Expression and Analysis of Recombinant Ion Channels

Edited by Jeffrey J. Clare and Derek J.Trezise Related Titles

Daunert, S., Deo, S. K. (eds.) Sibley, D. R. Photoproteins in Bioanalysis Receptor Biochemistry and approx. 300 pages Methodology Series Hardcover 6 Volume Set ISBN 3-527-31016-9 approx. 2500 pages 2002 Dingermann, T., Steinhilber, D., Hardcover ISBN 0-471-32258-X Folkers, G. (eds.) Molecular Biology in Medicinal Chemistry Novartis

435 pages with 113 figures and 32 tables Novartis Foundation 2004 Symposium 245 – Ion Channels – Hardcover From Atomic Resolution Physiology ISBN 3-527-30431-2 to Functional Genomics

284 pages Babine, R. E., Abdel-Meguid, S. S. (eds.) 2002 Protein Crystallography in Hardcover ISBN 0-470-84375-6 Drug Discovery

278 pages with 131 figures and 11 tables 2004 Hardcover ISBN 3-527-30678-1 Expression and Analysis of Recombinant Ion Channels

From Structural Studies to Pharmacological Screening

Edited by Jeffrey J. Clare and Derek J.Trezise The Editors & All books published by Wiley-VCH are carefully produced. Nevertheless, authors, editors, and Dr. Jeffrey J. Clare publisher do not warrant the information con- GlaxoSmithKline tained in these books, including this book, to be Department of Gene Expression free of errors. Readers are advised to keep in mind and Protein Biochemistry that statements, data, illustrations, procedural Gunnels Wood Road details or other items may inadvertently be inaccu- Stevenage, SG1 2NY rate. Great Britain Library of Congress Card No.: Dr. Derek J. Trezise applied for GlaxoSmithKline Department of Assay Development British Library Cataloguing-in-Publication Data Gunnels Wood Road A catalogue record for this book is available from Stevenage SG1 2NY the British Library. Great Britain Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at .

 2006 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim

All rights reserved (including those of translation into other languages). No part of this book may be reproduced in any form – by photoprinting, micro- film, or any other means – nor transmitted or translated into a machine language without written permission from the publishers. Registered names, trademarks, etc. used in this book, even when not specifically marked as such, are not to be considered unprotected by law.

Typesetting ProSatz Unger,Weinheim Printing Strauss GmbH, Mörlenbach Binding J. Schäffer GmbH i.G., Grünstadt

Printed in the Federal Republic of Germany Printed on acid-free paper

ISBN-13: 978-3-527-31209-2 ISBN-10: 3-527-31209-9 V

Contents

Preface XI List of Contributors XIII Color Plates XVII

1 Expression of Ion Channels in Xenopus Oocytes 1 Alan L. Goldin 1.1 Introduction 1 1.2 Advantages and Disadvantages of Xenopus Oocytes 2 1.3 Procedures for Using Oocytes 3 1.4 Types of Analyses 5 1.4.1 Electrophysiological Analysis 5 1.4.1.1 Two-electrode Whole Cell Voltage-clamp 5 1.4.1.2 Cut-open Oocyte Voltage-clamp 7 1.4.1.3 Macropatch Clamp 9 1.4.1.4 Single Channel Analysis 11 1.4.2 Biochemical Analysis 12 1.4.3 Compound Screening 13 1.4.3.1 Serial Recording Using the Roboocyte 14 1.4.3.2 Parallel Recording Using the OpusXpress 16 1.5 Examples of Use 17 1.5.1 Characterization of cDNA Clones for a Channel 17 1.5.2 Structure–Function Correlations 18 1.5.3 Studies of Human Disease Mutations 19 1.6 Conclusions 21 Acknowledgments 21 References 21

2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels 27 Louisa Stevens, Andrew J. Powell, and Dennis Wray 2.1 Introduction 27 2.2 Methods for cDNA Subcloning 28

Expression and Analysis of Recombinant Ion Channels. Edited by Jeffrey J. Clare and Derek J. Trezise Copyright # 2006 WILEY-VCH Verlag GmbH & Co. KGaA,Weinheim ISBN: 3-527-31209-9 VI Contents

2.2.1 Conventional Sub-cloning Using Restriction Enzymes and DNA Ligase 28 2.2.2 PCR-based cDNA Sub-cloning 31 2.2.3 Sub-cloning cDNA through Site-specific Recombination 33 2.3 Generation of Chimeric Channel cDNAs 36 2.3.1 Use of Restriction Enzymes to Generate Chimeric Channel cDNAs 36 2.3.2 PCR-mediated Overlap Extension for Chimera Generation 39 2.3.3 PCR-mediated Integration or Replacement of cDNA Fragments 43 2.4 Site-directed Mutagenesis 43 2.4.1 Examples of the Use of Site-directed Mutagenesis 45 2.4.2 Modification of the QuikChange Method for the Replacement of cDNA Fragments 50 2.5 Epitope-tagged Channels and Fusion Partners 50 2.6 Channel Subunit Concatamers 52 2.7 Concluding Remarks 53 References 54

3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology 59 Paul B. Bennett, Niki Zacharias, John B. Nicholas, Sue Dee Sahba, Ashutosh Kulkarni, and Mark Nowak 3.1 Introduction 59 3.2 Unnatural Amino Acid Mutagenesis Methodology 60 3.3 Unnatural Amino Acid Mutagenesis for Ion Channel Studies 64 3.4 Structure–Function Example Studies 65 3.4.1 Nicotinic Acetylcholine Receptor 65 3.4.2 Drug Interactions with the hERG Voltage-gated Potassium Ion Channel 67 3.5 Other Uses of Unnatural Amino Acids as Probes of Protein Structure and Function 72 3.6 Conclusions 73 Acknowledgements 74 References 74

4 Functional Expression of Ion Channels in Mammalian Systems 79 Jeff J. Clare 4.1 Introduction 79 4.2 cDNA Cloning and Manipulation 80 4.3 Choice of Host Cell Background 81 4.4 Post-translational Processing of Heterologous Expressed Ion Channels 85 4.5 Cytotoxicity 90 4.6 Transient Expression Systems 91 4.6.1 “Standard” Transient Expression 91 4.6.2 Viral Expression Systems 92 Contents VII

4.7 Stable Expression of Ion Channels 96 4.7.1 Bicistronic Expression Systems 96 4.7.2 Stable Expression of Multiple Subunits 100 4.7.3 Inducible Expression 101 4.8 Summary 103 Acknowledgements 103 References 104

5 Analysis of Electrophysiological Data 111 Michael Pusch 5.1 Overview 111 5.2 Introduction 111 5.3 Expression Systems and Related Recording Techniques 113 5.3.1 Expression in Xenopus Oocytes 113 5.3.2 Expression in Mammalian Cells 115 5.3.3 Leak and Capacitance Subtraction 116 5.4 Macroscopic Recordings 117 5.4.1 Analysis of Pore Properties – Permeation 118 5.4.2 Analysis of Fast Voltage-dependent Block – the Woodhull Model 121 5.4.3 Information on Gating Properties from Macroscopic Measurements 122 5.4.3.1 Equilibrium Properties – Voltage-gated Channels 124 5.4.3.2 Equilibrium Properties – Ligand Gated Channels 126 5.4.3.3 Macroscopic Kinetics 129 5.4.4 Channel Block 132 5.4.5 Nonstationary Noise Analysis 133 5.4.6 Gating Current Measurements in Voltage Gated Channels 135 5.5 Single Channel Analysis 136 5.5.1 Amplitude Histogram Analysis 136 5.5.2 Kinetic Single Channel Analysis 138 5.6 Summary 142 Acknowledgements 142 References 142

6 Automated Planar Array Electrophysiology for Ion Channel Research 145 Derek J Trezise 6.1 Introduction 145 6.2 Overview of Planar Array Recording 145 6.3 Experimental Methods and Design 147 6.3.1 Cell Preparation 148 6.3.2 Cell Sealing and Recording 149 6.3.3 Drug Application 152 6.3.4 Experimental Design and Data Analysis 155 6.4 Overall Success Rates and Throughput 158 6.5 Population Patch Clamp 159 6.6 Summary and Perspective 162 VIII Contents

Acknowledgments 162 References 162

7 Ion Flux and Ligand Binding Assays for Analysisof Ion Channels 165 Georg C. Terstappen 7.1 Introduction 165 7.2 Ion Flux Assays 166 7.2.1 Radioactive Ion Flux Assays 167 7.2.2 Nonradioactive Ion Flux Assays based on Atomic Absorption Spectrometry 168 7.2.2.1 Nonradioactive Rubidium Efflux Assay 168 7.2.2.2 Nonradioactive Lithium Influx Assay 174 7.2.2.3 Nonradioactive Chloride Influx Assay 174 7.2.2.4 Conclusions 174 7.3 Ligand Binding Assays 175 7.3.1 Heterogeneous Binding Assays Employing Radioligands 177 7.3.2 Homogeneous Binding Assays Employing Radioligands 178 7.3.3 Homogeneous Binding Assays Employing Fluorescent-Labeled Ligands and Fluorescence Polarization 180 7.3.4 Conclusions 181 Acknowledgements 182 References 182

8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes 187 Jesús E. González, Jennings Worley, and Fredrick Van Goor 8.1 Introduction 187 8.2 Membrane Potential Probes 188 8.2.1 Redistribution Probes 188 8.2.2 FRET Probes 190 8.2.3 Advantages and Limitations of Membrane Potential Probes 192 8.3 Ion-sensitive Fluorescent Probes 194 8.3.1 Calcium Dyes 194 8.3.2 Indicators of Other Ions 195 8.4 Fluorescence Assays for Ion Channels 196 8.4.1 Calcium Channels 196 8.4.2 Non-voltage-gated Calcium Permeable Channels 197 8.4.3 Sodium Channels 200 8.4.4 Potassium Channels 201 8.4.5 Chloride Channels 203 8.5 Assays for Monitoring Channel Trafficking 205 8.6 Summary 207 References 208 Contents IX

9 Approaches for Ion Channel Structural Studies 213 Randal B. Bass and Robert H. Spencer 9.1 Introduction 213 9.2 Expression of Membrane Proteins for Structural Studies 216 9.2.1 Mammalian Expression 216 9.2.2 Insect Expression 217 9.2.3 Yeast Expression 217 9.2.4 Bacterial Expression 218 9.3 The Detergent Factor 219 9.4 Purification 223 9.5 Crystallization 227 9.6 Use of Antibody Fragments 229 9.7 Generation of First Diffraction Datasets 230 9.8 Selenomethionine Phasing of Membrane Proteins 232 9.9 MAD Phasing and Edge Scanning 233 9.10 Negative B- factor Application (Structure Factor Sharpening) 234 9.11 Conclusions 235 References 235

10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels 241 Daniele Bemporad, Alessandro Grottesi, Shozeb Haider, Zara A. Sands, and Mark S.P. Sansom 10.1 Introduction 241 10.2 Computational Methods 242 10.3 Kir Channels 246 10.3.1 Structures 246 10.3.2 Molecular Modeling 247 10.3.3 Simulations 248 10.3.4 Filter Flexibility 248 10.3.5 M2 Helices and Hinge Motion 250 10.3.6 Intracellular Domain Dynamics 251 10.3.7 Interactions with Ligands 251 10.3.8 Towards an Integrated Gating Model 253 10.4 Kv Channels 254 10.4.1 Structures 254 10.4.2 S6 Helices, Hinges and Gating 256 10.4.3 The Barrier at the Gate 257 10.4.4 The Nature of the Voltage Sensor 258 10.4.5 A Possible Gating Model 260 10.5 Summary and Future Directions 261 Acknowledgements 262 References 262

Subject Index 269 XI

Preface

Given the exciting advances that have been made within the last few years there can seldom have been a more opportune time to collate a book on ion channel research methods. Molecular genetics has provided us with the sequence of the human ion channel genome and an understanding of ‘channelopathies’ – the link between ion channel gene mutations and pathology. Advances in gene expression and functional analysis methods have leveraged a greater understanding of ion channel structure/function as well as accelerating the quest for new ion channel drugs. Computational approaches are also providing further insight into the molecular dynamics of ion channels at the atomic level. Excitingly, in June 2005 through the pioneering work of Mackinnon and co-workers, large scale protein purification and X-ray crystallography have revealed for the first time the full structure of a eukaryotic ion channel, Shaker KV1.2, resolved to 2.9 Å. The opportunity for discovery in the ion channel field has never been greater. Presently, we know of approximately 350 ion channel encoding genes in humans. Of these, 143 comprise the voltage-gated ion channel superfamily making this the third largest gene family after the G-protein-coupled receptors and protein kinases (Yu & Catterall, 2004, Sci. STKE 253, 1-17). These channels are specialised for electrical signalling and ionic homeostasis and include the well characterised voltage-gated Na+,Ca2+ and K+ channel subfamilies as well as transient receptor potential (TRP) and cyclic-nucleotide-gated channels. Other ion channels are ligand-gated and well adapted for fast synaptic transmission, notably nicotinic acetylcholine receptors, GABA- and glutamate activated channels and ATP-gated P2X receptors. The remainder are currently categorised as ‘miscellaneous’. These include channels as structurally and functionally diverse as aquaporins, ABC-transporter-like proteins such as the Cystic-fibrosis transmembrane regulator (CFTR) and volume- and Ca2+-activated chloride channels. Heteromeric assembly of different ion channel subunits, and their regulation by alternative splicing and by cellular effectors such as kinases, nuclear receptors, integrins and GPCRs further amplifies this diversity in function. For the research scientist, the challenges ahead lie in rationalising which channel variants and complexes are physiologically and pathologically relevant and how these proteins function, down to the molecular and atomic level. For the drug discoverer the ability to use this knowledge-base and experimental techniques to rapidly find safe and efficacious new ion channel modulators is paramount.

Expression and Analysis of Recombinant Ion Channels. Edited by Jeffrey J. Clare and Derek J. Trezise Copyright # 2006 WILEY-VCH Verlag GmbH & Co. KGaA,Weinheim ISBN: 3-527-31209-9 XII Preface

This book is directed at both the new and the experienced ion channel researcher wishing to learn more about the considerations and methods for studying recombinant ion channels. Our aim is that it should be of interest to academic and industrial workers alike. Chapters 1 to 3 cover the use of the Xenopus oocyte expression system for structure-function studies, from basic approaches for manipulating ion channel cDNAs to more specialised but powerful techniques such as unnatural amino acid substitution. This is followed by reviews of strategies and methodologies available for expressing channels in mammalian cells and for their analysis by patch-clamp electrophysiology. In Chapters 6 to 8, the latest methodologies for ion channel drug discovery are reviewed, including high throughput screening using fluorescence and luminescence as well as automated planar array electrophysiology. The remaining 2 chapters focus on approaches for determining ion channel crystal structures and on computational approaches to understanding channel mechanisms at atomic resolution. Our goal is to have covered the spectrum of methods that are relevant to recombinant ion channel expression and analysis. Rather than provide detailed protocols, for which readers are directed to appropriate references in each chapter, the aim is to provide an overview of the techniques involved, reviewing underlying principles and providing working guidelines as well as an understanding of the key theoretical and practical considerations associated with each topic. In each case, this practical advice is illustrated by real life examples, taken either from the author’s own experience or from key examples in the literature. In summary, we hope to have compiled a compendium of practical ion channel information that will prove a valuable resource to the reader. We gratefully acknowledge the efforts of each of the expert authors who have provided contributions, without which this book would not be possible. We would also like to extend our sincere thanks to Jane Sanders from GSK for excellent ad- ministrative assistance and Steffen Pauly, Frank Weinreich and co-workers at Wiley for help and guidance through the publication process.

Stevenage, October 2005 Jeffrey J. Clare Derek J. Trezise XIII

List of Contributors

Randal B. Bass Alan L. Goldin Amgen Inc. Department of Microbiology Department of Analytical Sciences & Molecular Genetics 1201 Amgen Court West University of California Seattle Irvine WA 98119 CA 92697–4025 USA USA

Daniele Bemporad Jesu´s E. Gonza´lez Department of Biochemistry Vertex Pharmaceuticals, Inc. University of Oxford Discovery Biology South Parks Road 11010 Torreyana Road Oxford San Diego OX1 3QU CA 92121 UK USA

Paul B. Bennett Alessandro Grottesi Neurion Pharmaceuticals, Inc. Department of Biochemistry 180 N. Vinedo Avenue University of Oxford Pasadena South Parks Road CA 91107 Oxford USA OX1 3QU UK Jeff J. Clare Department of Gene Expression Shozeb Haider and Protein Biochemistry Department of Biochemistry GlaxoSmithKline University of Oxford Stevenage South Parks Road Herts Oxford SG1 2NY OX1 3QU UK UK

Ashutosh Kulkarni Zara A. Sands Neurion Pharmaceuticals, Inc. Department of Biochemistry 180 N. Vinedo Avenue University of Oxford Pasadena South Parks Road CA 91107 Oxford USA OX1 3QU UK John B. Nicholas Neurion Pharmaceuticals, Inc. Mark S.P. Sansom 180 N. Vinedo Avenue Department of Biochemistry Pasadena University of Oxford CA 91107 South Parks Road USA Oxford OX1 3QU Mark Nowak UK Neurion Pharmaceuticals, Inc. 180 N. Vinedo Avenue Robert H. Spencer Pasadena Department of Molecular Neurology CA 91107 Merck Research Laboratories USA P.O. Box 4 West Point Andrew J. Powell PA 19486 Department of Gene Expression and USA Protein Biochemistry GlaxoSmithKline Louisa Stevens Stevenage Faculty of Biomedical Sciences Herts University of Leeds SG1 2NY Leeds LS2 9JT UK UK

Michael Pusch Georg C. Terstappen Istituto di Biofisica Sienabiotech S.p.A. Consiglio Nazionale delle Ricerche Discovery Research Via de Marini 6 Via Fiorentina 1 16149 Genova 53100 Siena Italy Italy

Sue Dee Sahba Derek J. Trezise Neurion Pharmaceuticals, Inc. Department of Assay Development 180 N. Vinedo Avenue GlaxoSmithKline Pasadena Medicines Research Centre CA 91107 Gunnels Wood Road USA Stevenage SG1 2NY UK List of Contributors XV

Fredrick VanGoor Dennis Wray Vertex Pharmaceuticals, Inc. Faculty of Biomedical Sciences Discovery Biology University of Leeds 11010 Torreyana Road Leeds San Diego LS2 9JT CA 92121 UK USA Niki Zacharias Jennings Worley III Neurion Pharmaceuticals, Inc. Vertex Pharmaceuticals, Inc. 180 N. Vinedo Avenue Discovery Operations Pasadena 11010 Torreyana Road CA 91107 San Diego USA CA 92121 USA XVII

Color Plates

Fig. 2.1 Modification of multiple cloning site The annealed linker contains matching (MCS) in a vector by insertion of a synthetic BamHI and HindIII sticky ends, as well as the oligonucleotide linker. The figure shows, on sequences for the new restriction sites to be the left, part of the MCS of a vector, which is introduced (XbaI, EcoRI and NotI). After liga- cut with BamHI and HindIII, removing a tion of the linker into the linear vector (bot- SnaBI site. The linear vector is then dephos- tom of figure), the resulting vector contains phorylated to prevent religation. On the right, XbaI, EcoRI and NotI restriction sites instead two synthetic oligonucleotides are phosphory- of SnaBI. (This figure also appears on page lated with polynucleotide kinase (PNK) and 30.) annealed to form a double-stranded linker.

Expression and Analysis of Recombinant Ion Channels. Edited by Jeffrey J. Clare and Derek J. Trezise Copyright # 2006 WILEY-VCH Verlag GmbH & Co. KGaA,Weinheim ISBN: 3-527-31209-9 XVIII Color Plates tran- in vitro only the nonsense codon is createdacid with appended. the The unnatural gene amino of interestscribed (cDNA mRNA) or and the tRNA arewhere introduced the into protein a is cell expressed andphysiology. (This detected figure using also electro- appears on page 61.) system with electrophysiology read- The fundamental protocol for incorporating unna- in vivo” translation out. A special nonsense codon isof introduced interest into at the the cDNA position of interest. A tRNA that recognizes Fig. 3.1 tural amino acids through nonsense suppression,an showing “ Color Plates XIX

Fig. 3.3 Correlation between the change in in a decreased binding energy. Binding energy

EC50 and calculated binding energy in nicoti- was calculated in the gas phase and the abso- nic acetylcholine receptor binding pocket. lute values will be scaled down by the pre- Tryptophan 149 was systematically replaced sence of water and other factors but the trend with fluorinated (F) tryptophan (Trp, 1–4 F) is expected to remain the same. (This figure unnatural amino acids. Each added F further also appears on page 66.) deactivates the Trp p electron cloud, resulting XX Color Plates

Fig. 3.4 Homology models of hERG closed (left) and open ion conducting (right) states. Only the S5-P-S6 segments are shown. Many drugs enter the open channel and bind to residues along the S6 segment. (This figure also appears on page 68.) Color Plates XXI

Figure 3.5 hERG MAP astemizole: A hERG noncovalent binding interaction with the MAP elucidates the nature and relative impor- channel. Each compound displays a unique tance of specific drug–channel interactions. hERG MAP signature. In this example, The right hand bar graph shows the change in changes at serine 624 suggest H-bond inter- astemizole binding energy (kcal mol–1)at actions with the compound. Progressive each position of the channel when that posi- changes in binding as fluorine (F) is added to tion is altered. These changes in binding en- phenylalanine at position 652 indicate cation– ergy may be interpreted in terms of atomic le- p or p–p aromatic interactions. The chemical vel interactions such as hydrogen-bonding, structure of astemizole is shown. (This figure cation–p, hydrophobic, and ion pairing. Each also appears on page 69.) hERG mutant is designed to identify a specific XXII Color Plates

Fig. 4.4 Effect of reduced cell culture tempera- (see Fig. 6.2). Note that, in the HERG-expres- ture on ion channel expression. (A) Confocal sing cells, at 37 8C and 30 8C immunoreactivity images of stable HERG-expressing CHO cells is uniform and granular throughout the cyto- grown at different temperatures, following im- plasm whereas at 27 8C there is an accumula- munocytochemical staining with a HERG-spe- tion in giant lysosome-like bodies. (B) Quanti- cific antibody. The level of HERG immunoreac- tative analysis of HERG immunoreactivity by tivity is significantly increased in cells grown at flow cytometry confirms the increases seen at 27 8C and 30 8C compared to 37 8C. No immu- lower growth temperatures, as indicated by noreactivity is seen in untransfected CHO cells the rightward shift in peak fluorescence (X- (CHO-wt). The increase in total HERG protein axis) observed within the populations of cells at 30 8C seen here is also mirrored by an in- grown at 27 8C and 30 8C compared to 37 8C. crease in surface-localised functional protein (This figure also appears on page 89.) as measured by IonWorks electrophysiology Color Plates XXIII

Fig. 4.5 The BacMam expression system. ome when transfected into the appropriate (A) Map of the pFastBacMam1 shuttle vector E.coli host. (B) Workflow for the generation of which used to generate recombinant Bac- recombinant BacMam virus stock and trans- Mam viruses. The gene of interest is inserted duction into mammalian cells for expression. downstream of the CMV promoter where it is Reproduced with permission from Ref. [77]. flanked by Tn7 inverted repeats that direct (This figure also appears on page 94.) site-specific transposition into the viral gen- XXIV Color Plates

Fig. 8.4 hTRPV1 response in 3456 microtiter enlarged, represents an 11 point concentra- plate using fluorescent Ca2+ indicator Fluo-3/ tion response analysis to a single compound Fura-red readout. (A) HEK cells stably expres- in triplicate as well as capsaicin, positive sing hTRPV1 were plated in a 3456-nanoplate. (capsaicin plus antagonist) and negative Cells were loaded with Fluo-3AM/Fura-red and (DMSO) controls with N = 1. (B) Concentra- capsaicin (200 nM) was added to evoke a cal- tion–response curve of capsazepine block of cium transient measured using a fluorescent capscaicin-stimulated hTRPV1 is shown. (This plate reader. Each 6X6 square, one example is figure also appears on page 198.) Color Plates XXV

Fig. 8.7 CFTR membrane potential assay de- 100% wild-type CFTR-expressing cells. In the monstrates efficacy limitations compared to fluorescence-based assay, the half maximal re- flux assays: Response to CFTR activation be- sponse was observed at ~3% wild-type CFTR tween fluorescent-based and electrophysiolo- and was nonlinear as the concentration of gical assay formats. To monitor the response wild-type CFTR expressing cells was in- to increasing amounts of CFTR activation creased. In contrast, the half-maximal re- cells expressing wild-type CFTR were mixed sponse in the using chamber assay was with parental cells in the indicated propor- reached at ~60% wild-type CFTR and was lin- tions (% wild-type CFTR). The response to ear. These results highlight the nonlinearity of CFTR activation using a maximal concentra- the fluorescence-based assays, which can tion of forskolin was monitored in both the limit the SAR evaluation of agonist efficacybe- fluorescence assay (black circles) and in an cause the sensitive response saturates with electrophysiological assay (red circles) under low amounts of CFTR. (This figure also ap- voltage-clamp control. The response in both pears on page 204.) assays was normalized to the response using XXVI Color Plates Color Plates XXVII

3 Fig. 8.8 CFTR membrane potential assay de- genous K+ conductance. Only at very low CFTR

monstrates the dependence of channel density densities did the EC50 approximate the Kd for on agonist sensitivity: Effects of CFTR density forskolin. (C) Tomonitor the effects on CFTR on agonist activity in fluorescent membrane density on agonist stimulation in a fluores- potential assays. (A) Theoretical Michaelis– cence-based membrane potential assay, CFTR- Menten type increase in open probability of expressing cells were mixed with parental cells CFTR by forskolin (Kd = 5 mM). (B) Simulation at the indicated proportions and expressed as of the membrane potential response to CFTR % wild-type (wt) CFTR. As observed in the si- activation by forskolin. The concentration de- mulations, only at very low wt-CFTR propor-

pendent effects on membrane potential were tions did the EC50 for forskolin approximate its calculated using the Goldman–Hodgkin–Katz Kd. These results illustrate that the potency of equation incorporating the open probability at ion channel agonists in a fluorescence-based each agonist concentration and CFTR densities assay is highly sensitive to the channel density. of 0.01 to 100 times the background permeabil- (This figure also appears on page 206.) ity,which was assumed to be due to an endo- XXVIII Color Plates

Fig. 9.4 Examples of X-ray diffraction images. Note the significant improvement in the dif- (A) Initial diffraction pattern from a single fraction limit, as compared to (B), extending crystal of the MscL ion channel protein from to 3.5 Å. This image also illustrates the signifi- M. tuberculosis (Tb-MscL) prior to the optimi- cant anisotropy and intensity decay of the re- zation of crystallization and cryoprotection flections often observed with membrane pro- conditions. (B) Following optimization of the teins. (D) Ribbon diagram of the resulting crystallization and cryoprotection conditions 3.5 Å structure of the M. tuberculosis MscL, a for a crystal of Tb-MscL that diffracted to a mechanosensitive channel, reproduced with limiting resolution of 7 Å. (C) Diffraction pat- permission from Ref. [15]. (This figure also tern of a Tb-MscL crystal after soaking with appears with the color plates.) (This figure

the heavy atom compound, Na3Au(S2O3)2. also appears on page 231.) Color Plates XXIX

Fig. 10.7 Interactions of PIP2 with Kir6.2. sible binding site for PIP2. (B) Snapshot from (A) A molecular surface representation of a a simulation of three PIP2 molecules within a model of the Kir6.2 channel calculated using POPC bilayer. The PIP2 molecules are shown GRASP [115], color on electrostatic potential in space-filling format whilst the phosphorus (from –6.6 to 5.4 kT, red to blue). The region atoms of the POPC headgroups are shown as of positive electrostatic potential (blue) near green spheres. (This figure also appears on the intracellular membrane/water interface page 253.) (indicated by the circle) corresponds to a pos-

Fig. 10.13 The KvAP voltage sensor. (A) The Cas during the course of each simulation (on voltage sensor embedded in a detergent a scale from blue = 0.0 Å to red = 4.8 Å). The (DMA) micelle shown at the end (t=40 ns) of loss of helicity in S3a is evident. (C) KvAP S4 an MD simulation. (B) The structure of the helix hinge-swivelling about residue I130, as S2–S3 region of the VS at the end of a 40 ns revealed by eigenvector 1 of an MD simulation MD simulation at 368 K. The residues are co- at 368 K. The colors indicate the range of mo- lored according to the magnitude of root tions represented by this eigenvector. (This mean square fluctuations experienced by the figure also appears on page 260.) 1

1 Expression of Ion Channels in Xenopus Oocytes Alan L. Goldin

1.1 Introduction

Xenopus oocytes have been widely used for studying ion channels in a controlled in vivo environment since the system was initially developed for this purpose by Miledi and coworkers [1, 2]. There have been at least five major types of studies using oocytes to examine ion channel expression. The earliest use was to examine the properties of specific ion channels in a living cell free from other responses. The oocytes were injected with RNA isolated from whole brains, and the responses were analyzed using the two-microelectrode whole cell voltage- clamp [2, 3], the patch clamp [4], or a variety of biochemical techniques [5, 6]. Once the responses were isolated, Xenopus oocytes were then used in a second type of study as an assay system to isolate cDNA clones encoding the proteins involved. For example, cDNA clones encoding the 5-HT1C receptor were isolated using electrophysiological assays, both by hybrid depletion [7] and by directly transcribing RNA from a cDNA library and injecting the transcripts into oocytes [8]. These types of studies are much less commonly used now because of the large number of available heterologous expression systems and cDNA clones encoding ion channels. The third major type of study for which the Xenopus oocyte expression system has been, and continues to be, particularly useful is the correlation of molecular structure with electrophysiological function of a specific channel. The two basic approaches have been to construct defined mutations whose effects are determined by expression in oocytes, and to construct chimeric molecules between two closely related channels or receptors followed by expression in oocytes and electrophysiological analysis. These types of approach are still commonly used but with more sophisticated structural alterations and functional analyses. The fourth general approach utilizing expression in oocytes is to determine the functional effects of mutations that cause human diseases. These types of studies are also performed using expression in other heterologous systems such as mammalian cells, and the advantages and disadvantages of each approach will be discussed in Section 1.5.3.

The final approach that takes advantage of expression in Xenopus oocytes is to screen potential drugs to determine their relative efficacies against specific types of ion channels. These studies have been made feasible by the development of automated voltage-clamp devices. Two such devices are the Roboocyte from Multi- channel Systems and the OpusXpress from Axon Instruments, which is now part of Molecular Devices. The features of these instruments are described in Section 1.4.3.

1.2 Advantages and Disadvantages of Xenopus Oocytes

While the Xenopus oocyte system is a valuable tool for the study of ion channel function, there are a number of important factors to consider in deciding whether oocytes are the most appropriate system to use. One of the primary advantages of oocytes is that these cells do not express a large number of ion channels and receptors, so that the exogenous protein can be studied without contamination from endogenous channels. This advantage is not true in all cases, however, as oocytes do express some channels and receptors [9]. These responses are not usually a problem for two reasons. First, only some oocytes express the channels, so that it is frequently possible to obtain oocytes that do not have endogenous current. Second, current from the injected RNA is usually much larger than that through the endogenous channels, so that it is generally possible to record the expressed current without significant contamination from native oocyte currents. In some cases, the presence of an endogenous response can be used to advantage as a second messenger system that is coupled to the initial response that is being studied. Another advantage is that some channels can only be expressed in oocytes and not in mammalian cells. It is not possible to predict which channels fall into this category, and success using mammalian cells often depends very strongly on choosing the appropriate cell type (see Chapter 4). Even when the channels can be expressed in other systems, oocytes may still be advantageous for examination of the roles of different subunits. Since expression in oocytes involves injection of RNA, it is possible to adjust the ratio of RNA encoding each subunit and thus examine channels with a relatively well controlled composition. In contrast, it is more difficult to control the ratios of different subunits expressed by transfection in mammalian cells. There are also some advantages to the use of oocytes with respect to electrophysiological recording. The oocyte system is particularly well suited for the study of many different mutations because injection and two-electrode voltage-clamping can be carried out rapidly and in a semi-automated fashion. In addition, studies involving modulation by second messenger systems such as phosphorylation are particularly well suited to oocytes because it is possible to express multiple proteins in the same cell, and modulators can be injected while recording with the two-electrode voltage-clamp. Finally, some analysis techniques are unique to oo- 1.3 Procedures for Using Oocytes 3 cytes. For example, the cut-open oocyte voltage-clamp was developed specifically for high resolution electrophysiological recording from oocytes and it is particularly well-suited for analyzing fast ionic and gating currents [10, 11]. There are a number of disadvantages to using oocytes for expression of ion channels. First, as pointed out above, oocytes do express some endogenous channels and receptors. A second major disadvantage is that it has not been possible to express every channel in oocytes. On the other hand, most channels that have not been expressed in oocytes have not been expressed in other heterologous systems either, so that this problem is not unique to the use of oocytes. Another disadvantage of oocytes is that many pharmacological agents are less potent on channels in oocytes compared to the channels in mammalian cells or native tissues. This difference in potency most likely reflects decreased accessibility of the drug because of the large number of invaginations in the oocyte membrane, the vitelline membrane surrounding the oocyte surface, or the follicle cells around the oocyte. However, although the absolute concentration of drug that is required for block is often higher than that required in native tissues, the relative efficacies of drugs against different channels are generally representative of those in native tissues. Other potential disadvantages with the use of oocytes include the need for procedures and equipment beyond that usually found in a standard research laboratory, occasional wide variations in quality due to seasonal and other factors, and the fact that they are best maintained at ambient temperature which may lead to altered synthesis and processing of mammalian channels compared to physiological conditions. The most serious disadvantage of using expression in oocytes as an assay system is that the cells are not the native cells in which the channels are normally expressed. This can be reflected in two major ways. First, the functional properties that are observed may not be identical to those characterized in native tissues, although it is often difficult to make a direct comparison because the native cell usually contains multiple different types of channels. In addition, the functional properties may depend on the subunit composition, in which case the oocyte system can be used to determine which subunits are required for properties similar to those observed in vivo. The second consequence of oocytes not being the native tissue is that cellular trafficking is different, particularly in comparison to neurons. Because of this difference, some channels may not be expressed on the cell surface in oocytes and the effects of mutations that affect trafficking cannot be studied at all.

1.3 Procedures for Using Oocytes

The procedures for maintenance of Xenopus laevis, preparation of oocytes and injection with mRNA have been previously described [12, 13]. In addition, details concerning the maintenance of Xenopus laevis and the use of oocytes can be obtained from the Xenopus Express website (http://www.xenopus.com/links.htm). 4 1 Expression of Ion Channels in Xenopus Oocytes

A frog colony does not require elaborate equipment, although there are two important considerations. First, amphibians are sensitive to both chlorine and chlor- amine, so the water must be purified to remove both compounds. Second, although frogs can tolerate a wide range of temperature fluctuations, inconsistent temperature, and particularly elevated temperatures above 20 8C, greatly diminish oocyte viability. Surgery to remove oocytes is a relatively simple procedure that can be carried out on a bench top in a clean room. After preparation, the follicle cells are usually removed by treatment with collagenase, although oocytes can be injected and voltage-clamped with intact follicle cells around them. Some electrophysiological responses in oocytes either depend on the presence of follicle cells or occur in the follicle cells, which would be a reason to maintain the cells. However, all procedures are technically more difficult because the follicle cells are harder to pierce and it is time-consuming to separate out individual oocytes, so it is usually best to use defolliculated oocytes. It is possible to obtain oocytes that have been surgically removed and prepared for injection by commercial vendors. The advantages of this approach are that there is no need to maintain a frog colony, an animal use protocol is not required because no vertebrate animals will be used, and there is a considerable saving in time. The disadvantages are that the oocytes are significantly more expensive and they are only available in specific geographical areas, although the number of vendors may increase, depending on demand. One commercial source for oocytes is EcoCyte Bioscience in Germany (http://www.ecocyte.de). For most studies, oocytes are injected with RNA that has been transcribed in vitro from a cDNA clone using either a T7 or T3 promoter. Transcription is easily performed using commercially available kits, although there are some important considerations that affect expression levels. First, the RNA needs to be capped for optimal efficiency, which is part of the procedure in most kits. Second, inclusion of Xenopus b-globin 5' and 3' untranslated mRNA regions and a poly(A) tail at the 3' end usually enhances stability and translatability. Third, the length of the untranslated 5' region can have a dramatic effect on the level of current, with shorter regions generally resulting in greater efficiency of expression. It is also possible to inject DNA into the nucleus of oocytes. In this case, the cDNA should be cloned downstream of a eukaryotic promoter such as the commonly used Cytomegalovirus (CMV) promoter. The advantage of injecting DNA is that there is no need to perform in vitro transcription reactions, which saves both time and money. The disadvantage is that the procedure is more difficult and requires a more sophisticated injection apparatus [12]. Cytoplasmic injection is a relatively rapid and easy procedure. The basic requirements are a dissecting microscope, a micromanipulator, and an injector that can be a simple manual or motor driven dispenser (Fig. 1.1) [12]. Alternatively, oocytes can be injected using an automated device such as the Roboocyte from Multichan- nel Systems, as described in Section 1.4.3.1. This instrument was developed for automated two-electrode voltage-clamping of oocytes, but it can also function for both cytoplasmic and nuclear injection of oocytes. 1.4 Types of Analyses 5

Fig. 1.1 Apparatus for cytoplasmic injection of Xenopus oocytes. The oocytes are distributed on polypropylene mesh in a 35–mm tissue culture dish on the stage of a dissecting microscope. Injection needles are made using a pipette puller to draw out the glass bores that are normally used with the microdispenser. After pull- ing, the needles are broken off at a tip diameter of 20–40 mm, as measured with a reticle under a dissecting microscope. The injection needle is position over individual oocytes using a micromanipulator and the oocytes are injected with up to 100 nl of RNA solution. Using these procedures, it is possible to inject 20 oocytes with one sample in a few minutes.

1.4 Types of Analyses

1.4.1 Electrophysiological Analysis

The most sensitive approach for analyzing ion channel function in Xenopus oocytes is the use of electrophysiology. Essentially all of the standard electrophysiological techniques can be performed on oocytes, including whole cell recording and patch-clamp recording of both macroscopic and single channel currents. In addition, techniques such as cut-open oocyte voltage-clamp recording have been developed specifically for analyzing currents in Xenopus oocytes [10]. The goal of this chapter is to present the general considerations involved in using the various approaches, as detailed procedures have been previously described [14].

1.4.1.1 Two-electrode Whole Cell Voltage-clamp Whole cell voltage-clamping of oocytes involves using two electrodes inserted into the oocyte, rather than using one electrode to make a patch on the surface followed by rupturing the membrane, as is done in mammalian cells. One electrode is used to measure the internal potential of the oocyte and the other electrode is used to inject current (Fig. 1.2). The large size of the oocyte (about 1 mm in diameter and 0.5–1 ml in volume for stage V oocytes) makes this feasible, and is both the major advantage and disadvantage of the system. One advantage is that the procedure is easy to learn and fast to perform. The electrodes are simple to pre- 6 1 Expression of Ion Channels in Xenopus Oocytes

Fig. 1.2 Diagram of the two-electrode voltage- inject current for clamping the oocyte at differ- clamp. The oocyte is placed in a chamber un- ent potentials. The currents can be recorded der a dissecting microscope and two electro- either through the current electrode or sepa- des are gently inserted through the membrane rately through a virtual ground circuit in the using micromanipulators. One electrode is bath. The bath solution can be easily and used to measure the internal potential of the rapidly changed by continuous perfusion from oocyte and the other electrode is used to gravity flow reservoirs.

pare and it is generally possible to obtain records from all oocytes if they are healthy, so that there is very little time lost in preparation. In addition, perfusion of the external medium can be easily changed multiple times. These features make the two-electrode voltage-clamp ideal for screening purposes. A second advantage is that the recordings can be stable over long periods of time, which makes it particularly useful for analyzing properties that require long protocols, such as slow inactivation. A third advantage is that it is possible to insert multiple electrodes and injection needles into the same oocyte. Therefore, modulators of channel function can be injected inside the cell while recording, so that a rapid and direct response to an intracellular signal can be observed. The final advantage of the two-electrode voltage-clamp is that it records currents through channels present in the whole cell, so that it is very sensitive. For example, currents as small as 50 nA can be detected, which corresponds to only 56104 molecules if the single channel current is 1 pA. The two-electrode voltage-clamp can be used to record currents over a wide range of amplitudes from about 10 nA to over 100 mA, depending on the amount of RNA that is injected. However, it is important to adjust the amount of RNA to obtain currents that can be reliably clamped because it is difficult to accurately clamp the membrane potential of the oocyte if the currents are larger than 5 mA. 1.4 Types of Analyses 7

The major disadvantage of recording from the entire oocyte is that the large size and extensive membrane invaginations result in an extremely large membrane capacitance, approximately 150–200 nF. The large capacitance causes a slow clamp settling time when the membrane potential is changed. The capacity transient can be minimized by using electrodes with low impedances of 500 kO or less. This can be accomplished by filling the electrode tips with low-melting temperature agarose, making it possible to have a large tip opening without significant leakage of KCl into the oocyte [15]. However, even with the best electrodes it is difficult to obtain data during the first 1–2 ms of a depolarization, which is the time during which rapidly activating voltage-gated channels such as sodium channels are activated (Fig. 1.3). The large capacity transient is not a problem when recording slow responses or ligand-gated currents that do not require changes in voltage. A second major disadvantage is that there is no control of the internal cellular environment, so that it is difficult to perform quantitative studies, for example examining selective permeability.

1.4.1.2 Cut-open Oocyte Voltage-clamp The cut-open oocyte voltage-clamp was developed to circumvent many of the disadvantages involved in using the two-electrode whole cell voltage-clamp [10, 11]. In this procedure, the oocyte is inserted in a chamber that separates the surface into three regions (Fig. 1.4). The top portion of the oocyte membrane is the region that is clamped and is the section from which currents are actually recorded. The middle portion is a guard that is clamped to the same potential as the top to null leakage currents though the seals. The bottom portion is the region of the oocyte that is “cut-open”, either by permeabilization with saponin or by insertion of a cannula, thus making it possible to perfuse the internal environment and to inject current intracellularly through a low resistance pathway. The internal environment is clamped to ground, as measured by a low resistance electrode inserted through the top of the oocyte, which ensures that the region of the oocyte near the top portion is accurately held at ground. The bath surrounding the top portion of the oocyte is clamped to the command potential, which can be rapidly changed with minimal series resistance. Currents are recorded through low resistance electrodes in the top chamber. There are a number of advantages to the cut-open oocyte voltage-clamp compared to the two-electrode voltage-clamp. First, the capacity transient is minimized so that the clamp can settle in 50 ms, which is fast enough to study activation of even the fastest ion channels (Fig. 1.3). Second, current noise is quite low, approximately 1 nA RMS at 5 kHz bandwidth. Third, it is possible to control the solutions in both external and internal environments. The internal solution can be accurately adjusted with the initial perfusate, but is difficult to change because of the slow perfusion rate, even when using a cannula. In contrast, the external solution can be changed quickly and completely. Finally, the recordings can be stable for hours. These advantages make the clamp particularly well suited for studies involving fast ionic and gating currents. 8 1 Expression of Ion Channels in Xenopus Oocytes

Fig. 1.3 Representative traces of sodium currents using the different recording techniques. The two-electrode voltage-clamp records from channels throughout the entire oocyte membrane, which results in a large capacitive transient (the gap in the current records) so that data cannot be obtained for 1–2 ms after a depolarization. The cut-open oocyte voltage-clamp records from channels in approximately a third of the oocyte membrane. Therefore, the magni- tudes of the currents are comparable, but the time resolution is significantly faster because of the design of the clamp. The macropatch voltage-clamp records from channels in a small patch of oocyte membrane, resulting in fast time resolution but smaller current amplitudes (note the current scale is in pA rather than nA). Small patches can be used to record single channel activity with excellent time resolution. The amplitude of single sodium channels is approximately 1 pA at the potential used for these recordings. All of the data were obtained from oocytes injected with RNA encoding the rat

Nav1.2 sodium channel a subunit alone. The macroscopic current traces are shown for depolarizations from a holding potential of –100 mV to a range of potentials between – 30 and +30 mV in 10 mV increments. The single channel current trace is shown for a depolarization from –100 to –30 mV. The arrows indicate the start of the depolarization.

The major disadvantages of the cut-open oocyte voltage-clamp procedure are that it requires specialized equipment and that it is more difficult to use than the two- electrode voltage-clamp. However, all of the equipment for this procedure, including the voltage-clamp, recording chamber and temperature controller, are commercially available from Dagan Corporation (http://www.dagan.com/ca-1b.htm). 1.4 Types of Analyses 9

Fig. 1.4 Diagram of the cut-open oocyte vol- chamber is used to inject current intracellularly tage-clamp. The oocyte is inserted in a cham- through a low resistance pathway. The internal ber that separates the surface into three re- environment is clamped to ground as meas- gions. The top chamber is clamped to the ured by a low resistance electrode inserted command potential and is the section from through the top of the oocyte. A cannula in- which currents are recorded. The middle serted into the oocyte through the bottom chamber is a guard that is clamped to the chamber makes it possible to perfuse the inter- same potential as the top to null leakage nal environment. currents though the seals. The bottom

1.4.1.3 Macropatch Clamp An alternative method of circumventing the problems caused by the large size of the oocyte is to record from only a fraction of the membrane in an isolated patch [16]. To record macroscopic currents, an electrode with a relatively large opening of about 10 mm in diameter is used to establish a macropatch on the surface membrane (Fig. 1.5). Recording can be performed in either the cell-attached mode or excised inside-out mode. The cell-attached mode is technically easier and main- tains the normal cytoplasmic environment, but the intracellular potential needs to be determined. This potential can be measured by inserting electrodes into the oocyte, which is possible because of its large size. Either a single electrode can be used to measure the potential, or two electrodes can be used to clamp the oocyte at a fixed holding potential. The use of two electrodes has the advantage that channels throughout the remainder of the oocyte are held at the desired potential, minimizing slow inactivation of voltage-gated channels so that the same oocyte can be used for multiple patches. The excised patch technique allows complete control of the potential on both sides of the membrane, but it can be more diffi- 10 1 Expression of Ion Channels in Xenopus Oocytes

Fig. 1.5 Macropatch versus small patch recording. To record macroscopic currents using the patch clamp, an electrode with a large tip diameter of approximately 10 mm is used to make a seal with the oocyte membrane. Currents can be recorded while the electrode is still attached to the intact oocyte or the electrode can be gently pulled back to record currents through an excised portion of the membrane. More sensitive recordings can be obtained using the same approaches but with a smaller diameter electrode tip (approximately 1 mm).

cult to perform because the seal between the membrane and the large diameter electrode is less stable than with a small electrode. An advantage of using an excised patch is that the internal face of the membrane can be perfused with different solutions. There are a number of advantages in using the macropatch technique to record from ion channels in oocytes. First, the capacity transient is minimized because only a small region of the membrane is depolarized, which makes it possible to record fast ionic and gating currents (Fig. 1.3). Second, there is complete control of the solutions on both sides of the membrane if the patch is excised, although only the internal side can be altered with perfusion because the outside is fixed by the electrode composition. These characteristics of the macropatch technique are similar to those of the cut-open oocyte voltage-clamp. An advantage compared to the cut-open oocyte technique is that the equipment is not as specialized, so that a patch clamp amplifier from any supplier can be used. There are some disadvantages in using the macropatch technique compared to the cut-open oocyte voltage-clamp. First, macropatches are usually not as stable as oocytes in the cut-open oocyte voltage-clamp. The decreased stability is a function of the patch and the fact that it is necessary to remove the vitelline membrane from the oocyte to make a seal with the electrode. Oocytes without the vitelline membrane are less stable than intact oocytes and they will lyse if exposed to air. A second disadvantage of macropatch recording is that it usually requires a high level of expression in the oocyte. In this regard, it is possible to use smaller electrodes by increasing the level of expression, which in turn decreases the technical difficulty. Because the macropatch technique involves either cell-attached or excised inside-out patches, there is no access to the external surface of the portion of the membrane being studied. Therefore, macropatch recording is 1.4 Types of Analyses 11 not well suited for studying the interactions of toxins that directly bind to a channel from the external surface. On the other hand, it is an excellent system for studying modulation through second messenger systems, particularly in the cell- attached mode.

1.4.1.4 Single Channel Analysis Xenopus oocytes can also be used for conventional patch clamp recording, including single channel analysis. This approach differs from macropatch recording only in the size of the electrode, with a tip diameter of about 1 mm compared to about 10 mm for macropatches (Fig. 1.5). To make the seal, the vitelline membrane must first be removed from the oocyte, which is accomplished by placing the oocyte in a hypertonic solution (200 mM NaCl). The oocyte shrinks, leaving the vitelline membrane exposed so that it can be manually removed with forceps. Patches can then be obtained in the cell-attached or excised configuration, as in mammalian cells. One disadvantage of the cell-attached mode is that the intracellular potential must be determined. However, an advantage of using oocytes for this purpose is that additional electrodes can be inserted into the large oocyte, making it possible to accurately measure or clamp the intracellular potential. On the other hand, the additional electrodes and voltage-clamp increase the noise, which can be a significant problem for single channel recording. A second disadvantage of cell-attached recording is that oocytes express a high level of endogenous stretch- activated channels [17], and currents through these channels may interfere with the signal of interest. The patches can be excised in either the inside-out or outside-out configuration, just as with mammalian cells (Fig. 1.6). A difference compared to mammalian cells is that it is more difficult to rupture the oocyte membrane compared to a mammalian cell membrane. Positive pressure is generally the most effective technique, particularly with electrodes that have small electrode tip openings. In addition, the oocyte cannot be clamped with a single electrode after the membrane has been ruptured, unlike the situation with mammalian cells. Once the patch is excised, recording is comparable to recording from mammalian cell patches. Single-channel analysis in oocytes is generally equivalent to single channel analysis using mammalian cells, and it therefore has the same advantages and disadvantages. First, there is excellent time resolution because only a small portion of the membrane is depolarized. Therefore, it is possible to obtain detailed information about the opening and closing of individual channel molecules, which provides the most quantitative information for developing gating models (Fig. 1.3). The major disadvantage is that it is technically difficult to get high quality data, especially with small conductance channels. In addition, single channel recording is very time consuming for both acquisition and analysis. There are also some advantages and disadvantages to using oocytes for this purpose compared to mammalian cells. One advantage is that the level of expression can be adjusted by injecting different quantities of RNA. Therefore, it is possible to maximize the probability of obtaining patches that contain single channels. A 12 1 Expression of Ion Channels in Xenopus Oocytes

Fig. 1.6 Inside-out versus outside-out patch leaving the cytoplasmic face of the membrane recording. Patches of oocyte membrane can be exposed to the bath solution. For an outside- excised in either the inside-out or outside-out out patch, the oocyte membrane is first rup- configuration. In both cases, the electrode is tured, after which the electrode is gently pulled first placed against the oocyte membrane to back. The membrane reforms a vesicle at- obtain a tight seal. For an inside-out patch, tached to the electrode with the external face the electrode is then gently pulled back, exposed to the bath solution.

related advantage is that the oocytes can be screened first using a two-electrode voltage-clamp to determine the levels of expression so that only cells with an appropriate level of current are used to obtain patches. On the other hand, a disadvantage of oocytes is that the vitelline membrane must be removed before patch- ing, which adds an extra step and decreases the viability of the oocyte.

1.4.2 Biochemical Analysis

Ion channels expressed in oocytes can be studied using standard biochemical techniques such as immunoprecipitation or binding assays [18, 19]. The major disadvantage of these techniques in oocytes is that they are significantly less sensitive than the electrophysiological approaches. Using the whole cell voltage-clamp, it is possible to detect as few as 105 channel molecules in a single oocyte (less than 10–18 mole). On the other hand, biochemical techniques are generally reliable down to the level of 10–12 mole, although this depends strongly on the specific activity of the reagents being used. Because of this limitation, it is necessary to express the channels at a much higher level for biochemical analysis than is required for electrophysiological recording. 1.4 Types of Analyses 13

Another consideration in using biochemical techniques to analyze ion channel expression in oocytes is that solubilization results in isolation of both cytoplasmic and membrane proteins. Functional ion channels are located in the membrane, but there is generally a large intracellular pool of molecules [20]. An advantage of electrophysiological analysis is that it examines only the functional channels in the membrane, whereas immunoprecipitation of total oocyte proteins examines both membrane-bound and cytoplasmic proteins. This problem can be mitigated by using membrane preparations for solubilization [21], but it is difficult to remove all nonmembrane proteins and those bound to internal membranes. Ligand binding assays do not suffer from this limitation and thus they are well suited for the study of ion channels expressed in oocytes. Ligand binding assays using oocytes are very similar to those carried out using mammalian cells (see Chapter 7). The advantage of using oocytes is that it is possible to easily and quickly express many different channel mutations or variations. In addition, the assay can usually be carried out on a single oocyte so that the composition of channels and subunits is relatively homogenous. The use of a single cell is also the major disadvantage of oocytes, in that it is necessary to obtain a sufficiently high level of expression so that the ligand can be detected. Therefore, the assay requires a high affinity ligand that can be labeled to high specific activity.

1.4.3 Compound Screening

An application of Xenopus oocyte expression that has become more common in the past few years is screening the effectiveness of new pharmaceuticals and those in development. In this regard, the use of oocyte expression cannot really be considered high throughput screening but rather, moderate throughput screening. Even though the limitation on throughput is unlikely to be eliminated anytime soon, there are some advantages to using oocyte expression for this purpose. A major advantage of screening in oocytes is that the assay is electrophysiological, which is the most detailed and relevant response to the compound or drug. In this regard, screening ion channels in oocytes is comparable to screening with one of the automated patch clamp systems that have been developed. Because automated patch clamps are designed for screening cell lines that are usually constructed to be stably expressing the genes of interest, oocytes are better suited for analyzing multiple channel variations such as mutations or different compositions of subunits. For this reason, the oocyte system is particularly appropriate for target identification and optimization. There are also some situations for which the oocyte system has a clear advantage. The first is if the channels do not express well (or at all) in mammalian cells. The second is if the mammalian cells express native currents that interfere with detection of the expressed response. A major disadvantage of using oocytes for screening drugs is that the cell is not the physiological target and the responses may differ from those that occur in vivo. However, this is a criticism of most screening systems because the physiological target cells often cannot be used because they express many different chan- 14 1 Expression of Ion Channels in Xenopus Oocytes

nels and receptors. One difference between oocytes and mammalian cells is that higher drug concentrations are often required for effective block of ion channels in oocytes compared to mammalian cells. This difference probably reflects limited drug access to the oocyte membrane because of the large number of invagina-

tions. On the other hand, although the actual EC50 is usually higher in oocytes, the rank order of different drugs on the same channel or the relative efficacy of the same drug on different channels is likely to reflect the situation with native tissue. Two different automated systems have been developed for electrophysiological screening of ion channels using Xenopus oocytes. The Roboocyte system uses a serial approach in which oocytes are tested sequentially, and the OpusXpress system uses a parallel approach in which eight oocytes are tested simultaneously.

1.4.3.1 Serial Recording Using the Roboocyte The Roboocyte was developed by Multichannel Systems for the automated screening of oocytes in a serial fashion. Information about the Roboocyte can be obtained from the company web site at the following address: http://www.multi- channelsystems.com/products/roboocyte/robointro.htm. The instrument consists of a single head that moves vertically for both injection and recording, with the oocytes located in the chambers of 96 well dishes (Fig. 1.7). The head can be configured with an injection needle or with a recording assembly that contains both voltage and current electrodes and a perfusion needle. The dish sits in a carrier that moves in the X and Y directions to position each oocyte sequentially under the head. The entire instrument is computer-controlled with separate procedures for injection and electrophysiological recording. Perfu- sion can be controlled using a gravity based system that is part of the apparatus and contains either 8 or 16 valves, or the device can be connected to a Gilson liquid handler that can dispense compounds from multiwell plates. The Roboocyte is ideally suited to a situation in which the same sample will be injected into every oocyte, with alterations in the recording conditions or drug application. A major advantage is that injection can be automated after a single set- up configuration. On the other hand, it is not possible to inject different samples without going through the complete set-up procedure again. Oocytes can be obtained already prepared in multiwell plates from EcoCyte Bioscience (http:// www.ecocyte.de), although the availability is limited to specific geographic regions. A unique advantage of the Roboocyte is that it is relatively simple to perform nuclear injection. For this purpose, oocytes are allowed to settle in the 96 well plates, during which time the lighter nucleus rises towards the top surface of the oocyte. Because the oocytes are injected vertically, the instrument can be configured so that the needle pierces the nucleus by increasing the depth of injection. Nuclear injection avoids the time and expense of in vitro transcription. The coding region is cloned following a eukaryotic promoter rather than a T7 or T3 promoter, and the RNA is transcribed in the oocyte nucleus. A disadvantage of nuclear injec- 1.4 Types of Analyses 15

Fig. 1.7 The Roboocyte automated voltage- ized air above a magnetic steel plate. Perfusion clamp from Multichannel Systems. A, The in- can be controlled with a gravity-based system strument consists of a single head that moves containing 16 valves. B, Close-up view of an in- vertically for both injection and recording, with jection needle. C, Close-up view of the record- the oocytes located in the chambers of a 96 ing head, which contains both voltage and cur- well dish that moves on a cushion of pressur- rent electrodes and a perfusion needle. tion is that it is difficult to control the amount of RNA that is synthesized and hence the size of the current that is expressed. In addition, it is impossible to inject fixed ratios of different subunits. The Roboocyte performs automated two-electrode electrophysiological recording with semi-automated on-line and off-line analysis. The recording protocol is the same for each oocyte, and the oocytes are tested for viability before recording so that data are not obtained from dead oocytes. The sampling frequency is up to 2–kHz, which is a lower time resolution than that of the OpusXpress. The perfusion system that is included as part of the Roboocyte is limited to a maximum of 16 samples flowing by gravity. The manifold includes the outlet from all reservoirs so there is no lag time for perfusion, but there is a risk of cross-contamination at the tip. The instrument is designed to interface with a more sophisticated liquid handling system from Gilson, in which case samples can be stored in a variety of wells or tubes and the flow rate is controlled by a peristaltic pump. A disadvantage of this system is that there is a significant lag time for drug delivery to the recording chamber, so that the flow rate must be calibrated to determine when the compound reaches the oocyte. 16 1 Expression of Ion Channels in Xenopus Oocytes

1.4.3.2 Parallel Recording Using the OpusXpress The OpusXpress was developed by Axon Instruments, which is now part of Mole- cular Devices. This instrument is designed for automated analysis of oocytes in a parallel configuration. Information about the OpusXpress can be obtained from the company web site at the following address: http://www.axon.com/cs_Opus Xpress.html The instrument consists of eight individual recording chambers that are configured with perfusion and ground assemblies (Fig. 1.8). Separate voltage and current electrodes are positioned in each chamber for a total of 16 electrodes. Manip- ulation of the electrodes is controlled by 8 separate controllers and recording is carried out using 8 separate voltage-clamp modules. As with the Roboocyte, the entire instrument is computer-controlled. Perfusion is applied from one of two large reservoirs to all of the chambers simultaneously. Individual compounds are applied from 96 well dishes by an automated liquid handling system that uses 8 parallel pipette tips. The OpusXpress is designed purely as a recording instrument with no provision for automated injection, so that injection of the oocytes must be carried out separately. The OpusXpress is particularly well suited for examining oocytes injected with different types of RNA. The parallel design has the potential to increase throughput, as recordings are obtained from 8 oocytes simultaneously. The initial set-up time is longer than for the Roboocyte because 16 electrodes must be prepared and positioned in the holders. However, the electrodes can be reused for a number of days, so that the set-up for additional recordings only involves replacing the

Fig. 1.8 The OpusXpress automated voltage- the voltage electrodes on the left and the cur- clamp from Molecular Devices. A, The instru- rent electrodes on the right. C, Close-up view ment consists of eight individual two-electrode of the 8 voltage-clamp recording chambers, voltage-clamps and a liquid handling system. each of which is equipped with perfusion and B, Close-up view of the automated manipula- virtual ground assemblies. tors that control the 8 pairs of electrodes, with 1.5 Examples of Use 17 oocytes, which takes significantly less time. A major advantage of this approach is that compounds can be applied simultaneously to oocytes expressing different channels. Recordings are obtained using the same series of protocols for all 8 oocytes, with automated operation and real-time analysis. In addition, oocytes are tested for viability so that compounds are not delivered to chambers containing dead oocytes. An advantage of the OpusXpress is that the sampling frequency is up to 30 kHz, which is significantly faster than for the Roboocyte. The instrument can be programmed for a variety of recording protocols and solutions exchanges. The compounds are loaded into the chambers of 96 well dishes and relatively small volumes are required. Another advantage of the OpusXpress is that each compound is delivered via an individual pipette tip, so that there is no delay and no cross-contamination.

