human diseases. Mutations in the SH2 domain of SAP (or SLAM (signalling lymphocyte activation molecule) associated protein, also known as SH2D1A or DSHP) cause X-linked lymphoproliferative disease (XLP).14–16. This disorder is characterized by an extreme sensitivity to Epstein–Barr virus, with infection resulting in uncontrolled B cell proliferation. The average age of onset in affected young boys is 2.5 years, with 75 per cent mortality by age 10 and 100 per cent mortality by age 40. SAP functions in immune cell signalling, with its expression restricted primarily to T cells. It interacts with the protein SLAM, a 70 kDa glycoprotein expressed on many lymphocyte cell surfaces. SLAM is a self-ligand type of receptor that operates in bi-directional signalling between B and T cells, perhaps functioning in the switching of helper T cell phenotype.

SH2 DOMAINS AND HUMAN PATHOLOGIES

153

A point mutation in the SH2 domain of the Bruton tyrosine kinase has been shown to cause cases of X-linked agammaglobulinaemia. X-linked agammaglobulinaemia is the prototypical humoral immunodeficiency first described by Bruton in 1952.17 It is characterized by a paucity of circulating B cells and a drastic reduction in the serum concentrations of immunoglobulins. Studies analysing patterns of X chromosome inactivation showed that the genetic defect was intrinsic to the B cell lineage and mapping studies located the defect in the midportion of the long arm of the X chromosome at Xq225. Subsequent studies demonstrated that mutations in the cytoplasmic tyrosine kinase gene Btk (the gene for Bruton’s tyrosine kinase) are responsible for X-linked agammaglobulinaemia. Like other cytoplasmic tyrosine kinases, Bruton’s tyrosine kinase (Btk) contains a unique amino-terminal region, SH3 and SH2 domains (SH3 domains are homology domains that specifically bind to polyproline sequences), and a carboxy-terminal kinase domain. Saffran and co-workers17 identified a mutation (at residue 361) in the SH2 domain of Btk that causes the disease and that appears to destabilize the full-length protein or interfere directly with SH2 domain function. Finally, mutations in the N-terminal SH2 domain of the SHP2 phosphatase have been identified as causing a relatively common autosomal dominant disorder called Noonan’s syndrome.18 This disease is characterized by dysmorphic facial features, short stature, and heart disease. Webbed neck, chest deformity, and mental retardation are also frequently associated with this disease. Mis-sense mutations in PTPN1, a gene that encodes the nonreceptor protein tyrosine phosphatase SHP-2, which contains two SH2 domains, cause Noonan syndrome and account for more than 50 per cent of the cases examined. All mutations locate in the interacting portions of the amino-terminal SH2 domain, which is involved in down-regulating phosphatase activity. Thus, the disease state may result from excessive phosphatase activity. Although examples of direct implication of SH2 domains in diseases are few, their involvement in pathways altered by disease-causing mutations make them attractive targets for pharmaceuticals. Table 8.1 summarizes SH2 domain potentials as drug targets.19 As can be seen, a large array of pathological conditions can be targeted by SH2-domain-binding inhibitors, ranging from cancer, allergy, and AIDS to osteoporosis. The SH2 domains of the Grb2 adaptor, the Src kinase and Src kinase family of proteins (Hck, Lyn, Lck for example), the z-chain-associated protein kinase ZAP70, the p85 subunit of the phosphatidylinositol 3 kinase (PI3K), and the Syk kinase have been and still are the subjects of extensive drug design effort in the pharmaceutical industry and academic research for potential uses in cancer (Grb2, PI3K), autoimmune diseases (ZAP70), allergy and asthma (Syk), and osteoporosis (Src). Below, details of the cellular biology in which Src, Syk, and Grb2 are involved are provided.

154

THERMODYNAMICS OF SH2 DOMAIN BINDING

Table 8.1

Potential of SH2 domains as pharmaceutical targets19

Pathology

SH2 domain-containing target

AIDS Allergy and asthma Anaemia Autoimmune disease Breast cancer Cancer CML and ALL Eurythroleukaemia Inflammatory disease Prc-B cell leukaemia Myelodysplastic syndrome Osteoporosis

Lck, Hck Syk, Lyn SH-PTP1 Zap-70 Grb2, Grb7, Src p85, Shc, Grb2, Gap Grb2, CrkL Shc STATs Btk Tec Src

The Src kinase consists of four domains, a short myristylation domain, which locates Src to the membrane, an SH3 domain, which binds polyprolinerich sequences, an SH2 domain, and a kinase domain (Figure 8.1).20 Src has been shown to be involved in a variety of cellular processes including bone resorption, tumour growth and angiogenesis, and tumour metastasis.20 However, the major phenotype of an Src-knockout mutant mouse is an excessive accumulation of bone in marrow cavities during bone development due to excessive bone resorption (osteopetrosis).21 A more severe form of osteopetrosis is observed in the double Src, Hck knockout mice.22 Osteoclasts are the cells specialized in bone resorption and osteoclasts derived from Src-deficient mice are viable; however, they fail to form ruffled borders, and thus are unable to secrete acid proteases and bone-metabolizing enzymes.23,24 Src-deficient osteoblasts, the cells that produce bones, in contrast, appear to have increased activity,25 thus possibly contributing to the osteopetrotic phenotype observed in Src-deficient mice. Partial rescue of the osteopetrosis phenotype in Src-deficient mice by transgenic expression of the kinasedeficient K295M Src mutant suggests that the role of Src in osteoclasts is

Figure 8.1 Primary structure of the Src and Syk protein tyrosine kinases

STRUCTURE AND LIGAND BINDING

155

mediated by its protein–protein SH2 or SH3 interaction domains.26 Thus, an SH2-domain-binding inhibitor of the Src kinase could potentially inhibit bone resorption and serve as a treatment of osteoporosis, a condition that results in weakening of the bone structure.27 The Syk kinase consists of two SH2 domains in tandem (collectively named the ‘tandem-SH2 domain’) and a kinase domain (Figure 8.1).28 Syk is a member of the Syk family of cytosolic protein tyrosine kinases that also includes the ZAP-70 kinase.29,30 Syk is expressed ubiquitously in haematopoietic cells such as B cells, mast cells, polymorphonuclear leukocytes, platelets, macrophages, and immature T cells. Syk is also present in nonhaematopoietic cells such as epithelial cells, hepatocytes, fibroblasts, and neuronal cells. The absence of either Syk family kinase results in arrested T and B cell development and functional defects for a variety of immune receptors including the T cell receptor, the B cell receptor, and receptors for IgG and IgE.31–37 In addition, Syk family kinases have been implicated in activating NK cells,38,39 and signalling by non-immune receptors such as cytokine, thrombin, integrin, and G-protein coupled receptors.40–45 The involvement of Syk in immune cell biology suggests that SH2 domain inhibitors targeting the tandem SH2 domain of this enzyme could be used for treatment of allergy. Grb2 is a small adaptor protein consisting of one SH2 domain flanked by two SH3 domains.46 The function of the Grb2 was elucidated in a series of outstanding publications, where it was reported that Grb2 serves to link receptor-tyrosine kinases to Ras signalling.46–50 The SH2 domain of Grb2 binds to specific phosphorylated tyrosine residues on activated receptors while its SH3 domains bind to the guanine nucleotide exchange factor specific to the low-molecular-weight GTP-binding protein Ras, thereby activating Ras and the downstream Ras-dependent MAP kinases. As Ras is known to play a key role in the regulation of cellular growth and differentiation, one means to control Ras is through indirect inhibition of its activation by potential SH2domain-binding inhibition of its Grb2 activator. The effort expanded in designing efficacious Src and Grb2 SH2-domainbinding inhibitors has been met with some measure of success and will be described in the section ‘SH2-domain-binding inhibitors’ below.

8.3 Structure and ligand binding Because of their importance in some of the most essential biological processes in humans, SH2 domains have been the subject of extensive investigations aimed at understanding the structure/function relationship or/and designing SH2-domain-binding inhibitors.

156

THERMODYNAMICS OF SH2 DOMAIN BINDING

The structures of a large number of SH2 domains have been determined (reviewed in reference 51). These studies have revealed a common fold consisting of a central b-sheet flanked by two a-helices. Figure 8.2(a) shows the structure of the Src SH2 domain bound to the pYEEI peptide (see definition below). The SH2 domain is in ribbon representation. The peptide ligand pYEEI is shown in stick representation. Notations of secondary structures are according to references 60 and 61. A remarkable feature of the structure, which was revealed by the determination of the first structure of a binary complex of the Src SH2 domain bound to a tyrosine phosphorylated peptide, is the definition of two regions, each with a distinct functional role: the region between the first a-helix and the b-sheet is the site of phosphotyrosine recognition while the region between the b-sheet and the C-terminal helix forms the so-called ‘specificity-determining region’ as it forms the interacting surface for the residues C-terminal to the phosphotyrosine.52

Phosphotyrosine binding In vivo and in vitro studies have shown that phosphorylation of tyrosine residues is a requirement for binding (reviewed in reference 53). However, exactly how stringent the requirement for phosphate is has only recently been rigorously evaluated.54 This study showed that more than half of the binding free energy of the interaction of the SH2 domain of the Src kinase (Src SH2 domain) with its cognate tyrosyl phosphopeptide target (sequence pYEEI, where pY indicates phosphotyrosine; see below for more details) is contributed by phosphorylation. Other studies have shown that no other residue in the tyrosine phosphorylated ligand is nearly as important.55 Thus, as suspected and now rigorously determined, the phosphate group is by far the most essential structural requirement for binding. One exception is the SH2 domain of SAP, which appears to be much less phosphorylation dependent (see below). The region that binds phosphotyrosine forms a deep positively charged pocket, the so-called ‘phosphotyrosine (pY or pTyr or 0-pTyr)-binding pocket’.52 A vast array of interactions is observed between the pocket and the phosphotyrosine. They include a salt bridge between the phosphate and an arginine residue, ArgbB5, an amino-aromatic interaction between the guanidinium group of another arginine, ArgaA2, and the aromatic ring of the phosphotyrosine, and numerous hydrogen bonds contributed by residues in the BC loop (see notation for secondary structures in Figure 8.2(a)). Remarkably, studies by Bradshaw et al.56 have shown that, of the myriad of interactions the SH2 domain makes with the phosphate, only one contributes significant binding free energy: that of ArgbB5 with the phosphate.54 Figure 8.2(b) shows interactions at the pTyr-binding pocket. The pYEEI peptide is

STRUCTURE AND LIGAND BINDING

157

Figure 8.2 (a) Structure of the SH2 domain of the Src kinase bound to pYEEI peptide. (b) Interactions at the pTyr-binding pocket. (c) Interactions with residues in the ‘specificitydetermining region’

158

THERMODYNAMICS OF SH2 DOMAIN BINDING

shown in dark ball-and-stick representation. Residues in the SH2 domains are shown in light ball-and-stick representation. The binding interface in this region of the structure is a classical hot-spot interface where only one interaction appears to be critical.57 Interestingly, the SH2 domain of Src also contains a cysteine in its pY-binding pocket, the role of which remains obscure, as mutation to Ala or Ser of this residue increases binding affinity.54

Peptide-binding specificity of single SH2 domains First evidence that SH2-domain-containing proteins are recruited specifically to certain tyrosine phosphorylated sites on activated receptors emerged in early work on such proteins (reviewed in reference 53). Subsequent studies defined the residues C-terminal to the phosphotyrosine as being specifically involved in providing discrimination between phosphorylation sites.58,59 These studies defined binding motifs specific to classes of SH2 domains. For example, the SH2 domain of Src kinase family members have a binding preference for the motif pYEEI, i.e. two glutamates at the þ1 and þ2 position C-terminal to the phosphotyrosine followed by Ile at the þ3 position. The Nterminal SH2 domain of the p85 subunit of PI3 kinase has a distinct binding motif, pYXXM, where X may be any amino acid: thus, the þ3 position defines the selective recruitment of these SH2 domains. The SH2 domain of Grb2 distinguishes itself by requiring a Asn at the þ2 position, while the þ1 and þ3 positions appear to play minor roles.59 The structural characterization of selective peptide binding was first approached by the crystallization of Src family member SH2 domains with the pYEEI motif.60,61 A central feature of the contact interface for the +1Glu, +2Glu, and +3Ile residues is a deep hydrophobic pocket formed by residues of the central b-sheet, the C-terminal helix aB, and the EF and BG loops (Figure 8.2(c); Representations of residues in the peptide and in the SH2 domain are as in (b)). The þ3Ile of the pYEEI motif fits snugly into that pocket. The N- and C-terminal SH2 domains of the p85 subunit of the PI3 kinase are also characterized by a well defined þ3 binding pocket where the þ3Met of the pYXXM cognate motif inserts.62,63 However, such a feature is not universal. The structure of the C-terminal SH2 domain of phospholipase Cg in complex with a tyrosyl phosphopeptide shows a peptide binding interface that is extended to the residue þ5 positions C-terminal to the pTyr.64 A similar extended binding interface is seen in the N-SHP-2 SH2 domain.65 It might have been expected that the interactions between residues throughout this extended interface would be critical for high-affinity binding. However, binding experiments using a series of truncated peptides have indicated that only the residue immediately C-terminal to the pTyr is required to maintain the majority of the binding energy of the C-PLCg SH2-domain-binding

STRUCTURE AND LIGAND BINDING

159

interaction.66 Thus, the majority of interactions seen in the structure contribute little to binding. Interestingly, the side-chains which are of less importance for binding show greater mobility in the bound state than the residues of the pY-binding pocket, which are essential for binding.66,67 The fact that many interactions observed in the crystal or NMR structures appear to be of little importance is emphasized by a study aimed at assessing the importance of the þ1, þ2, and þ3 positions of the pYEEI motif.55 Analysis of peptide binding revealed that conservative substitutions at each peptide position induce only very minor, threefold or less, decreases in affinity compared with the wild-type pYEEI peptide. It was also found that the effect of alanine substitutions is generally small; the pYAEI, pYEAI, and pYEEA peptides bound only twofold, sixfold, and tenfold, respectively, weaker than the pYEEI peptide. A mutational investigation of the Src SH2 domain itself has probed the importance of the residues in the SH2 domain responsible for coordinating the EEI motif (the so-called specificity-determining residues) and determined that the majority of these residues make only small energetic contributions to binding.56 In this investigation, individual mutation to alanine of six SH2 domain residues (Arg bD’1, Leu BG4, Ile bE4, Thr EF1, Arg EF3, and Asp BG2) caused a threefold or less reduction in binding affinity. However, mutation of Tyr bD5 (Figure 8.2(c)), the residue that forms a platform for the peptide backbone and contributes part of the hydrophobic þ3 binding pocket as well as interaction with the Cb of the þ1 Glu, induced a greater than tenfold loss in affinity upon mutation. Also, mutation of Lys bD3, the residue which coordinates the þ1 Glu of the peptide, caused a sevenfold reduction in affinity. Interestingly, subsequent studies showed that the effect of LysbD3 mutation to Ala is actually caused by the indirect effect of two aspartate residues which are not part of the interface.68 Thus, LysbD3’s role appears to have nothing to do with peptide binding but rather to be to neutralize the negative charges emanating from other regions of the structure. Thus, based on studies of the pYEEI-motif binding by the Src SH2 domain, single SH2 domains appear to be unable to discriminate radically between tyrosyl phosphorylated ligands. However, the pYEEI motif is not a physiological substrate for the Src SH2 domain. Instead, one established binding site for Src recruitment is that located on the PDGF receptor.20,69–71 The site has been mapped to two tyrosines, Tyr 579 and 581 in the juxtamembrane region of PDGF receptor b and Tyr 572 and 574 of the areceptor isoform.72,73 Although phosphorylation of the second tyrosine is not essential for binding of the Src SH2 domain, it has been shown that tyrosine phosphorylation at the þ2 position considerably enhances the free energy window for specific interaction.74 In fact, the predominant contributors to binding free energy are no longer the 0 pTyr and the þ3 position but the 0 pTyr and the þ2 pTyr. Thus, in the physiologically relevant interaction, the two residues anchoring the ligand to the peptide binding interface are the 0

160

THERMODYNAMICS OF SH2 DOMAIN BINDING

pTyr and the þ2 position.74 One anchor, the 0 pTyr, inserts into the pYbinding pocket; the second anchor, the þ2 pTyr, lies on a flat surface between the 0 pTyr and þ3 binding pockets. How þ2 pTyr recognition occurs remains a puzzle: the crystal structure of the Src SH2 domain bound to its binding site on the PDGF receptor and further mutation studies showed that (1) an arginine nearby ideally positioned to form a salt bridge with the phosphate group plays no role and (2) most of the interactions between þ2 pTyr and the SH2 domain are mediated by water molecules.74–76 In the physiological substrate, the predominant role that 0 pTyr and þ2 pTyr play is further enhanced by the presence of residues at the þ1 and þ3 positions that are less than ideal contributors to binding, a þ1 Ile and a þ3 Val (the PDGF receptor binding motif is pYIpYV). For the most part, the specificity of SH2 domain binding is determined by the residues of the target C-terminal to the 0 pTyr. However, for some SH2 domains, the residues N-terminal to the pTyr can be important. For example, it has been determined that the N-SHP-2 SH2 domain has a requirement for a hydrophobic residue (Val, Ile, or Leu) at the position
2 (second position Nterminal to 0 pTyr) to the pTyr.77,78 This unusual requirement has been traced to the presence of a Gly at the aA2 position of the SH2 domain, a position that an Arg occupies in 80 per cent of known SH2 domains.78,79 Another example of
2 position requirement is that of the SH2 domain of SAP: however, in this case, such a requirement appears to be the basis for phosphorylation-independent interaction. The SAP SH2 domain is unusual in that it can bind targets in a phosphate-independent manner. The first hint at this non-canonical behaviour was that SAP–SLAM binding could be detected in a yeast two-hybrid screen.16 A peptide surrounding tyrosine 281 of SLAM has been found to immunoprecipitate SAP independently of whether Tyr 281 is phosphorylated, unphosphorylated, or mutated to phenylalanine. Fluorescence polarization binding studies have confirmed that tyrosine phosphorylation of the SLAM peptide is not absolutely required for SAP SH2 domain binding, although it does somewhat increase binding affinity.80 This investigation has also identified that SAP binding requires residues both N-terminal and C-terminal to Tyr 281 in its SLAM target. The structures of SAP in its unliganded form, as well as bound to both tyrosine phosphorylated and unphosphorylated peptides, have shown that the residues in the peptide N-terminal to Tyr 281 make specific contacts with SAP.81 Figure 8.3(a) shows the structure of the SH2 domain bound to the SLAM peptide. The SH2 domain is in ribbon representation. The peptide ligand (motif TIpYAQVQK) is shown in stick representation. Labelling draws attention to the
1 and
2 interactions that appear to be responsible for the ability of the SAP SH2 domain to bind the unphosphorylated SLAM motif. These interactions explain the requirement of the SAP SH2 domain for residues N-terminal to the pTyr.

STRUCTURE AND LIGAND BINDING

161

Figure 8.3 (a) Structure of the SAP SH2 domain bound to its cognate SLAM peptide. (b) Structure of the Grb2 SH2 domain bound to its cognate pYxNx motif

The Grb2 SH2 domain is relatively unique in having a strong requirement for a residue (Asn) at the position þ2 C-terminal to the 0 pTyr.59 A thermodynamic and mutational analysis of Grb2 SH2 domain binding has underscored the importance of this þ2 Asn for high-affinity tyrosyl phosphopeptide binding. Here it was demonstrated that substitution of the þ2 Asn to Ala in a high-affinity Grb2 tyrosyl phosphopeptide decreased the binding affinity by three orders of magnitude.82 Alanine mutation of the other residues had less than an order of magnitude effect on binding. The þ2 Asn requirement is caused by a Trp residue at position EF1 of the SH2 domain of Grb2 which not only fills a potential þ3 binding pocket but also redirects the peptide ligand at the þ2 position (labelled in Figure 8.3(b)).83 Figure 8.3(b) shows the structure of the SH2 domain bound to the pYxNx motif. The SH2 domain and peptide ligand (motif pYxNx) are shown as in (a). Mutation of the corresponding residue in the Src SH2 domain to Trp dramatically changes the binding properties of this SH2 domain and makes it remarkably similar to the Grb2 SH2 domain.84 The findings described above indicate that interactions between the 0 pTyr in the target and its binding pocket in the SH2 domain are those most crucial for the molecular recognition of SH2 domains; the interactions between other residues of the target and the specificity determining residues of the SH2 domain appear to be secondary in importance in terms of SH2 domain

162

THERMODYNAMICS OF SH2 DOMAIN BINDING

binding affinity. Furthermore, the level of specificity exhibited by SH2 domains, particularly those of the Src family, is modest. These results have profound implications for the mechanism by which SH2 domains specifically engage their targets within cells. In particular, they suggest that initial models of SH2 domain recognition, where single SH2 domains are able to selectively recognize their particular binding sites due only to a distinct set of interactions between the SH2 domain and the target, may be overly simplistic. Rather, a single SH2 domain alone may not contain all the required elements for selective recognition of targets. How, if the interactions involving the specificity determining residues are weak, can SH2 domains be targeted to specific pTyr containing sites within activated receptors? For single SH2-domain-containing proteins, specific recruitment may be facilitated by other domains. For example, Nmyristoylation of the Src kinase – and hence its association with the cell membrane – may facilitate the interaction of the Src SH2 domain with the most membrane proximal pTyr of the PDGF receptor b. Another possibility stems from the fact that many SH2 domain-containing proteins do not contain just a single SH2 domain, but rather SH2 domains located in tandem (Zap-70, Syk, PI-3’kinase, PLCg, etc.) or SH2 domains combined with other domains such as SH3 domains. For these proteins, the concerted binding of multiple domains probably enhances overall specificity. In the case of tandem SH2 domain binding, this would be achieved for two reasons. First, the binding of two SH2 domains may be positively coupled, which would result in an increase in binding specificity. Second, improper targeting to tyrosine phosphorylated sites may be reduced by the topological constraints imposed by the strict spacing between pTyr residues. Hence, although SH2 domains appear to have evolved only low levels of specificity, specific targeting of SH2domain-containing proteins to signalling complexes may be achieved through cooperative interactions between these domains and other partners in the signalling complex.

Peptide-binding specificity of tandem SH2 domains Syk family kinases contain two Src homology 2 (SH2) domains positioned in tandem (termed collectively as ‘tandem SH2 domain’).28 Syk family kinases are recruited to cell surface receptors through the interaction of their tandem SH2 domains with tyrosine phosphorylated sequence motifs termed ITAMs (immunoreceptor tyrosine-based activation motifs).85 ITAMs typically have the consensus sequence Yxx(L/I)–x7/8–Yxx(L/I) i.e. two tyrosines spaced by a 10/11 residue sequence and containing a hydrophobic residue (L or I) at the third position C-terminal to the phosphotyrosines.86 Remarkably, the tandem SH2 domain of Syk appears to be able to recognize a variety of doubly

STRUCTURE AND LIGAND BINDING

163

phosphorylated (dp) ITAMs, which vary considerably not only in sequence but also in the length of the spacer region between the two phosphotyrosines.40,87,88 For example, the g-chain of the FceRI receptor, an Fc receptor for IgE, contains an ITAM with a spacer region of 10 residues while the spacer region of the ITAM on the single-chain Fc receptor class IIA, an Fc receptor to IgG, is 15 residues in length.89 Their sequences are also considerably different. Yet, doubly phosphorylated peptides derived from these dpITAMs bind equally well and with high affinity (2–4 nM) to the tandem SH2 domain of the Syk kinase.88 The mechanism by which the tandem SH2 domain of Syk can accommodate such structural variability in the ITAM was hinted at when the crystal structure of a complex of the Syk tSH2 domain bound to a dpITAM peptide derived from the CD3-e chain of the T cell receptor (Kd ¼ 20 nM) was determined.90 The crystals contained 6 complexes in the asymmetric unit. Structural alignment of the six complexes revealed substantial flexibility in the relative orientation of the two SH2 domains. Figure 8.4(a) shows the flexibility in the relative orientation of the two SH2 domains. The extreme orientations of the SH2 domains in the tandem SH2 domain bound to the CD3-e dpITAM are shown in light and mid-grey. One intermediate is shown in dark grey. The orientations correspond to rigid body motion around the linker region between the two SH2 domains. This feature, coupled with the fact that the interface between the SH2 domains is relatively small, suggested that, in the unbound state, there may even be larger fluctuations in the relative orientations of the two SH2 domains: the Syk tandem SH2 domain may fluctuate between an ‘open’ state, where the two SH2 domains are far apart from each other, and a ‘closed’ state, where they are closer to one another. This hypothesis was supported by the results of a multifaceted biophysical

Figure 8.4 (a) Structural flexibility in the relative orientation of the two SH2 domains in the tandem SH2 domain of the Syk kinase. (b) Temperature dependence of the binding enthalpy of the Syk tandem SH2 domain–CD3-e dpITAM interaction

164

THERMODYNAMICS OF SH2 DOMAIN BINDING

study of ITAM binding to the Syk tandem SH2 domain in solution.91 In this study, an unusual non-linear temperature dependence of the binding enthalpy (i.e a temperature-dependent heat capacity change (DCp8)) was observed, which was interpreted as the thermodynamic signature of a pre-existing conformational equilibrium (between a closed and an open form) in the unbound state. Figure 8.4(b) shows the temperature dependence. The heat of binding is monitored as a function of temperature. The curvature obtained with the wild type protein (open diamonds) is the thermodynamic signature of a conformational equilibrium which is interpreted as taking place between a closed form where the SH2 domains are close to one another and an open form where the SH2 domains are apart. The additional heat observed at higher temperature is the result of not only binding but also of the conformational transition from open to closed (see reference 91 for details). These results provided a model for the ability of the Syk tandem SH2 domain to recognize dpITAMs with spacer regions of variable length: the two SH2 domains sample a continuum of relative orientations between an open and closed state, thus allowing the tandem SH2 domain to adapt and adjust to the various lengths of the inter-phosphotyrosine region. Although this model was consistent with the complexity of the binding data, direct confirmation was lacking. This model can be tested by restricting the conformational ensemble accessible to the tandem SH2 domain in the unbound state. This was achieved by introducing a disulfide bridge at the interface between the two SH2 domains (S. Kumaran and G. Waksman, unpublished). Remarkably, this double cysteine (2Cys) mutant displays all the binding properties that are expected from a Syk tandem SH2 domain locked into a closed conformation, i.e. the two SH2 domains locked close to one another and unable to open up: (a) the 2Cys mutant displays a temperature-independent heat capacity change (DCp8), which reflects a single-species population as opposed to the previously observed temperature-dependent DCp8 (Figure 8.4(b) closed circle), reflecting a multi-conformer equilibrium (Figure 8.4(b)), and (b) the affinity of this 2Cys mutant for a dpITAM sequence with a short spacer region is less affected than that for a dpITAM with a longer spacer region. Does the Zap-70 tandem SH2 domain also exhibit significant conformational flexibility? The orientation of the SH2 domains, as well as the inter-SH2 linker, in the Zap-70 structure suggests that a change in conformation of this module might accompany binding.92 Indeed, a change in the Zap-70 tandem SH2 domain conformation upon dpITAM binding has been characterized using sedimentation velocity analytical ultracentrifugation.93 Here a reduction in the frictional coefficient of the Zap-70 tandem SH2 domain is seen upon dpITAM binding, suggesting that the complex adopts a more compact structure than the unligated protein. Furthermore, a study that has used antibodies directed against the inter-SH2 domain region of Zap-70 has also

STRUCTURE AND LIGAND BINDING

165

detected conformational changes in the tandem SH2 domain upon binding to the z-chain of the T-cell receptor.94 Here, the inter-SH2 domain region was far more accessible in the tandem SH2 domain as compared to the full-length Zap-70 molecule; furthermore, the accessibility of the inter-SH2 domain region was significantly reduced upon dpITAM ligation. Finally, a thermodynamic study that has compared and contrasted the recognition properties of the tandem SH2 domains of Zap-70 and the p85 subunit of PI-3’kinase has also suggested that a conformational change occurs upon binding of the Zap70 tandem SH2 domain.95 It is not clear at present whether tandem SH2 domains demonstrate higher levels of sequence specificity. Grucza and colleagues have shown that Syk tandem SH2 domain binding to the non-cognate CD3-edpITAM peptide is only 20-fold weaker than binding to the biologically relevant (and similarly spaced) dpITAM derived from the g-chain of the FcgRI receptor.88 Also, mutations at the þ1, þ2, or þ3 positions C-terminal to both the N- and Cterminal pY residues do not dramatically affect binding to the Syk tandem SH2 domain (G. Waksman and R.A. Grucza, unpublished). Similarly, modest differences in affinity were observed when binding of the Syk and Zap-70 tandem SH2 domains to various dpITAM-based peptides was tested.96 However, when doubly phosphorylated and similarly spaced peptides derived from the PDGF receptor were tested, binding was found to be three to four orders of magnitude weaker.97 Therefore, it would appear from this example that SH2 domains in tandem may exhibit higher levels of sequence specificity than single SH2 domains. As discussed above, the inter-pY spacing seems to contribute only modestly to peptide-binding specificity, as Syk binding of the FceRI dpITAM peptide, which contains ten residues between the pTyr residues, is as tight as that of a dpITAM peptide based on the FceRIIA receptor, a class II Fc receptor to IgG, which contains an additional five residues in the inter-pTyr region.88 Consistent with these results, Ottinger et al.96 showed that a peptide based on the sequence of SHPS (a substrate for the SHP-2 phosphatase) that contains 13 additional residues in the inter-pTyr region binds with only a 10–20-fold weaker affinity to both Syk and Zap-70 tandem SH2 domains than a peptide containing only 10 residues in the inter-pTyr region. Depending on the degree of receptor activation, ITAM sequences within the cell can be either unphosphorylated, singly phosphorylated (mpITAM), or doubly phosphorylated. Hence, an additional issue regarding the specificity of tandem SH2 domain binding is the preferential binding of doubly phosphorylated, as opposed to singly or unphosphorylated, ITAM sequences. It has been experimentally observed that the Syk tandem SH2 domain displays dramatic differences in affinity between mpITAMs and dpITAMs.88 For instance, the Syk tandem SH2 domain binds to the FceRI ITAM with an affinity of about 2 nM whereas affinities are about 1 mM and 10 mM for the

166

THERMODYNAMICS OF SH2 DOMAIN BINDING

corresponding ITAM with the N-terminal or C-terminal phosphate removed, respectively; loss of one phosphate of the ITAM causes about a 1000-fold loss in affinity.88 This difference is similar to that observed for the binding of the Src SH2 domain to a phosphorylated versus dephosphorylated peptide.54 Hence, tandem SH2 domains are highly specific for doubly phosphorylated sequences.