1.5 Examples of Use

1.5.1 Characterization of cDNA Clones for a Channel

Originally, the oocyte expression system was especially useful for the isolation of cDNA clones encoding ion channels for which no sequence information was available. With the acquisition of complete genomic sequence information from many species, new ion channels are now usually identified as candidate genes based on their similarity with known family members. However, once the sequence has been determined, it is still necessary to demonstrate that the gene encodes a functional channel. This step is critical because the sequence may not correctly predict the properties of the encoded protein. An example of this situation is the characterization of the BSC1 channel from the German cockroach Blattella germanica [22]. BSC1 was originally identified as the orthologue of the DSC1 channel from Drosophila melanogaster [23], which was in turn identified by its sequence similarity to voltage-gated sodium channels [24]. However, neither DSC1 nor BSC1 had been functionally expressed, so that the assignment of these genes as voltage- gated sodium channels was based purely on sequence similarity. Zhou et al. [22] succeeded in expressing BSC1 in oocytes and demonstrated that it encodes a functional voltage-gated cation channel whose properties differ significantly in a number of ways from those of voltage-gated sodium channels. First, the channels are more selective for barium than for sodium. Second, the kinetics of activation and inactivation are significantly slower than the kinetics of sodium channel gating. Third, the channels deactivate very slowly with a substantial tail current. Finally, sodium currents through the channel can be blocked by low concentrations of calcium, resulting in an anomalous mole fraction effect. All of these properties are more similar to voltage-gated calcium channels than to voltage- gated sodium channels. BSC1 appears to be the prototype of a novel family of in- 18 1 Expression of Ion Channels in Xenopus Oocytes

vertebrate voltage-dependent, cation channels with a close structural and evolu- tionary relationship to voltage-gated sodium and calcium channels. The Xenopus oocyte expression system was both helpful and problematical for the characterization of BSC1. A critical advantage was that the BSC1 channel had only been expressed in oocytes, so that this system was essential to study the channel. Second, currents through BSC1 were too small to measure using isolated patch recording, which made it difficult to compare the permeability of different ions. The cut-open oocyte voltage-clamp made it possible to examine permeability because the ionic composition on both sides of the membrane could be altered while still recording macroscopic currents through most of the cell membrane. The major disadvantage of using oocytes was that they express a robust calcium- activated chloride current that was turned on by calcium entry through BSC1. The chloride current made it significantly more difficult to record calcium current through the slowly gating BSC1 channel, which made it necessary to record barium rather than calcium current.

1.5.2 Structure–Function Correlations

One of the most powerful uses for oocytes in the study of ion channels has been to correlate molecular structure with biochemical and electrophysiological functions. Studies of this type initially involved mutagenesis followed by relatively straightforward analysis using either the two-electrode voltage-clamp or patch clamp. These approaches identified many of the regions involved in activation [25, 26], inactivation [27–30], toxin-binding [31–34] and permeation [35, 36] of voltage- gated sodium and potassium channels. More sophisticated approaches have since been developed that take advantage of the features of oocyte expression to investigate the molecular mechanisms involved. One technique that has been particularly powerful has been the combination of fluorescent microscopy with electrophysiological recording to determine the movement of specific regions of the channel. This approach utilizes the substituted cysteine scanning accessibility method originally developed by Javitch et al. [37], which involves replacing amino acids individually with cysteine and then using cysteine-modifying reagents to determine the accessibility of those residues (see Chapter 2). For the fluorescence measurements, the cysteine residue is labeled with a fluorophore that can be used to detect movement of the specific region of the molecule. While cysteine scanning mutagenesis has been performed using a variety of expression systems, the combination of fluorescence microscopy with electrophysiological recording is a unique capability of the oocyte expression system. These approaches have been developed and used extensively by Isacoff and coworkers and Bezanilla and coworkers. The combination of fluorescence and electrophysiology measurements has revealed a great deal about the movement of the voltage sensor in the potassium channel. Cha et al. [38] used lanthanide-based resonance energy transfer to measure the voltage-dependent distance changes near the S4 subunit of the Shaker 1.5 Examples of Use 19 potassium channel, demonstrating that gating is accompanied by a rotation and possible tilt rather than a large transmembrane movement. Mannuzzu et al. [39] used voltage-clamp fluorometry of oocytes to measure gating rearrangements in the Shaker potassium channel. Their results demonstrated that there are two charge-carrying steps, the first of which takes place independently in each subunit whereas the second involves cooperative interactions between S4 segments. This approach has also been useful for testing structural models of voltage-gated ion channels. For example, Gandhi et al. [40] used accessibility probing and disulfide scanning experiments to demonstrate that the S4 voltage sensor in the bacterial

KvAP potassium channel lies in close apposition to the pore domain in the resting and activated state, in contrast to the predictions of the crystal structure for that channel [41, 42]. Similar approaches have revealed details about the interactions of the four S4 voltage sensors in the sodium channel. Chanda et al. [43] used the cut-open oocyte voltage-clamp to simultaneously record fluorescence signals and gating currents, demonstrating that the voltage-dependent movement of the S4 segment in domain IV is a late step in the activation process after the S4 segments in domains I–III have moved. They further showed that the S4 segment of domain III most likely moves at the most hyperpolarized potentials and that the S4 segments in domains I and II move at more depolarized potentials. Chanda et al. [44] used the same approach to provide direct evidence for coupling interactions between the voltage sensors. Their results indicate that movement of all four voltage sensors is coupled to varying extents, with energetic linkage between the voltage sensors in domains I and IV. The technology has been continually improved in various ways. Sonnleitner et al. [45] used total internal reflection fluorescence microscopy, which allowed them to measure the movement of single voltage-gated Shaker potassium channels rather than the movements of large ensembles of proteins. Asamoah et al. [46] utilized a novel fluorescent probe (Di-1-ANEPIA) to record dynamic changes in the electric field during the gating process of the Shaker potassium channel. Cohen et al. [47] developed a novel fluorescent probe (aminophenoxazone maleimide), which made it possible to track the motions of the side chains to which the probe was attached. These approaches have provided very detailed mechanistic and structural information about the movements of specific regions of ion channels, and they are uniquely suited to expression in Xenopus oocytes.

1.5.3 Studies of Human Disease Mutations

The oocyte expression system has been extensively used to characterize the effects of ion channel mutations that cause human diseases. An example of this use is the study of mutant voltage-gated sodium channels that cause diseases of the musculoskeletal, cardiovascular and nervous system. Mutations in the SCN4A gene encoding the Nav1.4 skeletal muscle sodium channel cause three neuromus- cular diseases, periodic paralysis, paramyotonia congenita and the potassium-ag- 20 1 Expression of Ion Channels in Xenopus Oocytes

gravated myotonias [48, 49]. Mutations in the SCN5A gene encoding the Nav1.5 cardiac channel cause long QT type 3, which predisposes to ventricular tachycardia (torsades de pointes), and Brugada syndrome, which is manifested as ventricular fibrillation [50]. Mutations in any of three neuronal sodium channel genes cause generalized epilepsy with febrile seizures plus (GEFS+). The mutations

have been identified in SCN1A encoding Nav1.1 [51–53], SCN2A encoding Nav1.2 [54] and SCN1B encoding the b1 subunit [55, 56]. These mutations have been analyzed using a variety of different expression systems, each of which has certain advantages and disadvantages. Most studies of the mutations in the skeletal muscle sodium channel have been carried out using transfected HEK or tsA201 cells [57, 58], although some studies have been carried out using oocytes [59]. However, neither expression system is a very good model for skeletal muscle fibers, in which the mutant channels are expressed in vivo.In fact, Cannon et al. [60] used theoretical reconstructions to demonstrate that the integrity of the muscle cell T-tubule system is required to produce myotonia. Simi- larly, mutations in the cardiac sodium channel have been studied using both oocytes [61–63] and mammalian cells [63, 64], with the same reservation that neither system is a good model for cardiac myocytes. Papadatos et al. [65] solved this prob-

lem by constructing mice in which the mouse Scn5a gene encoding the Nav1.5 cardiac sodium channel was disrupted, which made it possible to study both the electrophysiological properties of the ventricular myocytes and the electrocardio- graphic characteristics of the mice. Similar studies have been carried out to analyze the effects of mutations causing GEFS+ using both oocytes [66–69] and mammalian cells [70, 71]. The results in the two different systems are sometimes comparable and sometimes different. For example, using the oocyte expression system, Spampanato et al. [66, 67] demonstrated that R1648H dramatically accelerates recovery from inactivation, W1204R shifts the voltage-dependence of activation and inactivation in the negative direction, and T875M enhances slow inactivation. These results suggest that multiple different alterations in sodium channel function can lead to a similar seizure phenotype. Lossin et al. [70] examined the same three mutations using an HEK cell expression system and obtained different results. They observed a marked increase in persistent current for R1648H and a slight increase in persistent current for T875M and W1204R, and they hypothesized that the epileptic phenotype resulted from the persistent current in all cases. It is not known which alterations reflect the actual effects of the mutations in neuronal cells in vivo. These discrepancies emphasize the necessity to examine the effects of disease- causing mutations in the cell types in which they are normally expressed in vivo. There are certain instances in which the oocyte expression system is particularly well suited for the analysis of a disease-causing mutation. One mutation that causes GEFS+ is D1866Y, which alters an evolutionarily conserved aspartate residue in the C-terminal cytoplasmic domain of the sodium channel a subunit [72]. This mutation decreases modulation of the a subunit by b1, which normally causes a negative shift in the voltage-dependence of inactivation in oocytes. There is less of a shift with the mutant channel, resulting in a 10 mV difference between References 21 the wild-type and mutant channels in the presence of b1. This shift increases the magnitude of the window current, which results in more persistent current during a voltage ramp. Computational analysis suggests that neurons expressing the mutant channels would fire an action potential with a shorter onset delay in response to a threshold current injection, and that they would fire multiple action potentials with a shorter inter-spike interval at a higher input stimulus. The results suggest that the D1866Y mutation weakens the interaction between the a and b1 subunits, demonstrating a novel molecular mechanism leading to seizure susceptibility. The use of oocytes made it possible to quantitatively assess the effects of b1 by injecting different ratios of RNA encoding the a and b1 subunits, which is very difficult to accomplish using a mammalian cell expression system.

1.6 Conclusions

In summary, Xenopus oocytes have been widely used as a heterologous expression system for the study of ion channels. Most channels can be expressed in a variety of different cell types, each of which has its own advantages and disadvantages. Oocytes are particularly well suited for studying many different samples, such as multiple mutations or the effects of different compositions of subunits. In addition, they are excellent for correlating structure with function using a combination of molecular biological and electrophysiological techniques, some of which have been developed specifically for oocytes. Finally, oocytes represent the only heterologous system in which some channels have been expressed. On the other hand, oocytes are not the native cells in which the channels are normally expressed, and this caveat must be remembered when interpreting the results.

Acknowledgments

Work in the author’s laboratory is supported by grants from the NIH (NS48336), the National Multiple Sclerosis Society (RG3405) and The McKnight Endowment Fund for Neuroscience (34653).

References

1 Barnard E. A, Miledi R., Sumikawa K., nels in Xenopus oocytes, Nature 1984, Translation of exogenous messenger 308, 421–424. RNA coding for nicotinic acetylcholine 3 Miledi R., Parker I., Sumikawa K., Pro- receptors produces functional receptors perties of acetylcholine receptors trans- in Xenopus oocytes, Proc. R. Soc. London lated by cat muscle mRNA in Xenopus 1982, 215, 241–246. oocytes, EMBO J. 1982, 1, 1307–1312. 2 Gundersen C. B., Miledi R., Parker I., 4 Methfessel C.,Witzemann V.,Taka- Messenger RNA from human brain in- hashi T., Mishina M., Numa S., Sak- duces drug- and voltage- operated chan- mann B., Patch clamp measurements on 22 1 Expression of Ion Channels in Xenopus Oocytes

Xenopus laevis oocytes: currents through sed from rat brain cDNA, Eur. Biophys. J. endogenous channels and implanted ace- 1987, 14, 131–138. tylcholine receptor and sodium channels, 17 Yang X-C., Sachs F., Characterization of Pflügers Arch. 1986, 407, 577–588. stretch-activated ion channels in Xenopus 5 Thornhill W. B., Levinson S. R., Bio- oocytes, J. Physiol.(London) 1990, 431, synthesis of electroplax sodium channels 103–122. in Electrophorus electrocytes and Xenopus 18 Goldin, A. L., Expression of ion chan- oocytes, Biochemistry 1987, 26, 4381– nels by injection of mRNA into Xenopus 4388. oocytes, Methods Cell Biol. 1991, 36, 487– 6 Thornhill W. B., Levinson S. R., Bio- 509. synthesis of electroplax sodium chan- 19 Shih, T. M., Smith, R. D., Toro, L., Gol- nels, Ann. N. Y. Acad. Sci. 1986, 479, din, A. L., High-level expression and 356–363. detection of ion channels in Xenopus 7 Lübbert H., Hoffman B. J., Snutch oocytes, Methods Enzymol. 1998, 293, T. P.,VanDyke T., Levine A. J., Hartig 529–556. P. R., Lester H. A., Davidson N., cDNA 20 Schmidt J. W., Rossie S., Catterall

cloning of a serotonin 5-HT1C receptor W. A., A large intracellular pool of inac- by electrophysiological assays of mRNA- tive Na channel alpha subunits in develo- injected Xenopus oocytes, Proc. Natl. ping rat brain, Proc. Natl. Acad. Sci. USA Acad. Sci. USA 1987, 84, 4332–4336. 1985, 82, 4847–4851. 8 Julius D., MacDermott A. B., Axel R., 21 Colman A., Translation of eukaryotic Jessell T. M., Molecular characterization messenger RNA in Xenopus oocytes, of a functional cDNA encoding the sero- 1984, 1, 271–302. tonin 1c receptor, Science 1988, 241, 558– 22 Zhou W., Chung I., Liu Z., Goldin 564. A. L., Dong K., A voltage-gated calcium- 9 Weber W-M., Ion currents of Xenopus lae- selective channel encoded by a sodium vis oocytes: state of the art, Biochim. Bio- channel-like gene, Neuron 2004, 42, 101– phys. Acta 1999, 1421, 213–233. 112. 10 Taglialatela M., Toro L., Stefani E., 23 Liu Z., Chung I., Dong K, Alternative Novel voltage clamp to record small, fast splicing of the BSC1 gene generates tis- currents from ion channels expressed in sue-specific isoforms in the German Xenopus oocytes, Biophys. J. 1992, 61, cockroach, Insect Biochem. Mol. Biol. 78–82. 2001, 31, 703–713. 11 Stefani E., Bezanilla F., Cut-open 24 Salkoff L., Butler A.,Wei A., Scavarda oocyte voltage-clamp technique, Methods N., Giffen K., Ifune C., Goodman, R., Enzymol. 1998, 293, 300–318. Mandel G. Genomic organization and 12 Goldin A. L., Maintenance of Xenopus deduced amino acid sequence of a puta- laevis and oocyte injection, Methods Enzy- tive sodium channel gene in Drosophila, mol. 1992, 207, 266–279. Science 1987, 237, 744–748. 13 Goldin A. L., Sumikawa K., Preparation 25 Stühmer W., Conti F., Suzuki H., of RNA for injection into Xenopus oocy- Wang X., Noda M.,Yahagi N., Kubo H., tes, Methods Enzymol. 1992, 207, 279– Numa, S. Structural parts involved in ac- 297. tivation and inactivation of the sodium 14 Stühmer W., Electrophysiologic recor- channel, Nature 1989, 339, 597–603. dings from Xenopus oocytes, Methods 26 Papazian D. M., Timpe L.C., Jan Y.N., Enzymol. 1998, 293, 280–300. Jan L.Y., Alteration of voltage-depen- 15 Schreibmayer W., Lester H.A., Dascal dence of Shaker potassium channel by N.,Voltage clamping of Xenopus laevis mutations in the S4 sequence, Nature oocytes utilizing agarose-cushion electro- 1991, 349, 305–310. des, Pflügers Arch. 1994, 426, 453–458. 27 Hoshi T., Zagotta W. N., Aldrich 16 Stühmer W., Methfessel C., Sakmann R.W., Biophysical and molecular mecha- B., Noda M., Numa S., Patch clamp chanisms of Shaker potassium channel inac- racterization of sodium channels expres- tivation, Science 1990, 250, 533–538. References 23

28 Zagotta W. N., Hoshi T., Aldrich 39 Mannuzzu, L. M., Isacoff, E. Y. Inde- R.W., Restoration of inactivation in mu- pendence and cooperativity in rearrange- tants of Shaker potassium channels by a ments of a potassium channel voltage peptide derived from ShB, Science 1990, sensor revealed by single subunit fluore- 250, 568–571. scence, J. Gen. Physiol. 2000, 115, 257– 29 Patton D. E.,West J. W., Catterall 268. W. A., Goldin, A. L., Amino acid resi- 40 Gandhi, C. S., Clark, E., Loots, E., dues required for fast sodium channel Pralle, A., Isacoff, E. Y., The orientation inactivation. Charge neutralizations and and molecular movement of a K+ chan- deletions in the III-IV linker. Proc. Natl. nel voltage-sensing domain, Neuron Acad. Sci. USA 1992, 89, 10905–10909. 2003, 40, 515–525. 30 West J.W., Patton D. E., Scheuer T., 41 Jiang,Y., Lee, A., Chen, J., Ruta,V., Wang Y., Goldin A. L, Catterall W. A., Cadene, M., Chait, B. T., MacKinnon, A cluster of hydrophobic amino acid resi- R., X-ray structure of a voltage-dependent dues required for fast Na+ channel inacti- K+ channel, Nature 2003, 423, 33–41. vation, Proc. Natl. Acad. Sci. USA 1992, 42 Jiang,Y., Ruta,V., Chen, J., Lee, A., 89, 10910–10914. MacKinnon, R., The principle of gating 31 Yellen G., Jurman M. E., Abramson T., charge movement in a voltage-dependent MacKinnon R., Mutations affecting in- K+ channel, Nature 2003, 423, 42–48. ternal TEA blockade identify the probable 43 Chanda, B., Bezanilla, F., Tracking vol- pore-forming region of a K+ channel, tage-dependent conformational changes Science 1991, 251, 939–942. in skeletal muscle sodium channel du- 32 MacKinnon R., Miller C., Mutant po- ring activation, J. Gen. Physiol. 2002, 120, tassium channels with altered binding of 629–645. charybdotoxin, a pore-blocking inhibitor, 44 Chanda, B., Asamoah, O. K., Science 1989, 245, 1382–1385. Bezanilla, F., Coupling interactions 33 MacKinnon R.,Yellen G., Mutations between voltage sensors of the sodium affecting TEA blockade and ion permea- channel as revealed by site-specific meas- tion in voltage-activated K+ channels, urements, J. Gen. Physiol. 2004, 123, Science 1990, 250, 276–279. 217–230. 34 MacKinnon R., Heginbotham L., 45 Sonnleitner, A., Mannuzzu, L. M., Abramson T., Mapping the receptor site Terakawa, S., Isacoff, E. Y., Structural for a pore-blocking potassium channel rearrangements in single ion channels inhibitor, Neuron 1990, 5, 767–771. detected optically in living cells, Proc. 35 Yool A. J., Schwarz T. L., Alteration of Natl. Acad. Sci. USA 2002, 99, 12759– ionic selectivity of a K+ channel by muta- 12764. tion of the H5 region, Nature 1991, 349, 46 Asamoah, O. K.,Wuskell, J. P., Loew, 700–704. L. M., Bezanilla, F. A., Fluoremetric ap- 36 Hartmann H. A., Kirsch G. E., Drewe proach to local electric field measure- J. A., Taglialatela M., Joho R. H., ments in a voltage-gated ion channel, Brown, A. M., Exchange of conduction Neuron 2003, 37, 85–97. pathways between two related K+ chan- 47 Cohen, B. E., Pralle, A.,Yao, X. J., nels, Science 1991, 251, 942–944. Swaminath, G., Gandhi, C. S., Jan, 37 Javitch, J. A., Fu, D., Chen, J., Karlin, Y. N., Kobilka, B. K., Isacoff, E. Y. et A., Mapping the binding-site crevice of al., A fluorescent probe designed for stu- the dopamine D2 receptor by the substi- dying protein conformational change, tuted-cysteine accessibility method, Neu- Proc. Natl. Acad. Sci. USA 2005, 102, ron 1995, 14, 825–831. 965–970. 38 Cha, A., Snyder, G. E., Selvin, P. R., 48 Cannon, S. C., From mutation to myoto- Bezanilla, F., Atomic scale movement nia in sodium channel disorders. Neuro- of the voltage-sensing region in a potas- musc. Disord. 1997, 7, 241–249. sium channel measured via spectroscopy, 49 Cannon, S. C., Spectrum of sodium Nature 1999, 402, 809–813. channel disturbances in the nondystro- 24 1 Expression of Ion Channels in Xenopus Oocytes

phic myotonias and periodic paralyses, lies with periodic paralysis defines new Kidney Int. 2000, 57, 772–779. aspects of sodium channel slow inactiva- 50 Balser, J. R., Inherited sodium channelo- tion, J.Clin.Invest. 2000, 106, 431–438. pathies: models for acquired arrhyth- 58 Struyk, A. F., Scoggan, K. A., Bulman, mias? Am. J. Physiol. Heart Circ. Physiol. D. E., Cannon, S. C., The human skele- 2002, 282, H1175–H1180. tal muscle Na channel mutation R669H 51 Escayg, A., MacDonald, B. T., Meisler, associated with hypokalemic periodic pa- M. H., Baulac, S., Huberfeld, G., ralysis enhances slow inactivation, J.Neu- An-Gourfinkel, I., Brice, A., LeGuern, rosci. 2000, 20, 8610–8617. E. et al., Mutations of SCN1A, encoding 59 Featherstone, D. E., Fujimoto, E., a neuronal sodium channel, in two fami- Ruben, P. C., A defect in skeletal muscle lies with GEFS+2, Nat. Genet. 2000, 24, sodium channel deactivation exacerbates 343–345. hyperexcitability in human paramyotonia 52 Escayg, A., Heils, A., MacDonald, B. T., congenita, J. Physiol. (London) 1998, 506, Haug, K., Sander,T., Meisler, M. H., 627–638. A novel SCN1A mutation associated with 60 Cannon, S. C., Brown, R. H., Jr., Corey, generalized epilepsy with febrile seizures D. P., Theoretical reconstruction of myo- plus and prevalence of variants in pa- tonia and paralysis caused by incomplete tients with epilepsy, Am. J. Hum. Genet. inactivation of sodium channels, Biophys. 2001, 68, 866–873. J. 1993, 65, 270–288. 53 Wallace, R. H., Scheffer, I. E., 61 Chen, Q., Kirsch, G. E., Zhang, D., Barnett, S., Richards, M., Dibbens, L., Brugada, R., Brugada, J., Brugada, P., Desai, R. R., Lerman-Sadie, T., Lev, D. Potenza, D., Moya, A. et al., Genetic ba- et al., Neuronal sodium-channel a1-sub- sis and molecular mechanism for idiopa- unit mutations in generalized epilepsy thic ventricular fibrillation, Nature 1998, with febrile seizures plus, Am. J. Hum. 392, 293–296. Genet. 2001, 68, 859–865. 62 Rook, M. B., Alshinawi, C. B., Groene- 54 Sugawara, T., Tsurubuchi,Y., wegen,W. A., van Gelder, I. C., van Agarwala, K. L., Ito, M., Fukuma, G., Ginneken, A. C. G., Jongsma, H. J., Mazaki-Miyazaki, E., Nagafuji, H., Mannens, M. M. A. M.,Wilde, A. A. M., Noda, M. et al., A missense mutation of Human SCN5A gene mutations alter car- the Na+ channel aII subunit gene Nav1.2 diac sodium channel kinetics and are as- in a patient with febrile and afebrile sei- sociated with the Brugada syndrome, zures causes channel dysfunction, Proc. Cardiovasc. Res. 1999, 44, 507–517. Natl. Acad. Sci. USA 2001, 98, 6384– 63 Baroudi, G., Carbonneau, E., Pouliot, 6389. V., Chahine, M. SCN5A mutation 55 Wallace, R. H.,Wang, D. W., Singh, R., (T1620M) causing Brugada syndrome ex- Scheffer, I. E., George, A. L., Jr., hibits different phenotypes when expres- Phillips, H. A., Saar, K., Reis, A. et al., sed in Xenopus oocytes and mammalian Febrile seizures and generalized epilepsy cells, FEBS Lett. 2000, 467, 12–16. associated with a mutation in the Na+- 64 Abriel, H., Cabo, C.,Wehrens, X. H. T., channel b1 subunit gene SCN1B, Nat. Rivolta, I., Motoike, H. K., Memmi, M., Genet. 1998, 19, 366–370. Napolitano, C., Priori, S. G. et al., 56 Wallace, R. H., Scheffer, I. E., Novel arrhythmogenic mechanism revea- Parasivam, G., Barnett, S.,Wallace, led by a long-QT syndrome mutation in G. B., Sutherland, G. R., Berkovic, S. the cardiac Na+ channel, Circ. Res. 2001, F., Mulley, J. C., Generalized epilepsy 88, 740–745. with febrile seizures plus: mutation of 65 Papadatos, G. A.,Wallerstein, P. M. R., the sodium channel subunit SCN1B, Head, C. E. G., Ratcliff, R., Brady, Neurology 2002, 58, 1426–1429. P. A., Benndorf, K, Saumarez, R. C., 57 Bendahhou, S., Cummins, T. R., Hahn, Trezise, A. E. O. et al., Slowed conduc- A. F., Langlois, S.,Waxman, S. G., tion and ventricular tachycardia after tar- Ptácek, L. J., A double mutation in fami- geted disruption of the cardiac sodium References 25

channel gene Scn5a, Proc. Natl. Acad. hanced inactivation and acceleration of Sci. USA 2002, 99, 6210–6215. activation of the sodium channel associa- 66 Spampanato, J., Escayg, A., Meisler, ted with epilepsy in man, Eur. J. Neurosci. M. H., Goldin, A. L., Functional effects 2001, 13, 2171–2176. of two voltage-gated sodium channel mu- 70 Lossin, C.,Wang, D. W., Rhodes,T. H., tations that cause generalized epilepsy Vanoye, C. G., George, A. L., Jr., Mole- with febrile seizures plus type 2, J. Neuro- cular basis of an inherited epilepsy, Neu- sci. 2001, 21, 7481–7490. ron 2002, 34, 877–884. 67 Spampanato, J., Escayg, A., Meisler, 71 Lossin, C., Rhodes,T. H., Desai, R. R., M. H., Goldin, A. L., The generalized Vanoye, C. G.,Wang, D., Carniciu, S., epilepsy with febrile seizures plus type 2 Devinsky, O., George, A. L., Jr., mutation W1204R alters voltage-depen- Epilepsy-associated dysfunction in the

dent gating of Nav1.1 sodium channels, voltage-gated neuronal sodium channel Neuroscience 2003, 116, 37–48. SCN1A, J. Neurosci. 2003, 23, 11289– 68 Alekov, A. K., Rahman, M. M., 11295. Mitrovic, N., Lehmann-Horn, F., 72 Spampanato, J., Kearney, J. A., de Lerche, H., A sodium channel mutation Haan, G., McEwen, D. P., Escayg, A., causing epilepsy in man exhibits defects Aradi, I., MacDonald, B. T., Levin, S. I. in fast inactivation and activation in vitro, et al., A novel epilepsy mutation in the J. Physiol.(London) 2000, 529, 533–539. sodium channel SCN1A identifies a cyto- 69 Alekov, A. K., Rahman, M., Mitrovic, plasmic domain for b subunit interac- N., Lehmann-Horn, F., Lerche, H., En- tion, J. Neurosci. 2004, 24, 10022–10034. 27

2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels Louisa Stevens, Andrew J. Powell, and Dennis Wray

2.1 Introduction

Recombinant DNA manipulation and expression techniques have been used to extensively characterise the amino acid residues and domains that influence specific ion channel functions. Structure–function studies have been carried out to investigate all aspects of ion channel function, including channel gating, ligand binding, channel inactivation/desensitisation, voltage-dependence, ion selectivity, multimeric assembly, channel modulation by intracellular pathways or accessory subunits, membrane trafficking and toxin/drug binding sites. Over the last ten years, a plethora of molecular biology techniques and tools have been described to manipulate cDNA sequences in order to generate expression constructs for protein structure–function studies. In this chapter, our aim is to review some of these techniques and to describe how they have been applied to ion channels. Rather than reproduce a basic molecular biology book, we refer readers unfamiliar with recombinant DNA manipulation to molecular biology textbooks (e. g. Watson et al. [1]) and for practical details for specific techniques to Sambrook et al. [2]. A typical ion channel structure–function study will initially involve an analysis of the amino acid sequence of the channel of interest to identify regions that may be involved in the channel property under investigation. For example, studies to identify residues involved in ligand binding may focus on residues within an extracellular domain that are conserved between channel family members that respond to the same agonist or that are not conserved in family members that do not respond to the agonist. Once candidate domains have been identified, the role of these domains can be investigated experimentally through the generation of cDNAs for chimeric channels, where the candidate domain from the agonist-responsive channel is replaced with the corresponding domain from the non-agonist-responsive channel. By functionally characterising the agonist responses of these “domain-swap” chimeric channels in a recombinant expression system the specific regions that are involved in agonist sensitivity can be identified. These studies can then be followed up by using site-directed mutagenesis to introduce single amino acid substitutions of specific candidate residues within that domain

to pinpoint the key determinants involved. For example, characterisation of chimeric GluR5-GluR6 channels provided data supporting a critical role for the extracellular S2 domain in determining sensitivity to specific agonists [3]. Subsequent single amino acid mutagenesis within the S2 domain identified key amino acids involved in conferring sensitivity to receptor ligands such as AMPA, domoate and kainite. The methods that we will describe are those relating to the generation of cDNA clones for chimeras, amino acid insertions, deletions and substitutions, the addition of tags to enable channel characterisation and the generation of channel subunit concatamers to study homomeric or heteromeric channel assembly. Firstly, however, we will introduce the basic molecular biology methods that can be used to transfer full length or partial cDNA clones between different DNA constructs. The techniques described include the use of restriction enzymes, insertion of short linkers, creation of constructs by polymerase chain reaction (PCR) and the overlap extension method, and site-directed mutagenesis. Our aim is to pass on information about techniques which we ourselves have found useful in ion channel research.

2.2 Methods for cDNA Subcloning

The ability to carry out a detailed structure–function study on a particular channel requires the availability of a cDNA clone for that channel, in a plasmid construct appropriate for the expression system to be utilised for the functional characterisation experiments. The specialised techniques for de novo cDNA cloning are not covered here, but instead we describe techniques for sub-cloning cDNAs into alternate plasmid vectors. Historically, restriction enzymes, the basic tools of molecular biology, have been used to transfer cDNA fragments from one vector to another. Recently, however, several new techniques have emerged to expedite cDNA sub-cloning without the use of restriction enzymes [4]. These techniques are based on the use of cDNA amplification by PCR or the use of site-specific DNA recombination, each of which are also covered in this section.

2.2.1 Conventional Sub-cloning Using Restriction Enzymes and DNA Ligase

Restriction endonucleases cleave double stranded DNA at specific short sequences called restriction sites. Digestions of plasmid vectors with appropriate restriction enzymes allows the isolation of the required fragments, i.e. those encompassing the channel cDNA, which can be subsequently re-ligated into an alternate plasmid vector containing the desired regulatory sequences, for example, for recombinant expression (see Chapter 4). Over 500 restriction endonucleases have been isolated from several hundred bacterial strains and made commercially available. These include enzymes that recognise and cleave at 4-base, 6-base and 8-base sequences 2.2 Methods for cDNA Subcloning 29 in double-stranded DNA. Usually 6- and 8-base cutters are used for sub-cloning, since their recognition sequences are less likely to occur frequently in a cDNA sequence or plasmid construct backbone. Many enzymes cut leaving specific 5' or 3' single-strand DNA overhangs (“sticky ends”) while others leave no overhang (“blunt ends”). The advantage of using enzymes with single strand overhangs is that specific ligations can be carried out to enable precise assembly of constructs. On the other hand, blunt-end cutting enzymes are useful because any blunt ended fragment can be ligated with another blunt ended fragment produced by a different enzyme. Unique restriction enzyme recognition sites in channel cDNAs and in plasmid vector DNA sequences can be found using GCG software, other commercially available DNA analysis software packages, or on-line programs. The manufacturers always supply full information about reaction conditions and protocols. Ideally restriction enzymes should not be used at incubation temperatures above 37 8C because of DNA degradation; shopping around with different companies does help as some sell forms of the same enzyme with different incubation temperatures. Once the required restriction digest fragments have been isolated, following separation by agarose gel electrophoresis, in our hands we find it is best not to try to cut with two different enzymes at the same time (even if the manufacturer’s instructions allow it), as the efficiency of the enzymes may be compromised. Before the ligation step, it is best to dephosphorylate the vector (using shrimp alkaline phosphatase or antarctic phosphatase) to prevent self-ligation. The appropriate fragments can then be ligated using T4 DNA ligase and the ligated DNA transformed into E.coli bacteria. E.coli strains such as XL1-Blue (Stratagene) or DH10B (Invitrogen) can be used successfully in many cases; however strains that either decrease plasmid recombination (SURE, Stratagene or STBL2, Invitrogen) or decrease plasmid copy number (ABLE-C or ABLE-K, Stratagene) can be useful to sub-clone channel cDNAs that are difficult to propagate due to high GC content or long repetitive sequences (e.g. NaV or CaV channel cDNAs, see Chapter 4 for further details). Large clones, like calcium channels, do not always ligate easily, and it may be necessary to cut the clone into smaller fragments and then subclone these one at a time. The plasmid DNA can be extracted from individual bacterial colonies, correct clones identified by restriction digest analysis and sequenced to confirm the integrity of the new construct. It is sometimes necessary to insert a very short linker sequence into a clone or a vector. For instance, it may be required to insert a short tag into a sequence so that the resulting protein can be recognised, e.g. by antibodies directed against the tag, such as the FLAG epitope. Another example where a short linker may need to be inserted is in order to alter the multiple cloning site in a vector to enable the easy insertion of a cDNA using different restriction enzymes from those already present in the multiple cloning site of an existing vector. For this technique, two matched oligonucleotides (i.e. a sense oligonucleotide and corresponding anti- sense oligonucleotide) are designed so that when the pair are annealed they correspond to the required linker with the necessary sticky ends for the existing restriction sites. This is best illustrated by an example (Fig. 2.1), where we used this 30 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

Fig. 2.1 Modification of multiple cloning site The annealed linker contains matching (MCS) in a vector by insertion of a synthetic BamHI and HindIII sticky ends, as well as the oligonucleotide linker. The figure shows, on sequences for the new restriction sites to be the left, part of the MCS of a vector, which is introduced (XbaI, EcoRI and NotI). After liga- cut with BamHI and HindIII, removing a tion of the linker into the linear vector (bot- SnaBI site. The linear vector is then dephos- tom of figure), the resulting vector contains phorylated to prevent religation. On the right, XbaI, EcoRI and NotI restriction sites instead two synthetic oligonucleotides are phosphory- of SnaBI. (This figure also appears with the lated with polynucleotide kinase (PNK) and color plates.) annealed to form a double-stranded linker.

method to modify the multiple cloning site in pGEM-HE vector [5]. We needed to replace the SnaBI restriction site (between BamHI and HindIII) with XbaI, EcoRI and NotI sites to produce a new multiple cloning site; this modified vector was then successfully used to express the channel in Xenopus oocytes for voltage- clamp recording (Fig. 2.2) [5]. For this example, matched oligonucleotides were designed containing the XbaI, EcoRI and NotI restriction sites, together with sticky overhangs at each end carefully designed to allow the linker to be ligated into the vector using BamHI and HindIII at either end; the exact details of these oligonucleotides are shown in Fig. 2.1. Synthetic oligonucleotides are not normally synthesised in the phosphorylated form; phosphorylation at the 5' end needs to be carried out to enable subsequent ligation. The annealed double-stranded oligonucleotide linker was then ligated (with T4 ligase) into the vector, which had been similarly cut with BamHI and HindIII. 2.2 Methods for cDNA Subcloning 31

Fig. 2.2 Functional expression of the

CaV3.1 calcium channel in Xenopus oocytes. (A) A sample recording is shown for a voltage step from –80 mV to –30 mV. (B) The current–voltage relation is shown for the mean of 12 cells.

2.2.2 PCR-based cDNA Sub-cloning

It is often necessary to subclone a cDNA construct or section thereof into a vector. However, when no suitable restriction sites are available in the cDNA construct, it is easily possible to create new restriction sites at either end of the construct using primers with overhangs and PCR. For this, primers are designed partly complementary to the construct to be inserted (around 20 bases), but with additional bases as overhangs containing the restriction sites (around 10 bases) (Fig. 2.3). Following amplification, the PCR product is restriction digested with enzymes corresponding to the sites in the overhangs and ligated into the digested vector as described in Section 2.2.1. An advantage of the PCR-based cloning approach is that additional sequences can be introduced, for example, translation initiation signals (Kozak sequence, see Chapter 4.2) or epitope tags to enable channel detection or purification (see Section 2.5 and Chapter 9). The use of thermostable DNA polymerases with proof-reading activity is recommended when using PCR amplification to generate cDNA fragments for sub-cloning, since the resulting cDNA is less likely to contain mutations introduced during the polymerization. Many DNA polymerases exhibit 3' to 5' proof-reading exo- nuclease activity that allows the enzyme to check each nucleotide during DNA synthesis and excise mismatched nucleotides in the 3' to 5' direction. A number of thermostable proof-reading DNA polymerases are commercially available, for  example, Pfu polymerase (Stratagene), Pfx polymerase (Invitrogen) and Vent polymerase (New England BioLabs). 32 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

Fig. 2.3 Creating constructs using PCR. The and XhoI) (note that extra bases are added to figure shows the construction of an N-term- allow the enzymes to cut efficiently). The PCR inal GST fusion protein attached to the N- product was then cut with these enzymes. terminal residues (codons 2–181) of the The pGex-4T-3 vector, which contains the GST Kv2.1 potassium channel. The starting point tag (left-hand side of figure), was cut with the was to PCR amplify a fragment of Kv2.1 same enzymes, dephosphorylated, and li- (right-hand side of figure) using primers with gated with the PCR product to produce the

overhangs containing restriction sites (EcoRI required fusion protein (rKv2.1N2–181).

An alternative PCR-based approach, that removes the requirement for restriction digests and ligation, takes advantage of the fact that Ta q polymerase adds a single 3' adenine base to the end of the resulting DNA fragment [6, 7]. Several companies produce PCR fragment cloning kits that enable ligation into vectors which are provided as linear DNA with single 3'-thymidine overhang (TA cloning kit, Invitrogen, The T-Easy system, Promega). In many cases, the vectors contain the coding sequence for b-galactosidase which is interrupted by successful cloning of the PCR fragment. This gives white (rather than blue) colonies when transformed E.coli are plated onto agar plates with X-Gal, which would turn blue if broken down by b-galactosidase. Thus, clones that contain the vector and inserted PCR fragment are identified as white colonies. Since the TA sub-cloning approach does not determine the direction that the PCR product is inserted, it is necessary to analyse the resulting clones to determine the orientation of the cDNA in relation to the regulatory sequences contained in the vector. The TA vectors contain 2.2 Methods for cDNA Subcloning 33 restriction sites either side of the cloning site, to enable subsequent cDNA subcloning as required. An important consideration when TA subcloning PCR fragments is that Ta q polymerase lacks proof-reading activity and is therefore prone to make errors which can introduce mutations. On the other hand, proof-reading polymerases such as Pfu produce blunt ended PCR products and therefore cannot be directly sub-cloned; instead a common approach is to create the PCR fragment with Pfu, and then subsequently incubate with Ta q polymerase (72 8C for 10 min), which adds adenine overhangs at each blunt end. This product can then be cloned using the TA cloning systems. Alternatively, a system that allows high efficiency cloning of either blunt ended or 3'-A overhang PCR products has also been developed.  The TOPO cloning system (Invitrogen) utilises a novel approach, whereby each end of the linear cloning vector provided is covalently linked to a single topoisomerase I enzyme [8]. Topoisomerase I catalyzes the ligation of the PCR product into the vector, releasing the enzyme from the construct. The ligated DNAs are transformed into E.coli and positive clones identified in a manner similar to the other methods described. A range of TOPO cloning vectors are available to allow the generation of cDNA constructs for a number of purposes. This system has also been adapted to allow directional cDNA cloning, through the introduction of a short specific sequence at the 5'end of the PCR primer used to initiate polymeri-  zation of the sense-strand of the cDNA (Directional TOPO Kit, Invitrogen). As an example of TA subcloning, this technique was used when cloning the human EAG2 potassium channel from a cDNA library [9]. Primers were designed to amplify the channel from a human foetal brain cDNA library using Pfu polymer-  ase, and the PCR product was subcloned into pGEM -T-Easy (Promega) after treatment with Ta q polymerase to yield 3'A overhangs. The insert was then transferred to the Xenopus oocyte expression vector pGEMHE using EcoRI restriction sites, and complementary RNA prepared for injection into oocytes. A clone with the hEAG2 cDNA in the correct orientation was identified, the channel expressed in oocytes and characterised by two-electrode voltage clamp electrophysiology [9]. Figure 2.4 shows examples of currents observed by voltage-clamp in oocytes and the resulting current–voltage relation for the newly cloned hEAG2 potassium channel.

2.2.3 Sub-cloning cDNA through Site-specific Recombination

Sub-cloning using restriction enzymes and ligase or PCR-mediated modification and amplification both require isolation, by agarose gel electrophoresis, and subsequent purification of the cDNA restriction fragment or PCR product. A further complication with this approach is that sub-cloning strategies can become rather complicated as it is often difficult to find suitable restriction sites that are either absent or occur infrequently within the target cDNA. Cloning systems that involve site-specific recombination avoid these problems by removing the need to use restriction enzymes and to isolate and purify DNA fragments, since recombination 34 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

Fig. 2.4 Functional expression of hEAG2 potassium channel in Xenopus oocytes. (A) An example of a current family is shown during a series of depo- larising steps from –80 mV to +70 mV. (B) The current–voltage relation is shown for the mean of 17 cells, normalised to the current at +70 mV.

between complementary DNA sequences can take place directly between the par-  ental circular plasmid DNAs. One such system, the Gateway Technology cloning system (Invitrogen) provides a straightforward mechanism for sub-cloning cDNA inserts between different cloning and expression vectors without the use of restriction enzymes and ligases. The system utilises well-characterised site-specific sequences involved in recombination and crossover in lambda bacteriophage. Lambda recombination occurs between site-specific attachment (att) sites [10–12]. These att sites contain the binding sequences for the recombination proteins, and the recombination reactions are catalyzed by a mixture of enzymes that bind to these specific sequences, bringing the target sites together, cleaving and covalently attaching the DNA. In this system, attP sites recombine with attB sites to give attL and attR sites (and vice versa as the reaction can be driven in either direction), giving a high specificity to the reaction. The experimental steps in the process are as follows. The first step is to create a PCR product of the desired insert/construct using primers with overhangs of attB1 and attB2 at the 5' and 3' ends of the construct (Fig. 2.5). The PCR product (i.e. the construct flanked by attB sites), is then reacted with the “Donor vector” (which already contains attP sites) to produce an “Entry clone” that now contains the construct flanked by attL sites (Fig. 2.5); this reaction is called the BP reaction. After transformation and selection with kanamycin, the Entry clone can now be used to subclone into a whole range of expression vectors. This step involves reaction of the Entry clone with a “Destination vector” (containing attR sites) to produce the required expression clone (with attB sites), as shown in Fig. 2.5, the LR 2.2 Methods for cDNA Subcloning 35

Fig. 2.5 GatewayTM cloning technique. attB alyzed by integrase and integration host factor sites are introduced upstream and down- (IHF), and the product selected with kanamy- stream of the cDNA to be sub-cloned by PCR cin. The cDNA from the Entry clone can then (top left), using primers with attB sequence be integrated into a range of different “Desti- overhangs. The PCR product is then reacted nation vectors”, containing attR sites, in LR with a “Donor vector” (left) which contains ClonaseTM reactions (right), yielding the re- attP sites, to create an “Entry clone” in which quired expression clone. The LR reaction is the cDNA sequence is flanked by attL sites. catalysed by integrase, excisionase and IHF This reaction (“BP ClonaseTM reaction”) is cat- and the product selected with ampicillin. reaction. The expression clone is then ready for use after selection with ampicillin.  Gateway -mediated subcloning allows the straightforward generation of a range of destination vectors that can be used for expression studies. Modification of the sequences between the flanking att-sites and the channel open reading frame within the cDNA insert, to ensure maintenance of the open reading frame through the attB site formed post-recombination, will allow the use of vectors that can introduce either N- or C- terminal tags as required. A range of destination vectors are commercially available that allow expression in bacteria, yeast, baculovirus and mammalian cells. The Gateway system is becoming more widely used; for instance Obrdlik et al. [13] used the system in their study of binding partners for Arabidopsis potassium channels (KAT1, AKT1 and AKT2). The Gateway site-specific recombination approach for cDNA sub-cloning is amenable for robotic automation. Once an Entry vector has been generated with 36 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

the required channel cDNA flanked by attL sites, the processes required for Gate- way mediated sub-cloning into multiple destination vectors are readily automated (the ClonaseTM reactions, bacterial transformation, DNA mini-prep analysis and sequencing), allowing the parallel generation and validation of multiple channel cDNA constructs for multiple purposes [14]. The application of automation is particularly beneficial when the Entry vector is designed such that it allows in-frame recombination into Destination vectors. Using this approach a panel of constructs can be generated to investigate the utility of different N- or C-terminus tags in expression studies, for example, for structural work (see Chapter 9).

2.3 Generation of Chimeric Channel cDNAs

Chimeric cDNA sequences can be generated using several techniques. Restriction enzymes can be used if appropriate sites are fortuitously available within the channel cDNA at positions that allow replacement of the domain of interest. This can provide a relatively straightforward and rapid method to create a range of constructs, including those for deletion and insertion mutants as well as domain- swap chimeras. Often, however, restriction enzyme sites are not conveniently located within the channel cDNA to allow generation of the optimal chimera, deletion or insertion mutants needed to investigate the role of a specific domain in channel function. Thus, the chimeras generated are dictated by the position of the available restriction sites rather than by scientific consideration of the putative domain boundaries within the channel. To overcome this problem several methods have been developed over recent years that can be used to generate “seamless” chimeras. These approaches avoid the use of restriction sites through the use of specifically designed oligonucleotide primers and PCR amplification of the modified cDNA construct. In this section we will also give examples of the utility of various approaches in creating channel chimeras, insertion and deletion mutants.

2.3.1 Use of Restriction Enzymes to Generate Chimeric Channel cDNAs

Where restriction sites are conveniently available at the boundary of the domain to be replaced, these can be used to generate the chimeric channel cDNA, provided that ligation of the chosen restriction sites retains the open reading frame to produce the chimeric protein. An example of the use of restriction sites to create a chimeric channel is shown in Fig. 2.6. Restriction enzymes were used to create chimeras between rat and human forms of the voltage-gated potassium channel Kv2.1 [15]. This was done in order to locate the molecular domains that underlie the differences in activation kinetics between the two versions of the channel. As shown in Fig. 2.6, both rat and human Kv2.1 were digested with the enzymes BssHII and Bsp1407I (TypeII restriction endonucleases). The appropriate DNA fragments were isolated by purifi- 2.3 Generation of Chimeric Channel cDNAs 37

Fig. 2.6 Example of the use of restriction enzymes to create a chimera. A chimera was constructed between rat and human forms of the Kv2.1 potassium channel as shown. Restriction enzymes BssHII and Bsp1407I were used to cut the rat (gray) and human (striped) forms of the channel in their respective vectors. The required DNA fragments were isolated and ligated to form the chimeric cDNA with residues 108 to 528 of the rat channel replaced by human

(rKv2.1h108–528). 38 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

Fig. 2.7 Activation times of Kv2.1 channels. Activation times (time for current to rise from 10% to 90% of maximum current,

t10–90%) versus test potential for: (A) chimera rKv2.1h108–528 (residues 108–528 of rat replaced by human), (B) chimera rKv2.1h741– 853 (residues 741–853 of rat replaced by human). Human wild type (&, n = 4), rat wild type (*, n = 10), and chimera ~, n = 7).

cation from agarose gels and ligated in order to create the required chimeric channel with residues 108–528 of the rat channel replaced by corresponding residues

from human channel (rKv2.1h108–528) [15]. Characterisation of the chimera demonstrated that it retained the fast activation kinetics typical of the rat wild type channel [15] (Fig. 2.7A), suggesting that residues 108–528 do not contribute to the differing activation kinetics between rat and human Kv2.1. Another chimera with residues 741–853 of rat Kv2.1 replaced

by human Kv2.1 (rKv2.1h741–853) was created using the same technique (restriction enzymes BsmI and NotI). In this case, by contrast, the activation kinetics were similar to human Kv2.1 (Fig. 2.7B), showing that residues between amino acids 2.3 Generation of Chimeric Channel cDNAs 39

741–853 contribute to determining the rate of activation for the two potassium channels.

2.3.2 PCR-mediated Overlap Extension for Chimera Generation

Overlap extension PCR can be used to create chimeric, insertion or deletion mutants in specific regions by “seamlessly” fusing the required cDNA fragments [16, 17]. This powerful method is very useful for the precise creation of chimeric cDNAs when useful restriction sites are not available for the restriction-ligation method. The technique involves PCR amplification of fragments that encompass the required domains from each of the “parent” cDNAs. Assembly of these PCR fragments is achieved via combining the isolated fragments in a further PCR amplification; correct assembly of the fragments is enabled by the presence of complementary sequences (overhangs) incorporated within the primers used for the initial PCR. The example shown in Fig. 2.8 demonstrates the use of this method to construct a cDNA with a deletion of the C-terminus of the rat Kv2.1 potassium channel; in this example, residues 412 to 853 were deleted [18]. The initial step was to produce two separate PCR products for each part of the required construct, using primers with complementary DNA sequence at the join (i. e. just after codon 412 in one fragment, and just before the stop codon in the other fragment). In the second step, the two PCR products were then combined in a further PCR reaction. Joining of the two fragments is made possible by the overlaps which were introduced via the overhangs in the previous step, each fragment constituting a ‘mega- primer‘. The resulting DNA, constituting the fusion of the two initial PCR products, was then further amplified through polymerization initiated from two primers (1 and 4 in Fig. 2.8) complementary to the extreme ends of the fusion product. The final step involved cutting the PCR product at any nearby restriction sites (at sites outside the region being swapped, ApaI and NotI in this example) (Fig. 2.8), followed by ligation (with T4 DNA ligase) into the original clone, which was also cut with the same enzymes. This cDNA construct was used to express the deletion in COS-7 cells, followed by purification of the protein and electron microscope single particle studies [18]. Stepwise PCR amplification and overlap extension of two, three or more cDNA fragments can be carried out to create chimeras with specific fusion boundaries [19]. Thus, the technique can also be used to substitute domains within internal regions of a protein. A second example (Fig. 2.9) shows an extension of the method to construct a chimera between rat and human forms of the Kv2.1 potassium channel [15]. In this example amino acids 741 to 795 within the C-terminal activation domain of rat Kv2.1 were replaced by the corresponding residues in the human channel, therefore overlap extension of three cDNA sequences was required. As before, the initial step was to create PCR products with complementary sequence at the joins (Fig. 2.9). In the second and third steps, the PCR products were joined together by further PCR reactions, again facilitated by the comple- 40 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

Fig. 2.8 Example of a deletion by the overlap Primers are designed with overhangs as extension PCR method. In this example, the shown. The two PCR products were used in a overlap extension method was used to make further PCR reaction (“second round PCR”) to a deletion of the C-terminal domain (codons produce a fragment containing restriction 412–853) of the rat Kv2.1 potassium channel. sites (ApaI and NotI). After cutting the frag- The first step (“first round PCR”) was to PCR ment with these enzymes, the final product amplify part of the Kv2.1 channel that was to was ligated into the similarly cut wild type be included (up to codon 411, in black) along channel cDNA vector to generate the trun- with a fragment of the pMT3 vector (in gray) cated Kv2.1 channel cDNA. using a primer (3) to introduce a stop codon. 2.3 Generation of Chimeric Channel cDNAs 41

Fig. 2.9 Example of a chimera made by the be replaced to form the chimera. Primers are overlap extension PCR method. The example again used with overhangs as shown. The sec- shows the construction of a chimera between ond round PCR joins two of the fragments, rat and human Kv2.1 potassium channels and the third fragment is added in the third with codons 741–795 of the rat channel re- round PCR. The final PCR product is digested placed by corresponding residues of the hu- with restriction enzymes (Bsp1407I and NotI) man channel. For the first round PCR, pro- and sub-cloned into the similarly cut wild type ducts were made for rat Kv2.1 sections (black) rat Kv2.1 cDNA construct. and the human Kv2.1 section (gray) that is to 42 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

mentary sequences introduced by the primers overlying the joins in the initial PCRs (Fig. 2.9). Digestion of the final PCR product is required, followed by ligation into the similarly digested wild type rat Kv2.1 clone. This chimera, with residues 741–795 of the rat Kv2.1 channel replaced by corresponding residues in the

human channel (rKv2.1h741–795), was used in expression studies in oocytes to determine the effect of these residues on the activation kinetics of the channel [15]. The results shown in Fig. 2.10 show that these residues alone did not affect the activation kinetics of the rat channel, as the chimeric channel retained fast activation kinetics similar to wild type rat Kv2.1.

Fig. 2.10 Examples of data obtained using constructs made by overlap extension. Mean activation times

(t10–90%) are shown for chimera rKv2.1h741–795 (residues 741–795 of rat replaced by human). Human wild type (&, n = 5), rat wild type (*, n = 4), and chimera (~, n = 6).

Successful use of the overlap extension technique relies on careful primer design to ensure that the correct chimeric product will be expressed from the cDNA. It is advisable first to assemble the DNA sequence for the required construct in si- lico (using one of the available DNA sequence analysis programs) and to translate the sequence to make sure the new reading frame expresses the expected chimeric channel protein. For each PCR fragment, the overhangs in the primers should be 10 bases and the template-specific sequence for the fragment about 20 bases. Where possible, primers should be designed such that they have the same calculated melting temperature (for the template-specific part of the primer); 60 8C is optimal to allow the annealing temperature in the PCR reaction to be set at 55 8C. Large fragments (<3 kb) do not join very well to small fragments; it is worth trying to keep the length of the PCR fragments similar to help the stability of the PCR when it is annealed. It is possible to carry out the sequential PCR steps to produce the final PCR product in one reaction containing all the primers and cDNA template mix for all of the fragments to be fused. However, we have found that better yields are obtained if the PCR reactions are carried out in separate steps and the PCR products isolated on an agarose gel and checked for size prior to the next step. Increasing the amounts of DNA template or polymerase can be tried if the overlap extension is not initially successful. In addition, adding DMSO at a final concentration of 5% v/v may also help, particularly when GC-rich cDNA se-  quences are being amplified. Proof reading polymerases (e. g. Pfu, Pfx or Vent ) 2.4 Site-directed Mutagenesis 43 should again be used to minimize the possibility of introducing unwanted sequence errors during the PCR amplification and, to ensure errors have not been introduced, the entire sequence of the resulting chimeric cDNA insert should be confirmed by sequencing. For the final step, restriction enzymes should be chosen so that at least 500 bases are removed, making it easy to detect correct cutting of the final fragment when run on a gel.

2.3.3 PCR-mediated Integration or Replacement of cDNA Fragments

Another powerful PCR-based approach applicable for the generation of channel chimeras has recently been described by Geiser et al. [20]. This technique is an adaptation of the Stratagene QuikChange site-directed mutagenesis method [21, 22]) which is covered in the next section, and so we will return again to a discussion of this useful approach after the Quik-Change method has been described.

2.4 Site-directed Mutagenesis

Once a specific ion channel domain that influences a particular channel function has been identified through the characterisation of chimeric channel mutants, the role of individual amino acids within the domain can be characterised through the introduction of specific small (single, double or more) amino acid deletions, insertions or substitutions (“site-directed mutagenesis”). Generally, amino acid substitutions are preferred, since the role of specific amino acid side chains can be investigated while avoiding gross changes to the polypeptide backbone. The influence of charged, neutral or bulky amino acids in mediating a specific channel function can be determined through replacement with residues with either the opposite or intermediate characteristics. The most common technique for site-directed mutagenesis, commercialised by Stratagene (QuikChangeTM Mutagenesis), can be used quickly and easily to introduce mutations into a cDNA construct, and is illustrated in Fig. 2.11. To generate the mutation, this technique uses two primers that are each complementary to the opposite strands of the plasmid DNA template but contain the appropriate base substitutions to give rise to the desired mutation. After separation of the double stranded circular DNA at high temperature, primers are annealed at the target site at lower temperature and the template extended with Pfu polymerase, producing the required mutant strands. The procedure is repeated in a thermocycler some 25 times. It is worth noting that, although PCR-like processing is used, this is not strictly a PCR reaction because exponential amplification cannot occur. Next, the DNA products are treated with the enzyme DpnI, an endonuclease which cleaves dam-methylated DNA. Only DNA that has been previously isolated from E. coli will be dam-methylated and so the original parental (i.e. wild type) template will be digested, while the synthesised mutant product will remain. The mutant is 44 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

then transformed into E. coli (which repairs a nick in the mutant strand, Fig. 2.11), and the mutated plasmid DNA extracted from several individual colonies and sequenced to confirm that the mutagenesis has been successful. As usual, primer design is crucial for successful results. The desired mutation should be in the middle of the primers to aid accurate annealing; primers should be between 25 and 45 bases long, and the theoretical annealing temperature should be greater than or equal to 78 8C, and they should have a minimal GC content of

Fig. 2.11 Site directed mutagenesis. Primers containing the required mutant sequence (in gray) are annealed to the cDNA construct (in black) and extended by thermocycling using Pfu DNA polymerase. The enzyme DpnI is added to digest the original dam-methylated plasmid template leaving the required mutant construct. The resulting mutated DNA is transformed into E. Coli (which repairs the nicks shown), and clones verified by sequencing. 2.4 Site-directed Mutagenesis 45

40%. A control reaction should be carried out alongside the mutant, with the control mix omitting the polymerase. Upon transformation of control and mutant reaction products, colonies should be present for mutant but not for control (since the DpnI reaction should have removed the control wild type clones). In order to decrease the risk of unwanted mutations being introduced by the polymerase, it can be beneficial to sub-clone a short section of cDNA containing the site-directed mutation(s) back into the original wild type clone using appropriate restriction sites close to the mutations. In any case, sequence validation of the resulting cDNA clone is essential.

2.4.1 Examples of the Use of Site-directed Mutagenesis

The technique of site-directed mutagenesis has been widely used. One example is in the creation of silent mutations (i. e. mutations that alter bases but not the corresponding amino acid). Introduction of silent mutations can be useful in order to produce restriction sites that make subsequent sub-cloning easier. This site- directed mutagenesis technique was used to introduce Bsu36I and BsmI sites into the human Kv2.1 channel [15]. The rat form of this channel already contained these restriction sites at homologous positions, and so introduction of the same sites enabled us to create the required chimeras between rat and human channels using these restriction enzymes. There are many examples in the literature describing the construction of site-specific mutations to meet the requirements of each individual study [e.g. 23–26]. For example, studies on Kir channels have focused on the role of native cysteine residues on channel function and membrane trafficking [25, 26]. Studies on P2X receptors to investigate the extracellular domain structures involved in ATP-binding, channel gating and modulation have utilised site-directed mutagenesis in order to mutate conserved positive residues [27–29], aromatic residues [28, 30], cysteines (to investigate disulfide bonding, [31, 32]), histidines (to investigate Zn and pH modulation, [33]) and proline residues (to investigate secondary structure, [34]). In contrast to studies where specific residues are substituted with others on the basis of specific expectations within the study, some mutagenesis approaches for investigation of protein structure and function involve the systematic replacement of residues in the domain of the channel under study by specific amino acids, followed by characterisation of the mutant’s functions. One such approach is alanine-scanning mutagenesis, where the sequential residues within the domain under investigation are each replaced by an alanine residue to generate a panel of mutant cDNAs for functional characterisation. Other frequently used scanning mutagenesis approaches involve replacement of single amino acid residues with cysteine, lysine, proline, tryptophan or histidine. Each approach enables determination of the role of specific amino acid properties in the function of the channel domain being studied. For example, mutation to alanine removes the amino acid side chain and substitutes a small hydrophobic group, which would be expected to interact favorably with membrane lipids or hydrophobic regions [35]. When ala- 46 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

nine scanning was carried out on S1, S2, S3 and S4 regions of a potassium channel, periodic effects were observed on channel gating, consistent with a-helical structures of these regions and the clustering of gating effects on faces of the helices, leading to a model for the packing of these helices [36]. The technique has also been widely used to determine toxin and drug binding sites, either through mutagenesis of the channel to determine the key determinants of a channel binding site, such as that used to define the drug binding site on the hERG channel [37], or through mutagenesis of a toxin to identify the epitope on the toxin involved in binding, such as that carried out to identify key residues on o-atraco- toxin-Hv1a required for selectivity against calcium channels [38]. Alanine-scan-

ning mutagenesis of the rat brain NaV1.2 channel identified specic amino acid residues in the IVS6 transmembrane domain involved in binding the local anes- thetic, etidocaine, as several alanine substitutions significantly reduced voltage- and frequency-dependent block [39] by the compound. Cysteine-scanning mutagenesis is particularly beneficial in studies of structure–function relationships in membrane proteins. The introduced cysteines can be covalently modified by charged, hydrophilic, sulfhydryl reagents to probe the accessibility of the substituted cysteines in transmembrane domains. The effects of the covalently bound reagent on channel function can then be characterised. This approach is termed the substituted cysteine accessibility method (SCAM, [40]) and has been used to study the dynamics of channel function in potassium channels [41, 42], calcium channels. [43], nicotinic acetylcholine receptors [44–46], P2X receptors [47–49] and GABA-A receptors [50, 51], amongst others. Typical sulfhydryl reagents used for these studies are parachloromercuribenzensulfonate (PCMBS) or derivatives of methane thiosulfonate (MTS), either positively charged methane thiosulfonate ethylammonium (MTSEA) and methane thiosulfonate- ethyltrimethylammonium (MTSET), negatively charged methane thiosulfonate- ethylsulfonate (MTSES) or neutrally charged methyl methane thiosulfonate (MTSM). In some studies, cysteine binding reagents that are not membrane permeable are applied in the extracellular solution, and the effects of the reagent will only be noted if the introduced cysteine residue is exposed and accessible to the extracellular solution. These cysteine accessibility studies have proved extremely useful in investigating ion channel structure and function, since they provide a method whereby dynamic changes in structure during channel activation or inactivation can be analysed. For example, the approach was used to demonstrate that the S4 transmembrane region of the prototypical Shaker voltage-gated potassium channel moves outwards upon membrane depolarisation, confirming the role of this region as the voltage sensor [52, 53]. Cysteine substitutions were introduced into the S4 domain, the mutants expressed in Xenopus oocytes and their sensitivity to extracellularly applied sulfhydryl reagents characterised by two-electrode voltage-clamp electrophysiology. Upon membrane depolarisation, movement of the S4 domain during channel gating exposes introduced cysteines (that were inaccessible prior to depolarisation) to the sulfhydryl reagent added in the external solution. Covalent modification of the exposed cysteine residues was evident through the appearance 2.4 Site-directed Mutagenesis 47 of cumulative current block upon repetitive activation of the channel. The data demonstrated that, upon depolarisation, the S4 transmembrane segment undergoes a conformational change that leads to channel opening. Similar results have been obtained for calcium channels using cysteine-scanning mutagenesis techniques [43]. Cysteine mutants were generated using Quik- ChangeTM site-directed mutagenesis at positions V263, A265, L266, A268, F269, and V271 within the S4 segment of domain I in a chimeric CaV1.2/CaV3.1 calcium channel. The cysteine-binding reagent PCMBS was applied extracellularly to each of the mutants expressed in oocytes and depolarizing pulses applied. As can be seen in Fig. 2.12, the introduced cysteines in mutants V263C, A265C, L266C, and A268C reacted with the PCMBS to give a reduction in the recorded current, with mutants F269C and V271C remaining unaffected. This shows that, upon depolarisation, movement of the S4 transmembrane domain S4 exposes residues 263–268 to the extracellular solution, while residues 269 and 271 remain buried in the membrane. Further experiments demonstrated that movement of S4 is voltage dependent because PCMBS gave a voltage-dependent block of CaV3.1 current, indicating outward movement of the S4 region upon depolarization. The differing charge on various cysteine binding reagents can be exploited in order to probe the influence of side chain charge on channel properties. For example, substituted cysteine accessibility mutagenesis identified a residue within the P2X2 ATP-gated channel (I67) that gives normal function when the mutant channel was exposed to neutral or positive methanethiosulfonates while exposure to negative MTSES decreased agonist potency, demonstrating that the maintenance of positive charge close to I67 (residues K69 and K71) is required for ATP binding and channel activation [29], and introduction of negative charge at residue 67 dis- rupts this. On the other hand, substituted cysteines that are accessible to modification by a neutral methanethiosulfonate were identified at positions within the first and second transmembrane domains of this channel; introduction of positive and negative charges at these positions using MTSETor MTSES has allowed investigation of the ion permeation pathway and the aqueous environment around the intracellular and extracellular ends of the transmembrane domains [29, 47, 48, 54]. Cysteine scanning mutagenesis has also been used without cysteine modifying reagents to investigate the proximity of residues through disulfide formation between engineered cysteines. This approach was used to address the subunit arrangement of homomeric P2X2 and P2X4 receptors and heteromeric P2X2/3 receptors [55]. Engineered cysteines in the outer ends of the first transmembrane domain of one subunit and the second transmembrane domain of the second subunit demonstrated the proximity of these domains in the heteromeric channel through their ability to form a disulfide bond. Using specific combinations of cysteine-substituted P2X2 and P2X3 mutants and measuring the DTT-sensitive reduction of channel function (caused by the formation of disulfide bonds between the cysteines) demonstrated that the heteromeric trimer formed by P2X2 and P2X3 receptor subunits consists of two P2X3 subunits and one P2X2 subunit. An alternative method of determining the accessibility of mutated residues is to carry out histidine scanning mutagenesis. Substituted histidine mutants are used 48 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels 2.4 Site-directed Mutagenesis 49

in combination with acidic buffers to investigate accessibility to hydrogen ion of mutated residues [56]. Changes in the solvent accessibility of the introduced histidines that accompany conformation changes in the channel voltage sensor can be detected as pH-dependent changes in the gating currents, due to the protonation of the exposed histidine. The advantages of histidine scanning mutagenesis are that protons are accessible to more confined spaces within the protein (compared to the more bulky sulfhydryl reagents), protonation rates are faster than channel gating rate constants so labelling is effectively instantaneous, and replacement of basic residues (such as the lysines and arginines in the S4 voltage sensor of voltage-gated channels) with histidine is less disruptive than neutralisation by cysteine because the native charge is maintained when protonated. Scanning mutations can be made instead to introduce tryptophan [57] or lysine [58], which have more bulky side chains and so, in principle, allow the detection of functional effects of residues that may be missed by the milder mutations introduced by alanine scanning. Also, tryptophan has a hydrophobic side chain while lysine has a hydrophillic one, so these residues can be used to discriminate between the requirement for hydrophobic or hydrophilic surfaces within a channel domain. For tryptophan scanning mutagenesis it is assumed that side chains involved in protein packing would be sensitive to substitution by tryptophan, whereas those exposed to the lipid membrane would be relatively tolerant to this mutation. In studies on the transmembrane domains of voltage-gated potassium channels [59–62] and ligand-gated channels [57, 63–66], this approach has yielded a pattern whereby every third or fourth residue exhibits sensitivity to an introduced tryptophan – suggesting an a-helical structure with a significant lipid-exposed surface. Hydrophillic lysine mutations can be used to determine the orientation of residue side chains towards core hydrophobic or surface hydrophillic environments. These mutations disrupt subunit assembly at overlapping hydrophobic regions and so may be useful in identifying such regions of overlap between subunits [67]. Lysine-scanning mutagenesis has also been used to characterise hydrophillic surfaces involved in channel–toxin [58] interactions. Proline scanning, though not widely used, has also given some interesting results for ion channels. This residue introduces a kink in the peptide chain, and studies using proline scanning have identified channel gating regions, for example, in an inward rectifier potassium channel [68].

3 Fig. 2.12 Cysteine scanning mutagenesis of the S4 segment of domain I of a calcium channel. (A) The effect of application of PCMBS (100 µM, indicated by the bars) on normalised currents for cysteine mutants V263C (n = 3), A265C (n = 4), L266C (n = 3), A268C (n = 3), F269C (n = 4), and V271C (n = 3) during repetitive depolarisations to +10 mV from a holding potential of –80 mV. (B) Membrane topology of domain I showing the positions of the residues mutated to cysteine (shaded). Movement outwards during depolarization is as far as A268. 50 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

2.4.2 Modification of the QuikChange Method for the Replacement of cDNA Fragments

The QuikChange technique has been adapted to allow the construction of chimeras [20]. The approach involves two main steps. Firstly, a PCR product must be generated for the region to be integrated into a channel cDNA sequence. For this, primers need to be designed with overhangs at each end of the region to be inserted, such that the overhangs are complementary to the sequences at the intended sites of integration to the channel cDNA. Secondly, following isolation and purification of the PCR product, it is then mixed with the channel cDNA in its plasmid vector, and then thermocycling with Pfu is applied as for the Quikchange method described above. During this process, the primers anneal to the channel cDNA at the chosen insertion sites, determined by the PCR overhangs, and extension occurs from the 5' end of the product overhang along the circular channel cDNA plasmid until polymerization reaches the 3' end of the PCR insert, so integrating the PCR insert into the channel cDNA plasmid. As in the QuikChange method, the methylated parent plasmid is then digested with DpnI, and the reaction product is transformed into bacteria. Colonies containing the chimeric cDNA plasmid with the insertion or replacement are identified through restriction digest or PCR analysis, and the chimeric cDNA confirmed by sequencing. This approach was used to generate P2X receptor chimeras in order to define the determinants of differences in ATP analog sensitivity between the P2X2 and P2X4 receptor subtypes [69]. Functional chimeras were generated in which the extracellular domain of the P2X2 subtype was replaced with the P2X4 extracellular domain. Characterisation of the ligand potencies and receptor desensitisation rates of these chimeras defined the regions of the channels that determine ATP analogue sensitivity.

2.5 Epitope-tagged Channels and Fusion Partners

Structure–function studies of ion channels are often facilitated through the use of tagged ion channel protein or fusion of the channel to a reporter protein. The use of tags or fusions can be applied to monitor channel expression, to localise the channel at the subcellular level, to purify the channel (see Chapter 9) and for analysis of channel topology, dynamics and interactions. Tagging involves the addition of a short peptide sequence to the wild type ion channel polypeptide, which can be used along with specific antibodies or affinity reagents to detect or purify the channel [70]. Examples of frequently used affinity tags are a 10 amino acid fragment of the c-myc protein, an 8 amino acid peptide based on the enterokinase cleavage termed FLAG, a 9 amino acid peptide from the haemaglutinin HA1 protein of human influenza virus or a hexa-His polypeptide. Such peptide tags are usually introduced either at the N- or C-termini of the channel where they are less likely to interfere with the function of the channel; however in some cases, they 2.5 Epitope-tagged Channels and Fusion Partners 51 are placed within a specific intrapolypeptide domain, such as extracellular or intracellular loops to allow detection of specific channel functions. For example a HA tag was placed in the extracellular S1-S2 loop of the hERG potassium channel in order to detect channel trafficking to the plasma membrane [71]. Channel fusions to larger reporter polypeptides, such as green fluorescent protein, have also been used to analyse sub-cellular localisation of ion channels (see for example Refs. [72–74]). There are several approaches that can be used to generate cDNA that introduce short tags into a channel protein. One technique uses two matched oligonucleotides in a similar manner to the approach used to introduce new restriction digest sites described in Section 2.2.1. The oligonucleotide pair contains the coding sequence for the required tag flanked by a unique restriction site that can be used to insert the short sequence into the open reading frame of the channel cDNA at the required position (in this case the oligonucleotide pair is flanked by the same restriction site overhangs at either end). However, this method is generally limited due to the requirement for restriction sites at appropriate positions within the cDNA. In these cases, tags can be incorporated using PCR-based approaches. Another approach involves the introduction of a tag by placing the channel cDNA into a vector which already contains the necessary tag. The subcloning can be carried out using the available restriction sites in the vector together with matching restriction sites at either side of the channel cDNA insert. For this, the required restriction sites can be introduced into the channel cDNA by the PCR method described in Section 2.2.2, using PCR with primer overhangs containing the restriction sites. Careful primer design is required to ensure that the product is in the correct frame (translation of the required fusion/tagged protein from the resulting cDNA sequence is essential). For instance, this technique was used [15] to sub-clone the N-terminal region (residues 2–181) of a KV2.1 potassium channel into a pGex vector (Amersham Biosciences) which contains a glutathione-S transferase (GST) sequence, such that expression leads to a GST-tagged fusion protein.

The GST-N-terminal KV2.1 fusion protein was purified utilising the GSTag and used in studies to show binding with the radiolabelled C-terminal region of the

KV2.1 channel [15]. The PCR primers were designed such that only the N-terminus-encoding fragment of the cDNA was amplified and the 3'primer introduced an in-frame stop codon (Fig. 2.3). The 5'primer introduced a flanking EcoRI restriction site, while the 3'primer introduced a XhoI site, to allow in-frame subcloning of the PCR product into the similarly digested pGex vector. The overlap extension technique described in Section 2.3.2 can also be used to create tagged proteins, for instance to create cDNA constructs to express ion channels fused with GFP or its variants, YFP (yellow) or CFP (cyan). Here, the tag or fusion partner cDNA and the channel cDNA are first amplified by PCR using primers containing overhangs corresponding to the overlapping sequence for the intended fusion site. The PCR products are joined by further PCR, made possible by the primer overhangs, and the final PCR product is then cut with restriction enzymes and ligated back into the wild type channel. Putting both N- and C- terminal fluorescent tags on the Kv2.1 potassium channel by this technique enabled 52 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

Fig. 2.13 Example of data obtained after insertion of a tag using overlap extension. Current-voltage curve for rKv2.1 wild type

(!, n = 4), and the chimera rKv2.1N-YFP-C-CFP (*, n = 5) with fluorescent tags on both the N terminus (YFP) and C terminus (CFP).

the use of FRET experiments to investigate movement of the intracellular N- and C-terminal regions of the expressed channel during channel gating (work in progress, [75]). The tags themselves did not affect the electrophysiological properties of the channel (Fig. 2.13).

2.6 Channel Subunit Concatamers

Another aspect of ion channel structure–function analysis that merits discussion is the analysis of homomeric and heteromeric subunit assembly and stoichiometry. Many classes of ion channel consist of multi-subunit complexes, which can be either homomeric or heteromeric. Examples of functional homomeric channel complexes are the P2X1 receptor channel [76], the Shaker potassium channel [77] and the a7 nicotinic acetylcholine receptor channel [78]. Heteromeric examples are the P2X2/3 channel [55], the epithelial sodium channel (ENaC) [79] or the GABA-A receptor [80]. Structure–function studies, including mutagenesis, have helped to resolve the stoichiometry of subunit expression required to form heteromeric channels, and have identified specific domains involved in channel multi- merisation. However, an additional approach that has proved useful to investigate the subunit composition required for functional channel formation is the generation and characterisation of channel concatamers, whereby two or more subunits are joined together to make a single recombinant protein. Usually consecutive subunits within these concatamers are separated through the inclusion of short, flexible peptide sequences (linkers); a short polyglutamine peptide (5–8 amino acids) is 2.7 Concluding Remarks 53 frequently used [79, 81, 82]. Studies using subunit concatamers can enable characterisation of heteromeric channels while avoiding interference from homomeric channels that would be present in co-expression studies using the individual subunits, for instance to allow studies of inter-subunit binding sites [83, 84]. A prerequisite for successful use of concatameric channels is that the N- and C-termini of the subunits are either both extracellular or both intracellular. For channel subunits with extracellular N-termini and signal peptide sequence, the signal sequence should be removed from the downstream subunit(s) to favor replication of the correct transmembrane topology in the concatameric channel [86]. The channel concatamers cDNAs can be generated using several of the techniques described in this chapter. For example, restriction sites can be utilised, along with a synthetic DNA linker encoding a short peptide linker, to couple two channel open reading frames, in a manner similar to that described to introduce a short peptide tag [85]. Alternatively, site-directed mutagenesis (preferably employing silent mutations) can be used to introduce appropriate restriction sites at the 3' and 5' ends of the open reading frames of the two subunits to be ligated. For example, this approach was used to generate a concatamer of three P2X2 subunits, and allowed characterisation of an intersubunit zinc binding site [83]. Overlap extension PCR can also be used to generate the concatameric channel cDNA when distinct subunits are to be coupled. The requirement for distinct PCR primer sequences at the 5' and 3' ends of the open reading frames to be coupled precludes the use of this approach to link identical subunits. As discussed previously, the overlap extension PCR technique removes the necessity to incorporate restriction enzyme sites into the concatameric open reading frame yielding “seamless” cDNA constructs. Several studies have highlighted the need to accompany functional characterisation of channel concatamers with biochemical validation, to ensure that the integrity of the engineered concatamer is maintained when expressed. For instance, biochemical characterisation of concatameric P2X1 channels demonstrated that unexpected monomer (or dimer) byproducts were readily expressed and could account for the functional responses observed rather than the concatamer itself [82]. Also, experiments with nicotinic acetylcholine receptor subunits showed that subunit concatamers can form functional channels through the unexpected incorporation of tandem-expressed subunits and that these were only revealed through analysis of the expressed channels using protein biochemical techniques (sucrose gradient sedimentation) [81, 85].

2.7 Concluding Remarks

Extensive structure-function data, generated by molecular biology techniques, described in this chapter, has already produced much vital information about ion channels. Recent progress in determining the three-dimensional structures of several ion channels has allowed models to be developed that account for multiple 54 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

aspects of channel function, including activation, inactivation, ligand binding, subunit interactions, toxin or drug binding and channel regulation. Generation of structure–function data to support these models will continue to be required in order to resolve the specific channel domains and individual residues involved in particular aspects of ion channel function. We hope that the information presented in this chapter will guide investigators in their selection of molecular biology approaches for making ion channel cDNA constructs for structure–function studies.

References

1 Watson, J. D., Gilman, M.,Witkowski, 10 Weisberg, R. A., Landy, A., Site-specific J., Zoller, M., Recombinant DNA, Scien- recombination in phage lambda, in tific American Books, 2nd edn., 1992. Lambda II, Cold Spring Harbor Press, 2 Sambrook, J., Fritsch, E. F., Maniatis, New York, 1983, pp. 211–250. T., Molecular Cloning, A Laboratory Ma- 11 Landy, A., Dynamic, structural, and re- nual, Cold Spring Harbor Laboratory gulatory aspects of lambda site-specific Press, New York, 2nd edn., 1989. recombination, Annu. Rev. Biochem. 3 Swanson, G. T., Gereau IV, R. W., 1989, 58, 913–49. Green,T., Heinemann, S.F., Identifica- 12 Hartley, J. L., Temple, G. F., Brasch, tion of amino acid residues that control M. A., DNA Cloning using in vitro site- functional behaviour in GluR5 and specific recombination, Genome Res. GluR6 kainate receptors, Neuron 1997, 2000, 10, 1788–1795. 19, 913–926. 13 Obrdlik, P. et al, K+ channel interacti- 4 Lu, Q., Seamless cloning and gene fu- ons detected by a genetic system optimi- sion, Trends Biotechnol. 2005, 23,199–207. zed for systematic studies of membrane 5 Li, J., Stevens, L., Klugbauer, N.,Wray, protein interactions, Proc. Natl. Acad. Sci. D., Roles of molecular regions in deter- USA 2004, 101, 12242–12247. mining differences between voltage de- 14 Luan, C-H. et al., High-throughput ex-

pendence of activation of CaV3.1 and pression of C.elegans proteins, Genome CaV1.2 calcium channels, J. Biol. Chem. Res. 2004, 14, 2102–2110. 2004, 279, 26858–26867. 15 Ju, M., Stevens, L., Leadbitter, E., 6 Clark, J. M., Novel non-templated nu- Wray, D., The Roles of N- and C-terminal cleotide addition reactions catalyzed by determinants in the activation of the procaryotic and eucaryotic DNA polyme- Kv2.1 potassium channel, J. Biol. Chem. rises, Nucleic Acids Res. 1998, 16, 9677– 2003, 278, 12769–12778. 9686. 16 Horton, R. M., Hunt, H. D., Ho, S. N., 7 Mead, D., Pey, N. K., Herrnstadt, C., Pullen, J. K., Pease, L. R., Engineering Marcil, R. A., Smith, L. M., A universal hybrid genes without the use of restric- method for the direct cloning of PCR am- tion enzymes: gene splicing by overlap plified nucleic acid, Biotechnology 1991, 9, extension, Gene 1989, 77, 61–68. 657–663. 17 Ho, S.N., Hunt, H.D., Horton, R.M., 8 Shuman, S., Novel approach to molecu- Pullen, J.K., & Pease, L.R. Site-directed lar cloning and polynucleotide synthesis mutagenesis by overlap extension using using vaccinia DNA topoisomerase, J. the polymerase chain reaction, Gene Biol. Chem. 1994, 269, 32678–32684. 1989, 77, 51–59. 9 Ju, M.,Wray, D., Molecular identification 18 Stevens, L., Structure and function of and characterisation of the human eag2 potassium and calcium channels, PhD potassium channel, FEBS Lett. 2002, 524, Thesis, University of Leeds, 2005. 204–210. 19 Shevchuk, N. A., Bryksin, A. V., Nusin- ovich,Y. A., Cabello, F. C., Suther- References 55

land, M., Ladisch, S., Construction of of aromatic and basic amino acids revea- long DNA molecules using long PCR-ba- led using mutagenesis and partial ago- sed fusion of several fragments simulta- nists, J. Biol. Chem. 2004, 279, 9043– neously, Nucleic Acids Res. 2004, 32, e19. 9055. 20 Geiser, M., Cebe, R., Drewello, D., 29 Jiang, L-H., Rassendren, F., Schmitz, R., Integration of PCR frag- Surprenant, A., North, R.A., Identifica- ments at any specific site within cloning tion of amino acid residues contributing vectors without the use of restriction en- to the ATP-binding site of a purinergic zymes and DNA ligase, BioTechniques P2X receptor, J. Biol. Chem. 2000, 275, 2001, 31(1), 88–92. 34190–34196. 21 Weiner, M. P., Costa, G. L., Schoett- 30 Nakazawa, K., Ojima, H., Ohno,Y., lin,W., Cline, J., Mathur, E., Bauer, J. A highly conserved tryptophan residue B., Site-directed mutagenesis of double- indispensable for cloned rat neuronal stranded DNA by the polymerase chain P2X receptor activation, Neurosci. Lett. reaction, Gene 1994, 151,119–123. 2002, 324, 141–144. 22 Wang,W. Malcolm, B. A., Two stage 31 Clyne, J. D.,Wang, L-F., Hume, R. I., PCR protocol allowing introduction of Mutational analysis of the conserved resi- multiple mutations, deletions and inser- dues of the rat P2X2 receptor, J. Neurosci. tions using QuikChangeTM site-directed 2002, 22, 3873–3880. mutagenesis. BioTechniques 1999, 26, 32 Ennion, S. J., Evans, R. J., Conserved 680–682. cysteine residues in the extracellular loop 23 Wilson, G. G., O’Neill, C. A., Sivapra- of the human P2X1 receptor form disul- sadarao, A., Findlay, J. B.,Wray, D., fide bonds and are involved in receptor Modulation by protein kinase A of a clo- trafficking to the cell surface, Mol. Phar- ned rat brain potassium channel expres- macol. 2002, 61, 303–311. sed in Xenopus oocytes, Pflugers Arch. 33 Clyne, J. D., LaPointe, L. D., Hume, 1994, 428, 186–193. R. I., The role of histidine residues in 24 Wilson, G. G., Sivaprasadarao, A., modulation of the rat P2X2 purinoceptor Findlay, J. B.,Wray, D., Changes in acti- by zinc and pH, J. Physiol. 2002, 539, vation gating of IsK potassium currents 347–359. brought about by mutations in the trans- 34 Roberts, J. A., Evans, R. J., Mutagenesis membrane sequence, FEBS Lett. 1994, studies of conserved proline residues of 353, 251–254. human P2X1 receptors for ATP indicate 25 Bannister, J. P.,Young, B. A., Sivapra- that proline 272 contributes to channel sadarao, A.,Wray, D., Conserved extra- function, J. Neurochem. 2005, Epub cellular cysteine residues in the inwardly ahead of print. rectifying potassium channel Kir2.3 are 35 Morrison, K. L.,Weiss, G. A., Combina- required for function but not expression torial alanine-scanning, Curr. Opin. in the membrane, FEBS Lett. 1999, 458, Chem. Biol. 2001, 5, 302–307. 393–399. 36 Li-Smerin,Y., Hackos, D. H., Swartz, K. 26 Bannister, J.P.,Young, B.A., Main, J., a-Helical structural slements within M.J., Sivaprasadarao, A.,Wray, D. The the voltage-sensing domains of a K+ effects of oxidizing and cysteine-reactive Channel, J. Gen. Physiol. 2000, 115, 33– reagents on the inward rectifier potas- 50. sium channels Kir2.3 and Kir1.1. Pflugers 37 Perry, M., de Groot, M. J., Helliwell, Arch. 1999, 438, 868–878. R., Leishman, D., Tristani-Firouzi, M., 27 Ennion, S., Hagan, S., Evans, R. J., The Sanguinetti, M. C., Mitcheson, J., role of positively charged amino acids in Structural determinants of HERG chan- ATP recognition by human P2X1 recep- nel block by clofilium and ibutilide, Mol. tors, J. Biol. Chem. 2000, 275, 29361– Pharmacol. 2004, 66, 240–249. 29367. 38 Tedford, H.W., Gilles, N., Menez, A., 28 Roberts, J. A., Evans, R. J., ATP binding Doering, C. J., Zamponi, G.W., King, at human P2X1 receptors: contribution G. F., Scanning mutagenesis of o-Atraco- 56 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

toxin-Hv1a reveals a spatially restricted 49 Haines,W. R.,Voigt, M. M., Migita, K., epitope that confers selective activity Torres, G. E., Egan, T. M., On the contri- against insect calcium channels, J. Biol. bution of the first transmembrane do- Chem. 2004, 279, 44133–44140. main to whole-cell current through an 39 Ragsdale, D. S., McPhee, J. C., ATP-gated ionotropic P2X receptor, J. Scheuer,T., Catterall,W. A., Molecular Neurosci. 2001, 21, 5885–5892. determinants of state-dependent block of 50 Xu, M., Covey, D. F., Akabas, M. H., In- Na+ channels by local anesthetics, Science teraction of picrotoxin with GABA(A) re- 1994, 265, 1724–1728. ceptor channel-lining residues probed in 40 Karlin, A., Akabas, M. H., Substituted- cysteine mutants, Biophys. J. 1995, 69, cysteine accessibility method, Methods 1858–1867. Enzymol. 1998, 293, 123–145. 51 Xu, M., Akabas, M. H., Amino acids li- 41 Milligan, C. J.,Wray, D., Local move- ning the channel of the g-aminobutyric ment in the S2 region of the voltage-ga- acid type A receptor identified by cy- ted potassium channel hKv2.1 studied steine substitution, J. Biol. Chem. 1993, using cysteine mutagenesis, Biophys. J. 268, 21505–21508. 2000, 78, 1852–1861. 52 Yusaf, S. P.,Wray, D., Sivaprasadarao, 42 Pascual, J. M., Shieh, C. C., Kirsch, G. A., Measurement of the movement of the E., Brown, A. M., Multiple residues spe- S4 segment during the activation of a vol- cify external tetraethylammonium blok- tage-gated potassium channel, Pflugers kade in voltage-gated potassium chan- Arch. 1996, 433, 91–97. nels, Biophys. J. 1995, 69, 428–434. 53 Larsson, H. P., Baker, O. S., Dhillon, 43 Li, J., Stevens, L.,Wray, D., Molecular D. S., Isacoff, E.Y., Transmembrane regions underlying the activation of low- movement of the Shaker K+ channel S4, and high-voltage activating calcium chan- Neuron 1996, 16, 387–397. nels, Eur. Biophys. J. 2005, DOI: 10.1007/ 54 Jiang, L-H., Rassendren, F., Spelta,V., s00249–005-0487–7 [Epub ahead of Surprenant, A., North, R.A., Amino print]. acid residues involved in gating identi- 44 Akabas, M. H., Stauffer, D. A., Xu, M., fied in the first membrane-spanning do- Karlin, A., Acetylcholine receptor chan- main of the rat P2X2 receptor, J. Biol. nel structure probed in cysteine-substitu- Chem. 2001, 276,14902–14908. tion mutants, Science 1992, 258, 307–310. 55 Jiang, L-H., Kim, M., Spelta,V., Bo, X., 45 Akabas, M. H., Karlin, A., Identification Surprenant, A., North, R.A., Subunit of acetylcholine receptor channel-lining arrangement in P2X receptors, J. Neuro- residues in the M1 segment of the a-sub- sci. 2003, 23, 8903–8910. unit, Biochemistry 1995, 34,12496–12500. 56 Starace, D.M., Bezanilla, F., Histidine 46 Akabas, M. H., Kaufmann, C., Arch- scanning mutagenesis of the basic resi- deacon, P., Karlin, A., Identification of dues of the S4 segment of the Shaker K+ acetylcholine receptor channel-lining re- channel, J. Gen. Physiol. 2001, 117, 469– sidues in the entire M2 segment of the a- 490. subunit, Neuron 1994, 13, 919–927. 57 Silberberg, S. D., Chang, T. H., 47 Rassendren, F., Buell, G., Newbolt, A., Swartz, K. J., Secondary structure and North, R. A., Surprenant, A., Identifi- gating rearrangements of transmem- cation of amino acid residues contribu- brane segments in rat P2X4 receptor ting to the pore of a P2X receptor, EMBO channels, J. Gen. Physiol. 2005, 125, 347– J. 1997, 16, 3446–3454. 359. 48 Egan, T. M., Haines,W.R.,Voigt, M. M., 58 Li-Smerin,Y., Swartz, K. J., Helical A domain contributing to the ion chan- structure of the COOH terminus of S3 nel of ATP-gated P2X2 receptors identi- and its contribution to the gating modified by the substituted cysteine accessibi- fier toxin receptor in voltage-gated ion lity method, J. Neurosci. 1998, 18, 2350– channels, J. Gen. Physiol. 2001, 117, 205– 2359. 218. References 57

59 Monks, S. A., Needleman, D. J., Miller, ligand binding domain, J. Biol. Chem. C., Helical structure and packing orienta- 2002, 277, 29210–29223. tion of the S2 segment in the Shaker K+ 68 Jin, T., Peng, L., Mirshahi, T., channel, J. Gen. Physiol. 1999, 113, 415– Rohacs, T., Chan, K.W., Sanchez, R., 423. Logothetis, D.E., Identification of criti- 60 Choe, S., Stevens, C. F., Sullivan, J. cal residues controlling G protein-gated M., Three distinct structural environ- inwardly rectifying K+ channel activity ments of a transmembrane domain in through interactions with the bg sub- the inwardly rectifying potassium chan- units of G proteins, Mol. Cell. 2002, 10, nel ROMK1 defined by perturbation, 469–481. Proc. Natl. Acad. Sci USA 1995, 92, 69 He, M-.L., Zemkova, H., Stojilkovic, 12046–12049. S. S., Dependence of purinergic P2X re- 61 Hong, K. H., Miller, C., The lipid– ceptor activity on ectodomain structure, protein interface of a Shaker K+ channel, J. Biol. Chem. 2003, 278, 10182–10188. J. Gen. Physiol. 2000, 115, 51–58. 70 Jarvik, J. W.,Telmer, C. A., Epitope tag- 62 Li-Smerin,Y., Hackos D. H., Swartz, ging, Annu. Rev. Gen. 1998, 32, 601–618. K. J., A localized interaction surface for 71 Ficker, E., Dennis, A. T.,Wang, L., voltage-sensing domains on the pore Brown, A. M., Role of the cytosolic cha- domain of a K+ channel, Neuron 2000, perones Hsp70 and Hsp90 in maturation 25, 411–423. of the cardiac potassium channel hERG, 63 Jenkins, A., Andreason, A., Trudell, Circulation Res. 2003, 92, E87–E100. J. R., Harrison, N. L., Tryptophan scan- 72 Grabner, M., Dirksen, R.T., Beam, ning mutagenesis in TM4 of the GA- K. G., Tagging with green fluorescent BA(A) receptor a1 subunit: implications protein reveals a distinct subcellular dis- for modulation by inhaled anaesthetics tribution of L-type and non-L-type Ca2+ and ion channel structure, Neuropharma- channels expressed in dysgenic myotu- cology 2002, 43, 669–678. bes, Proc. Natl. Acad. Sci. USA. 1998, 64 Guzman, G. R., Santiago, J., Riccardo, 95,1903–1908. A., Marti-Arbona, R., Rojas, L. V., 73 Makhina, E. N., Nichols, C. G., Inde- Lasalde-Dominicci, J. A., Tryptophan pendent trafficking of K-ATP channel scanning mutagenesis in the aM3 trans- subunits to the plasma membrane, J. membrane domain of the Torpedo califor- Biol. Chem. 1998, 273, 3369–3374. nica acetylcholine receptor: functional 74 Grailhe, R., de Carvalho, L. P., Paas, and structural implications, Mol. Phara- Y., Le Poupon, C., Soudant, M., Brege- macol. 2003, 61, 12243–12250. stovski, P., Changeux, J. P., Corrin- 65 Pamchenko,V. A., Glasser, C. R., ger P. J., Distinct subcellular targeting Mayer, M.L., Structural similarities be- of fluorescent nicotinic a3b4 and seroto- tween glutamate receptor channels and ninergic 5-HT3A receptors in hippocam- K+ channels examined by scanning mu- pal neurons, Eur. J. Neuroscience. 2004, tagenesis, J. Gen. Physiol. 2001, 117, 345– 19, 855–862. 360. 75 Kobrinsky, E., Stevens, L., Bentil, S., 66 Tamamizu, S., Guzman, G. R., Wray, D., Soldatov, N.M., Molecular Santiago, J., Rojas, L.V., McNamee, rearrangements associated with the vol-

M.G., Lasalde-Dominicci, J. A., Functio- tage gating of the KV2.1 channel, FASEB nal effects of periodic tryptophan substi- J. 2004, 18(A226), Suppl. S. tutions in the aM4 transmembrane do- 76 Nicke, A., Baumert, H.G., Rettinger, main of the Torpedo californica nicotinic J., Eichele, A., Lambrecht, G., Mutsch- acetylcholine receptor, Biochemistry. 2000, ler, E., Schmalzing, G., P2X1 and P2X3 39, 4666–4673. receptors form stable trimers – a novel 67 Sine, S. M.,Wang, H. L., Bren N., structural motif of ligand-gated ion chan- Lysine scanning mutagenesis delineates nels, EMBO J. 1998, 17, 3016–3028. structural model of the nicotinic receptor 77 MacKinnon R., Determination of the subunit stoichiometry of a voltage-activa- 58 2 Molecular Biology Techniques for Structure – Function Studies of Ion Channels

ted potassium channel, Nature 1991, 350, 82 Nicke, A., Rettinger, J., Schmalzing, 232–235. G., Monomeric and dimeric by-products 78 Drisdel, R.C., Green,W. N., Neuronal are the principal functional elements of a-bungarotoxin receptors are a7 subunit higher order P2X1 concatamers, Mol. homomers, J. Neurosci. 2000, 20, 133– Pharmacol. 2003, 63, 243–252. 139. 83 Nagaya, N., Tittle, R. K., Saar, N., 79 Firsov, D., Gautschi, I, Merillat, Dellal, S. S., Hume, R. I., An intersubu- A. M., Rossier, B. C., Schild, L., The he- nit zinc binding site in rat P2X(2) recep- terotetrameric architecture of the epithe- tors, J. Biol. Chem. 2005, 280, 25982– lial sodium channel (ENaC), EMBO J. 25993. 1998, 17, 344–352. 84 Yuill, K., Ashmole, I., Stanfield, P. R., 80 McKernan, R. M.,Whiting, P.J.,Which The selectivity filter of the tandem pore GABAA-receptor subtypes really occur in potassium channel TASK-1 and its pH- the brain? Trends Neurosci. 1996, 19, 139– sensitivity and ionic selectivity, Pflugers 143. Arch. Eur. J. Physiol. 2004, 448, 63–69. 81 Groot-Kormelink, P. J., Broadbent, 85 Zhou Y., Nelson, M. E, Kuryatov, A., S. D., Boorman, J. P., Sivilotti, L.G., Choi, C., Cooper, J., Lindstrom, J., Incomplete incorporation of tandem sub- Human alpha4 beta2 acetylcholine units in recombinant neuronal nicotinic receptors formed from linked subunits, receptors, J. Gen. Physiol. 2004, 123, 697– J. Neurosci. 2003, 23, 9004–9015. 708. 59

3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology Paul B. Bennett, Niki Zacharias, John B. Nicholas, Sue Dee Sahba, Ashutosh Kulkarni, and Mark Nowak

3.1 Introduction

Ion channel proteins are the fundamental electrical signaling molecules of living systems. Among other things they function as electrical field effect transistors that permit human beings to perform complex reasoning and such purposed actions as gene cloning, aircraft design and space travel. Protein structure, function and synthesis are complex processes programmed in the genetic code. A gene (sequence of DNA archived in human chromosomes) is transcribed into an mRNA which is read by the ribosomes whose task it is to assemble appropriate amino acids according to the code. The amino acids are each shuttled to the ribosomal complex by an amino acid-specific (transfer) tRNA. Proteins consist of an assort- ment of the 20 natural amino acids. Site directed mutagenesis, the purposeful substitution of a codon specifying an alternative amino acid at a specific position in the protein, has been widely employed to probe the role of amino acids in proteins both for structure–function and to elucidate drug binding to protein receptors (see Chapter 2). Although extremely powerful for elucidating some aspects of these relationships, conventional site-directed mutagenesis is limited to substitution of the original amino acid with one of the other 19 naturally occurring amino acids. Binding of a drug to an ion channel protein requires a very specific three-dimensional arrangement between the shape and chemical functionality of the drug and the amino acids in the protein binding site. Multiple, rather subtle types of interactions can occur, including hydrogen bonding and dipole attractions. Although the natural amino acids are used to perturb protein structure and function, they provide a limited range of structural and chemical diversity. As such, it is not possible with conventional mutagenesis to subtly or specifically probe the role of side chains or the protein carbonyl backbone. More recently, methods have been developed to permit incorporation of amino acids that do not occur in nature (unnatural amino acids) [1–4]. For example, it is possible to site-specifically incorporate any number of unnatural amino acids into almost any protein through the approach termed nonsense suppression. The terminology of nonsense suppression

comes from the field of microbial genetics. A nonsense mutation in nature is a base change that converts a codon within the gene sequence into a stop codon. In microbes, a suppressor is a second mutation that restores a function lost by a primary nonsense mutation. In the technique of nonsense suppression, a non-native nonsense (stop) codon is inserted into a cDNA and is reverted or suppressed by introduction of a novel amino acid-tRNA that recognizes the stop codon. This approach has enormous potential and permits systematic structure–function analyses with precision far beyond conventional site-directed mutagenesis. Nonsense suppression methodology has enabled a much wider range of protein structure– function studies with more detailed information [5–7]. Natural amino acid variants with subtle perturbations in side chain chemistry as well as probes such as spin labels, FRET (fluorescence resonance energy transfer) pairs or fluorescent groups can be incorporated [6, 8–11]. Using this approach insights can be gained that are not possible using conventional site-directed mutagenesis (see below). There are a number of excellent reviews on this subject [6, 8–12]. The goal of this chapter is to focus first on methodological issues that a potential user must recognize; and secondly, on recent applications of the methodology to illustrate general strategies and specific information obtainable using unnatural amino acids.

3.2 Unnatural Amino Acid Mutagenesis Methodology

The basic approach to incorporation of an unnatural amino acid is outlined in Fig. 3.1. First, a stop codon (TAG, “amber”) is incorporated into the recombinant cDNA at the position coding for the amino acid of interest using conventional site-directed mutagenesis (see Chapter 2). From this cDNA a UAG-containing mRNA is generated, either by in vitro transcription or within cells. Secondly, a tRNA that recognizes the amber stop codon is prepared and chemically acylated with the desired unnatural amino acid. The DNA encoding the gene of interest or the in vitro transcribed mRNA and the tRNA carrying the unnatural amino acid are then added to an appropriate expression system. Several groups have contribu- ted important steps in the development of these methods [1–3, 13–18]. Methodological issues focus primarily on the tRNA. When considering nonsense suppression, care must be taken to ensure that the tRNA is orthogonal in the chosen expression system. This means that the exogenous suppressor tRNA must not be recognized by the native aminoacyl tRNA synthetases (RS), the enzymes that charge tRNAs with their cognate amino acids. Unless this is achieved, the suppressor tRNA, once it has delivered its unnatural amino acid to a protein, will be charged with a ‘natural’ amino acid and return to the protein synthesis cycle. This will produce a mixture of proteins with natural and unnatural amino acids and would be of limited value. Another methodological issue that must be addressed when using nonsense suppression is the fidelity of the system. To test this at least two control experi- 3.2 Unnatural Amino Acid Mutagenesis Methodology 61 tran- in vitro only the nonsense codon is createdacid with appended. the The unnatural gene amino of interestscribed (cDNA mRNA) or and the tRNA arewhere introduced the into protein is a expressed cell andphysiology. (This detected figure using also electro- appears with the color plates.) system with electrophysiology read- The fundamental protocol for incorporating unna- in vivo” translation Fig. 3.1 tural amino acids through nonsense suppression,an showing “ out. A special nonsense codon isof introduced interest into at the the cDNA position of interest. A tRNA that recognizes 62 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

ments are performed. In one control experiment, UAG-mRNA only is added to the expression system. This control tests if functional protein will be synthesized without suppressor tRNA present and if truncated protein will interfere with the assay. A functional protein could be synthesized if an endogenous tRNA misreads the UAG codon (read through) [2]. It is also possible for the ribosome to stop translation when it reaches the UAG codon, creating a truncated protein that could interfere with a functional assay. Another control experiment to confirm the fidelity of the system is to compare the function of wild-type protein expressed using nonsense suppression (UAG mRNA and tRNA acylated with the wild-type amino acid) and protein expressed using wild-type mRNA. Because one has placed the natural wild type amino acid back into the protein at its correct position, the protein expressed using nonsense suppression or wild-type mRNA should function identically [2]. We rarely see expression of protein using UAG- mRNA, only in our in vivo nonsense suppression experiments in Xenopus oocytes, and the channels/receptors display indistinguishable characteristics when expressed using nonsense suppression or mRNA. Figure 3.2 illustrates some of the possible side chain modifications and their use.

Fig. 3.2 Unnatural amino acid mutations that backbone and the –OH side chain. Substitu- probe H-Bonds, cation–p, and p–p interaction of phenylalanine (Phe) for tyrosine (Tyr) tions. Replacement of a serine by an alanine removes an H-bonding group. Additions of removes an –OH and the potential for hydro- fluorine (F) deactivate the ring and affect p–p gen bonding; substitution of threonine steri- or cation–p interactions. Cyclohexyl-alanine cally alters the –OH; replacement of the pep- replaces an aromatic ring with an aliphatic tide backbone amide with an ester in hydroxy- ring to probe hydrophobic interactions with- threonine impacts hydrogen bonding by the out greatly altering side chain volume. 3.2 Unnatural Amino Acid Mutagenesis Methodology 63

The originally described yeast-derived suppressor tRNA [1] is not useful in some expression systems such as Xenopus oocytes because the endogenous aminoacyl tRNA synthetases (RS) reacylate the tRNA. Using established rules of RSs– tRNA recognition, we generated a new suppressor tRNA [13] by taking advantage of the fact that Tetrahymena thermophila has a nonstandard genetic code: TAG is not a stop codon but instead codes for glutamine. Thus, a naturally occurring suppressor tRNA exists and a slight variant of this (THG73) is a much improved suppressor in the Xenopus oocyte system [19]. A limitation of the nonsense suppression methodology is that the aminoacyl tRNA is consumed in the process and is not regenerated. The synthetic methods for generating the acylated tRNA reagent are well known, but still require a significant effort. Furthermore the suppression process is approximately 15–20% as efficient as the natural process [2]. As such it requires excess aminoacyl tRNA for each molecule of protein synthesized. In order to sustain the presence of the altered protein, this reagent must be delivered in large excess or continually delivered. A solution to these problems is to synthesize sufficient quantities of aminoacyl tRNA required to generate the protein of interest in an expression system. For example, Xenopus oocytes can be injected with nanogram quantities of tRNA or in vitro assays can be scaled up. However, for assays with limited sensitivity such as NMR it can still be very difficult to synthesize enough protein, particularly membrane proteins. This limitation may be a primary reason limiting the use of nonsense suppression methodology. In principle, novel RNA synthases (RS) can be evolved and selected to recognize a given unnatural amino acid and thus overcome the requirement for the large quantities of exogenous suppressor aminoacyl-tRNA discussed above. Methods have been developed to generate orthogonal tRNA-RS pairs specific for a given unnatural amino acid [20–23]. In this approach, a novel orthogonal RS catalyses the acylation of the suppressor tRNA with the unnatural amino acid. Importantly, the suppressor tRNA can be re-acylated, avoiding the need to deliver large quantities of exogenous aminoacylated suppressor tRNA. To generate orthogonal tRNA-RS pairs directed evolution is employed [21, 22]. With directed evolution, all possible mutations are introduced at key residues in the RS important for recognition of the cognate amino acid. The mutant RS library is then screened for aminoacylation of only the corresponding suppressor tRNA (and not the endogenous tRNAs of the host expression system) with only the desired unnatural amino acid (and not any natural amino acids). Using this approach, specific orthogonal tRNA-RS pairs have been developed for a variety of unnatural amino acids and milligram quantities of unnatural amino acid protein mutants have been obtained in E. coli, yeast and mammalian expression systems [23–25]. The limitation of this approach is that a suppressor tRNA-RS pair must be developed for each unnatural amino acid and, therefore it is not generally applicable. 64 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

3.3 Unnatural Amino Acid Mutagenesis for Ion Channel Studies

An alternative strategy is to use a highly sensitive assay where only very small amounts of protein are required. Ion channels are complex, integral membrane proteins that lower the energy barrier for ion translocation across the hydrophobic lipid bilayer. They are highly dynamic proteins whose function (gating) and pharmacology depend on the channel conformational state. For the most part these proteins are not readily amenable to high-resolution structural methods such as X-ray crystallography and NMR (see Chapter 9). Because these methods provide only limited or no temporal information on the protein dynamics, they provide a description of one particular protein conformational state. Thus ion channels are good candidates for studies using unnatural amino acids as these methods provide information on structure and function. This approach can be readily adopted with ion channel proteins because there are exquisitely sensitive methods for measuring protein function. Ion channels can be easily expressed in a number of systems including HEK and CHO mammalian cells (Chapter 4) or in Xenopus oocytes (Chapter 1) which are a well-established general vehicle for heterologous expression and characterization of ion channel proteins [26, 27]. Using whole cell two-electrode voltage- clamp electrophysiology in oocytes, as little as 10–15 M of protein can be detected. Using the patch-clamp method of single channel recording a single molecule can be analyzed. Xenopus oocytes are injected with UAG-mRNA for the ion channel of interest and the suppressor tRNA. Usually, after 24 to 48 h the cells have synthesized the ion channel and have embedded it into their cell membrane. The ion channels can then be studied using electrophysiology methods. The generality of the nonsense suppression method has been established with a large number of different unnatural amino acids (including those shown in Fig. 3.2) incorporated at a comparable number of sites in dozens of proteins. Scientists at Neurion Pharmaceuticals and the California Institute of Technology have successfully incorporated over 85 different amino and hydroxy acids (including 15 of the 20 natural amino acids) into at least 14 different receptors or ion channels (e.g. the M2 muscarinic acetylcholine (ACh) receptor, the nicotinic ACh receptor a1, a2, a4, and b1 subunits, inward rectifier and G-protein coupled potassium

channels Kir1.1 (ROMK1), Kir2.1, GIRK1, GIRK4, the 5-HT3A receptor subunit, the MOD1 receptor, the NR2A subunit of the NMDA receptor, delayed rectifier (Kv and Shaker) and hERG potassium channels, CFTR, and the GAT1 GABA transporter) for a total of 454 combinations of amino acids/sites [5, 6, 28]. It has been generally observed that hydrophobic amino acids are incorporated more efficiently than charged ones [5]. Proteins are generally made up of l-amino acids (the levorotatory optical isomer). d-amino acids are not compatible with nonsense suppression, nor are b-amino acids [29]. Nevertheless, some fairly elaborate unnatural amino acids have been employed [5]. For example, a biotin derivative was incorporated consecutively in several positions in an ion channel to identify surface-exposed residues in the channel [30]. We have incorporated caged amino acids, unnatural amino acids that mimic ligands of ion channels, amino acids 3.4 Structure–Function Example Studies 65 that when exposed to light rearrange and break the backbone of the ion channel, and several hydroxy acids [31–36]. In addition to introducing amino acids with unnatural side chains, it is also possible to insert a-hydroxy acids in place of a-amino acids [37]. This mutation replaces the backbone amide bond with an ester, thus replacing amide NH with an ester O, eliminating a hydrogen bonding donor. In addition, the adjacent backbone carbonyl is a weaker hydrogen bond acceptor in the ester than in the amide. a-Hydroxy acid substitutions are useful as probes of secondary structure, because they disrupt hydrogen bonding that stabilizes a-helices and b-sheets. More important for our work, drugs often hydrogen bond to backbone carbonyls that are not already involved in hydrogen bonds with neighboring amino acids [28]. This will be described further in the hERG section of this chapter.

3.4 Structure–Function Example Studies

3.4.1 Nicotinic Acetylcholine Receptor

Studies employing conventional site-directed mutagenesis often suggest a particular role for a given amino acid in drug binding, yet the evidence is incomplete and does not lend itself to a strict physical chemical interpretation for a mechanism. This is due to the limited number of natural amino acids which precludes a detailed investigation of the relevant interactions between the protein and the drug. Additional systematic subtle changes, made possible by nonsense suppression, can provide the definitive evidence for a mechanism. For instance, in the nicotinic acetylcholine receptor (nAChR), early photo-affinity labeling studies identified many of the amino acids that constitute the binding site for acetylcholine (ACh, the agonist for the receptor) [38–40]. However, the exact amino acids that bind ACh and the specific binding interactions were not known. Recently, using nonsense suppression with unnatural amino acids, it was demonstrated that aTrp149 in the mouse muscle nAChR interacts with the quaternary ammonium of acetylcholine in a cation–p interaction [35, 41]. A cation–p bond is the noncovalent interaction that occurs between the p electrons of a conjugated system (like an aromatic ring) and a cation. In proteins, a cation–p interaction can occur between a cation and the aromatic rings of tyrosine, phenylalanine, neutral histidine, or tryptophan [42, 43]. The cation–p interaction in nAChR was determined by incorporating the unnatural amino acids 4-F-Trp (addition of fluorine in the 4 position of the tryptophan ring), 4,5-Trp, etc. into the ACh binding site [41]. Progressive fluorination decreases the ability of the tryptophan ring to interact with a cation while having minimal effect on the sterics of the amino acid (Fig. 3.3). A linear relationship was observed between the theoretically predicted change in the cation–p interaction energy upon successive fluorination and the experimentally observed change in receptor affinity for ACh binding. 66 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

DDG = 0.6 ln (KD-mutant /KD-WT) (1)

This finding indicates that, upon binding, the quaternary ammonium of ACh makes van der Waals contact with the indole side chain of aTrp149, which allows us to infer the localization of the binding between ACh and Trp149 [41] to within 0.5 Å.

Fig. 3.3 Correlation between the change in in a decreased binding energy. Binding energy

EC50 and calculated binding energy in nicoti- was calculated in the gas phase and the abso- nic acetylcholine receptor binding pocket. lute values will be scaled down by the pre- Tryptophan 149 was systematically replaced sence of water and other factors but the trend with fluorinated (F) tryptophan (Trp, 1–4 F) is expected to remain the same. (This figure unnatural amino acids. Each added F further also appears with the color plates.) deactivates the Trp p electron cloud, resulting

In marked contrast, substitution of fluorinated Trp residues at a149 did not ap- preciably affect the interaction of nicotine with the mouse nAChR. Instead, it was determined that nicotine interacts with the muscle nAChR via a hydrogen bond between the positively charged nitrogen in the pyrrolidine ring of nicotine and the backbone carbonyl of Trp a149. This was demonstrated by insertion of the unnatural amino acid a-OH-threonine into a150. This changes the a149 backbone amide bond to an ester, weakening the H-bonding capability of the carbonyl. This interaction had been previously suggested by modeling studies on the ACh binding protein [36]. Previous to these findings, ACh and nicotine were thought to be represented by a single pharmacophore. These studies clearly demonstrate that unnatural amino acid mutagenesis can be used to determine distinct ligand binding interactions, even between molecules with similar structures. 3.4 Structure–Function Example Studies 67

3.4.2 Drug Interactions with the hERG Voltage-gated Potassium Ion Channel

In order to implement this approach and focus it on a practical problem of importance to the medical community and the pharmaceutical industry, we are determining the structural requirements for the interactions of various agents with the hERG potassium channel. hERG potassium ion channels govern the repolarization phase of human ventricular action potentials [44, 45]. Many drugs and other xenobiotic agents or their metabolites can inhibit hERG potassium channels and lead to cardiac arrhythmias and sudden death [46–49]. As a consequence, tremendous effort has gone toward creating novel drugs that do not possess hERG channel interactions. Avoiding unintended blockade of the hERG potassium ion channel is a costly and time-consuming challenge for the pharmaceutical industry and health care in general. Present discovery programs often rely upon brute force screening efforts that attempt to empirically eliminate hERG channel activity. These efforts commonly rely on assays using only the wild-type hERG channel.

From IC50 data in electrophysiology or radioligand binding/displacement studies medicinal chemists attempt to elucidate the hERG binding interactions and eliminate them from the molecule. However, we have discovered that there are multiple binding modes and without knowledge of the precise manner in which a given molecule binds to hERG, it may be difficult to synthesize a molecule that binds to the target but not to hERG [28]. We have employed natural and unnatural amino acid mutagenesis to reveal the structural determinants governing drug interactions with hERG. We believe with an understanding of how and where a compound is binding to hERG, medicinal chemists will then have the ability to design out hERG blockade for that particular compound. Figure 3.2 shows some of the natural and unnatural amino acids incorporated into hERG for that purpose. Previous studies have employed natural amino acid mutagenesis, e.g. alanine scanning to probe drug binding [50–52]. These experimental mutation studies demonstrated that the likely hERG binding region for most drugs is the cavity formed by the S6 helices under the selectivity filter [51, 53]. These studies identified two amino acids, Tyr652 and Phe656 puta- tively involved in drug binding, and the authors suggested the nature of interactions (cation–p, p–p and hydrophobic) based on the amino acid side chains involved [50–52]. Figure 3.4 shows homology models of hERG based on the bacterial channels KcsA (closed) and KvAP (open conformation) crystal structures [51, 54]. Studies have shown that drugs such as MK-499 bind within the large pore volume lined by the S6 helices that serves as the binding pocket for drugs that block hERG [51, 53]. From mutation data and models such as these, we are able to determine the amino acids likely to be involved in hERG block. Our work has focused on the roles of Thr623, Ser624, Tyr652, and Phe656, and the backbone carbonyl of Leu622. By substituting other amino acids at the positions 622, 623, 624, 652, and 656, we can probe in detail how these amino acids participate in the drug binding pocket. The amino acids that are substituted were chosen to provide the most information while not perturbing the overall structure or function of the protein. 68 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

Our mutational studies of more than 60 drugs that block hERG show that they all bind to two or more of these five key residues within the hERG channel [28]. Each of these amino acids projects a side chain functionality into the hERG pore to which drug molecules can bind. For example, Tyr652 can bind drugs through cation–p or p–p interactions with the aromatic ring, or through hydrogen bonding with the 4' hydroxy group [42]. By making subtle but informative mutations to Tyr652 we can determine how it is binding drugs. For example, if we mutate Tyr652 to Phe using site-directed mutagenesis, we eliminate the hydrogen bonding ability. If a drug’s binding is decreased by this mutation, we can conclude that hydrogen bonding is important to this drug’s binding to hERG. We can use similar mutations to test for cation–p or p–p interactions. We have made a number of subtle pore mutations that probe the steric, hydrophobic and electronic features that determine drug binding at amino acid positions Thr623, Ser624, Tyr652, and Phe656. We have also explored carbonyl backbone hydrogen bonding interactions by mutating the peptide amide bond at Leu622 to an ester linkage. We expressed the mutated channels in Xenopus oocytes and measured K+ currents by using two-electrode voltage-clamp methods while perfusing the oocytes with various known hERG blockers. By comparing the change in hERG currents between the WT channel and the mutant channels, we determined the energetics of the specific interactions that govern drug binding (see Eq. (1) above). We typically generate data for a total of 11 mutants at these 4 amino acid positions. This data set, which we call a Mutant Activity Panel (MAPTM) is then plotted as shown 3.4 Structure–Function Example Studies 69 in Fig. 3.5. Our data indicate that the channel binding interactions vary greatly, even for structurally related hERG blockers. Figure 3.5 shows the hERG MAP for astemizole. The figure shows the structure of the molecule and the positions of the amino acid mutations on the S6 helix of hERG. The bar graph shows the change in binding energy, relative to WT, for each of the 11 mutant channels. From the data we can see that mutation of Ser624 to Ala greatly reduces drug block. The Ser624Ala mutation eliminates the hydrogen bonding ability of serine, thus indicating that hydrogen bonding at this position is important for hERG block by astemizole. The progressive decrease in drug binding when we mutate Phe656 to 4-F-Phe, and then 3,5-F2-Phe indicates that astemizole makes a cation–p or p–p interaction with Phe656. We have generated the hERG MAPs of more than 60 pharmaceuticals using Xenopus oocytes. Consequently, we have developed an extensive database of compound structure–activity relationships (SAR). We have determined the binding geometries of a variety of compounds with the hERG channel. For example, ris- peridone and haloperidol showed cation–p interactions at Tyr652, whereas amperozide, fluphenazine, and triflupromazine showed cation–p interactions at

Figure 3.5 hERG MAP astemizole: A hERG noncovalent binding interaction with the MAP elucidates the nature and relative impor- channel. Each compound displays a unique tance of specific drug–channel interactions. hERG MAP signature. In this example, The right hand bar graph shows the change in changes at serine 624 suggest H-bond inter- astemizole binding energy (kcal mol–1)at actions with the compound. Progressive each position of the channel when that posi- changes in binding as fluorine (F) is added to tion is altered. These changes in binding en- phenylalanine at position 652 indicate cation– ergy may be interpreted in terms of atomic le- p or p–p aromatic interactions. The chemical vel interactions such as hydrogen-bonding, structure of astemizole is shown. (This figure cation–p, hydrophobic, and ion pairing. Each also appears with the color plates.) hERG mutant is designed to identify a specific 70 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

Phe656. Most of the drugs also exhibited hydrogen bonding interactions with the backbone of Leu622 and the side chain of Ser624. Figure 3.6 highlights the unique patterns of different interactions seen with different compounds. For example amperozide and pimozide rely on hydrogen bonding at Leu622 carbonyl (623- OH-Thr mutation), whereas haloperidol and astemizole do not. Only pimozide hydrogen bonds to the tyrosine hydroxy group, the other drugs do not. The substitution of cyclohexyl-alanine at position 656 (substitution with a nonaromatic ring) greatly affected amperozide, but not pimozide block. This information enables rational modification of these molecules. These insights cannot be obtained from

IC50 measurements against wild-type hERG only, as is now commonly done.

Fig. 3.6 Differentiation of binding by hERG blockers. The mutation at 623 affects H-bonding. Both pimozide and amperozide interact at this site, whereas block by haloperidol was unaffected. Conversion of a Tyr to Phe at position 652 removes an H-bonding site on the ring. Cyclohexyl-alanine is a nonaromatic ring substituted for Phe at position 656. This has no effect on pimozide block but greatly affects amperozide.