8.4 SH2-domain-binding inhibitors As discussed above, the central role of SH2 domains in many tyrosine kinase signalling pathways makes them at least potentially attractive targets of pharmaceuticals,12,13 and hence there have been significant efforts aimed at discovering SH2-domain-binding inhibitors.

Difficulties in targeting pharmaceuticals to SH2 domains Many of the studies of SH2 domain binding described here have revealed significant obstacles, which complicate the discovery of SH2-domain-targeted pharmaceuticals. For instance, mutational studies have indicated that the SH2-domain-binding mechanism is such that a charged interaction in the pTyr binding pocket, that between ArgbB5 and the phosphate of the pTyr, is most crucial for binding. Hence, charged ligands would be optimal to inhibit SH2 domain–target interactions. Unfortunately, highly charged ligands frequently have difficulty penetrating the cell membrane, and hence charged interactions are difficult to target with pharmaceuticals. Furthermore, the primacy of the ArgbB5–phosphate interaction in SH2 domain binding also suggests that targeting particular pharmaceuticals solely to one particular SH2 domain or class of SH2 domains will be difficult. This will especially be true if the ArgbB5–phosphate interaction is as crucial for binding in other SH2 domains as it is in the Src SH2 domain (which it is likely to be, given the universal conservation of ArgbB5). Despite these difficulties, significant progress has been made in discovering pharmaceuticals that target particular SH2 domains.

Src SH2-domain-binding inhibitors As described earlier, the Src SH2 domain has a cysteine residue (CysbC3) located within its pTyr binding cavity that is unique to Src; other SH2 domains have either a serine, threonine, valine, or alanine at this position. Cysteine residues such as CysbC3 are potentially good targets for pharmaceuticals due to their nucleophilic character: these residues can in

SH2-DOMAIN-BINDING INHIBITORS

167

certain cases be induced to form covalent bonds with aldehyde groups on pharmaceutical targets, as has been shown for cysteine proteases.97 In particular, CysbC3 might be particularly reactive due to the fact that the positively charged environment of the pTyr binding cavity lowers its pKa .98 It has indeed been shown using x-ray crystallography and titration microcalorimetry that CysbC3 can form a covalent bond with aldehyde groups on pharmaceutical-like compounds, implying that it might be possible to selectively target the Src SH2 domain with specific compounds in vivo.99,100 This possibility has been further advanced by studies that have examined the effect of a CysbC3 targeting compound in living cells. As indicated above, mice lacking the src gene demonstrate osteopetrosis (thickened bones), indicating that Src plays a role in bone formation and suggesting that Src might be a useful target for treatment of osteoporosis. A compound that targets CysbC3 of the Src SH2 domain has been shown to preferentially bind to the Src SH2 domain in vitro.101 This compound inhibits Src SH2 domain– phosphopeptide interactions within cells and also diminishes the resorptive activity of rabbit osteoclasts.101 More recently, second-generation compounds have been designed that appear to have been inspired by the crystal structure of a citrate-bound Src SH2 domain complex.102 In this structure, citrate appears to make hydrogenbonding interactions with residues that had not previously been exploited for design of phosphotyrosine mimics.103 Incorporation of diphosphonomethyl phenylalanine into the design of inhibitory peptides targeting the Src SH2 domain led to the discovery of another potent active molecule exhibiting bone-targeting properties and anti-resorptive activity in a cell-based assay utilizing osteoclasts and bone.103

Grb2 SH2-domain-binding inhibitors The SH2 domain of Grb2 is perhaps the most selective SH2 domain in terms of target recognition: there is a strict requirement for Asn two residues Cterminal to the pTyr.59,82 Furthermore, only in the Grb2 SH2 domain structure does the peptide bind in a b-turn conformation to accommodate interactions between the þ2 Asn and the protein.83 These findings may partially explain why non-tyrosine phosphorylated peptidomimetics have been successfully designed and reported for the Grb2 SH2 domain, but no other SH2 domain.104 These non-phosphorylated peptide inhibitors are very desirable given that the negatively charged phosphate group of phosphopeptides is a detriment to the cell permeability of pharmaceutical compounds. For instance, a cyclic, 11-residue non-phosphorylated peptide that contains the crucial Asn has been shown to bind specifically and with micromolar affinity to the Grb2 SH2 domain.104 Other non-cyclic inhibitors of the Grb2 SH2

168

THERMODYNAMICS OF SH2 DOMAIN BINDING

domain have also been found. Studies with these compounds have highlighted how Grb2 couples activation of the Ras signalling pathway to cellular transformation.105 Furthermore, Grb2 inhibitors have also demonstrated that this protein plays a role in the motility of tumour cells.106 These results have highlighted how inhibition of the Grb2 SH2 domain could be an effective anticancer treatment. Further readings on SH2-domain-binding inhibitors may be found in the reviews in references 27 and 107–111.

8.5 Conclusions Studies of SH2 domains have revealed their mode of action. Clearly, the most important binding property of SH2 domains is their selectivity for the phosphorylation state of their targets. Other recognition processes appear to play only minor roles. As receptor activation induces phosphorylation of numerous tyrosine residues, inducing the recruitment of a number of SH2containing proteins, one would have expected some degree of specificity apparently essential to prevent signal transduction pathways from crossing. However, based on binding data, a back-of-the-envelope calculation of the degree of occupancy of any particular tyrosine phosphorylated site by the Src SH2 domain shows that all sites would be at least partially occupied by this SH2 domain regardless of the sequence surrounding the site. Thus, a certain level of cross-talk between signal transduction pathways appears to be happening. Such cross-talk was actually revealed in a study that used genearray technology to monitor genes induced by phosphorylation of any particular tyrosine of the PDGF receptor upon activation by PDGFb.112,113 Using global expression monitoring, the relationship between receptor tyrosine kinase (RTK) activated signalling pathways and the transcriptional induction of immediate early genes (IEGs) was investigated. 66 fibroblast IEGs induced by platelet-derived growth factor b-receptor signalling were identified. Mutant receptors lacking binding sites for activation of the PLCg, PI3K, SHP2, and RasGAP pathways still retain partial ability to induce 64 of these IEGs. Removal of the Grb2-binding site further broadly reduced induction. These results suggest that the diverse pathways exert broadly overlapping effects on IEG induction. The fact that SH2 domains are involved in multiple cellular processes, some of which may be altered in diseases, has made them obvious targets for pharmaceuticals. Unfortunately, the enormous investment that the pharmaceutical industry has made in the last 10–15 years towards finding suitable SH2-domain-binding inhibitors has not been met with the anticipated success. Part of the problem is that the effort was not guided by a deep understanding of the relative importance of the various constituents of the phosphorylated

REFERENCES

169

sites that SH2 domains target. Although it had been realized early in the SH2 domain research that phosphorylation was an important determinant of binding affinity, the extent of the predominance of the phosphorylated tyrosine in binding versus any other residues in the SH2 domain recognition sites was revealed only recently. It became then apparent that (1) specificity in the targeting of the drug would be difficult to achieve as a phosphotyrosine derivative would need to be present, thus running the risk of targeting the drug to SH2-domain-containing proteins that play no role in the diseased state targeted for therapy, and (2) the large electrostatic field emanating from such an essential drug component would present an obstacle for the entire molecule in penetrating the hydrophobic environment of the cell membrane, and thus delivery to the cell may be prevented.114 Whether a drug targeting an SH2 domain will appear on the market is still uncertain. However, the search for an efficacious and viable SH2-domain-binding inhibitor is actively continuing. Added specificity may be obtained by targeting sites that combined the recruitment of more than one SH2 domain, such as the tandem SH2 domain of the ZAP-70 and Syk kinases. A binding inhibitor targeting the two pYbinding pockets in each SH2 domain with no other sequence specificity requirement would probably bind predominantly dpITAM-recognition domains, thereby possibly providing efficacious treatments of some immune diseases. In this regard, the details emerging on the binding properties of tandem SH2 domains will greatly facilitate the development of a successful strategy for designing binding inhibitors specific to the targeted immune receptor-dependent pathways.

References 1. Bostro¨m H, Willetts K, Pekny M, Leve´en P, Lindahl P, Hedstrand H, Pekna M, Hellstro¨m M, Gebre-Medhin S, Schalling S. et al. (1996) Cell 85: 863–873. 2. Leve´en P, Pekny M, Gebre-Medhin S, Swolin B, Larsson E and Betsholtz C. (1994) Genes Dev. 8: 1875–1887. 3. Soriano P. (1994) Genes Dev. 8: 1888–1896. 4. Soriano P. (1997) Development 124: 2691–2700. 5. Heldin C-H and Westermark B. (1999) Physiol. Rev. 79: 1283–1316. 6. Claesson-Welsh L, Eriksson A, Westermark B and Heldin C-H. (1989) Proc. Natl Acad. Sci. USA 86: 4917–4921. 7. Matsui T, Heidaran M, Miki T, Toru M, Popescu N, La Rochelle W, Kraus M, Pierce J and Aaronson SA. (1989) Science 243: 800–803. 8. Yarden Y, Escobedo JA, Kuang W-J, Yang-Feng TL, Daniel TO, Tremble PM, Chen EY, Ando ME, Harkins RN, Francke U et al. (1986) Nature 323: 226–232. 9. Pawson T and Schlessinger J. (1993) Current Biol. 3: 434–442. 10. Sadowski I, Stone JC and Pawson T. (1986) Mol. Cellular Biol. 6: 4396–4408. 11. Pawson T and Gish GD. (1992) Cell 71: 359–362. 12. Brugge JS. (1993) Science 26: 918–919.

170

THERMODYNAMICS OF SH2 DOMAIN BINDING

13. Sawyer TK. (1998) Peptide Sci. 47: 243–261. 14. Coffey AJ, Brooksbank RA et al. (1998) Nature Genetics 20: 129–135. 15. Nichols KE, Harkin DP, Levitz S, Krainer M, Kolquist KA, Genovese C, Bernhar A, Ferguson M, Zuo L, Snyder E et al. (1998) Proc. Natl Acad. Sci. USA 95: 13 765– 13 770. 16. Sayos J, Wu C, Morra M, Wang N, Zhang X, Allen D, van Shalk S, Notarangelo L, Geha R, Roncarolo MG et al. (1998) Nature 395: 462–469. 17. Saffran DC, Parolini O, Fitch-Hilgenberg M, Rawlings DJ, Afar D, Witte ON and Conley ME. (1994) N. Engl. J. Med. 330: 1488–1491. 18. Tartaglia M, Mehler EL, Goldberg R, Zampino G, Brunner HG, Kremer H, van der Burgt I, Crosby AH, Ion A, Jeffery S et al. (2001) Nature Genetics 29: 465–468. 19. Botfield MC and Green J. (1995) Annu. Rep. Med. Chem. 30: 227–237. 20. Thomas SM and Brugge JS. (1997) Annu. Rev. Cell Dev. Biol. 13: 513–609. 21. Soriano P, Montgomery C, Geske R and Bradley A. (1991) Cell 64: 693–702. 22. Lowell CA, Niwa M, Soriano P and Varmus HE. (1996) Blood 87: 1780–1792. 23. Boyce BF, Yoneda T, Lowe C, Soriano P and Mundy GR. (1992) J. Clin. Invest. 90: 1622–1627. 24. Lowe C, Yoneda T, Boyce BF, Chen H, Mundy GR and Soriano P. (1993) Proc. Natl Acad. Sci. USA 90: 4485–4490. 25. Marzia M, Sims NA, Voit S, Migliaccio D, Taranta A, Bernardini D, Faraggiana T, Yoneda T, Mundy GR, Boyce BF et al. (2000) J. Cell Biol. 151: 311–320. 26. Schwartzberg PL, Xing L, Hoffmann O, Lowell CA, Garrett L, Boyce BF and Varmus HE. (1997) Genes Dev. 11: 2835–2844. 27. Sawyer T, Boyce BF, Dalgarno D and Iuliucci J. (2001). Expert Opin. Invest. Drugs 10: 1327–1344. 28. Chan AC, and Shaw AS. (1995) Curr. Opin. Immunol. 8: 394–401. 29. Kurosaki T. (1997) Curr. Opin. Immunol. 9: 309–318. 30. Reth M and Wienands J. (1997) Annu. Rev. Immunol. 15: 453–479. 31. Chan AC, Kadlecek TA, Elder ME, Filipovich AH, Kuo W-L, Iwashima M, Parslow TG and Weiss A. (1994) Science 264: 1599–1601. 32. Cheng AM, Negishi I, Anderson SJ, Chan AC, Bolen J, Loh DY and Pawson T. (1997) Proc. Natl Acad. Sci. USA 94: 9797–9801. 33. Cheng AM, Rowley B, Pao W, Hayday A, Bolen JB and Pawson T. (1995) Nature 378: 303–306. 34. Costello PS, Turner M, Walters AE, Cunningham CN, Bauer PH, Downward J and Tybulewicz VLJ. (1996) Oncogene 13: 2595–2605. 35. Crowley MT, Costello PS, Fitzer-Attas CJ, Turner M, Meng F, Lowell C, Tybulewicz VLJ and DeFranco AL. (1997) J. Exp. Med. 186: 1027–1039. 36. Poole A, Gibbins JM, Turner M, van Vugt MJ, van de Winkel JGJ, Saito T, Tybulewicz VLJ and Watson SP. (1997) EMBO J. 16: 2333–2341. 37. Zhang J, Berenstein EH, Evans RL and Siraganian RP. (1996) J. Exp. Med. 184: 71–79. 38. Brumbaugh KM, Binstadt BA, Billadau DD, Schoon RA, Dick CJ, Ten RM and Leibson PJ. (1997) J. Exp. Med. 186: 1965–1974. 39. Lanier LL, Corliss BC, Wu J, Leong C and Phillips JH. (1998). Nature 391: 703–707. 40. Corey SJ, Burkhardt AL, Bolen JB, Geahlen RL, Tkatch LS and Tweardy DJ. (1994) Proc. Natl Acad. Sci. USA 91: 4683–4687. 41. Gao J, Zoller KE, Ginsberg MH, Brugge JS and Shattil SJ. (1997) EMBO J. 16: 6414– 6425. 42. Higuchi M, Asao H, Tanaka N, Oda K, Tikeshita T and Sugamura K. (1997) Leukemia 11: 416–417.

REFERENCES

171

43. Lin TH, Rosales C, Mondal K, Bolen JB, Haskill S and Juliano RL. (1995) J. Biol. Chem. 270: 16 189–16 197. 44. Wan Y, Kurosaki T and Huang X-Y. (1996) Nature 380: 541–544. 45. Yanaga F, Poole A, Asselin J, Blake R, Schieven GL, Clark EA, Law CL and Watson SP. (1995) Biochem. J. 311: 471–478. 46. Lowenstein EJ, Daly RJ, Batzer AG, Li W, Margolis B, Lammers R, Ullrich A, Skolnick EY, Bar-Sagi D and Schlessinger J. (1992) Cell 70: 431–442. 47. Egan SE, Giddings BW, Brooks MW, Buday L, Sizeland AM and Weinberg RA. (1993) Nature 363: 45–51. 48. Olivier JP, Raabe T, Henkemeyer M, Dickson B, Mbamalu G, Margolis B, Schlessinger J, Hafen E and Pawson T. (1993) Cell 73: 179–191. 49. Rozakis-Adcock M, McGlade J, Mbamalu G, Pelicci G, Daly R, Li W, Batzer A, Thomas S, Brugge J, Pelicci PG et al. (1992) Nature 360: 689–692. 50. Simon MA, Dodson GS and Rubin GM. (1993) Cell 73: 169–177. 51. Kuriyan J and Cowburn D. (1997) Annu. Rev. Biophys. Biomol. Struct. 26: 259–288. 52. Waksman G, Kominos D, Robertson SR, Pant N, Baltimore D, Birge RB, Cowburn D, Hanafusa H, Mayer BJ, Overduin M et al. (1992) Nature 358: 646–653. 53. Pawson T. (1992). Curr. Opin. Struct. Biol. 2: 432–437. 54. Bradshaw JM, Mitaxov V and Waksman G. (1999) J. Mol. Biol. 293: 971–985. 55. Bradshaw JM and Waksman G. (1999) Biochemistry 38: 5147–5154. 56. Bradshaw JM, Mitaxov V and Waksman G. (2000) J. Mol. Biol. 299: 521–535. 57. Clackson T and Wells JA. (1995) Science 267: 383–386. 58. Songyang Z, Shoelson SE, Chaudhuri M, Gish G, Pawson T, Haser WG, King F, Roberts T, Ratnofski S, Lechleider RJ et al. (1993) Cell 72: 767–778. 59. Songyang Z, Shoelson SE, McGlade J, Olivier P, Pawson T, Bustelo XR, Barbacid M, Sabe H, Hanafusa H, Yi T et al. (1994) Mol. Cellular Biol. 14: 2777–2785. 60. Eck MJ, Shoelson SE and Harrison SC. (1993) Nature 362: 87–91. 61. Waksman G, Shoelson SE, Pant N, Cowburn D and Kuriyan J. (1993) Cell 72: 779– 790. 62. Breeze AL, Kara BV, Barratt DG, Anderson M, Smith JC, Luke RW, Best JR and Cartlidge SA. (1996) EMBO J. 15: 3579–3589. 63. Nolte RT, Eck MJ, Schlessinger J, Shoelson SE and Harrison SC. (1996) Nature Struct. Biol. 3: 364–373. 64. Pascal S, Singer AU, Gish G, Yamazaki T, Shoelson SE, Pawson T, Kay LE and Forman-Kay JD. (1994) Cell 77: 461–472. 65. Lee CH, Kominos D, Jacques S, Margolis B, Schlessinger J, Shoelson SE and Kuriyan J. (1994) Structure 2: 423–438. 66. Kay LE, Muhandiram DR, Farrow NA, Aubin Y and Forman-Kay JD. (1996) Biochemistry 35: 61–368. 67. Kay LE, Muhandiram DR, Wolf G, Shoelson SE and Forman-Kay JD. (1998) Nat. Struct. Biol. 5: 156–163. 68. Lubman OY and Waksman G. (2002) J. Mol. Biol. 316: 291–304. 69. Erpel T and Courtneidge SA. (1995) Curr. Opin. Cell Biol. 7: 176–182. 70. Kypta RM, Goldberg Y, Ulug ET and Courtneidge SA. (1990) Cell 62: 481–492. 71. Ralston R and Bishop JM. (1985) Proc. Natl Acad. Sci. USA 82: 7845–7849. 72. Alonso G, Koegl M, Mazurenko N and Courtneidge SA. (1995) J. Biol. Chem. 270: 9840–9848. 73. Mori S, Ro¨nnstrand L, Yokote K, Engstro¨m A, Courtneidge SA, Claesson-Welsh L and Heldin C-H. (1993) EMBO J. 12: 2257–264. 74. Lubman OY and Waksman G. (2003) J. Mol. Biol. 328: 655–668.

172

THERMODYNAMICS OF SH2 DOMAIN BINDING

75. Ladbury JE. (1996) Chem. Biol. 3: 973–980. 76. Renzoni DA, Zvelebil MJJM, Lundback T and Ladbury JE. (1997) In Structure-Based Drug Design, Ladbury JE and Connelly PR (eds), Landes, Austin, TX. Chapter 6. 77. Burshtyn DN, Yang W, Yi T and Long EO. (1997) J. Biol. Chem. 272: 13 066–13 072. 78. Huyer G, Li ZM, Adam M, Huckle WR and Ramachandran C. (1995) Biochemistry 34: 1040–1049. 79. Huyer G and Ramachandran C. (1998) Biochemistry 37: 2741–2747. 80. Li SC, Gish G, Yang D, Coffey AJ, Forman-Kay JD, Ernberg I, Kay LE and Pawson T. (1999) Curr. Biol. 9: 1355–1362. 81. Poy F, Yaffe MB, Sayos J, Saxena K, Morra M, Sumegi J, Cantley LC, Terhorst C and Eck MJ. (1999) Mol. Cell 4: 555–561. 82. McNemar C, Snow ME, Windsor WT, Prongay A, Mui P, Zhang R, Durkin J, Le HV and Weber PC. (1997) Biochemistry 36: 10 006–10 014. 83. Rahuel J, Gay B, Erdmann D, Strauss A, Garcia-Echeverria C, Furet P, Caravatti G, Fretz H, Schoepfer J and Gru¨tter MG. (1996) Nature Struct. Biol. 3: 586–589. 84. Marengere LE, Songyang Z, Gish CD, Schaffer MD, Parsons JT, Stern MJ, Cantley LC and Pawson T. (1994) Nature 369: 502–505. 85. Reth M. (1989) Nature 338: 383–384. 86. Cambier JC. (1995) J. Immunol. 155: 3281–3285. 87. Chacko GW, Brandt JT, Coggeshall KM and Anderson CL. (1996) J. Biol. Chem. 271: 10 775–10 781. 88. Grucza RA, Bradshaw JM, Mitaxov V and Waksman G. (2000) Biochemistry 39: 10 072–10 081. 89. Daeron M. (1997) Annu. Rev. Immunol. 15: 203–234. 90. Futterer K, Wong J, Grucza RA, Chan AC and Waksman G. (1998) J. Mol. Biol. 281: 523–537. 91. Grucza RA, Fu¨tterer K, Chan AC and Waksman G. (1999). Biochemistry 38: 5024– 5033. 92. Hatada MH, Lu X, Laird ER, Green J, Morgenstern JP, Lou M, Marr CS, Phillips TB, Ram MK, Theriault K et al. (1995) Nature 377: 32–38. 93. Labadia ME, Jakes S, Grygon CA, Greenwood DJ, Schembri-King J, Lukas SM, Warren TC and Ingraham RH. (1997) Arch. Biochem. Biophys. 342: 117–125. 94. Grazioli L, Germain V, Weiss A and Acuto O. (1998). J. Biol. Chem. 273: 8916–8921. 95. O’Brien R, Rugman P, Renzoni D, Layton M, Handa R, Hilyard K, Waterfield MD, Driscoll PC and Ladbury JE. (2000) Protein Sci. 9: 570–579. 96. Ottinger EA, Botfield MC, and Shoelson SE. (1998). J. Biol. Chem. 273: 729–735. 97. Rich DH. (1986) Proteinase Inhibitors, Elsevier, Amsterdam. 98. Bradshaw JM and Waksman G. (1998) Biochemistry 37: 15 400–15 407. 99. Alligood KJ, Charifson PS, Crosby R, Consler TG, Feldman PL, Gampe RT, Gilmer TM, Jordan SR, Milstead MW, Mohr C et al. (1998) Bioorg. Med. Chem. Lett. 8: 1189–1194. 100. Charifson PS, Shewchuck LM, Rocque W, Hummel CW, Jordan SR., Mohr C, Pacofsky GJ, Peel MR, Rodriguez M, Sternbach DD and Consler TG. (1997) Biochemistry 36: 6283–6293. 101. Violette SM, Shakespeare WC, Bartlett C, Guan W, Smith JA, Rickles RJ, Bohacek RS, Holt DA, Baron R and Sawyer TK. (2000) Chem. Biol. 7: 225–235. 102. Bohacek RS, Dalgarno DC, Hatada MH, Jacobsen VA, Lynch BA, Macek KJ, Merry T, Metcalf CA III, Narula SS, Sawyer T et al. (2001) J. Med. Chem. 44: 660–663.

REFERENCES

173

103. Shakespeare W, Yang M, Bohacek R, Cerasoli F, Stebbins K, Sundaramoorthi R, Azimioara M, Vu C, Pradeepan S, Metcalf C III et al. (2000). Proc. Natl Acad. Sci. USA 97: 9373–9378. 104. Oligino L, Lung F-DT, Sastry L, Bigelow J, Cao T, Curran M, Burke TR, Wang S, Krag D, Roller PP and King CR. (1997) J. Biol. Chem. 272: 29 046–29 052. 105. Gay B, Suarez S, Caravatti G, Furet P, Meyer T and Schoepfer J. (1999) Int. J. Cancer 83: 235–41. 106. Gay B, Suarez S, Weber C, Raheul J, Fabbro D, Furet P, Caravatti G and Schoepfer J. (1999) J. Biol. Chem. 274: 23 311–23 315. 107. Dunican DJ and Doherty P. (2001) Biopolymers 60: 45–60. 108. Sawyer T, Bohacek RS, Dalgarno D, Eyermann CJ, Kawahata N, Metcalf CA III, Shakespeare W, Sundaramoorthi R, Wang Y and Yang MG. (2002) Mini Rev. Med. Chem. 2: 475–488. 109. Shakespeare W. (2001) SH2 domain inhibition: a problem solved? Curr. Opin. Chem. Biol. 5, 409–415. 110. Burke TR Jr, Yao Z-J, Liu D-G, Voigt J and Gao Y. (2001) Biopolymers 60: 32–44. 111. Lung F-DT and Tsai J-Y. (2003) Biopolymers 71: 132–140. 112. Fambrough D, McClure K, Kazlauskas A and Lander ES. (1999) Cell 97: 727–741. 113. Pawson T and Saxton TM. (1999) Cell 97: 675–678. 114. Bradshaw JM and Waksman G. (2002) In Advances in Protein Chemistry, Janin J (ed.), Academic, New York, pp. 161–210.