Not only is it possible to manipulate amino acid side channels, but also one of the more remarkable possibilities of this approach is the possibility to insert a-hydroxy acids in place of a-amino acids, replacing the amide peptide linkage with a backbone ester bond [34, 55]. The ribosome can apparently mediate ester formation and the nonsense suppression methodology is well suited to this tactic. In fact, in the in vivo nonsense suppression methodology we have often observed that a-hydroxy acids are more efficiently incorporated than analogous a-amino 3.4 Structure–Function Example Studies 71 acids [33]. This observation, coupled with the facts that the synthetic protocols associated with a-hydroxy acids are generally simpler [33, 56] and the resulting hy- droxyacyl tRNA is more stable than an aminoacyl tRNA, makes a-hydroxy acids very appealing for many types of nonsense suppression experiments. An a-hydroxy acid substitution can be placed at many sites in a protein without being disruptive. An a-hydroxy acid substitution makes the carbonyl a weaker hydrogen- bond acceptor. The weakening of the carbonyl as an acceptor was determined to have a larger effect (0.89 kcal mol–1) than the deletion of the NH (0.72 kcal mol–1) [34]. We have used a-hydroxy acids as a means to probe the role of H-bonds in drug binding in the hERG pore region. A backbone ester cannot form hydrogen bonds of the sort found in an a-helix. Figure 3.7 documents the distribution of compounds that utilize H-bonding at the 622/623 carbonyl. Some compounds hydrogen bond while others do not. Only with this type of information can meaningful changes in a molecule to eradicate hERG interactions be made.

Fig. 3.7 Role of peptide backbone carbonyl at Thr623 in drug– hERG interactions. Bars show relative change in binding affinity (DDG) for compounds. Positive ÄÄG indicates decreased affinity relative to WT hERG. A DDG of 1 kcal mol–1 corresponds to an

~5-fold decrease in IC50. Some compounds interact with this position strongly (e.g. pimozide) while others do not (e.g. astemizole). 72 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

We clustered these compounds into distinct classes based on their binding interactions with the hERG channel. We then made pharmacophore models for each cluster. Our data demonstrate that different molecules bind to the hERG channel in a variety of ways, and there is no universal hERG binding pharmacophore. We are using this knowledge about hERG–drug interactions to syntheti- cally eliminate the hERG binding ability of new therapeutic compounds.

3.5 Other Uses of Unnatural Amino Acids as Probes of Protein Structure and Function

A promising application of the nonsense suppression methodology is the site-specific incorporation of biophysical probes such as fluorescent groups or spin labels. Several groups have incorporated side chains with fluorescent reporter groups [7, 11, 57, 58]. Fluorescent groups can allow static and dynamic investigations of protein function in several ways. Chollet et al. [59, 60] incorporated an unnatural amino acid based on the nitrobenzoxadiazole (NBD) fluorophore into the neurokinin-2 receptor using the in vivo protocol and Xenopus oocyte expression. The neurokinin-2 receptor is a G-protein-coupled receptor that binds tachykinin. Fluorescence from the NBD group could be monitored in the membranes from ~10 oocytes [59]. Fluor- escence resonance energy transfer between the fluorescent agonist and the labelled receptor was used to obtain distance information relating the location of the tachykinin agonist-binding site to selected residues in the receptor [59, 60] Similar studies could be carried out to investigate ligand–binding site interactions of ion channels. Spin labels are another potentially useful probe that could be incorporated by nonsense suppression [7]. Another strategy is to use nonsense suppression to incorporate side chains that permit post-translational modifications, such as phosphorylation or glycosylation [11, 61–63]. This could provide a useful adjunct to existing methods for evaluating the significance of these important processes. Incorporation of amino acids with photoreactive side chains into proteins is an especially useful application of the nonsense suppression methodology in neurobiology [64]. Applications have involved ‘caged’ side chains where a heteroatom is protected. After incorporation of the unnatural amino acid into the protein, photolysis removes the protective group and reveals the previously caged functionality. Photo-decaging also offers the potential for time-resolved studies in which the photolysis is the triggering event. This aspect of using caged unnatural amino acids was demonstrated using a caged tyrosine incorporated into several locations at the agonist-binding site of the nAChR [31]. Caged tyrosine, serine and threonine, hold promise as tools for intracellular cell signaling research. These amino acids are targets for protein kinases. Phosphorylation of these amino acids plays a role in the regulation of a number of proteins and signaling pathways. Kinase dependent modulation of ion channels plays a role in synaptic plasticity associated with learning and memory. Caged tyrosine was used to study the phosphorylation of Kir2.1 (inward rectifier K+ channel) in Xenopus oocytes [65]. When tyrosine kinases were active, flash deca- 3.6 Conclusions 73 ging of a caged tyrosine at position 242 in Kir2.1 led to decreased K+ currents and 15–26% decrease in capacitance, implying net membrane endocytosis. It was discovered that decaging initiated two kinase dependent pathways, one involving direct modulation of the channel and the other involving endocytosis of the channel through a clathrin mediated mechanism [65]. This example illustrates how nonsense suppression can be used to deduce signaling pathways and provide new insights into pathways of biological signalling.

3.6 Conclusions

The nonsense suppression for unnatural amino acid incorporation into proteins is an important new tool for protein structure–function studies. Important scientific problems have been addressed, producing knowledge that is complementary to other methods and in some cases impossible to gain with conventional methods. The approach has been used to determine specific interactions governing drug–receptor interactions at the level of individual bonds and side chains [28]. It can be used to incorporate bioprobes to ascertain protein function in real time [5, 11]. It is possible to utilize a second suppressor tRNA to allow for incorporation of two different unnatural amino acids to evolve more sophisticated assays (e.g. FRET pairs) [66–69]. Much of the work discussed in this chapter was carried out using Xenopus oocytes which are a robust system industrialized at Neurion Pharmaceuti- cals. We are currently pursuing the routine incorporation of unnatural amino acids into proteins expressed in mammalian cells [27]. In addition, some groups are extending the nonsense suppression technique to use a four-base codon instead of the amber codon to incorporate unnatural amino acids [70–72]. In this chapter we have briefly discussed several methodological issues that must be considered including: mRNA read through, orthogonality of the suppressor tRNA, the problem of introducing large quantities of exogenous tRNA into the system. These issues have now been overcome in different expression systems including Xenopus oocytes, wheat germ lysate, and E coli [2, 73–75]. We have verified, using electrophysiology, a sensitive measure of protein function, that protein modified through nonsense suppression is functionally indistinguishable from protein expressed conventionally [2]. Such studies help validate unnatural amino acid mutagenesis as an important new tool for protein structure function studies. An existing challenge is the quantity of protein that can be generated and assayed. This is not a problem when studying ion channels by electrophysiology, but is a limitation in some other cases. The approach has been used to determined specific interactions governing drug–receptor interactions. It can be used to incorporate bioprobes to ascertain protein function in real time [5, 11]. 74 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

Acknowledgements

Much of the pioneering work that made some of these studies possible was done at Caltech in the groups of Prof. Henry Lester, Division of Biology and Prof. Den- nis Dougherty, Division of Chemistry. We gratefully acknowledge their experimental and intellectual contributions. We are also grateful for the experimental contributions of the Neurion scientists including Alisha Goodwin, David Paisner, Elisha Mackey, and Heinte Lesso.

References

1 C. J. Noren, S. J. Anthony-Cahill, amino acids. Methods Mol. Biol. 1998, 77, M. C. Griffith, P.G. Schultz, A gen- 325–354. eral-method for site-specific incorpora- 9 D. Mendel,V. W. Cornish, P.G. tion of unnatural amino acids into pro- Schultz, Site-directed mutagenesis with teins, Science 1989, 244, 182–188. an expanded genetic code, Annu. Rev. 2 M. W. Nowak, J. P. Gallivan, S. K. Biophys. Biomol. Struct. 1995, 24, 435– Silverman, C. G. Labarca, D. A. 462. Dougherty, H.A. Lester, In vivo incor- 10 L. Wang, P. G. Schultz, Expanding the poration of unnatural amino acids into genetic code, Angew. Chem. Int. Ed. Engl. ion channels in Xenopus oocyte expres- 2005, 44, 34–66. sion system, Methods Enzymol. 1998, 293, 11 T. A. Cropp, P.G. Schultz, An expand- 504–529. ing genetic code, Trends Genet. 2004, 20, 3 T. G. Heckler, L. H. Chang,Y. Zama, 625–630. T. Naka, M. S. Chorghade, 12 T. Hohsaka, M. Sisido, Incorporation of S. M Hecht, T4 RNA ligase mediated non-natural amino acids into proteins, preparation of novel “chemically misacy- Curr. Opin. Chem. Biol. 2002, 6, 809–815. lated” tRNAPheS, Biochemistry 1984, 23, 13 M. W. Nowak et al., Nicotinic receptor 1468–1473. binding site probed with unnatural 4 J. D. Bain, C. G. Glabe, T. A. Dix, A. R. amino acid incorporation in intact cells, Chamberlin, Biosynthetic site-specific Science 1995, 268, 439–442. incorporation of a non-natural amino 14 J. D. Bain, D. A Wacker, E. E Kuo, acid into a polypeptide, J. Am. Chem. Soc. R. A Chamberlin, Site-specific incor- 1989, 111, 8013–8014. poration of non-natural residues into 5 D. L. Beene, D. A. Dougherty, H. A. peptides: effect of residue structure on Lester, Unnatural amino acid mutagen- suppression and translation efficiencies, esis in mapping ion channel function, Tetrahedron 1991, 47, 2389–2400. Curr. Opin. Neurobiol. 2003, 13, 264–270. 15 J. D. Bain, E. S. Diala, C. G. Glabe, 6 D. A. Dougherty, Unnatural amino D. A. Wacker, M. H. Lyttle, T. A. Dix, acids as probes of protein structure and A. R. Chamberlin, Site-specific incor- function, Curr. Opin. Chem. Biol. 2000, 4, poration of nonnatural residues during 645–652. in vitro protein biosynthesis with semi- 7 V. W. Cornish, D. R. Benson, C. A. synthetic aminoacyl-tRNAs, Biochemistry Altenbach, K. Hideg,W. L. Hubbell, P. 1991, 30, 5411–5421. G Schultz, Site-specific incorporation of 16 S. A. Robertson, J. A. Ellman, P. G. biophysical probes into proteins, Proc. Schultz, A general and efficient route Natl. Acad. Sci. U S A 1994, 91, 2910– for chemical aminoacylation of transfer 2914. RNAs, J. Am. Chem. Soc. 1991, 113, 8 L. E. Steward, A. R. Chamberlin, Pro- 2722–2729. tein engineering with nonstandard 17 F. A. Laski, R. Belagaje, U. L. RajBhandary, P. A. Sharp, An amber References 75

suppressor tRNA gene derived by site- expressed in mammalian cells, Chem. specific mutagenesis: cloning and func- Biol. 2003, 10, 573–580. tion in mammalian cells, Proc. Natl. 28 M. W. Nowak et al., hERG Mutant Panel Acad. Sci. U S A 1982, 79, 5813–5817. for Lead Optimization of Compunds 18 C. Kohrer, L. Xie, S. Kellerer, U. with hERG Liability, Abstracts of Papers of Varshney, U. L. RajBhandary, Import the American Chemical Society 2005, 229. of amber and ochre suppressor tRNAs 29 L. M. Dedkova, N. E. Fahmi, S. Y. Golo- into mammalian cells: a general ap- vine, S. M. Hecht, Enhanced D-amino proach to site-specific insertion of amino acid incorporation into protein by modi- acid analogues into proteins, Proc. Natl. fied ribosomes, J. Am. Chem. Soc. 2003, Acad. Sci. U S A 2001, 98, 14310–14315. 125, 6616–6617. 19 M. E. Saks et al., An engineeredTetra- 30 J. P. Gallivan, H. A. Lester, D. A. hymena tRNAGln for in vivo incorpora- Dougherty, Site-specific incorporation tion of unnatural amino acids into proof biotinylated amino acids to identify teins by nonsense suppression, J. Biol. surface-exposed residues in integral Chem. 1996, 271, 23169–23175. membrane proteins, Chem. Biol. 1997, 4, 20 F. Hamano-Takaku et al., A mutant Es- 739–749. cherichia coli tyrosyl-tRNA synthetase 31 J. P. Miller, S. K. Silverman, P. M. utilizes the unnatural amino acid aza- England, D. A. Dougherty, tyrosine more efficiently than tyrosine, H. A. Lester, Flash decaging of tyrosine J. Biol. Chem. 2000, 275, 40324–40328. sidechains in an ion channel, Neuron 21 D. Kiga et al., An engineered Escherichia 1998, 20, 619–624. coli tyrosyl-tRNA synthetase for site-spe- 32 P. M. England, H. A. Lester, N. David- cific incorporation of an unnatural son, D. A. Dougherty, Site-specific, amino acid into proteins in eukaryotic photochemical proteolysis applied to ion translation and its application in a wheat channels in vivo, Proc. Natl. Acad. Sci. U germ cell-free system, Proc. Natl. Acad. SA1997, 94, 11025–11030. Sci. U S A 2002, 99, 9715–9720. 33 P. M. England, H. A. Lester, D. A. 22 D. R. Liu, P. G. Schultz, Progress to- Dougherty, Mapping disulfide connec- ward the evolution of an organism with tivity using backbone ester hydrolysis, an expanded genetic code, Proc. Natl. Biochemistry 1999, 38, 14409–14415. Acad. Sci. U S A 1999, 96, 4780–4785. 34 P. M. England,Y. Zhang, D. A. Dough- 23 S. W. Santoro, L. Wang, B. Herberich, erty, H. A. Lester, Backbone mutations D. S. King, P. G. Schultz, An efficient in transmembrane domains of a ligand- system for the evolution of aminoacyl- gated ion channel: implications of the tRNA synthetase specificity, Nat. Biotech- mechanism of gating, Cell 1999, 96, 89– nol. 2002, 20, 1044–1048. 98. 24 Z. Zhang, L. Alfonta, F. Tian, 35 L. Li ,W. Zhong, N. Zacharias, B. Bursulaya, S. Uryu, D. S. King, C. Gibbs, H. A. Lester, D. A. P. G. Schultz, Selective incorporation of Dougherty, The tethered agonist ap- 5-hydroxytryptophan into proteins in proach to mapping ion channel pro- mammalian cells, Proc. Natl. Acad. Sci. U teins–toward a structural model for the SA2004, 101, 8882–8887. agonist binding site of the nicotinic acet- 25 J. W. Chin, T. A. Cropp, S. Chu, ylcholine receptor, Chem. Biol. 2001, 8, E. Meggers, P. G. Schultz, Progress to- 47–58. ward an expanded eukaryotic genetic 36 A. L. Cashin, E. J. Petersson, H. A. code, Chem. Biol. 2003, 10, 511–519. Lester, D. A. Dougherty, Using physi- 26 H. A. Lester, Heterologous expression of cal chemistry to differentiate nicotinic excitability proteins: route to more speci- from cholinergic agonists at the nicotinic fic drugs? Science 1988, 241, 1057–1063. acetylcholine receptor, J. Am. Chem. Soc. 27 S. L. Monahan, H. A. Lester, D. A. 2005, 127, 350–356. Dougherty, Site-specific incorporation 37 J. T. Koh,V. W. Cornish, P. G. Schultz, of unnatural amino acids into receptors An experimental approach to evaluating 76 3 Unnatural Amino Acids as Probes of Ion Channel Structure – Function and Pharmacology

the role of backbone interactions in pro- 49 M. T. Keating, M. C. Sanguinetti, Mo- teins using unnatural amino acid muta- lecular and cellular mechanisms of car- genesis, Biochemistry 1997, 36, 11314– diac arrhythmias, Cell 2001, 104, 569– 11322. 580. 38 P. N. Kao, A. J. Dwork, R. R. J. Kaldany, 50 J. A. Sanchez-Chapula, R. A. Navarro- M. L. Silver, J. Wideman, S. Stein, Polanco, C. Culberson, J. Chen, M. C. A. Karlin, Identification of the a subunit Sanguinetti, Molecular Determinants half-cystine specifically labeled by an affi- of Voltage-dependent Human Ether-a- nity reagent for the acetylcholine recep- Go-Go Related Gene (HERG) K+ Chan- tor binding site, J. Biol. Chem. 1984, 259, nel Block, J. Biol. Chem. 2002, 277, 11662–11665. 23587–23595. 39 A. Karlin, Chemical modification of the 51 J. S. Mitcheson, J. Chen, M. Lin, active site of the acetylcholine receptor, J. C. Culberson, M. C. Sanguinetti, Gen. Physiol. 1969, 54, s245–s264. A structural basis for drug-induced long 40 H. R. Arias, Topology of ligand binding QT syndrome, Proc. Natl. Acad. Sci. U S sites on the nicotinic acetylcholine recep- A 2000, 97, 12329–12333. tor, Brain Res. Brain Res. Rev. 1997, 25, 52 D. Fernandez, A. Ghanta, G. W. Kauff- 133–191. man, M. C. Sanguinetti, Physicochem- 41 W. Zhong, J. Gallivan,Y. Zhang, L. Li, ical features of the HERG channel drug H. Lester, D. A. Dougherty, From ab binding site, J. Biol. Chem. 2004, 279, initio quantum mechanics to molecular 10120–10127. neurobiology: A cation-p binding site in 53 M. C. Sanguinetti, J. S. Mitcheson, the nicotinic receptor, Proc. Natl. Acad. Predicting drug-hERG channel interac- Sci. USA 1998, 95, 12088–12093. tions that cause acquired long QT syn- 42 D. A.Dougherty, Cation–p interactions drome,Trends Pharmacol. Sci. 2005, 26, in chemistry and biology: A new view of 119–124. benzene, Phe, Tyr, and Trp, Science 1996, 54 R. A. Pearlstein et al., Characterization 271, 163–168. of HERG potassium channel inhibition 43 N. Zacharias, D. A. Dougherty, using CoMSiA 3D QSAR and homology Cation–p interactions in ligand recogni- modeling approaches, Bioorg. Med. tion and catalysis, Trends Pharmacol. Sci. Chem. Lett. 2003, 13, 1829–1835. 2002, 23, 281–287. 55 T. Lu, A.Y. Ting, J. Mainland, L. Y. Jan, 44 J. Warmke, B. Ganetzky, A Family of Po- P. G.Schultz, J. Yang, Probing ion per- tassium Channel Genes Related to eag in meation and gating in a K+ channel with Drosophila and Mammals, PNAS 1994, backbone mutations in the selectivity 91, 3438–3442. filter, Nat. Neurosci. 2001, 4, 239–246. 45 M. C. Sanguinetti, C. Jiang, M. E. 56 P. M. England, H. A. Lester, D. A. Curran, M.T. Keating, A mechanistic Dougherty, Incorporation of esters into link between an inherited and an ac- proteins: Improved synthesis of hydro- quired cardiac arrhythmia: HERG en- xyacyl tRNAs, Tetrahedron Lett. 1999, 40,

codes the IKr potassium channel, Cell 6189–6192. 1995, 81, 299–307. 57 L. E. Steward, C. S. Collins, M. A. 46 R. R. Fenichel et al., Drug-induced tor- Gillmore, J. E. Carlson, J. B. A. Ross, sades de pointes and implications for A. R. Chamberlain, In vitro site-specific drug development, J. Cardiovasc. Electro- incorporation of fluorescent probes in physiol. 2004, 15, 475–495. b-galactosidase, J. Am. Chem. Soc. 1997, 47 J. I. Vanderberg, B. D. Walker,T. J. 119, 6–11. Campell, HERG K+ channels: friend 58 A. V. Karginov, M. Lodder, S. M. and foe, TIPS 2001, 22, 240–246. Hecht, Facile characterization of trans- 48 J. S. Mitcheson, M. D. Perry, Molecular lation initiation via nonsense codon sup- determinants of high-affinity drug bind- pression, Nucleic Acids Res. 1999, 27, ing to HERG channels, Curr. Opin. Drug 3283–3290. Discov. Dev. 2003, 6, 667–674. References 77

59 G. Turcatti et al., Probing the structure lian cells, Mol. Cell Biol. 1986, 6, 3059– and function of the tachykinin neuroki- 3067. nin-2 receptor through biosynthetic in- 68 C. Kohrer, J. H. Yoo, M. Bennett, corporation of fluorescent amino acids at J. Schaack , U. L. RajBhandary, A Pos- specific sites, J. Biol. Chem. 1996, 271, sible Approach to Site-Specific Insertion 19991–19998. of Two Different Unnatural Amino Acids 60 G. Turcatti, K. Nemeth, M. D. Edger- into Proteins in Mammalian Cells via ton, J. Knowles, H. Vogel, A. Chollet, Nonsense Suppression, Chem. Biol. 2003, Fluorescent labeling of NK2 receptor at 10, 1095–1102. specific sites in vivo and fluorescence en- 69 C. Kohrer, E. L. Sullivan, U. L. ergy transfer analysis of NK2 ligand-re- RajBhandary, Complete set of orthogo- ceptor complexes, Receptors Channels nal 21st aminoacyl-tRNA synthetase-am- 1997, 5, 201–207. ber, ochre and opal suppressor tRNA 61 N. E. Fahmi, S. Golovine, B. Wang , pairs: concomitant suppression of three S. M. Hecht, Studies toward the site spe- different termination codons in an cific incorporation of sugars into pro- mRNA in mammalian cells, Nucleic Acids teins: synthesis of glycosylated aminoa- Res. 2004, 32, 6200–6211. cyl-tRNAs, Carbohydr. Res. 2001, 330, 70 T. Hohsaka, M. Sisido, Incorporation of 149–164. nonnatural amino acids into proteins by 62 D. M. Rothman, E. J. Petersson, M. E. using five-base codon-anticodon pairs, Vazquez, G. S. Brandt, D. A. Dough- Nucleic Acids Symp. Ser. 2000, 99–100. erty, B. Imperiali, Caged phosphopro- 71 M. Sisido, Extension of biochemical teins, J. Am. Chem. Soc. 2005, 127, 846– functions by the introduction of nonna- 847. tural amino acids and artificial nucleic 63 Z. Zhang et al., A new strategy for the acid analogs, Nucleic Acids Symp. Ser. synthesis of glycoproteins, Science 2004, 2000, 275–276. 303, 371–373. 72 J. C. Anderson, N. Wu, S. W. Santoro, 64 E. J. Petersson, G. S. Brandt, N. M. V. Lakshman, D. S. King, P. G. Zacharias, D. A. Dougherty, H. A. Schultz, An expanded genetic code with Lester, Caging proteins through unna- a functional quadruplet codon, Proc. tural amino acid mutagenesis, Methods Natl. Acad. Sci. U S A 2004, 101, 7566– Enzymol. 2003, 360, 258–273. 7571. 65 Y. Tong et al., Tyrosine decaging leads to 73 J. Ellman, D. Mendel, S. Anthony- substantial membrane trafficking during Cahill, C. J. Noren, P.G. Schultz, modulation of an inward rectifier potas- Biosynthetic method for introducing un- sium channel, J. Gen. Physiol. 2001, 117, natural amino acids site-specifically into 103–118. proteins, Methods Enzymol. 1991, 202, 66 J. P. Capone, P. A. Sharp, U. L. 301–336. RajBhandary, Amber, ochre and opal 74 D. R. Liu, T. J. Magliery, M. Pastrnak, suppressor tRNA genes derived from a P. G. Schultz, Engineering a tRNA and human serine tRNA gene, Embo J. 1985, aminoacyl-tRNA synthetase for the site- 4, 213–221. specific incorporation of unnatural 67 J. P. Capone, J. M. Sedivy, P. A. Sharp, amino acids into proteins in vivo, Proc. U. L. RajBhandary, Introduction of Natl. Acad. Sci. U S A 1997, 94, 10092– UAG, UAA, and UGA nonsense muta- 10097. tions at a specific site in the Escherichia 75 D. R. Liu, T. J. Magliery, P.G. Schultz, coli chloramphenicol acetyltransferase Characterization of an ‘orthogonal’ sup- gene: use in measurement of amber, pressor tRNA derived from E. coli ochre, and opal suppression in mamma- tRNA2(Gln), Chem. Biol. 1997, 4, 685– 691. 79

4 Functional Expression of Ion Channels in Mammalian Systems Jeff J. Clare

4.1 Introduction

The expression of recombinant ion channels in heterologous systems is an invaluable tool for detailed analysis of their molecular and biophysical properties. This powerful approach compliments the analysis of endogenous channels in their native tissues by providing an opportunity to study and manipulate them in a defined and electrically inactive cell environment. This is advantageous for a wide range of studies, particularly for detailed definition of biophysical and pharmacological properties, for analysing structure–function relationships, and for characterising the role of individual subtypes and/or component subunits. An expression system that has been extensively used in this way is based on oocytes of the South African clawed frog, Xenopus laevis. This system has the advantage of being easy to manipulate while giving robust expression and stable recordings for a broad range of recombinant ion channels in an environment that is largely free of endogenous electrical activity (see Chapter 1). However, despite the utility of Xeno- pus oocytes for ion channel functional analysis, certain disadvantages are inherent. The maintenance of frogs and preparation of oocytes requires additional procedures and equipment beyond that usually found in a standard research laboratory. Oocyte quality is critical for success and this can be subject to wide variation due to seasonal and other factors. More importantly, mammalian channels can exhibit biophysical properties that are somewhat unexpected [1, 2], possibly due to an absence of regulatory, accessory and other factors that normally interact with the channels in their native environment. Similarly, pharmacological properties can be altered in oocytes. Typically, lower drug potencies are observed and this has been ascribed to a reduction of the free intracellular concentration of drug due to absorption by the large amounts of membrane and yolk particles found inside the oocyte [3]. Many of these drawbacks can be avoided, at least partially, by the use of mammalian cells as an expression host for studying recombinant mammalian ion channels. The aim of this chapter is therefore to review approaches that are currently available for heterologous ion channel expression in mammalian systems. For re-

search applications, expression of the channel in a functional form is normally a prerequisite. Similarly, although alternative formats exist (e. g. radioligand binding, chapter 7), the most reliable read-outs for ion channel drug screening are those based, either directly or indirectly, on functional activity of the channel in a cell-based assay. However, this can often present significant challenges given the huge diversity and organisational complexity found within the ion channel superfamily. Ion channel subunits generally contain multiple transmembrane spanning segments and functional channels are usually complex multi-subunit proteins. Correct membrane localisation and orientation, as well as appropriate subunit composition and stoichiometry, are often vital for faithful reproduction of the pharmacological and biophysical properties of the native channel. In addition, ion channel cDNAs can be difficult to manipulate due to high instability and frequently cause cytotoxicity to the host cell when over-expressed. Finally, even for well expressed channels, the generation of reagents that are sufficiently robust for use in a high throughput screening environment can be extremely problematic. In this chapter a number of strategies that can be used to address these challenges will be described. A number of case studies will be presented that highlight some of the theoretical and practical considerations for obtaining robust expression of functionally active channels.

4.2 cDNA Cloning and Manipulation

cDNA instability can be a fairly common phenomenon with certain ion channels, which can make cloning and manipulation using standard techniques difficult. This is presumably due to the presence of sequences that are either toxic or re- combinogenic in E.coli (e. g. direct repeats, Z DNA, etc.) causing negative selection pressure and resulting in a strong orientation bias and/or few viable clones. Clones that do arise are frequently found to contain mutations ranging from large rearrangements and deletions to single base changes that cause mis-sense or frame shift mutations. Certain channel subfamilies, e.g. voltage-gated sodium

(NaV) channels and ATP-binding cassette (ABC) transporters such as CFTR, sulfonylurea receptors, seem to be particularly susceptible to this, but the problem can also occur more sporadically in other subfamilies, e.g. voltage-gated calcium chan-

nels (CaV). These problems can be minimised using a variety of approaches, either separately or, if necessary, in combination. The use of “directional” cloning strategies (i.e. where insertion of the DNA fragment in the required orientation is forced by the use of incompatible restriction sites) is recommended wherever possible. Vectors that are “transcriptionally silent” (i.e. containing no E.coli promoter sequences, cyptic or otherwise, in the vicinity of the cloning site) and/or that replicate at low copy number in E.coli (e.g. pBR322-based) are also useful. In addition, E.coli host strains are commercially available that allow propagation of standard ColE1-based vectors at reduced copy number (ABLE, Stratagene). The use of commercially available recombination-deficient E.coli host strains (SURE, Stratagene, 4.3 Choice of Host Cell Background 81

STABL2, Invitrogen) is also recommended. Finally, propagation of E.coli clones at reduced temperature (e.g. 30 8C) is often found essential for problematic channels – the colonies may take several days to appear and will generally be much smaller than normal. Despite taking some or all of the above precautions, for channels that are known to be unstable, it is essential to rigorously validate final constructs by complete DNA sequencing and it frequently proves necessary to correct mutations by one or more rounds of site-directed mutagenesis. In extreme cases, it may also be necessary to resequence the channel-encoding region each time a new batch of DNA is prepared. Finally, for optimal expression, it is common to engineer the region upstream of the start codon to contain sequences compatible with efficient initiation of translation (Fig. 4.1). The consensus sequence associated with genes that are efficiently expressed in mammalian cells, identified by Kozak [4], is often used, and this can normally be inserted during construction of the expression vector.

4.3 Choice of Host Cell Background

As indicated in the introduction to this chapter, the choice of host cell background is often critical for efficient expression as well as for faithful reproduction of the properties of the native channel. Two major concerns are the existence of endogenous channels in the host cell that may be similar to the channel of interest and mask its expression, and the presence or absence of accessory factors that may normally modify the properties or expression of the channel in its native environment. Clearly, these parameters will vary between different cell types and are dependent on the nature of the channel of interest. Thus, it is advisable to screen a range of potential cell lines for their suitability as hosts before embarking on an expression project. This can be done by functional screening, preferably using the assay format that will ultimately be used for analysing the channel of interest, or alternatively by molecular screening, for example by quantitative mRNA analysis (e.g. Taqman). The latter approach is often more convenient and is useful as a primary screen in that the endogenous expression of wide range of channels and accessory proteins can be screened simultaneously (e.g. using reverse Taqman). However, this requires a reasonable level of background knowledge of the channel family and its likely modulatory or interacting partners, which may not always be available, especially for more novel channels. In addition, the detection of mRNA for a particular channel or accessory factor does not necessarily mean that the active protein is expressed, and functional assays are usually necessary as a secondary screen to establish this. Although pre-screening of potential host cell lines may indicate a lack of con- founding background activities, it is worth remembering that it is formally possible for normally silent endogenous channel proteins to become up-regulated during the process of generating an expression reagent, perhaps in response to over expression of the target channel. Thus, it is prudent to validate expression of the 82 4 Functional Expression of Ion Channels in Mammalian Systems

Fig. 4.1 Quantitative comparison of different cells measured by IonWorks high throughput expression constructs and screening of stable electrophysiology. (B) Screening of stable clones using automated high throughput clones for functional hERG channel expres- electrophysiology. (A) Comparison of repre- sion. 24 different clonal CHO cell lines gener- sentative hERG K+ channel expressing CHO ated with construct C (above), were analysed cell lines that were generated using different by Ionworks electrophysiology. The graphs types of expression construct. A and D = non- show a comparison, for each clone, of the per- optimised mRNA leader sequence, B = opti- centage of cells giving usable seals (>80 MO, mised leader, C = optimised leader in IRES ex- bottom), percentage cells expressing currents pression vector. Each point represents peak above background (middle) and mean current current (left panel) or percentage of cells ex- amplitude (top). Data are averaged from 64 pressing (right panel) averaged from ~300 cells for each clone. 4.3 Choice of Host Cell Background 83 target protein in the final reagent by a sequence dependent method, for example by immunological validation, if selective antibodies are available, or by a quantitative PCR (Taqman). In a drug discovery setting another factor that should be considered when choosing a host background is the physical characteristics of the cells, which must be compatible with the intended assay format. Thus, electrophysiological assays depend on the formation of high resistance electrical seals with the plasma membrane, and microtitre plate-based assays (see Chapters 7 and 8) often require good adherence of cells to the plastic substrate. It is worth stating that these qualities can deteriorate in cells over-expressing ion channels compared to untransfected parent cells, especially over prolonged passage. A variant of the HEK293 cell line that has been engineered to be more adherent to plastic surfaces by expressing the macrophage scavenger receptor is available (Invitrogen). A variety of mammalian cells lines have been successfully used as hosts for recombinant ion channel expression. Standard hosts that are in general use for heterologous expression have also been successfully used for ion channels, including CHO, COS7, CV1, HEK293 etc. As indicated above, variants of these lines which are designed for specific applications have also been engineered and their use will be highlighted throughout this chapter. These commonly used hosts are relatively (though not completely) free of background conductances and are suitable for many channel types (e.g. see Refs. [5, 6]). However, it is worth noting that different isolates can vary in this respect and may behave differently in different laboratories, so it is still advisable to functionally screen for background currents of interest. In situations where there is relatively little background knowledge of the channel and its potential accessory factors, or where expression proves problematic in “standard hosts”, it is often worth searching for less common immor- talised lines that are derived from tissues that natively express the channel of interest. For example, various neuroblastoma cell lines have proved useful as hosts for neuronal ion channel expression, such as SH-SY5Y, ND7–32 , NG108 [7–9]. In practise, the precise reason(s) for this have not been fully explored but it is likely that the intracellular environment and processing machinery are more favorable in a host cell derived from the tissue where the channel is normally expressed. Other factors that can be important probably include the presence of cognate accessory proteins and better tolerance to a wider range of resting membrane potential than nonexcitable cells. An example of this is provided by the sen- + sory neuron specific NaV1.8 Na channel. This is only poorly expressed in commonly used host cells (e.g. HEK293) but is more efficiently expressed in an im- mortalised dorsal root ganglion-derived cell line (ND7–23, Fig. 4.2a). One possible explanation is that this relates to the accessory b3 subunit since this is endogenously expressed in ND7–23 cells but, not in HEK293 (Fig. 4.2c). In support of this, it was found that the recombinant b3 subunit enhances the level of

NaV1.8 specific currents when co-expressed in HEK293 cells (Fig. 4.2b). Further- more, this effect is also seen in ND7–23 cells, suggesting that the effect of endogenous b3 in these cells can be enhanced by over-expression of the recombinant subunit. 84 4 Functional Expression of Ion Channels in Mammalian Systems

Fig. 4.2 Effect of cell background and auxili- Bars represent the mean peak currents ob-

ary subunits on efficiency of ion channel ex- tained. (C) NaVb3 subunit mRNA is endogen- pression. (A) Comparison showing represen- ously expressed in ND7–23 but not in tative current traces recorded from HEK293 HEK293 cells. RT-PCR analysis of cell extracts and ND7–23 cell lines stably expressing the using subtype selective primers is shown. + NaV1.8 channel. Peak Na currents in Lanes labelled “-” indicate –ve controls where HEK293 stable clones were smaller compared reverse transcriptase was omitted; +ve con- to ND7–23 and declined during passage or trol reactions, using the cloned b subunit following freeze–thawing of cells. (B) Transi- cDNAs as template, are shown in the right

ent co-expression of the NaVb3 subunit en- hand lanes. Reproduced with permission hances NaV1.8 expression in HEK293 cells. from Ref. [89]. 4.4 Post-translational Processing of Heterologous Expressed Ion Channels 85

As illustrated above, it is possible to engineer host cell lines to make them more favorable for ion channel expression by over-expressing accessory factors that may be required for efficient expression. It may also be desirable to generate cell lines that express reporter genes (e.g. Aequorin [10], halide-sensitive YFP [11]) for use as generic hosts for ion channel assays. An additional approach to engineering cell lines to make them more favorable as hosts is to manipulate their electrical properties by introducing additional channels. This is particularly useful when generating reagents for microtitre plate-based functional assays in which the readout is dependent on voltage-sensitive dyes, but can also apply to assays based on Ca2+ binding fluorescent dyes. A good example of this is shown in Fig. 4.3, which illustrates the expression of the R-type voltage-gated calcium channel in non-neuronal cells (HEK293). In these cells only small Ca2+ responses can be detected in fluorescence imaging plate reader (FLIPR, Molecular Devices) assays when they are depolarised, even though good sized currents can be detected electrophysiolo- gically. This is apparently due to the resting potential of HEK293 cells which is more positive than that typically found in neurons (e. g. –10 to –20 mV compared with –60 to –80 mV) resulting in the majority of channels being inactivated. How- ever, it is possible to increase the magnitude of depolarisation-induced Ca2+ responses by manipulating the resting potential by co-expressing a constitutively active K+ channel, KCNK2 (TREK-1). This leads to a more negative resting potential, resulting in more Ca2+ channels being available to open when the cells are depolarised. This dual expression system can also be used in reverse to measure activity of KCNK2 in FLIPR assays using the Ca2+ channel as a reporter.

4.4 Post-translational Processing of Heterologous Expressed Ion Channels

Since, by nature, they are complex multimeric transmembrane proteins, ion channels present particular challenges for efficient heterologous expression. In order to obtain functional activity a number of post-translational steps are required to ensure that channel proteins are correctly folded, processed, assembled and trans- ported to the appropriate membrane compartment [for review see Ref. [12]]. Since, each of these steps requires interaction of the heterologous channel protein with homologous host factors, there is potential for incorrect processing at every stage. In addition, abnormally high levels of expression can potentially overwhelm the capacity of the endogenous processing machinery, again leading to unpro- cessed or aberrant channel protein. Similarly, for any given channel, a particular processing step may require specific factors that might not be endogenous to the cell type being used (see below for examples), again leading to aberrant, nonfunc- tional protein. Processing “bottlenecks” such as these can result in the aberrant channel protein either accumulating and then aggregating within the cell [13], or alternatively being targeted for degradation [14, 15]. In extreme cases both of these outcomes can be difficult to diagnose, since the former can lead to cytotoxicity and low viability (see discussion below), and the latter can appear as a total failure 86 4 Functional Expression of Ion Channels in Mammalian Systems 4.4 Post-translational Processing of Heterologous Expressed Ion Channels 87

of expression unless this is being monitored by measuring mRNA levels. In less extreme cases a low level of the mature functional protein may be detected, but the amount produced is likely to be sensitive to both the level and rate at which expression is being driven. Paradoxically, attempting to improve yields in this situation by increasing expression, for example with the use of stronger promoters, may actually be counterproductive [16] since this could exacerbate blockage of the processing pathway. Instead, the use of titratable (e.g. viral) and/or inducible systems is recommended in order to try to optimise the balance between the total protein expressed and the proportion that is correctly processed and functional (see Sections 4.6.1 and 4.7.3). An alternative approach for increasing the proportion of mature channel protein is to augment the level of processing factors that may be present only in limiting amounts or absent altogether. As discussed previously, an empirical way of achieving this is to identify and use an alternative host cell type that is more related to the tissue in which the native channel is expressed. However, another approach is to overproduce factors that are likely to be, or are known to be, involved in channel maturation by co-expression of their cloned cDNA. Proteins with this activity broadly fall into two categories: those that are thought to play a general role in quality control and promoting protein folding and transport within most cells (e. g. chaperones), and those with more specialised trafficking roles for particular types of ion channel. Chaperones act by binding to nascent proteins in order to promote correct folding and targeting. It is known that proteins with large regions of exposed hydrophilic regions are toxic to cells, possibly by promoting membrane damage [17], and indeed this is believed to contribute to the pathogenic mechanism of certain “viral” proteins (e.g. prions [14, 15]). It is likely that the production of large amounts of misfolded recombinant membrane proteins can also cause cytotoxicity by similar mechanisms. Thus, co-expression of factors with chaperone-like activity should be beneficial, not only by increasing the proportion of mature channel protein produced, but also by counteracting this toxic effect. Examples of chaperones that have been shown to boost the surface expres-

sion of various recombinant ion channels (e. g. Kv, nAChR, HERG, CFTR, NaV) include calnexin, heat shock proteins (e. g. Hsp70, Hsp90) and fibroblast growth factor homologous factor 2B (FHF2) [18–23]. However, as well as their beneficial

3 Fig. 4.3 Manipulation of the host cell back- tion of KCl) of cells stably expressing the

ground to optimise functional expression of R-type CaV channel (CaV2.3 and b3 subunits). ion channels. (A) Co-expression of the B) Similar functional responses can be ob- + TREK1 K channel is required in order to hy- served when all three subunits (CaV2.2, b3and perpolarise the resting potential of HEK293 TREK1) are delivered using BacMam. Little or 2+ cells so that functional R-type CaV channels no Ca influx is seen in the absence of TREK1 can be detected in a FLIPR-based fluorescence virus. (C) The magnitude of Ca2+ influx assay. Only when the TREK1 K+ channel is co- (change in fluorescence) can be manipulated expressed, in this case using a BacMam viral by varying the amount of TREK1 virus added. vector, can robust Ca2+ influx (measured by an (X, no TREK1 virus added; ~,56106 plaque increase in fluorescence of the Ca2+ binding forming units (pfu) ml–1 of TREK1 virus dye Fluo4) be induced by depolarisation (addi- added, &,107 pfu ml–1; ^,26107 pfu ml–1.) 88 4 Functional Expression of Ion Channels in Mammalian Systems

effects, it must be remembered that, in some circumstances, binding of chaperones can actually target proteins for degradation [22]. In addition to chaperones, certain cytoskeletal scaffold proteins which interact with channel proteins at the cell surface can also promote surface localisation of the channel [24]. As well as proteins that have general chaperone-like activity within the cell, a number of channel-specific maturation factors have been identified. Examples include the PACS proteins, annexin II light chain, and KChap which promote sur-

face localisation of Trp, Kv and NaV1.8 channels respectively [25–27]. In addition, numerous other channel-associated proteins are known that have dual functions, both acting as channel-specific modulatory subunits as well as enhancing channel

trafficking. Examples include the b subunits of Kv, NaV (see Fig. 4.2B) and CaV channels [28–30], a2d subunits of CaV channels [31], calmodulin [32, 33] and stargazin [34] (which promote trafficking of IK/SK and AMPA channels respectively). Some of these are thought to act by masking specific ER retention motifs present in the channel protein. There is evidence for these retention signals in a wide range of channels, including Kir, SK/IK, Kv, Eag, HERG, NMDA, kainate, nAChR [35–44], and it is thought that these motifs are exposed in the monomeric subunit and become inaccessible upon assembly of the mature multimeric channel (see for example Ref. [44]). Thus, for optimal expression of any given channel, it is worth co-expressing any known channel-specific accessory subunits or interacting factors, particularly in cases where known ER retention signals are present. Another strategy that can be used to increase the amount of functional channel protein found at the cell surface is to lower the growth temperature at which the cells are maintained. This has been shown to give higher levels of surface expression for a number of different channel types. For several of these examples (e.g.

CFTR, P2X7, HERG, NaV1.5 [45–49]) the effect was first noted with variants of the channel that are defective for trafficking. Incubation of cells expressing these mutants at reduced temperature (e.g. 25–30 8C) was found to rescue the defect. This appears to be a general phenomenon since it has also been observed with nonmu- tated channels (e.g. ROMK1, nAChR’s, KAT1 [50–55], and HERG – see Fig. 4.4 and 6.2). There are several possible explanations for this effect, for instance, lower temperatures could result in increased steady-state levels of transcript or synthesis of channel protein, reduced protein turnover, improved protein folding, assembly or transport. The best studied example is probably the CFTRD508 mutant that is the most prevalent cause of cystic fibrosis. Recent studies indicate that the deleted residue normally has a major contribution to the proper folding of the channel [56, 57]. Thus, at physiological temperature (37 8C) folding of the mutant is impaired and the protein is targeted for degradation, probably due to binding of the Hsp70/CHIP chaperone complex [22]. However, at lower temperature (30 8C) folding kinetics are more favorable, protein degradation is greatly reduced, and the mutant protein can then be detected at the cell surface. A similar mechanism probably accounts for the increased surface expression of nonmutant channels at lower temperature. Consistent with this, reduced protein degradation has been reported in several cases (ROMK1, a4b2nAChR [50, 53] and HERG – see Fig. 4.4b). Evidence suggests that degradation can occur via mechanisms that are both proxi- 4.4 Post-translational Processing of Heterologous Expressed Ion Channels 89

Fig. 4.4 Effect of reduced cell culture tempera- (see Fig. 6.2). Note that, in the HERG-expres- ture on ion channel expression. (A) Confocal sing cells, at 37 8C and 30 8C immunoreactivity images of stable HERG-expressing CHO cells is uniform and granular throughout the cyto- grown at different temperatures, following im- plasm whereas at 27 8C there is an accumula- munocytochemical staining with a HERG-spe- tion in giant lysosome-like bodies. (B) Quanti- cific antibody. The level of HERG immunoreac- tative analysis of HERG immunoreactivity by tivity is significantly increased in cells grown at flow cytometry confirms the increases seen at 27 8C and 30 8C compared to 37 8C. No immu- lower growth temperatures, as indicated by noreactivity is seen in untransfected CHO cells the rightward shift in peak fluorescence (X- (CHO-wt). The increase in total HERG protein axis) observed within the populations of cells at 30 8C seen here is also mirrored by an in- grown at 27 8C and 30 8C compared to 37 8C. crease in surface-localised functional protein (This figure also appears with the color as measured by IonWorks electrophysiology plates.) 90 4 Functional Expression of Ion Channels in Mammalian Systems

mal (proteosomes) and distal (endocytic recycling) to the secretory pathway ([58], Fig. 4.4A). Thus, maintenance of cell cultures at lower temperatures would be expected to improve functional expression of any channel that is prone to misfold- ing when over-expressed.

4.5 Cytotoxicity

Many ion channels are poorly tolerated when over-expressed in mammalian cells, leading to poor growth and viability, and low levels of expression which tend to de- cline during cell passage. Usually this becomes evident during stable cell line generation when only a low number of viable clones can be selected, with few of those that do arise giving measurable levels of expression. Among these clones only a low proportion of cells actually express and there are large cell-to-cell variations. This is a hallmark of toxicity induced by heterologous protein expression, but often the situation can be at least partially rescued by the use of appropriate expression vectors (e.g. IRES vectors, see Section 4.7.1). Other strategies to address this problem include the use of transient or inducible expression systems (see Sec- tions 4.6 and 4.7.3). However, it may also be possible to improve the situation by attempting to tackle the root cause(s) of the problem. As discussed above, one possible reason for toxicity is saturation of the host processing machinery – this may lead to deficiencies in the processing of essential host proteins, with consequent toxic effects, as well as the accumulation of large amounts of misfolded protein which may have membrane damaging effects. In this case co-expression of chaperones or channel accessory subunits may be beneficial. Another possible mechanism for toxicity is that functional activity of the channel itself has cytotoxic consequences for the host cell. Channels that are permeable to calcium may be particularly susceptible to this (e.g. NMDA-NR2a and NR2b [59]), since calcium signalling regulates many fundamental processes within the cell and is normally tightly regulated. Clearly, over-expression of large conductance calcium-permeable channels, or those that normally have a role in calcium homeostasis, is more likely to cause this type of problem. A potential solution to this problem is to use known channel blockers during reagent generation (e.g. during selection of stable clones) and also during routine maintenance of the recombinant cells [60]. In the light of this it is worth bearing in mind that aminoglycoside antibiotics, which are often used for selection during stable cell line generation (e.g. neomycin, geneti- cin), have been found to have inhibitory activity at several types of ion channel

(e.g. TRPV1, CaV, NMDA [61–63]). This may be beneficial during cell line generation for toxic channels. On the other hand, it has also been found the aminoglycoside antibiotics can also potentiate channel activity (e.g. NMDA-NR2b [63]), which may actually exacerbate toxicity. 4.6 Transient Expression Systems 91

4.6 Transient Expression Systems

Transient expression systems are commonly used in ion channel studies for a wide variety of applications. They are particularly valuable where rapid and/or high throughput analysis is required, for example in alanine scanning, cysteine scanning, site-directed or random mutagenesis studies and for the evaluation of expression constructs, channel variants and subunit combinations. Transient expression provides a flexible way of titrating the level of expression, controlling stoichiometry of multi-subunit channels and introducing reporters or interacting factors that may be required. In a drug screening environment, where many different ion channel assays may be run in parallel, transient expression can be more convenient than stable expression since this reduces cell culture demands by removing the need to continuously maintain multiple cell lines in culture. As discussed already, transient systems are also useful where toxicity is known to be a problem. Several alternative approaches for transient expression are available and each has different advantages and disadvantages.

4.6.1 ‘‘Standard’’ Transient Expression

“Standard” transient expression approaches usually involve the treatment of cells with chemical agents that promote the uptake and internalisation of naked DNA. Different types of agent are available, and these work with varying efficiencies on different cell types [64]. Among the most commonly used for ion channels are calcium phosphate [65] and various cationic lipid reagents which are available from commercial suppliers (e.g. Lipofectamine, Invitrogen). Although the latter come with detailed protocols, there are particular considerations for ion channels. For example, in electrophysiology experiments an appropriate balance is needed between the transfection rate (i. e. proportion of cells expressing) and the health of the cells following transfection. Thus, in order to achieve the best results for any given channel it may be worth optimising the transfection conditions (amount and time of reagent exposure, recovery time etc.). An alternative methodology to chemical reagents sometimes used for transient expression of ion channels is electroporation [66], though this tends to be less efficient. This is because the magnitude of the electrical pulse that can be given to promote DNA uptake is limited by deleterious effects on cell viability and once again a balance between transfection efficiency and cell health has to be struck. The efficiency of transient transfection and gene expression is highly dependent on the cell type and expression vector used. HEK293 and COS cells are among the more competent for DNA uptake and are commonly used for ion channel expression, though other cell types can also be used (e.g. CHO). Variants of HEK293 are available which enable plasmid replication when using vectors containing the origin of replication from SV40 due to the expression of SV40 T antigen (HEK293T, ATCC) [67, 68]. This “episomal” vector system should give enhanced expression 92 4 Functional Expression of Ion Channels in Mammalian Systems

due to the increased copy number of the channel cDNA and possibly due to other activities of T-antigen. Another episomal vector system uses an analogous trans- acting DNA replication factor (EBNA1) from Epstein Barr virus (EBV) which enables replication of vectors containing the replication origin from EBV (oriP) [69]. Expression may also be boosted via a reported EBNA1-dependent enhancer like activity of oriP [70]. Different variations of this system have been developed where EBNA1 is encoded either by the expression vector (e.g. pCEP4, Invitrogen) or by the host cells (HEK293E, Invitrogen). The former has the advantage of being compatible with any host cell type, whereas the latter may be more convenient since a smaller vector can be used which may aid insertion of the channel encoding cDNA. Streamlined oriP-containing episomal vectors of this type have been developed that contain optimised promoters and that have been shown to give very high levels of expression [71].

4.6.2 Viral Expression Systems

Standard transient transfection is the most rapid means of mammalian cell expression. However, the efficiency is highly dependent on the host cell background and many cell types that are of interest for ion channel expression can be relatively resistant to transfection methodologies (e.g. neuroblastoma cell lines). A potential solution to this problem is to make use of virus-based systems for gene delivery, and two examples that have been successfully applied to the heterologous expression of ion channels and membrane-bound receptors in mammalian cells will be discussed here. The first system is derived from the single-stranded RNA virus, Semliki Forest Virus (SFV), and the second is based on the insect cell baculovirus, Autographa California nuclear polyhedrosis virus (AcMNPV). The SFV expression system comprises two plasmid vectors which together contain a cDNA copy of the viral genome [72]. The expression vector contains the non- structural genes, the strong SFV 26S promoter for expression of the heterologous gene and a viral packaging signal. The helper vector encodes the structural proteins, including the capsid and envelope genes, but no packaging signal. A stock of recombinant viral particles is produced by first generating single-stranded helper and expression vector RNAs using in vitro transcription and then co-transfecting these into baby hamster kidney (BHK) cells. These particles can then be used to efficiently infect a broad range of mammalian cell lines (including BHK, COS, HEK293), as well as primary cells, and are capable of giving rapid, high-level expression. However, no virus progeny is produced because the recombinant particles contain no helper RNA and are therefore replication deficient. This system has been used extensively to express G-protein coupled receptors (GPCRs) [73] often giving good expression levels and correctly folded protein, as determined by ligand binding assays and in some cases by coupling to G-protein signalling pathways. The potential for ion channel expression has also been explored and several channels have been expressed, including 5HT3, P2X1, 2 and 4 [74], GABAa [75] and Kv [76]. Functional activity was demonstrated for 5HT3, P2X receptors and GABAa. 4.6 Transient Expression Systems 93

Recombinant baculoviruses have been routinely used for efficient protein expression in insect cells for many years. However, although baculovirus virions are able to enter mammalian cells, no viral expression can normally be detected. Re- cently baculovirus-based “BacMam” vectors have been engineered to allow expression in mammalian cells by insertion of promoters that are active in mammalian hosts (for a review see Ref. [77]). The gene of interest is inserted into a transfer vector downstream of the mammalian promoter (e.g. CMV) in which, similar to the baculovirus system, it is flanked by regions that promote insertion into the baculovirus genome, either by homologous recombination in insect cells or by trans- poson-mediated recombination in E.coli cells (see Fig. 4.5). A stock of virus particles is generated by propagation of the recombinant virus genome in insect cells and this can then be used to deliver the gene for transient expression in a range of different mammalian cell types, including many of those commonly used for recombinant protein production. The BacMam transfer vector commonly used also contains a selectable marker conferring resistance to the antibiotic G418, thus allowing the additional option of selecting stable cell lines from virus-transduced cells [78]. The BacMam system works particularly well with certain cell types including HEK293, and the human osteosarcoma derived cell line, U2OS in which transduction and expression rates are high [79, 80]. It has been used with good success for membrane protein expression, including a large range of GPCRs and ion channels [81, 82]. In common with other transient systems it offers advantages for expression of toxic channels and for multi-subunit and multi-channel expression [81]. For example, the NMDA-NR2b glutamate receptor, which requires expression of both the NR2 b and NR1 subunits for functional activity, is extremely toxic when expressed in non-neuronal mammalian cells [59]. In the author’s laboratory, repeated attempts to generate NR2b-expressing stable cell lines were unsuccessful, even using IRES vectors (see Section 4.7.1) and even in the presence of a channel blocker (ketamine). However, by generating separate BacMam viruses for each subunit, and transducing a mixture of these into HEK293 cells, robust glutamate-induced calcium influx can be observed that is not present in mock-transduced cells (Fig. 4.6). The magnitude of the response is proportional to the total amount of virus added and the stoichiometry of the two subunits can be readily investigated and optimised by varying the dose of virus for each subunit. An example of multi-channel expression using BacMam viruses is shown in Fig. 4.3. In addition to viruses for two different subunits of the R-type calcium channel + (CaV2.3, a2d), a virus encoding the TREK-1 K channel was also introduced, in order to hyperpolarise the resting potential of the host HEK293 cells and thus prevent inactivation of the R-type channel. In this way, robust calcium influx can be measured when the TREK-1 virus is added, but not in its absence, and the magnitude of the response can be varied by altering the amount of TREK-1 virus added. In summary, the BacMam system is a versatile tool for the robust analysis of many ion channels and is particularly suitable for toxic or multi-subunit channels. It allows flexible control over subunit or channel stoichiometry simply by varying the dose of the corresponding viral constructs used for transduction. It is also a 94 4 Functional Expression of Ion Channels in Mammalian Systems

Fig. 4.5 The BacMam expression system. ome when transfected into the appropriate (A) Map of the pFastBacMam1 shuttle vector E.coli host. (B) Workflow for the generation of which used to generate recombinant Bac- recombinant BacMam virus stock and trans- Mam viruses. The gene of interest is inserted duction into mammalian cells for expression. downstream of the CMV promoter where it is Reproduced with permission from Ref. [77]. flanked by Tn7 inverted repeats that direct (This figure also appears with the color site-specific transposition into the viral gen- plates.) 4.6 Transient Expression Systems 95

Fig. 4.6 Expression of a cytotoxic channel NR1A and NR2b subunits, respectively. The using the BacMam system. (A) Functional ex- arrow indicates the point of glutamate addi- pression of the NMDA-NR2b receptor using tion. (B) The magnitude of the peak response BacMam. Glutamate-induced Ca2+ influx, as is dependent on the total amount of virus measured by increased fluorescence of the added, as measured by MOI (multiplicity of Ca2+ binding dye Fluo4, can be observed in infection – the number of virus particles HEK293 cells transduced with a 1:5 mixture added per cell). of BacMam viruses expressing the NMDA- convenient means of introducing additional factors into existing ion channel stable cell lines, for example chaperones or reporter genes that might be required for use in alternative assay formats (e.g. aequorin, halide sensitive YFP). Little or no cytopathic effects of virus transduction are observed and cells can normally be used for electrophysiology assays the day after transduction. BacMam viruses are also safe since they are unable to replicate in mammalian cells and the budded form of the virus is noninfectious for the natural insect host [77]. 96 4 Functional Expression of Ion Channels in Mammalian Systems

4.7 Stable Expression of Ion Channels

For many applications stable expression of the target ion channel may be desirable, for example for detailed biophysical studies or extended compound screening campaigns. Unlike transient systems, this avoids the need to continually carry out transfections or transductions, and there is no requirement for repeated re-synthesis of vector DNA or virus reagents. In theory, stable expression would also be expected to be more reproducible than transient expression. Nevertheless, a certain degree of day-to-day and cell-to-cell variation of expression in stable cell lines is li- able to occur and this can be considerable for some ion channels. In practise, the worst stable cell lines can be more variable than some transient systems, e.g. Bac- Mam viruses,which are capable of giving perfectly acceptable reproducibility.

4.7.1 Bicistronic Expression Systems

The generation of robust ion channel expressing stable cell lines can be problematic. As discussed already, since many ion channels are poorly tolerated when over-expressed, difficulties can arise due to poor growth and loss of expression during passage, even when selection is maintained. Often this is accompanied by large cell-to-cell variability in expression level even within clonally isolated lines, which can be a particular problem for single-cell format electrophysiology assays. A related problem during cell line generation is that, following transfection and selection, only a low proportion of clones actually express measurable levels of channel activity. This situation can be improved by using vectors where expression of the channel is closely coupled to that of the selectable marker. The best way of achieving this is to use a viral internal ribosome entry site (IRES) which promotes translation of both proteins from a single bicistronic mRNA [83]. The IRES element is placed between the two coding regions, with the channel upstream and the selection marker immediately downstream (Fig. 4.7). Translation of the upstream channel is initiated, as normal, by signals in the 5' untranslated region, whereas translation of the downstream selection marker is initiated by binding of ribosomes to the IRES element. Thus, both proteins are produced from a single transcript with the advantage that expression of the two is very tightly linked. This is unlike “standard” vectors where the selection marker and gene of interest are in separate expression cassettes and it is not uncommon for expression of the two to become separated, either during clonal selection or subsequent maintenance of the cell line, giving rise to the problems described above. This is especially true with toxic genes where the growth disadvantage conferred on expressing cells is an additional selection pressure for this to occur. Thus, an additional benefit with IRES vectors is that expression should remain relatively stable over extended periods as long as selection is maintained (Fig. 4.8). The advantages of IRES vectors for stable cell line generation become particularly apparent when using high-throughput electrophysiology assays. Automated 4.7 Stable Expression of Ion Channels 97

Fig. 4.7 Bicistronic expression vectors for ing generation and propagation of the cells. stable ion channel expression. (A) Using an (B) Plasmid map of a typical bicistronic ex- IRES element the expression of the ion chan- pression vector, pCIN5L, that can be used to nel of interest is co-translationally linked to generate stable cell lines. The ion channel the selection marker (neoR), thus avoiding cDNA is inserted upstream of the IRES ele- loss of expression and instability problems ment between the Not I and Nhe I restriction that can occur with conventional expression sites. (IVS, intervening sequence or intron). vectors due to independent segregation dur- patch-clamp instruments are planar array instruments (see Chapter 6), and place great demands on the quality of the cell line being used since they are single cell assays in which the cells are randomly selected (though this problem is now being addressed by the latest generation instruments, e.g. IonWorks Quattro). Thus, a very high proportion of cells in the population that express currents of usable size is essential in order to avoid low success rates and unacceptably high costs resulting from compound wastage and reduced throughput. On the other hand, these instruments provide invaluable tools for functional screening during cell line generation. They enable a vast increase in the number of clones that can be screened compared to that previously possible using conventional patch-clamp (Fig. 4.2B). Using a combination of IRES vectors and screening by high throughput automated electrophysiological screening, very high quality stable cell lines can readily be obtained with a variety of different ion channels, which typically express peak currents in the nA range in greater than 90–95% of the cell population (Fig. 4.9). High throughput patch-clamp instruments also add further precision to the 98 4 Functional Expression of Ion Channels in Mammalian Systems

Fig. 4.8 Quantitative analysis of multiple sub- pressors comprise >80% of the cell popula- unit expression using automated high tion. (B) Stability of expression over passage throughput electrophysiology. (A) Population showing percentage of cells expressing (trian- histogram of current amplitudes in individual gles) and mean peak current in expressing cells (n=3180) co-expressing KCNQ1 and the cells (circles). Each point represents the accessory subunit minK. The fitted line is a 2 mean of >250 cells. Kinetic analysis of the cur- peak gaussian fit indicating the presence of rents (see Fig. 6.5) indicates that virtually two populations – nonexpressors (currents every cell expresses the minK subunit as well <0.5 A, mean = 0.33 nA), and expressors as KCNQ1. (currents >0.5 A, mean = 1.05 nA). The ex- 4.7 Stable Expression of Ion Channels 99

Fig. 4.9 Detailed evaluation of ion channel sentative inward Na+ current evoked by depo- stable cells lines using automated high larisation to –10mV, showing blockade by te- throughput electrophysiology. HEK293 cells trodotoxin (TTX, 300 nM). Mean peak cur- were transfected with an expression vector for rents (B) and distribution of peak current (C) human NaV1.7 and multiple stable clones measured from 270 cells are shown. (D) were selected, expanded and then screened Greater than 80% of cells gave usable seals for expression of Na+ currents using Ion- (>80 MO) and expressed currents above Works. The best expressing clone was identi- background. fied and then characterised further. (A) Repre- screening and functional validation of ion channel reagents by providing the capability for statistically analysing expression in large numbers of individual cells. This is highly advantageous for monitoring the often very subtle effects of auxiliary subunits or for carrying out detailed comparative studies, for example, of different expression constructs; vectors, host cells etc. (see Fig. 4.1, 4.8, 4.9 and 6.5). IRES-containing vectors carrying the CMV promoter to drive transcription of the bicistronic mRNA and the IRES element derived from Encephalomyocarditis virus (EMCV) are commercially available (BD Biosciences Clontech). Versions with different variants of the EMCV IRES are also available. In the original pCIN vector 100 4 Functional Expression of Ion Channels in Mammalian Systems

(and similar vectors, e.g. pCIN3 [84]) the efficiency of expression of the downstream neomycin-resistant selection marker is relatively low since the distance between the neomycin phosphotransferase initiation codon and the IRES element is nonoptimal (i.e. 35bp less than is found between the IRES element and the major site of initiation of the viral polyprotein in the native EMCV virus) [85]. This is likely to be advantageous for many applications because a relatively low efficiency of translation of neomycin phosphotransferase should favor the selection of clones expressing high levels of the bicistronic mRNA, therefore giving relatively high levels of the upstream ion channel gene also. However, for reasons already discussed, this is not necessarily optimal for functional expression of ion channels and an alternative vector, pCIN5, that gives lower expression is also available [85, BD Biosciences Clontech]. In this vector the initiation codon of neomycin phosphotransferase has been engineered to be in precisely the same position with respect to the IRES as in the downstream ORF of the native EMCV virus. This is likely to give more efficient expression of the selectable marker resulting in lower levels of bicistronic mRNA and correspondingly lower levels of channel expression from the upstream cistron. The pCIN5 vector has been used successfully for a variety of different channels (see Fig. 4.2, 4.4, 4.8, and 4.9 and also Refs. [7, 83–88]).

4.7.2 Stable Expression of Multiple Subunits

As well as those with different IRES variants, vectors with different selection markers are available [85, BD Biosciences Clontech] and have been used successfully for a variety of multi-subunit and multi-channel examples [87, Fig. 4.8]. It is possible to successfully co-express up to three subunits or channels using these different vectors with different selection markers for each. However, most cell types grow relatively poorly under such extreme selection pressure in the presence of three antibiotics. Thus, for channels containing more than three subunits, and also where channel expression itself has deleterious effects on host cells, a different strategy is sometimes required. One possibility is to link two of the required subunits on the same expression vector, either as separate expression cassettes or via an IRES element. As previously discussed, a concern with using two separate expression cassettes on the same vector is that expression of the two subunits could become segregated during cell line selection or maintenance. Similarly, a concern with linking the two subunits via an IRES element is that, unless arranged in a tricistronic configuration (see below), the selection marker must be expressed as a separate expression cassette, again with the risk that it can become separated from expression of the channel subunits. Another potential problem with the latter approach is that, because expression of cDNAs downstream of the IRES element is considerably lower than that of cDNAs positioned upstream, the stoichiometry of subunits expressed may be inappropriate. Nevertheless, this system has been used successfully for functional expression of the P2X2/3 channel [89]. Another application where this arrangement has been successful is a variation where the downstream cistron is a reporter that can be used for fluorescence- 4.7 Stable Expression of Ion Channels 101 activated cell sorting (FACS), e.g. green fluorescent protein (GFP). In this case, GFP fluorescence can be used as a more convenient surrogate for monitoring expression of the upstream gene of interest and for selecting a subpopulation of transfected cells that are enriched for expression [90]. Obtaining an acceptable stoichiometry of subunits is also a consideration when using tricistronic expression constructs. Since this arrangement involves the expression of three different cDNAs (e.g. two channel subunits plus selection marker) linked using two IRES elements, the expression of the cDNA at the 3'end will be very low indeed. Stability of expression with tricistronic constructs is also a potential issue due to intrinsic instability of such lengthy transcripts, especially for larger channel subunits e.g. NaV or CaV a subunits. In addition, unless IRES elements from different sources (i. e. having different sequences) are used, there is also the potential for deletions and rearrangements to occur by homologous recombination. Despite these caveats there are reported examples of successful expression using tricistronic constructs [85, 89, 91]. Another interesting strategy for multi-subunit channels that may be applicable to some channel types is the expression of tandemly concatenated subunits. This approach has been shown to be feasible in a number of instances [92] and is particularly useful for investigating or fixing the stoichiometry and assembly of complex channels, since covalently linking subunits predetermines their composition and arrangement in the mature protein.

4.7.3 Inducible Expression

While the use of IRES vectors has proven effective for the selection of stable cell lines for a wide variety of ion channel types, sometimes expression of the channel can be so cytotoxic that the growth rate of antibiotic resistant clones is prohibi- tively slow or, in extreme cases, no clones actually survive antibiotic selection. As reviewed above, one option is to use transient expression to address this problem but, if stable expression is required, an alternative strategy is to use inducible expression. As with transient expression, inducible systems offer the possibility of regulating the level of channel expression, in this case by titration of the inducer. Several inducible different systems have been described, but probably those most commonly used for ion channels make use of the bacterial tetracycline resistance operon. Alternative versions are available depending on whether regulation occurs via a derepression or transactivation mechanism. In the original system, expression of the gene of interest is driven by a mini CMV promoter containing a tet-repressor binding site [93, BD Biosciences Clontech]. This is activated by a fusion protein which is a hybrid between the tet repressor and the activation domain of the VP16 protein from Herpes Simplex virus. When tetracycline is added this hybrid tet/VP16 repressor protein no longer binds the promoter and no expression of the gene of interest can occur. This “Tet-off” system has been used for several ion channels including NMDA-NR2a [94], NaV1.7 and NaV1.8 [95]. A“Tet-on” variation of this has been developed which uses a mutant tet repressor/VP16 hybrid 102 4 Functional Expression of Ion Channels in Mammalian Systems

activator that binds to the promoter only when tetracycline is present [96]. An ion channel example using this system is NR2 b [97]. The other tetracycline regulated system (“T-REx”, Invitrogen) uses a wild-type tet repressor protein (i. e. not fused to VP16) which normally binds to tet operator sequences within the CMV promoter silencing expression of the downstream gene of interest. When tetracycline is added the repressor protein no longer binds and expression of the gene of interest is derepressed [98]. A potential disadvantage of this derepression approach, compared to transactivation, is that production of the tet repressor needs to be very efficient in order to effectively silence the powerful CMV promoter. Thus, basal levels of expression can be relatively high, which may be a problem for highly toxic channels. Nevertheless, this system has been successfully used for several ion

channels including TrpM2, CaV3.2 and Kv2.1 [99–101]. On the other hand the level of expression when fully induced is higher with the derepression system since this uses a stronger CMV promoter which does not rely on the activity of a hybrid viral transactivator. Other types of inducible system that have been used for ion channels involve the use of transactivating steroid hormone receptors. One of these is based on the insect hormone, ecdysone, and uses a heteromeric receptor that is functional in mammalian cells [102]. This receptor is a heteromer between the mammalian reti- noid X receptor (RXR) and the insect ecdysone receptor (EcR) that binds to an engineered promoter containing specific recognition sequences and tightly represses

Fig. 4.10 Generation and analysis of stable non-inducing (filled bars) and inducing condi- clones with inducible channel expression. Ex- tions (mifepristone added, open bars) is pression screening of 16 HERG-GeneSwitch shown. Clones 9 and 16 showed substantial clones by flow cytometry using a HERG-speci- increases in HERG immunoreactivity follow- fic antibody. Clones were generated by trans- ing induction, though the induced level of ex- fection of a pGENE-HERG vector into pression was lower than that seen in 4 repre- HEK293 cells expressing the GeneSwitch reg- sentative clones generated with an expression ulatory protein (pSwitch) and selection with vector that uses the CMV promoter (pCIN5, zeocin. For each clone expression in both lanes shown at far left). Acknowledgements 103 expression of the downstream gene. In the presence of synthetic ecdysone analogue inducers (muristerone A or ponasterone A) the conformation of the receptor is altered and transactivation of the downstream gene then occurs. A similar system, “GeneSwitch” (Invitrogen), uses a receptor that is a hybrid between the DNA binding domain from the yeast Gal4 transactivator, the ligand binding domain from the human progesterone receptor and the activation domain from the NFkb transcription factor (p65AD). This binds to an engineered GAL4-adenovirus E1b promoter upstream of the gene of interest. In the absence of the inducer, mifepristone, the hybrid GeneSwitch receptor represses the promoter but when mifepristone is added it binds to the receptor and induces a conformational change that causes activation of expression. An example of an ion channel expressed using this system is shown in Fig. 4.10. Both the ecdysone and the GeneSwitch systems should give relatively low basal expression levels since the receptors are able to repress promoter activity in the absence of inducer, so may be suitable for the expression of highly toxic channels. For example, the ecdysone system has been used to express the cytotoxic NMDA-NR2a and NR2b receptors [103, 104]. However, the fully induced level of expression may not be as high as with vectors that use the CMV promoter (see Fig. 4.10).

4.8 Summary

In conclusion, this chapter has attempted to highlight some of the many potential difficulties that functional expression of heterologous ion channels in mammalian cells can present. These challenges can be magnified even further within a drug discovery setting, due to the stringent demands for robust, highly reproducible expression over a long term and on a large scale. Nevertheless, as has been outlined, a variety of approaches to address these problems are available, each with their own set of advantages and disadvantages. The choice of expression system will ultimately be governed by the nature of the channel of interest and by the final assay format that is required. However, it is often not possible to predict with any certainty how successful a particular system will be and in many cases, particularly for more novel channels, a combination of approaches may be necessary. Thus, it is often worth exploring several different strategies in parallel in order to develop a final expression reagent that is most appropriate for the particular channel-assay combination(s) required.

Acknowledgements

The author would like to acknowledge numerous colleagues at GSK who have given support and input, particularly Mark Chen, Yuhua Chen, Pat Condreay, Dave Grose, Bruce Hamilton, Andy Powell, Andy Randall, Steve Rees, Mike Romanos, Derek Trezise and Simon Tate. 104 4 Functional Expression of Ion Channels in Mammalian Systems

References

1 G. Seebohm, C. Lerche, A. E Busch, tyrosine kinase receptors, Receptors A. Bachmann, Dependence of I(Ks) bio- Channels 2002, 8, 319–30. physical properties on the expression sys- 11 L. J. V. Galietta, M. F. Springsteel, tem, Pflugers Arch. 2001, 442, 891–895. E. Masahiro, E. J. Niedzinski, K. By, 2 A. Shcherbatko, F. Ono, G. Mandel, M. J. Haddadin, M. J. Kurth, M. H. P. Brehm,Voltage-dependent sodium Nantz, A. S. Verkman, Novel CFTR channel function is regulated through Chloride Channel Activators Identified membrane mechanics, Biophys. J. 1999, by Screening of Combinatorial Libraries 77, 1945–1959. Based on Flavone and Benzoquino.lizi- 3 H. Witchel, J. T. Milnes, J. S. Mitche- nium Lead Compounds, J. Biol. Chem. son, J. C. Hancox, Troubleshooting pro- 2001, 276, 19723–19728. blems with in vitro screening of drugs 12 C. Deutsch, The birth of a channel, Neu- for QT interval prolongation usung ron 2003, 40, 265–276. HERG K+ channels expressed in mam- 13 J. Johnston, C. L. Ward, R. R. Kopito, malian cell lines and Xenopus oocytes, J. Aggresomes: A cellular response to Pharmacol. Toxicol. Methods 2002, 48, 65– misfolded proteins, J. Cell Biol. 1999, 80. 143, 1883–1898. 4 M. Kozak, The scanning model for trans- 14 C. M. Dobson, Protein folding and mis- lation: an update, J. Cell. Biol. 1989, 108, folding, Nature 2003, 426, 884–890 229–241. 15 A. L. Goldberg, Protein degradation and 5 P. H. Lalik, D. S. Krafte,W. A. Volberg, protection against misfolded or damaged R. B. Ciccarelli, Characterisation of en- proteins, Nature 2003, 426, 895–899. dogenous sodium channel gene ex- 16 M. K. Higgins, M. Demir, C. G. Tate, pressed in Chinese hamster ovary cells, Calnexin co-expression and the use of Am. J. Physiol. 1993, 264, C803–C809. weaker promoters increase expression of 6 G. J. Stephens, K. M. Page, J. R. Burley, correctly assembled Shaker potassium N. S. Berrow, A. C. Dolphin, Func- channel in insect cells, Biochim. Biophys. tional expression of rat brain cloned a1e Acta 2003, 1610, 124–132. calcium channels in COS7 cells, Pflugers 17 J. I. Kourie, C. L. Henry, Ion channel Arch. 1997, 433, 523–532. formation and membrane linked pathol- 7 V. H. John et al., Heterologus expression ogies of misfolded hydrophobic proteins:

and functional analysis of rat NaV1.8 the role of dangerous unchaperoned mo- (SNS) Voltage-gated sodium channels in lecules, Clin. Exp. Pharmacol. Physiol. the dorsal root ganglion neuroblastoma 2002, 29, 741–753. cell line ND7–23, Neuropharmacol 2004, 18 L. N. Manganas, J. S Trimmer, Calnexin 46, 425–438. regulates mammalian Kv1 channel traf- 8 J. Chemin, A. Monteil, S. Dubel, ficking, Biochem. Biophys. Res. Commun. J. Nargeot, P. Lory, The alpha1I T-type 2004, 322, 577–584. calcium channel exhibits faster gating 19 W. Chang, M. S Gelman, J. M Prives, properties when overexpressed in neuro- Calnexin-dependent Enhancement of blastoma/glioma NG 108–15 cells, Eur. Nicotinic Acetylcholine Receptor Assem- J. Neurosci. 2001, 14, 1678–86. bly and Surface Expression, J. Biol. 9 T. Ljungstrom, M. Grunnet, B. S. Jen- Chem. 1997, 46, 28925–28932. sen, S. P. Olesen, Functional coupling 20 S. H. Keller, J. Lindstrom, P. Taylor, between heterologously expressed dopa- Involvement of the chaperone protein mine D(2) receptors and KCNQ chan- calnexin and the acetylocholine receptor nels, Pflugers Arch. 2003, 446, 684–94. beta-subunit in the assembly and cell sur- 10 V. J. Dupriez, K. Maes, E. Le Poul, face expression of the receptor, J. Biol. E. Burgeon, M. Detheux, Aequorin- Chem. 1996, 37, 22871–22877. based functional assays for G-protein- 21 E. Ficker, A. T Dennis, L. Wang, A. M coupled receptors, ion channels, and Brown, Role of the Cytosolic Chaper- ones Hsp70 and Hsp90 in Maturation of References 105

the Cardiac Potassium Channel hERG, 33 W-S. Lee, T. J Ngo-Anh, A. Bruening- Circ. Res. AHA 2003, 92, e87–e100. Wright, A. Maylie, J. P. Adelman, 22 M. D. Amaral, CFTR and Chaperones, Small Conductance Ca2-activated K+ J. Mol. Neurosci. 2003, 23, 0895–8696. Channels and Calmodulin, J. Biol. Chem. 23 E. K. Wittmack, A. M Rush, M. J 2003, 278(28), 25940–25946. Craner, M. Goldfarb, S. G Waxman, 34 W. Vandenberghe, R. A. Nicoll, D. S. S. D Dib-Hajj, Fibroblast growth factor Bredt, Stargazin is an AMPA receptor homologous factor 2B: association with auxiliary subunit, PNAS. 2005, 102(2),

NaV1.6 and selective colocalization at 485–490. nodes of Ranvier of dorsal root axons, 35 D. Ma, N. Zerangue,Y-F. Lin, J. Neurosci. 2004, 30, 6765–6775. A. Collins, M. Yu,Y. N. Jan, L. Y. Jan, 24 C. Altier et al., Trafficking of L-type Role of ER Export Signals in Controlling Calcium Channels Mediated by the Post- Surface Potassium Channel Numbers, synaptic Scaffolding Protein AKAP79, Science 2001, 291. J. Biol. Chem. 2002, 277(37), 33598– 36 M. M. Zarei, M. Eghbali, A. Alioua, 33603. M. Song, H. G. Knaus, E. Stefani, 25 M. Köttgen et al., Trafficking of TRPP2 L. Toro, An endoplasmic reticulum traf- by PACS proteins represents a novel ficking signal prevents surface expres- mechanism of ion channel regulation, sion of a Voltage- and Ca2(+)- activated EMBO J. 2005, 24(4), 705–716. K+ channel splice variant, Proc. Natl. 26 K. Okuse, M. Malik-Hall, M. D Baker, Acad. Sci. USA 2004, 101(27), 10072– W.Y Poon, H. Kong, M. V. Chao, J. N. 10077. Wood, Annexin II light chain regulates 37 R. Roncarti, I. Decimo, G Fumagalli, sensory neuron-specific sodium channel Assembly and trafficking of human expression, Nature 2002, 6, 417(6889), small conductance Ca2(+)- activated K+ 653–656. channel SK3 are governed by different 27 Y. A. Kuryshev,T. I Gudz, A. M Brown, molecular domains, Mol. Cell. Neurosci. B. A Wible, KChAP as a chaperone for 2005, 28(2), 314–325. specific K (+) channels, Am. J. Physiol. 38 H. M. Jones, K. L. Hamilton, G. D. Pap- Cell. Physiol. 2000, 278(5), C863–C864. worth, C. A. Syme, S. C. Watkins, N. A. 28 O. T. Pongs et al., Functional and mole- Bradbury, D. C. Devor, Role of the NH2 cular aspects of voltage-gated K + chan- terminus in the assembly and trafficking nel beta subunits, Ann. N.Y. Acad. Sci. of the intermediate conductance Ca2(+)- 1999, 868, 344–355. activated K+ channel hIK1, J. Biol. Chem. 29 L. L. Isom, Sodium channel beta subu- 2004, 279(15),15531–15540. nits: anything but auxiliary, Neuroscientist 39 D. Li, K. Takimoto, E. S. Levitan, Sur- 2001, 1, 42–54. face expression of Kv1 channels is gov- 30 M. W. Richards, A. J. Butcher, A. C. erned by a C-terminal motif, J. Biol. Dolphin, Ca2+ channel ß-subunits: Chem. 2000, 275(16), 11597–11602. structural insights AID our understand- 40 M. Jenke, A. Sánchez, F. Monje, ing, Trends Pharm. Sci. 2005, 12, 0165– W. Stühmer, R. M. Weseloh, L. A. 6147. Pardo, C-terminal domains implicated 31 N. Klugbauer, E. Marais, F. Hofmann, in the functional surface expression of Calcium channel alpha2delta subi- potassium channels, EMBO 2003, 22(3), nuts: differential expression, function, 395–403. and drug binding, J. Bioenerg. Biomembr. 41 A. Akhavan, R. Atanasiu, A. Shrier, 2003, 35(6), 639–647. Identification of a COOH-terminal seg- 32 W. J. Joiner, R. Khanna, L. C. Schlich- ment involved maturation and stability ter, L. K. Kaczmarek, Calmodulin Regu- of human ether-a-go-go-related gene po- lates Assembly and Trafficking of SK4/ tassium channels, J. Biol. Chem. 2003, IK1 Ca2+-activated K+ Channels, J. Biol. 278(41), 40105–40112. Chem. 2001, 276(41), 37980–37985. 42 D. B. Scott, I. Michailidis,Y. Mu, D. Logothetis, M. D. Ehlers, Endocyto- 106 4 Functional Expression of Ion Channels in Mammalian Systems

sis and Degradative Sorting of NDMA lowered temperature of cell-surface a7- Receptors by Conserved Membrane-Prox- nicotinic acetylcholine receptor expres- imal Signals, J. Neurosci. 2004, 24(32), sion on transfected SH-EP1 cells, J. Neu- 706–7109. rochem. 2003, 85, 581–591. 43 Z. Ren, N. J. Riley, E. P. Garcia, J. M. 52 M. E. F. Nelson, F. Wang, A. Kuryatov, Sanders, G. T. Swanson, J. Marshall, C. H Choi,V. Gerzanich, J. Lindstrom, Multiple Trafficking Signals Regulate Functional Properties of Human Nicoti- Kainate Receptor KA2 Subunit Surface nic AChRs Expressed by IMR-32 Neuro- Expression, J. Neurosci. 2003, 23(16), blastoma Cells Resemble Thise of a3â4 6608–6616. AChRs Expressed in Permanently Trans- 44 J-M.Wang, L. Zhang,Y. Yao, N. Viroon- fected HEK Cells, J. Gen. Physiol. 2001, chatapan, E. Rothe, Z-Z. Wang, A 118, 563–582. transmembrane motif governs the sur- 53 S. T. Cooper, P. C. Harkness, E. R. face trafficking of nicotinic acetylcholine Baker, N. S. Millar, Up-regulation of receptors, Nature 2002, 5, No. 10 Cell-surface a4b2 Neuronal Nicotinic Re- 10.1038/nn198. ceptors by Lower Temperature and Ex- 45 G. M. Denning, M. P. Anderson, J. F. pression of Chimeric Subunits, J. Biol. Amara, J. Marshall, A. E. Smith, M. J. Chem. 1999, 17, 27145–27152. Welsh, Processing of mutant cystic fi- 54 M. E. Nelson, A. Kuryatov, C. H. Choi, brosis transmembrane conductance reg- Y. Zhou, J. Lindstrom, Alternate Stoi- ulator is temperature-sensitive, Nature chiometries of a4b2 Nicotinic Acetylcho- 1992, 27,358(6389), 761–764. line Receptors, Mol. Pharm. 2003, 63(2), 46 L. C. Denlinger et al., Mutation of a Di- 2035–1038020. basic Amino Acid Motif Within the C Ter- 55 I. Szabò, A. Negro, P. M. Downey, minus if the P2X7 Nucleotide Receptor M. Zoratti, F. I. Schiavo, G. M. Giaco- Results in Trafficking Defects and Im- metti, Temperature-Dependent Func- paired Function, J. Immunol. 2003, 171, tional Expression of a Plant K+ Channel 1304–1311. in Mammalian Cells, Biochem. Biophys. 47 Z. Zhou, Q. Gong, C. T. January, Cor- Res. Commun, 2000, 274, 130–135. rection of Defective Protein Trafficking of 56 K. Du , M. Sharma, G. L. Lukas, The a Mutant hERG Potassium Channel in DeltaF508 cystic fibrosis mutation im- Human Long QT Syndrome, J. Biol. pairs domain-domain interactions and Chem. 1999, 274(44), 31123–31126. arrests post-translational folding of 48 A. Paulussen, A. Raes, G. Matthijs, CFTR, Nat. Struct. Mol. Biol. 2005, 12(1), D. J. Snyders, N. Cohen, J. Aerssens, A 17–25. Novel Mutation (T65P) in the PAS Do- 57 P. H. Thibodeau, C. A. Brautigam, main of the Human Potassium Channel M. Machius, P. J. Thomas, Side chain hERG Results in the Long QT Syndrome and backbone contributions of Phe508 to by Trafficking Deficiency, J. Biol. Chem. CFTR folding, Nat. Struct. Mol. Biol. 2002, 277(50), 48610–48616. 2005, 12(1), 10–6. 49 C. R. Valdivia, M. J. Ackerman, D. J. 58 M. Gentzsch et al., EndocyticTrafficking Tester,T. Wada, J. McCormack, B. Ye, Routes of Wild Type and ÄF508 Cystic Fi- J. C. Makielski, A novel SCN5A arrhyth- brosis Transmembrane Conductance mia mutation, M1766L, with expression Regulator, Mol. Biol. Cell 2004, 15, 2684– defect rescued by mexiletine, Cardiovasc. 2696. Res. 2002, 55(2), 279–289. 59 N. J. Anegawa, D. R. Lynch, T. A. Ver- 50 M. Brejon, S. Le Maout, P. A. Welling, doorn, D. B. Pritchett. Transfection of J. Merot, Processing and Transport of N-methyl-D-aspartate receptors in a non- ROMK1 Channel is Temperature-Sensi- neuronal cell line leads to cell death, J. tive, Biochem. Biophy. Res. Commun. Neurochem. 1995, 64(5), 2004–2012. 1999, 261, 364–371. 60 N. J. Anegawa, R. P. Guttman, E. R. 51 K. Schroeder, J. Wu, L. Zhao, R. J. Grant, R. Anand, J. Lindstrom, D. R. Lukas, Regulation by cycloheximide and Lynch, N-Methyl-D-aspartate receptor References 107

mediated toxicity in non-neuronal cell 71 Y. Durocher, S. Perret, A. Kamen, lines: characterization using fluorescent High-level and high-throughput recom- measures of cell viability and reactive binant protein production by transient oxygen species production, Brain Res. transfection of suspension-growing hu- Mol. Brain. Res. 2000, 77(2), 163–175. man 293-EBNA1 cells, Nucleic Acids Res. 61 M. Raisinghani, L. S. Premkumar, 2001, 30(2), e9. Block of native and cloned vanilloid re- 72 P. Liljeström, H. Garoff, A New Gen- ceptor (TRPVI) by aminoglycoside anti- eration of Animal Cell ExpressionVectors biotics, J.Pain 2004, 0304–3959. Based on the Forest Vorus Replicon, Bio/ 62 D. Dobrev, U. Ravens, Therapeutically Tech. 1991, 9. relevant concentrations of neomycin se- 73 K. Lundstrom, Semliki Forest virus vec- lectively inhibit P-type Ca2+ channels in tors for rapid and high-level expression rat striatum, Eur. J. Pharmacol. 2003, of integral membrane proteins, Biochim. 461(2–3), 105–11. Biophys. Acta. 2003, 90–96. 63 T. Masuko, T. Kuno, K. Kashiwagi, 74 K. Lundstrom, A. Michel, H. Blasey, T. Kusama, K. Williams, K. Igarashi, A. R. Bernard, R. Hovius, H. Vogel, Stimulatory and Inhibitory Properties of Surprenant A. Expression of ligand-gated Aminoglycoside Antibiotics at N-Methyl- ion channels with the Semliki Forest D-Asparate Receptors, J. Pharm. Exp. virus expression system, J. Recept. Signal Ther. 1999, 290(3),1026–1033. Transduct Res. 1997, 17(1–3), 115–126. 64 A. Colosimo et al., Transfer and expres- 75 G. H. Gorrie,Y. Vallis, A. Stephenson, sion of foreign genes in mammalian J. Whitfield, B. Browning, T. G. cells, Biotechniques 2000, 2, 314–318, Smart, S. J. Moss, Assembly of GABBA 320–322, 324 passim. receptors composed of alpha1 and beta2 65 C. Chen, H. Okayama, High-efficiency subunits in both cultured neurons and fi- transformation of mammalian cells by broblasts, J. Neurosci. 1997, 17(17), 6587– plasmid DNA, Mol. Cell Biol. 1987, 7, 6596. 2745–2752. 76 O. Shamotienko, S. Akhtar, C. Sidera, 66 J. Gehl, Electroporation: theory and F. A. Meunier, B. Ink, M. Weir, J. O. methods, perspectives for drug delivery, Dolly, Recreation of Neuronal Kv1 Chan- gene therapy and research, Acta. Physiol. nel Oligomers by Expression in Mamma- Scand. 2003, 177(4), 437–447. lian Cells using Semliki Forest Virus, 67 RE. B. DuBridge, P. Tang, H. C. Hsia, Biochemistry. 1999, 38, 16766–16776. P. M. Leong, J. H. Miller, M. P. Calos, 77 T. A. Kost, J. P. Condreay, Recombinant Analysis of mutation in human cells by baculoviruses as mammalian cell gene- using an Epstein-Barr virus shuttle sys- delivery vectors,Trends Biotechnol. 2002, tem, Mol. Cell. Biol. 1987, 1, 379–387. 20(4). 68 P. Mellon,V. Parker,Y. Gluzman, 78 J. P. Condreay, S. M.Witherspoon, T. Maniatis, Identification of DNA se- W. C. Clay , T. A. Kost, Transient and quences required for transcription of the stable gene expression in mammalian human alpha 1-globin gene in a new cells transduced with a recombinant ba- SV40 host-vector system, Cell. 1981, 2(1), culovirus vector, Proc. Natl. Acad. Sci. 279–288. USA. 1999, 96, 127–132. 69 C. R. Sclimenti, M. P. Calos, Epstein- 79 W. C. Clay, J. P. Condreay, L. B. Moore, Barr virus vectors for gene expression S. L. Weaver, M. A.Watson, T. A. Kost, and transfer, Curr. Opin. Biotechnol. 1998, J. J. Lorenz, Recombinant baculoviruses 9, 476–479. used to study estrogen receptor function 70 B. Sugden, N. Warren, A promoter of in human osteosarcoma cells, Assay Drug Epstein-Barr virus that can function dur- Dev. Technol. 2003, 1(6), 801–810. ing latent infection can be transactivated 80 R. Ames, P. Nuthulaganti, J. Forn- by EBNA-1, a viral protein required for wald, U. Shabon, H. van-der-Keyl, viral DNA replication during latent infec- N. Elshourbagy, Heterologous expres- tion, J. Virol. 1989, 63(6), 2644–2649. sion of G protein-coupled receptors in 108 4 Functional Expression of Ion Channels in Mammalian Systems

U-2 OS osteosarcoma cells, Receptors 90 E. Kawashima et al., A novel and effi- Channels. 2004, 10(3–4), 117–124. cient method for the stable expression of 81 J. L. Pfohl, J. F. Worley 3rd, J. P. heteromeric ion channels in mammalian Condreay, G. An, C. J. Apolito, T. A. cells, Receptors Channels 1998, 5(2), 53– Kost, J. F. Truax, Titration of KATP 60. channel expression in mammalian cells 91 D. A. Persons et al., Enforced expression utilizing recombinant baculovirus trans- of the GATA-2 transcription factor blocks duction, Receptors Channels. 2000, 8(2), normal hematopoiesis, Blood 1999, 93(2), 99–111. 488–499. 82 R. Ames et al.. BacMam recombinant 92 J. Zhu, M. L. Musco, M. J. Grace, baculoviruses in G protein-coupled re- Three-color flow cytomentry analysis of ceptor drug discovery, Receptors Channels. tricistronic expression of eBFP, eGFP, 2004, 10(3–4),99–107. and eYFP using EMCV-IRES linkages, 83 S. K. Jang, H. G. Krausslich, M. J. Cytometry 1999, 37(1), 51–59. Nicklin, G. M. Duke, A. C. Palmen- 93 F. Minier, E. Sigel, Techniques: Use of berg, E. Wimmer, A segment of the 5' concatenated subunits for the study of li- nontranslated region of encephalomyo- gand-gated ion channels, Trends Pharm. carditis virus RNA directs internal entry Sci. 2004, 25(9). of ribosomes during in vitro translation, 94 M. Gossen, H. Bujard, Tight control of J. Virol. 1988, 62(8), 2636–2643. gene expression in mammalian cells by 84 S. Rees, J. Coote, J. Stables, S. Good- tetracycline-responsive promoters, Proc. son, S. Harris, M. G. Lee, Bicistronic Natl. Acad. Sci. USA 1992, 89, 5547– Vector for the Creation of Stable Mam- 5551. malian Cell Lines that Predisposes all 95 S. Renard, C. Drouet-Pétré, M. Parti- Antibiotic-Resistant Cells to Express Re- seti, S. Z. Langer, D. Graham, F. Bes- combinant Protein, Biotechniques 1996, nard, Development of an inducible 20(1), 102–110. NMDA receptor stable cell line with an 85 S. Rees. intracellular Ca2+ reporter, Eur. J. Pharm. 86 Y-H. Chen, T. J. Dale, M. A. Romanos, 1999, 366, 319–328. W. R. J. Whitaker, X. M. Xie, J. J. Clare, 96 I. Akibu et al., Stable Expression and Cloning, distribution and functional ana- Characterization of Human PN1 and lysis of the type III sodium channel from PN3 Sodium Channels, Receptors Chan- human brain, Eur. J. Neurosci. 2000, 12, nels 2003, 9, 291–299. 4281–4289. 97 M. Gossen, S. Freundlieb, G. Bender, 87 M. J. Main, J. E. Cryan, J. R. B. Dupere, G. Muller,W. Hillen, H. Bujard,Trans- B. Cox, J. J. Clare, S. A. Burbidge, criptional activation by tetracyclines in Modulation of KCNQ2/3 Potassium mammalian cells, Science 1995, Channels by the Novel Anticonvulsant 268(5218), 1766–1769. Retigabine, Mol. Pharm. 2000, 58, 253– 98 A. Rossignoli, M.G. Giribaldi, S Cor- 262. azza, Establishment of a CHO cell line 88 S. A. Burbidge, T. J. Dale, A. J. Powell, expressing the NMDA NR1a/NR2b sub- W. R. J. Whittaker, X. M. Xie, M. A. units for functional analysis of receptor Romanos, J. J. Clare, Molecular cloning, ligands, presented at The Society for Bio- distribution and functional analysis of molecular Screening Conference, 2002.

the NaV1.6. Vol.tage-gated sodium chan- 99 F. Yao,T. Svensjo, T. Winkler, M. Lu, nel from human brain, Mol. Brain. Res. C. Eriksson, E. Eriksson, Tetracycline 2002, 103, 80–90. repressor, tetR, rather than the tetR- 89 V. H. John et al., Heterologous expres- mammalian cell transcription factor fusion and functional analysis of rat sion derivatives, regulates inducible gene

NaV1.8 (SNS) voltage-gated sodium chan- expression in mammalian cells, Hum. nels in the dorsal root ganglion neuro- Gene Ther. 1998, 9(13), 1939–1950. bastoma cell line ND7–23, NeuroPharm. 100 A-L. Perraud et al., ADP-ribose gating 2004, 46, 425–438. of the calcium-permeable LTRPC2 chan- References 109

nel revealed by Nudix motif homology, lian cells and transgenic mice, Proc. Natl. Nature 2001, 411. Acad. Sci. USA. 1996, 93(8), 3346–3351. 101 M. Xia et al., Generation and Characteri- 104 D. Kurkó, P. Dezsõ , A.Boros, zation of a Cell Line with Inducible Ex- S. Kolok, L. Fodor, J. Nagy, Z Szom-

pression of CaV3.2 (T-Type) Channels, As- bathelyi, Inducible expression and phar- say Drug. Dev. Tech. 2003, 1(5). macological characterization of recombi- 102 J. G. Trapani, S. J. Korn, Control of ion nant rat NR1a/NR2A NMDA receptors, channel expression for patch clamp re- Neurochem. Int. 2005, 46, 369–379. cordings using an inducible expression 105 J. Nagy, A. Boros, P. Dezsõ, S. Kolok, system in mammalian cell lines, BMC L. Fodor, Inducible expresión and phar- Neurosci. 2003, 4, 15. macology of recombinant NMDA recep- 103 D. No, T. P. Yao, R. M. Evans, Ecdysone- tors, composed of rat NR1a/NR2B subu- inducible gene expression in mamma- nits, Neurochem. Int. 2003, 43, 19–29. 111

5 Analysis of Electrophysiological Data Michael Pusch

5.1 Overview

This chapter outlines electrophysiological methods for extracting biophysical parameters that describe two fundamental properties of ion channels: gating and permeation. The Introduction provides a broad overview of the general concepts of ion channel biophysics and the text a review of the kind of information that can be extracted from electrophysiological recordings. The later sections introduce several methods for the analysis of electrophysiological experiments on heterologously expressed ion channels. Many parts are explicit and can be directly applied “at the bench”. Other, more advanced topics (gating current measurements, single-channel kinetic analysis) are touched upon only superficially since their application requires further background that can be found in the references. A number of simple and useful mathematical equations are included which should assist the reader in the design, execution, and analysis of electrophysiological experiments. Readers interested in further theoretical background to the concepts outlined in this chapter are referred to Ref. [1] and references therein.

5.2 Introduction

The function of ion channels is to rapidly pass – in a passive but selective manner –a large number of ions across biological membranes. This electrogenicity is exploited by excitable cells to quickly change the transmembrane voltage allowing, for example, the conduction of the action potential and the postsynaptic electrical response to chemical neurotransmission. Other cells and organelles exploit the large capacity of charge transport for ion homeostasis and transepithelial transport. For the researcher the electrogenicity is interesting and useful because it allows the measurement of ion channel functioning in “real-time”. It is possible to monitor the action of ion channels both in vivo or in appropriate simplified in vitro systems like brain slices. Intracellular electrodes directly measure the membrane

voltage of individual cells while extracellular electrodes monitor cellular electrical activity indirectly. However, the recording of the physiological voltage provides little information regarding the biophysical parameters of the underlying channels, for two main reasons. First, all cells possess a complement of different ion channels and other electrogenic transporters that make it difficult to tease apart their respective contribution to the electrical response. Second, the physiological recordings are current-clamp measurements in which the membrane voltage is “freely floating” and the concentration of neurotransmitters or other ligands is un- controlled. In order to reliably define a physical parameter of an ion channel (ligand affinity, slope of the voltage-dependence etc.) these key variables have to be fixed, or “clamped”. A further important step is to isolate, as far as possible, a single type of ion channel. In physiological preparations this can be partially achieved using appropriate solutions and blocking agents. Heterologous systems are even better in this respect, even if they carry the risk that some physiologically important molecular component is missing. From the perspective described above the application and interpretation of the voltage-clamp measurements by Hodgkin and Huxley [2] revolutionized the approach to ion channel analysis. Today’s description and interpretation of ion channel function still draws heavily on the principles embodied in this work. One of the most important concepts for ion channels is the “gate”. A gate can be either open or shut (closed). There is no intermediate, half-open gate. At the time of Hodgkin and Huxley, there was no direct evidence for this as practically nothing was known at the level of the single channel. The assumption of a simple open or closed gate allowed a convenient description, as a two-state Markov process associated with a linear differential equation at a fixed voltage or ligand concentration. For the classical “m-gate”, the activation gate of the voltage-gated Na+ channel, this equation reads

dm tÀ m t 1 À m t dt

where m(t) is the probability of the gate being open at time t. The opening and closing rate constants, a and b, respectively, are voltage-dependent and represent a model for the underlying molecular rearrangement leading to gate opening. A ver- ification of the open–closed dichotomy of most ion channels became possible with the patch clamp technique [3, 4] that allowed the real-time visualization of single channel opening and closing in almost all types of animal, plant, and bacterial cells (for simplicity here, the existence of “subconductance” states, “flickering”, and other probably ubiquitous complicated single-channel behavior is ignored). Of course, the description of gating as a two-state process is an idealization, opening is not instantaneous. Also, microscopically, a single “open” state does not exist, but rather an almost infinite number of possible molecular configurations that are macroscopically lumped together into “the open” state, because functionally, they are almost indistinguishable from each other. The transitions from closed to open (and vice versa) appear “instantaneously” on single channel 5.3 Expression Systems and Related Recording Techniques 113 recordings, but in reality take several 100 ns. However, despite the availability of several ion channel structures (see for example Refs. [5–7]) and large computing power it is not yet possible to explore channel gating by molecular dynamics. Even ion permeation that occurs on a much faster time scale cannot be fully simulated, even though some theoretical progress has been made, especially for K+ channels (see for example Ref. [8] and Chapter 10). Quantitative functional measurements are still essential for a detailed insight into the mechanisms of ion channels. It can be expected that many more structural data for ion channels will be available in the near future. The structures will guide computational studies and rational mutagenesis in order to understand the mechanisms of function at a molecular model, to obtain high affinity ligands, and possibly to exploit ion channels as molecular devices in applied technological systems. Computational predictions and structure-based hypotheses have to be tested experimentally with functional data. The present chapter aims to provide an aid to designing, analyzing and interpreting such measurements.

5.3 Expression Systems and Related Recording Techniques

Each type of heterologous expression system determines a range of possible recording techniques. The most popular expression systems are Xenopus oocytes and mammalian cell lines, like HEK293 or CHO cells. A more rarely used system is the incorporation of relatively crude vesicles or purified proteins into planar lipid bilayers.

5.3.1 Expression in Xenopus Oocytes

The expression in Xenopus oocytes represents an extremely versatile system that allows the application of many different electrophysiological and biochemical methods [see Chapter 1 and Refs. 9–11]. Normally, in vitro transcribed cRNA is microinjected but expression can also be achieved using nuclear injection of eukaryotic expression plasmids. The oocyte system is popular because electrophysiological recordings can be easily performed by nonexperts employing the two-electrode voltage-clamp technique (TEV) [9]. This method allows a relatively high throughput compared to patch-clamp techniques and is thus often used, for example, for drug screening. A commonly underestimated problem of the TEV technique that is relevant also for qualitative measurements concerns the error introduced by the so-called series resistance (see for example Ref. [12]). The series resistance is caused by a finite conductivity of the oocyte cytoplasm, leading to a voltage drop within the cytoplasm and thus to a voltage error (see Fig. 5.1). Typical values of the series resistance are of the order of 0.5–1 kO. Thus a current of 10 µA will cause a voltage error of the order of 5–10 mV, a value that cannot always be neglected. 114 5 Analysis of Electrophysiological Data

Fig. 5.1 The intracellular series resistance in Xenopus oocytes. Current flowing through the interior of the oocyte leads to a voltage drop caused by the finite resistance of the cytoplasm.

Furthermore, even when the series resistance error is accounted for, the TEV technique has a limited time resolution of the order of almost 1 ms in most real- world applications. The apparent time resolution can be enhanced but the oocyte nevertheless provides a nonideal space-clamp. “Fast” kinetic parameters that are derived from TEV measurements are therefore seldom comparable to the same parameters measured with the patch clamp technique. Another disadvantage of whole oocytes is that their cytoplasmic content cannot be controlled. This may lead to a significant variability of measurements from different oocytes if the channel properties depend on the cytosolic composition. Also one would often like to change, or at least to fix, the intracellular solution. Furthermore, in the case of large expression the intracellular ion concentrations can be significantly altered by the voltage-clamp measurements. For example it is very difficult to handle a large expression (>10 µA) of the Cl– selective muscle channel CLC-1, because its kinetics depends strongly on the intracellular Cl– concentration. The disadvantages described above are partially overcome by the “cut-open” oocyte technique [13]. With this method only a small part of the surface area of the oocyte is clamped and the intracellular solution can be exchanged. However, the method is low throughput, necessitates considerable skill, and perfusion of the interior solution is very slow. Thus, this method finds a narrow range of special applications. A general problem with the expression of Ca2+-permeable channels or channels that depend on intracellular Ca2+ is that Xenopus oocytes endogenously express a 2+ – 2+ large Ca dependent Cl current, ICl– (Ca ) [14,15]. With maximal stimulation 2+ 2+ ICl– (Ca ) can reach several tens of µA of current. Thus activation of Ca permeable channels that leads to an influx of Ca2+ will inevitably activate this endogenous current and confound the measurements. It is also practically impossible to manipulate the intracellular Ca2+ concentration in order to study its effect on expressed channels. Nevertheless, the endogenous current can be exploited as a Ca2+ sensor to test for a possible Ca2+ permeability and also to test if the activation of (expressed) receptors and/or G-proteins results in an increase in the intracellular Ca2+ concentration (see for example Ref. [16]). One other advantage of the oocyte system is that several cRNAs coding for different subunits of an ion channel or other interacting proteins can be co-injected 5.3 Expression Systems and Related Recording Techniques 115 at defined proportions. For example dominant heterozygous genotypes of channelopathies can be simulated by a one to one co-expression of WTand mutant subunits, and possible dominant negative effects can be quantified (see for example Ref. [17]). Co-expression of different proteins can also be achieved in transfected cells. However, with the oocyte injection it is easier to precisely control the relative expression of each protein. Finally, electrical data recorded with the TEV can be correlated for the same oocyte with the surface expression of the expressed protein, using for example an introduced extracellular epitope [18, 19]. This is of particular importance for channelopathies because many disease-causing mutations are pathogenic because they effectively reduce or enhance the plasma membrane expression of the channel (see for example Refs. [18, 20]). The patch clamp technique can also be applied to Xenopus oocytes after the vitelline membrane has been removed [10]. Recordings can be performed in the cell- attached, the inside out and the outside out configuration (see Ref. [4] for a description of these methods). The electrical properties of the obtainable seal are ex- ceptional – seal resistances >100 GO can be achieved, which allow very high resolution recordings. The size of the patch pipette range from very small to “giant” [21], allowing single-channel or macroscopic recordings. Rapid solution exchanges can be applied to excised patches allowing a precise investigation of, for example, transmitter activated channels (see for example Ref. [22]). The excellent electrical properties of the cell-free patch-clamp configuration represent a significant advantage over whole cell recordings of small cells that can suffer from limited time-resolution due to the access (series) resistance (see below).

5.3.2 Expression in Mammalian Cells

Another popular expression system is “transfected” mammalian cell lines like HEK, CHO or many others (see Chapter 4 and Ref. [23]). The expression of one or more proteins is induced by the introduction of the DNA in an appropriate eukaryotic (often mammalian) expression vector by various chemical or physical methods. Cells can be either transiently or stably transfected. Stable transfection generally requires the integration of one or more copies of the expression construct into the genome and is initially more labor intensive than transient transfection. It is, however, convenient for long term studies on a particular channel or for drug screening where large numbers of cells may be required. Many molecular biological methods exist that increase transfection efficiency and the level of expression. Expression can also be induced with several different kinds of viruses [24]. The patch-clamp technique is the method of choice for studying the function of channels expressed in these small cells. All configurations (cell attached, whole cell, inside out, outside out) can be applied but the whole cell configuration is the most straightforward and widely used. Indeed, several different technical approaches have been taken to automate the whole cell patch-clamp for high throughput drug screening [see Chapter 6 and Ref. [25]]. Several factors have to be 116 5 Analysis of Electrophysiological Data

considered for the analysis of whole cell data. The time resolution in voltage-jump experiments is limited by the time that is necessary to charge the membrane capa-

citance, Cm, across the access resistance, Ra, given by

t = Cm Ra

Typical values of Cm = 20 pF, Ra =5MO yield a charging time constant of 0.1 ms. This is adequate for most applications but can lead to problems for very fast kinetics observed for example in voltage-gated Na+ or Cl– channels [26, 27]. The access resistance leads also to an error in the voltage-reading similar to the series resistance problem in the two-electrode voltage-clamp. For a membrane cur-

rent Im the voltage error amounts to

DV = Im Ra

which for typical values Ra =5MO, Im = 1 nA amounts to 5 mV and becomes worse for larger currents and/or access resistances. In voltage-jump experiments with large currents and fast kinetics the two kinds of errors described combine to create a complex dynamic error that can render certain measurements uninterpre- table. Most amplifiers provide an access (series) resistance compensation that com-

pensates both kinds of errors based on an estimate of Cm and Ra. Since the compensation involves positive feedback elements it increases the noise and is prone to oscillations. Care must therefore be taken with its application (see Ref. [28]). Similar to the oocyte system cell-free patches allow a much better voltage-clamp and also faster solution changes. However, because most mammalian cells are quite small, “macroscopic” recordings are more difficult to achieve in excised patches since large pipette diameters are poorly tolerated.

5.3.3 Leak and Capacitance Subtraction

In heterologous expression systems unwanted currents can arise either from true leak caused by the recording electrodes or from endogenous ion channels and transporters. These have to be carefully avoided using appropriate solutions and protocols. The subtraction of currents remaining after application of a specific blocker, if available, at a saturating concentration, is a very good but often tedious method. For ligand-gated channels the subtraction of currents at zero concentration of ligand is obviously a good method, because the spontaneous open probability is very small in most cases. For voltage-gated channels studied by step-protocols the responses are additionally distorted by the capacitive transients. These can be assumed to be linear, meaning that their size is proportional to the voltage step but independent of the voltage from which the pulse is delivered. Thus smaller steps applied in a voltage range where channels are closed or steps to the reversal potential can be used to subtract capacitive transients after appropriate scaling (see for example Refs. [29,30]). The most commonly used protocol for the subtrac- 5.4 Macroscopic Recordings 117 tion of linear leak and capacity currents when measuring voltage-gated sodium and potassium channels is called the “P/4 method” [29], that is a standard feature of most data acquisition programs. For this method four small voltage pulses with a quarter of the size of the main voltage pulse are delivered before or after the main pulse. The small pulses are subthreshold and elicit exclusively leak and capacitive currents. The response to the four quarter-sized pulses is summed and subtracted from the main response. Linear components are therefore practically completely subtracted.

5.4 Macroscopic Recordings

In the following sections it is assumed that the ion channel of interest is expressed in a heterologous system and represents the major contribution to the total membrane conductance. The methodologies to extract various biophysical parameters that are useful for a characterization of ion channels are ex- plained. Single channel measurements, if recorded at a sufficient bandwidth, contain, in principle, more information than macroscopic ensemble measurements. How- ever, they are significantly more technically demanding and the very small single channel currents for many ion channels renders their analysis virtually impossible. In addition, fast kinetics are difficult to measure at the single channel level because of the lower signal to noise ratio that has to be compensated by heavier filtering. Thus, for many applications macroscopic recordings represent the only practical approach. The most important relation regarding macroscopic currents is given by

I = Nip (1) that describes the total current, I, through a homogenous population of N independent channels. A basic assumption is that the channel under investigation possesses a single open state with current level i, and is without subconductance states. Of course this is an oversimplification for many channels. Without this assumption, however, a practical interpretation of macroscopic currents is almost impossible, because it is very hard to tease apart different conductance levels of a single channel from macroscopic recordings. The parameter p represents the open probability of the channel, that is the time- and/or voltage- and/or ligand-dependent probability of the channel being in a conducting state with an associated current, i. The single parameter p in Eq. (1) summarizes the combined action of all gates of the channels. Often it is useful to think of the gates as independent devices. For example the voltage-gated Na+ channel of Hodgkin and Huxley has 3 m- gates and one h-gate, all independent from each other such that the parameter p is equivalent to p = m3 h. While the independence of different gates is seldom realistic, it is very useful conceptually. 118 5 Analysis of Electrophysiological Data

Eq. (1) incarnates one of the dogmas of ion channel biophysics: permeation through the open channel, characterized by the parameter i, is independent of channel gating, characterized by the parameter p. This is a very useful conceptual distinction, although in real life it breaks down immediately: the occupancy of the pore by permeating ions generally stabilizes the protein structure and thereby influences gating. However, such effects are generally relatively small with some ex- ceptions. For example, in CLC type Cl–-channels, permeation and gating are strongly coupled [31, 32]. On the other hand, gating has practically no influence on permeation through the open pore, because, by definition “open” implies an open gate. Any influence of closed gates on ion occupancy of pore binding sites vanishes rapidly after opening the gate because the two processes occur on vastly different time scales. This assumption becomes questionable only if a rapid “flicker type” gate is present that opens and closes on a time scale more similar to that of ion conduction. For example in KvLQT1 (KCNQ1) K+ channels a rapid flicker type gate seems to be present that leads to drastic effects of Rb+ ions on the macroscopic current amplitude [33]. In this case it is difficult to distinguish between an effect of Rb+ on permeation or gating because the concept of a “gate” becomes questionable.

5.4.1 Analysis of Pore Properties – Permeation

Two basic parameters are important when considering permeation properties. One is ion selectivity and the other conductance. Ion selectivity is the ability to favor one ion over another and is expressed as the “permeability” ratio of the two

ions, for example PK/PNa for potassium and sodium. It is, in practice, determined from the reversal potential measured in mixtures of the two ions. The simplest situation is when the ions are present at equal concentrations, one on one side of the membrane and the other on the other side – so called bi-ionic conditions. We consider the example of a cationic channel measured with 150 mM NaCl intracel-

lular and 150 mM KCl extracellular. Then the reversal potential, Erev, is given by

Erev = RT/(zF)ln(PK/PNa) (2)

where R is the gas constant, T the absolute temperature, F the Faraday constant and z the valence (z = 1 for the example above). At room temperature the factor RT/F amounts to ~25 mV, a value that is very useful to remember for electrophy- siologists. Inverting Eq. (2) yields

zFErev PK/PNa =e /(RT)

+ Thus, for example, Erev = 25 mV indicates an e-fold higher permeability of K versus Na+ (e ~ 2.718). Bi-ionic conditions are preferable but are not easily achieved in some experimental systems. For example in TEV recordings from oocytes the intracellular solution cannot be changed. In this case, the difference of the reversal potential that arises by changing from one extracellular solution to an- 5.4 Macroscopic Recordings 119 other is measured. To obtain a permeability ratio, the Goldman–Hodgkin–Katz equation is used

Erev = RT/F ln((PK [K]ext + PNa [Na]ext)/(PK [K]int + PNa [Na]int)) (3)

+ + where [K]ext is the extracellular K concentration and similarly for Na (for anions the sign of the reversal potential has to be inverted). We consider the case where + + the extracellular concentrations of Na and K are changed from [Na]0 to [Na]1 and from [K]0 to [K]1, respectively, and assume that the measured reversal potential changes from E0 to E1. From Eq. (3) it follows that the permeability ratio is given by

f f PK/PNa = ([Na]1 – [Na]0 e )/([K]0 e – [K]1) where f = F(E1–E0)/(RT)~(E1–E0)/(25 mV). The assumption is that the intracellular concentrations do not change. When ions of different valence are compared (for example Na+ and Cl– or Na+ and Ca2+) the equations change slightly [1] but the basic type of measurement remains the same. One important and often overlooked problem of reversal potential measurements is the presence of liquid junction potentials that invariably arise when a solution is exchanged for another. The liquid junction potential is caused by the different mobility of different ions and is most pronounced when small inorganic mobile ions (for example Na+,Cl–) are exchanged by large organic quite immobile ions (for example NMDG+, gluconate–; see Ref. [34] for how to determine and correct for liquid junction potentials). When the Cl– concentration is changed care must be taken because most reference electrodes are Ag/AgCl electrodes that must be shielded from the solution exchange, for example by agar bridges. It may be difficult to determine the reversal potential because the current flowing through the channel is small, such that endogenous background conductances or a leak conductance dominate the effective reversal potential. One reason for this might be that the gates are closed at the expected reversal potential. This is especially a problem for voltage-gated channels. In this case so-called tail-current analysis can be applied to determine the reversal potential and the shape of the single channel current–voltage relationship, as illustrated in Fig. 5.2. The currents shown in Fig. 5.2c were simulated based on the two-state scheme shown in Fig. 5.2a using the pulse-protocol illustrated in Fig. 5.2b. Opening and closing rate constants are exponentially voltage-dependent such that the channel closes at negative voltages and opens maximally at positive voltages. The channel was assumed to have a linear single-channel current–voltage relationship (i–V) with an imposed reversal potential of Erev = –70 mV. However, from the steady state current–voltage relationship, obtained at the end of the variable pulse (see box in Fig. 5.2c, squares in Fig. 5.2d), the reversal potential cannot be obtained, because the open probability is too low. In contrast, the initial current, “immediately” after the end of the activating voltage-step (arrow in Fig. 5.2c), is measurable at these negative voltages. This “instantaneous tail current” and the resulting in- 120 5 Analysis of Electrophysiological Data

Fig. 5.2 Tail current analysis. Macro- scopic currents were simulated for the two state scheme shown in A. The reversal potential was –70 mV. From the holding potential of –80 mV a 0.2 s prepulse to 60 mV was followed by variable pulses ranging from –140 to 80 mV as illustrated in B. At negative voltages currents deactivate quickly such that the reversal potential cannot be reliably obtained from the steady-state currents (squares in D). The initial tail currents (circles in D) faithfully reproduce the linear single-channel current–voltage relationship and allow a precise determination of the reversal potential.

stantaneous current–voltage relationship (Fig. 5.2d, circles) reflect the shape of the single channel current–voltage relationship. This can be seen from the equation

Itail(Vt)=Npend i (Vt) (4)

where Itail is the instantaneous current at the tail-voltage,Vt, pend is the open-probability at the end of the activating prepulse, and i (Vt) is the single channel current at Vt. It is assumed that the open probability immediately after the voltage-step remains at the value it had at the end of the prepulse (pend). Then, the factor Npend is independent of the tail voltage, and Itail (Vt) is proportional to i (Vt). If the purpose is only to determine the reversal potential it is not very important to get exactly the initial current, and an average over a short stretch of currents shortly after the voltage jump can be calculated. For measuring the exact shape of the single channel i–V, care must be taken to determine the “correct” initial value. This may be complicated if the deactivation is fast and the voltage jump is associated 5.4 Macroscopic Recordings 121 with a large capacitance transient. In this case the time course of the deactivating current (after the capacitance transient) is fitted by a suitable function (for example an exponential function) that is then back-extrapolated to time “zero”. The determination of the permeability of a blocking ion can also be difficult. For example the muscle Cl– channel CLC-1 is blocked by iodide while iodide has a significant permeability with a permeability ratio of PI/PCl ~ 0.2. If extracellular Cl– is completely exchanged by iodide, however, CLC-1 is totally blocked and the reversal potential is dominated by endogenous background conductances, resulting in a wrong estimate of PI/PCl. The problem can be solved by partially exchanging extracellular Cl– for iodide (for example change from 100 mM Cl– to a solution containing 20 mM Cl– and 80 mM iodide) leaving enough Cl– to allow for a significant conductance. Eq. (3) can then be used again to quantify the permeability ratio.