9 Titration Calorimetry as a Tool to Determine Thermodynamic and Kinetic Parameters of Enzymes M. Lucia Bianconi

9.1 Introduction Although life is driven by spontaneous reactions, the action of biological catalysts such as enzymes is essential for these reactions to take place in a biologically relevant time-scale. Enzymes allow many chemical reactions to occur within the homeostasis constraints of a living organism, acting as catalysts that significantly increase the rate of a reaction by lowering the activation energy. They are very specific in nature, and while most of them are proteins a few ribonucleoprotein enzymes known as ribozymes have been described.1 The kinetic characterization of biological enzymes has been the subject of great interest as their clinical and biotechnological applications have increased in the past few decades. The use of enzymes in the diagnosis of several diseases is one of the most important benefits derived from the intensive research in clinical chemistry since the 1940s.2 The determination of abnormal levels of different enzymes in serum has been used, for instance, to diagnose miocardial infarction,3 acute pancreatitis,4 and liver diseases5 or in the evaluation of the severity of pre-eclampsia6 as well as cancer.7–9 Enzymes are also important targets in drug development for several life threatening diseases such as malaria, cancer, diabetes, heart disease, and Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

176

THERMODYNAMIC AND KINETIC PARAMETERS OF ENZYMES

stroke, among others. Malaria is a major health problem in developing countries and the increasing resistance of the parasite Plasmodium falciparum to the available drugs10 is a subject of considerable concern. The development of new targets for chemotherapy is based on anti-malarial drug development using models of enzyme structure.11,12 Enzyme research can also advance anticancer drug development.13,14 Many other clinical examples of the use of enzymes, from the replacement therapy for metabolic disorders such as Gaucher disease15 to the treatment of acute lymphoblastic leukaemia by using bacterial asparaginase16 have been reported. Industrial applications of enzymes have been related to the production of cheese, beer, and wine for many years. Today, with the methods of recombinant production by microorganisms, the applications of well characterized enzymes have enormously enhanced a variety of industrial processes including the detergent and textile industries.17 Therefore, the proper determination of enzyme kinetic parameters can be helpful in both the clinical and technological fields. Nevertheless, the direct measurement of either reactants or products of an enzyme catalysed reaction by optical or electrochemical assays is not always possible. In such cases, it is frequent to use a modified substrate by introducing a UV active or fluorescent group to follow the reaction kinetics. Fluorescence-based assays usually require the incorporation of a large fluorophor onto the substrate.18,19 However, substrate modification will almost certainly affect the affinity of the enzyme for the substrate, leading to misinterpretation of the data. It is also possible to use a coupled enzyme assay in order to study one particular reaction in which the product can be used as a reactant by an additional enzyme.20,21 The disadvantage of this method is the need for large amounts of the additional enzyme and the required cofactors to avoid unintended errors in the characterization of the enzyme of interest. ITC provides a direct and accurate assay for determining kinetic and thermodynamic parameters of an enzyme catalysed reaction, based on the measurement of the heat absorbed or released during catalysis. ITC has been used to study enzyme catalysed reactions since the 1950s22 and it is already well established in the literature.23–28 The great advantage of using ITC is that neither reactants nor the products need to have optical activity. This allows direct measurement of activity without changing the substrates or using coupled reactions. In addition, there is no need for the sample to be clear since the technique does not involve absorption or emission of light. In this Chapter, the methodology and data analysis for obtaining thermodynamic and kinetic parameters will be presented, together with the results from two different enzymes. All the thermograms were obtained in an MCS-ITC from MicroCal Llc (Northampton, MA), and the samples were prepared as described in the text. The data were analysed using either version 6.0 or 7.0 of the Origin software provided by MicroCal.

ITC OF ENZYME CATALYSED REACTIONS

177

9.2 ITC of enzyme catalysed reactions For a given reaction, the heat (Q) evolved is proportional to the molar enthalpy of the reaction (DH) and, therefore, to the amount of product ([P]) formed in a determined reaction volume (V): Q ¼ n DH ¼ ½Ptotal V DH

ð9:1Þ

The ITC method for enzyme assays is based on the fact that the thermal power reflects the heat flow, i.e. the dissipation of heat (dQ) as a function of time (dt). The heat flow (dQ/dt) is directly proportional to the rate of product formation (d[P]/dt), and can be described as dQ=dt ¼ DH Vðd½P=dtÞ

ð9:2Þ

Equation (9.2) can be rearranged in terms of the reaction rate: d½P=dt ¼ dQ=dt=ðDH VÞ

ð9:3Þ

Equation (9.3) shows that in order to calculate the rate of an enzyme catalysed reaction by ITC it is necessary to determine not only the heat flux, but also the enthalpy of the reaction. Under experimental conditions where the reaction can be described in terms of first-order kinetics the heat flow is given by dQ=dt ¼ DH Vk½C0  expð
ktÞ

ð9:4Þ

where k is the rate constant, and [C0] is the initial concentration of substrate. In conditions where the substrate is totally consumed, the observed calorimetric enthalpy (DHcal) for the reaction can be determined by dividing the total heat (QT) generated in the reaction by the amount of product formed. QT can be calculated by integrating the area under the peak of the calorimetric thermograms (dQ/dt as a function of time). In this chapter the results obtained with the hexokinases PI (Hxk1) and PII (Hxk2) from yeast and with glucose-6-phosphate dehydrogenase (G6PD) from Leuconostoc mesenteroides are presented. Hexokinases catalyse the transfer of a phosphoryl group from ATP to hexoses, where the reaction is usually studied by a coupled enzyme assay with glucose-6-phosphate dehydrogenase.20 Since hexokinases use two substrates, glucose and ATP, in the reaction, it is essential to use limited amounts of one substrate, maintaining the other in excess in order to find QT and, consequently, DHcal. This is necessary since one of the substrates should be totally consumed to obtain DHcal. It is worth noting that DHcal is independent of which substrate is used in excess/limited concentration. Figure 9.1 shows a typical calorimetric trace for the reaction catalysed by Hxk1 from yeast at 258C in Tris buffer. The reaction started by the injection of a small volume (14 ml) of the enzyme solution into the sample cell (V ¼ 1.38 ml) loaded with the reaction medium containing an excess of glucose and MgCl2,

178

THERMODYNAMIC AND KINETIC PARAMETERS OF ENZYMES

Figure 9.1 Calorimetric thermogram for the yeast Hxk1 reaction at 258C, in 50 mM Tris buffer containing 125 mM NaCl, 10 mM glucose, 5 mM MgCl2, and 0.1 mM ATP. The arrows indicate the time of the injections of enzyme

and ATP at a low concentration, and a total consumption of the ATP occurs, as observed by the return of the heat flux to the baseline level. The area under each peak gives the total heat (QT, in mcal) for the reaction (first injection) and for the dilution of the enyzme (Qd, second injection). The final Hxk1 concentration was 1.6 nM, which corresponded to 0.2 units ml71. The downward displacement of the baseline after the injection indicates the exothermic nature of the reaction. As the concentration of ATP decreases, the heat flow returns to the baseline level. A second injection of enzyme was always obtained after completion of the reaction to determine the heat of dilution (Qd) of the enzyme solution, which is subtracted from the total heat of reaction. It is important to note that the second injection is also necessary in order to obtain the correct baseline unless the Origin 7.0 software is being used for data analysis. The heat of dilution can be minimized if the enzyme is dialysed against the same buffer as used in the reaction medium. In some cases, dilution effects can be minimized by injections of enzyme rather than the substrate, which usually gives larger Qd values since the concentrations of enzyme used in kinetic studies are quite small. For example, the final concentration of Hxk1 isozyme used in the reaction shown in Figure 9.1 was 1.6 nM. However, the calorimetric enthalpy is the sum of different heat effects taking place during any reaction. If the reaction involves the release (or uptake) of protons, for instance, DHcal will be a combination of the intrinsic enthalpy of reaction (DHR) and the enthalpy of protonation (or ionization) for each proton absorbed (or released) by the buffer. DHR can be calculated from the linear relationship between DHcal and the enthalpy of buffer protonation/ ionization (DHP) expressed as

ITC OF ENZYME CATALYSED REACTIONS

DHcal ¼ DHR þ nDHP

179

ð9:5Þ

where n represents the number of protons absorbed or released in the reaction. It is important to check the pH of the reaction medium before and after the titrations. In the hexokinase reaction the phosphorylation of glucose is accompanied by a release of H+, which, in turn, is absorbed by the buffer. By using buffer systems with different DHP values the calorimetric traces will present a different area as shown in Figure 9.2 for the Hxk1 reaction in Tris, imidazole, Mops and Hepes. The correlation between DHcal and DHP described in equation (9.5) was applied (Figure 9.3), and although exothermic for both isozymes a small, but significant (p50.0001), difference in DHR was found in the reaction catalysed by Hxk1 (DHR ¼ 75.1+0.2 kcal mol71) or Hxk2 (DHR ¼ 73.3+0.3 kcal mol71). Nevertheless, the number of protons released during glucose phosphorylation is essentially the same for Hxk1 (n ¼ 0.94) and Hxk2 (n ¼ 0.96). It should be expected to obtain the same intrinsic enthalpy for the glucose phosphorylation reaction when using either isozyme. Therefore, the difference in DHR suggests that in addition to the glucose phosphorylation another side reaction is taking place, such as the enzyme phosphorylation and/or ATP hydrolysis.29

Steady-state kinetics An enzyme catalysed reaction can be followed at the steady state after injection of the enzyme solution into the calorimetric cell containing an excess

Figure 9.2 Calorimetric scans for Hxk1 reaction in Hepes (dots), Mops (dashes), imidazole (dash–dot–dot) and Tris (solid). The enthalpies of buffer protonation (DHP) are from the work of Takahashi and Fukada.29 At 258C, DHP is 75.02 kcal mol71 for Hepes, 75.22 kcal mol71 for Mops, 78.75 kcal mol71 for imidazole, and 711.36 kcal mol71 for Tris

180

THERMODYNAMIC AND KINETIC PARAMETERS OF ENZYMES

Figure 9.3 Correlation between calorimetric and buffer protonation enthalpy for Hxk1 (open symbols) and Hxk2 (solid symbols) determined in buffer systems with different enthalpies of protonation as described in Figure 9.2. The symbols represent the mean+S.E. (n ¼ 8), and the curves were obtained by linear regression of the experimental data with linear correlation r ¼ 0.9980 for Hxk1 and r ¼ 0.9973 for Hxk2, respectively. DHP for phosphate buffer is 71.22 kcal mol71 at 258C.29 (From reference 28, used by permission)

of substrate. In this way, reaction rates can be calculated since the heat flow (dQ/dt) is directly proportional to the rate of reaction (see Equation (9.3)). This is shown in Figure 9.4 for a reaction with G6PD from Leuconostoc mesenteroides. This enzyme catalyzes the oxidation of glucose-6-phosphate (G6P) and presents an unusual dual-coenzyme specificity that plays an important role in the metabolism of the bacteria.30 The G6PD from Leuconostoc mesenteroides is inhibited by ATP.30 The thermogram in Figure 9.4(a) was obtained using a reaction medium with high concentrations of both glucose-6-phosphate and NAD+. Two injections of enzyme at very low concentration were made and the reaction could be followed in the steady state. For this kind of experiment it is necessary to obtain a good baseline that precedes the first injection so that this can be subsequently subtracted. The integration of the calorimetric thermogram shows how the heat evolves as a function of time (Figure 9.4(b)). Once dQ/dt is obtained, it is possible to convert it into the enzyme turnover number (kcat) if the enthalpy of reaction DHR and the amount of enzyme used are known. In order to determine Km, a single experiment can be performed where several injections of substrate into the sample cell containing the enzyme are made. Figure 9.5 shows the raw calorimetric data for a series of 2.5 ml injections of NAD+ into the cell filled with a solution of glucose-6-phosphate and G6PD. After the first injection, there is an initial exothermic peak corresponding to the heat of NAD+ dilution (Qd). As soon as Qd is completed,

ITC OF ENZYME CATALYSED REACTIONS

181

Figure 9.4 Steady-state kinetics of the reaction catalysed by G6PD from Leuconostoc mesenteroides at 258C. (a) The raw data were obtained at 258C using a reaction medium containing 10 mM glucose-6-phosphate and 5 mM NAD+ in 50 mM Tris, pH 7.8, with two 5 ml injections of G6PD. Each injection corresponded to the addition of G6PD at a final concentration of 0.3 nM in the cell. (b) Integrated heat showing two slopes analysed by linear regression

the baseline stabilizes at a negative value of heat flux in relation to the initial baseline (Figure 9.5). This new level of the baseline indicates that an exothermic reaction is taking place. Another 2.5 ml injection of NAD+ is made after 3 min, and a similar response is observed. The new baseline level after Qd is twice as negative as the first injection. A similar response is obtained for the following injections, indicating that the rate of heat generated (dQ/dt) is proportional to the substrate concentration injected and, therefore, to the rate of reaction. Figure 9.6 shows the changes in dQ/dt as a function of substrate concentration, in which NAD+ varies from 0.01 to 0.31 mM in the cell. Figure 9.6 also shows the rates of reaction obtained in the presence of ATP, which acts as a competitive inhibitor of the G6PD.29 The Km-value for NAD+

182

THERMODYNAMIC AND KINETIC PARAMETERS OF ENZYMES

Figure 9.5 Raw calorimetric data for Km determination at 258C. The figure shows the first five injections (2.5 ml each) of 5.6 mM NAD+ into the sample cell containing 5 mM glucose6-phosphate and 0.10 units ml71 G6PD in 50 mM Tris, pH 7.8. The first exothermic peak corresponds to the heat of NAD+ dilution and the new baseline level after each injection to dQ/dt, which is proportional to the substrate concentration

obtained in the ITC experiments (Km ¼ 89 mM) is very close to that reported by Levy et al. (Km ¼ 106 mM).30 ITC also allows an accurate determination of thermodynamic parameters of activation since the actual temperature of the catalysed reaction is known. The temperature dependence of the reaction rate can be described by both the Arrhenius and the Eyring equations. The Eyring equation is a theoretical

Figure 9.6 Heat flux (dQ/dt) as a function of NAD+ concentration for the reaction with G6PD as described in Figure 9.5. The raw data were obtained with 20 injections (1062.5 ml, plus 1065.0 ml injections) of 5.6 mM NAD+ into the cell containing enzyme and 5 mM glucose-6-phosphate, in the absence (*) and in the presence (*) of 2 mM ATP

ITC OF ENZYME CATALYSED REACTIONS

183

construct based on the transition-state theory, where the rate constant (k) for the complete reaction is described as k¼

kB T exp
DH] =RTÞexpð
DS] =RÞ h

ð9:6Þ

where kB is Boltzmann’s constant (1.381610723 J K71), h is the Planck constant (6.626610734 J s), and DH# and DS# are the activation enthalpy and entropy respectively. A plot of ln(k/T) versus 1/T will be a straight line with a slope of 7DH#/R. From the y-intercept, one can calculate DS# by yðx ¼ 0Þ ¼ lnðkB =hÞ þ DS] =R

ð9:7Þ

Therefore, the free energy of activation can be calculated for the appropriate reaction temperatures, according to DG] ¼ DH]
T DS]

ð9:8Þ

Figure 9.7 shows the Eyring plot for the G6PD reaction, using kcat-values obtained from calorimetric thermograms. Since the MCS-ITC was used, the lower temperature of the reaction was 228C. It can be seen from the error bars for the x-axis that the variation in temperature from one experiment to another is very small. However, the possibility of having the exact temperature for the reaction gives a more precise determination of the thermodynamic parameters of activation.

Figure 9.7 Eyring plot for the reaction with G6PD from 22 to 458C. The kcat-values were calculated from kinetic data obtained by ITC as described in the text. The error bars represent SD values for the variations found in kcat and in the actual temperature of the experiment

184

THERMODYNAMIC AND KINETIC PARAMETERS OF ENZYMES

9.3 Summary In this chapter, the description of the experimental design for the determination of kinetic and thermodynamic parameters of enzyme catalysed reactions has been achieved by describing some examples with the yeast hexokinase isozymes and glucose-6-phosphate from Leuconostoc mesenteroides. When studying an enzyme of interest, it is important to remember that different factors can affect the reaction and the appropriate choice of the medium is necessary for a better understanding of the mechanisms of catalysis. Therefore, the right choice of buffers, enzyme and substrate concentration, pH, temperature, or even the ionic strength can be important when kinetic mechanisms are being studied. ITC is a very convenient tool for such studies since it provides the direct measurement of the reaction rate without the need for modified substrates or a coupled enzyme reaction. Previous studies showed a very good agreement between kinetic data obtained by spectroscopy and ITC.24,28 The determination of thermodynamic parameters of enzyme catalysed reactions provides valuable information about the nature of the transition state, and consequently about the reaction mechanism. The entropy of activation, for instance, gives a measure of the intrinsic probability of the transition state. Large and negative values of DS# indicate that the transition state requires that the reacting molecules adopt a highly ordered conformation where the degrees of freedom (translational, rotational, or vibrational) are restricted. In short, the methodology described here for ITC presents several advantages in the characterization of biological enzymes, for either the development of new drugs or new biotechnological applications.

Acknowledgements The author thanks CNPq for a research fellowship used in this work, MicroCal, Llc. for technical support and advice, Dr L. de Meis for calorimetric facilities (maintained by funds from PRONEX-CNPq), and Mr A. C. Miranda for technical assistance.

References 1. Cech TR. (2002) Biochem. Soc. Trans. 30: 1162–1166. 2. Kaplan A, Jack R, Opheim KE, Toivola B and Lyon AW. (1995) In Clinical Chemistry – Interpretation and Techniques, 4th edn, Harris JM, (ed), Williams and Wilkins, Malvern, PA.

REFERENCES

185

3. Lee TH and Goldman L. (1986) Ann. Intern. Med. 105: 221–233. 4. Wereszczynska-Siemiatkowska U, Jedynak M, Mroczko B and Siemiatkowski A. (2003) Arch. Immunol. Ther. Exp. 51: 195–200. 5. Reichling JJ and Kaplan MM. (1988) Dig. Dis. Sci. 33: 1601–1604. 6. Yoneyama Y, Suzuki S, Sawa R, Otsubo Y, Miura A, Kuwabara Y, Ishino H, Kiyokawa Y, Doi D, Yoneyama K, Kobayashi H and Araki T. (2002) Gynecol. Obstet. Invest. 54: 168–171. 7. Killian CS, Emrich LJ, Vargas FP, Yang N, Wang MC, Priore RL, Murphy GP and Chu TM. (1986) J. Natl Cancer Inst. 76: 179–185. 8. Laidler P, Kowalski D and Silberring J. (1991) Clin. Chim. Acta 204: 69–77. 9. Turkmen S, Oner P, Cinar F, Kocak H, Guvenen G, Altun H and Eryavuz Y. (2001) Cancer Lett. 166: 95–101. 10. White NJ. (1998) Br. Med. Bull. 54: 703–715 11. Rosenthal PJ, Olson JE, Lee GK, Palmer JT, Klaus JL and Rasnick D. (1996) Antimicrob. Agents Chemother. 40: 1600–1603. 12. Sharma SK, Kapoor M, Ramya TN, Kumar S, Kumar G, Modak R, Sharma S, Surolia N and Surolia A. (2003) J. Biol. Chem. 278: 45 661–45 671. 13. Schindler T, Bornmann W, Pellicena P, Miller WT, Clarkson B and Kuriyan J. (2000) Science 289: 1857–1959. 14. Sharma S, Raymond E, Soda H, Sun D, Hilsenbeck SG, Sharma A, Izbicka E, Windle B and Von Hoff DD. (1997) Ann. Oncol. 8: 1063–1074. 15. Grabowski GA, Leslie N and Wenstrup R. (1998) Blood Rev. 12: 115–133. 16. Mu¨ller HJ and Boos J. (1998) Crit. Rev. Oncol. Hematol. 28: 97–113. 17. Kirk O, Borchert TV and Fuglsang CC. (2002) Curr. Opin. Biotechnol. 13: 345–351. 18. Szollosi J, Damjanovich S and Matyus L. (1998) Cytometry 34: 159–179. 19. Boonacker E and Van Noorden CJ. (2001) J. Histochem. Cytochem. 49: 1473–1486. 20. Guerra R and Bianconi ML. (2000) Biosci. Rep. 20: 41–49. 21. McCarthy JK, O’Brien CE and Eveleigh DE. (2003) Anal. Biochem. 318: 196–203. 22. Sturtevant JM. (1955) J. Am. Chem. Soc. 77: 255–258. 23. Barclay K and Jespersen N. (1975) Anal. Lett. 8: 33–40. 24. Morin PE and Freire E. (1991) Biochemistry 30: 8494–8500. 25. Williams BA and Toone EJ. (1993) J. Org. Chem. 58: 3507–3510. 26. Lonhienne T, Baise E, Feller G, Bouriotis V and Gerday C. (2000) Biochim. Biophys. Acta 1545: 349–356. 27. Todd MJ and Gomez J. (2001) Anal. Biochem. 296: 179–187. 28. Bianconi ML. (2003) J. Biol. Chem. 278: 18 709–18 713. 29. Fukada H and Takahashi K. (1998) Proteins 33: 159–166. 30. Levy HR, Christoff M, Ingulli J and Ho EML. (1983) Arch. Biochem. Biophys. 222: 473–488.

Part III Differential Scanning Calorimetry

10 Energetics of Site-Specific DNA Recognition by Integrase Tn916 Stoyan Milev, Hans Rudolf Bosshard and Ilian Jelesarov

10.1 Introduction Site-specific protein binding to short DNA duplexes is often accompanied by heat capacity changes (DCp), whose magnitude cannot be accounted for by the amount and type of surface buried at the protein–DNA interface. Surfacebased parametrization schemes assume that the heat capacities of the free components do not differ in the free and bound states. However, even if binding is not coupled to pronounced conformational changes (structural adaptation), the heat capacities of all the participants in the association reaction are not invariant over the temperature range of isothermal titration calorimetry (ITC) experiments. This may be a typical feature of protein–DNA association. DNA-binding domains are often very flexible or partly unfolded in the free state and their heat capacity increases rapidly with temperature.1 The heat capacity of short DNA duplexes also increases significantly below 30– 408C.2,3 On the other hand, association can cause redistribution of thermally accessible states. Therefore, ITC-measured DCp may contain contributions arising from non-parallel changes of the heat capacities of the associated and dissociated states of the interacting system. Differential scanning calorimetry (DSC) provides the methodology for dealing with such complications, but heat capacity analysis has only occasionally been applied to protein–DNA complexes.4,5

Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

190

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

The advantage of linking information from DSC and ITC experiments is exemplified in this chapter for the case of the integrase–DNA complex. The DNA-binding domain (DBD) of the transposon integrase protein (INT-DBD) binds its DNA target site by positioning the face of a three-stranded antiparallel b-sheet within the major groove.6 Interestingly, the protein partially unfolds upon association. INT-DBD binding to DNA was studied by ITC. The observed DCp is a non-linear temperature function increasing from 71.4 to 72.9 kJ K71 mol7 between 4 and 308C. The expected DCp due to dehydration of molecular surface buried at the complex interface amounts to only 71.2 kJ K71 mol71. A precise analysis of the partial molar heat capacities of free protein, free DNA duplex and protein–DNA complex reveals that the thermal motions of the components are restricted upon specific binding. After correction for this effect, DCp becomes temperature independent but its magnitude of 71.8 kJ K71 mol71 still overestimates the semi-empirical, surface-based prediction. We propose that incomplete dehydration of polar groups at the complex interface is the reason for the discrepancy between experiment and semi-empirical predictions of binding energetics. Indeed, packing density analysis identifies at the complex interface cavities large enough to accommodate *10 water molecules. The thermodynamic consequences of incomplete dehydration can roughly reconcile the enthalpy and heat capacity change characterizing ‘rigid-body’ binding with the structural features of this particular protein–DNA complex.

10.2 Conformational stability of INT-DBD and the target DNA duplex ITC estimates of DCp are reliable only if DH measurements can be performed over a broad temperature range where the binding partners are in their binding-competent state. Modern titration calorimeters operate between 5 and 60–708C. Unfortunately, DNA-binding domains and short DNA duplexes are usually not very stable and start to unfold at lower temperatures. In order to determine the temperature range suitable for ITC experiments we have measured the conformational stability of INT-DBD and the target 13 bp DNA duplex.

Thermal unfolding of INT-DBD followed by DSC Partial molar heat capacity Figure 10.1(a) shows the temperature dependence of the partial molar heat capacity of the protein (heavy line). The transition maximum and the shape of

CONFORMATIONAL STABILITY OF INT-DBD

191

Figure 10.1 Thermal unfolding of the protein observed by DSC. (a) Temperature dependence of the partial molar heat capacity function. (b) Deconvolution of the apparent partial molar heat capacity

the calorimetric trace are independent of the protein concentration and the rate of heating, in agreement with a reversible monomolecular conformational transition. Hence, slow kinetic steps are not affecting the conformational transition. The absolute Cp at 258C is 12.25+0.50 kJ K71 mol 71, corresponding to 1.43+0.12 J K71 per gram of protein. This is well within the range of specific heat capacities for other small globular proteins and equal to that of barnase.7,8 The initial slope of the pre-transitional portion of the specific heat capacity below 158C is small, 6–761073 J K72 g71, but increases above *158C to 12.561073 J K72 g71. This increase may point to a small conformational transition preceding the main unfolding reaction, as will be discussed below. The heat capacity of the denatured protein is constant above 658C and has a value of 18.9+0.7 kJ K71 mol 71 (2.13+0.08 J K71 g71). The expected heat capacity of the fully unfolded protein calculated according to Privalov and Makhatadze10 (–6–) and Hakin and Hedwig9 (–&–) is also indicated. Experiments were performed with 65 mM protein in standard buffer of pH 6.0. The experimental heat capacity trace of the protein is enveloped by the two Cp functions, indicating that the unfolded chain is fully hydrated above 658C.

Heat capacity change from DSC melting curve The difference between the partial heat capacity of the unfolded and the folded state represents the direct calorimetric estimate of the unfolding heat capacity 71 mol 71 at 258C (53 J K71 mol residue71) change, DCcal p . It is equal to 3.9 kJ K 71 71 and 4.1 kJ K mol at 108C (56 J K71 mol residue71). Considering the are in uncertainties of the model calculations, the two values of DCcal p agreement with DCp of 5.0+0.8 kJ K71 mol 71 obtained by van’t Hoff analysis of spectroscopic unfolding data (not shown).