5.4.2 Analysis of Fast Voltage-dependent Block – the Woodhull Model

Tail current analysis is useful to analyze voltage-dependent block by fast blockers. A commonly used model to describe voltage-dependent block is the Woodhull model [1, 35]. In this model it is assumed that the charged blocking particle enters the channel pore to a certain distance, and senses therefore part of the transmembrane electric field. Block is quantified by

I c 1 (5) I 0 c 1 exp zdVF= RT KD 0

Here I(c) is the current in presence of blocker at concentration c, KD(0) is the dissociation constant at zero voltage, z the valence of the blocking ion and d the “electrical” distance of the binding site from the bulk solution, that stands for the fraction of the electric field from the bulk solution to the binding site. Eq. (5) describes simultaneously the concentration and the voltage-dependence of block with just two parameters, KD(0) and d (see Fig. 5.3). Another way to describe the Woodhull model is in terms of an exponentially voltage-dependent dissociation constant (Fig. 5.3c). The simplicity of the Woodhull model makes it attractive and it is often a good initial model. Several assumptions are, however, seldom truly sa- tisfied. Firstly, “blocking” ions are often permeable to some extent and may “punch through” at large voltages. Also, almost all ion channels have multi-ion pores in which the ions interact. A blocking ion could displace a permeable ion present at the blocking site within the pore. Such effects add to the intrinsic voltage dependence of block (described by d) and complicate the picture. The incorporation of such features into mechanistic models is beyond the scope of this chapter. See Ref. [1] for a comprehensive description of blocking mechanisms. 122 5 Analysis of Electrophysiological Data

Fig. 5.3 Illustration of the Woodhull model of (dashed lines in A). The concentration depen- channel block. A linear single-channel i–V dence of block is illustrated in B for different and a zero reversal potential was assumed for voltages (–100, –30, 30, and +100 mV). The the unblocked channel (solid line in A). The exponential voltage-dependence of the appar-

parameters of the block were KD(0) = 1 mM ent KD, determined by fitting the curves in B and zd = 0.4. Application of increasing with a simple 1:1 binding curve is illustrated amounts of blocker (0.5, 1, and 5 mM) pro- in C. duces the increasing voltage-dependent block

5.4.3 Information on Gating Properties from Macroscopic Measurements

Macroscopic currents can be used to obtain an estimate about the open-probability. According to Eq. (1) the macroscopic current is proportional to the open probability, p, but further information or assumptions about the number of channels, N, and the single-channel current, i, are necessary to estimate p from the measured current, I. The number of channels can be assumed to be constant in a typical experiment, as long as its duration is relatively short and no particular maneuvers are undertaken to enhance or decrease protein turnover. Some ion channels can be drastically affected by co-expression and acute manipulation of co-expressed or endogenous regulating proteins, like the ubiquitin-protein ligase Nedd4 [36]. The number of channels can also be nonspecifically affected by agents that lead to a general retrieval of plasma membrane. For example treating Xenopus oocytes with phorbol esters, activators of protein kinase C, leads to an unspecific reduction of expressed conductances via a reduction of the plasma membrane surface [37]. Nevertheless, in most cases, N can be regarded as fixed. If measurements are performed at a fixed voltage, as in the case of ligand activated channels or in “isochronous tail-current” measurements for voltage-gated channels (see below) the single channel current, i, is also fixed. Otherwise, the shape of the single- channel current–voltage relationship (i – V) has to be taken into account. For the 5.4 Macroscopic Recordings 123 purpose of extracting information about the open probability, the i – V is parametrized in a phenomenological manner, without necessarily interpreting the corresponding parameters mechanistically. In the absence of direct information the i – V is often assumed to be linear

i(V)=g(V – Erev) with a single-channel conductance, g, and a reversal potential, Erev. This assumption is particularly appropriate if the reversal potential is relatively close to 0 mV, because in this case the intrinsic “Goldman”-rectification [38] has little influence on the shape of the i–V. If the reversal potential is far from zero, and if the channel is highly selective for one ion species present in the solutions, the Goldman– Hodgkin–Katz equation is more appropriate to describe the i – V because it takes the concentrations of the permeable ion on the two sides of the membrane into account [1]:

exp z À À 1 i VK r (6) exp zÀ1

Where K is a constant depending on the ionic concentrations, f = VF/(RT) and

fr = ErevF/(RT). The nonlinear shape of the Goldman current–voltage relationship is illustrated in Fig. 5.4 for a monovalent cation with a reversal potential of – 60 mV. It is also not uncommon that some voltage-dependent block or strong rectification is present that has to be taken into account for the description of the i – V. Such a block can be phenomenologically described by a factor that is derived from the Woodhull model. For a Goldman-type rectification with additional block the i – V is described by

exp z À À 1 1 i VK r (7) exp zÀ1 1 exp V À V1=V2

where V1 and V2 are empirical parameters describing the block; an example is shown in Fig. 5.4 (dashed line).

Figure 5.4 The Goldman–Hodgkin–Katz equation. The

solid line is drawn according to Eq. (6) with Erev = –60 mV. The dashed line is drawn according to Eq. (7) with

Erev = –60 mV and V1 = V2 =50mV. 124 5 Analysis of Electrophysiological Data

5.4.3.1 Equilibrium Properties – Voltage-gated Channels Two types of pulse protocols are most often used to extract the overall-open probability from macroscopic measurements. They are illustrated in Fig. 5.5, together with simulated currents based on the 2-state scheme of Fig. 5.2. In the direct pulse protocol (Fig. 5.5a and b) pulses are delivered to various potentials and the maximum current during the pulse is plotted versus voltage (Fig. 5.5c, squares). Usually, the open-probability of voltage-gated channels is parametrized with a Boltzmann distribution, here written in two different versions

1 1 popen (8) 1 exp V1=2 À V=k 1 exp zg V1=2 À VF= RT

In both forms the voltage of half-maximal activation,V1/2, describes the voltage at which popen = 0.5. The steepness of the voltage dependence is described either by the so-called “slope-factor”, k (in mV), or the “apparent gating valence” zg (di- mensionless). These two quantities are inversely related by

RT k zgF

Fig. 5.5 Determination of the open probability for voltage gated channels. Currents were simulated according to the 2-state scheme shown in Fig. 5.2A assuming a reversal potential of 0 mV and applying the pulse protocol shown in A. Steady state currents measured at the end of the variable-voltage pulse (see arrow in B) are plotted in C as squares. The tail currents at the beginning of the constant tail pulse to 60 mV are shown as circles. 5.4 Macroscopic Recordings 125

The Boltzmann equation derives from the statistical Boltzmann equilibrium that relates the ratio of the probability to be in one of two microscopic states, O and C that differ in free energy by a certain amount DG:

p O exp ÀDG= RT pC

The Boltzmann distribution of voltage-gated channels (Eq. (8)) stems from a simple model for voltage-gated channels that assumes that the free energy difference between the open and the closed state is additively composed of an electrical term, determined by the electrical charge, denoted by QC and QO, respectively, and a purely chemical term, DG0, such that DG is given by

DG DG0 V QO À QCDG0 VzgF

where zg is the apparent gating valence. The larger the charge difference between the closed and open state, the more sensitive is the channel to voltage, and the steeper the popen(V) curve. To extract the gating component (p) from the permeation component (i) for currents obtained from the “direct” I – V (Fig. 5.5c, squares) the I – V is fitted by the product of the i – V term and the Boltzmann term:

exp z À À 1 1 I VK r (9) exp zÀ1 1 exp zg V1=2 À VF= RT

Here, in Eq. (9), the Goldman–Hodgkin–Katz equation (Eq. 6) was used for a description of the i – V. The four parameters, K, fr,V1/2, and zg are obtained from a fit to the macroscopic data, while only the two parameters V1/2, and zg are relevant for gating. The tail-pulse protocol illustrated in Fig. 5.5a and b (see arrow) is often called “isochronous tail protocol” because the fixed tail pulse is applied after a fixed amount of time. The initial tail current is a measure of the open probability at the end of the (variable) pre-pulse (see Eq. (4)). As for the instantaneous I–V (see Sec- tion 5.4.1) a correct determination of the initial tail current may be hindered by the capacitive artifact. A careful back-extrapolation of the time course of the tail current to “time 0” may be necessary to obtain a reliable estimate of the initial tail current. The tail voltage should be chosen such that the relaxations are well resolved with the employed voltage-clamp technique. The resulting initial tail currents are then plotted versus the pre-pulse voltage (Fig. 5.5c, circles) and fitted by

I I V max (10) 1 exp zg V1=2 À VF= RT 126 5 Analysis of Electrophysiological Data

Here, Imax, is the maximal current obtained at saturating voltages. It can be determined by normalization with the measured currents or it can be left as an independent parameter. The latter possibility is particularly appropriate if the employed voltage range is not sufficient to saturate channel gating. In this case the plateau of the Boltzmann distribution is not reached and currents can not be normalized by the maximally measured value. Sometimes it happens that currents do not tend to zero at voltages where the channel should close according to Eq. (8). This may be an intrinsic property of the gating mechanism of the channel or may represent an uncompensated leak component. If such a “residual” open probability is an intrinsic property, a description with a Boltzmann distribution (Eq. (8)) is, strictly speaking, not adequate. Never-

theless the shape of the popen(V) curve can often be described phenomenologically by a modified Boltzmann distribution 1 À pmin I VImax pmin 1 exp zg V1=2 À VF= RT

Where pmin describes the minimal open probability reached at saturating voltages where channels are maximally closed. V1/2 is no longer the voltage at which popen = 0.5 but where popen = pmin + 0.5(1 – pmin). The isochronous tail-current protocol is, in principle, superior to the direct I – V because it is not influenced by the shape of the i–V. However, in certain cases a direct method has to be employed. For example, voltage-gated Na+ channels are governed by two main gating processes of opposite voltage dependence and one wants to determine separately their respective voltage dependence. Na+ channels inactivate with a voltage-dependent time-course after an activating voltage step. The steady state, isochronous tail current I – V would determine only the “window” current, the region where activation and inactivation gates are both open. To separate the two gates of the Na+ channel two different protocols are used to assess the voltage dependence of the activation and the inactivation gate, respectively. The “peak current” of the direct I – V is generally used as a measure of the activation gate (Fig. 5.6a and b). This is justified because the time constant of activation is considerably faster than that of inactivation. The inactivation is measured with a two-pulse protocol similar to the isochronous tail current protocol (Fig. 5.6c and d).

5.4.3.2 Equilibrium Properties – Ligand Gated Channels Conceptually, the determination of equilibrium properties of ligand activated channels is similar to that of voltage activated channels. The energy driving the conformational change is not supplied by the membrane voltage but by the chemical energy of ligand binding. Accordingly, the relevant intensive variable is the ligand concentration, here denoted by [L]. However, for ligand activated channels the allosteric action of the ligand is much more evident and explicit than is the voltage for voltage gated channels [39] (even though most quantitative models of 5.4 Macroscopic Recordings 127

Fig. 5.6 Activation and inactivation of vol- as shown by the lines. C shows currents from tage-gated sodium channels. Currents were a different cell evoked by a two-pulse protocol recorded from tsA201 cells transfected with as shown in the inset. The response to the the cardiac sodium channel and measured 100 ms long prepulse to voltages from –120 using the whole cell configuration of the to –10 mV is not shown. The currents repre- patch clamp technique. In A currents were eli- sent the response to the fixed tail pulse to – cited by 10 ms voltage steps from –80 to 10 mV that assays channel availability. The 50 mV (see inset). In B the peak currents are peak currents are plotted in D (symbols) to- plotted versus the test voltage (symbols) gether with a fit of Eq. (10). Note that activa- superimposed with a fit of the equation tion (A, B) and inactivation (C, D) have an opposite voltage-dependence. 1 I VGNa V À Erev 1 exp zg V1=2 À VF= RT voltage gated channels often have an allosteric character). Thus instead of assuming a scheme of the form

(Scheme 1) where binding of a ligand directly opens the channel, the minimal scheme applied for ligand activated channels assumes that binding of the ligand favors opening but does not directly open the channel. This scheme is given by

(Scheme 2) with a closed, unliganded state U, a liganded (bound) but closed state, B, and an open state, O [40]. Ligand binding occurs with second-order association rate kon 128 5 Analysis of Electrophysiological Data

and dissociation rate koff, while the allosteric transition is described by rate constants a and b. More complex schemes are necessary to include the possibility that unliganded channels are able to open [39]. Allosteric schemes and equations become even more complex if more than one binding site is present, the usual case in real life. Therefore, mostly for reasons of simplicity, the phenomenological description of ligand activated channels is often expressed in terms of the Hill equation

pmax popen (11) K n 1 D L

Where KD is the apparent dissociation constant of the binding site(s) and n the Hill coefficient, an estimate of the number of binding sites [1]. For the simple Scheme 2, equilibrium properties can indeed be expressed in the form of Eq. (11)

1 r= 1 r popen (12) 1 1 K KÃ 1 D 1 D r r L L

* Where r = a/b, KD = koff/kon, and the apparent affinity KD = KD/(1 + r). The maximum open probability is pmax = r/(1 + r)=a/(a + b). Thus, even though the concentration dependence of the macroscopic current strictly follows a 1:1 binding isotherm, the measured apparent affinity can be very different from that of the * true affinity of the ligand binding site. KD is always smaller than KD and only if r is very small (a 5 b) is the apparent affinity equal to the true affinity. But in this case

the ligand is not very effective (pmax 5 1). If r is very large, the ligand is very effective (pmax ~ 1) but the apparent affinity is much higher than the true affinity * (KD 5 KD). Since the absolute open probability is difficult to determine from macroscopic equilibrium measurements alone, additional information is necessary to determine true affinities and ligand efficacies. These can stem from kinetic macroscopic measurements, noise analysis, or single-channel analysis (see below). A fundamental difference between voltage gated channels and ligand gated channels is that the latter can be more or less efficiently activated by different types of ligands (for example certain glutamate receptors can be activated also by NMDA), while there is only one stimulus (voltage) for voltage gated channels. In terms of the simple model shown by Scheme 2, quantified by Eq. (12), different li-

gands have generally a different (true) affinity, and a different efficacy (pmax). For ligands that occupy the “same” binding sites the true number of binding sites is the same. The Hill coefficient in Eq. (11) may nevertheless be different, since Eq. (11) is an approximate phenomenological description of channel activation. Cer- tain ligands might also counteract the effect of a more potent activator. Further- more, certain receptors possess different kinds of binding sites. For example certain glutamate receptors need glycine as a co-factor for full activation by glutamate 5.4 Macroscopic Recordings 129

[41]. A good overview over equilibrium properties of ligand gated ion channels with further reference is given by [39]. Many ligand gated channels exhibit desensitization: currents decrease despite the continuous presence of ligand. The degree and kinetics of desensitization vary wildly between different channel types (see Ref. [42] and references therein). This phenomenon is conceptually similar to the “inactivation” of voltage gated channels. The presence of desensitization complicates the determination of activation properties. Experimentally, the most difficult problem, particularly if desensitization is an issue, is the fast application of the ligand (see Section 5.3.2).

5.4.3.3 Macroscopic Kinetics Channel kinetics can be evoked by a variety of stimuli. Sometimes, biophysical analysis alone is not sufficient to determine the physiological effect in a complex cellular system, as for example a cardiac myocyte, where numerous channel types contribute to the various phases of the action potential. In such cases, a “physiological” stimulation with an action potential waveform or other stimuli might reveal if, for example, a given mutation leads to action potential shortening (see for example Ref. [43]). However, such physiological stimuli are not well suited to unco- vering the underlying mechanistic effect. As outlined in the Introduction, for this purpose clamping the relevant intensive physical parameter (voltage, ligand concentration, temperature, pressure, light intensity,) to a fixed value and performing jump experiments are more informative. Like practically all kinds of conformational changes of proteins, current relaxations of a homogenous population of channels induced by a step-wise change of an intensive parameter, can be described by the sum of a constant term (the steady-state current, I?) and one or more exponential functions

Xn I tI1 ai exp Àt=i (13) i1

With amplitudes, ai, and time constants, ti (t is the time after the jump). The kinetics is thus determined by 2n+1 parameters. The time constants, ti, depend only on the actual value of the relevant physical parameter (the voltage or the ligand concentration after the jump), while the coefficients, ai, depend on the state occupancy before the jump. The exponential time dependence is a mathematical consequence from the Markov property of the conformational changes: once the channel undergoes a conformational change it loses “immediately” the memory of from what state it arrived (if there is more than one possibility to arrive in a certain state) and it has no memory of how long it has already been in a given state. One of the underlying assumptions is, of course, that there exists a finite set of de- finable, stable “states”. Actually, it is more scientific to turn the argument around: the experimental result that relaxation kinetics for most channels can be well described by Eq. (13) (with a reasonably small and reproducible number, n, of components) suggests that the Markov assumption is valid for ion channels. What is a 130 5 Analysis of Electrophysiological Data

reasonably small number, n? This is indeed a difficult question. In principle, a gating scheme with N states predicts exponential relaxation with N-1 components (for example a two-state system has single-exponential kinetics). In practice it is very difficult to reliably fit more than two or three exponential components. How- ever, gating schemes often require more than four states. Indeed it is extremely difficult to define a “correct” gating scheme based on fitting of current relaxations. Often gating schemes that have a large number of states can be simplified with symmetry arguments and the number of free parameters can be reduced. A beau- tiful example are the Hodgkin–Huxley equations (see Ref. [1]), but also recent models of Shaker K+ channels have a relatively small number of parameters, despite a large number of states (see Refs. [44, 45]). Furthermore, current relaxations are often approximately single- or double exponential under certain conditions, even though a full kinetic model requires many states, because most components are negligible. For example the deactivation of Na+ channels at very negative voltages after a brief activating pulse can be well described by a single exponential, even though the general gating involves very many states. Often, kinetics of ion channels is fitted and time constants are determined phenomenologically without necessarily wanting to define a molecular mechanism of gating. In many circumstances, this is the only practical choice, because the underlying mechanism is too complex to be determined reliably from the measurements. One of the most difficult problems in curve fitting with exponentials is to separate components with time constants that differ less than, let us say three-fold. Extreme care has to be taken in such cases and reproducibility has to be tested extensively. Often data can be reasonably well described with the sum of two exponentials, even though the “true” mechanism would require at least three. In such cases, the time constants (and relative weights) of the exponential components determined from the double-exponential fit can be almost meaningless. If the true time course is distorted slightly, for example by voltage-clamp errors (see above), the kinetic analysis becomes even more difficult. Under certain conditions it is impossible to directly follow the kinetics of a process measuring the ionic current. For example, the channel may be closed quickly by one kind of gate while another gate is slowly changing its status. Another reason that renders impossible a direct measurement might be that the voltage is close to the reversal potential or that a strong block occurs. Also, a large capacitive artifact may obscure fast relaxations [27]. In these cases so-called “envelope” protocols can often be applied, as illustrated in Fig. 5.7 that shows the classical protocol to study the recovery from inactivation of the voltage gated Na+ channel or, completely analogously it illustrates the measurement of the recovery from desensitization of a ligand gated channel. It is insightful to explicitly consider the kinetics of the most simple, two state system, because often more complex schemes can also be simplified to it, allowing an easy quantitative description.

(Scheme 3) 5.4 Macroscopic Recordings 131

Fig. 5.7 Envelope protocol to study recovery from inactivation. The pulse protocol is illustrated in A, current traces are shown in B. Cur- rents during the recovery period are not shown. In C the peak current at the final test pulse is plotted versus the recovery time together with a single exponential fit. Currents were simulated with a simplified Hodgkin–Huxley model.

Opening occurs with rate-constant a, closing with rate constant b. These rate constants depend on the intensive physical parameters (voltage, ligand concentration,). The equilibrium open probability is

a p1 a b

While the relaxation time constant is given by

t 1 a b

Knowing, both p and t allows the determination of a and b:

a p1=t; b 1 À p1=t

Relaxations that start from a given value of open probability, p0, proceed in time as

p tp1 p0 À p1 exp Àt=t

The Del Castillo–Katz model for ligand gated channels (Scheme 2) can be reduced to an effective two-state system if the ligand binding/unbinding is much faster than the opening isomerization transition (described by a and b). In this case the receptor is always in equilibrium with the ligand (see Ref. [1]): 132 5 Analysis of Electrophysiological Data

(Scheme 4)

Where the effective opening rate is given by

1 aeff a K 1 D L

Combining steady state-measurements (Eq. (12)) and relaxation measurements with step-changes in ligand concentration allows the determination of all three

parameters (KD, a, b) of the model of Scheme 4: for jumps into zero concentration –1 of ligand, the relaxation rate is t = because aeff = 0 for [L] = 0. For jumps into sa- –1 turating concentration of ligand the relaxation rate is t = a + b because aeff = a for [L]4KD. These kinetic experiments thus provide estimates for a and b. From * the equilibrium measurements the apparent affinity, KD is determined (Eq. (12)) and using the values for a and b the true affinity, KD, can be calculated from KD = * KD (1 + a/b). This simple example illustrates how the combined use of equilibrium and kinetic measurements in addition to simplifying assumptions can be used to obtain quantitative information about a molecular mechanism.

5.4.4 Channel Block

All ion channels interact with a variety of smaller molecules (peptides, small organic molecules) that directly or indirectly reduce ion permeation. Such substances are called blockers or inhibitors and are the bread and butter of the pharmaceutical industry interested in ion channel targets. Blockers may directly block the pore and physically impede ion flow. Inhibitors may reduce current flow by stabilizing the closed state of a channel gate, for example, in which case the binding site may be far away from the ionic pore. Such inhibitors are often called “gating modifiers”. This distinction between these two kinds of modulators is actually not so strict – many pore blockers additionally alter the gating by binding more tightly to one or another gating state (state-dependent block). Such blockers may be useful tools to study the properties of channels [46, 47]. Many pore blockers exert a voltage-dependent block that can often be described by the Woodhull model (see Section 5.4.2). For most practical purposes channel block is quantified by the Hill equation

I B 1 I 0 B n 1 KD

that quantifies the ratio of current in the presence of blocker at concentration [B] to

the current in its absence with an apparent affinity KD and the Hill coefficient, n. 5.4 Macroscopic Recordings 133

5.4.5 Nonstationary Noise Analysis

The opening and closing of ion channels is a random process that renders current registrations “noisy”, in particular if few channels are present. The statistical properties of the noise can be used to infer some characteristics of the underlying elementary events. A prerequisite is that the channel-induced noise is significantly above the background noise of the recording system. This condition is, for example, generally not fulfilled in TEV recordings from Xenopus oocytes. The patch- clamp technique, on the other hand, is exceptionally well suited for noise analysis. However, background noise may be large if the series resistance in whole cell recordings is highly compensated (see Section 5.3.2). For stationary noise analysis the system is recorded for a prolonged time at fixed external conditions (voltage, ligand concentration). The power spectrum is then fitted with the sum of Lorent- zian functions. While this method can yield important information [48], nonstationary noise analysis is often faster and easier to perform. Under appropriate conditions, each of the elementary parameters of Eq. (1) can be determined assuming a single open conductance level. This should be verified with single channel analysis. For standard nonstationary noise analysis a step-protocol (that may be a voltage- step or stepwise change in ligand concentration) is applied repeatedly, with enough time passing between individual stimulations to ensure identical initial conditions for each step. Each current response, Ii(t), is recorded (i =1,…,n) (Fig. 5.8a). From these recordings the mean can be calculated by

Xn 1  Ii t (14) n i1

This is now a much smoother curve than the individual traces (Fig. 5.8 b) and because of this it can be written a

 Nip t where p(t) is the time course of the (“true”) open probability. The variance of the response, s2(t), for each time point, t, is given by

Xn 2 1 2 s t Ii tÀ (15) n i1 the standard statistical definition (Fig. 5.8c). However, the most important “trick” in nonstationary noise analysis is to calculate the variance not as suggested by Eq. (15) but as the squared difference of consecutive records:

XnÀ1 2 1 2 s t I tÀI t (16) 2 n À 1 i 1 i i1 134 5 Analysis of Electrophysiological Data

(Note the scaling by a factor of 1/2 in Eq. (16)). For a perfectly stable system the results obtained from Eq. (15) and Eq. (16) are identical. However, in reality, small drifts of the total current amplitude (“run-down”, “run-up”) or of the reversal potential or other parameters are practically unavoidable. These small drifts are well cancelled out using Eq. (16) while they artificially increase the variance when calculated by Eq. (15) and may render the noise analysis meaningless, particularly if the single channel conductance is small [49]. Meaningful results can be obtained even with run-down up to a factor of two or more. Since the variance is obtained from the differences of records, leak currents and capacity transients also cancel out (if they are associated with negligible inherent noise). That means that the individual records do not have to be, and should not be, leak-corrected for the application of Eq. (16), in contrast to the calculation of the mean (Eq. 14). A nice property of the variance is that independent noise source adds independently to it. Thus, total variance is given by

s2 s2 s2 tot channel background

and the background variance can usually be assumed to be independent of voltage. It can thus be determined at a voltage where no ion current flows through the channels (for example at the reversal potential or in the absence of ligand for ligand activated channels or at a voltage where channels are closed for voltage gated channels). Having determined the variance and the mean it is now possible to proceed with the “variance-mean analysis”. For this we use the equality

s2 Ni2p 1 À p (17)

This fundamental equation [48, 49] can be understood intuitively: for p = 0 the variance is null because no current flows. Also for p = 1 (the maximum value it can attain) there is no fluctuation in the current because all channels are permanently open. The largest fluctuations are present for p = 0.5 when channels are half open and half closed on average. Note that (Eq. (17)) is independent of the kinetics of the fluctuation. However, the bandwidth of the recording system must be sufficiently large to resolve the fastest transitions. Combining Eq. (17) with Eq. (1) yields

s2 iI À I2=N (18)

that relates the two macroscopic quantities, s2, and mean current I, in a parabolic function that depends on two parameters, the single channel current i, and the number of channels, N, that can be determined by a least-squares fit (Fig. 5.8d). Often the whole traces are not plotted and fitted against each other but they are first binned in an appropriate manner [49] (see Fig. 5.8d). The two parameters, i

and N, are best defined if a large interval of popen is covered by the relaxation. If only a very small interval of popen is sampled, the two parameters cannot be determined independently: a small variance can be caused by a small number of chan- 5.4 Macroscopic Recordings 135 nels and/or by a small or large popen and vice versa. It may be that the current ex- cursion in the relaxation is substantial but that popen remains significantly smaller than 0.5. In this case the second term in Eq. (18) is negligible, the relationship is linear, and only the single channel current can be determined. If both, i and N, can be determined the absolute popen during the relaxation can be calculated from

I t p t Ni

Thus, using a simple experimental protocol (Fig. 5.8) allows a quite precise determination of fundamental channel parameters, without the need to perform single channel analysis.

Fig. 5.8 Nonstationary noise analysis. Currents were repeatedly evoked by a test pulse and individual responses are shown in A (currents were simulated with a simplified Hodgkin–Huxley model). The mean and the variance are shown in B and C, respectively. In D the (binned) variance is plotted versus the respective mean together with a fit of Eq. (18). The horizontal line in D marks the level of the background variance.

5.4.6 Gating Current Measurements in Voltage Gated Channels

Another way of obtaining additional information about molecular gating mechanisms is to measure the so-called “gating currents” associated with the molecular rearrangements of voltage gated cation channels [50]. These are transient currents, similar to capacitive currents, that reflect the movement of the gating charges within the electric field. Of course, any conformational rearrangement of 136 5 Analysis of Electrophysiological Data

the channel protein that is associated with charge redistribution gives rise to “transient” gating currents, even if they do not reflect the movement of a “voltage sensor”. However, in voltage gated K+,Na+ and Ca2+ channels the voltage-sensor movements clearly dominate the gating currents. In order to resolve gating currents that are of small magnitude it is necessary to eliminate the normal ionic currents by applying blockers or by eliminating permeant ions. However, it is necessary to ensure that such maneuvers of completely eliminating ion flow through the pore does not alter significantly the gating process itself, an often difficult task. It is beyond the scope of this chapter to describe the design and the analysis of gating current measurements (see Ref. [51] for review). However, the estima- tion of the total gating charge of a single voltage gated K+ channel provides a nice example of this approach [52]. The authors determined first the number of channels, N, in an inside out patch using nonstationary noise analysis (see Section 5.4.5). Then they replaced intracellular K+ with TEA+, a blocker of K+ channels, to eliminate the ionic currents and measured the total gating charge, Q, by integrating the gating currents. The ratio Q/N, the gating charge per channel, was about 12 elementary charges, consistent with more indirect measurements. Another nice result was obtained by Conti and Stühmer [53], who estimated the size of the charge of a single voltage sensor in Na+ channels. Voltage gated cation channels possess four voltage sensors that move more or less independently of each other. The movement of each sensor produces a spike-like tiny current. The ensemble of many sensors is the random superposition of many such spikes, filtered at the recording frequency. Nonstationary noise analysis yielded an elementary charge of individual spikes of about 2.3 elementary charges, a very reasonable value.

5.5 Single Channel Analysis

The possibility to observe and analyze the opening and closing of single ion channel molecules marked a revolution in ion channel research and remains one of the few techniques that allow a true single molecule measurement in real time [3, 4]. Single channel recordings can provide a wealth of information and numerous numerical methods for single channel analysis have been developed. The reader is referred to “Single channel Recording” by Sakmann and Neher for detailed information [54]. Here a very broad overview of a typical single channel analysis is provided. Recent papers utilizing more advanced techniques can also be found [55, 56].

5.5.1 Amplitude Histogram Analysis

The first step of a single channel analysis is usually to construct an amplitude histogram. Already at this point one ever returning aspect of single channel analysis becomes important: adequate filtering. Of course the data should be filtered with 5.5 Single Channel Analysis 137 a good filter (for example an 8-pole Bessel filter) with a cut-off of at least at one half the sample frequencies to avoid aliasing [57]. In order to get, at least in principle, as much information as possible, the sampling rate must be sufficiently high. It is however useless to acquire data at a low signal-to-noise ratio. A dilemma is sometimes that one would like to see online the highly filtered data in order to get an immediate impression of its quality, but one also wants to acquire at a higher frequency for quantitative offline analysis. One possibility is to divert the signal after a primary anti-alias filter into two separately sampled channels. One signal is acquired after only the first filter, while in the second the current signal is sub- jected to further filtering for immediate inspection. This can be done on an oscil- loscope, or, if the acquiring software allows the simultaneous sampling of two channels, the highly filtered signal can be acquired as a second input channel and visualized on the computer screen. For the construction of the amplitude histogram all sample points that fall in a given “current-bin” are counted, resulting in the number of events per bin that are then plotted versus the mid-point of the current bin. This is illustrated in Fig. 5.9. To each current level of the channel (“closed” and “open” in Fig. 5.9) corresponds a “peak” in the amplitude histogram. The peaks may not be well separated because noise is large (Fig. 5.9b). In this case the data can be digitally filtered by a Gaussian filter that has various convenient properties [57] (Fig. 5.9c). If the baseline is not stable the amplitude histogram becomes distorted. Excessive baseline drift can make the single channel analysis very difficult and must be corrected. Several analysis programs are available that allow baseline correction and many other features. Once an acceptable amplitude histogram has been constructed it is fitted with the sum of Gaussian functions, Gi, one for each peak, i X X 2 a I À H I G I i exp À i (19) i s s2 i i i 2 i

Where each Gaussian functions is characterized by a mean mi, a width, si, and amplitude, ai. The inclusion of the width, si, in the prefactor ai/si in the Gaussian fit (Eq. (19)) facilitates the calculation of the relative area, Ai, that is occupied by each Gaussian component:

Pai Ai 100% aj j

The relative area, Ai, is a measure of the probability to dwell in the conductance state associated with mean mi. Often it happens that the membrane patch contains an unknown number, N, of identical channels, leading to equidistant peaks of the amplitude histogram at levels ni (n = 0, 1, …), where i is the amplitude of a single channel. Even though the absolute open probability cannot easily be determined in such a situation a useful parameter to evaluate effects of drug application or other maneuvers is the so-called “NpO”, i.e. the product of the (unknown) number 138 5 Analysis of Electrophysiological Data

Fig. 5.9 Amplitude histogram analysis. Cur- the right amplitudes (0 pA and 1 pA), and the rents were simulated based on a simple 2- fit with the sum of two Gaussian functions state scheme with an open conductance level (thick line in B) correctly predicts the ampli- of 1 pA and an added Gaussian noise of tude and area of each conductance level. The 0.4 pA SD (panel A). The baseline and the trace shown in C is strongly filtered and the open conductance level are indicated by hori- histogram in D shows two well separated zontal lines. The noisy trace in A seems to be peaks of correct amplitude and weight. The almost useless. However, the amplitude his- bin width used for the histograms was 10f.A. togram shown in B clearly shows two peaks at

of channels and their open probability. From the histogram fit the “NpO” can be calculated as

P1 jAj j 0 NpO P1 Aj j0

where the “areas” Aj are obtained from the Gaussian fits as described above, and A0 corresponds to the baseline.

5.5.2 Kinetic Single Channel Analysis

A comprehensive description of the kinetic analysis of single channel data is beyond the limits of this chapter and only general directions can be given. The stra- tegic decisions that can be taken for kinetic analysis are outlined schematically in Fig. 5.10. The most direct way of analysis is depicted on the left of the figure and 5.5 Single Channel Analysis 139

Fig. 5.10 Flow diagram of strategies for kinetic single channel analysis. consists of directly fitting a “hidden Markov model” to the “raw” data [58–60]. A Markov model is a kinetic scheme like those described above (for example, Scheme 2) with possibly many states of various conductance levels connected by rate constants. The rate constants and the conductance levels of the various states are the parameters fitted in this approach. The model is “hidden” for two reasons. First, several kinetic states may be associated with the same conductance. Transi- tions among these states are therefore not directly visible. Second, the noise can hide shortlived dwell times or low conductance states. One advantage of the hidden Markov model approach is that it takes the noise into account explicitly [59, 60]. Algorithmically, for the hidden Markov model the following question is raised: for a given set of parameters (these parameters include the rate constants of the model, the conductance level of each state, and parameters that describe the noise) what is the probability of observing the currents that have been measured? The parameters of the model and the characteristics of the noise are then 140 5 Analysis of Electrophysiological Data

adapted to maximize this probability. The calculation of this probability is a for- midable task but efficient algorithms have been developed that allow the analysis of quite long data sets. One drawback of the method is that a reasonable kinetic model should a priori be known. The method can be applied for example to study the effect of mutations of an ion channel for which a kinetic scheme has been established previously. Another drawback is that the method is a kind of black box with little possibility of visual evaluation to check if the results are “reasonable”. A further serious problem may occur if the general properties of the noise are not adequately treated [60]. Thus, while the hidden Markov modeling can be a powerful tool, its use requires some experience and results should be checked with other methods. The more traditional methods of analysis do not work directly on the raw data trace but this is first “idealized” (Fig. 5.11). In the idealization process the events of channel opening and closing are detected either completely automatically or in- teractively using specially designed computer programs. The noisy raw data trace is thus substituted by the idealized smooth trace (Fig. 5.11) that can be easily represented as a list with two numbers for each entry in the list: the duration of each dwelling together with the corresponding current level. For idealization one must decide if a current fluctuation represents a transition to another conductance level or if it is just noise. As a criterion the 50% level criterion is most often

Fig. 5.11 Dwell-time analysis. In A a short cumulative distribution equals 1 for a dura- stretch of a simulated trace of a two-state tion of 0. A single exponential dwell-time dis- scheme is shown. In B the corresponding tribution is represented by a straight line in idealized trace is displayed. The cumulative the logarithmic scaling of Fig. 5.11. Other re- dwell-time histograms shown in C and D are presentations of dwell-time histograms are based on a total of 408 events each and repre- probably more common [62] but require a sent the relative frequency of events to be much larger number of events for a satisfac- longer than a given duration. By definition the tory graphical display. 5.5 Single Channel Analysis 141 employed [57]. To apply this criterion, the possible conductance levels are estimated first by the fitting of the amplitude histogram (see above). In addition, events are only accepted if they are of a minimal length. For a reliable assignment of the transitions data usually have to be filtered more than for the hidden Markov approach, at least if the signal to noise level is low. The basic problem in the idealization process is that short events are easily missed (the missed events problem) but short events can also be artifactually introduced if noise is excessive. The problem is double: missing a short closure not only leads to the loss of a closed event but it also lengthens the opening time of the event during which the closure occurred. Similarly, the artifactual introduction of a short closure not only alters the closed times but also shortens, and divides into two the underlying opening. To deal with this problem in generality is not easy. First, a reasonable cut-off is defined such that no events shorter than this are rigorously accepted and the resulting error in the final fitting procedure can then be compensated [56, 61]. For simplicity, we ignore here the missed events problem. The idealized trace can then be analyzed in several ways. A first step is to construct and inspect dwell-time histograms. Several kinds of binning procedures and histograms to construct have been proposed (see for example [62]). In Fig. 5.11 so-called cumulative dwell time histograms for the two conductance levels are displayed. These histograms can then be fitted with the sum exponential components in order to extract kinetic information from the single channel data. The construction and fitting of histograms is always a first step in data analysis if little is known about the underlying channel and if one wants to obtain an impression about its kinetic behavior, like for example an estimate of the number of open and closed states. Histogram fitting may also provide a purely empirical set of parameters whose variation under the influence of ligands or mutation can be studied. If a good working hypothesis for a kinetic Markov model for the channel is available it may instead be a good idea to fit directly the likelihood of the idealized channel trace. This approach is similar to the hidden Markov approach described above in that the full kinetic information including possible correlations are exploited. The maximum-likelihood fitting overcomes one of the biggest problems of histogram fitting: it is not clear how the time constants and coefficients extracted from closed and open time histograms have to be weighted in fitting a concrete Markov scheme. In the maximum-likelihood approach the rate constants defining the Markov scheme are directly optimized [56, 61]. Recordings obtained under different conditions can also be fitted simultaneously. This approach thus allows on the one hand an objective estimate of physical parameters similar to the hidden Markov fitting and on the other hand the results can be easily judged vi- sually by comparing the predictions for all kinds of dwell-time histograms [56]. 142 5 Analysis of Electrophysiological Data

5.6 Summary

This chapter provides a broad overview of current concepts and methods of analysis of electrophysiological data. Several of the methods are incorporated into the free analysis program written by the author that is available for download at http://www.ge.cnr.it/ICB/conti_moran_pusch/programs-pusch/software-mik.- htm. The simulation program used to generate several of the figures can also be found there.

Acknowledgements

I thank Armando Carpaneto for critically reading the manuscript. The financial support by Telethon Italy (grant GGP04018) and the Italian Research Ministry (FIRB RBAU01PJMS) is gratefully acknowledged.

References

1 B. Hille, Ion Channels of Excitable Mem- 8 S. Bernèche, B. Roux, A microscopic branes, Sinauer, Sunderland, MA, 2001. view of ion conduction through the K+ 2 A. L. Hodgkin, A. F. Huxley, A quanti- channel, Proc. Natl. Acad.Sci. U. S. A. tative description of membrane current 2003, 100, 8644–8648. Epub 2003 Jul 01. and its application to conduction and ex- 9 W. Stühmer, Electrophysiological re- citation in nerve, J. Physiol. 1952, 117, cording from Xenopus oocytes, Methods 500–544. Enzymol. 1992, 207, 319–339. 3 E. Neher, B. Sakmann, Single-channel 10 W. Stühmer, Electrophysiological re- currents recorded from membrane of de- cordings from Xenopus oocytes, Methods nervated frog muscle fibres, Nature 1976, Enzymol. 1998, 293, 280–300. 260, 799–802. 11 M. F. Romero,Y. Kanai, H. Gunshin, 4 O. P. Hamill, A. Marty, E. Neher, M. A. Hediger, Expression cloning B. Sakmann, F. J. Sigworth, Improved using Xenopus laevis oocytes, Methods patch-clamp techniques for high-resolu- Enzymol. 1998, 296, 17–52. tion current recording from cells and 12 G. Nagel, T. Szellas, J. R. Riordan, cell-free membrane patches, Pflugers. T. Friedrich, K. Hartung, Non-specific Arch. 1981, 391, 85–100. activation of the epithelial sodium chan- 5 D. A. Doyle et al., The structure of the nel by the CFTR chloride channel, potassium channel: molecular basis of EMBO Rep. 2001, 2, 249–254. K+ conduction and selectivity, Science 13 E. Stefani, L. Toro, E. Perozo, F Beza- 1998, 280, 69–77. nilla, Gating of Shaker K+ channels: I. 6 Y. Jiang et al., The open pore conforma- Ionic and gating currents, Biophys. J. tion of potassium channels, Nature 2002, 1994, 66, 996–1010. 417, 523–526. 14 M. E. Barish, A transient calcium-de- 7 R. Dutzler, E. B. Campbell, M. Cadene, pendent chloride current in the imma- B. T. Chait, R. MacKinnon, X-ray structure Xenopus oocyte, J. Physiol. 1983, ture of a ClC chloride channel at 3.0 Å re- 342, 309–325. veals the molecular basis of anion selectivity, Nature 2002, 415, 287–294. References 143

15 C. Hartzell, I. Putzier, J. Arreola, Cal- 27 A. Accardi, M. Pusch, Fast and slow cium-activated chloride channels, Annu. gating relaxations in the muscle chloride Rev. Physiol. 2005, 67, 719–758. channel CLC-1, J. Gen. Physiol. 2000, 116, 16 T. M. Moriarty et al., Beta gamma subu- 433–444. nits of GTP-binding proteins inhibit 28 F. J. Sigworth, in Single-channel Record- muscarinic receptor stimulation of phos- ing, B. Sakmanmn, E. Neher (Eds.), Ple- pholipase C, Proc. Natl. Acad. Sci. U. S. num Press, New York, 1995, pp. 95–127. A. 1988, 85, 8865–8869. 29 F. Bezanilla, C. Armstrong, Inactiva- 17 M. Pusch, K. Steinmeyer, M. C. Koch, tion of the sodium channel. I. Sodium T. J. Jentsch, Mutations in dominant hu- current experiments, J. Gen. Physiol. man myotonia congenita drastically alter 1977, 70, 549–566. the voltage dependence of the CIC-1 30 C. Saviane, F. Conti, M. Pusch, The chloride channel, Neuron 1995, 15, 1455– muscle chloride channel ClC-1 has a 1463. double-barreled appearance that is differ- 18 D. Firsov et al., Cell surface expression entially affected in dominant and reces- of the epithelial Na channel and a mu- sive myotonia, J. Gen. Physiol. 1999, 113, tant causing Liddle syndrome: a quanti- 457–468. tative approach, Proc. Natl. Acad. Sci. U. 31 M. Pusch, U. Ludewig, A. Rehfeldt, S. A. 1996, 93, 15370–15375. T. J. Jentsch, Gating of the voltage-de- 19 N. Zerangue, B. Schwappach,Y. N. Jan, pendent chloride channel CIC-0 by the L.Y. Jan, A new ER trafficking signal reg- permeant anion, Nature 1995, 373, 527– ulates the subunit stoichiometry of 531. plasma membrane K(ATP) channels, 32 T-Y. Chen, Structure and function of Neuron 1999, 22, 537–548. CLC channels, Annu. Rev. Physiol. 2005, 20 M. Schwake, M. Pusch, T. Kharkovets, 67, 809–839. T. J. Jentsch, Surface expression and 33 M. Pusch, L. Bertorello, F. Conti, Gat- single channel properties of KCNQ2/ ing and flickery block differentially af- KCNQ3, M-type K+ channels involved in fected by rubidium in homomeric epilepsy, J. Biol. Chem. 2000, 275, 13343– KCNQ1 and heteromeric KCNQ1/ 13348. KCNE1 potassium channels, Biophys. J. 21 D. W. Hilgemann, C. C Lu, Giant mem- 2000, 78, 211–226. brane patches: improvements and appli- 34 E. Neher, Correction for liquid junction cations, Methods Enzymol. 1998, 293, potentials in patch clamp experiments, 267–280. Methods Enzymol. 1992, 207, 123–131. 22 D. J. Maconochie, D. E. Knight, A 35 A. M. Woodhull, Ionic blockage of so- method for making solution changes in dium channels in nerve, J. Gen. Physiol. the sub-millisecond range at the tip of a 1973, 61, 687–708. patch pipette, Pflugers. Arch. 1989, 414, 36 O. Staub et al., Regulation of stability 589–596. and function of the epithelial Na+ chan- 23 R. J. Kaufman, Overview of vector design nel (ENaC) by ubiquitination, EMBO J. for mammalian gene expression, 1997, 16, 6325–6336. Methods. Mol. Biol. 1997, 62, 287–300. 37 M. S. Awayda, Specific and nonspecific 24 G. P. Nolan, A. R. Shatzman, Expres- effects of protein kinase C on the epithe- sion vectors and delivery systems, Curr. lial Na+ channel, J. Gen. Physiol. 2000, Opin. Biotechnol. 1998, 9, 447–450. 115, 559–570. 25 P. B. Bennett, H. R. Guthrie, Trends in 38 D. E. Goldman, Potential, impedance, ion channel drug discovery: advances in and rectification in membranes, J. Gen. screening technologies, Trends Biotechnol. Physiol. 1943, 27, 37–60. 2003, 21, 563–569. 39 J. Krusek, Allostery and cooperativity in 26 W. Stühmer et al., Structural parts in- the interaction of drugs with ionic chan- volved in activation and inactivation of nel receptors, Physiol. Res. 2004, 53, 569– the sodium channel, Nature 1989, 339, 579. 597–603. 144 5 Analysis of Electrophysiological Data

40 J. Del Castillo, B Katz, Interaction at 52 N. E. Schoppa, K. McCormack, M. A. end-plate receptors between different Tanouye, F. J. Sigworth, The size of gat- choline derivatives, Proc. R. Soc. London, ing charge in wild-type and mutant Sha- Ser. B. Biol. Sci. 1957, 146, 369–381. ker potassium channels, Science 1992, 41 J. D. Clements, G. L. Westbrook, Acti- 255, 1712–1715. vation kinetics reveal the number of glu- 53 F. Conti,W. Stühmer, Quantal charge tamate and glycine binding sites on the redistributions accompanying the struc- N-methyl-D-aspartate receptor, Neuron tural transitions of sodium channels, 1991, 7, 605–613. Eur. Biophys. J. 1989, 17, 53–59. 42 J. P. Changeux, S. J Edelstein, Allos- 54 Single-channel Recording, B. Sakmann, E. teric receptors after 30 years, Neuron Neher (Eds.), Plenum Press, New York, 1998, 21, 959–980. 1995. 43 G. Berecki et al., HERG Channel (Dys)- 55 F. Qin, Restoration of single-channel function Revealed by Dynamic Action currents using the segmental K-means Potential Clamp Technique, Biophys. J. method based on Hidden Markov Model- 2005, 88, 566–578. ing, Biophys. J. 2004, 86, 1488–1501. 44 W. N. Zagotta, T. Hoshi, R. W. 56 D. Colquhoun, C. J. Hatton, A. G. Aldrich, Shaker potassium channel gat- Hawkes, The quality of maximum likeli- ing. III: Evaluation of kinetic models for hood estimates of ion channel rate con- activation, J. Gen. Physiol. 1994, 103, stants, J. Physiol. (London) 2003, 547, 321–362. 699–728. 45 J. Zheng, F. J. Sigworth, Selectivity 57 D. Colquhoun, F. J. Sigworth, in Sin- Changes during Activation of Mutant gle-channel Recording, B. Sakmann, Shaker Potassium Channels, J. Gen. E. Neher (Eds.), Plenum Press, New Physiol. 1997, 110, 101–117. York, 1995, pp. 483–585. 46 E. D. Koch, B. M. Olivera, H. Terlau, 58 S. H. Chung, J. B. Moore, L. G. Xia, F. Conti, The Binding of {kappa}-Cono- L. S. Premkumar, P. W. Gage, Character- toxin PVIIA and Fast C-Type Inactivation ization of single channel currents using of Shaker K+ Channels are Mutually Ex- digital signal processing techniques clusive, Biophys. J. 2004, 86, 191–209. based on Hidden Markov Models, Philos. 47 M. Pusch et al., Mechanism of block of Trans. R. Soc. London, Ser. B Biol. Sci. single protopores of the Torpedo chloride 1990, 329, 265–285. channel ClC-0 by 2-(p-chlorophenoxy)bu- 59 F. Qin, A. Auerbach, F. Sachs, Hidden tyric acid (CPB), J. Gen. Physiol. 2001, Markov modeling for single channel ki- 118, 45–62. netics with filtering and correlated noise, 48 F. Conti, Noise analysis and single- Biophys. J. 2000, 79, 1928–1944. channel recordings, Curr. Top. Membranes 60 L. Venkataramanan, F. J. Sigworth, Transport 1984, 22, 371–405. Applying hidden Markov models to the 49 S. H. Heinemann, F. Conti, Nonsta- analysis of single ion channel activity, tionary noise analysis and application to Biophys. J. 2002, 82, 1930–1942. patch clamp recordings, Methods Enzy- 61 F. Qin, A. Auerbach, F. Sachs, Estimat- mol. 1992, 207, 131–148. ing single-channel kinetic parameters 50 C. M. Armstrong, F. Bezanilla, Cur- from idealized patch-clamp data contain- rents related to movement of the gating ing missed events, Biophys. J. 1996, 70, particles of the sodium channels, Nature 264–280. 1973, 242, 459–461. 62 F. J. Sigworth, S. M. Sine, Data trans- 51 F. Bezanilla, The voltage sensor in vol- formations for improved display and fit- tage-dependent ion channels, Physiol. ting of single-channel dwell time histo- Rev. 2000, 80, 555–592. grams, Biophys. J. 1987, 52, 1047–1054. 145

6 Automated Planar Array Electrophysiology for Ion Channel Research Derek J. Trezise

6.1 Introduction

Electrophysiological methods have long been a cornerstone of biomedical science. Basic extracellular field potential recording, for example, has proved invaluable in numerous preclinical and clinical scenarios, as wide ranging as neural network analysis in invertebrates to human electrocardiomyography. Capacitance and impedance measurements, intracellular voltage monitoring, and short circuit current detection have also been extensively used. The patch clamp electrophysiology technique, pioneered in the late 1970s [1–3], allows the direct recording of ionic currents in either a patch or the entire plasmalemmal membrane of a cell. It is capable of resolving single channel gating events on the submillisecond time scale. This exquisite sensitivity and temporal resolution has truly revolutionized ion channel functional analysis, and patch clamp electrophysiology is now well established as the gold standard method. However, the technically demanding nature of this method has restricted the extent and speed at which data can be gathered, and thus almost 30 years on, its full potential has yet to be realized. The advent of automated planar array electrophysiology has begun to address this bottleneck [4, 5]. New commercially available systems offer operator de-skilling and, in some cases, quantal increases in throughput. This chapter introduces automated planar array electrophysiology methods and work practices and aims to highlight key considerations and challenges in the emerging field. Applications ranging from ion channel reagent validation, characterization and high throughput screening will be introduced. The reader is referred elsewhere for reviews that cover automated nonplanar array electrophysiology [5–7].

6.2 Overview of Planar Array Recording

Patch clamp recordings are generally made one at a time via an electrolyte filled glass pipette (1–2 mm tip diameter) positioned on the surface of the cell with the aid

of microscopy and micromanipulation. Providing that a tight seal between the electrode tip and the cell membrane is formed, electrical access for current injection and recording can be obtained that can permit good voltage control and isolation of the biological current. Full details of this method can be found elsewhere [2, 3, 8, 9]. Fertig and co-workers [10, 11] from Nanion technologies provided the first detailed description of the replacement of the micropipette with a planar, microstructured quartz chip. Cell recording sites were created as small apertures (1 mm diameter, 3–5 MO) in the quartz substrate, by irradiation with accelerated gold ions followed by wet etching. A cell in suspension was attracted to the aperture by suction, and once a high electrical resistance seal had formed, further suction was applied to rupture the membrane to provide access to the cell interior. In this way whole cell recordings were made without cell visualization and pipette micromanipulation (see Fig. 6.1). By creating multiple recording sites in parallel on a chip, quantal increases in patch clamp throughput were promised. Another advantage was that the bioelectric properties of the planar substrates were such that extremely low electrical noise recordings could be achieved. The use of apertures in planar substrates now forms the basis for most commercially available higher throughput patch clamp systems. As of 2005, there are

Fig. 6.1 Schematic diagram of the planar brane to form the ‘whole-cell’ configuration, patch clamp principle. In step 1, cells in sus- evident by the large increase in the capaci- pension are attracted to the aperture via suc- tance spikes (step 3). After the internal solution from beneath the substrate – the meation has fully dialysed the cell, the biological sured current evoked by the voltage step signal (in this case an inward Na+ current) yields a hole resistance of 2–5 MO. Once a can be studied in the absence (step 4) and single cell finds the aperture (step 2) the re- presence (step 5) of drug (in this case a Na+ sistance markedly increases (the ‘giga-ohm’ channel blocker). seal’). Further suction ruptures the mem- 6.3 Experimental Methods and Design 147 four established multichannel systems for mammalian cells on the market. Three of these, PatchXpress7000A (Molecular Devices Corporation, www.moleculardevi- ces.com/), Q-patch16 (Sophion Biosciences, www.sophion.dk/) and NPC-16p (Na- nion technologies, www.nanion.de) use disposable chips with 16 recording sites whilst the fourth, IonWorksHT (Molecular Devices Corporation) has 384 sites per plate (the ‘Patch plate’). Whilst the precise details of the fabrication methods for these consumables have not been disclosed, it is known from patent literature that glass, polyimide/polyethylene terephthalate and silicon wafer substrates are used in conjunction with etching or laser drilling (see Ref. [5] for review). Seal rate (i.e. fraction of successful recordings), seal stability, and cost are key parameters for the consumable. The 16-channel devices have chips that form tight seals with cell membranes of 1–10 GO resistance and use suction from beneath the aperture to break into the cell for electrical access. In contrast, the Ion- WorksHT system achieves only 50–300 MO resistance seals to the substrate and uses a membrane permeabilising reagent (e.g. amphoterocin) for ‘perforated- patch’ access [12]. Overall success rates for recording range from 25–95% depending on the criteria, cell line and experimental design (see Section 6.5.4). Gener- ally, long lasting recordings (>45 min) can be made with the planar array method since vibration in the recording system does not disturb the electrical seal between the aperture and cell, as occurs with the pipette and cell membrane in conventional electrophysiology. Capacitance and series resistance artifacts arising from the substrate and cell can be fully compensated for within the software of the PatchXpress, Q-patch16 and NPC-16p instruments, akin to conventional patch clamp, but not in IonWorksHT. The major trade off for these different properties is the cost per well of the consumables – the 16-channel systems work out at somewhere between $5–$10 per well whilst the IonWorks patch plates are 50c–$1 per well.

6.3 Experimental Methods and Design

Planar array electrophysiology differs fundamentally from conventional electrophysiology in many ways, perhaps the most important of which are the cell preparation, experimental design and data analysis methods. In conventional electrophysiology cells are plated on glass coverslips or Petri-dishes and visualized by microscopy. A trained electrophysiologist will increase his/her likelihood of success by selecting healthy looking cells with a ‘clean’ cell surface and good morphology. With planar array systems this is not possible – cells are placed in suspension and recorded from ‘blind’. Often the goals of an automated electrophysiology experiment may differ from those tackled with conventional methods, creating new challenges for protocol design and analysis. This is especially true in the pharmaceutical setting where very pragmatic approaches may be taken in high(er) throughput drug screening. The ability to record from greater and greater numbers of cells per se opens up possibilities of ion channel experiments that would 148 6 Automated Planar Array Electrophysiology for Ion Channel Research

not otherwise have been attempted. These and other considerations are dealt with in the following sections.

6.3.1 Cell Preparation

To maximize the probability of generating a ‘useful’ recording at a given aperture it is essential to obtain healthy, isolated cells in suspension at an appropriate density. Optimisation of the suspension method can have a major impact on final assay quality, and the key to consistent assays is the reproducibility of these cell preparation steps. Typically, using a CHO-cell line stably expressing the ion channel of interest, cells are first grown to 80% confluence in cell culture media in a T75 flask and then washed with phosphat-buffered saline and incubated for 5 min in a cell dissociation solution (e.g. versene EDTA). The lifted cells are then centrifuged (4000 rpm for 2 min) and the cell pellet is washed again and resuspended in the external recording solution. A clean cell preparation minimizes the possibility of cell debris contacting the aperture and preventing a good recording. For an Ion- Works experiment the suspension is gently triturated for 60 s with a Gilson pipette to obtain viable, dissociated cells. Final cell densities of (0.5–2)6106 cells ml–1 are optimal. Shorter trituation times appear to work better for the 16-channel systems. For HEK cells, a trypsin-based cell dissociation solution generally works better than versene and more trituration is required to avoid cells forming clumps. Common problems are either insufficient or excessive enzymatic digestion or trituration, which can detrimentally impact both seal rate and length of the recording. In most cases the cell preparation is improved by a short settling period (3–10 min) prior to use. Of course, the creation of good cell suspensions is of little use if the fraction of cells that exhibit the ionic current of interest is unacceptably low. In the extreme, certain transient transfection methods and host cell combinations may yield <5% of expressing cells in recombinant systems (see Chapter 4). In this case, planar array recording can be disadvantageous compared to conventional electrophysiology where cell markers such as GFP can be used to guide selection [13]. Experiments with native cell preparations on dorsal root ganglion or hippocampal neurons may also be complicated by the presence of glial cells or astrocytes unless methods are applied to exclude these. In contrast, with good stable cell lines >95% of cells can be made to express the channel of interest. For this reason these have been the major focus for most automated electrophysiology work. Indeed, the higher throughput systems like IonWorksHT are extremely useful tools for validat- ing clonal cell lines per se by functional expression. Large numbers of clones (50– 100) can be grown and prepared for testing within the same day. Promising clones, based on expression criteria such as the fraction of ‘expressors’ and the median current amplitudes, can be rapidly identified and selected for more detailed profiling (see Chapter 4, Fig. 4.2). For some cell lines there is a strong relationship between the growth conditions and ion channel expression, highlighting the need for attention to detail in cell 6.3 Experimental Methods and Design 149

Fig. 6.2 Upregulation of hERG expression at or 30 8C (B). The fraction of cells with cur- low temperature, quantified by automated rents >0.1 nA increased from 74.8% to 95.8% electrophysiology. The histograms show the (P<0.001) by lowering the culture tempera- cell population distributions for peak tail cur- ture, and the mean current amplitude in- rents (measured at –40 mV) from a CHO cell creased from 0.28 nA to 0.93 nA (P < 0.001; line stably expressing hERG. Bin size 0.1 nA. n = 657 at 37 8C and 686 at 30 8C). M. X. Cells were cultured for 24 h at either 37 8C(A) Chen and D. J. Trezise, unpublished data. culture. For example, hERG K+ channel stable cell lines appear to lose expression when the confluency rises above 90%, presumably as a consequence of specific regulation on cell to cell contact or as part of the growth cycle [14,15]. Conversely, if the same cells are grown at 30 8C for 24 h prior to the experiment a marked increase in channel expression occurs (Fig. 6.2). From anecdotal observations different media, serum and even plasticware can also affect ion channel expression, although to date automated electrophysiology data has yet to be described.

6.3.2 Cell Sealing and Recording

A prerequisite for high-fidelity patch clamp is the high resistance seal that forms between the cell and the pipette tip, or the substrate in the case of planar electrophysiology. The microscopic processes that occur at this interface are not fully understood [10, 16, 17], but what is clear is that a clean contact surface is essential. Thus, the IonWorks patch plates come dry-packed and Aviva SealChips are stored in double distilled water to avoid contamination with dust or salts. It is equally important to ensure that all recording solutions that will be in contact with the patch wells are filtered (e.g. 0.22 mM). Microscopic particles can block the apertures and are a major cause of poor performance. Another common problem is blocking of 150 6 Automated Planar Array Electrophysiology for Ion Channel Research

the apertures by air bubbles, and it is strongly advised to degas solutions too. Gen- erally, prior to addition of cells and in filtered and degassed ionic buffers, >95% of holes on commercially supplied chips/plates are open (unblocked) and have acceptable resistances (1–5 MO). As with glass pipettes in conventional electrophysiology it appears that recording chips cannot be easily reused. Once positioned in the instrument, suction from beneath the substrate is applied to attract a cell to the aperture and help form the seal. Electro-positioning, whereby an electrical field is created to attract the polarized cell to the recording aperture, has been attempted but appears to offer no advantage [18, 19]. In systems such as PatchXpress and Q-patch the suction parameters are entirely user configurable and can be optimized depending on the cell type that is being used. For more fragile cells gentle suction ramps appear to work best whilst for sturdier cells rapid GO seals can be achieved with steeper ramps or steps (e.g. negative pressure ramp –10 to –32 mmHg at 1.6 mmHg s–1 for 25 s [20]). In a detailed analysis of the SealChips used with the PatchXpress device, Xu and coworkers [19] found that for CHO-Kv cells >1 GO seals could be obtained on >90% occasions. 82% of seals occurred within 15 s of the cell landing at the recording site. With HEK-hERG cells, both Dubin et al. [20] and Quinn [21] had lower success rates with GO seals on 50–60% occasions. Depending on the cell type, >1 GO seals were obtained for 40–95% of trials with a prototype of the Sophion Q-patch system [22]. For IonWorksHT, the suction parameters cannot be edited from the graphical user interface. Somewhat lower resistances are observed (100–300 MO) but >95% seal rates can routinely be achieved [12, 23]. Taken together, these and other data illustrate that cell sealing methods to apertures in planar substrates are now well established and no longer represent the technical challenge they once were. To attain the whole cell recording configuration, further suction is applied to rupture the cell membrane in the aperture. Suction to each well/cell is controlled individually so that it can be disengaged immediately after the ‘whole cell’ configuration is achieved. Excessive suction can destroy the seal and recording and this is avoided by continuous monitoring of the cell capacitance coupled to a feedback circuit to the suction control. One common problem is ‘premature breakthrough’ where membrane rupture occurs before a high resistance membrane seal is achieved [20, 21]. Overall, whole cell recordings with membrane resistances >1 GO can be achieved for between 40–60% of cells depending on the system and/or cell type [19–22, 24]. Whilst this is clearly impressive, since perhaps 6–10 whole recordings may be obtained on a 16-channel seal chip, there is scope for improvement here. It will be interesting to see whether this single cell suction method for whole cell access proves truly scalable with ever larger parallel arrays. ‘Perforated-patch’ clamp with membrane permeabilising agents such as amphoterocin, gramicidin or b-escin is arguably a more elegant solution commensurate to higher throughput planar array electrophysiology [12, 25, 26]. Cells can be treated homogeneously by exposure of the cell surface at the aperture to a fixed concentration of agent, removing any requirement for capacitance/suction feedback circuitry. Provided that sufficient time is allowed for permeabilisation, stable low access resistance pathways (<15 MO) can be consistently obtained. Indeed, in the 6.3 Experimental Methods and Design 151 authors laboratory and from published findings the success of gaining long term (>45 min) stable electrical access once a seal is achieved is >95%. For some experiments, the low permeability of the ‘perforation’ to large species and ions (e.g. Ca2+) restricts flexibility in the study of certain ion channels i.e. Ca2+-activated K+ channels. However, incomplete dialysis of the cell is generally advantageous since it preserves key regulatory pathways and components in the cell, which leads to more stable current measurements. This is a major consideration when characterizing channels that rapidly run down (e.g. voltage-gated Ca2+ channels, KCNQ K+ channels) and when conducting pharmacological studies (see Section 6.3.4). Minimising potential errors in voltage command and control and isolating extra- neous nonbiological ionic currents from the final measured signal are other prere- quisites for high quality patch clamp recordings [3, 27]. Theoretically, the challenges of signal filtering, sampling frequency, leak and offset corrections, and capacitance and series resistance compensation are no different to conventional patch clamp. Many of the amplifiers used in planar array electrophysiology are very similar or identical to those found on a conventional electrophysiology system, and the correction algorithms applied in software are the same. For the systems that treat each cell individually, criteria for several membrane parameters can be configured (e.g. access resistance <15 MO, membrane resistance >500 MO) for quality assurance. If any parameters drift outside the acceptable range corrective action such as compensation readjustment or the application of further suction is triggered. If the problem persists the recording is terminated. In practice, with planar array recordings, membrane parameters and compensation comparable to conventional electrophysiology can be achieved. In the IonWorksHT system a fundamentally different philosophy to data capture is applied. Certain error sources are deemed insignificant and are ignored (e.g. capacitance transients) whilst for others, corrections are made across the entire plate for all cells rather than on a cell by cell basis. For many experiments demanding higher throughput, especially expression profiling and drug screening this pragmatic approach is perfectly acceptable. The initial voltage difference (Voff) between the head and ground Ag/AgCl electrodes in the open circuit solutions, for example, is measured in every well and then the median value from 384 is taken and applied to all wells. Provided that the electrodes are well chlorided so that the standard deviation of VOff across the plate is low (<3 mV) in the vast majority (>95%) of the cells <10 mV error will occur. Cells with unacceptably large VOff errors (i. e. >15 mV) can be removed by post-hoc filtering. Leak correction is not performed 1 by the traditional ‘P/4’ method in which a scaled signal /4 of the size (and the same length) of the test waveform is applied 4 times and summed for subtraction from the acquired signal. Rather, the current obtained from a fixed duration and size step signal (e.g. –10 mV, 40 ms) is scaled assuming totally linear leak prior to subtraction. In practice, in most cases <50 pA of poorly corrected leak is observed. Given that comparatively low seal resistances are generally obtained (100– 300 MO), and hence the relative contribution of leak is high, this method works remarkably well. The main exception to this is when working with leak-like chan- + nels (e.g. twin-pore K channels, Kirs) where discrimination of the biological sig- 152 6 Automated Planar Array Electrophysiology for Ion Channel Research

nal from the nonbiological leak can prove troublesome. Series resistance is not compensated for in the IonWorks system, and thus large ionic currents (>2 nA) should be studied with caution to avoid unacceptable voltage errors.