192

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

Evidence for a minor conformational transition preceding the main unfolding reaction The intrinsic heat capacity change is modelled in two ways: connection of the heat capacity of the folded state between 5 and 158C with the heat capacity of the unfolded state above 658C (thin solid line); connection of the heat capacity of the folded state between 15 and 258C with the heat capacity of the unfolded cal state above 658C (dotted line). The ratio DHvH m /DHm is 1.04+0.07 if the excess heat capacity function is constructed using the data above 158C, in accord with a single two-state transition. However, if the slope of Cp below 158C is considered to represent the heat capacity of the native state, DHvH m / is 0.92+0.07. Although this ratio is perhaps not significantly below DHcal m unity, it indicates that the unfolding reaction is not perfectly two state and is best described by a minor and a major transition. A more rigorous analysis makes no assumption about the ‘baseline’ connecting the pre- and posttransitional heat capacity traces but uses a general value for the heat capacity function of a folded protein.11,12 Analysing the trace of Figure 10.1(a) with Cp,N ¼ 6.7+1.061073 J K72 g71 (set of 12 proteins; Table III of Gomez et al.7) one obtains the minimal model shown in Figure 10.1(b), which accurately describes the experimental heat capacity trace by two transitions. The two component reactions corresponding to a minor and a major transition are indicated by broken lines. The experimental function (solid line, corresponding to the trace of panel (a)) and the function synthesized using the two component reactions (dotted line) match perfectly. The main transition has a Tm of 43.7+0.38C and a DHm of 245+19 kJ mol 71. The minor transition has a Tm of 28.1+5.08C and a DHm of 29.7+10.4 kJ mol 71. DCp of the main transition is 2.6+0.5 kJ K71 mol 71 and that of the minor transition is 1.9+0.1 kJ K71 mol 71. Values of Tm, DHm and DCp obtained from the different analyses of the DSC traces in Figure 10.1 as well as from thermal unfolding monitored by circular dichroism (CD) and fluorescence spectroscopy (not shown) are put together in Table 10.1. The thermodynamic parameters deduced from independent calorimetric and spectroscopic experiments agree well. The unfolding free energy is 14 kJ mol 71 at 258C and the protein is thus fully folded up to 308C.

Conformational stability of target DNA duplex Figure 10.2(a) presents the partial molar heat capacity function of DNA melting obtained by DSC. Dotted lines connecting the pre-transitional and post-transitional traces model the apparent intrinsic heat capacity change. The calculated heat capacity of the folded duplex is indicated by the lowermost (dashed) line. The solid line corresponds to the baseline. The heat capacity

193

CONFORMATIONAL STABILITY OF INT-DBD

Table 10.1 Thermodynamic parameters for thermal unfolding of INT-DBD, 13 bp DNA duplex and their 1:1 complexa Tm (8C)

DHm (kJ mol71)

43.7

28+5 43.7

255+13 (cal) 236+13 (vH) 0.92+0.07 238+13 (cal) 248+13 (vH) 1.04+0.07 30+10 245+19

43.4 43.7

258+25 230+25

Trp fluorescence mean DNA (65 mM) DSCe

44.2 43.8+0.3

275+30 255+18

48.0

CD INT-DBD-DNA complex (65 mM) DSC

47.7

332+4 (cal, peak) 335+5 (vH) 334

INT-DBD DSC single transitionc (data below 158C) single transitiond (data above 158C) two transitionse CD 219 nm 285 nm

44.1

52

DCp (kJ K mol71) 71

@Cp/@Tb (kJ K72 mol71)

4.9+0.6 (at 258C) 0.055 (5–158C) 5.8+0.6 (at Tm) 3.9+0.6 (at 258C) 0.107 (15–308C) 4.0+0.6 (at Tm) 1.9+0.1 2.6+0.5 5.7+0.6

5.0+0.8 *0 (at Tm)

0.192 (average)

*0 (at Tm)

0.24

Sum INT-DBD+DNA

0.30 (average)

a

Conditions: 50 mM Na phosphate, 100 mM NaCl, pH 6. ‘Average’ indicates the average temperature slope of Cp (4–308C). c Calculated with the solid ‘baseline’ in Figure 10.1(a). d Calculated with the dotted ‘baseline’ in Figure 10.1(a). e Statistical–mechanical deconvolution analysis.11,12 b

peak develops from a rather low temperature and is asymmetric; Tm increases with increasing DNA concentration. This is typical for unfolding linked to dissociation. The heat capacity of the duplex increases steeply below the onset of the main transition. The average initial slope is 0.192+0.017 kJ K72 mol 71, corresponding to 2561073 J K72 per gram of DNA. Linear extrapolation of the initial and final heat capacities into the transition zone (dotted lines in Figure 10.2(a)) shows a very small difference at Tm between the heat capacity of the duplex and the sum of the heat capacities of the unfolded complementary strands. The mean apparent transition enthalpy obtained by integration of the four heat capacity peaks delimited by the dotted ‘baselines’ is 332+4 kJ mol 71. A very similar value is obtained from from van’t Hoff analysis of CD melting experiments (Table 10.1) and from plotting 1/Tm versus the concentration of DNA duplex and (not shown).

194

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

Figure 10.2 Thermal unfolding of the duplex DNA followed by DSC. (a) Melting of 33, 65, 90 and 119 mM duplex DNA (left to right). (b) Excess heat capacity function calculated from the trace for 65 mM duplex DNA and the thin solid ‘baseline’ of panel (a)

Deconvolution analysis of the excess heat capacity function yields two different, partially overlapping phases: an initial gradual heat absorption, followed by a phase of intense heat absorption associated with the disruption of bulk packing interactions and bimolecular strand dissociation. The first phase is centred around 208C and proceeds with absorption of 15–20 per cent of the total heat (Figure 10.2(b)). The low- and high-temperature components indicated by dashed lines corresponds, respectively, to a broad pre-transitional accumulation of thermal energy and to cooperative melting accompanied by strand dissociation. Addition of the two components perfectly matches the experimental data (overlaying solid lines). Extrapolating the total melting enthalpy back to room temperature with the help of DCp of 0.21 kJ K71 (mol bp)71 (a typical value for DNA unfolding) one obtains DH, DS and DG of duplex formation at 258C of 27 kJ (mol bp)71, 76 J K71 (mol bp)71 and 4.1 kJ (mol bp)71, respectively. These values are typical for natural and synthetic DNA molecules. The dissociation constant of the 13 bp duplex DNA increases from 1610713 M at 58C to 161078 M at 308C. At 308C and 30 mM concentration, the duplex is still more than 98 per cent populated.

10.3 Thermodynamics of complex formation measured by titration calorimetry Dissociation constant, enthalpy of association and complex stoichiometry deduced from ITC Addition of small aliquots of protein to DNA in the calorimetric cell produces measurable heat effects, which saturate as the molar ratio of protein to DNA increases (Figure 10.3(a)). Symbols represent the integrated heats after correction for non-specific heat effects and normalization for the molar

UNFOLDING OF THE PROTEIN–DNA COMPLEX

195

concentration. Continuous lines are non-linear fits for a 1:1 binding model. Experimental temperatures are indicated. From the shape of the titration curve the binding stoichiometry, n, the association constant, KA, and the apparent calorimetric enthalpy of association, DHA, are calculated.13 The stoichiometry is 1.02+0.15 (mean+SD of 20 experiments), in agreement with a 1:1 complex. Within experimental error, KA is temperature independent between 4 and 308C. The average KA is (7.1+5.0)6106 M71, in agreement with KA ¼ (8.3+0.3)6106 M71 from fluorescence titration (not shown). The apparent enthalpy and entropy changes of association vary with temperature (Figure 10.3(b)). Complex formation is endothermic at low temperature and exothermic at high temperature, DHA changing sign at around 138C. The same DHA is measured in buffers of different heats of protonation, indicating that there is no change in the protonation state of the protein or the DNA upon binding.14 The entropy of complex formation is positive below about 298C and negative at higher temperature (Figure 10.3(b) and Table 10.2).

The heat capacity change of association determined by ITC The temperature variation of the binding enthalpy, dDHA/dT, represents the heat capacity change of association, DCp,A. Linear regression of the data shown in Figure 10.3(c) yields DCp,A of 72.3+0.2 kJ K71 mol 71 (dotted line). However, the data are not well described by a straight line. Fitting statistics improve significantly if a second-order polynomial is used (heavy line in Figure 10.3(c)). The curvature of the dDHA/dT function indicates a temperature variation from 71.4 kJ K71 mol 71 at 48C to 72.9 kJ K71 mol 71 at 308C. The thin solid line shows a linear fit of dDHA,corr/dT, yielding DCp,A,corr of 71.8+0.3 kJ K71 mol71 for a hypothetical rigid-body association reaction.

10.4 Thermal dissociation and unfolding of the protein–DNA complex The temperature-induced conformational transitions of the complex are highly reversible at pH 6 and 0.16 M ionic strength. The partial molar heat capacity function is shown in Figure 10.4 together with the traces recorded for the isolated components. Thermograms are recorded in standard buffer of pH 6.0 at a heating rate of 1 K min71 with sample concentrations of 65 mM (heavy line, complex; thin line, free protein; dotted line, DNA; dashed line, calculated sum of partial molar heat capacities of free protein and free DNA). The DSC trace demonstrates that the components are thermally stabilized in the complex. Melting of the complex produces a single, sharp heat absorption

196

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

Figure 10.3 Energetics of protein–DNA association measured by ITC. (a) Binding isotherms from titration of 30 mM DNA placed in the calorimetric cell with protein from a 300 mM stock solution in 50 mM Na phosphate, 100 mM NaCl, pH 6.0. (b) Changes of enthalpy, entropy and free energy in the range of 4–308C. (c) Determination of the heat capacity change DCp,A from @DHA/@T. Filled squares, DHA; asterisks, DHA,corr for a hypothetical rigid-body association reaction

197

UNFOLDING OF THE PROTEIN–DNA COMPLEX

Table 10.2 Thermodynamic parameters of the association of the integrase DNA binding domain with a 13 bp duplex DNA observed by ITC (sample data of 20 experiments)a T (8C) 4.3 5.7 7.5f 8.5f 9.3 17.6 25.1 27.1g 30.0

DHA (kJ mol 71) 15.7+1.0 10.9+4.4 11.8+0.6 9.8+0.8 7.6+0.2 710.4+1.0 725.4+1.3 740.1+4.2 747.8+4.0

DHA,corrb (kJ mol 71)

DGAc (kJ mol 71)

T DSAd (kJ mol 71)

15.8 11.2 12.4 10.7 8.9 75.5 714.3 726.5 726.2

736.1+2.1 733.2+1.4 735.4+1.0 737.7+2.0 734.6+1.2 738.1+3.2 740.0+4.0 741.5+4.0 737.8+1.4

51.8+2.3 44.1+4.6 47.8+1.5 47.5+1.7 42.2+1.2 27.7+3.3 14.6+4.2 1.4+5.6 710.0+4.2

T DSA,corre (kJ mol 71) 51.9 44.4 48.4 48.4 43.4 32.6 25.7 15.0 9.6

a Conditions: 50 mM Na phosphate, 100 mM NaCl, pH 6.0. Errors of DHA and DGA reflect errors of protein and DNA concentration determination, which were obtained as standard errors of the mean from three to five rounds of fitting with either the protein or the DNA concentration as the adjustable parameter of a 1:1 complex. The fitting error was less than 5 per cent for DHA and less than 10 per cent was 2–3 kJ mol 71 and that of DGA 3–4 kJ mol 71. for KA. From triplicate experiments, the error of DHq A ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi The error zTDS of T DSA was calculated as zTDS ¼ ðzDH Þ2 þ ðzDG Þ2 . b DHA,corr is the enthalpy change of a hypothetical rigid-body association reaction calculated from Equation (10.2); see the text for details. c From 7RT ln KA. d From DHA 7DGA. e From DHA,corr7DGA. f Experiment in HEPES buffer. g Experiment in ACES buffer.

peak whose temperature of maximum heat absorption is higher than the melting temperatures of protein and DNA alone. The shape of the DSC trace of the complex is asymmetric, which indicates a cooperative unfolding process accompanied by subunit dissociation.15 The dissociation of the complex and the concurrent unfolding of the protein and the DNA take place within a narrow temperature interval.

Figure 10.4 Heat capacity curves of the complex and the isolated components obtained by DSC

198

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

10.5 The heat capacity change of protein–DNA association Analysis of the partial molar heat capacities explains the temperature dependence of DCp,A A peculiar feature of the protein–DNA association reaction studied here is the non-linear temperature dependence of the association enthalpy seen in Figure 10.3(c). This behaviour can be explained in terms of temperatureinduced heat capacity changes of the protein, DNA and their complex occurring in the temperature interval of the ITC experiments. Such changes are clearly seen between 4 and 308C in Figure 10.4. The calculated sum of the heat capacity slopes of protein and DNA (dashed line in Figure 10.4) increases non-linearly from 0.24 kJ K72 mol71 near 48C to about 0.40 kJ K72 mol71 near 308C, the average over the ITC temperature range being 0.32 kJ K72 mol71. At the same time, the heat capacity of the complex displays a nearly constant temperature slope of 0.24 kJ K72 mol71. This behaviour indicates the following. (a) In the temperature range before the main thermal transition, the free protein, the free DNA and the complex exhibit minor changes of conformation, of thermal fluctuations and of vibrational content, including changes in the water shell around the molecules. (b) In the complex and its components the changes do not occur in parallel. (c) The enthalpy fluctuations of free protein and free DNA duplex are attenuated when association has taken place. These observations explain in a qualitative way why dDHA/dT is curved (heavy line in Figure 10.3(c)). If we were able to account in a quantitative way for the different thermal properties of the system in its associated and dissociated states, we could predict the enthalpic behaviour of a hypothetical rigid-body association reaction between the protein and the DNA in their binding competent conformations (see Scheme I below). The necessary procedure has been developed by Privalov and colleagues.5 The enthalpy of association at any temperature can be written as DHA ðTÞ ¼ DHA ðTR Þ þ DCp,A ðTR ÞðT
TR Þ þ l

ð10:1Þ

where l is defined as ðT ( TR



) X i Cp ðTÞ
Cp ðTR Þ dT Cp ðTÞ
Cp ðTR Þ
c

i

Superscript i denotes free protein and free DNA, respectively, and superscript denotes the complex. DCp,A(TR) is defined as

c

199

THE HEAT CAPACITY CHANGE OF PROTEIN–DNA ASSOCIATION

DCp;A ðTR Þ ¼ Cp ðTR Þc


X

Cp ðTR Þi

i

The integral l represents the difference between the temperature dependence of Cp of the complex and the summed temperature slopes of Cp of free protein and DNA. Thus, we can define a ‘corrected’ enthalpy change of rigid-body association as DHA;corr ¼ DHA
l

ð10:2Þ

Values of DHA,corr were calculated from Equation (10.1). They are added as asterisks to Figure 10.3(c) and are listed in Table 10.2. Most satisfying, the change of DHA,corr with temperature is linear, as expected for rigid-body association. The corrected heat capacity change, DCp,A,corr ¼ @DHA,corr/@T, is 71.8+0.08 kJ K71 mol 71 (slope of thin straight line in Figure 10.3(c)).

Correlation of the heat capacity change of association with structural features DCp,A,corr of the hypothetical rigid-body association reaction can now be judged against the structure of the complex, that is with regard to the molecular surface buried at the complex interface. The association reaction can be formally divided into two steps (Scheme I): (a) the transition of protein and DNA from the free to the binding-competent conformations (marked by asterisks in Scheme I) and (b) the association of the binding-competent molecules to the complex. The heat capacity changes can be estimated from the dehydration of non-polar and polar surface.8,16,17 The protein and the DNA unfold slightly to expose more binding surface in their bindingcompetent conformation (+450 A˚2 of non-polar surface and 780 A˚2 of polar surface). Therefore, DCp,conf is positive with an estimated value of 0.83 kJ K71 mol 71. Association of the binding-competent molecules buries 790 A˚2 of aliphatic, 540 A˚2 of aromatic and 1040 A˚2 of polar surface, leading to a negative value of DCp,rb of 71.2 kJ K71 mol 71. The estimated overall heat capacity change of association is DCp,A ¼ DCp,conf+DCp,rb and is very much

conformational transition DCp,conf = 0.83+0.17

"

protein*+DNA* rigid-body association DCp,rb = 71.2+0.1

DCp,A = 70.37+0.12

Scheme 1

"

"

protein+DNA

protein–DNA complex

200

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

smaller than any of the measured values. The analysis of the DSC data leading to DCp,A,corr eliminates the contributions arising from conformational adaptation and thermal fluctuations. Therefore, DHA,corr represents the enthalpy of inter-molecular contacts plus the changes of hydration upon association of rigid molecular surfaces. Hence, DCp,A,corr should be equal to the structure-based estimate of DCp,rb. This is not the case: DCp,A,corr is 71.8 kJ K71 mol 71 and DCp,rb is 71.2 kJ K71 mol 71. How can we explain the ‘extra’ 70.6 kJ K71 mol 71? We propose that it arises from incomplete dehydration of the complex interface.

Water is probably present at the protein–DNA interface Since the ‘extra’ heat capacity change is negative, it can be attributed to partial dehydration of polar surface (the heat capacity of polar hydration is negative). Thus, the 70.6 kJ K71 mol 71 can be accounted for by assuming that the buried polar surface is only *60 per cent dehydrated. This estimation is based on an elementary contribution of hydration of 71.27 J K71 mol 71 A˚72.8 Dehydration is even less than 60 per cent if one takes a value of 71.09 J K71 mol 71 A˚72 from the parametrization advanced by Freire and colleagues.18 Incomplete dehydration means that water is trapped in the complex. From packing density calculations using the NMR structures of complex, protein and DNA, we identify a total of 140+40 A˚3 of ‘empty space’ distributed over six to seven cavities at the complex interface. These cavities are mainly polar and large enough to contain together about 10 water molecules that are inaccessible to the bulk solvent. Five to seven water molecules bridge polar protein and DNA groups throughout a 2 ns MD simulation of the complex. According to statistical structural analysis 15 water molecules are expected at the interface of an ‘average’ protein–DNA complex which buries 2300 A˚2.19

Incomplete dehydration can be reconciled with the enthalpy of association The enthalpy of an association process approximating rigid-body binding can be formally decomposed into four contributions: pol apol pol DHA ¼ DHapol int þ DHint þ DHdehydr þ DHdehydr

ð10:3Þ

where the terms DHint represent the intermolecular contacts in vacuum and the terms DHdehydr the enthalpy effects of surface dehydration; superscripts apol and pol refer to contacts between non-polar and polar surfaces, respectively. Using the total amount of buried surface to predict the total enthalpy effect of dehydration following the parametrization of Privalov and colleagues,8 we

DISCUSSION

201

pol obtain about 1460 kJ mol 71 for DHapol dehydr þ DHdehydr , of which about 90 per cent arises from the very unfavourable enthalpy of burying polar groups at the complex interface. The enthalpy of van der Waals contacts involving aliphatic 71. Since the total and aromatic groups, DHapol int , is only about 7195 kJ mol association enthalpy of the rigid-body association reaction at 258C, DHcorr A , is pol 71 71 (Figure 10.3(c)), one obtains DHint of 71270 kJ mol from 719 kJ mol insertion in Equation (10.3). It is reasonable to assume that the bulk of DHpol int originates from hydrogen bonds at the complex interface. Again with the parametrization from reference 8, we take a value of 745 to 760 kJ mol 71 for the enthalpic content of a single hydrogen bond and predict about 21–28 hydrogen bonds at the complex interface. There are only 13 residues capable of forming H-bonds between the protein and the DNA. Hence the predicted number of H-bonds is too large. However, if only 60 per cent of the buried polar surface were dehydrated upon binding, DHpol int would be only about 7760 kJ mol 71 and would agree with 13–17 H-bonds. This estimate is corroborated by NMR structural data.6 Thirteen H-bonds can be observed in at least two or three conformers of the NMR ensemble. Up to 16 hydrogenbonds can be identified in some conformers after optimization of the hydrogen positions. Nine residues are hydrogen-bonded to DNA bases and backbone groups throughout a 2 ns MD simulation.20 On average, the ‘H-bond density’ in protein–DNA complexes is one per 125 A˚2 of buried surface, corresponding to 18+6 ‘average’ H-bonds in this complex.19 Furthermore, if water molecules are present at the interface, they also may serve as donors or acceptors of additional H-bonds. Altogether, the enthalpy parsing analysis reconciles structural features of the complex with the observed heat capacity changes if one assumes that only 60 per cent of the complex interface is dehydrated.

10.6 Discussion Sequence-specific protein binding to DNA has been investigated by calorimetry in several systems and representative data have been discussed.5,21 It appears that significant negative heat capacity changes are the only common feature of several protein–DNA association reactions. This is reasonable since protein–DNA complexes bury a large amount of apolar molecular surface. As for other systems, DCp,A of the integrase–DNA complex is large and exceeds by far the expected heat capacity effect of surface dehydration. Interestingly, the plot of DHA versus T is curved. We propose that a non-linear change of association enthalpy with temperature is caused by the non-parallel and non-linear change of the heat capacity Cp of the free components and the complex before the main thermal transition. There are only two other protein–DNA systems that have been analysed so far in terms of heat capacity differences, namely, specific DNA targeting by

202

ENERGETICS OF SITE-SPECIFIC DNA RECOGNITION

the mouse SOX-5 HMG box and by the first three zinc fingers of the X. laevis transcription factor TFIIIA.4,5 In the case of SOX-5 binding, thermal fluctuations were significantly increased in the bound state, while the heat capacities of the free components added up almost exactly to the heat capacity of the zinc fingers–DNA complex between 0 and 408C. The different thermodynamic behaviour of the present and two previously analysed protein–DNA complexes indicates that the heat capacity increments caused by local folding/unfolding, by appearance of new vibrational modes and by redistribution of structural water molecules and ions, can be very differently balanced. We present a useful and promising way to deal with this complication, namely to link ITC measurements of protein–DNA association with the independent DSC analysis of the complex and its free components.

References 1. Dyson HJ and Wright PE. (2002) Curr. Opin. Struct. Biol. 12: 54–60. 2. Holbrook JA, Capp MW, Saecker RM and Record MT Jr. (1999) Biochemistry 38: 8409–8422. 3. Jelesarov I, Crane-Robinson C and Privalov PL. (1999) J. Mol. Biol. 294: 981–995. 4. Liggins JR and Privalov PL. (2000) Proteins Suppl. 4: 50–62. 5. Privalov PL, Jelesarov I, Read CM, Dragan AI and Crane-Robinson C. (1999) J. Mol. Biol. 294: 997–1013. 6. Wojciak JM, Connolly KM and Clubb RT. (1999) Nature Struct. Biol. 6: 366–373. 7. Gomez J, Hilser VJ, Xie D and Freire E. (1995) Proteins 22: 404–412. 8. Makhatadze GI and Privalov PL. (1995) Adv. Prot. Chem. 47: 307–425. 9. Hakin AW and Hedwig GR. (2001) Biophys. Chem. 89: 253–264. 10. Privalov PL and Makhatadze GI. (1990) J. Mol. Biol 213: 385–391. 11. Crane-Robinson C, Read CM, Cary PD, Driscoll PC, Dragan AI and Privalov PL. (1998) J. Mol. Biol. 281: 705–717. 12. Freire E. (1994) Methods Enzymol. 240: 502–530. 13. Wiseman T, Williston S, Brandts JF and Lin LN. (1989) Anal. Biochem. 179: 131–137. 14. Baker BM and Murphy KP. (1998) Methods Enzymol. 295: 294–315. 15. Freire E. (1989) Comments Mol. Cell Biophys. 6: 123–140. 16. Murphy KP and Freire E. (1992) Adv. Prot. Chem. 43: 313–361. 17. Spolar RS, Livingstone JR and Record MT, Jr. (1992) Biochemistry 31: 3947–3955. 18. Luque I and Freire E. (1998) Methods Enzymol. 295: 100–127. 19. Nadassy K, Wodak SJ and Janin J. (1999) Biochemistry 38: 1999–2017. 20. Gorfe AA, Caflish A and Jelesarov I. (2004) J. Mol. Recognit. 17: 121–131. 21. Jen-Jacobson L, Engler LE and Jacobson LA. (2000) Structure 8: 1915–1923.

11 Linkage Between Temperature and Chemical Denaturant Effects on Protein Stability: the Interpretation of Calorimetrically Determined m Values Beatriz Ibarra-Molero, Raul Perez-Jimenez, Raquel Godoy-Ruiz and Jose M. Sanchez-Ruiz

11.1 Introduction Two experimental approaches are commonly employed to characterize the thermodynamic stability of proteins: (1) thermal denaturation studies, often monitored by differential scanning calorimetry (DSC); (2) chemical denaturation studies, in which denaturation is induced by urea or guanidine. In principle, DSC studies on thermal denaturation offer several advantages. Provided that the DSC profiles accurately reflect an equilibrium denaturation process, the equilibrium thermodynamics analysis may lead to the number and thermodynamic parameters (Gibbs energy, enthalpy, entropy, heat capacity) of the protein states significantly populated during denaturation.1–5 On the other hand, analysis of chemical denaturation profiles (value of a physical property versus urea or guanidine concentration)6,7 is to a large Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

204

TEMPERATURE AND CHEMICAL DENATURATION OF PROTEINS

extent empirical and based upon extrathermodynamic assumptions. Thus, evidence supporting two-state denaturation must often be derived from the comparison of several denaturation profiles obtained using different physical properties or from direct measurements of the concentration of native protein using time-consuming double-jump assays.8–10 In addition, the value of the denaturation Gibbs energy change (DG) in the absence of denaturant is usually calculated by assuming that the denaturant concentration dependence of DG is linear outside the narrow transition zone. This socalled linear extrapolation method (LEM) has been the subject of considerable debate in the literature.9,11–30 Despite the obvious advantages of the thermal denaturation DSC approach, it must be recognized that one of the parameters derived from chemical denaturation profiles has found widespread application in protein folding studies: the slope of the DG versus denaturant concentration plot, referred to as the m value,11 which has been found empirically to be related to the amount of protein surface exposed upon denaturation.30,31 On this basis, equilibrium m values derived from chemical denaturation profiles have been used as probes of solvent exposure and residual structure in protein denatured states16,32–35 and, combined with kinetic m values, have provided information about transition states in protein folding processes.9,36–38 The main purposes of this work are (a) to show that denaturant m values can also be obtained from DSC experiments using a procedure that is straightforward and, to a large extent, model independent and (b) to discuss briefly the interpretation of these calorimetrically determined m values. We will focus mainly on two-state equilibrium transitions; we will show, however, that our approach can be easily extended to deal with multistate equilibrium denaturation. On the other hand, we will not consider here kinetic distortions of the DSC thermograms, either associated with irreversibility4,39,40 or with slow equilibrium.4,41–45 In connection with the latter possibility, it must be recalled that plots of logarithm of folding–unfolding rate constant versus denaturant concentration display a characteristic ‘V’ shape, usually referred to as a ‘chevron’; thus, folding–unfolding may in some cases become very slow at denaturant concentrations corresponding to the ‘bottom’ of the chevron plot.9,10 We will assume in the theoretical treatments given below that any potential ‘slow equilibrium’ distortions have been analysed and corrected for using the known procedures.4,41 Finally, we will assume that experiments have been performed in a pH region in which the pH effect on denaturation energetics is small or negligible; accordingly, we will not need to concern ourselves with the effect of denaturant on the pK values or residues critical for stability or with uncertainties associated with the definition of the pH scale in chemical denaturant solutions.46,47

205

LINKAGE

11.2 Linkage between temperature and chemical denaturant effects on protein stability The linkage between temperature and chemical denaturant effects on protein stability has been addressed in several articles in the literature.42,48,49 Here, we will follow the approach of Ibarra-Molero and Sanchez-Ruiz9 (see also reference 50). Assume a two-state, equilibrium denaturation process, N$D

ð11:1Þ

where N is the native state and D is the denatured (i.e. unfolded to an unspecified extent) state. All thermodynamic changes for the process are taken to be functions of, both temperature (T) and chemical denaturant concentration (C); thus, the denaturation change in Gibbs energy is expressed as DGðC, TÞ

ð11:2Þ

The values of T and C for which DG ¼ 0 (and, consequently, the equilibrium constant for the process is unity) define an equilibrium line in C–T or T–C plots. Note that this equilibrium line can be viewed in two entirely equivalent ways: (1) as the effect of denaturant concentration on the denaturation temperature (effect of C on Tm ); (2) as the effect of temperature on the midpoint denaturant concentration (effect of T on C1=2 ). The partial derivatives of DG with respect to T and C give, respectively, the denaturation entropy change and the denaturant m value, 
@DG ¼
DSðC, TÞ ð11:3Þ @T C 
@DG ¼
mðC, TÞ ð11:4Þ @C T both of which are functions of C and T, as indicated in the right-hand-sides of Equations (11.3) and (11.4). We will refer to m values corresponding to the equilibrium line (i.e., to C and T conditions for which DG ¼ 0) as m1=2 values. Likewise, denaturation entropy changes corresponding to equilibrium line conditions will be termed DSm ; they are given by DSm ¼ DHm =Tm

ð11:5Þ

where DHm is the denaturation enthalpy change at the same equilibrium line conditions. The denaturant concentration effect on DG (for a given temperature) can be expressed as a Taylor expansion from the C1=2 value:

206

TEMPERATURE AND CHEMICAL DENATURATION OF PROTEINS


DGðCÞ ¼ DGðC1=2 Þ þ

@DG @C



1 ðC
C1=2 Þ þ 2 C1=2




@ 2 DG @C 2

 ðC
C1=2 Þ2 þ . . . C1=2

ð11:6Þ or 1 DGðCÞ ¼
m1=2 ðC
C1=2 Þ þ 2




@ 2 DG @C 2

 ðC
C1=2 Þ2 þ . . .

ð11:7Þ

C1=2

since C1=2 implies equilibrium-line conditions and, hence, DGðC1=2 Þ ¼ 0 and
ð@DG=@CÞ for C1=2 is an m1=2 value. In chemical denaturation experiments, DG is available only in a narrow denaturant concentration range around C1=2 . Within that narrow range we can safely neglect the higher order terms in the Taylor expansion and write DGðCÞ ¼
m1=2 ðC
C1=2 Þ

ð11:8Þ

It is important to note that the above equation can be considered as virtually exact within the narrow concentration range of the chemical denaturant induced transition. It is also clear that m values determined from chemical denaturation experimental profiles can be rigorously considered as equilibrium-line m1=2 values (i.e. m values corresponding to T, C1=2 conditions under which DG ¼ 0). On the other hand, Equation (11.8) is often assumed to hold outside the narrow range of the transition; thus, substituting C ¼ 0 into Equation (11.8) yields the linear extrapolation method estimate of the denaturation Gibbs energy change in the absence of denaturant: ¼ m1=2 C1=2 DGLEM W

ð11:9Þ

Consider now a series of DSC experiments carried out with different concentrations of the chemical denaturant. Straightforward application of the mathematics of partial differentiation (for a review, see reference 51) leads to the following expression for the effect of denaturant concentration on denaturation temperature: 
m1=2 @T dTm ð@DG=@CÞT ¼
¼ ¼
ð11:10Þ dC ð@DG=@TÞC DSm @C DG¼0 where we have used Equations (11.3) and (11.4) and we have taken into account that, since at Tm the value of DG is zero, the m and DS values in the above equation belong to the equilibrium line (hence, they are denoted m1=2 and DSm ). Using now Equation (11.5) and solving for m1=2 we arrive at 
DHm dTm ð11:11Þ m1=2 ¼
Tm dC

LINKAGE

207

which allows us to calculate m1=2 values from experimental DSC profiles obtained at different denaturant concentrations. Several important features of Equation (11.11) must be emphasized. (1) The denaturant m value obtained using Equation (11.11) belongs to the equilibrium line (m1=2 value). (2) Equation (11.11) is rigorous (for two-state equilibrium denaturation) and, in particular, is not based on the linear extrapolation approximation (LEM would only be invoked if the m1=2 were used in Equation (11.9) to estimate DG at zero denaturant concentration). (3) The calculation of m1=2 from Equation (11.11) only requires values of the denaturation enthalpy change (equilibrium-line DHm values) and the effect of denaturant concentration on denaturation temperature (so that the derivative dTm =dC can be computed). In particular, the value of the denaturation heat capacity change is not required. (4) Equation (11.11) can be used to calculate m1=2 in the limit of zero denaturant concentration (that value would also correspond to a temperature equal to the Tm value for C ¼ 0). Thus, the slope dTm =dC can be calculated for C ¼ 0 and combined with the DHm and Tm values obtained in the absence of denaturant to yield m1=2 at C ¼ 0. (5) Although Equation (11.11) has been derived for a two-state process, it can also be applied to multistate equilibrium denaturation provided that the appropriate characteristic temperatures and transition enthalpies are used. Assume a DSC transition (not necessarily a two-state transition) with clearly defined high temperature and low temperature baselines. We will refer here to the states populated at low and high temperature as A and B states, respectively (for instance, A could be the native state, N, and B the denatured state, D). As we have recently pointed out,52 the characteristic temperature for the conversion A$B (temperature at which the equilibrium constant for the process is unity) can be derived from the experimental DSC data without assuming any specific denaturation model. Thus, partition functions taking states A and B as reference (QA and QB ) can be calculated as a function of temperature using the Freire–Biltonen double-integration procedure1 and the characteristic temperature is calculated as the temperature at which QA ¼ QB (for further details, see reference 52). Characteristic temperatures and total enthalpies for the transition as a function of denaturant concentration can be used in conjunction with Equation (11.11) to obtain the m1=2 values. In general, for a complex DSC profile that has been interpreted in terms of several transitions, Equation (11.11) can be used to obtain the m1=2 value for each individual transition, provided that the corresponding transition temperatures and enthalpies are used.

208

TEMPERATURE AND CHEMICAL DENATURATION OF PROTEINS

11.3 Calorimetrically determined urea m values The calorimetric determination of urea m values will be illustrated using the extensive DSC data published by Johnson and Fersht24 for barnase. Figure 11.1 gives a summary of the experimental parameters relevant to the m1=2 calculation: the denaturation temperature versus urea concentration profile (the equilibrium line, lower panel) and the denaturation enthalpy values along the equilibrium line (the DHm values, upper panel). Note that the DHm values shown correspond to different temperatures and urea concentration conditions, but, in all cases, conditions under which DG ¼ 0. To highlight this fact we give in the figure the urea concentrations for some of the

Figure 11.1 Energetic parameters for barnase (taken from Johnson and Fersht24), HEW lysozyme and ribonuclease A (taken from Makhatadze and Privalov48) used in the calculation of the urea m values given in Figure 11.2

CALORIMETRICALLY DETERMINED UREA M VALUES

209

DHm values. The dependence of Tm on urea concentration has been fitted with a second-order polynomial (lower panel), from which the values of the slope dTm =dC have been calculated (see the inset). In the case of lysozyme and ribonuclease A, this slope was calculated assuming a linear dependence of Tm with C within the 0–1 M urea concentration range. These slopes are then combined with the corresponding Tm and DHm values to yield (using Equation (11.11)) the m1=2 values shown in Figure 11.2. Note that, although these m1=2 values are plotted in Figure 11.2 versus urea concentration, they belong to the equilibrium line; that is, they correspond to different C and T conditions (but, in all cases, to C and T conditions under which DG ¼ 0). To highlight this fact, we give in Figure 11.2 the temperatures for some selected m1=2 values. We also include in Figure 11.2 a urea m value for barnase denaturation at 258C, determined from the ‘traditional’ chemical denaturation experiment.24 Note that this is also an equilibrium-line m1=2 value, as it corresponds to the midpoint urea concentration at 258C (C1=2 ¼ 4:6 M). There is an excellent agreement between the calorimetrically determined m1=2 values and the value obtained from chemical denaturation. Furthermore, it appears that m1=2 changes only slightly along the equilibrium line and thus the m1=2 value at C ¼ 0 M and T ¼ 54:98C (m1=2 ¼ 9:2 kJ mol
1 M
1 ) is close to the value at C ¼ 4:6 M and 258C (m1=2 ¼ 8 kJ mol
1 M
1 ). This behaviour is also observed for HEW lysozyme and ribonuclease A: the m1=2 values at C ¼ 0 and high T (calculated from the DSC data reported by Makhatadze and Privalov48; see legends to Figures 11.1 and 11.2 for details) are close to the values at high C and 258C determined from chemical denaturation experiments.53

Figure 11.2 Closed symbols: calorimetrically determined urea m1=2 values for barnase, lysozyme and ribonuclease A (i.e. calculated using the energetic parameters given in Figure 11.1 and Equation (11.11)). Open symbols: urea m1=2 values obtained from chemical denaturation experiments for barnase (taken from Johnson and Fersht24), lysozyme and ribonuclease A (taken from Ahmad and Bigelow53)

210

TEMPERATURE AND CHEMICAL DENATURATION OF PROTEINS

11.4 Calorimetrically determined guanidine m values The calorimetric determination of guanidine m values will be illustrated using the DSC data published for HEW lysozyme9 and ubiquitin.26 Figure 11.3 shows the Tm versus guanidine concentration profiles for lysozyme (pH 4.5) and ubiquitin (pH 2), together with the corresponding denaturation enthalpy values along the equilibrium lines (DHm values, upper panel). The polynomial fittings used to calculate the values of the derivative dTm =dC are also shown in Figure 11.3. In the case of ubiquitin we could not find a satisfactory polynomial fit to the entire Tm range and we used piecewise polynomial fits, as shown in the figure. These values, together with the corresponding values of Tm and DHm , lead (Equation (11.11)) to the m1=2 values given in Figure 11.4. Note that, although the guanidine m1=2 values shown are plotted versus guanidine concentration, they belong to the equilibrium lines (DG ¼ 0 lines) and, accordingly, correspond to different guanidine concentrations and

Figure 11.3 Energetic parameters for HEW lysozyme (data at pH 4.5 taken from IbarraMolero and Sanchez-Ruiz9) and ubiquitin (data at pH 2 taken from Ibarra-Molero et al.26) used in the calculation of the guanidine m values given in Figure 11.4

CALORIMETRICALLY DETERMINED GUANIDINE M VALUES

211

Figure 11.4 Closed symbols: calorimetrically determined guanidine m1=2 values for HEW lysozyme and ubiquitin (i.e. calculated using the energetic parameters given in Figure 11.3 and Equation (11.11)). Open symbols: guanidine m1=2 values obtained from chemical denaturation experiments for HEW lysozyme (data at pH 4.5 taken from Ibarra-Molero and Sanchez-Ruiz9) and ubiquitin (data at pH 2 taken from Ibarra-Molero et al.26)

temperatures. To highlight this fact we give in the figure the temperatures for some of m1=2 values. The inset shows the entire range of m1=2 values for ubiquitin, including the values at very low guanidine concentration, which are very large in absolute value and negative. As was the case with the urea m values (Figure 11.2), we find here an excellent agreement between the calorimetrically determined m1=2 and those derived from chemical denaturation experiments (Figure 11.4). Unlike the urea case, however, we find that m1=2 changes strongly along the equilibrium line, in particular for guanidine concentrations below approximately 1 M (note that, for ubiquitin at pH 2, the m1=2 value becomes very large in absolute value and negative for very low guanidine concentrations). Since m is the (minus) slope of the DG versus C dependence (Equation (11.4)), the abrupt changes in m1=2 below *1 M guanidine very likely correspond to strong deviations from linearity in plots of DG versus C at low guanidine concentration. As we have discussed previously,9,26 these deviations from linearity may reflect the contribution to DG that arises from charge–charge electrostatic interactions; this contribution is expected to be screened out at moderate concentrations of guanidine (about 1 M, according to the data

212

TEMPERATURE AND CHEMICAL DENATURATION OF PROTEINS

shown in Figure 11.4), since this denaturant is a salt (guanidinium chloride). Note that the DG contribution arising from charge–charge interactions may be positive or negative, depending on the pH value and the charge distribution on the protein surface,9,26,54 which rationalizes the different behaviour observed in Figure 11.4 for lysozyme at pH 4.5 and ubiquitin at pH 2 (for further discussion see reference 26).

11.5 Concluding remarks We have shown that denaturant m values can be calculated from DSC experiments in a straightforward and, to a large extent, model-independent manner. The calculation requires knowledge of the effect of denaturant concentration on denaturation temperatures and enthalpies. It does not require, however, values of denaturation heat capacity change. The illustrative calculations reported here suggest the following. . The calorimetrically determined urea m values change little along the temperature–concentration equilibrium line and, consequently, can be interpreted on the same basis as the ‘traditional’ m values derived from chemical denaturation experiments.31 Thus, they may potentially provide information about solvent exposure and residual structure for states populated during thermal denaturation. . The calorimetrically determined guanidine m values may change strongly along the equilibrium line, in particular for low guanidine concentration (below approximately 1 M guanidine). These abrupt changes may provide information regarding the contribution to protein stability that arises from charge–charge electrostatic interactions. This contribution is of considerable interest, since recent work suggests that it can be rationally modified to obtain proteins of enhanced stability.26,55–62

Acknowledgement Research in the authors’ laboratory is supported by grant BIO2000-1437 from the Spanish Ministry of Science and Technology.

References 1. Freire E and Biltonen RL. (1978) Biopolymers 17: 463–479. 2. Privalov PL. (1982) Adv. Prot. Chem. 35: 1–104.

REFERENCES

213

3. Sturtevant JM. (1987) Annu. Rev. Phys. Chem. 38: 463–488. 4. Sanchez-Ruiz JM. (1995) In Subcellular Biochemistry, Volume 24. Proteins: Structure, Function, and Engineering, Biswas BB and Roy S (eds), Plenum, New York, pp. 133– 176. 5. Freire E. (1995) Methods Enzymol. 259: 144–168. 6. Pace CN. (1986) Methods Enzymol. 131: 266–280. 7. Pace CN, Shirley BA and Thomson JA. (1989) In Protein Structure, a Practical Approach, Creighton TE (ed.), IRL Press at Oxford University Press, Oxford, pp. 311– 330. 8. Mu¨cke M and Schmid FX. (1994) Biochemistry 33: 12 930–12 935. 9. Ibarra-Molero B and Sanchez-Ruiz JM. (1996) Biochemistry 35: 14 689–14 702. 10. Ibarra-Molero B and Sanchez-Ruiz JM. (1997) Biochemistry 36: 9616–9624. 11. Greene RF and Pace CN. (1974) J. Biol. Chem. 249: 5388–5393. 12. Schellman JA. (1978) Biopolymers 17: 1305–1322. 13. Schellman JA. (1987) Annu. Rev. Biophys. Chem. 16: 115–137. 14. Santoro MM and Bolen DW. (1988) Biochemistry 27: 8063–8068. 15. Bolen DW and Santoro MM. (1988) Biochemistry 27: 8069–8074. 16. Pace CN, Laurents DV and Thomson JA. (1990) Biochemistry 29: 2564–2572. 17. Alonso DO and Dill KA. (1991) Biochemistry 30: 5974–5985. 18. Santoro MM and Bolen DW. (1992) Biochemistry 31: 4901–4907. 19. Hu CQ, Sturtevant JM, Thomson JA, Erickson RE and Pace CN. (1992) Biochemistry 31: 4876–4882. 20. Hagihara Y, Aimoto A, Fink AL and Goto Y. (1993) J. Mol. Biol. 231: 180–184. 21. Mayr LM and Schmid FX. (1993) Biochemistry 32: 7994–7998. 22. Ahmad F, Taneja S, Yavad S and Haque SE. (1994) J. Biochem. 115: 322–327. 23. Monera OD, Kay CM and Hodges RS. (1994) Protein Sci. 3: 1984–1991. 24. Johnson MJ and Fersht AR. (1995) Biochemistry 34: 6795–6804. 25. Yao M and Bolen DW. (1995) Biochemistry 34: 3771–3781. 26. Ibarra-Molero B, Loladze VV, Makhatadze GI and Sanchez-Ruiz JM. (1999) Biochemistry 38: 8138–8149. 27. Makhatadze GI. (1999) J. Phys. Chem. B 103: 4781–4785. 28. Pace CN and Shaw KL. (2000) Proteins: Struct. Funct. Genet. 41(S4): 1–7. 29. Bolen DW and Yang. (2000) Biochemistry 39: 15 208–15 216. 30. Courtenay ES, Capp MW, Saecker RM and Record MT. (2000) Proteins: Struct. Funct. Genet. 41(S4): 72–85. 31. Myers JK, Pace CN and Scholtz JM. (1995) Protein Sci. 4: 2138–2148. 32. Pace CN, Laurents DV and Erickson RE. (1992) Biochemistry 31: 2728–2734. 33. Shortle D. (1995) Adv. Protein Chem. 46: 217–247. 34. Wrabl J and Shortle D. (1999) Nat. Struct. Biol. 6: 876–883. 35. Pace CN, Alston RW and Shaw KL. (2000) Protein Sci. 9: 1395–1398. 36. Matouscheck A and Fersht AR. (1993) Proc. Natl Acad. Sci. USA 90: 7814–7818. 37. Matthews JM and Fersht AR. (1995) Biochemistry 34: 6805–6814. 38. Matouscheck A, Otzen DE, Itzhaki LS, Jackson SE and Fersht AR. (1995) Biochemistry 34: 13 656–13 662. 39. Sanchez-Ruiz JM. (1992) Biophys. J. 61: 921–935. 40. Plaza del Pino IM, Ibarra-Molero B and Sanchez-Ruiz JM. (2000) Proteins: Struct. Funct. Genet. 40: 58–70. 41. Freire E, van Osdol WW, Mayorga OL and Sanchez-Ruiz JM. (1990) Annu. Rev. Biophys. Biophys. Chem. 19: 159–188. 42. Plaza del Pino IM, Pace CN and Freire E. (1992) Biochemistry 31: 11 196–11 202.

214 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62.

TEMPERATURE AND CHEMICAL DENATURATION OF PROTEINS

Yu Y, Makhatadze GI, Pace CN and Privalov PL. (1994) Biochemistry 33: 3312–3319. Rosengarth A, Rosgen H and Hinz HJ. (1999) Eur. J. Biochem. 264: 989–995. Kaushik JK, Ogasahara K and Yutani K. (2002) J. Mol. Biol. 316: 991–1003. Garcia-Mira MM and Sanchez-Ruiz JM. (2001) Biophys. J. 81: 3489–3502. Acevedo O, Guzman-Casado M, Garcia-Mira MM, Ibarra-Molero B and SanchezRuiz JM. (2002) Anal. Biochem. 306: 158–161. Makhatadze GI and Privalov PL. (1992) J. Mol. Biol. 226: 491–505. Nicholson EM and Scholtz JM. (1996) Biochemistry 35: 11 369–11 378. Plaza del Pino IM and Sanchez-Ruiz JM. (1995) Biochemistry 34: 8621–8630. Blinder SM. (1966) J. Chem. Ed. 43: 85–92. Tho´rolfsso´n M, Ibarra-Molero B, Fojan P, Petersen SB, Sanchez-Ruiz JM and Martinez A. (2002) Biochemistry 41: 7573–7585. Ahmad F and Bigelow CC. (1982) J. Biol. Chem. 257: 12 935–12 938. Yang AS and Honig B. (1993) J. Mol. Biol. 231: 459–474. Loladze VV, Ibarra-Molero B, Sanchez-Ruiz JM and Makhatadze GI. (1999) Biochemistry 38: 16 419–16 423. Grimsley GR, Shaw KL, Fee LR, Alston RW, Huyghues-Despointes BM, Thurlkill RL, Scholtz JM and Pace CN. (1999) Protein Sci. 8: 1843–1849. Spector S, Wang M, Carp SA, Roblee J, Hendsch ZS, Fairman R, Tidor B and Raleigh DP. (2000) Biochemistry 39: 872–879. Perl D, Mueller U, Heinemann U and Schmid FX. (2000) Nat. Struct. Biol. 7: 380–383. Pace CN. (2000) Nat. Struct. Biol. 7: 345–346. Sanchez-Ruiz JM and Makhatadze GI. (2001) Trends Biotechnol. 19: 132–135. Perl D and Schmid FX. (2001) J. Mol. Biol. 313: 343–357. Ibarra-Molero B and Sanchez-Ruiz JM. (2002) J. Phys. Chem. B 106: 6609–6613.

12 Thermodynamic Indications of the Molten Globule State of Cytochrome c Induced by Hydrophobic Salts Ali A. Moosavi-Movahedi and Jamshid Chamani

12.1 Introduction The first investigations into the different states of proteins date as far back as the 1920s, when Hsien Wu devised one of the first theories of protein denaturation.1,2 There are still large gaps in our current understanding of the mechanisms by which proteins fold. For a long time it appeared that the folding of small single-domain proteins could be approximated to a two-state cooperative transition between an ensemble of unfolded state and the folded state. However, it soon became apparent that intermediate states exist in many instances, and their properties were seen to be common to many proteins. Protein folding/unfolding is generally a highly cooperative process in which only the completely folded and unfolded states become significantly populated. Under certain circumstances, however, several proteins have been shown to exhibit a folding intermediate known as a ‘molten globule’.3–8 The molten globule (MG) intermediates found for different proteins appear to have some common features, primarily a native-like secondary structure, a high degree of compactness, and a disrupted tertiary structure.4,9,10 The MG state has also been shown to exhibit a rotational correlation time and viscosity Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

216

THERMODYNAMICS OF MG STATE OF CYTOCHROME

close to that of the native state, indicating that the molten globule is highly compact.11 Ewbank and Creighton have shown that the MG state is compatible with a variety of disulfide bonding pairings, suggesting the absence of many specific tertiary structure interactions.8 The relevance of this intermediate state to the folding reaction was supported by the demonstration that it could be populated in both equilibrium and kinetic experiments. However the significance of the molten globule has been put into question.12–15 It has been argued that folding intermediates investigated so far are either partly folded protein,12 i.e. they contain several domains and one domain folds to the fully folded state before another (as a sequence of equivalent thermodynamic steps), or they form misfolded structures, which can be considered to be ‘off-pathway’ species.15,16 A large part of the folding literature has recently been dedicated to defining the nature of a folding intermediate in an attempt to resolve these arguments. It is important to elucidate the structure and the stabilizing mechanism of the MG state as an intermediate between the native and denatured states, in order to understand the principles of constructing a three-dimensional protein structure. X-ray angle scattering studies have shown that the MG states of various proteins have a wide range of structures from relatively disordered to highly ordered.17–19 This implies that the MG state is a largely fluctuating ensemble of various energy minima. Moreover, the stability of the MG state is determined by a delicate balance of interactions such as electrostatic repulsion between the charged residues and opposing forces such as hydrophobic interaction. A significant influence of salts or charges on the stability of the MG state reveals that the main driving force of the MG state is the reduction of the electrostatic repulsion between the charged residues that favour the unfolded conformations.20–22 There is a lack of substantial evidence regarding the contribution of the hydrophobic interactions to the stability of the MG state. However, such interactions have been suggested for the positive heat capacity changes of the thermal unfolding of the MG state of apomyoglobin23,24 and cytochrome c.25–27 Realizing which interactions are responsible for the stabilization of the MG state is crucial to an understanding of specificity in the protein folding.28–30 A unifying observation emerging from a number of detailed studies on the better characterized MG states from a-helix proteins such as cytochrome c, myoglobin, a-lactalbumin, or staphylococcal nuclease is that their folding process intermediates, at least under equilibrium conditions, can be made to populate a number of different structural states by altering the solvent conditions or ligand binding effects. The MG state can be obtained from two processes. When native protein is unfolded by different kinds of ligand or solvent perturbation, it goes through a transition state that is called the MG; in the other process unfolded protein is refolded at different conditions and can be shown to be an MG state.31–33 The effects of various polyols, sugars, and so on were investigated on the structure of the acid

INTRODUCTION

217

unfolded state of cytochrome c at pH 2 by several techniques, in order to address the contribution of hydrophobic interactions to the stability of the MG of proteins.27,34 The usual techniques utilized to recognize the MG state of proteins are circular dichroism (CD) and visible and fluorescence spectroscopies. (1) Circular dichroism (CD). Far UV CD spectra determine the secondary structure contents, whereas near UV CD spectra changes show the tertiary structure disruption. It is important to note that the MG state has a similar secondary structure to native state but its tertiary structure curve in near UV CD is different from the native form. For example, the interaction between n-octyl-b-D-glucopyranoside (octyl glucoside) as a non-ionic surfactant and bovine liver glutamate dehydrogenase (GDH) showed that this protein retains its secondary structure in the presence of octyl glucoside, but loses a degree of its tertiary structure by acquiring a more extended tertiary structure.35 (2) Visible spectroscopy. Visible spectroscopy has been used for determining the MG states of haem proteins, such as cytochrome c, myoglobin, horseradish peroxidase, and apomyoglobin.36–39 The haem absorption of protein in the vicinity of  400 nm reflects the spin state of the iron, which is dependent on the conformational state of the protein. The distinction of visible spectra of native, acid unfolded, and MG states can be manifested via the effect of salts and n-alkyl sulfates (hydrophobic salts) on cytochrome c in acidic condition.40,41 (3) Spectrofluorimetry. The MG state, because of its increased surface hydrophobicity, binds nonpolar molecules in solution more strongly than the native protein. The binding of the hydrophobic fluorescence probe ANS (8-anilino naphthalene sulfunate) to the MG state is an easy test with which to prove the manifestation of the described state. Since this probe binds to solvent-accessible clusters of nonpolar groups in the native state,42 it can also bind more strongly to the MG.43,44 For example the interaction of ANS with GDH in the presence of various concentrations of octyl glucoside was investigated by spectrofluorimetry. It is clear that the interaction between ANS and the native state of GDH is accompanied by fluorescence enhancement as well as by a considerable blue shift in lmax .35 It is known that GDH has at least one hydrophobic patch, which serves as a part of its binding site for the reduced nicotinamide ring.45 Thus, it seems that ANS binds to the hydrophobic path in the native state of the protein (as judged by the fluorescence enhancement), and the presence of octyl glucoside induces conformational changes in the GDH tertiary structure, which extends its hydrophobic clusters; consequently, ANS binding is promoted.

218

THERMODYNAMICS OF MG STATE OF CYTOCHROME

Accordingly, a few studies have been reported on the thermodynamic characterization of the MG state of proteins. Jennifer et al. showed that the helix–helix interaction stabilizes the MG state and the native state to the same degree. They determined that the heat capacity change for MG state denaturation is approximately 60 per cent of the value for native state denaturation, indicating that the MG state interior is native-like.46 On the other hand, the enthalpy change of the native to MG transition is similar to the enthalpy change of the transition of native a-lactalbumin to the completely unfolded form.47–53 No thermal transition of the MG state could be found in the literature by means of differential scanning microcalorimetry. The main goal of this chapter is to provide an overview of the thermodynamic characterization of the MG state of cytochrome c, which is induced by n-alkyl sulfates as hydrophobic salts.