6.3.3 Drug Application

There are several key considerations in the design of drug application systems for automated electrophysiology. First, how much compound sample is likely to be available? In a pharmaceutical setting, or when peptides or toxins are being studied, often only a few mL (2–200) of test compound can be afforded. Designs that incorporate low volume (10–50 mL) wells have evolved for this reason (see e.g. Ref. [28]). Next, how rapidly and for how long do drugs need to be applied? Fluid exchange times of <30 ms are required to fully resolve fast-ligand gated ion channel events such as nAChR and NMDA receptor activation, otherwise receptor desensitization prematurely curtails the signal. Short application times (1–10 s) are also desirable to avoid longer term desensitization or tachyphylaxis. In contrast, long incubation times (2–20 min) may be required for small molecule ion channel blockers to fully equilibrate or when long voltage-command protocols are being studied (e.g. steady-state inactivation curves) The third consideration is how many different compounds or concentrations of the same compound need to be applied to the same cell? Cumulative concentration-response curves for single test compounds are often constructed with conventional electrophysiology methods by applying increasing concentrations of drug immediately after the previous concentration has reached equilibrium. The long recordings (>45 min) achievable with the patch clamp method provide this option. If different test compounds are to be studied on the same cell the ability to rapidly and fully washout the previous drug is important. For studying antagonists or modulators of ligand-gated channels it is necessary to preincubate and then co-apply with the reference agonist. In practice, there is often a strong trade off in speed between a desire to maximize the data that can be captured from a single cell versus the ensuing complex logis- tics of the experimental design and data analysis (see next section). This is compounded by the unpredictability of novel drug behavior (e.g. differences in association rates and washout times). Finally, in what format are the compounds of interest likely to be supplied? If only low numbers of compounds are to be profiled this is largely of academic interest. However, in almost every drug screening environment compounds are supplied on standard geometry microtitre plates (96 well, 384 well), for which the liquid handling of the automated electrophysiology system must be compatible. With these considerations in mind, different vendors have implemented either perfusion systems based on microfluidics, exchange methods in which the external buffer can be largely replaced, or static wells in which additional volume (drug) can be added followed by a mixing step. In the Sophion Q-patch, the consumable contains a sophisticated laminar flow channel system driven by a passive capillary pump – this requires <10 mL drug and claims liquid exchange times 6.3 Experimental Methods and Design 153 commensurate with the study of most ligand-gated ion channels (e.g. <100 ms). Washout is possible and multiple drug applications can be made to the same cell. The 16 recording channels are supplied with drugs via a 4-channel pipettor which aspirates from either a 96- or 384-well drug plate. The Nanion NPC-16p also uses microfluidics for drug application and receives low volume (<15 mL) samples from a pipetting arm. Fluid exchange times of <50 ms are claimed. Seal chips used in the PatchXpress instrument have wells of larger volume (up to 100 mL) – a single pipettor is used to first aspirate buffer from the recording well and then rapidly replace it with a drug containing solution – the time to 90% exchange is several hundred milliseconds. All three systems use control software to schedule drug addition according to when cells have achieved pre-set membrane and signal quality assurance criteria. For IonWorksHT there is considerably less flexibility in the drug application system, which is based on a ‘mix and read’ method. Compounds are applied to the recording wells according to a predetermined schedule via a 12-channel fluidics- head, which aspirates from a 96- or 384- well drug plate. 3.5 mL of drug is added to the 7 mL of buffer in the well and mixed to create a 1:3 final dilution. Optimisa- tion of the mixing conditions (i.e. number of mixes, mix volume, pipettor speed) is important and can negatively impact the observed pharmacology if done incorrectly. Currently it is not possible to make more than one drug application to a given cell, or to read during the drug addition time. In its current guise, this precludes the study of even slow ligand-gated channels since the minimum time between drug addition and data acquisition is >30 s The philosophical difference here is that with so many recordings on the plate (up to 384) it is possible to accu- mulate single (drug) point data on many cells for amalgamation (e.g. to construct concentration–response curves) more quickly and robustly than it is to apply multiple concentrations or drugs to individual cells. An inventive design feature is the 1:4 transfers from the drug plate to the patch plate in which each drug well is applied to 4 different patch wells. This approach circumvents any requirement to only add drugs to ‘good cells’. If >75% of recordings are acceptable then probability theory states that >99% of the compounds from a 96-well drug plate will see one or more good cells [12, 23]). It is only since the first generation planar array automated electrophysiology systems have been used in ‘real world’ drug screening that certain issues regarding compound adherence and adsorption to recording substrates, compound plates and pipetting systems have come to light. When comparing with conventional patch clamp data, an underestimation of the potency of certain compounds has been observed. This is especially true for highly lipohilic, ‘sticky’ drugs such as astemizole and terfenadine, for which 10–100-fold drop in potency for block of hERG channels has been observed [20, 23, 24, 29]. For PatchXpress, adding the compound three times to the recording well appears to resolve some of this dis- crepancy – whilst the exact reason for this is unclear it may be that compound in the sample addition first occupies binding sites on the substrate or plastic rather than the cell. Dispensing from glass-coated rather than polypropylene drug plates goes someway toward mitigating this and can increase compound potency 3–5- 154 6 Automated Planar Array Electrophysiology for Ion Channel Research

Fig. 6.3 Effect of compound plate type on in- glass- (o) coated 96-well plates. Note the 6- hibition of Kv7.1+minK K+ currents by L- fold increase in potency and shallowing of the 735281 [36] measured using IonWorksHT.(A) Hill slope from 2.37 to 0.90 with the glass- shows mean log concentration–response coated plates. (B) shows the individual de-

curves for L-735281 obtained when the drug rived pIC50 values from 39 (glass) and 20 was plated on either polypropylene- (.)or (plastic) concentration–response curves.

fold (see e.g. Ref. [21]; Molecular Devices technical note). This has been observed both on the IonWorks and PatchXpress platforms. As the technology becomes more widely adopted it will be interesting to see whether fewer complications are observed with the Q-patch system in which the laminar flow channels are made of glass. Carry-over of compounds on the pipette tips from one well or plate to the next is another pitfall to be aware of – with some systems that use fixed tips 100% ethanol and water washes between drug additions are required to avoid this. Returning briefly to the challenge of fast drug application and the study of ligand-gated ion channels, it is fair to say that the existing platforms do not fully meet the requirements of ultrafast addition (<30 ms) coupled to significant parallelization. With a perfusion system and single cell recording, experimental design will always be complex if interactions between multiple concentrations of gating- ligand and modulator/blocker are to be studied. This is compounded by the need to make ultrashort agonist applications with rapid washout. One interesting nonplanar array automation innovation in this area is the DynaflowTM microfluidic chip [30, 31], designed to integrate with the recording stage of a conventional patch clamp set up. This device is a multi-channel (currently 48) laminar flow system that creates an array of continuous, controlled liquid environments through which a cell can be ‘scanned’. The rise time when moving from one flow channel to the next can be a few milliseconds. Preloading of the chip to create the flow channels markedly simplifies experiments with multiple combinations of ligand and modulator. Integration of this technology with a parallel patch clamp system could provide a solution to most of the challenges of electrophysiology drug screening at fast ligand-gated ion channels. 6.3 Experimental Methods and Design 155

6.3.4 Experimental Design and Data Analysis

On the whole the experimental design and analysis methods deployed in the 16- channel systems are very similar to those used in conventional electrophysiology, albeit on a larger scale. The acquisition and analysis softwares provide great flexibility and few constraints. For biophysical characterization of heterologously expressed ion channels it is relatively straightforward to script diverse voltage command protocols that can be concatenated. Once the experiment is started the instruments can gather data unattended. The same is true for simple drug addition protocols. In contrast, by virtue of the extent of the parallelization and the drug application philosophy, the IonWorksHT approach presents a new set of challenges for experimental design and analysis. These are a combination of many that are continually faced by the high throughput screening community with others more immediately recognizable to the classically trained electrophysiologist. This is best exemplified by the manner in which data are ‘quality controlled’. All patch clampers are used to applying good judgment as to whether data from a given cell are valid – cells can be rejected for many reasons such as the seal becoming unstable over time, the holding current or access resistance increasing, or the biological signal proving too small to measure reliably. In many cases this is done more subjectively than objectively. With IonWorksHT, data from >2500 cells can be gathered in a single day – far too many to go though one by one, even for the most dedicated researcher. Nevertheless, it is important to exclude poor recordings from subsequent analysis to avoid inappropriate conclusions. This is achieved by setting simple filters in the software such that, for example, any cells with seal resistances <50 MO or holding currents >50 pA are excluded. Another useful filter is on current amplitude to eliminate poorly expressing cells (i.e. peak current <100 pA). More complex ‘stability’ filters can be created such that differences between the pre-drug and post-drug membrane parameters are monitored, e.g. pre- minus post- holding current > or <50 pA. What may be less familiar are the methods used to monitor and control for time-dependent changes in the biological signal of interest, and/or effects of drug vehicles. Commonly, in conventional patch clamp an ionic current may be evoked repeatedly by a voltage step, for example, and monitored until it is stable. At this point the drug vehicle may be applied. Providing that only a small change in signal (e. g. <10%) is observed, the operator moves on to applying test compounds to the cell. The contribution of time-dependent (not drug-induced) changes in the signal (if any) is controlled either by washout of the drug back to the original baseline, or in some cases by a correction factor from extrapolation of the signal vs. time plot. With IonWorksHT, long term repeated channel gating, multiple drug addition and washout are not possible and hence a different technique is required. In the author’s laboratory, two columns of the drug plate, corresponding to 1/6 of the total wells, are used for assay plate quality control. The first (low, column 11) contains the drug vehicle, usually 0.3% final dimethylsulfoxide (DMSO) and con- 156 6 Automated Planar Array Electrophysiology for Ion Channel Research

trols any vehicle and time-dependent changes between the post- and pre-addition signals. The second (high, column 12) contains a supramaximal concentration of reference compound (e.g. a blocker of the ion channel under study) to ensure that the biological signal is not contaminated by other conductances or leak. At the end of the experiment, and after filtering of cells with poor membrane parameters (see above), a Z’ value can be calculated for the pre- and post-addition signals

using the equation Z’=1– (3 (s.d.low + s.d.high)/(meanlow–meanhigh)) [32]. Z’ values are a universally accepted statistical measure of assay quality and reflect signal dynamic range and the data variation associated with the signal measurements. As- say plates can thus be accepted or rejected based on Z’ criteria – in our hands Z’ values >0.3 are suitable for single concentration screening, and for optimized assays values of 0.6–0.8 can be routinely achieved. For concentration–response curve analysis on IonWorksHT, data must be amal- gamated from a sufficient number of cells from the plate to provide a precise

quantitation of the pIC50 (or pEC50) value of the drug of interest. The drug plate format needs to account for this and the practicalities of liquid handling for generating compound plates. Of course, a major consideration is also cost, for which a goal in a pharmaceutical setting is to find the optimum balance between throughput and data quality. We originally tried ‘cross-plate’ dilutions whereby the highest concentration of 8 different test compounds was positioned in A1, B1, C1 through to H1, and serial dilutions were made across the plate. However, row-to-row carry over with certain potent “sticky” compounds occurred which proved difficult to eliminate through tip washing. By plating different compounds at the highest test concentration in row H (1–10), and diluting 1:3 up the plate, the sequencing of the fluidics head is such that potential carry over cannot occur between compound columns (i.e. the same dispensing needle sees only a single compound, and moves from low concentrations to higher ones). Moreover, 10 rather than 8 concentration–response curves comprised of data from up to 32 cells could be obtained, with columns 11 and 12 as controls. In practice, after filtering, 20–30 cells usually remain for inclusion in the four parameter equation logistic fitting for

IC50. It is at this point that the balance between over-filtering (too few data included in the curve fit) versus under filtering (poor data is included which skews the curve fit) is most apparent. After filtering, we normalize the post: pre drug values to the high and low controls in columns 11 and 12, and then use all points, rather than the mean of the 0–4 cells obtained at each drug concentration, for curve fitting. In this way, each raw data point can be equally weighted. As with other screening methods, it is good practice to include a reference compound con-

centration–response curve for additional quality control for which if the pIC50 falls outside acceptable limits (e.g. >0.5 log units from a historical mean value) the plate is rejected. Another analysis challenge and opportunity with the higher throughput systems is kinetic determinations. Patch clamp electrophysiology provides detailed temporal information, which can be used to intepret conformational state transitions, mechanism of drug action and the function of regulatory subunits, as examples (see Chapter 5 and Ref. [33] for review). Typically, a kinetic is quantified by fit- 6.3 Experimental Methods and Design 157

Fig. 6.4 Compound plate format and curve rent amplitude. Compound wells for which fitting example for concentration–response >50% block of the signal occurred are curve analysis using IonWorksHT. A, shows a shaded, such that active compounds can be summary view from the software of the 96 readily visualized. The processed data for a well plate with ‘up-plate dilutions’ for 10 dif- single compound (column 9) is shown in B. ferent compounds arrayed at the highest con- The IC50 value is obtained by fitting a four- centration in H1, H2 etc and diluted up the parameter logistic function to the data from plate 1:3 (see text for further explanation). all of the post filtered cells, in this case 29. The values in each well indicate the number One ‘outlier’ (circled) is detected by the curve of cells that have ‘passed’ exclusion filters fitting software and is excluded from the such as low seal resistance and low peak cur- curve fit. ting a component of the current waveform to an appropriate function (e. g. single or bi-exponential decay functions often well describe the tail currents observed when voltage-gated channels deactivate). However, this process usually requires some ad hoc interaction with the data by the analyst, which, for thousands of cells, would constitute a significant computational overhead. It should also be borne in mind that any ultrafast kinetic determination is likely to be imperfect given the absence of capacitance (and series resistance) compensation. Nevertheless, depending on the question in mind, pragmatic approximations to kinetic behavior can prove extremely valuable. This is best illustrated by example. Activation of Kv7.1 (KCNQ1) K+ channels is markedly slowed when the regulatory subunit minK is co-expressed [34, 35]. In that minK is not pore forming per se (it is a single trans-membrane domain protein) the best way to detect its presence in a given cell is to use the slower activation, compared to Kv7.1 alone, as a marker. In building stable cell lines for drug screening we confirmed the presence of minK by setting simple measures at a fixed time point (150 ms) after the gating step and at full activation (4 s). The ratio of these two values (I150 :IPeak) was markedly lower (<0.4) in the presence than in the absence of minK (>0.6), in- dicative of slower activation. The power of this approach is that we could rapidly and confidently determine, based on a population analysis (n>200 cells), what fraction of cells within the Kv7.1+minK stable cell line were not expressing the minK subunit (i. e. exhibited a high I150 :IPeak value). We found that <2% of KCNQ1 expressing cells had rapid kinetics and were therefore low minK expressors (see Fig. 6.5). 158 6 Automated Planar Array Electrophysiology for Ion Channel Research

Fig. 6.5 Kinetic estimations to determine the as measured ‘classically’ by fitting single expo- presence of regulatory subunits in cell lines nential functions to individual cells (n=8) and using the IonWorksHTelectrophysiology plat- also by using simple metrics and population form. (A) shows data from a Kv7.1 CHO cell line statistics (see text for full explanation). Abs- and (B) shows data from a Kv7.1 line stably cissae: ratio of current amplitude at 150 ms to transfected with the minK subunit. Expression the peak current. Ordinate: cell count bin size of minK markedly slowed the activation of Kv7.1 0.01. The time/amplitude bar is 1 s and 0.5 nA.

As noted earlier, the IonWorks system cannot simultaneously voltage clamp and apply drugs. The 48-channel electronics (E-) head must move away from the wells, transiently unclamping the cells, whilst the fluidics (F-) head dispenses and mixes. For certain voltage-gated channels, the impact of this transient unclamping requires consideration in the experimental design. Channels that are strongly inactivated at unclamped membrane potentials (e.g. –40 to –5 mV) may need a long recovery period once clamping resumes. Certain drugs with strongly voltage-dependent binding may also block more potently during the unclamped (depolarized) phase and may only slowly unblock upon repolarization. Other compounds may require prolonged repeated channel gating (‘use-dependence’) to block fully, which is not possible during the mixing phase. Despite these challenges, with understanding and care it has proved possible to configure robust, meaningful assays for most members of the voltage-gated super family attempted so far.

6.4 Overall Success Rates and Throughput

As has been highlighted, there are a number of reasons why a given planar array recording may prove unsuccessful or yield unsatisfactory data. Blocked apertures, unhealthy cells, premature whole cell access, inappropriate or unstable mem- 6.5 Population Patch Clamp 159 brane parameters, small or unstable ionic currents, or slow drug washout can all, either individually or in combination, reduce the final output. Of course, the ‘success’ criteria themselves will be strongly dependent on the experimental objective. Entire plates, rather than individual cells might be failed based on Z’ statistics or the discrepant pharmacology of a standard compound in a drug screening mode. There is currently little published data on the overall attrition that these factors contribute, and less still on the overall throughput benefit afforded by the current systems. For sure, the ‘real world’ does not always match the vendor claims. For the PatchXpress system, overall success rates of 25–36% were obtained for compound screening at the hERG K+ channel [20, 21, 24]. This particular assay comprised stringent QC criteria around membrane parameters, current stability and, in the Dubin study, the activity of an internal control compound on each cell. There was no compromise in the data quality compared to conventional electrophysiology and 3–4 point IC50 values were generated from each cell. The overall increase in throughput was 3- to 4-fold. For this type of study, the other 16-channel systems (Sophion Q patch, Nanion NPC-16p) are likely to offer similar efficiency gains. For screening cell lines, biophysical characterisation and for less demanding pharmacological protocols 5–10-fold increases may be obtainable. With IonWorksHT, 8 plate runs of 45 min each are easily achievable in a working day, corresponding to 86384 = 3072 attempted recordings. For most cell or compound profiling applications with high expressing cell lines, using success criteria of cells >50 MO, Ihold <50 pA, peak current > 200 pA and a plate Z' value of >0.3 for stability, useful data can be gathered on >300 cells per run, or >2400 cells per day ([12] and Trezise and Dale, unpublished data). If the application is screening cell clones for expression this equates to >100-fold efficiency gain compared to conventional patch clamp. For compound profiling, once the 1:4 transfers, data amalgamation across cells and control wells are accounted for, 640 single test compounds or 64 10-point concentration–response curves can be generated within a day. This corresponds to a throughput increase of 30–80-fold. Whilst certain detailed biophysical analyses may expose the limitations in working with low seal resistances and without capacitance or Rs compensation, for most screening applications this trade off for the quantal throughput increase is perfectly toler- able. Indeed, for the first time, this level of throughput is sufficient to support iterative structure–activity relationship screening by electrophysiology for a modern medicinal chemistry drug discovery program.

6.5 Population Patch Clamp

In an exciting new development in the field, Jan Hughes and Alan Finkel at Mole- cular Devices have shown that planar array electrophysiology methods can be applied to multiple cells in a single well, a principle termed ‘population patch clamp’ (PPC). In modified IonWorks plates with 64 apertures microfabricated in each well, robust ensemble ionic currents were observed with Kv1.5, NaV1.5 and hERG 160 6 Automated Planar Array Electrophysiology for Ion Channel Research

Fig. 6.6 Screening performance for a voltage- the average number of usable currents per gated Na+ channel assay using IonWorksHT to plate and peak current amplitude. In D the Z’

exemplify assay stability and quality control. value and pIC50 value for a standard com- Each panel shows data taken from 64 inde- pound added to each plate are shown. Plate pendent runs (assay plates), from which dif- 62 was the only failure based on the QC cri-

ferent variables were monitored. A, shows the teria of Z’ < 0.3 and standard pIC50 value number of blocked wells and the number of within ±0.5 of the rolling mean value (i.e. good seals (or recordings) from a possible 4.7 ± 0.5). T. J. Dale and D. J.Trezise, unpub- 384. B, shows the median seal resistances at lished data. different time points in the experiment and C,

expressing stable cell lines. By theory, with resistors in parallel, the inverse of the measured resistance across the entire well will simply be the sum of the inverse of

the resistance of the individual apertures (1/Rmeasured = S 1/Rindividual). Provided that a high fraction of the apertures are occupied by cells (which display high resistance) the ‘short-circuit’ contribution of the unoccupied (low resistance) wells becomes insignificant. This relationship is modeled in Fig. 6.7. For robust leak correction seal resistances >80 MO are required, which corresponds to an aperture occupancy of >95%, assuming a single cell resistance of 120 MO. The measured currents become the sum of the individual currents arising from each cell, and can be simply scaled (i.e. divided by 64) to provide quasi-average single cell currents. As far has been determined, ionic currents measured using this approach do not markedly differ biophysically or kinetically from single cell recordings. For certain applications in the screening arena, this simple idea of PPC offers enormous potential advantages over single cell automated electrophysiology. The 6.5 Population Patch Clamp 161

Fig. 6.7 Principle of population patch clamp- fraction of occupied holes and the measured ing with IonWorks Quattro. Recordings are resistance across the well. For these resis- made from a population of cells on planar tance parameters in the model >95% occu- substrates with multiple apertures in the pancy is required for a final equivalent resis- same well. These behave as resistors in paral- tance of >75 MO (C), which is adequate for lel (A). Assuming that any individual aperture acceptable leak subtraction and voltage con- is either occupied or not, with respective re- trol. Remade from Molecular Devices Cor- sistances of 120 MO and 10 MO, the equation poration with permission. in (B) describes the relationship between the requirement to build cell lines in which >90% of cells express sufficient channels of interest will become less stringent. As long as the mean current is measurable it will be of less importance if a higher fraction of cells are low expressors. Consis- tency of data from well to well should also be markedly improved, as the impact of cell to cell variability is minimized by data averaging. Indeed, with optimized assays a good signal should be obtained in every well, as is the case with the vast majority of plate readers (e.g fluorescence and luminescence). Compounds can now be added 1:1 to assay wells rather than 1:4 as in the original IonWorksHT platform yielding an immediate 4-fold increase in drug screening throughput. With high quality assays it may be possible to dispense with the pre-drug read altogether (since they will all be very similar) thus shortening plate read times significantly. Channel multiplexing may also be easier – by mixing cell lines that express ion channels with different phenotypes (e.g. a voltage-gated Na+ channel and voltage-gated K+ channel) two (independent) signals could be resolved in the same well, without the need to generate a single cell line that expressed both channels (see Chapter 4). This approach is very powerful for selectivity profiling, and further increases the ‘value’ from each well screened. Overall, PPC should bring automated electrophysiology ever closer to the true ‘mix and read’ screening paradigm that is so desirable in high throughput screening. 162 6 Automated Planar Array Electrophysiology for Ion Channel Research

6.6 Summary and Perspective

Automated planar array recording is now an established variant of patch clamp electrophysiology, offering operator de-skilling and higher throughput. Its primary use so far has been for rapid drug screening in the pharmaceutical and biotechnology sectors where it has proved truly enabling. Undoubtedly, as the capital and operational costs fall it will become more widely adopted in the academic community where more fundamental questions in the ion channel field will be addressed. As with any new technology, challenges and considerations arise with use, some foreseen and others completely unexpected. The major differences between automated planar array recording and conventional patch clamp lie in the cell preparation, drug application and experimental design and analysis. With understanding, thought and care it has proved possible in several drug screening laboratories to exploit the potential of the first generation commercially available instruments. A major area of current unmet need is for integrated fast drug application coupled to a quantal increase in throughput – this will hopefully be addressed in the next generation of instruments. Excitingly, population patch clamp, which extends the patch clamp method to a multicell recording paradigm, should bring further improvements in quality and throughput. In the medium term, automated planar array electrophysiology should help fulfil the remaining untapped potential of the patch clamp method, and accelerate the search for new ion channel therapeutics.

Acknowledgments

The author gratefully acknowledges the support and input of colleagues at GSK, notably Tim Dale and Claire Townsend, who have made significant contributions to the deployment of high throughput electrophysiology methods.

References

1 Neher E., Sakmann B., Single-channel 4 Sigworth F. J., Klemic K. G., Patch currents recorded from membrane of clamp on a chip, Biophys. J. 2002, 82(6), denervated frog muscle fibres, Nature 2831–2832. 1976, 260(5554), 799–802. 5 Wang X., Li M., Automated electrophy- 2 Hamill O. P., Marty A., Neher E., siology: high throughput of art, Assay Sakmann B., Sigworth F. J., Improved Drug Dev. Technol. 2003,1(5), 695–708. patch-clamp techniques for high-resolu- 6 Wood C.,Williams C.,Waldron G. J., tion current recording from cells and Patch clamping by numbers, Drug Dis- cell-free membrane patches, Pflugers covery Today 2004, 9(10), 434–441. Arch. 1981, 391(2), 85–100. 7 Trezise, D. J., Patch clamp electrophy- 3 Sakmann B., Neher E., Patch clamp siology steps up a gear, Eur. Pharm. Rev. techniques for studying ionic channels 2005, 2, 69–73. in excitable membranes, Annu Rev. Phy- siol. 1984, 46, 455–472. (Review.) References 163

8 Cahalan M., Neher E., Patch clamp 19 Xu J., Guia A., Rothwarf D., techniques: an overview, Methods Enzy- Huang M., Sithiphong K., Ouang J., mol. 1992, 207, 3–14. Tao G.,Wang X.,Wu L., A benchmark 9 Molleman, A., Patch Clamping: an In- study with sealchip planar patch-clamp troductory Guide to Patch Clamp Electro- technology, Assay Drug Dev. Technol. physiology, John Wiley & Sons Ltd, 1992. 2003, 1(5), 675–684. 10 Fertig N., Blick R. H., Behrends 20 Dubin A. E. et al., Identifying modula- J. C.,Whole cell patch clamp recording tors of hERG channel activity using the performed on a planar glass chip, Bio- PatchXpress planar patch clamp, J. Bio- phys. J. 2002, 82(6), 3056–3062. mol. Screen. 2005, 10(2), 168–81. 11 Fertig N., George M., Klau M., 21 Quinn, C., Rapid ICE (Rapid Ion Chan- Meyer C., Tilke A., Sobotta C., Blick nel Electrophysiology): an ion channel R. H., Behrends J. C., Microstructured screening service using the PatchXpress apertures in planar glass substrates for 7000A system, Presentation at Molecu- ion channel research, Receptors Chan- lar Devices Electrophysiology User’s nels 2003, 9(1), 29–40. Meeting, Long Beach CA, February 12 Schroeder K., Neagle B., Trezise D. 2005. J.,Worley J., Ionworks HT: a new high- 22 Kutchinsky J. et al., Characterization throughput electrophysiology measure- of potassium channel modulators with ment platform, J. Biomol. Screen. 2003, QPatch automated patch-clamp technol- 8(1), 50–64. ogy: system characteristics and perfor- 13 Marshall J., Molloy R., Moss G. W., mance, Assay Drug Dev. Technol. 2003, Howe J. R., Hughes T. E., The jellyfish 1(5), 685–93. green fluorescent protein: a new tool 23 Kiss L., Bennett P. B., Uebele V. N., for studying ion channel expression Koblan K. S., Kane S. A., Neagle B., and function, Neuron 1995, 14(2), 211– Schroeder K., High throughput ion- 215. channel pharmacology: planar-array- 14 Crociani O., Guasti L., Balzi M., based voltage clamp, Assay Drug Dev. Becchetti A.,Wanke E., Olivotto M., Technol. 2003, 1(2), 127–35. Wymore R. S., Arcangeli A., Cell 24 Guo L., Guthrie H., Automated elec- cycle-dependent expression of HERG1 trophysiology in the preclinical evalua- and HERG1B isoforms in tumor cells, tion of drugs for potential QT prolonga- J. Biol. Chem. 2003, 278(5), 2947–2955. tion, J. Pharmacol. Toxicol. Methods 15 Cherubini A. et al., Human ether-a-go- 2005, in press. go-related gene 1 channels are physi- 25 Finkelstein A., Holz R., Aqueous cally linked to {beta}1 integrins and pores created in thin lipid membranes modulate adhesion-dependent signal- by the polyene antibiotics nystatin and ing, Mol. Biol. Cell 2005, 16(6), 2972– amphotericin B, Membranes 1973, 2, 2983. 377–408. 16 Corey D. P., Stevens, C. F., Science 26 Fan J. S., Palade P., Perforated patch and technology of patch recording elec- recording with beta-escin, Pflugers Arch. trodes, in Single Channel Recording, 1998, 436(6), 1021–1023. Sakmann, B., Neher, E. (Eds.), Plenum 27 Armstrong C. M., Gilly W. F., Access Press, New York 1983, pp. 53–86. resistance and space clamp problems 17 Opsahl L. R.,Webb W.W., Lipid-glass associated with whole-cell patch clamp- adhesion in giga-sealed patch-clamped ing, Methods Enzymol. 1992, 207, 100– membranes, Biophys. J. 1994, 66(1), 75– 122. 79. 28 Bruggemann A., George M., Klau M., 18 Worley J., Li M., Location is key – Beckler M., Steindl J., Behrends recent progress in single-cell-based J. C., Fertig N., High quality ion chan- high-throughput assays, Drug Discovery nel analysis on a chip with the NPC Today 2001, 6(9), 454. technology, Assay Drug Dev. Technol. 2003, 1(5), 665–673. 164 6 Automated Planar Array Electrophysiology for Ion Channel Research

29 Tao H. et al., Automated tight seal elec- says, J. Biomol. Screen. 1999, 4(2), 67– trophysiology for assessing the potential 73. hERG liability of pharmaceutical com- 33 Hille, B., Ion Channels of Excitable pounds, Assay Drug Dev. Technol. 2004, Membranes, Sinauer Verlag, Sunderland 2(5), 497–506. MA, 2001. 30 Sinclair J., Pihl J., Olofsson J., 34 Barhanin J., Lesage F., Guillemare Karlsson M., Jardemark K., Chiu E., Fink M., Lazdunski M., Romey G., D. T., Orwar O., A cell-based bar code K(V)LQT1 and lsK (minK) proteins as- reader for high-throughput screening of sociate to form the I(Ks) cardiac potas- ion channel-ligand interactions, Anal sium current, Nature 1996, 384(6604), Chem. 2002, 74(24), 6133–6138. 78–80. 31 Olofsson J., Pihl J., Sinclair J., 35 Sanguinetti M. C., Curran M. E., Sahlin E., Karlsson M., Orwar O., A Zou A., Shen J., Spector P. S., Atkin- microfluidics approach to the problem son D. L., Keating M. T., Coassembly of creating separate solution environ- of K(V)LQT1 and minK (IsK) proteins ments accessible from macroscopic vo- to form cardiac I(Ks) potassium chan- lumes, Anal Chem. 2004, 76(17), 4968– nel, Nature 1996, 384(6604), 80–83. 4976. 36 Seebohm G., Chen J., Strutz N., 32 Zhang J. H., Chung T. D., Olden- Culberson C., Lerche C., Sangui- burg K. R., A Simple Statistical Para- netti M. C., Molecular determinants of meter for Use in Evaluation and Valida- KCNQ1 channel block by a benzodiaze- tion of High Throughput Screening As- pine, Mol. Pharmacol. 2003, 64(1), 70– 77. 165

7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels Georg C. Terstappen

7.1 Introduction

Ion channels are pore-forming membrane proteins which enable the rapid passage (flux) of ions across cell membranes. Their ion conductivity is often highly specific and has been used for their general classification into sodium, potassium, calcium, chloride and nonselective cation channels. As opposed to active transport by membrane pumps such as the Na+/K+ ATPase, ion channels allow only passive transport of ions along a concentration gradient. Their opening and closing (‘gating’) is regulated by a range of different stimuli including transmembrane voltage, ligand binding, mechanical stress and temperature. The first two stimuli are the most common and therefore these membrane proteins are broadly grouped into voltage-gated and ligand-gated ion channels. Ion channels are central to many biological and disease processes and are particularly important for regulating electrical properties of excitable cells such as neurons and myocytes. In many other cell types they contribute to important physiological processes such as hormonal secretion and blood pressure regulation. Although ion channels constitute a complex gene family one common feature is a pore-forming region which determines ion selectivity and mediates ion flux across cell membranes. The recent sequencing of the human genome has revealed around 400 pore-forming ion channel genes corresponding to about 1.3% of the human genome [1]. These pore-forming ion channel subunits (a-subunits) contain a minimum of 2 trans-membrane domains, as in the case of the inward rectifying K+ channel Kir, and up to 24 transmembrane domains, as in the case of voltage-gated Na+ and Ca2+ channels. Some K+ channels contain two pore-forming regions in tandem. Additional complexity is generated since functional ion channels are often homo- or heteromeric protein complexes which can co-assemble with accessory (b- and further) subunits, thus creating a vast number of physiological ion channel complexes with different functions and pharmacology. A significant number of disease relevant ion channels have been identified and in conjunction with novel assay technologies [2] and mechanistic insights into channel function the development of selective and state-dependent drugs is on the horizon.

7.2 Ion Flux Assays

Activation of ion channels leads to a movement of charged molecular species across the cell membrane. This ion flux along a concentration gradient leads to changes in membrane potential that can be quantified by the Nernst equation (Fig. 7.1). If ions that pass through the channel under study can be radiolabeled, radioactive flux assays can be developed for functional analysis and screening of these ion channels. If such ions are not available, or the use of radioactive isotopes needs to be avoided, ions that pass through the channel can also be analyzed by atomic absorption spectrometry, a technique traditionally used for the detection of trace elements in environmental, biological and medical samples. In any case, a cellular system is necessary that either natively or recombinantly expresses the ion channel of interest. Typically, mammalian cells such as HEK293 or CHO- K1 are employed for recombinant expression of ion channels although some channels need other cellular ‘backgrounds’ for proper functional expression [3] (see Chapter 4). Since functional ion flux assays represent a direct measure of channel activity, they are robust and insensitive to disturbances. Compared to electrophysiological methods which can be considered the ‘gold standard’ for functional analysis of ion channels, their temporal resolution is limited to the sec-

Fig. 7.1 In cellular systems activation of ion can be determined by measurement of radio- channels leads to ion flux along a concentra- activity or atomic absorption spectrometry, re- tion gradient and concomitant changes in spectively. Both measurements represent the transmembrane potential which can be quan- basis for the development of functional as- tified by the Nernst equation (shown on the says to monitor ion channel activation. right). If radioactively labeled ions are added Shown free intra- and extracellular ion con- or if K+ or Na+ ions are exchanged with Rb+ or centrations have been taken from Ref. [12]. Li+ (see text), changes in ion concentrations 7.2 Ion Flux Assays 167 onds/minutes range and the membrane potential cannot be controlled precisely. Thus, these assays cannot be employed for screening of bona fide state-dependent ion channel modulators.

7.2.1 Radioactive Ion Flux Assays

The application of radioactive isotopes of ions that pass through the channel under study – and thus can serve as tracers for these ions in cellular assay systems – has long been used. Radioactive isotopes of the naturally conducting ion species, such as 22Na+ [4], 45Ca2+ [5] and 36Cl– [6], can be employed as tracers as can other radioactive ion species which are conducted by the channel. For instance, 86Rb+ has been used for the study of potassium and nonselective cation channels [7] and 14C-guanidinium for analysis of sodium channels [4]. Based on transmembrane concentration gradient and ion conductivity (Fig. 7.1) influx of radiotracer is usually measured for sodium, calcium and chloride channels, whereas efflux is measured for potassium channels upon activation. Cells expressing the ion channel of interest are typically grown in standard cell culture compatible microplates. Voltage-gated channels are activated by adding a ‘depolarizing’ concentration of 650 mM KCl to the cell medium whereas other channels are, for instance, activated by adding an appropriate concentration of ligand. Measurement of radioactivity using standard equipment is either carried out in the cell supernatant, the cell lysate or both matrices by direct Cerenkov counting (e. g. 86Rb) or liquid scintillation counting (e.g. 45Ca). Measurements of both matrices allow calculation of the relative flux of radiotracer thus eliminating potential well-to-well differences in cell densities and tracer loading. A homogeneous radioactive ion flux assay format is possible using CytoStar-T scintillating technology (GE Healthcare) employing b-emitting isotopes. Based on the principles of scintillation proximity assay (SPA) technology (see Section 7.3) each well of a special microplate is coated at the bottom with scintillant that will only detect radioisotopes in close vicinity to it. Thus, if cells are cultured as mono- layers on these plates and channels are activated in the presence of radiotracer, the influx or efflux can be measured as a change of light using any plate-based scintillation counter. With this technique, no wash steps and no separation of initially applied radiotracer are necessary thus making this assay format more amenable to the requirements of high throughput screening (HTS). A few examples of measuring influx of 14C-guanidinium for sodium channels [8], influx of 45Ca2+ for ionotropic glutamate receptors [9] and efflux of 86Rb+ for potassium channels [10] have been described. A major advantage of radioactive ion flux assays is that no special apparatus is necessary if the laboratory is equipped for the measurement of radioactivity, which is usually the case in both academic and industrial environments. The main disadvantage of flux assays is the use of radioisotopes which is associated with significant costs, safety hazards and environmental (e.g. disposal) problems. This is the 168 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

main reason why radioactive ion flux assays, which were frequently used in the 1990s in the pharmaceutical industry, have largely been abandoned and replaced by nonradioactive alternatives such as nonradioactive ion flux assays based on atomic absorption spectrometry (Section 7.2.2) and fluorescence-based assays (see Chapter 8).

7.2.2 Nonradioactive Ion Flux Assays based on Atomic Absorption Spectrometry

Atomic absorption spectrometry (AAS) is a well established technology, traditionally used for the detection of trace elements in environmental, biological and medical samples, that uses thermal energy to generate free ground state atoms in a vapor phase which absorb light of a specific wavelength. In practice, atomi- zation is achieved by spraying a sample containing the element to be measured into the flame of the burner of an atomic absorption spectrometer. Absorption of light, which is typically emitted by hollow cathode lamps, is measured with a photomultiplier (Fig. 7.2). Thus, an atomic absorption spectrometer can be imagined as a photometer where the cuvette is replaced by a burner generating the flame (”flame photometry”). The law of Lambert–Beer–Bouger applies and can be employed to determine the concentration of an element by measuring its absorption. In practice, however, this is usually done by comparing the light absorption of a sample with a standard curve obtained under identical experimental conditions.

Fig. 7.2 Schematic diagram of an atomic absorption spectrometer. For details refer to the text.

7.2.2.1 Nonradioactive Rubidium Efflux Assay An AAS-based rubidium efflux assay for functional analysis of potassium and nonselective cation channels was established in the 1990s [11]. Rubidium is an al- kali metal with atomic number 37 and an ionic radius of 1.61 Å which is not present in eukaryotic cells. Its similarity to K+ leads to a high permeability in potassium and nonselective cation channels [12]. It can easily be detected by using atomic absorption spectrometry with a sensitivity (‘characteristic concentration’) of 0.11 mg l–1 measuring absorption at 780 nm. 7.2 Ion Flux Assays 169

In general, the experimental protocol for a nonradioactive Rb+ efflux assay consists of two parts, cell biology and physical determination of the tracer rubidium by AAS. First, cells expressing the ion channel under study are cultured in cell compatible microplates and loaded with rubidium by simply exchanging potassium in a cell compatible buffer solution with the same concentration of rubidium. This loading phase, which is usually finished within 2–4 h, can be inhibited by the cardiac glycoside oubain, pointing to the involvement of Na+/K+-AT- Pases in transporting Rb+ into the cells. Prior to starting efflux experiments it is necessary to remove excess Rb+ by a series of quick wash steps with buffer containing KCl. The frequency and buffer volumes used for these wash steps mainly depend on the cell type, cell density, microplate formats and washing devices employed and should be optimized on a case-by-case basis since appropriate removal of excessive Rb+ is essential in order to obtain good signal-to-background ratios. Activation of the ion channel under study leads to Rb+ efflux into the cell supernatant due to the established concentration gradient for this tracer ion (see also Fig. 7.1). For voltage-gated potassium channels activation can be achieved by adding a depolarizing concentration of KCl (typically 650 mM) to the cells and for ligand- gated channels by adding an appropriate concentration of ligand. The incubation time with the channel activator has to be optimized empirically in order to achieve optimal efflux results but in most cases a period of 610 min is sufficient. When compounds are screened for channel blocking effects they should be added prior to channel activation (e.g. 10 min) because of kinetic considerations. Cell supernatants which contain the ‘effluxed’ Rb+ are removed and collected along with the cell lysates. Both of these Rb+-containing matrices can be stored at room temperature prior to AAS analysis which is not disturbed by cell debris. In principle, rubidium determinations can be carried out with any type of flame atomic absorption spectrometer and measurements should be carried out according to the instructions of the manufacturer. The recent development of an innovative AAS instrument for ion channel analysis (ICR 12000, Aurora Biomed Inc., Vancouver, Canada), featuring a sophisticated microsampling process utilizing 96- or 384-well microplates and simultaneous measurements of 12 samples at a time, allows the measurement of up to 60 000 samples per day (http://www.aurorabiomed.com/ICR12000.htm), making the nonradioactive Rb+ efflux assay compatible with the throughput requirements of HTS in drug discovery. Typically, the relative amount of rubidium in the supernatant is calculated as [Rb in supernatant/Rb in supernatant + Rb in cell lysate] thus eliminating potential well-to-well differences in cell densities and Rb+ loading. This relative rubidium efflux is a robust and direct measure of ion channel activity and a 6twofold increase of Rb+ efflux upon channel activation over basal efflux levels is usually sufficient for the configuration of good quality HTS assays [13] since the standard deviations for rubidium measurements by AAS are low [11]. If sample throughput needs to be increased, under highly standardized experimental conditions it might be possible to measure rubidium only in the supernatant.

The following assay protocol for the analysis of calcium-activated BKCa channels [14] stably expressed in mammalian CHO-K1 cells was successfully used for func- 170 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

tional selection of stable recombinant clones expressing this ion channel and subsequent screening for the identification of channel modulators. This protocol can serve as a basis for the development of such assays for other potassium and nonselective cation channels. Cells are grown at 37 8C in 96-well cell culture compatible microplates for 48 h to a final cell density of about 16104 cells per well in standard cell culture medium. After aspirating the medium, 0.2 ml cell buffer containing RbCl is added

(5.4 mM RbCl, 150 mM NaCl, 2 mM CaCl2, 0.8 mM NaH2PO4, 1 mM MgCl2, 5 mM glucose, 25 mM HEPES, pH 7.4) and cells are incubated for 4 h at 37 8C. Cells are then quickly washed three times with buffer (same as above, but containing 5.4 mM KCl instead of RbCl) to remove extracellular Rb+. Subsequently, 0.2 ml buffer containing a saturating concentration of 25 µM of the Ca2+ specific

ionophore A23187 (Fig. 7.3) is added to the cells in order to activate BKCa channels via a ‘molecular Ca2+ injection’ and after incubation for 10 min the supernatant is carefully removed and collected for rubidium measurements. Cells are lysed by the addition of 0.2 ml 1% Triton X-100 and also collected for rubidium determinations. AAS measurements are carried out with a flame atomic absorption spectrometer following the instructions of the manufacturer. The stimulated relative Rb+ efflux [Rb in supernatant/Rb in supernatant + Rb in cell lysate] with these recombinant cells amounts to 80% (Fig. 7.4), which represents a five-fold increase over basal conditions. The specificity of the induced Rb+ efflux is further demonstrated

by the use of the specific BKCa channel ligand iberiotoxin isolated from the scorpion Buthus tamulus [15] which blocks the channel in a concentration-dependent

manner with an IC50 of 15 nM (Fig. 7.5). If this protocol is used for the analysis of other potassium or nonselective cation channels, channel activation and specificity analysis have to be adapted appropriately. In addition to the above mentioned example, a screening assay was developed

for blockers of BKCa channels (Table 7.1) recombinantly expressed in HEK293 cells [16]. The diphenylurea analogue NS1608 was used to activate the channels

Fig. 7.3 Activation of BKCa channels recombinantly expressed in CHO- K1 cells with increasing concentrations of the calcium-specific ionophore A23187 leads to increasing Rb+ efflux. For details refer to the text. 7.2 Ion Flux Assays 171

Fig. 7.4 The Rb+ efflux induced with 25 mM of the ionophore A23187 is not observed in CHO-K1 cells which were

used to generate the recombinant BKCa channel expressing cell line. For details refer to the text.

Fig. 7.5 Rb+ efflux induced with 25 mM of the ionophore A23187 in recombinant

CHO-K1 cells expressing BKCa channels is inhibited in a concentration-dependent manner with the selective ligand iberiotoxin. For details refer to the text.

leading to a three- to four-fold Rb+ efflux which was completely blocked by the specific ligand iberiotoxin (IC50 = 12 nM). A pharmacological profile obtained with a series of known openers and blockers of BK channels compared very well with results obtained with a radioactive 86Rb efflux assay [16], demonstrating the utility of the nonradioactive Rb+ efflux assay for high throughput screening campaigns as well as SAR studies. SK3 channels (Table 7.1) are one of three members of small conductance calcium-activated potassium (SKCa) channels [17] which are all activated by submi- cromolar increases in intracellular Ca2+ concentration mediated by calmodulin [18]. SK3 channels were recombinantly expressed in HEK293 cells [19] and the above described assay protocol was employed, activating channels by increasing intracellular Ca2+ concentrations with thapsigargin, an inhibitor of endoplasmatic Ca-ATPase which leads to a release of Ca2+ from intracellular stores (Fig. 7.6). Cal- cium activation of SK3 channels led to a two- to three-fold Rb+ efflux which could 172 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

Table 7.1 Published examples of recombinant ion channels that have been analyzed employing nonradioactive Rb+ efflux assay technology.

Voltage-gated Ca2+-activated Ligand-gated nonselective K+ channels K+ channels cation channels

Kv1.1 SKCa (SK3) Nondisclosed nonselective cation channel

Kv1.3 BKCa Kv1.4 Kv1.5 Kv7.2 (KCNQ2) Kv7.2/3 (KCNQ 2/3) Kv11 (hERG)

Fig. 7.6 Recombinant HEK293 cells expressing SK3 channels are activated by thapsigargin, an inhibitor of endoplasmatic Ca- ATPase which leads to release of Ca2+ from intracellular stores. Rb+ efflux as measure of channel activity is completely blocked by the SK channel selective ligand apamin. For details refer to the text.

be completely blocked with the specific SK channel ligand apamin, a peptide present in bee venom toxin. Nonradioactive Rb+ efflux assays were developed for several recombinantly expressed voltage-gated potassium channels (Table 7.1). In the case of Kv1.1 and Kv1.4 channels expressed in HEK293 cells, activation with 50 mM KCl led to a less than twofold Rb+ efflux in 10 min, which was blocked by the nonspecific potassium channel blocker TEA [11]. The relatively low KCl-induced Rb+ efflux which was attributed to low expression levels and/or inactivation properties of the 7.2 Ion Flux Assays 173 channels was not sufficient to configure a robust screening assay. More recently, an assay was described for Kv1.3 channels expressed in CHO-K1 cells (http:// www.aurorabiomed.com/New-Pro/CHO_Kv13_CellLine.pdf). Depolarization with 63 mM KCl led to a fourfold Rb+ efflux in 15 min which could be blocked with an

IC50 value of 0.66 nM by Agitoxin-2, a peptide blocker isolated from the venom of the scorpion Leiurus quinquestriatus. Although no further details were disclosed, the generation of a Kv1.5 expressing cell line for the establishment of a Rb+ efflux assay was noted by Merck Research Laboratories at an ion channel conference in 2004 (http://www.aurorabiomed.com/retreat2004.htm). For Kv7.2 (KCNQ2) channels stably expressed in HEK293 cells stimulation with 50 mM KCl led to a four- fold Rb+ efflux [20]. Calculated Z’ factors [13] were 0.73 for a 96-well plate format and 0.6 for a 384-well format, respectively, indicating the high suitability for screening and SAR studies. The pharmacological profile of Kv7.2 defined by electrophysiology was faithfully reflected by the Rb+ efflux assay which allowed measuring 1000 data points per day in a 96-well plate format [20]. For the identification of heteromeric Kv7.2/3 (KCNQ2/3) channel (M-current) modulators a recombinant CHO-K1 cell line expressing these channels was employed [21]. Channels were activated with 20 mM KCl in the presence of the channel opener Way-1 and an average Z’ value of 0.81 for the 96-well format was calculated from a total of 20 experiments. A throughput of about 40 compounds per day for obtaining EC50 values (8-point curves) was achieved if AAS determinations of rubidium were only carried out in cell supernatants which gave results consistent with calculating relative Rb+ efflux by measuring rubidium contents in both supernatants and cell lysates. Specificity was demonstrated by using the known M-current blocker linopir- + dine which inhibited Rb efflux with an IC50 of 2.85 mM, a value in close agreement with results obtained from electrophysiological analysis. The voltage-gated potassium channel Kv11 (hERG) seems to be particularly susceptible to inhibition by many xenobiotics and drugs leading to potentially lethal arrhythmias [22]. In fact, several drugs have recently been withdrawn from the market due to hERG channel activity. Thus, in drug discovery hERG channel liability of novel compounds is a major concern. A nonradioactive Rb+ efflux assay was developed using hERG channels stably expressed in CHO-K1 cells [23]. Channels were activated by addition of 50 mM KCl for 10 min which resulted in an about two-fold Rb+ efflux. Although the signal-to-background ratio was relatively low, a Z’ value of 0.53 was calculated for a 96-well plate format thus meeting HTS stan- dards. A pharmacological characterization employing a series of known hERG channel blockers (dofetilide, terfenadine, sertindole, astemizole, cisapride) showed the same rank order of potency as electrophysiology. Absolute IC50 values were 5–20-fold higher when compared to electrophysiological results obtained with mammalian cells but were similar to data in Xenopus oocytes. These results indicate the suitability of the Rb+ efflux assay for hERG compound profiling. A similar recombinant cell line was also employed by AstraZeneca who disclosed results at an ion channel conference in 2003 (http://www.aurorabiomed.com/ main-1.htm). In this case, a four-fold Rb+ efflux was measured in a 384-well plate format after 30 min incubation with 50 mM KCl. The calculated Z’ value was 174 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

60.5 utilizing the ICR 12000 atomic absorption spectrometer (Aurora Biomed, Vancouver, Canada). Studies on native ligand-gated nonselective cation channels (nicotinic acetyl choline receptors and purinergic P2X receptors) in PC12 cells have also been described [11]. The development and application of nonradioactive Rb+ efflux assays for such recombinantly expressed ion channels has, however, yet to be fully described. In a recent presentation from Amgen at an ion channel conference in 2003a Rb+ efflux assay for a non-disclosed ligand-gated cation channel expressed in CHO-K1 cells was overviewed (http://www.aurorabiomed.com/main-1.htm). Exposing the recombinant cells to 3–10 mM of a nondisclosed agonist resulted in an about four-fold + Rb efflux. This efflux was blocked by a nondisclosed antagonist with an IC50 of 344 nM which was in very good agreement with electrophysiological results, again demonstrating the reliability of such functional ion flux assays.

7.2.2.2 Nonradioactive Lithium Influx Assay Due to the high sensitivity of AAS for the determination of Li+ ions (0.035 mg l–1), influx experiments for screening of sodium channels which display a high conductivity for Li+ [12] should also be possible. Although no data have been published yet in scientific journals, at an ion channel conference in 2003 promising results obtained with SH-SY5Y cells were presented by AstraZeneca (www.aurorabiomed.- com/main-1.htm). Cells were differentiated for 3–5 days with retinoic acid which induces the expression of tetrodotoxin-sensitive voltage-gated sodium channels. Cells were washed with buffer in which NaCl was replaced with choline chloride in order to remove free Na+. After incubation for 10 min at 37 8C in wash buffer, cells were treated with 5.4 mM KCl (basal) or 120 mM KCl (depolarization) in buffer containing LiCl for 15 min. After three wash steps at room temperature, cells were lysed with 1% Triton X-100 and cell lysates analyzed for Li+ concentrations with AAS. Un- der these conditions, a three-fold Li+ influx over basal levels was obtained which could be completely blocked by preincubation for 5 min with 1 mM tetrodotoxin. Re- sults obtained with recombinantly expressed sodium channels are awaited.

7.2.2.3 Nonradioactive Chloride Influx Assay A more indirect application of AAS-based methods for analysis of ion channels was briefly noted for the investigation of chloride channels [24]. In this case, Cl– flux is measured after precipitating these ions with silver nitrate as AgCl and determining free silver by AAS. The utility of this indirect method remains to be established.

7.2.2.4 Conclusions Taken together, nonradioactive ion flux assays based on AAS have largely dis- placed radioactive flux assays in drug discovery [25]. These assays represent a direct measure of channel activity, are HTS compatible if special equipment is used 7.3 Ligand Binding Assays 175 and do not require radioactive isotopes. To date, their application has mainly been limited to the functional analysis and screening of potassium and nonselective cation channels. As compared to the ‘gold standard’ electrophysiology their temporal resolution is relatively low (seconds–minutes) and the transmembrane potential cannot be controlled precisely. Compared to other screening technologies such as fluorescence-based methods, these robust assays are less prone to identifying ‘false positive’ hits in drug screening programs and are thus highly reliable.

7.3 Ligand Binding Assays

Binding assays employing radiolabeled ligands have a long tradition in the study of receptors. For the investigation of ion channels, in particular for drug screening purposes, these assays were frequently used in the 1980s/1990s before more information-rich functional cell-based assays became available. The use of radiolabeled neurotoxins has revealed the existence of multiple binding sites for sodium channels [26] whereas the use of radiolabeled glibenclamide has been instrumental for the discovery of first and second generation sulfonylurea compounds (blockers of

KATP channels) for treatment of type-2 diabetes [27]. More recently, radioligand binding assays have also been employed to investigate a potential hERG (Kv11) channel activity of novel compounds (see above) which might lead to potentially lethal arrhythmias [22]. The known hERG channel blockers dofetilide and MK- 499 were used in tritiated form and as a 35S analogue, respectively, to identify a hERG liability of compounds [28]. For the configuration of radioligand binding assays a radiolabeled ligand binding to the ion channel under study is necessary. Table 7.2 summarizes major ion channel radioligands that have been described to date. Furthermore, a source of the channel is also needed which can be organs, tissues or cells. Although binding assays can be performed with living cells cultured in microplates [3], usually cell membranes are prepared and utilized since they can be stored at –80 8C and are much easier to handle. In saturation experiments the affinity (Kd) of a radioactive ligand for the ion channel is determined whereas in competition (displacement) experiments the affinity of unlabeled ligands (Ki) is measured [29]. For screening purposes, which aim at the identification of novel chemical entities acting on ion channels, displacement experiments are carried out which can identify compounds that bind to the same site as the radioligand or sites allosterically coupled to it. Thus, such assays do not provide bona fide information about the effects of compounds on channel function. Usually the concentration of a compound necessary to displace 50% of the radioligand (IC50 value) is determined as a measure of the compound’s affinity for the channel. If Ki values are necessary, these can be calculated from IC50 values by using the Cheng–Prusoff equation [30]. Despite their ease and compatibility with the requirements of HTS, such ligand binding assays suffer from the need for radioisotopes such as 3H and 125I and the associated costs, safety hazards and environmental (e. g. disposal) problems. 176 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

Table 7.2 Major ion channel radioligands.

Channel Type Radioligand Ref. /Current

calcium

3 Cav 1L [H]-isradipine 37–39 [3H]-devapamil [3H]-diltiazem [3H]-DTZ323

125 Cav 2P/Q[ I]-o-conotoxin GVIA 40–42 N [125I]-o-conotoxin MVIIC

sodium

3 Nav [ H]-saxitoxin 43–49 [3H]-batrachotoxin [125I]-scorpion toxins [3H]-tetrodotoxin [3H]-brevetoxin [3H]-PbTx-3

potassium

Kv 1 Shaker-related [125I]-DT 50, 51 [125I]-BgK [125I]-DTX [125I]-HgTX1 [125I]-MgTx

Kv 11 erg [3H]-astemizole 52, 53 [3H]-dofetilide

Kir 6 ATP-sensitive [3H]-glibenclamide 54, 55 potassium [125I]-glibenclamide channel [125I]-A312110 [3H]-PKF217–744

KCa 1 BK [125I]-charybdotoxin 56–58 [125I]-iberiotoxin [19F]-BMS204352

KCa2 SK [125I]-apamin 59

nicotinic AChR [3H]-nicotine 60–63 [3H]-epibatidine [3H]-cytisine [3H]-MLA [3H]-bungarotoxin [3H]-tetracaine [3H]-TCP [3H]-ethidium [14C]-amobarbital [125I]-TID 7.3 Ligand Binding Assays 177

Table 7.2 (continued)

Channel Type Radioligand Ref. /Current glutamate NMDA [3H]-MK801 64

AMPA [3H]-AMPA 65–67 [3H]-LY395153 [3H]-Ro48–8587

Kainate [3H]-kainic acid 68, 69 [3H]-NBQX

[3H]-L-glutamate 70, 71 [3H]-CPP

3 GABAA [ H]-BIDN 72, 73 [3H]-muscimol [3H]-SR 95531 [3H]-flunitrazepam [3H]-zolpidem [3H]-Ro151788(flumazenil) [3H]-TBPS [3H]-Ro15–4513 [3H]-indiplon

5HT3 [3H]-zacopride 74 [3H]-BRL 43694 [3H]-GR65630 [3H]-LY278584 glycine [3H]-strychnine 75

Although many fluorescent-labeled ligands for ion channels, in particular peptide ligands, are commercially available or could easily be synthesized, to date ligand binding assays for screening purposes have largely been limited to using radiolabels due to the often reduced affinity of fluorescent-labeled ligands.