12.2 Thermodynamic description of molten globule state The complex and interrelated reactions that make up living processes are dependent upon the presence of proteins, not only as catalysts, but also as structural molecules, as storage and carrier molecules, and as molecular motors. All of these functions require that the nascent amino-acid chain correctly folds into the biologically active, three-dimensional structure of the native state. In pioneering studies, Anfinsen showed that all the necessary information for the nascent chain to fold into the native structure is contained in the sequence of amino acids.54 In order for proteins to fold spontaneously into their native state, the native state must be lower in Gibbs energy than the unfolded state. From this observation, Anfinsen proposed that the native and unfolded states are in equilibrium, and that the native state is the global minimum in the Gibbs energy that can be obtained without breaking covalent bonds. This assumption is known as the thermodynamic hypothesis, because it suggests that protein folding is under thermodynamic rather than kinetic control. The MG state can be induced by salts, n-alkyl sulfates as hydrophobic salts, and polyols from the acid unfolded state of cytochrome c. If the MG and acid unfolded states are in equilibrium, the contribution of various interactions to these thermodynamic quantities can also be explored as a mean of understanding how they contribute to molten globule stability. The equilibrium between the acid unfolded and MG states is defined by the equilibrium constant, K, as K ¼ ½MG=½U

ð12:1Þ

The difference in free energy between the MG and acid unfolded states is based on Pace theory:55

THERMODYNAMIC DESCRIPTION OF MOLTEN GLOBULE STATE

DG8 ¼
RT ln K ¼
RT lnðYobs
YU Þ=ðYMG
Yobs Þ

219

ð12:2Þ

where R is the universal gas constant, T is the absolute temperature in Kelvin, YU , YMG , and Yobs are the physical parameters such as extinction coefficient, molar ellipticity, and percentage of fluorescence of acid unfolded, MG, and any observed states respectively. The free energy is composed of both enthalpic and entropic contributions: DG8 ¼ DH8
T DS8

ð12:3Þ

where DH8 is the enthalpy change and DS8 is the entropy change upon refolding. However, as with other reactions that take place in water, both DH8 and DS8 are strongly temperature dependent, so the free energy change is better written as56 DG8 ¼ DH8R
TDS8R þ DCp fðT
TR Þ
T lnðT=TR Þg

ð12:4Þ

where the subscript R indicates the value of DH8 and DS8 at a reference temperature, TR ; DCp is the heat capacity change. The result of the significant DCp is that the free energy is not a linear function of temperature. Rather, it shows significant downward curvature and has a maximum value at some temperature, often near physiological conditions. The enthalpy change, DH8, corresponds to the binding energy (dispersion forces, electrostatic interactions, van der Waals potentials, and hydrogen binding), while hydrophobic interactions are described by the entropy term, DS8. Proteins become more stable with increasingly positive values of DG8, i.e. free energy of the unfolded protein (G8U ) relative to the free energy of the folded or native protein (G8N ). In other words, as the binding energy increases or the entropy difference between the two states decreases, the folded protein becomes more stable. The folded conformation of a domain is apparently in a relatively narrow free energy minimum and substantial perturbations of that folded conformation require a significant increase in free energy. The large heat capacity change upon protein unfolding causes there to be a temperature at which the stability of the folded state is at a maximum. Measured by free energy, the maximum occurs when DS8 ¼ 0, while that measured by the equilibrium constant occurs when DH8 ¼ 0. These maximum stabilities can occur at quite different temperatures, but both are used in different situations. Regardless of which one is used, however, the stability of the folded state decreases at both higher and lower temperatures. While factors such as binding interactions do obviously play a part in stabilizing the protein, they cannot account for a very significant portion of stabilization effects since similar phenomena occur in the unfolded state (although the interaction between folded protein and solvent would be expected to be stronger than the interaction between the unfolded protein coil and the solvent); the hydrophobic effect is probably the major stabilizing effect.57

220

THERMODYNAMICS OF MG STATE OF CYTOCHROME

The most common way to estimate the conformational stability of a protein is to tilt the equilibrium to conditions where the folded and unfolded states are almost equally populated, measure the stability, and extrapolate to obtain the stability under native conditions. The conformational stability of the MG state induced by n-alkyl sulfates can be measured from a sigmoidal curve (physical parameters such as molar ellipticity, absorbance, fluorescence intensity versus n-alkyl sulfates at low concentrations) and plot of DG8 against n-alkyl sulfate concentration (see Reference 41). The free energies of MG formation in the absence of n-alkyl sulfates, DG8(H2O), were calculated by the least-squares method from the following equation:55 DG8 ¼ DG8ðH2 OÞ
m½ligand

ð12:5Þ

where m is a measure of the dependence of DG8 on denaturant, and DG8(H2O) is an estimate of the conformational stability that assumes that the linear dependence of DG8 on denaturant observed in the transition region continues to 0 M denaturant. The equilibrium between the molten globule and unfolded states can be monitored by calorimetry,58 or by using spectroscopic techniques such as UV absorbance spectroscopy, fluorescence, and circular dichroism that monitor the changes in conformational states of a protein.59 There are several problems with these techniques. Traditional methods are usually unable to detect unfolded forms under ambient conditions because their concentrations are far too low. These methods for determining the conformational stability also depend on some basic limiting assumptions. They require that the folding transition of a protein be fully reversible, with a known and finite number of observable states. However, the Pace method55 is the best method for determining the stability of folded states in the protein folding process. One of the best criteria for determining the protein stability is free energy in the absence of ligand, DG8(H2O), and m-values. Table 12.1 shows the m-values and DG8(H2O) for the MG state of cytochrome c upon the addition of n-alkyl sulfates such as SOS (sodium octyl sulfate), SDeS (sodium decyl sulfate), SDS (sodium dodecyl sulfate), and STS (sodium tetradecyl sulfate). Table 12.1 shows the increment of DG8(H2O) and m-values that corresponds to the extent of the hydrophobic chains. Table 12.1 also indicates the appearance of a higher stabilized MG state, which corresponds to the interaction of cytochrome c with n-alkyl sulfates. Therefore, the formation of MG states of cytochrome c induced by n-alkyl sulfates is stabilized by the presence of hydrophobic tails of different lengths.41 In this chapter we analyse the stabilized form of the MG state of cytochrome c that is induced by n-alkyl sulfates as hydrophobic salts. This is evidence that hydrophobic interactions substantially contribute to the stabilization of the MG state, as expected from the positive heat capacity change of thermal unfolding.60,61

221

MOLTEN GLOBULE STRUCTURE DETECTION

Table 12.1 DG (H2O) and m-values for the MG state of cytochrome c at 208C upon interaction with n-alkyl sulfates. The data are taken from Reference 41 n-alkyl sulfates SOS SDeS SDS STS

DG (H2O) (kJ mol1 )

m (kJ mol1 M1 )

7:41  0:01 10:10  0:01 11:63  0:01 11:90  0:01

10.7 71.4 409.2 490.0

12.3 Molten globule structure detection by isothermal titration calorimetry (ITC) Isothermal titration calorimetry (ITC) has been very informative in the study of the interactions between biomacromolecules, proteins, and ligands.62–76 The increased sensitivity and reliability of ITC allows, under suitable conditions, the detection of a reaction with a small enthalpy change. ITC can provide an enthalpy change at each injection of ligand to protein. The enthalpy change includes different events that occur in the protein solution, i.e the enthalpy changes related to ligand binding, conformational change of biomacromolecules, ionization titratable groups, and so on. The ITC data can be used for interpretation of protein folding. Various basic proteins, including cytochrome c, are significantly unfolded at pH 2 in the absence of ligand, but addition of anions, from either salt or acid, causes stabilization of the compact MG state.22,77 The mechanism can be interpreted in terms of preferential binding of the anions to the compactly folded MG state compared with the expanded unfolded state.77 Hamada et al. reported, using horse cytochrome c, that ITC provides a reliable calorimetric enthalpy change for the formation of the salt induced MG state from the acid unfolded state. A number of proteins show salt-dependent conformational changes similar to that of cytochrome c.78 To understand the mechanism of protein folding, it is critically important to interpret the thermodynamic parameters of the molten globule state in comparison with the native state. In addition to salt, n-alkyl sulfates as hydrophobic salts can extend the stabilization of the MG states of cytochrome c from the acid unfolded state. Studies of the n-alkyl sulfate effects with identical polar heads but different nonpolar tails allow the determination of the contribution of the electrostatic and the hydrophobic forces in protein denaturation. Figure 12.1 shows the plots of calorimetry enthalpy changes (DH) versus concentration of n-alkyl sulfates (SOS, SDS, and STS) affecting the acid unfolded state of cytochrome c. This figure indicates the deepest minima (lowest negative values of DH) belong to SOS (1.2 mM), SDS (0.08 mM) and STS (0.05 mM) respectively.79

222

THERMODYNAMICS OF MG STATE OF CYTOCHROME

Figure 12.1 DH plotted against concentration of n-alkyl sulfates at pH 2 for cytochrome c, obtained after subtracting the enthalpy of n-alkyl sulfate dilution. DH values versus concentration of SOS (a), SDS (b), and STS (c). The figures were directly taken from Reference 79

Hagihara et al.77 proposed that, because anions can interact with positive charges of both the unfolded and MG states, the salt induced formation of the MG state should be explained in terms of preferential binding of anions to the compactly packed MG state compared with the expanded unfolded state. At

ENTHALPIC RELATIONSHIP WITH ELECTRON TRANSFER

223

pH 2, most titratable groups are protonated and hence the net charge of the MG state is essentially the same as that of the fully unfolded state at the same pH. However, because the anion binding arises from an electrostatic interaction, the compact MG state with higher charge density binds anions more tightly than the unfolded state. Thus the MG state is stabilized with an increase in salt concentration. In the presence of cytochrome c, the titration pattern indicated negative heat effects, in addition to the heat of dilution of nalkyl sulfates. These results suggested that the net exothermic heat change arose from the formation of the MG states and that they represented the enthalpy change for the formation of the MG states. Alkyl sulfates play a role like salts (hydrophobic salts), which have similar, negative polar heads with different hydrophobic tail groups. On one hand, their negative polar heads can interact with positive charges and decrease electrostatic charge repulsion.80 On the other hand, hydrophobic interactions between nonpolar tail groups of n-alkyl sulfates and amino-acid hydrophobic side-chains of the acid unfolded state of cytochrome c induce compaction states. Therefore, to understand the mechanism of protein folding, it is critically important to elucidate the thermodynamic mechanism responsible for the stability of the intermediate conformational states when compared with the native state. Accordingly, ITC becomes an useful tool for investigating the MG state in protein folding.

12.4 Enthalpic relationship with electron transfer of molten globule state One fruitful way of investigating redox properties of proteins is the electrochemical approach to electron transfer exchange. In this respect, it appears that the electroactivity of a protein is dependent on several factors such as the nature of the electrode, type of medium used, structure of the electroactive molecule, and extent of adsorption into the electrode material. The latter factor, in particular, seems to be essential to explain the overall exchange process.81 Unfolding can be considered as a prerequisite step of the electron exchange. From these considerations it emerges that the study of the folding–unfolding process of cytochrome c is fundamental to a better understanding of the electron transfer and, moreover, to the elucidation of interrelationships between protein structure and biological function. Figure 12.2 shows the linear plots between enthalpy change of calorimetry, ITC (DH), and current intensity (electron transfer) for cytochrome c upon interaction with SOS, SDS, and STS. As n-alkyl sulfates were added voltametric cathodic peaks were observed due to addition of n-alkyl sulfates to unfolded state of cytochrome c (see Figure 12.2(b)).41,82 As can be seen from

224

THERMODYNAMICS OF MG STATE OF CYTOCHROME

Figure 12.2(b), the values of electron transfer from the unfolded to MG state are significantly higher than values of electron transfer from MG to native state, as well as indicating an easier electron transfer occurring with the native state compared to other states.41,82 It is important to note that the induction of

Figure 12.2 (a) Electron transfer intensity versus enthalpy changes of isothermal titration calorimetry (DH) at various concentrations of n-alkyl sulfates. Arrows show the MG state indication, which is induced by SOS (1.2 mM, *), SDS (0.08 mM, ~) and STS (0.05 mM, &). (b) Voltamogram of cytochrome c at a Cys-gold modified electrode as a function of SDS concentration (as a prototype), in the conditions of 20 mM HCl, pH 2, at 208C. U, acid unfolded state; 1, 0.02 mM SDS; 2, 0.04 mM SDS; 3, 0.06 mM SDS; 4, 0.08 mM SDS; N, native state in 25 mM phosphate buffer at pH 7. The data were taken directly from Reference 82

INDICATION OF MOLTEN GLOBULE STATE

225

the MG states of cytochrome c by n-alkyl sulfates at low concentrations demonstrates a linear relationship between the higher value of exothermic calorimetric enthalpy (DH) and current intensity (electron transfer). This indicates that the MG states with different compactnesses (lowest value of exothermic enthalpy) reach near identical amounts of electron transfer.41,82

12.5 Indication of molten globule state by differential scanning calorimetry (DSC) Calorimetry acquires special significance in studies of temperature induced changes in the state of the protein, since temperature and enthalpy are coupled intensive and extensive variables, respectively. All temperature induced changes in macroscopic systems always proceed with a corresponding change of enthalpy; i.e., they are accompanied by heat absorption if the process is induced by a temperature increase, or by the evolution of heat if it is caused by a temperature decrease. The functional relation between enthalpy and temperature actually includes all the thermodynamic information on the macroscopic states accessible within the considered temperature range, and this information can be extracted from the enthalpic function by thermodynamic analysis. DSC has established itself as the prime technique for the study of the thermal stability of proteins, especially following the availability of ultrasensitive microcalorimeters and convenient deconvolution algorithms. Frequently, unfolding protein systems obey the two-state van’t Hoff model and hence can be described completely in terms of the thermodynamic properties of the folded and unfolded forms.83–86 There are two important forms of enthalpy as far as protein unfolding is concerned, the van’t Hoff enthalpy, from the temperature dependence of the equilibrium constant, DHVH , and that measured calorimetrically (the area under the peak), DHcal ; if these are equal it suggests there are no populated intermediates present at Tm , i.e. the system is a two-state one. Knowledge of the heat capacity value of the protein as a function of temperature permits the calculation of its molecular partition function and, through the appropriate algorithm, the deconvolution of the unfolding thermal profile into a two-state process.87,88 This analysis has already been applied to several proteins of various sizes and complexities, leading to the definition and characterization of submolecular cooperative blocks as structural domains of the macromolecule.86,89 One of the applications of DSC is determination of the conformational stability and also cooperativity of protein. The MG state is usually less stable and less cooperative relative to the native state of the protein. However, the DSC thermogram of the native state is a sharp heat absorption peak relative to the MG state.90 The Tm and enthalpy of unfolding state (DHcal ) values

226

THERMODYNAMICS OF MG STATE OF CYTOCHROME

obtained by DSC for the MG state of cytochrome c with eight acetylated amino groups and also induced by salts in acidic conditions confirm the lower stability and cooperativity of MG relative to native state.90 Here, Figure 12.3 shows the DSC curves belonging to the native state of cytochrome c and the MG state that is induced by SDS at low concentrations as a hydrophobic salt. The DSC curves demonstrate that the native form is more cooperative relative to the MG state. On the other hand, the transition point (Tm ) and calorimetric enthalpy change of unfolding (DHcal ) for the MG state are smaller than the native form and show the higher stability of the native structure of cytochrome c. Previously, it was reported in the literature that the thermal denaturation of haem proteins, including cytochrome c, myoglobin, and horseradish peroxidase, at pH 2 gave rise to a well defined DSC transition whose apparent Tm depended on the scan rate.91–95 Our experiments confirm the effect of the scan rate on the calorimetric profiles of cytochrome c that

Figure 12.3 DSC thermograms for the various conformational states of cytochrome c. U, acid unfolded state of cytochrome c at pH 2; MG, molten globule state of cytochrome c induced by SDS at pH 2; N, native state of cytochrome c at pH 7. The dashed curve shows the repeated scan of the MG state, that is keeping the same MG state sample solution in the cell. The data were directly taken from Reference 95

INDICATION OF MOLTEN GLOBULE STATE

227

corresponds to irreversible, kinetically controlled transitions.95 It is worth noting that the DSC curve for the MG state induced by SDS shows a reversible thermogram.95 This is advantageous for the MG state, allowing deconvolution of the thermogram for cytochrome c and more rigorous determination of the Tm and enthalpy of each energetic subdomain. In the acidic state of cytochrome c, haem is removed from the hydrophobic core prior to unfolding of the protein. The stability of the MG state is determined by a delicate balance of interactions such as electrostatic repulsion between charged residues and opposing forces such as hydrogen bonds and hydrophobic interaction.4,21 SDS is an amphipatic material and has a negative polar head. Addition of SDS reduces the electrostatic repulsion and reinforces the hydrophobic interaction twin at low concentrations, producing a net force favourable for MG formation. The spectroscopic data demonstrate that the addition of SDS to the acid unfolded state of cytochrome c cause haem to move to an essential place (data are not shown). Therefore, although the MG state of cytochrome c is less stable than the native state, its haem junction to the hydrophobic core is tight. It is noted that the haem plays an important role in the reversibility of the DSC thermogram, thus allowing deconvolution of the thermal profile of the MG state of cytochrome c induced by SDS and obtaining additional information about energetic subdomains for cytochrome c.95

Acknowledgement The financial support provided by the Research Council of the University of Tehran is gratefully appreciated.

References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

Edsall JT. (1995) Adv. Protein Chem. 45: 1–5. Wu H. (1995) Adv. Protein Chem. 46: 6–26. Kim PS and Baldwin RL. (1982) Annu. Rev. Biochem. 51: 459–489. Kuwajima K. (1989) Proteins: Struct. Funct. Genet. 6: 87–103. Baum J, Dobson CM, Evans PA and Hanley C. (1989) Biochemistry 28: 7–13. Hughson FM, Wright PE and Baldwin RL. (1990) Science 249: 1544–1548. Christensen H and Pain RH. (1991) Eur. Biophys. J. 19: 221–229. Ewbank JJ and Creighton T. (1991) Nature 350: 518–520. Hildebrandt P and Stockburger M. (1989) Biochemistry 28: 6710–6721. Ptitsyn OB. (1987) J. Protein Chem. 6: 272–293. Dolgika DA, Gilmanshin RI, Brazhnikov EV, Bychkova VE, Semisotnov GV, Venyaminov SY and Ptitsyn OB. (1981) FEBS Lett. 136: 311–315. 12. Privalov PL. (1996) J. Mol. Biol. 258: 707–725. 13. Ptitsyn OB. (1996) Nature Struct. Biol. 3: 488–490.

228

THERMODYNAMICS OF MG STATE OF CYTOCHROME

14. Sali A, Shakhnovich E and Karplus M. (1994) J. Mol. Biol. 235: 1614–1636. 15. Kiefhaber T. (1995) Proc. Natl Acad. Sci. USA 92: 9029–9033. 16. Sosnick TR, Mayne L, Hiller R and Englander SW. (1994) Nature Struct. Biol. 1: 149– 156. 17. Kataoka M, Hagihara Y, Mihata K and Goto Y. (1993) J. Mol. Biol. 229: 591–596. 18. Kataoka M, Nishii I, Fujisawa T, Ueki T, Tokunaga F and Goto Y. (1995) J. Mol. Biol. 249: 215–228. 19. Nishii I, Kataoka M, Tokunaga F and Goto Y. (1994) Biochemistry 33: 4903–4909. 20. Hamada D, Hoshino M, Kataoka M, Fink AL and Goto Y. (1993) Biochemistry 32: 10 351–10 358. 21. Goto Y and Fink AL. (1989) Biochemistry 28: 945–952. 22. Goto Y and Nishikori S. (1991) J. Mol. Biol. 222: 679–686. 23. Griko YV and Privalov PL. (1994) J. Mol. Biol. 235: 1318–1325. 24. Hagihara Y, Oobatake M and Goto Y. (1994) Protein Sci. 3: 1418–1429. 25. Potekhin S and Pfeil W. (1989) Biophys. Chem. 34: 55–62. 26. Yoshihisa H, Motohisa O and Goto Y. (1994) Protein Sci. 3: 1418–1429. 27. Kamiyama T, Sadahide Y, Nogusa Y and Gekko K. (1999) Biochim. Biophys. Acta 1434: 44–57. 28. Gasset M, Baldwin MA, Lloyd DH, Gabriel JM, Hotzman DM, Cohen F, Fletterick R and Prusiner SB. (1995) Proc. Natl Acad. Sci. USA 94: 10 940–10 945. 29. Zhange H, Kaneko K, Nguyen JT, Livshits TL, Baldwin MA, Cohen FE, James TL and Prusiner SB. (1995) J. Mol. Biol. 250: 514–526. 30. Wood SJ, Maleeff B, Hart T and Wetzel R. (1996) J. Mol. Biol. 256: 870–877. 31. Anil KL and Poonam K. (1992) J. Biol. Chem. 267: 19 914–19 918. 32. Greenfield NJ. (1996) Anal. Biochem. 235: 1–10. 33. Greenfield N and Easman GD. (1969) Biochemistry 8: 4108–4116. 34. Davis-Searles PR, Morar AS, Saonderes AJ, Erie DA and Pielak GJ. (1998) Biochemistry 37: 17 048–17 053. 35. Ghobadi S, Safarian S, Moosavi-Movahedi AA and Ranjbar B. (2001) J. Biochem. 130: 671–677. 36. Goto Y, Hagihara Y, Hamada D, Hoshino M and Nishii I. (1993) Biochemistry 32: 11 878–11 885. 37. Tsaprailis G, Wing Sze CD and English AM. (1998) Biochemistry 37: 2004–2016. 38. Bismuto E, Sirangelo I and Irace G. (1992) Arch. Biochem. Biophys. 298: 624–629. 39. Nishii I, Kataoka M and Goto Y. (1995) J. Mol. Biol. 250: 223–238. 40. Goto Y, Takahashi N and Fink AL. (1990) Biochemistry 29: 3480–3487. 41. Moosavi-Movahedi AA, Chamani J, Goto Y and Hakimelahi GH. (2003) J. Biochem. 133: 93–102. 42. Stryer L. (1965) J. Mol. Biol. 13: 482–495. 43. Semisotnov GV, Rodinova NA, Razgulyaev OI, Uversky VN, Gripas AF and Gilmanshin RI. (1991) Biopolymers 31: 119–128. 44. Semisotnov GV, Rodinova NA, Kutyshenko VP, Ebert B, Blank J and Ptitsyn OB. (1987) FEBS Lett. 198: 9–13. 45. Fisher HF. (1985) Methods Enzymol. 113: 16–27. 46. Jennifer LM, Melissa L and Pielak GJ. (1998) J. Mol. Biol. 275: 379–388. 47. Ohgushi M and Wada A. (1983) FEBS Lett. 164: 21–24. 48. Markovic-Housley Z and Garavito RM. (1986) Biochim. Biophys. Acta 869: 158–170. 49. Dejongh HHJ, Killiam JA and de Kruijff B. (1992) Biochemistry 31: 1636–1643. 50. Pfeil W, Bychkova VE and Ptitsyn OB. (1988) FEBS Lett. 224: 287–291. 51. Xie D, Bhakuni V and Freire E. (1991) Biochemistry 30: 10 673–10 678.

INDICATION OF MOLTEN GLOBULE STATE

52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85.

229

Yutani K, Ogasahara K and Kuwajima K. (1992) J. Mol. Biol. 228: 347–350. Griko YV, Freire E and Privalov PL. (1994) Biochemistry 33: 1889–1899. Anfinsen CB. (1973) Science 181: 223–230. Pace CN. (1986) Methods Enzymol. 131: 266–280. Murphy KP. (2001) Protein Structure: Stability and Folding. Humana, Totowa, NJ, pp. 1–16. Creighton TE. (1990) Biochem. J. 270: 1–16. Freire E. (1995) Methods Enzymol. 259: 144–168. Pace CN and Scholtz YM. (1997) In Protein Structure. A Practical Approach, Creighton TE (ed.), Oxford University Press, New York, pp. 299–321. Gekko K and Timashef SN. (1981) Biochemistry 20: 4667–4670. Gekko K and Morikawa T. (1981) J. Biochem. 90: 39–50. Connely PR, Varadarajan R, Sturtevan’t JM and Richard FM. (1990) Biochemistry 29: 6108–6114. Griko YV and Remeta DP. (1999) Protein Sci. 8: 554–561. Matulis D, Rouzina I and Bloomfield VA. (2000) J. Mol. Biol. 296: 1053–1063. Vippagunta SR, Dorn A, Ridley RG and Vennerstrom JL. (2000) Biochim. Biophys. Acta 1475: 133–140. Terada TP and Kuwajima K. (1999) Biochim. Biophys. Acta 1431: 169–281. Lopez MM, Chin DL, Baldwin RL and Makhatadze GI. (2002) Proc. Natl Acad. Sci. USA 99: 1298–1302. Leavitt S and Freire E. (2001) Curr. Opin. Struct. Biol. 11: 560–566. Kornblatt JA, Rajotte I and Heitz F. (2001) Biochemistry 40: 3639–3647. Menze M, Hellman N, Deker H and Grieshber M. (2001) J. Exp. Biol. 204: 1033–1038. Saboury AA, Moosavi-Movahedi AA and Bordbar AK. (1996) J. Chem. Thermodynamics 28: 1077–1082. Rowshan H, Bordbar AK and Moosavi-Movahedi AA. (1996) Thermochim. Acta 285: 221–229. Bathaie SZ, Moosavi-Movahedi AA and Saboury AA. (1999) Nucleic Acid Res. 27: 1001–1005. Ataie JG, Moosavi-Movahedi AA, Saboury AA, Hakimelahi GH, Hwu Jru and Tsay SC. (2000) Int. J. Biol. Macromol. 27: 29–34. Ajloo D, Moosavi-Movahedi, Hakimelahi GH, Saboury AA and Gharibi H. (2002) Colloid Surf. B: Biointerfaces 26: 185–196. Racker E. (1983) Fed. Proc. 42: 2899–2909. Hagihara Y, Aimoto S, Fink AL and Goto Y. (1993) J. Mol. Biol. 231: 180–184. Hamada D, Kidokoro S. Fukada H. Takahashi K and Goto Y. (1994) Proc. Natl Acad. Sci. USA 91: 10 325–10 329. Chamani J, Moosavi-Movahedi AA, Saboury AA, Gharanfoli M and Hakimelahi GH. (2003) J. Chem. Thermodynamics 35: 199–207. Moosavi-Movahedi AA. (2002) Encyclopedia of Surface and Colloid Science. Dekker, New York, pp. 5344–5354. Albery WJ, Edowes MJ, Hill HAO and Hillman AR. (1981) J. Am. Chem. Soc. 103: 3904–3912. Moosavi-Movahedi AA, Chamani J, Ghourchian H, Shafiey H, Sorenson CM and Sheibani N. (2003) J. Protein Chem. 22(1): 23–30. Huange ZX, Feng M, Wang YH, Cui J and Zho DS. (1996) J. Electroanal. Chem. 416: 31–40. Lojou E and Bianko P. (2000) J. Electroanal. Chem. 485: 71–80. Brandts JF. (1964) J. Am. Chem. Soc. 86: 4291–4297.

230

THERMODYNAMICS OF MG STATE OF CYTOCHROME

86. Privalov PL and Khechinashvili NN. (1974) J. Mol. Biol. 86: 665–685. 87. Freire E and Bilton RL. (1978) Biopolymers 17: 463–479. 88. Privalov PL, Mateo PL, Khechinashvili NN, Stepanov V and Revina L. (1981) J. Mol. Biol. 152: 445–449. 89. Privalov PL and Potekhin SA. (1986) Methods Enzymol. 131: 4–51. 90. Hagihara Y, Tan Y and Goto Y. (1994) J. Mol. Biol. 237: 336–348. 91. Pina DG, Shnyrova AV, Gavilanes F, Rodrigez A, Leal F, Roig MG, Sokhorov IY, Zhadan GG, Villar E and Shnyrov VL. (2001) Eur. J. Biochem. 267: 1–8. 92. Zhange F and Rowe ES. (1994) Biochim. Biophys. Acta 1193: 219–225. 93. Liggins JR, Lo TP, Brayer GD and Nall BT. (1999) Protein Sci. 8: 2645–2654. 94. Freire E, Vanosdol WW, Mayorga OL and Sanches Ruiz JM. (1990) Annu. Rev. Biophys. Chem. 19: 159–172. 95. Moosavi-Movahedi AA, Chamani J, Gharanfoli M and Hakimelahi GH. (2004) Thermochim. Acta 409: 137–144.

13 Microcalorimetry as Applied to Psychrophilic Enzymes Salvino D’Amico, Daphne´ Georlette, Tony Collins, Georges Feller and Charles Gerday

13.1 Introduction Cold-adapted, or psychrophilic, organisms have successfully colonized permanently cold environments and are able to grow efficiently at sub-zero temperatures. At the enzymatic level, such organisms have to cope with the reduction in chemical reaction rates induced by low temperatures in order to maintain adequate metabolic fluxes. In most cases, the adaptation to cold is achieved through a reduction in the activation energy, which possibly originates from an increased flexibility of either a selected area or of the overall protein structure. This evolution has two important consequences: (a) an increased catalytic efficiency at low temperatures and (b) a low thermal stability of the enzymes. In this context, we used microcalorimetry (DSC and ITC) in order to investigate the possible relationship existing between stability, flexibility and specific activity and to understand the possible adaptive strategies used by several psychrophilic enzymes from Antarctic bacteria.

13.2 Cold adaptation Psychrophiles are able to thrive at low temperatures in permanently cold environments such as polar regions or deep-sea waters.1 Recently, the unique Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

232

MICROCALORIMETRY AS APPLIED TO PSYCHROPHILIC ENZYMES

properties of psychrophilic organisms and of their enzymes have attracted attention due to their potential use in biotechnology.2,3 Temperature is one of the most important environmental factors for life as it influences most biochemical reactions. Therefore, a vast array of structural and physiological adjustments are required by psychrophiles in order to grow efficiently at nearzero temperatures. These organisms produce cold-evolved enzymes that are partially able to cope with the reduction in chemical reaction rates induced by low temperatures.4 The relative effect of temperature on the activity of psychrophilic and mesophilic enzymes is illustrated in Figure 13.1, which reveals three basic features: (a) psychrophiles synthesize enzymes with higher specific activity at low temperatures; (b) the apparent maximal activity is shifted toward low temperatures, reflecting their weak thermal stability, and (c) the adaptation does not seem complete, as the specific activity displayed by the psychrophilic enzyme around 08C remains lower than that of the mesophilic homologue at 378C. As a rule, cold-active enzymes display a high catalytic efficiency, associated, however, with a low thermal stability. In most cases, the adaptation to cold is achieved through a reduction in the activation energy, which possibly originates from an increased flexibility of either a selected area or of the overall protein structure. This enhanced plasticity seems in turn to be induced by the weak thermal stability of psychrophilic enzymes. Recently, eight crystallographic structures of cold-active enzymes have been solved. From the comparison with their mesophilic homologues, the main adaptations displayed by psychrophilic enzymes are (a) a decrease in the number or the strength of weak interactions stabilizing their folded state and (b) an optimization of their surface electrostatic potentials in order to direct the substrate towards the active site. The decrease in activation energy is

Figure 13.1 Thermal dependence of the specific activity of the psychrophilic a-amylase from Pseudoalteromonas haloplanktis (closed circles) and of its thermostable homologue from Bacillus amyloliquefaciens (open circles)

UNIFORMLY UNSTABLE ENZYMES

233

structurally achieved by a decrease of enthalpy-driven interactions that have to be broken during the activation steps. These interactions initially contribute to the stability of the active site conformation and, as a corollary, their alteration presumably gives rise to an increase in the flexibility of the structural domain of the enzyme involved in catalysis.5

13.3 Uniformly unstable enzymes Microcalorimetry has been very useful in characterizing different homologous enzymes adapted to various environmental temperatures. Thermograms of heat-induced unfolding of different enzymes recorded by differential scanning calorimetry are shown in Figure 13.2, from which several observations can be drawn:6,7 (a) the low thermal stability of the cold-adapted proteins is clearly illustrated by their low melting temperature (Tm ), (b) the unfolding enthalpy (DHcal , calculated from the area below the curve) reflects the enthalpy of disruption of bonds involved in maintaining the protein structure and there is a clear increase from cold-adapted to thermostable enzymes and (c) the psychrophilic a-amylase unfolds reversibly and without any stable intermediate. It is actually the largest known protein undergoing a totally reversible thermal unfolding.8 By contrast, more stable a-amylases unfold irreversibly and the denaturation peaks exhibit distinct thermodynamic units or domains with different stabilities. The same behaviour is observed for DNA ligases except that denaturation of the cold-adapted enzyme is not reversible due to protein aggregation (Georlette D., unpublished data). The contribution of individual weak interactions to the behaviour of the psychrophilic a-amylase was analysed by site-directed mutagenesis.7 Fourteen mutants of this enzyme were constructed, each bearing an engineered residue forming a weak interaction in mesophilic a-amylases but absent from the coldactive a-amylase. It was found that single amino acid side-chain substitutions can significantly modify the melting point Tm , the calorimetric enthalpy DHcal , the cooperativity and reversibility of unfolding and also the kinetic parameters kcat and Km (Figure 13.3) towards the behaviour and values found in heatstable enzymes. From the analysis of our available data, it is clear that the adaptation to cold of the psychrophilic a-amylase, DNA ligase9 and xylanase10 consists of a weakening of intramolecular interactions, thereby leading to an overall decrease in the thermostability of the protein. This in turn provides the appropriate plasticity around the catalytic residues necessary to adapt the catalytic efficiency of the enzyme to the low temperature of the environment. It is also clear that in the case of the a-amylase6 the limit of stability of the protein has been reached, precluding any further decrease of this stability, and precluding therefore any further improvement of the flexibility of the protein

234

MICROCALORIMETRY AS APPLIED TO PSYCHROPHILIC ENZYMES

Figure 13.2 Thermal unfolding of psychrophilic (P), mesophilic (M) and thermophilic (T) enzymes recorded by differential scanning calorimetry. (a) From left to right: a-amylases from Pseudoalteromonas haloplanktis, pig pancreas and Bacillus amyloliquefaciens. (b) From left to right: DNA ligases from Pseudoalteromonas haloplanktis, Escherichia coli and Thermus scotoductus. Deconvolution of the peaks into two or three domains for the mesophilic and thermophilic enzymes respectively is shown as thin curves

by this strategy and thus also any further adjustment of the catalytic efficiency. This latter remains, therefore, at the usual environmental temperature of the Antarctic microorganism, well below that displayed by the mesophilic counterpart at its own environmental temperature, as mentioned for Figure 13.1.

13.4 Enzymes with local flexibility Of course, not all cold-adapted enzymes exhibit a similar strategy. The analysis of the thermostability of psychrophilic enzymes shows that the adaptation strategy can be different from that leading to an uniformly

ENZYMES WITH LOCAL FLEXIBILITY

235

Figure 13.3 Thermal unfolding of mutants introducing new weak interactions in the coldadapted a-amylase and recorded by differential scanning calorimetry. (a) Mutants increasing the melting point and the calorimetric enthalpy. The wild-type enzyme is represented by the left-hand peak. (b) Mutants inducing the appearance of two calorimetric domains (thin curves). (c) Mutants inducing an alteration of the unfolding cooperativity

unstable protein (Figure 13.4). The calorimetric curve for a psychrophilic phosphoglycerate kinase appears to be composed of a heat-labile and a heatstable domain (two denaturation peaks)11 when compared with the fused transition observed in the case of the mesophilic counterpart from yeast. It has been hypothesized that the heat-labile domain provides the required flexibility around the active site and favours the reaction rate by reducing the energy cost of induced-fit mechanisms, whereas the heat-stable domain could improve substrate binding (as indicated by low Km values) as a result of its rigidity. This is a strategy of compromise between the necessity to increase the specific activity at low temperatures (engineering the flexibility is apparently the simplest way to achieve this goal) and the necessity to retain a high affinity for the substrate, a mandatory requirement in the case of low substrate concentrations as is typical of the intracellular compartment. This fact is further emphasized by calorimetric experiments carried out on a psychrophilic chitobiase.12 The calorimetric curve obtained from differential scanning calorimetry also displays a heat-labile and a heat-stable domain (Figure 13.4). Different ligands were used to evaluate the binding capacities of the enzyme and their effects on the Tm of either the first or the second denaturation peak. These experiments allowed us to unequivocally assign transition 1 (low Tm ) to the catalytic domain and transition 2 (high Tm ) to the

236

MICROCALORIMETRY AS APPLIED TO PSYCHROPHILIC ENZYMES

Figure 13.4 Thermal unfolding of psychrophilic enzymes demonstrating local flexibility. (a) Phosphoglycerate kinase. (b) Chitobiase

galactose binding domain. These results demonstrate that heat lability is not simply the result of a lack of selective pressure in cold environments, but seems to be the consequence of the improved plasticity required around the active site.

13.5 Thermal inactivation recorded by isothermal titration calorimetry Determination of inactivation rate constants usually requires the recording of the residual enzyme activity after incubation at high temperatures. In the case of the psychrophilic a-amylase and its mutants7, the model of Lumry and Eyring13 is valid: k1

k2

NÐU!I k
1

ð13:1Þ

where k2 is the first-order rate constant of thermal inactivation. However, the various degrees of reversibility noted for the mutants (kinetically characterized by k
1 ) strongly impaired the validity of the results obtained by conventional methods. We have therefore devised a new method involving isothermal titration calorimetry and recording activity at 458C as the heat released by the hydrolysis of starch glycosidic bonds. This method provides a direct and continuous monitoring of activity at the denaturing temperature (Figure 13.5). As a prerequisite, the same experiment was followed by fluorescence (without substrate), showing that the structural modifications (corresponding to

THERMAL INACTIVATION

237

Figure 13.5 Thermal inactivation recorded by isothermal titration calorimetry. Upper panel: following enzyme injection (arrow) of either P. haloplanktis a-amylase (AHA) or pig pancreas a-amylase (PPA), activity at 458C is recorded as the heat flow of the reaction. Lower panel: semilogarithmic plot of residual activity versus time, illustrating the lower inactivation rate of two mutants and the corresponding double mutant (NV ¼ V196F þ N150D) compared with the wild-type cold-adapted a-amylase (AHA)

transition N!U) accompanying a temperature shift from 15 to 458C are one order of magnitude faster than the transition U!I, and the disappearance of the active state N can be taken as a measure of k2 . The heat released by the enzyme activity induces a decrease of the microcalorimeter signal. It is clear that at 458C the activity of the mesophilic a-amylase from the pig pancreas is constant for a long period whereas the activity of the psychrophilic enzyme progressively returns to zero. The hyperbolic curves observed in the case of thermal inactivation can be represented on semilogarithmic plots (Figure 13.5), the slopes of which correspond to k2 , the inactivation rate constant. It was significant that 10 out of 14 a-amylase mutants were protected against thermal inactivation and that mutants carrying newly introduced electrostatic interactions displayed the lowest inactivation rate constants.7

238

MICROCALORIMETRY AS APPLIED TO PSYCHROPHILIC ENZYMES

Figure 13.6 Calorimetric trace of the cold-adapted chitinase assay on soluble chitin recorded by isothermal titration calorimetry. Arrows correspond to the injection of equal amounts of enzyme into the reaction cell. Note that each injection induce the same heat release

13.6 Microcalorimetric determination of enzyme kinetic parameters Isothermal titration calorimetry was initially designed to record the enthalpy change associated with the binding of a ligand to a specific molecule. Recently, however, this method has been used to record enzyme activity and determine its kinetic parameters.14,15 A steady state can be reached by using a low concentration of enzyme and saturating substrate concentration. In these conditions, the enzyme reaction rate is directly proportional to the endothermic or exothermic heat exchange of the reaction per unit of time. Practically, the reaction rate is calculated by subtracting the power generated by the calorimeter to maintain a constant temperature (Pi ) from the power generated in the reaction cell (Pr ) (Figure 13.6). Several enzyme injections made under saturating substrate concentration demonstrate the linear relationship between the reaction rate and the enzyme concentration. In order to transform the reaction power into the number of bonds cleaved per unit time, the reaction enthalpy has to be known. A good estimation can be obtained with isothermal titration calorimeters by recording the complete hydrolysis of a known amount of substrate and calculating the reaction enthalpy from the total heat evolved. The number of bonds cleaved per unit of time can then be normalized for the enzyme concentration in order to determine the reaction rate constant kcat ðs
1 ). In microcalorimetry, if the enzyme concentration is sufficiently low, it is possible to record a stable activity at various decreasing substrate

CONCLUSION

239

Figure 13.7 Calorimetric trace of the cold-adapted chitinase assay on insoluble chitin recorded by isothermal titration calorimetry. Arrows correspond to the injection of equal amounts of enzyme into the reaction cell. Note that the heat released by a same enzyme injection decreases progressively due to the saturation of the insoluble substrate

concentrations. Therefore, it is possible to construct the rectangular hyperbolic saturation curve that obeys Michaelis–Menten kinetics. The main advantage of this method is its precision as a whole range of experimental values covers the saturation curve and no a priori knowledge of the Km value is required. However, this is a time-consuming method because each record lasts 30–40 min, as a result of the long thermal equilibrium delay of the calorimeter. In the case of insoluble substrates such as chitin, the successive injections of chitinase aliquots into the substrate suspension showed a substrate saturation phenomenon14 (Figure 13.7). As the number of identical enzyme injections increased, the corresponding heat release detected by the microcalorimeter decreased progressively to reach zero when chitin was totally saturated by the enzyme. Any further injection of enzyme did not increase the heat release. The activity had attained a maximum level and it was demonstrated that this was not limited by substrate depletion. In such cases, working with lower enzyme concentrations can restore a linear relationship between the heat flow released and the enzyme concentration, therefore allowing the determination of kcat values.

13.7 Conclusion Even though important advances have been made in the field of enzyme cold adaptation, the debate as to why psychrophilic enzymes systematically associate high activity with low thermal stability remains a hot topic. However, it seems clear at this stage that cold-adapted enzymes require an

240

MICROCALORIMETRY AS APPLIED TO PSYCHROPHILIC ENZYMES

improved flexibility of the structural domain involved in the catalytic activity, whereas other regions not implicated in catalysis can retain a certain rigidity. Microcalorimetry, using both DSC and ITC, allows a thermodynamic comparison of homologous enzymes adapted to different environments with a precision never reached before. This is due to the microcalorimeter’s ability to directly record several thermodynamic parameters related, for example, to protein stability. Extrapolations from indirect data are no longer required.

References 1. Deming JW. (2002) Curr. Opin. Microbiol. 5: 301–309. 2. Gerday C, Aittaleb M, Bentahier M, Chessa JP, Claverie P, Collins T, D’Amico S, Dumont J, Garsoux G, Georlette D, Hoyoux A, Lonhienne T, Meuwis M-A and Feller G. (2000) Trends Biotechnol. 18: 103–107. 3. Cavicchioli R, Siddiqui KS, Andrews D and Sowers KR. (2002) Curr. Opin. Biotechnol. 13: 253–261. 4. D’Amico S, Claverie P, Collins T, Georlette D, Gratia E, Hoyoux A, Meuwis MA, Feller G and Gerday C. (2002) Phil. Trans. R. Soc. Lond. B 357: 917–925. 5. Lonhienne T, Gerday C and Feller G. (2000) Biochim. Biophys. Acta 1543: 1–10. 6. Feller G, d’Amico D and Gerday C. (1999) Biochemistry 38: 4613–4619. 7. D’Amico S, Gerday C and Feller G. (2001) J. Biol. Chem. 276: 25 791–25 796. 8. Kumar S, Tsai CJ and Nussinov R. (2002) Biochemistry 41: 5359–5374. 9. Georlette D, Jonsson ZO, Van Petegem F, Chessa J, Van Beeumen J, Hubscher U and Gerday C. (2000) Eur. J. Biochem. 267: 3502–3512. 10. Collins T, Meuwis MA, Stals I, Claeyssens M, Feller G and Gerday C. (2002) J. Biol. Chem. 27: 27. 11. Bentahir M, Feller G, Aittaleb M, Lamotte-Brasseur J, Himri T, Chessa JP and Gerday C. (2000) J. Biol. Chem. 275: 11 147–11 153. 12. Lonhienne T, Zoidakis J, Vorgias CE, Feller G, Gerday C and Bouriotis V. (2001) J. Mol. Biol. 310: 291–297. 13. Lumry R and Eyring H. (1954) J. Phys. Chem. 58: 110–120. 14. Lonhienne T, Baise E, Feller G, Bouriotis V and Gerday C. (2001) Biochim. Biophys. Acta 1545: 349–356. 15. Todd MJ and Gomez J. (2001) Anal. Biochem. 296: 179–187.

14 An Autosampling Differential Scanning Calorimeter for Study of Biomolecular Interactions Valerian Plotnikov, Andrew Rochalski, Michael Brandts, John F. Brandts, Samuel Williston, Verna Frasca and Lung-Nan Lin

14.1 Introduction to DSC Microcalorimetry is a widely used technique in the study and characterization of biomolecular interactions, and is important in drug discovery research, as demonstrated by other articles in this book and other recent publications.1–3 DSC measures the heat changes which occur in a macromolecule when there is a controlled increase in temperature.3,4 A biopolymer solution (e.g. protein, DNA) is placed in a sample cell, matched buffer is placed in a reference cell, and both cells are heated at a constant rate. The instrument has a cell feedback network, which differentially measures the heat produced or absorbed between the sample and reference cell during the heating.4 If the biopolymer thermally denatures from the low-temperature native form to the hightemperature denatured form, this unfolding is seen in the differential power at each temperature. Thermograms of proteins and DNA typically demonstrate an endothermic peak. The centre of the peak is the transition midpoint (TM ). The enthalpy (DH) and heat capacity of unfolding (DCp ) can be determined by data analysis. Stabilizing additives and buffers increase the TM of the

Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

242

AUTOSAMPLING DIFFERENTIAL SCANNING CALORIMETER

biopolymer, so comparison of TM in different solutions can provide valuable information in formulation development.5,6 DSC can also be used to study protein–small-molecule interactions.3,7,8 If a ligand binds more strongly to the native form of the biopolymer, the ligand has stabilized the native state of the biopolymer and the TM of the biopolymer–ligand complex is greater than that of the biopolymer alone. The binding constant of the ligand to protein can then be determined. DSC can be used to screen drug binding to target protein. This method can also be used to study ultratight binding interactions, where binding constants cannot be determined by isothermal titration calorimetry or other methods. DSC can also be used for protein–protein interactions7 and DNA–ligand interactions.9 Available DSCs have relatively low sample throughput, due to temperature equilibration and cell cleaning between experiments. For studies requiring multiple DSC scans, higher throughput is important. The VP-Capillary DSC Platform was designed with shorter equilibration times, faster scan rates, and a fully-integrated autosampler to provide continuous, hands-off operation.10

14.2 Description of instrument Figure 14.1 is a schematic diagram of the thermal core of the VP-Capillary DSC Platform. Sample and reference cells are fabricated from 1.5 mm i.d.

Figure 14.1 Schematic diagram of thermal core of DSC instrument (reprinted from reference 10)

DESCRIPTION OF INSTRUMENT

243

tantalum tubing. Each cell (working volume 125 ml) is formed by two helical turns of the tubing around a silver cylinder, and fixed in place inside a silver adiabatic jacket. The inlet and outlet tubes are formed by constricting the tantalum tubing to 0.8 mm i.d. The tubes pass through the adiabatic jacket and communicate with the distributor valve. The two cells are in thermal contact with a differential temperature detector, which provides a DT
1 voltage signal between sample and reference cell. The DT
1 voltage signal is amplified and fed back to the heater on either the sample or reference cell (whichever is cooler) to nullify the signal. A thermopile is between the cells and adiabatic jacket, providing a DT
2 voltage signal, which is amplified and fed back to the Peltier device on the jacket, which then cools or heats the jacket to nullify the signal. These two feedback mechanisms work during DSC operation to maintain the two cells at the same temperature and the jacket at the same temperature as the cells. As the temperature in the cells increases, the jacket is also heated up at the same rate controlled by DT
2 feedback. The differential power of the DT
1 feedback circuit is stored with the temperature, and these are the experimental data output. The DSC thermal core is fully integrated to an autosampling system. A 10port distributor valve is connected to the inlet and outlet ports of the sample and reference cells.10 There are four configurations for operation (Figure 14.2): I (1–3, 6–8). The inlet tube of the sample cell is open to the injection port and the outlet tube of the sample cell is open to waste. This is for cell draining, rinsing, or filling the sample cell.

Figure 14.2 Schematic diagram of distributor valve, syringe injection port, and access ports to sample and reference cells, waste, buffer, and external pressure source (reprinted from reference 10)

244

AUTOSAMPLING DIFFERENTIAL SCANNING CALORIMETER

II (1–4, 6–9). The same operations as configuration I for the reference cell. III (1–2, 6–7). Pressurizing the sample and reference cells with nitrogen gas. Both outlet tubes are open to the 30 psi pressure source, and both inlet tubes are closed. IV (1–5, 6–10). The injection syringe withdraws buffer from the reservoir. The autosampling assembly also has an injection syringe mounted on a slide bar above the distributor valve. The XYZ position of the injection needle and the syringe plunger operation are controlled by a series of software-driven stepping motors. The samples are placed in 96-well plates, and stored in three thermostatted drawers. The drawers are automatically opened and closed as required. The autosampling assembly has reservoirs (non-thermostatted) for rinse buffer/reference buffer/water storage, which can be accessed by the injection syringe. The loaded syringe (air, sample, reference buffer, or wash buffer) hermetically mates to the injection port, the distributor valve moves into the proper configuration, and the syringe injects the appropriate volume. VPViewerTM software (MicroCal, LLC) is used for experimental set-up and instrument operation of multiple samples. Samples can be identified and experimental conditions can be described (concentration, buffer, scan rate, scan temperature range, comments) for each sample. The following sequence of steps is automatically performed by the VPCapillary DSC Platform prior to each new sample.10 1. The DSC instrument begins the cooling process, using power to the Peltier device on the adiabatic jacket. 2. The injection syringe is filled with air; the air is pushed through the sample and reference cells to drain contents to waste. 3. Both cells are rinsed with water or buffer, then drained to waste. 4. The sample cell is filled with sample from the 96-well plate, and the reference cell is loaded with buffer from the 96-well plate. 5. Pressure from the nitrogen tank is applied to the cells, to increase the boiling point. 6. Cooling continues until the cells are equilibrated to the start temperature of the next scan. 7. The scan begins. Many design features of the instrument act to improve overall performance and throughput.10 The autosampler injects sample solution directly into the cells via the distributor valve, allowing for different options such as cleaning and filling. This also eliminates the need for long feed lines from sample

MATERIALS AND METHODS

245

injection to cell, allowing for low dead volumes. The use of inlet and outlet tubes allows for efficient cell cleaning and filling using the autosampler. The high (metal surface)/(cell volume) ratio in the capillary cell and the small cell volume of 125 ml permits thermal equilibration of sample solution primarily by conduction of heat from metal to solution, rather than by convection through the solution, thereby shortening thermal equilibrium. The instrument is capable of upscan rates of up to 2508C h
1 , about three times faster than other conventional DSC instruments, which shortens run times for each sample analysis. Down time between scans is also reduced, due to cooling rates greater than 3008C h
1 .10 Using this capillary cell design, it has been found that a buffer rinse between samples is sufficient for cell cleaning, and detergent washes are less frequently needed. This applies even when proteins precipitate due to thermal denaturation. With conventional DSC instruments, precipitated proteins need to be removed with detergent or acid washes. Post-run data analysis is performed with OriginTM software (OriginLab, Inc.), with modifications by MicroCal, LLC. Routines are available for multiple sample data handling, and multiple data files are opened and plotted in a single Origin graph. A designated buffer–buffer instrument baseline is subtracted from each sample scan, each scan normalized for concentration, and TM values are determined. Binding constants (KB ) can also be determined as described below. All information is contained in a worksheet.

14.3 Materials and methods Materials Bovine ribonuclease A (catalogue number R5500), cytidine 2’-monophosphate (2’CMP) (catalogue number C7137), cytidine 3’-monophosphate (3’CMP) (catalogue number C1133), uridine 3’-monophosphate (3’UMP) (catalogue number U1126), sodium phosphate monobasic (catalogue number S9638), and sodium pyrophosphate (catalogue number S6422) were purchased from Sigma Chemical Company (St Louis, MO). All other chemicals were of reagent grade.

Preparation of ribonuclease A and inhibitors Ribonuclease A solution (in 50 mM potassium acetate, pH 5.5) was dialysed against two changes of 500 ml of buffer solution at 48C for 3 hours each (using MWCO 6000–8000 tubing) prior to DSC. Ribonuclease concentration was determined spectrophotometrically at 278 nm, using an extinction coefficient

246

AUTOSAMPLING DIFFERENTIAL SCANNING CALORIMETER

of 9800 cm
1 M
1 . Nucleotides were prepared and absorbances were measured at 260 nm. Concentrations were determined using coefficients of 7400 cm
1 M
1 for 2’CMP and 3’CMP, and 10 000 cm
1 M
1 for 3’UMP. Concentrations for phosphate and pyrophosphate were estimated by weighing.

DSC analysis of inhibitor binding to ribonuclease A 400 ml aliquots of 0.090 mM ribonuclease A were placed in wells of the 96-well plate, with a final concentration of 10 mM inhibitor. Four aliquots were scanned for each experimental condition. Reference buffer was also placed in the 96-well plate. Samples were scanned in the VP-Capillary DSC Platform at 2008C h
1 with a filter time of 2 s. All samples were analysed for TM , DH, and DCp using standard methods.

14.4 Results for ribonuclease binding to anionic inhibitors Experimental results are shown in Figure 14.3. These inhibitors are known to bind to the active site of ribonuclease A. Experiments were performed as described in Section 14.3 and the figure legend. The numbered arrows indicate

Figure 14.3 Raw data of ribonuclease A in the absence and presence of inhibitor. Six different solutions of ribonuclease A (0.09 mM protein, in 50 mM acetate buffer, pH 5.5) with four aliquots of each solution were analysed to produce overlapping DSC scans. The transition midpoints are indicated by arrows for (1) no added ligand, (2) 10 mM phosphate, (3) 10 mM 3’CMP, (4) 10 mM 3’UMP, (5) 10 mM pyrophosphate, and (6) 10 mM 2’CMP (reprinted from Reference 10)

RESULTS FOR RIBONUCLEASE BINDING TO ANIONIC INHIBITORS

247

Table 14.1 Experimental estimates for the transition midpoint TM , heat of unfolding DH, and heat capacity of unfolding DCp for ribonuclease A in the presence of five different ligands at a concentration of 10 mM (reprinted from Reference 10)

Ligands at 10 mM Control – no ligand Phosphate 3’CMP 3’UMP Pyrophosphate 2’CMP

TM (8C) 61.08 62.95 66.68 67.47 68.24 70.71

DH DCp (kcal mol1 ) (kcal mol1 deg1 ) 109 127 145 137 140 154

1.3 1.4 1.4 1.5 1.4 1.4

KB a (M1 )

DHB (kcal mol1 )

– 155 1 560 2 380 3 580 13 100

– 226 232 221 222 232

For each parameter, the recorded values are the average for four separate determinations. The binding constant KB and the heat of binding DHB for each ligand were calculated as explained in the text. a The corresponding dissociation constants Kd may be obtained as the reciprocal of the binding constants, i.e. Kd ¼ 1=KB .

average TM for each solution, and a summary of the results is in Table 14.1. The average deviation of the peak position from the mean was 0:028C. All 24 scans in Figure 14.3 were analysed using standard methods3,7,10 to obtain heats of unfolding, and heat capacities of unfolding at TM , and mean values are shown in Table 14.1. Average deviation from the mean was 1.5 kcal mol
1 deg
1 for DH values, and 0.2 kcal mol
1 deg
1 for DCp values. DH became more positive as TM increased. This is normally seen for protein unfolding reactions when TM is increased by changing the pH of the solution.3,10 This temperature dependence of DH is from the following thermodynamic relationship: 
@DH ð14:1Þ DCp ¼ @T p DCp is positive for all samples in Figure 14.3 ( 1:4 kcal mol
1 deg
1 ), thereby making DH larger as TM becomes higher. However, the results are complicated since ligand release occurs during protein unfolding and this will also contribute to DH differences in the presence and absence of ligand, over and above the usual contributions from DCp . Ligand release makes a positive contribution to experimental DH values.10 The binding constant KB for each ligand can be estimated from DH, DCp and TM in the absence of ligand, and TM in the presence of ligand, using Equations (14.2) and (14.3) in the appendix. The KB values are shown in Table 14.1. These values are for KB at the experimental TM , which is different for each ligand. The heat of binding, DHB , can also be estimated from the data in Table 14.1, using Equations (14.4) and (14.5) in the appendix. These estimates have a high degree of uncertainty, since they are the small difference after

248

AUTOSAMPLING DIFFERENTIAL SCANNING CALORIMETER

subtracting one large number from another large number. The change in the entropy of binding, DSB , can also be estimated, and values for the ligands in Table 14.1 are all large and negative.10 The binding in the cases shown here is heat driven, with entropy acting to favour the unbound state.

14.5 Discussion The VP-Capillary DSC Platform integrates the convenience of autosampling with high sensitivity, and produces data with high reproducibility. This instrument is designed for applications requiring higher sample throughput than is obtained with other commercially available ultrasensitive microcalorimeters. The application shown here with ribonuclease A inhibitor binding shows that this instrument can be used to estimate binding constants for protein–ligand interactions, making the instrument very suited for screening drug binding to target protein. For protein–small-molecule interactions, ligands with small binding constants must be studied at high concentration to produce a measurable shift in TM , while ligands with tighter binding can be studied at lower concentrations. The range of KB values that can be determined from DSC results is very broad, as shown by these results and other data with DSC.7,10 Results with ribonuclease A binding to inhibitors show very high repeatability and precision of TM and DH values as seen in Table 14.1. However, the absolute accuracy of KB values cannot be determined since there are no independent results using other techniques under identical conditions available for comparison. Anderson et al.11 performed binding studies of inhibitors to ribonuclease A using spectrophotometric methods. They also observed that 2’CMP had a larger binding constant than 3’CMP and 3’UMP. Also, use of enzymatic assays showed pyrophosphate had a higher inhibitor constant than phosphate. The temperature and buffers were different, however, so quantitative comparison is not possible. Wiseman et al.12 used ITC to study binding of 2’CMP to ribonuuclease A in the same buffer as used here, but at 288C. A direct comparison of KB from ITC and DSC is not possible, due to different temperatures for experiments. When the KB value at 288C (7  106 M
1 ), determined from ITC, was extrapolated to 70.78C using DH of binding of
17 kcal mol
1 and DCp of
140 cal deg
1 mol
1 (both measured by ITC) the resulting KB of 12  103 M
1 compared well to the value of 13  103 M
1 as determined here by DSC. The determination of KB by DSC is an indirect method, since the estimation is made from measurements made on the equilibration between folded and unfolded biopolymer, rather than equilibrium between unbound and bound forms.7,8 The model of Brandts and Lin7 which describes protein–ligand binding studies with DSC assumes that protein unfolding is two state and

REFERENCES

249

reversible, and that ligand binds exclusively to either the native or denatured form of protein but not to both. These assumptions are not always correct. Though DSC is useful for comparing relative affinities of ligand binding to protein, the KB values are estimates only.7,10 It is important to note that DSC is very useful in determination of KB of ultratight interactions (up to 1020 M
1 ), which cannot be measured by ITC or other binding studies.7,10 This is because the TM shift with increasing ligand concentration continues to occur even after 100 per cent binding saturation has taken place.3,7,10 The transition midpoints for protein–ligand DSC scans are also typically at a higher temperature than used for ITC and other binding studies. It is estimated that if DH is
10 kcal mol
1 , KB will increase by a factor of two with a reduction of 158C in temperature.10 It is possible to obtain the rank ordering of relative binding strength if DH values are similar for different ligands. However, if ligands have very different DH values, this relationship will not necessarily hold.8,10 If DH and DCp of binding are known from ITC, KB at temperatures lower than TM can then be determined (see Equation (14.7) in the appendix).

Summary This paper describes use of the VP-Capillary DSC Platform in the study of protein–ligand interactions, and protein–protein7 and DNA–ligand binding studies9 can also be performed with DSC. The VP-Capillary DSC Platform is useful in screening binding of ligands to potential target.8,10 This instrument is also useful for biopharmaceutical formulation screening, stability studies of mutant proteins, analysis before protein crystallization, and other applications requiring TM analysis of multiple samples.

References 1. Jelesarov I and Bosshard HR. (1999) J. Mol. Recognit. 12: 3–18. 2. O’Brien R, Chowdhry BZ and Ladbury JE. (2001) In Protein–Ligand Interactions: Hydrodynamics and Calorimetry, Harding SE and Chowdhry BZ (eds), Oxford University Press, Oxford, pp. 263–285. 3. Cooper A, Nutley M and Wadood A. (2001) In Protein–Ligand Interactions: Hydrodynamics and Calorimetry, Harding SE and Chowdhry BZ (eds), Oxford University Press, Oxford, pp. 287–318. 4. Plotnikov VV, Brandts JM, Lin LN and Brandts JF. (1997) Anal. Biochem. 250: 237– 244. 5. Remmele RL, Nightlinger NS, Srinivasan S and Gombotz WR. (1998) Pharm. Res. 15: 200–208. 6. Remmele RL and Gombotz WR. (2000) BioPharm 13: 36–46. 7. Brandts JF and Lin LN. (1990) Biochemistry 29: 6927–6940.

250

AUTOSAMPLING DIFFERENTIAL SCANNING CALORIMETER

8. Waldron TT and Murphy KP. (2003) Biochemistry 42: 5058–5064. 9. Leng F, Priebe W and Chaires JB. (2003) Biochemistry 37: 1743–1753. 10. Plotnikov V, Rochalski A, Brandts M, Brandts JF, Williston S, Frasca V and Lin LN. (2002) Assay Drug Dev. Technol. 1: 200–208. 11. Anderson DG, Hammes GG and Walz FG. (1968) Biochemistry 7: 1637–1645. 12. Wiseman T, Williston S, Brandts JF and Lin LN. (1989) Anal. Biochem. 179: 131–137.

Appendix: data analysis (reprinted from Reference 10) Often, specific ligands will bind much more strongly to the native, folded form of a protein than they will to the thermally unfolded form. The treatment below assumes only binding to the native form and no binding to the unfolded form. This being the case, the thermal midpoint of the unfolding transition will necessarily occur at a higher midpoint temperature, TM , when the ligand is present in solution than when the ligand is absent, T0 . For two-state reversible transitions, the equation relating the shift in midpoint temperature to the binding constant KB has been derived earlier7 and is

    1
DH8 1 1 DCp TM T0 ln KðTM Þ ¼
þ
1
1 exp þ nR T0 TM ½LTM nR TM T0 ð14:2Þ where KB (TM ) is the binding constant at TM , DH8 and DCp are the enthalpy and heat capacity change at T0 for the transition in the absence of ligand, ½LTM is the concentration of unbound ligand at TM , and n is the number of identical binding sites on the protein. Transition temperatures are expressed in K. The above equation is most useful when TM and T0 are well separated in temperature. A very large KB insures there will be large separation in midpoint even at low ligand concentration, whereas measuring TM in the presence of high ligand concentration will act to produce large separation even for weaker binding. In the former case involving large binding constants, addition of lessthan-saturating ligand concentration results in separate transition peaks for the liganded and unliganded proteins so that T0 and TM may be obtained in the same experiment.7 The value of free ligand concentration at the midpoint TM can be calculated from the equations ½LTM ¼ Ltot
½LTM ¼

nPtot 2

nPtot ðLtot 5nPtot Þ 2

ð14:3Þ

ðLtot 4nPtot Þ

where Ptot and Ltot are the total concentrations of protein and ligand, respectively, in the solution. These equations assume total saturation of the

APPENDIX: DATA ANALYSIS

251

binding sites on the native protein (Ltot 5nPtot ) or assume there is no unbound ligand ðLtot 4nPtot ) until the temperature where unfolding begins. Using the above equations, it is not necessary to know the heat of binding of ligand to protein even though ligand release will contribute to the experimental heat at TM . In principle, it is possible to determine the heat contribution from ligand release at TM as the difference between the experimental heat and the heat expected at TM based on the measured heat and heat capacity at T0 in the absence of ligand. The equation for calculating the heat of binding DHB (the negative of the heat of release) is heat for ligand release ¼
nf DHB ¼ DHðTM Þ
½DH8 þ DCp ðTM
T0 Þ ð14:4Þ where DH(TM ) is the experimental heat at TM in the presence of ligand, and DH and DCp are the heat and heat capacity change at T0 in the absence of ligand. The second term on the right-hand side of Equation (14.4) is the expected heat at TM with no bound ligand, whereas the first term is the actual heat at the same temperature in the presence of ligand. In cases where the binding sites are not completely saturated, the fraction of sites occupied, f, must be obtained from Equation (14.5) before DHB is determined. f ¼ KB ½L=ð1 þ KB ½LÞ

ð14:5Þ

When the binding constant and heat of binding are known, then the entropy of binding DSB may also be obtained by solving the equation DGB ¼
RT ln K ¼ DHB
T DSB

ð14:6Þ

If the heat and heat capacity of binding are known accurately (e.g. from ITC studies), it then becomes possible to extrapolate the KB value, obtained from Equations (14.2) and (14.3) at TM , to another temperature TF using the equation7

  
DHB 1 1 DCp TF TF ln KB ðTF Þ ¼ KB ðTM Þexp
þ1
þ R R TM TM TF TM ð14:7Þ where DHB and DCp are the heat and heat capacity of ligand binding at TF .

Index Page numbers in italics indicate figures and tables. DSC ¼ differential scanning calorimetry; ITC ¼ isothermal titration calorimetry. accessible surface area (ASA) 138, 141, 145–6 acetazolamide 109, 111, 112, 113, 114, 115, 116, 117, 118, 119, 122, 123, 124, 126, 127, 128 acidic fibroblast growth factor (aFGF/ FGF-1), heparin/MHS binding 133–50 acute lymphoblastic leukaemia 176 acyl carrier protein enoyl reductase (ACPER) 68 allergy 153, 155 Alzheimer’s disease 8, 26 p-aminobenzyl sulfonamide 109, 113, 114, 115, 116, 122, 124, 131 a-amylase 233, 236–7 amyloid plaques 26 ANS binding 217 antithrombin III 144 apomyoglobin, molten globule state 216, 217 ArgbB5 156, 166 Arrhenius equation 182 asparaginase 176 asthma 153 autoimmune disease 153 barnase 191, 208–9 basic fibroblast growth factor (bFGF/ FGF-2) 133, 144, 145 Bevington method 137 bile salt 8 binding, DSC 25 binding entropy (DSbind) 86–7

binding site mapping, scanning alanine mutagenesis 12 bioavailability 63 biocalorimetry pioneer studies 4 scope 7 bioisosteric replacement 63 BME 47 Bruton tyrosine kinase (Btk) 153 buffers, ITC 44–5, 46–7 C value 45 calmodulin, cation binding 8 calorimeters 4 cancer enzymes 175, 176 protein–nucleic acid interactions 15 SH2 domains 53, 153, 168 telomerase 28 carbonic anhydrase, sulfonamide inhibitor binding 107–32 ‘anion’ binding thermodynamics 123–5 protonation events 108–10, 119–23 structure/thermodynamic correlations 125–31 thermodynamics 111–19 4-carboxybenzene sulfonamide 109, 113, 122, 124, 131 cell signalling 8 chaperones 8 charge–charge interactions, ribonuclease binding 96–103 charged–charged hydrogen bonds 84–6

Biocalorimetry 2. Edited by John E. Ladbury and Michael Doyle. & 2004 John Wiley & Sons, Ltd. ISBN 0 470 84968 1

254

INDEX

chemical denaturation 203–4 protein stability 205–7 chevron plot 204 chitin 239 chitobiase 235 circular dichroism, molten globule state 217 cloning 4 conformational changes 10 cooperativity, transition 22 Crestor 75–6 CysbC3 167 cysteine residues, drug targets 167 cytochrome c, molten globule state 216, 217, 218–27 cytoskeleton 152 detergents 8, 25, 176 diabetes 53 dichlorophenamide 109, 111, 112, 113, 114, 115, 116, 117, 118–19, 121, 122, 123, 124, 125 differential scanning calorimetry (DSC) applications 22–3 autosampling 25, 241–51 binding 25 detergents 25 DNA–ligand interactions 242 DNA–lipid interactions 25 DNA–PNA interactions 25 drug–DNA interactions 25 formulations 26 gelation 26 high-sensitivity 22 intramolecular protein domain interactions 25 irreversible system 24 KB 248–9 m values 203–14 macromolecular structures 22, 25 micellar systems 25 molten globule state 225–7 nucleic acid helix!coil transitions 26–30 oligomerization 25–6 peptide–ligand binding 25 principles 21–2 protein–protein interactions 242 protein–small molecule interactions 25, 242

RNase Sa stability and salt concentration 100–2 screening 25 shelf life 24–5 Tm 24–6 thermal denaturation 203 thermal unfolding of INT-DBD 190–2 thermodynamic stability 23, 24–5 VP-Capillary DSC Platform 18, 242–9 diphenylfuran compounds 15 DMSO 47 DNA gyrase inhibitors 74 DNA–ligand interactions, DSC 242 DNA ligase, cold adaptation 233 DNA–lipid interactions, DSC 25 DNA melting 18 DNA–PNA binding, DSC 25 DNA–protein interactions see protein– DNA interactions double-jump assays 204 drug discovery 59–79 active compounds 61, 70–2 assay validation 68–9 bioavailability 63 enthalpy–entropy compensation 77 enzymes 175–6 functional efficacy 63 hit identification 61 identification of new classes 12 ITC 67–77 lead identification 61–2 lead optimization 62–3 overview 60–3 screening 61, 242 structure-based design 72–6 target characterization 67–8 target selection 60–1 target validation 60–1, 68 drug–DNA interactions DSC 25 ITC 8, 13–15 osmotic stress 18 drug–receptor binding 12 DSHP 152 duplex DNA 18, 27–8, 192–4 electrostatic interactions 12, 216 embryogenesis 151

INDEX

enthalpy change (DH) 6, 9, 38, 39, 53–4, 65, 66, 67 calorimetric transition (DHcal) 22, 23 ionization 44–5, 54, 56 temperature dependence 43–4, 54 van’t Hoff (DHvH) 22, 23, 66 enthalpy–entropy compensation 12, 77 entropy change (DS) 6, 38, 53–4, 65, 67 DSr+t 14 binding (DSbind) 86–7 configurational 39 hydrophobic interactions 10 NMR 87, 88, 89 translational/cratic 39 enzymes assays 176 catalysis, ITC 18–21, 175–85 diagnostic role 175 drug development 175–6 industrial applications 176 see also psychrophilic enzymes epilepsy 108 equilibrium binding constant (Kb) 5–6, 38, 45, 52, 65, 66, 248–9 equilibrium dialysis 6 equilibrium dissociation constant (Kd) 6 equilibrium line 205 Eyring equation 182 FceRI receptor 163 FceRIIA receptor 163, 165 fibroblast growth factors acidic see acidic fibroblast growth factor basic (bFGF/FGF-2) 133, 144, 145 cys-type binding 134, 146–7 heparin affinity 133–4 heparin-induced oligomerization 134, 147–8 trans-type binding 134, 146–7 fluorescence assays 176 folding intermediates 215–16 formulations, DSC 26 Freire–Biltonen double-integration procedure 207 G3 28–9 G quadruplexes 28 Gaucher disease 176

255

GCN4, bZIP domain 82, 83, 88–9 gelation, DSC 26 Gibbs free energy (DG) 6, 22, 38–9, 65, 66, 67 DGconf 13 DGhyd 13, 14 DGmol 13 DGpe 13 DGr+t 13 additive contributions 39 linear extrapolation method 204 Gill titration calorimetry 135 glaucoma 108 glucocorticoid receptor DNA-binding domain (GR DBD) 82, 83–4, 85 glucose-6-phosphate dehydrogenase (G6PD) 177, 180–3 glutamate dehydrogenase, octyl glucoside interaction 217 3’-GMP, RNase Sa binding 96–103 Grb2, SH2 domains 153, 155, 158, 161 binding inhibitors 167–8 groove binding drugs 12, 13–14, 15 guanidine, m values 210–12 3’-guanosine monophosphate (3’-GMP), RNase Sa binding 96–103 haem proteins, molten globule state 217 heat capacity change (DCp) ITC 38, 40, 66 protein–DNA association 195, 198–201, 202 SASA relationship 9 heat of dilution 42, 48–9, 50 heparin, acidic fibroblast growth factor binding 133–50 heparin-induced oligomerization 134, 147–8 HEPES buffer 45 HEW lysozyme 210–12 hexokinases PI and PII 177–9 high-throughput screens 61 histidine, ionization enthalpy 85 HIV, protease inhibitors 12 HMG-CoAR 76 Hoechst 33258 13–14 homeostasis 152 horseradish peroxidase, molten globule state 217

256

INDEX

hydration effects 17–18 hydrogen bonds 10, 84–6 hydrophobic interactions DNA–protein interaction comparison 83–4 entropy change 10 molten globule state 216, 217, 219, 220 hydrophobic surface area burial 39 hydrophobicity 12 immobilization effects, ITC 68 immune response 152 immunodeficiency 53 ‘induced fit’ 10 integrase-DNA binding domain (INTDBD), DNA binding 189–202 conformational stability 190–4 heat capacity change 195, 198–201 thermal dissociation 195, 197–8 thermal unfolding 190–2, 195, 197–8 thermodynamics 194–5 interatomic interactions 10 intercalating drugs 12, 15 intramolecular protein domain interactions, DSC 25 ionization, DH 44–5, 54, 56 irreversible systems 24 isothermal titration calorimetry (ITC) 37– 58 adiabatic jacket 40 advantages 8 applications 8–10, 53–7 buffers 44–5, 46–7 chemically-modified proteins 9 control experiments 48–50 data analysis 50–2, 64–6 data collection 64–6 data fitting 65 detectable signal/heat of interaction 43–5 drug discovery 67–77 drug–DNA interactions 8, 13–15 enthalpy change (DH) 9, 38, 39, 53–4, 65, 66, 67 enthalpy–entropy compensation 77 enzyme catalysis 18–21, 175–85 equilibrium binding constant (Kb) 38, 45, 52, 65, 66 experimental set-up 43–50

heat capacity change (DCp) 38, 40, 66 heat of dilution 42, 48–9, 50 hydration effects 17–18 immobilization effects 68 instrumentation 40–2, 63–4 integrase-DNA binding domain (INTDBD), DNA binding 194–5 machine blank 49, 50 metal ions 8 micellar systems 8 model systems 68–9 molten globule state 221–3 multiple binding events 9 Nano ITC 64 parameter range and uncertainty 67 principles 7–8 protein–carbohydrate interactions 8 protein–DNA interactions 8, 15–17, 81– 91, 194–5 protein–drug interactions 8 protonation effects 15–17, 54, 56 quality control 8 raw data 42–3 recombinant proteins 68 sensitivity 64 stoichiometry data 8, 56–7, 65 thermal inactivation, a-amylase 236–7 thermodynamic characterization 38–40 thermodynamic discrimination 10–13 thermodynamic signatures 10 titration curve 45–6 ultrasensitive 9 VP-ITC 64 ITAMs 162–6 kinases 8 kinetic equivalence 68 Le Chatalier’s principle 26 lectins, glycoconjugates 9 leukaemia 176 linear extrapolation method 204 linkers 8 liver disease 175 m values 203–14 guanidine 210–12 urea 208–9

INDEX

m1/2 values 205, 206–7 machine blank, ITC 49, 50 macromolecular structures, DSC 22, 25 malaria 176 MCS titration calorimeter 138 metal ion interactions 8 methazolamide 109, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 126, 127, 129 mevaldyl-CoA 76 micellar systems DSC 25 ITC 8 pressure perturbation calorimetry 18 Michaelis–Menten plot 20 model systems ITC 68–9 protein–DNA interactions 82–3 molten globule state 215–30 ANS binding 217 apomyoglobin 216, 217 circular dichroism 217 compact 216 cytochrome c 216, 217, 218–27 disulfide bond pairing 216 DSC 225–7 electron transfer exchange 223–5 electrostatic repulsion 216 haem proteins 217 helix–helix interaction 218 horseradish peroxidase 217 hydrophobic interactions 216, 217, 219, 220 ITC 221–3 myoglobulin 217 significance 216 spectrofluorimetry 217 stability 216, 217, 219, 220 thermodynamics 218–21 visible spectroscopy 217 X-ray angle scattering 216 mucus proteinase inhibitor 144 multiple binding events 9 MunI 15–17 MurC inhibitor 71–2 muscle contraction 8 myocardial infarction 175 myoglobulin, molten globule state 217

257

myo-inositol hexasulfate (MHS), acidic fibroblast growth factor binding 133– 50 N-SHP-2, SH2 domain 158, 160 Nano ITC 64 NMR, entropy 87, 88, 89 NMT 70 Noonan’s syndrome 153 nucleic acid binding 4 helix!coil transitions 26–30 stability 4 synthesis 4 octyl glucoside, glutamate dehydrogenase interaction 217 ‘off-pathway’ species 216 oligomerization DSC 25–6 heparin-induced 134, 147–8 OMTKY3, PPE binding 94–6 optical spectroscopy 6 order–disorder transitions 6–7 OriginTM software 245 osmotic stress 17–18 osteopetrosis 154 osteoporosis 153, 155, 167 Pace theory 218–19, 220 pancreatitis 175 Parkinson’s disease 108 parvalbumins, cation binding 8 pegylation 8 peptide–lipid binding, DSC 25 peptide synthesis 4 pH changes binding enthalpies 85–6 electrostatic interactions 12 ITC 15–17 phosphate binding, protein–protein complex 94–6 phosphoglycerate kinase 235 phosphotyrosine binding 53–4, 156–8 phosphotyrosine-binding pocket 156 PI3 kinase p85 subunit 69, 153, 158 Zap 70 subunit 9

258

INDEX

Plasmodium falciparum 176 platelet-derived growth factor (PDGF) receptors 151, 159–60, 168 Poisson–Boltzmann calculations 97, 102 polyelectrolyte effects 14, 87, 88 porcine pancreatic elastase (PPE), OMTKY3 binding 94–6 PP1-g 20, 21 PPNP 20, 21 pre-eclampsia 175 pressure perturbation calorimetry (PPC) 18 protease inhibitors 12, 94–6 protein binding 4 chemical modification 9, 205–7 construct 8–9 denaturation 24 stability 4, 25, 205–7 protein–carbohydrate interactions, ITC 8 protein–DNA interactions binding entropy 86–7 charged–charged hydrogen bonds 84–6 heat capacity change (DCp) 195, 198–201, 202 hydrophobic effect 83–4 integrase DNA binding domain 189–202 ITC 8, 15–17, 81–91, 194–5 model systems 82–3 protonation 84–6 protein–drug interactions, ITC 8 protein–ligand interaction 62, 96 protein–protein complex, anion binding 94–6 protein–protein interactions, DSC 242 protein–small-molecule interactions, DSC 25, 242 protonation effects carbonic anhydrase inhibitors 108–10, 119–23 ITC 15–17, 54, 56 protein–DNA interactions 84–6 psychrophilic enzymes 231–40 a-amylase 233, 236–7 catalytic efficiency 232 chitobiase 235 DNA ligase 233 flexibility 234–6 ITC 236–7

kinetic parameters 237–9 phosphoglycerate kinase 235 thermal inactivation 236–7 thermal stability 232, 233–4 xylanase 233 PTPN1, mis-sense mutation 153 pY 53 pYEEI 53, 54, 156, 158 pYEEV 54 quadruplex DNA 28–30 quality control procedures 8 quantitative SAR (QSAR) 11, 12 receptor binding, conformational changes 10 recombinant proteins 68 reporter groups 9 restriction endonucleases 15 ribonuclease A (RNase A) crystal structure 98 inhibitor binding 245–8 ribonuclease–ligand interactions 93–105 ribonuclease Sa (RNase Sa), 3’-GMP binding 96–103 ribozymes 175 ‘rigid body’ interactions 10 rosuvastatin 75–6 saline effects 12 acidic fibroblast growth factor, heparin/ MHS binding 143–5 ribonuclease–ligand interactions 93–105 SAP 152, 156, 160 SAP–SLAM binding 160 scanning alanine mutagenesis 12 screening 25, 61, 242 SH2 domains 151–73 binding inhibitors 166–8 cellular process regulation 152 discovery 152 drug target 153, 154, 166–9 ITC 53–4 pathologies 53, 152–5 peptide-binding specificity 158–66 phosphotyrosine binding 156–8 specificity-determining region 156 structure and ligand binding 155–66

INDEX

tandem 155, 162–6 tyrosine residues 53, 151–2 SH2D1A 152 shelf life 24–5 SHP2 phosphatase 153 Shrake–Rupley algorithm 138 signal transduction 151 signalling pathways, conformational changes 10 Simplex algorithm 137 SLAM 152, 160 solvent accessible surface area (SASA), DCp relationship 9 SOX-5 HMG box 202 spectrofluorimetry, molten globule state 217 Src homology 2 domains see SH2 domains Src kinase, SH2-domain 53–4, 153, 154–5, 156–8, 159, 161 binding inhibitors 166–7 Sso7d 82, 83, 87–8 statin inhibitors 76 stoichiometry data, ITC 8, 56–7, 65 Streptomyces griseus protease B (SGPB) 94 structure–activity relationships (SARS) drug discovery 11, 73 quantitative (QSAR) 11, 12 sulfanilamide 109, 111, 112, 113, 114, 115, 116, 121, 122, 124, 126, 127, 129 sulfate binding, protein–protein complex 96 surface area 138, 141, 145–6 Syk kinase, SH2 domains 153, 155, 162–4, 165–6 Tm 24–6 Tanford–Kirkwood calculations 102 TAT binding protein–DNA interactions, osmotic stress 18 Taylor expansion 205–6 TCEP 47 telomerase 28 telomeres, G quadruplexes 28 temperature 12 enthalpy change (DH) 43–4, 54 protein stability 205–7 textiles 176 TFIIIA 202 thermal denaturation 203

259

thermodynamic characterization, ITC 38– 40 thermodynamic discrimination, ITC 10–13 thermodynamic hypothesis 218 thermodynamic signatures 10 thermodynamic stability, DSC 23, 24–5 thermodynamics 4–5 function/structure correlations 9–10 thrombin 144 titration curve, ITC 45–6 transcription 8 transition cooperativity 22 triclosan 70–1 trifluoromethane sulfonamide 109, 111, 112, 113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 126–7, 130 triplex DNA, melting 18 TRIS buffer 45, 85 troponin C, cation binding 8 turkey ovomucoid third domain (OMTKY3), PPE binding 94–6 tyrosine residues 53, 151–2 ubiquitin 210–12 UHBD program suite 97, 102 urea, m values 208–9 van’t Hoff enthalpy (DHvH) 22, 23, 66 van’t Hoff relationship 6 visible spectroscopy, molten globule state 217 VP-Capillary DSC Platform 18, 242–9 VP-ITC 64 VPViewerTM software 244 water 17 wound healing 151 X-linked agammaglobulinaemia 153 X-linked lymphoproliferative disease (XLP) 152 X-ray angle scattering, molten globule state 216 xylanase, cold adaptation 233 Zap 70 9, 153, 155, 164–5 zinc fingers–DNA complex 202

Index compiled by Christine Boylan