7.3.1 Heterogeneous Binding Assays Employing Radioligands

Since ion channels are integral membrane proteins, a suitable ion channel containing membrane preparation has to be obtained. Nowadays, recombinant cell lines expressing the ion channel under study are often employed due to the high ion channel densities that can be achieved. After incubating the membrane preparation with a high affinity radioligand (^nM), which should exhibit a high specific radioactivity (630 Ci mmol–1), until equilibrium has been reached (typically minutes to hours), channel-bound radioligand is quickly separated from free radioligand, usually by filtration and washing. For screening assays typically glass 178 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

fiber filter-mounted 96- or 384-well plates are utilized which retain ion channel- bound radioactivity on the filter. Subsequently, filter-bound radioactivity is measured using scintillation or g-counting, depending on the isotope used. If compounds are screened in displacement experiments they are usually added to the channel preparation during the incubation phase. Plotting bound radioactivity (e.g. bound cpm) against the concentration of displacing compound allows easy

calculation of an IC50 value for this compound as measure of its affinity for the channel. The GraphPad Prism software (GraphPad Software Inc., San Diego) which was developed for analysis of radioligand binding is most useful for data analysis and visualization. The bee venom peptide toxin apamin is a high affinity ligand of small conductance calcium-activated potassium (SK) channels which are expressed in the cortex of rat brain [19]. The following protocol which was successfully applied for the characterization of SK channels can also serve as the basis for the configuration of other ligand binding assays with a filtration step for removing free ligand. All steps of tissue preparation were performed at 4 8C unless otherwise indicated. Cerebral cortex from rats was homogenized in 10 volumes of ice cold buffer solution [0.32 M sucrose, 5.4 mM KCl and 25 mM HEPES pH 7.0 supplemented with a protease inhibitor cocktail (CompleteTM, Roche)] using a glass/teflon homogeni- zer (1000 g min–1 for 12 cycles). The homogenate was centrifuged at 1000g for 10 min and the supernatant re-centrifuged at 48 000g for 20 min. The resultant pellet was washed twice with the above solution without sucrose. The final pellet was re-suspended in this buffer and the membrane suspension divided in aliquots and stored frozen at –80 8C until further use. [125I]-Apamin binding experiments were performed in 20 mM HEPES, 5.4 mM KCl, 0.2% BSA (pH 7.4) using 200 pM radioligand (specific activity 81.4 TBq mmol–1) and 70 µg protein equivalent of rat cortex membrane preparations in a total volume of 500 µl per test tube. Nonspecific binding was determined in the presence of an excess of 1 µM unlabeled apamin. For displacement experiments the respective concentrations of compounds to be tested were added to the test tubes. After incubation for 60 min at 4 8C unbound radioligand was removed by rapid filtration through glass fiber filters (Whatman GF/C filters, pre-soaked in 0.3% polyethyleneimine for 1 h) utilizing a cell harvester (Brandel) followed by three wash steps with ice-cold buffer. Filters were placed in scintillation vials with 3.5 ml scintillation cocktail (Filter CountTM, Packard) and bound radioactivity was measured using a b-counter (Packard).

7.3.2 Homogeneous Binding Assays Employing Radioligands

In order to avoid the separation of bound from free radioligand and associated necessary wash steps, a homogeneous assay format was developed which is more compatible with the requirements of HTS, largely due to reduced complexity of the experimental protocol (‘mix & measure’). This scintillation proximity assay (SPA) format [31] is based on solid microspheres (‘beads’) containing scintillant 7.3 Ligand Binding Assays 179 which are chemically modified at their surfaces to enable the coupling of molecules (Fig. 7.7). A commonly used bead type for applications with receptors and ion channels contains the lectin wheat germ agglutinin (WGA) at the surface which immobilizes membrane preparations by binding to glycosyl residues [32]. If a specific radioligand is added, it will bind to the ion channel contained in the immobilized membrane fraction and hence its emitted radiation will be in close enough proximity to activate the scintillant. The resulting emission of light around 400 nm (Fig. 7.7) can be measured with a scintillation counter. The energy released from unbound free radioligand is absorbed by the aqueous environment before it reaches the bead and hence does not activate the scintillant (Fig. 7.7). Since b-particles emitted by 3H and Auger electrons released from 125I have very short pathlengths in aqueous environments (<1 mm and about 17 mm), radioligands labeled with these isotopes are best suited for these assays. Due to their homogeneous format such assays can easily be automated and adapted to 384- well and higher density microplate formats which makes it especially useful for HTS. One disadvantage of ‘homogeneity’ is assay interference of test compounds due to quenching, which leads to reduced scintillation counting efficiency and hence reduced assay signals. In particular, yellow colored compounds will lead to quenching of the blue light emitted from the beads. A ‘second generation’ type of beads (SPA Imaging Beads, GE Healthcare) which contain a different scintillant that produces a red-shifted signal has been developed in order to avoid such quenching problems. The protocol described below was routinely employed for the characterization of SK channels (Fig. 7.8) and serves as a basis for the development of other SPAs. Cell membranes were prepared from recombinant HEK293 cells stably expressing SK3 channels [18] by detaching cells cultured in 175 cm2 T-flasks by PBS/EDTA treatment, homogenization with a Polytron at full speed for 3 bursts of 10 s in 25 mM HEPES, pH 7.4 supplemented with 5.4 mM KCl (SPA buffer) and centrifugation for 30 min at 48000 g. Membrane pellets were re-suspended in SPA buffer (1–2 ml per 175 cm2 T-flask) and membrane suspensions stored as aliquots at

Fig. 7.7 Schematic diagram of SPA technology. Left: Binding of radioligand to an ion channel immobilized at the bead surface leads to light emission via the scintillant incorporated in the beads. Right: Free radioligand is not in close enough proximity to the beads and hence does not stimulate light emission. 180 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

–80 8C until further use. Binding experiments were carried out in 96-well microplates in a final volume of 200 ml using 10 pM [125I]-apamin and 15 mg per well of SK3 channel containing membrane preparation. Each well also contains 1 mg of WGA-coated SPA beads (GE Healthcare). Nonspecific binding was determined in the presence of 1 mM unlabeled apamin. For displacement experiments the respective concentrations of compounds to be tested were added to the wells. After 15 min of incubation at room temperature with gentle shaking, the microplates were left to stand over night. Subsequently, bound radioactivity was measured with a b-counter (TopCount, Packard).

Fig. 7.8 [125I]-apamin binding to membranes prepared from recombinant HEK293 cells expressing SK3 channels analyzed employing an SPA format. (A) total binding (TB) and nonspecific binding (NSB) measured in the presence of 1 mM unlabeled apamin, (B) the concentration-dependent displacement of [125I]- apamin binding to SK3 channels by the antidepressant drug fluoxetine.

7.3.3 Homogeneous Binding Assays Employing Fluorescent-Labeled Ligands and Fluorescence Polarization

Since bound and free ligands show differences in molecular rotation, binding assays employing fluorescent-labeled ligands can be configured using fluorescence polarization. This read-out system is a well established analytical technique in the field of diagnostics [33]. The principle of this method is based on the observation that excitation of a fluorophore with polarized light leads to the emission of light, which will retain the initial degree of polarization depending on the rotation that occurred during the fluorescence lifetime, typically on the nanosecond time scale 7.3 Ligand Binding Assays 181

[34]. At constant temperature and viscosity, the rotational relaxation time of a molecule is directly proportional to its molecular volume. Hence, if a fluorescent- labeled ligand binds to a macromolecular receptor the increase in molecular volume and concomitant decrease in rotation result in an increase in fluorescence polarization which can be measured. Advances in appropriate instrumentation over the last ten years have enabled the configuration of assays employing 96- and 384-well microplate formats. A fluorescence polarization reader can be imagined as a standard filter fluorometer with the addition of polarizing filters for the generation and detection of polarized fluorescence light. Samples are excited with polarized light and the emitted light is passed through both horizontal and vertical polarizing filters, prior to detection with a photomultiplier. Hence, the degree of polarization of the emitted light is determined in form of a ratiometric measurement, thus eliminating interferences such as ‘inner filter effects’. Although di- mensionless numbers are the result of such measurements, the unit P (polarization unit) was introduced for convenience and fluorescence polarization readers usually present numbers as milli P (mP). Since fluorescence polarization is a homogeneous ‘mix and measure’ technology, it can easily be automated. For screening purposes assays are usually carried out in the form of competition (displacement) binding experiments (see above), determining the decrease in fluorescence polarization as a function of the concentration of competing compounds to be identified. Although significant improvements in the sensitivity of instrumentation have been achieved, high expression levels of receptors (61 pmol (mg protein)–1) and high ligand binding affinities

(Kd ^ 10 nM) are still necessary in order to configure robust fluorescence polarization assays. To date binding assays employing fluorescence polarization have not been widely used for the analysis of ion channels [35]. The main reason for  this is that fluorescent labels such as fluorescein, Bodypy , Texas RedTM, Oregon  Green and Rhodamine RedTM are quite bulky chemical entities whose attachment to ligands often results in steric hindrance and concomitant reduction of binding affinity [36]. Moreover, coupling chemistry of such fluorescent labels for peptide ligands is quite advanced whereas considerable efforts might be necessary to create fluorescent derivatives of small organic molecules. In addition, since a substantial fraction of the fluorescent ligand needs to be bound in order to configure a robust fluorescence polarization assay, this might lead to significant ligand depletion and thus affect measured absolute IC50 values [35].

7.3.4 Conclusions

For primary drug screening purposes aimed at identifying novel active chemical molecules, ligand binding assays have largely been abandoned in favor of functional cell-based assays (see above and Chapters 6 and 8). The main reason for this is the limited information content of such assays which allows no distinction between agonists and antagonists. Moreover, compounds identified in binding assays will typically reflect the mode of action of the ligand used. Thus, compounds 182 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

with novel mechanisms of channel modulation cannot be readily identified. A more technical limitation relates to the fact that for many ion channels there are no selective high-affinity ligands. Nowadays, ligand binding assays are more often employed as secondary assays in screening cascades in order to compare the molecular pharmacology of new compounds with known ligands of the channel. Furthermore, the mode of action of new compounds can be investigated if they can be labeled and used as ligands in binding assays themselves. This might also support the molecular and structural characterization of the ion channel under study since new active binding sites might be identified. The latter aspects may become increasingly important as more ion channel structures and novel modulators are discovered, and guaran- tee that ligand binding assays will remain an important tool for ion channel analysis.

Acknowledgements

The author would like to thank Dr. Renza Roncarati (Sienabiotech S.p.A., Discov- ery Research) for assistance in generating Table 7.2 and helpful discussions during the preparation of the manuscript.

References

1 J. C. Venter et al., The sequence of the ide channels in cultured cerebral neu- human genome, Science 2001, 291, rons, J. Biol. Chem. 1984, 259, 1753– 1304–1351. 1757. 2 G. C. Terstappen, Ion channel screen- 7 S. W. Weir, A. H. Weston, The effects ing technologies today, Drug Discovery of BRL 34915 and nicorandil on electri- Today: Technologies 2005, 2, 133–140. cal and mechanical activity and on 86Rb 3 C. Virginio, A. Giacometti, L. Aldegheri, efflux in rat blood vessels, Brit. J. Phar- J. M. Rimland, G. C. Terstappen, Phar- macol. 1986, 88, 121–128. macological properties of rat alpha7 ni- 8 J. Denyer, J. Worley, B. Cox, cotinic receptors expressed in native G. Allenby, M Banks, HTS approaches and recombinant cell systems, Eur. J. to voltage-gated ion channel drug dis- Pharmacol. 2002, 445, 153–161. covery, Drug Discovery Today 1998, 3, 4 M. Willow, E. A. Kuenzel, Catterall, 323–332. Inhibition of voltage-sensitive sodium 9 A. Cushing, M. J. Price-Jones, channels in neuroblastoma cells and sy- R. Graves, A. J. Harris, K. T. Hughes, naptosomes by the anticonvulsant D. Bleakman, D. Lodge, Measurement drugs diphenylhydantoin and carbama- of calcium flux through ionotropic glu- zepine, Mol. Pharm. 1984, 25, 228–234. tamate receptors using Cytostar-T scin- 5 C. A. Rasmussen, J. L. Sutko,W. H. tillating microplates, J. Neurosci. Barry. Effects of ryanodine and caffeine Methods 1999, 90, 33–36. on contractility, membrane voltage, and 10 J. W. Ford, E. B. Stevens, J. M. calcium exchange in cultured heart Treherne, J. Packer, M. Bushfield, cells, Circ Res. 1987, 60, 495–504. Potassium channels: gene family, thera- 6 K. G. Thampy, E. M. Barnes Jr., peutic relevance, high-throughput Gamma-aminobutyric acid-gated chlor- References 183

screening technologies and drug dis- and SK3, Neuroscience Lett. 2003, 346, covery, Prog. Drug Res. 2002, 58, 133–68. 85–88. 11 G. C. Terstappen, Functional analysis 20 C. W. Scott, D. E. Wilkins, S. Trivedi, of native and recombinant ion channels D. J. Crankshaw, A medium-through- using a high-capacity nonradioactive put functional assay of KCNQ2 potas- rubidium efflux assay, Anal. Biochem. sium channels using rubidium efflux 1999, 272, 149–155. and atomic absorption spectroscopy, 12 B. Hille, Ionic Channels of Excitable Anal. Biochem. 2003, 319, 251–257. Membranes, Sinauer Associates, Sunder- 21 K. Wang, et al.,Validation of an atomic land, MA, 1992 absorption rubidium ion efflux assay 13 J-H. Zhang, T. D. Y. Chung, for KCNQ/M- channels using the Ion K. R. Oldenburg, A simple statistical Channel Reader 8000, Assay Drug Dev. parameter for use in evaluation and Technol. 2004, 2, 525–534. validation of high throughput screening 22 J. I. Vandenberg, B. W. Walker,T. J. assays, J. Biomol. Screen. 1999, 4, 67–73. Campbell, HERK K+ channels: friend 14 M. Garcia-Calvo, H-G. Knaus, O. B. and foe. Trends Pharmacol. Sci. 2001, 22, McManus, K. M. Giangiacomo, 240–246. G. Kaczorowski, M. L. Garcia, Purifi- 23 W. Tang, J. Kang, X. Wu, D. Rampe, cation and reconstitution of the high- L. Wang, H. Shen, Z. Li, D. Dunning- conductance, calcium-activated potas- ton, T. Garyantes, Development and sium channel from tracheal smooth evaluation of high throughput func- muscle, J. Biol. Chem. 1994, 269, 676– tional assay methods for hERG potas- 682. sium channel, J. Biomol. Screen. 2001, 6, 15 V. K. Gribkoff, et al., Effects of chan- 325–331. nel modulators on cloned large-conduc- 24 L. Stankovich, D. Wicks, S. Des- tance calcium-activated potassium potovski, D. Liang, Atomic absorption channels, Mol. Pharmacol. 1996, 50, spectroscopy in ion channel screening, 206–217. Assay Drug Dev. Technol. 2004, 2, 569– 16 A. S. Parihar, D. R. Groebe,V. E. 574. Scott, J. Feng, X-F. Zhang, 25 G. C. Terstappen, Nonradioactive Rb+ U. Warrior, M. Gopalakrishnan, efflux assay and its applications in drug C-C. Shieh, Functional analysis of discovery and development, Assay Drug large conductance Ca2+ activated K+ Dev. Technol. 2004, 2, 553–559. channels: ion flux studies by atomic 26 W. A. Catterall,V. Trainer, D. G. Ba- absorption spectrometry, Assay Drug den, Molecular properties of the so- Dev. Technol. 2003, 1, 647–654. dium channel: a receptor for multiple 17 M. Kohler, B. Hirschberg, C. T. neurotoxins, Bull. Soc. Path. Exp. 1992, Bond, J. M. Kinzie, N. V. Marrion., 85, 481–485. J. Maylie, J. P. Adelman, Small-con- 27 J. L. Challinor-Rogers, D. C. Kong, ductance Ca2+ activated K+ channels M. N. Iskander, G. A. McPherson, from mammalian brain, Science 1996, Structure-activity relationship of gliben- 273, 1709–1714. clamide analogs: A comparison of po- 18 X. M. Xia, et al., Mechanism of Ca2+ tency as levcromakalim antagonists in gating in small conductance Ca2+-acti- rat aorta vs. affinity for [3H]-glibencla- vated K+ channels, Nature 1998, 395, mide binding to membranes from rat 503–507. cerebral cortex, J. Pharmacol. Exp. Ther. 19 G. C. Terstappen, A. Pellacani, 1995, 273, 778–86. L. Aldegheri, F. Graziani, 28 K. Finlayson, L. Turnbull, C. T. C. Carignani, G. Pula, C. Virginio, January, J. Sharkey, J. S. Kelly, [3H]- The antidepressant fluoxetine blocks Dofetilide binding to HERK in trans- the human small conductance calcium- fected membranes: A potential high activated potassium channels SK1, SK2 throughput preclinical screen, Eur. J. Pharmacol. 2001, 430, 147–148. 184 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

29 D. B. Bylund, M. L. Toews, Radioligand toxin-sensitive human N-type calcium binding methods: practical guide and channel, Science 1992, 257, 389–395. tips, Am. J. Physiol. 1993, 265, L421– 41 P. Favreau, et al., A new omega-cono- 429. toxin that targets N-type voltage-sensi- 30 R. Cheng, G. Prusoff, Relationship tive calcium channels with unusual spe-

between the inhibition constant (Ki) cificity, Biochemistry 2001, 40, 14567– and the concentration of inhibitor 14575. which causes 50 per cent inhibition 42 R. J. Lewis, et al., Novel omega-conotox-

(I50) of an enzymatic reaction, Biochem. ins from Conus catus discriminate Pharmacol. 1973, 22, 3099–4006. among neuronal calcium channel sub- 31 N. D. Cook, Scintillation proximity as- types, J. Biol. Chem. 2000, 275, 35335– say: a versatile high-throughput screen- 35344. ing technology, Drug Discovery Today 43 J. W. West, T. Scheuer, L. Maechler, 1996, 1, 287–294. W. A. Catterall, Efficient expression of 32 F. Graziani, A. Aldegheri, G. C. rat brain type IIA Na+ channel subunits Terstappen, High throughput scintilla- in a somatic cell line, Neuron 1992, 8, tion proximity assay for the identifica- 59–70. tion of FKBP-12 ligands, J. Biomol. 44 J. C. Rogers,Y. Qu, T. N. Tanada, Screen. 1999, 4, 3–7. T. Scheuer,W. A. Catterall, Molecular 33 W. B. Dandliker,V. A. de Saussure, determinants of high affinity binding of Fluorescence polarization in immuno- ù-scorpion toxin and sea anemone toxin chemistry, Immunochemistry 1970, 7, in the S3-S4 extracellular loop in do- 799–828. main IV of the Na+ channel subunit, J. 34 P. Perrin, Polarisation de la lumière de Biol. Chem. 1996, 271, 15950–15962. fluorescence, J. Phys. Radium 1926, 7, 45 S. Cestèle,Y. Qu, J. C. Rogers, 390–401. H. Rochat, T. Scheuer,W. A. Catter- 35 M. Allen, J. Reeves, G. Mellor, High all,Voltage sensor-trapping: enhanced throughput fluorescence polarization: a activation of sodium channels by homogeneous alternative to radioligand o-scorpion toxin bound to the S3-S4 binding for cell surface receptors, J. Bio- loop in domain II, Neuron 1998, 21, mol. Screen. 2000, 5, 63–69. 919–931. 36 P. Wu, M. Brasseur,W. Schindler, A 46 R. S. Sheldon, N. J. Cannon, H. J. high throughput STAT binding assay Duff, Binding of [3H]-batrachotoxinin using fluorescence polarization, Anal. A benzoate to specific sites on rat car- Biochem. 1997, 249, 29–36. diac sodium channels, Mol. Pharmacol. 37 H. Glossmann, J. Striessnig, Molecu- 1986, 30, 617–623. lar properties of calcium channels, Rev. 47 M. Taouis, R. S. Sheldon, R. J. Hill, Physiol. Biochem. Pharmacol. 1990, 114, H. J. Duff, Cyclic AMP-dependent reg- 1–105. ulation of the number of [3H]-batracho- 38 A. Koschak, D. Reimer, I. Huber, toxinin benzoate binding sites on rat M. Grabner, H. Glossmann, J. Engel, cardiac myocytes, J. Biol. Chem. 1991,

J. Striessnig, Cav1.3 subunits can form 266, 10300–10304. L-type calcium channels activating at 48 A. J. Bourdelais, S. Campbell, negative voltages, J. Biol. Chem. 2001, H. Jacocks, J. Naar, J. L. Wright, 276, 22100–22106. J. Carsi, D. G. Baden, Brevenal is a nat- 39 M. Hagiwara, S. Adachi-Akahne, ural inhibitor of brevetoxin action in so- T. Nagao, High-affinity binding of [3H]- dium channel receptor binding assays, DTZ323 to the diltiazem-binding site of Cell. Mol. Neurobiol. 2004, 24, 553–63. L-type Ca2+ channels, Eur. J. Pharmacol. 49 M. Yotsu-Yamashita, K. Nishimori, 2003, 466, 63–71. Y. Nitanai, M. Isemura, A. Sugimoto, 40 M. E. Williams, et al., Structure and T. Yasumoto, Binding properties of functional expression of an o-cono- [3H]-PbTx-3 and [3H]-saxitoxin to brain membranes and to skeletal muscle References 185

membranes of puffer fish Fugu pardalis cific activity radioligand for high-con- and the primary structure of a voltage- ductance calcium-activated potassium gated Na+ channel alpha-subunit channels, Biochemistry 1997, 36, 1943– (fMNa1) from skeletal muscle of F. par- 1952. dalis, Biochem. Biophys. Res. Commun., 58 D. O. Kiesewetter, E. M. Jagoda, J. E. 2000, 267, 403–412. Starrett Jr,V. K. Gribkoff, P. Hewa- 50 O. Shamotienko, S. Akhtar, C. wasam, N. Srinivas, D. Salazar,W. C. Sidera, F. A. Meunier, B. Ink, Eckelman, Radiochemical synthesis M. Weir, J. O. Dolly, Recreation of and biodistribution of a novel maxi-K

neuronal Kv1 channel oligomers by ex- potassium channel opener, Nucl. Med. pression in mammalian cells using Biol. 2002, 29, 55–59. Semliki Forest virus, Biochemistry 1999, 59 M. Hugues, D. Duval, P. Kitabgi, 38, 16766–16776. M. Lazdunski, J. P. Vincent, Prepara- 51 J. Racape, A. Lecoq, R. Romi-Lebrun, tion of a pure monoiodo derivative of J. Liu, M. Kohler, M. L. Garcia, the bee venom neurotoxin apamin and A. Menez, S. Gasparini, Characteriza- its binding properties to rat brain sy- tion of a novel radiolabeled peptide se- naptosomes, J. Biol. Chem. 1982, 257, lective for a subpopulation of voltage- 2762–2769. gated potassium channels in mamma- 60 D. J. Anderson, S.P Arneric, Nicoti- lian brain, J. Biol. Chem. 2002, 277, nic receptor binding of [3H]cytisine, 3886–3893. [3H]-nicotine and [3H]-methylcarbamyl- 52 P. J. Chiu, et al.,Validation of a [3H]-as- choline in rat brain, Eur. J. Pharmacol. temizole binding assay in HEK293 cells 1994, 253, 261–267. expressing HERG K+ channels, J. Phar- 61 D. Gnadisch, E. D. London, P. Terry, macol. Sci. 2004, 95, 311–319. G. R. Hill, A. G. Mukhin, High affi- 53 K. Finlayson, A. J. Pennington, J. S. nity binding of [3H]-epibatidine to rat Kelly, [3H]-dofetilide binding in brain membranes, Neuroreport 1999, 10, SHSY5Y and HEK293 cells expressing a 1631–1636. HERG-like K+ channel, Eur. J. Pharma- 62 H. A. Navarro, D. Zhong, P. Abra- col. 2001, 412, 203–212. ham, H. Xu, F. I. Carroll, Synthesis 54 R. Davis-Taber, et al., [125I]A-312110, a and pharmacological characterization of novel high-affinity 1,4-dihydropyridine [125I]-iodomethyllycaconitine ([125I]- ATP-sensitive K+ channel opener: char- iodo-MLA). A new ligand for the alpha acterization and pharmacology of bind- 7 nicotinic acetylcholine receptor, J. ing, Mol. Pharmacol. 2003, 64, 143–153. Med. Chem. 2000, 43, 142–145. 55 P. W. Manley, C. Loffler-Walz, 63 H. R. Arias, J. R. Trudell, E. Z. Bayer, U. Russ, A. Hambrock,T. Moenius, B. Hester, E. A. McCardy, M. P. Blan- U. Quast, Synthesis and characteriza- ton, Noncompetitive antagonist bind-

tion of a novel tritiated KATP channel ing sites in the torpedo nicotinic acetyl- opener with a benzopyran structure, Br. choline receptor ion channel. Structure- J. Pharmacol. 2001,133, 275–85. activity relationship studies using ada- 56 M. Hanner, R. Vianna-Jorge, mantane derivatives, Biochemistry 2003, A. Kamassah.,W. A. Schmalhofer, 42, 7358–7370. H. G. Knaus, G. J. Kaczorowski, 64 F. Murray, J. Kennedy, P. H. Hutson, M. L. Garcia, The beta subunit of the J. Elliot, I. Huscroft, K. Mohnen, high conductance calcium-activated po- M. G. Russell, S. Grimwood, Modula- tassium channel. Identification of resi- tion of [3H]-MK-801 binding to NMDA dues involved in charybdotoxin binding, receptors in vivo and in vitro, Eur. J. J. Biol. Chem. 1998, 273, 16289–16296. Pharmacol. 2000, 397, 263–270. 57 A. Koschak, R. O. Koch, J. Liu, G. J. 65 L. Bunch, T. H. Johansen, H. Brau- Kaczorowski, P. H. Reinhart, M. L. ner-Osborne,T. B. Stensbol, T. N. Jo- Garcia, H. G. Knaus, [125I]-Iberiotoxin- hansen, P. Krogsgaard-Larsen, D19Y/Y36F, the first selective, high spe- U. Madsen, Synthesis and receptor 186 7 Ion Flux and Ligand Binding Assays for Analysis of Ion Channels

binding affinity of new selective GluR5 71 R. H. Porter, R. S. Briggs, P. J. Ro- ligands, Bioorg. Med. Chem. 2001, 9, berts, Modulation of [3H]3-((+-)-2-car- 875–879. boxypiperazin-4-yl)propyl-1-phosphonic 66 A. M. Linden, H. Yu, H. Zarrin- acid ([3H]CPP) binding by ligands act- mayeh,W. J. Wheeler, P. Skolnick, ing at the glycine and the polyamine Binding of an AMPA receptor potentia- sites of the rat brain NMDA receptor tor ([3H]-LY395153) to native and re- complex, Eur. J. Pharmacol. 1992, 227, combinant AMPA receptors, Neurophar- 83–88. macology 2001, 40, 1010–1018. 72 J. J. Rauh, et al., Effects of [3H]-BIDN, 67 V. Mutel, et al., Binding characteristics a novel bicyclic dinitrile radioligand for of a potent AMPA receptor antagonist GABA-gated chloride channels of in- [3H]Ro 48–8587 in rat brain, J. Neuro- sects and vertebrates, Br. J. Pharmacol. chem. 1998, 71, 418–426. 1997, 121, 1496–1505. 68 N. Crawford, T. K. Lang, D. S. Kerr, 73 S. K. Sullivan, R. E. Petroski, D. J. de Vries, High-affinity [3H] kainic G. Verge, R. S. Gross, A. C. Foster, acid binding to brain membranes: a re- D. E. Grigoriadis, Characterization of evaluation of ligand potency and selec- the interaction of indiplon, a novel pyra- tivity, J. Pharmacol. Toxicol. Methods zolopyrimidine sedative-hypnotic, with 1999, 42, 121–125. the GABAA receptor, J. Pharmacol. Exp. 69 K. K. Dev,V. Petersen,T. Honore, Ther. 2004, 311, 537–546. J. M. Henley, Pharmacology and regio- 74 S. C. Lummis, J. Baker, Radioligand nal distribution of the binding of 6-[3H]- binding and photoaffinity labelling stu- nitro-7-sulphamoylbenzo[f]-quinoxa- dies show a direct interaction of phe- line-2,3-dione to rat brain, J. Neurochem. nothiazines at 5-HT3 receptors, Neuro- 1996, 67, 2609–2612. pharmacology 1997, 36, 665–70. 70 L. H. Martini, C. R. Souza, P. B. Mar- 75 R. J. Vandenberg, C. R. French, P. H. ques, J. B. Calixto, R. A. Yunes, D. O. Barry, J. Shine, P. R. Schofield, An- Souza, Compounds extracted from tagonism of ligand-gated ion channel Phyllantus and Jatropha elliptica inhibit receptors: two domains of the glycine the binding of [3H]-glutamate and [3H]- receptor alpha subunit form the strych- GMP-PNP in rat cerebral cortex mem- nine-binding site, Proc. Natl. Acad. Sci. brane, Neurochem. Res. 2000, 25, 211– USA. 1992, 89, 1765–1769. 215. 187

8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes Jesús E. González, Jennings Worley, and Fredrick Van Goor

8.1 Introduction

Ion channels are critical for physiological signaling and are the targets of several drugs. Most of these drugs were discovered using in vivo pharmacology models or directly from observation in humans and only later was it determined that their mechanism of action involved modulation of ion channel activity. Ion channels are widely recognized as “druggable” targets due to their modulation by a wide diversity of small molecules. Despite their promise, these targets have historically been difficult to pursue because of limited structural information, low expression levels, requirement of a membrane environment for proper folding and pharmacology, and limitations or absence of high-throughput screening methods. In the last 15 years drug discovery approaches have increasingly relied on high throughput methods to profile a larger number of candidate compounds in increasing numbers of in vitro assays aimed at providing insight into which molecules will be more likely to be efficacious, safe, and able to be administered in humans. Assays for primary targets, safety counter-screens, chemical properties, and in vitro metabolism are key components of modern discovery processes and approaches. Ion channel assays play as an important role and are particularly challenging as they generally require cellular expression. Fluorescence readouts are commonly use for cell-based assays because of their high sensitivity, compatibility with readily available instrumentation and microtiter plates, and the availability of a variety of probes and reagents [1]. The application of physiological indicators of intracellular calcium and membrane potential has made possible functional fluorescence based ion channels assays with the requisite throughput, sensitivity, and reliability required for large scale profiling. In this chapter we will review the range of fluorescence probes, approaches and concepts for assaying activity for sodium, calcium, potassium, and chloride channels, including those activated and regulated by ligands and voltage. Within each class of channels the utility and challenges will be discussed. Where appropriate, relevant reviews will be cited and illustrative examples will be given. Finally, we will

provide examples and describe areas of where novel fluorescence assays are being developed, with an emphasis on drug discovery applications and new approaches for measuring ion channel trafficking.

8.2 Membrane Potential Probes

Fluorescent probes of cellular membrane potential were initially developed as physiological indicators and are divided into three major categories. The first includes fast electrochromic probes, such as Di-4-ANEPPS, that are capable of detecting microsecond voltage changes yet have low voltage sensitivity (~1–10%

DF/Fo per 100 mV)[2]. The second class is comprised of environment-sensitive dyes that distribute between cellular sites and extracellular solution according to

the membrane potential. These probes have good sensitivity (1–100% DF/Fo per 100 mV) yet relatively slow time resolution, requiring minutes to attain maximal fluorescence response [2]. In the last five years new redistribution dyes have been introduced by Molecular Devices that respond substantially faster, in tens of seconds instead of minutes [3, 4]. The third category involves probes based on fluorescence resonance energy transfer (FRET) readout of rapid transmembrane translocation of fluorescent hydrophobic ions [5, 6]. These probes have good sensitivity

(20–150 % DF/Fo per 100 mV) and respond much more rapidly than redistribution dyes since the temporal response results from facile transmembrane redistribution of a mobile voltage-sensitive ionic dye and not slow diffusion across multiple membrane/water interfaces. Ion channels in cells throughout the body open and close in micro- to milliseconds to produce rapid, <5 s, changes in membrane potential which are important for cell signaling and are not, in many cases, readily detected with slow redistribution dyes. For this reason most applications requiring the highest temporal resolution, such as neuronal signaling, require electrochromic or FRET dyes. Redistribution and FRET probes are generally used for most ion channel drug discovery applications because of their relatively high sensitivity and ease of use.

8.2.1 Redistribution Probes

Fluorescent redistribution probes are environment-sensitive charged dyes that equilibrate between intracellular hydrophobic sites and extracellular solution according to membrane potential. Fig. 8.1A is a schematic illustrating the redistribution mechanism and Fig. 8.1B shows a fluorescence micrograph of cells stained

with the oxonol DiSBAC4(3). Generally, the fluorescence quantum yield is high when the probe binds to hydrophobic sites such as proteins and membranes and very low when in aqueous solution. Consequently positively charged probes such as cyanines and rhodamines result in bright fluorescence at negative membrane potentials and are relatively dim at less negative or positive potentials. The oppo- 8.2 Membrane Potential Probes 189

Fig. 8.1 Oxonol probes that operate via redis- on the right panel, more dye accumulates in tribution. (A) Schematic of voltage-sensitive the cell and the total fluorescence increases. redistribution mechanism. Negatively- (B) A fluorescence micrograph of CHO-K1 charged oxonol molecules (small circles) dis- cells stained with DiSBAC4(3) showing exten- tribute into a cell (large circles) according to sive intracellular staining. The extracellular the membrane potential Vm. The dye fluores- solution is relatively dark because the quan- cence is greatly enhanced, represented by the tum efficiency in aqueous solution is neglible. small filled circles, when bound to hydro- (C) Structures of bis-barbiturate trimethine phobic intracellular protein and membrane oxonols. sites. At less negative potentials, represented site is true for negatively charged probes such as oxonol dyes, which make it more challenging to detect significant oxonol fluorescence from cells that hyperpolarize to very negative potentials. Anions translocate across the membranes and into cells more easily than cations because of the dipole that originates from the lipid carbonyl groups. The rate difference due to charge can be orders of magnitude, as has been shown for the isostructural borates and phosphonium ions [7]. This is an important reason why negatively charged probes, such as oxonols, are routinely used and positively charge probes are not. The structures of commonly used bis- barbiturate oxonol dyes are shown in Fig. 8.1C. One of the first fluorescent assays specifically designed for identifying ion channel modulators used the redistribution oxonol probe DiBAC4(3) to develop a higher throughput assay for identifying openers of KATP channels [8]. This work played a major role in the development of the fluorometric imaging plate reader

(FLIPR) [9]. DiBAC4(3)’s high sensitivity to temperature and excitation and emission properties were key considerations in the original FLIPR design specifica- tions. Key issues with DiBAC4(3) are temperature sensitivity, slow temporal response, interference from fluorescent compounds and relatively high false positive rate. Even at 37 8C, the time response often requires 15 min or more to reach the maximal fluorescence change. Despite these limitations, DiBAC4(3) has been broadly applied and has also been used with a microfluidic device, in 190 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

combination with a cationic dye, to measure ion channel activity in human T lymphocytes [10]. Recently Molecular Devices have introduced a new FLIPR membrane potential (FMP) kit that includes a redistribution dye that offers advantages in temporal re-

sponse and reduced fluorescence interference compared to DiBAC4(3). The probe has fluorescence properties similar to known thiobarbiturate oxonols and has re-

distribution kinetics much faster than DiBAC4(3) with maximal response achieved in approximately 10–15 s. The kit also includes a second quenching dye that reduces extracellular fluorescence and improves signal to background and potentially reduces the effect of some fluorescent compounds. Another FMP kit advantage for HTS applications is that the dyes can be added directly to cells and do not require additional washing steps, streamlining the process and resulting in greater throughput.

8.2.2 FRET Probes

Membrane potential sensors based on FRET are useful for high-throughput screening (HTS) of ion channel targets [11–13]. They are comprised of two fluorescent components. The first is a highly fluorescent, hydrophobic ion that binds to the plasma membrane and “senses” the transmembrane electric field. The ion sensor rapidly redistributes between two binding sites on opposite sides of the membrane, establishing a Nernstian equilibrium (10-fold concentration ratio for ~60 mV). In response to a membrane potential change, the hydrophobic ions electrodiffuse across the membrane and establish a new equilibrium corresponding to the new membrane potential. The voltage-dependent redistribution is con- verted into a ratiometric fluorescent readout with a second fluorescent molecule that binds specifically to one face of the plasma membrane and functions as a FRET partner to the mobile voltage-sensing ion. A schematic of the mechanism is shown in Fig. 8.2A. A variety of fluorescent membrane-bound molecules have been designed to function as voltage-sensitive FRET partners with different voltage sensitivities, temporal responses, and wavelengths. In the most commonly used configuration, the two fluorescent dye components are a ChloroCoumarin-labeled phospholipid (e.g. CC1-DMPE and CC2-DMPE)

and a bis-(1,3-dialkylthiobarbituric acid) trimethine oxonol (DiSBACn(3)), where n corresponds to the number of carbon atoms in the n-alkyl group, shown in Fig. 8.2B. Both are very bright fluorophores, the properties are shown in Table 8.1. CC1/2-DMPE selectively partitions into the outer leaflet of the plasma membrane and acts as a fixed FRET donor to the mobile voltage-sensitive and negatively charged oxonol acceptor. CC1/2-DMPE does not cross the bilayer because of two negative charges from the coumarin and the phosphate groups. Cells have negative (inside) resting membrane potentials and under these conditions the majority of the negatively charge oxonols populate the relatively positive extracellular leaflet, resulting in efficient FRET (i.e. quenching of the coumarin do- nor and increase in the oxonol acceptor emission). As illustrated in Fig. 8.2A, depo- 8.2 Membrane Potential Probes 191

Fig. 8.2 Voltage-sensitive FRET probes. (A) Schematic of voltage- sensitive FRET mechanism. Fluores- cent donor molecules bind selectively to the extracellular leaflet of the plasma membrane, represented by a circle with hatching. Negatively charged acceptor, bold circles, distribute across the plasma membrane according to the membrane potential in a Nernstian manner. At negative resting potentials the acceptors are predominantly at the extracellular surface and FRET is efficient, as shown on the left panel. Upon depolarization, the transmembrane acceptor equilibrium changes so that more oxonols are at the intracellular side. This causes a decrease in FRET and results in an increase and decrease in donor and acceptor fluorescence, respectively. (B) Structures of FRET donor CC1-DMPE and thiobarbiturate oxonols.

Table 8.1 Fluorescence and sensitivity properties of FRET membrane potential probes.

kex kem Tc e } Vm sensitivity (nm) (nm) (ms) (M–1 cm–1) Q.Y. %DR per mV

Donor CC1/2-DMPE 405 460 na 40 000 1.0 na

Acceptors

DiSBAC2(3) 540 560 500 200 000 0.44 1–3 DiSBAC4(3) 540 560 20 200 000 0.44 0.6–1 DiSBAC6(3) 540 560 2 200 000 0.44 0.4–0.8 DiSBAC2(5) 640 660 50 225 000 0.67 0.5–2 DiSBAC4(5) 640 660 2 225 000 0.67 <0.4 DiSBAC6(5) 640 660 0.40 225 000 0.67 <0.2 larization causes translocation of the oxonol to the inner surface of the plasma membrane and an increase in the mean donor and acceptor distance. Because FRET is very sensitive to donor–acceptor distance, this charge movement causes a simultaneous increase and decrease in the CC1/2-DMPE and oxonol fluorescence, respectively. The donor and acceptor fluorescence emission changes are reversed upon repolarization. The oxonol moves reversibly in the membrane with sub-second kinetics, allowing voltage-sensitive FRET to report fast voltage changes. 192 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

Simultaneous patch-clamp and rapid optical recording have been used to demonstrate the speed, sensitivity and ratiometric nature of voltage-sensitive FRET in cells [5, 6]. The coumarin donor to oxonol acceptor fluorescence emission ratio is independent of the excitation intensity, the number of cells being detected, and the optical path length, providing fewer experimental artifacts compared to single wavelength probes. These dyes load well in about 20 min at room temperature and cells can be maintained for hours with the dyes without significantly degrad- ing voltage responses. Analysis of data from a wide variety of cell types gives a sen-

sitivity of about ~1–3% ratio change per mV for DiSBAC2(3) over the relevant physiological range. Oxonols differ in their physical, optical, and voltage-sensing properties. DiS-

BAC2(3) is more water soluble and is left in the extracellular media during assay. DiSBAC2(3) is often the first choice of ion channel screening applications because of its ease of loading, assay stability, and sensitivity. With a response time constant

of ~500 ms, a FRET assay using DiSBAC2(3) as acceptor is 2 orders of magnitude faster than DiBAC4(3) redistribution assays [14] and 10-fold faster than the FMP probes, making this dye well suited for liquid addition protocols which are often used to trigger membrane potential changes in HTS assays. FRET assays with even higher temporal resolution are possible by using more hydrophobic oxonols,

such as DiSBAC6(3) [5]. These more hydrophobic oxonols require using excipi- ents, for example Pluronic F-127, to assist cellular loading and are fast enough to measure millisecond voltage changes. In the case of hexyl substituted penta- methine oxonol the response time constant was measured at 400 ms [6]. A summary of fluorescence and voltage-sensing properties of various FRET membrane potential probes is given in Table 8.1. The recent development of rapid electrical stimulation methods has enabled FRET ion channel assays that utilize the millisecond probes (Huang et al., submitted).

8.2.3 Advantages and Limitations of Membrane Potential Probes

Ion channels pose unique challenges for functional assay development due to their variety and complex kinetics. Direct sensing of membrane potential is an attractive ion channel readout because it is sensitive, generic, and is applied in intact living cells. Fluorescent membrane potential assays can be as, or more, sensitive as other ion channel assays methods including isotopic flux assays and patch clamping while being more amenable to high-throughput profiling. Detection sensitivity of ion channel agonists can often be comparable to or better than voltage-clamp in whole cells, depending on channel density and background cellular conductances. Consistent with drug–receptor models where the assay response is

nonlinear with receptor occupancy [15], the observed EC50s for agonist are often shifted to lower concentrations because only a small fraction of the channels have to be activated to elicit sufficient current to change the membrane potential. For antagonist assays, this property works against detection sensitivity and generally

causes shifts in IC50s to higher compound concentrations. Only ~30 million ions 8.2 Membrane Potential Probes 193 need to pass through channels to change the membrane potential 60 mV in a 50 mm diameter cell [16]. Except for Ca2+, this small number of ions will not ap- preciably change the intracellular or extracellular ion concentrations under normal physiological conditions. In the case of a neuronal voltage-activated Na+ channel, only a small number of channels (10s to 100s) need to be opened to cause enough ions to enter into the cell and rapidly depolarize a cell with a high resistance. The approach is generic because changes in membrane potential can be generated by any electrogenic ionic flux across the membrane. Voltage-sensitive fluorescent probes therefore provide a basis for a universal HTS-compatible assay platform for ion channels. Researchers applying voltage-sensitive fluorescent probes must also be aware of some limitations, potential sources of experimental artifacts and important optimization parameters. Selective functional assays for the target of interest require a host cell line with low background currents that are not activated by the stimulation protocol. Occasionally, small currents that seem insignificant in voltage- clamp recording can cause large effects on membrane potential assays. Once sufficient functional expression and stimulation conditions are established, it is important to optimize the cellular dye staining. Generally, staining conditions are selected for maximum dynamic range, signal over background, and response stability. For HTS applications, where reproducibility and efficiency is critical, it is important to select conditions that are not overly sensitive. Another potential issue is drug–dye interactions. The susceptibility of various membrane potential probes to such artifacts has been evaluated and it has been reported that FRET probes and

FMP probes have decreased compound–dye issues than DiBAC4(3)[17]. The relatively high optical interference of DiBAC4(3) was also reported in a comparison with FMP in hERG [18] and KATP [4] channel assays. Normally these effects are fast and occur immediately after compound addition and do not affect the probes ability to respond to voltage changes. However, if the fluorescence intensities are significantly changed then the apparent voltage changes can be proportionately offset, causing significant error in the measured compound activity. For FRET probes, the ratiometric readout reduces some of these artifacts relative to single intensity dyes, and these effects can be readily identified as a large change in the donor or acceptor intensities compared to control wells in the absence of test compound. These interactions are to a large extent not observed at test compound concentrations below 3 mM, so potent compounds that interfere at higher concentration can usually be cleanly observed at lower concentrations. Finally, cell issues that affect membrane potential as well as most cell-based assays are confluence, health and stability of the cells. Sick cells or those cultured under variable conditions often have different functional channel expression and membrane properties that cause increased false positives or negatives and variability. 194 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

8.3 Ion-sensitive Fluorescent Probes

8.3.1 Calcium Dyes

Calcium dyes have been used extensively for probing Ca2+ signaling in cells and tissue. All currently used exogenous small molecules indicators trace their origin to the seminal work from Roger Tsien and coworkers. These cellular probes are fluorescent Ca2+ chelators based on EGTA that have different excitation and emission wavelengths, ratiometric readouts, and Ca2+ affinities. Fluo3 and its progeny, such as Ca-Green and Fluo4 [19] have been the probes of choice for calcium channel assays using single wavelength excitation, which is typical for most plate readers. Fluo3, which is based on a fluorescein, is conveniently excited with the 488 nm argon ion laser line and its fluorescence can increase 100-fold upon bind- 2+ ing Ca (KD = ~400 nM) [20]. Fura2 has been favored for ratiometric measurements and requires instrumentation capable of dual excitation such as the plate- imaging fluorimeter developed by the former SIBIA Neurosciences and SAIC group [21]. These intracellular indicators are typically loaded into cells as acetoxy- methyl (AM) esters which are subsequently hydrolyzed via intracellular esterases to the tetraacid forms which are trapped within cells. The structures of commonly used fluorescent ion indicators are shown in Fig. 8.3.

Fig. 8.3 Structures of fluorescent ion probes. 8.3 Ion-sensitive Fluorescent Probes 195

8.3.2 Indicators of Other Ions

Cellular fluorescent indicators of other common ions that permeate through ion channels exist and have been successfully applied in a variety of cells and tissue, however they are not as commonly employed as Ca2+ indicators. This is particularly true for HTS applications that generally require a larger fluorescent change to ensure robustness, and compatibility with screening plate readers. Specific ion channel dependent fluorescence changes are modest because of limited ion selectivity and the difficulty of generating sufficiently large ion concentration changes due to target ion channels. All fluorescent indicators for Na+ and K+ are based on crown ethers, which con- fer ion selectivity, conjugated with two fluorescent probes. Upon binding to the macrocycle, the fluorescence intensity is enhanced. The sodium, SBFI, and potassium, PBFI, indicators use a benzofuran fluorphore [22]. These probes produce ratiometric changes in their excitation spectra that require similar detection instrumentation as Fura2. Sodium green was developed by Molecular Probes using fluorescein as the fluorophore which enables excitation in visible wavelengths similar to Fluo3 [23]. A key issue that limits the utility of these probes for sodium channels, as an example, is that the selectivity over K+ is not sufficiently high to eliminate competition with high intracellular K+. As a result large, >10 mM, intracellular Na+ concentrations are required to elicit significant fluorescence changes. This is not feasible for many channels that inactivate rapidly and/or express at low levels. For K+ channels the intracellular concentrations are essentially fixed, however, the use of thallium as a surrogate in combination with the low affinity Ca2+ indicator BTC has been shown to be capable of large K+ channel dependent signals[24] and is discussed later. Chloride indicators are primarily quinolinium or acridinium dyes that have decreased fluorescence when exposed to higher halide concentrations due to increasing collisional quenching. Different dyes, including SPQ and MQAE, have been applied to monitor CFTR activity [25]. An iodide-sensitive pteridine dye with improve optical properties has been reported [26] and applied to studying CFTR function [27]. An important property of these dyes is that their fluorescence is essentially insensitive to other physiological relevant anions and nitrate. This feature has been used to develop CFTR assays. CFTR expressing cells are loaded with Cl– or I– and a dye such as SPQ. The extracellular Cl– is replaced with nitrate and upon CFTR activation with forskolin intracellular Cl– leaves the cell via CFTR and is exchanged with nitrate, which causes the cellular fluorescence to increase. While routinely used for CFTR these approaches have not been generally used because they have limited brightness and signal changes, require the channel to stay open for a long time (like CFTR), and assay development complications that arise from other endogenous conductance pathways for halides and anion substitutes, can also be problematic. 196 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

8.4 Fluorescence Assays for Ion Channels

In the remainder of the chapter, we will describe specific examples of how fluorescence probes have been applied to assay various ion channel sub-families. A large number of these applications are directed toward drug discovery because these assays can be automated and used cost effectively to profile large numbers of compounds. This is currently a necessity because of the inability to design modulators de novo. A summary of available and preferred options for the individual sub-families is listed in Table 8.2.

Table 8.2 Fluorescent probes and assays for ion channel classes: the approach most often used is highlighted in bold.

Target Membrane Ion indicators Fluorescent Comments channel class potential proteins

Na+ FRET SBFI, Na-green No Fast FRET probes and Redistribution E-VIPR use-dependence without gating modifiers

Ca2+ FRET Fluo-3&4, Fura2 Calcium-sen- Co-expression of inward Redistribution sitive GFPs rectifier used to assay voltage-dependence

K+ FRET Tl+ sensitive No Thallium as a substitute Redistribution probes for K+ used for fluorescent flux assay

Cl– FRET Quenching dyes Halide-sen- Trafficking assays for Redistribution (e.g. SPQ) sitive GFPs CFTR

Ligand-gated FRET If Ca2+ permeable yes TRP nicotinic channels Redistribution use both Ca2+ and membrane potential dyes

8.4.1 Calcium Channels

The ability to develop functional fluorescent calcium channel assays greatly bene- fited from the development of fluorescent intracellular calcium probes and screening instrumentation, such as the FLIPR [9, 28] and the voltage ion probe reader (VIPR) [12]. Typically these assays take advantage of the ~105-fold Ca2+ concentration gradient across the plasma membrane and high sensitivity and selectivity of probes such as Fura2 [29] and Fluo-3 [20]. The channels are typically activated with high K+ or ligand and the probes sensitively detect intracellular concentration changes from ~50–200 nM basal levels. Since the basal intracellular calcium con- 8.4 Fluorescence Assays for Ion Channels 197 centrations levels are so low very few activated channels are required to elicit a substantial fluorescence change. Voltage-gated calcium channels are the targets for established drugs such as dihydropyridines, phenylalkylamines, and benzothiazepines that block L or CaV1 channels and are used to treat cardiovascular disease, in particular hypertension and angina. The N-type or CaV2.2 channel is the target for the recently launched drug PrialtTM, which is used to treat chronic morphine resistant pain [30]. Func- tional calcium channel assays have been developed for most members of this target class. Velicelebi and coworkers have developed functional Ca2+ assays against

CaV2 family channel complexes using stably expressing cell lines [21]. CaV-dependent calcium influx was stimulated with high K+ depolarization and the assay was able to reproduce the known molecular pharmacology differences of peptide toxins against these CaV subtypes. For example, they found that the m-conotoxin GVIA selectively blocked CaV2.2 over CaV2.1 and CaV2.3. This approach has also been applied to the screening of plant extracts for modulators of neuronal channels [31]. One of the limitations of assays designed to identify blockers for voltage-gated channels is that many blockers are state and voltage dependent. Xia and coworkers co-expressed the inward rectifier Kir2.3 with a CaV1 or L-type channel complex in order to probe the voltage dependence of block by incubating compound and cells at two extracellular K+ concentrations before applying a maximal high K+ stimulation [32]. They showed that blockers such as nimodopine, which are known to be voltage-dependent blockers with higher affinity at less negative membrane potentials, are more sensitively detected when the cells are K+-clamped at less negative potentials. In the case of nimodipine, they determined an IC50 of 3 nM at 30 mM external K+ versus 60 nM at 6 mM external K+. The ability to assay the relative state- dependent activity of channel modulators in a HTS compatible format is a significant improvement over traditional high K+ induced assay protocols.

8.4.2 Non-voltage-gated Cation Permeable Channels

In addition to voltage-induced channel responses, fluorescent indicators have been widely used to measure activity of cation permeable channels that are activated by voltage-independent mechanisms. These channels can be classified as ligand-gated or ligand-activated, a characteristic property used to activate or induce channel activity at a defined time in the assay. For example exogenous application of a ligand to trigger external calcium entry has been essential to implement calcium indicator assays to aid the discovery of small molecules that pharmacologically modulate channel function. One important type of ligand-gated channel is the transient receptor potential (TRP) protein channel family. This class of proteins has gained considerable attention over the past several years as a novel and rapidly expanding channel superfamily. These primarily voltage-independent channels are most notable for their ability to respond to a variety of physical and chemical stimuli, such as temperature, mechanical force, phospholipids, oxidants and many others [33–35]. Because of these 198 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

properties and where they are expressed, TRP channels have been implicated in a wide range of physiological processes including sensory physiology and immune response and related diseases. The calcium permeable nature of many TRP channels has been exploited to advance the field by assisting the identification and determination of the function of novel family members, for example TRPM3[36], as well as the search for new physiological stimuli and novel small molecules. TRPV1 (vanilloid receptor sub-family) has been defined by its activation by va- nilloids such as capsaicin and noxious heat. Figure 8.4 demonstrates the use of the Fluo3 and Fura-Red calcium indicators to measure hTRPV1 response to capsaicin in a 3456-well nanoplate [37]. Capsazepine, a known TRPV1 antagonist dose-dependently inhibits the fluorescent response and is shown in the right panel. Using a similar assay, Rami and coworkers [38] used Fluo-3 to search for compounds that inhibited the capsaicin induced calcium-dependent fluorescent responses. In this study, they characterized a potent series of compounds that bind to an extracellular site to produce inhibition of temperature-induced activation. In a side-by-side comparison of the responses of the cold responsive channel TRPM8 and TRPV1 Behrendt and coworkers [39] found selective action from a library of odorants, providing new tools to help dissect the details of complex sensory signaling pathways that discriminate among different thermo-sensations. Fluorescent calcium indicators have also been used to identify compounds that modulate native channel complexes thought to be related to or comprised by members of the TRP family. Ishikawa et al. [40] examined the ability of small molecules to inhibit induced Fura-2 responses in human Jurkat T-cells to identify a potent inhibitor of

Icrac, which was shown to alter immunosuppressant activity in a mouse contact hypersensitivity model. Appendino and colleagues have reported another example of extending studies from bench to preclinical stage. In this study, potent sub-nanomolar TRPV1 agonists were identified as well as sub-micromolar antagonists from channels exogenously expressed in HEK293 cells using Fluo-3 [41]. These agents were shown to pharmacologically modulate dorsal root ganglion neurons, bladder smooth muscle contraction and significantly altered micturition volume and frequency in an obstructed bladder function model. Thus the use of calcium indicators has greatly facilitated the identification of agents that modify in vivo physiology and are now being tested in clinical trials. In addition to elevating intracellular calcium levels, calcium ion entry can also produce membrane depolarization or hyperpolarization, via calcium-activated K+ channels. The nictoinic receptor, a calcium permeable cation selective channel is an example of a channel that has multiple functional roles. It is activated by acetylcholine and is important in excitation–contraction coupling as well as many central and peripheral neuronal functions. The cell signaling events that a channel participates in can vary and can lead to complex function. Recently, a number of different cell lines using both calcium and membrane potential probes to measure nicotinic receptor channel activity have been used to dissect and exploit these differences [42]. Data from both formats were complementary but demonstrated distinct differences in ligand-induced channel kinetics. Membrane potential measurements demonstrated greater sensitivity and potency to activation by agonists, while antagonists were slightly less potent as compared to calcium indicator measurements in the same cells. While some differences may be addressed by channel density or assay sensitivity importantly, this represents the ability to use fluorescent indicators to differentiate between channel functions. For voltage-gated calcium channels, participation of these channels in calcium entry versus membrane depolarization is also of interest as both functions have been shown to impact and regulate cellular physiology. Figure 8.5 compares the measured fluorescence membrane potential versus the fluorescence calcium esponse resulting from applying an electrical field to cells containing a CaV channel. Note the differences in waveform of the calcium and Vm signal. The membrane potential has an abrupt large transient change in signal that then reaches a plateau, however, the calcium indicator signal increases slowly building to a sustained level. The cellular property of calcium Ca2+ release (CICR) provides an additional complexity for the application of calcium dyes to the study of channel activity. CICR can greatly amplify the initial calcium entry through the plasma membrane-bound assay target. While providing high sensitivity, CICR can also limit the detection of slowly acting blockers and cause displacement of the dose response relationship relative to the actual compound activity on the target channel. The use of fast membrane potential probes, such as FRET probes, can be designed to provide complementary data that is not dependent on CICR. Using both formats as a measure of CaV activity, differences in pharmacology have been revealed and may be pursued to identify small molecules that differentiate between these CaV functions. 200 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

Fig. 8.5 Membrane potential versus calcium signal. Comparative measurements from

CaV containing cells evoked using electric field stimulation using either an indicator of changes in membrane potential or intracellular calcium ions. HEK cells stably expres-

sing a CaV channel were plated in a 96-well E-VIPR. The stimulation period is indicated. Cells were stained with membrane potential FRET dye pair (top) or loaded with Fluo- 3AM and Fura RedAM (bottom). All responses were blocked by the calcium channel blocker Mibefradil (10 mM, not shown).

8.4.3 Sodium Channels

Voltage-gated sodium channels (Navs) historically have been one of the most difficult ion channel classes for which to develop high-throughput assays. The difficulty has arisen from a combination of challenges including cellular expression, voltage dependence of activation and inactivation, millisecond activation and inactivation kinetics, and voltage and state dependence of many blockers. The use of fluorescence membrane potential probes is the primary approach for this target class because of its high sensitivity to small currents and high-speed readout. Be- cause these channels open and inactivate in milliseconds most assays require the use of gating modifiers such as veratridine and batrachotoxin (BTX) to lock the channel in an open conformation to prolong Na+ influx and cause sufficient current to depolarize cells. The use of toxins in assays is generally not desired because test compounds must compete for these sites and can limit the sensitivity of the assays to the detection of competitive binders. Typically cells are incubated in the absence of extracellular Na+, substituting with organic cations such as tetra- methylammonium, choline, or N-methyl-d-glucamine which do not permeate + through open NaV channels. With the channel locked open, addition of Na results in an Nav-dependent inward current that rapidly depolarizes the cell [12]. Blockers are identified by their ability to inhibit this depolarization. The most commonly used fluorescent membrane potential probes are Di-

BAC4(3), FMP and FRET-based voltage sensor probes. Some recent examples of these approaches have been published. Vickery and coworkers have applied FMP

dye and NaV activator evoked depolarizations to compare the molecular pharmacology of activators and blockers against rNaV1.8, rNaV1.2 and hNaV1.5 [43]. Using these assays they observed that different activators or gating-modifiers, including 8.4 Fluorescence Assays for Ion Channels 201 veratridine and the type II pyrethroids deltamethrin and venfalerate, had efficacies that were subtype-dependent. They reported that veratridine induced NaV-dependent depolarizations in rNaV1.2 and hNaV1.5 but was not effective against rNaV1.8. The type II pyrethroids were most effective in hNaV1.5 and rNaV1.8 and were potentiated with veratridine. Of the approximately15 known blockers examined, all blocked the three subtypes with approximately equal concentrations, except for the neurotoxin TTX. Felix and coworkers have used voltage-sensitive

FRET probes to develop assays for NaV1.7 and NaV1.5. They have reported that pre-incubation of compounds prior to addition of activators enabled compounds to pre-bind to the native conformation and then subsequently added the activator prior to Na+ addition [44]. They observed that this protocol more accurately reported blocking activities for some classes of compounds. The use of electrical stimulation and high-speed FRET probes has been introduced recently and enables repetitive millisecond stimulation and detection of

NaVs in standard 96 and 384 well plates. The electrical stimulation voltage ion probe reader (E-VIPR) eliminates the need for using pharmacological modifiers and enables repetitive stimulation. E-VIPR is very flexible and has been shown to robustly identify and characterize NaV blocker activity, including assessment of use-dependence (Huang et al. submitted). Figure 8.6 shows E-VIPR concentration–response traces of 5 drugs using an E-VIPR assay and stimulating at 5 Hz. The response includes contributions from both voltage and frequency. At intermediate concentrations use-dependence is detected as differential blockade of the first (tonic) and last (use-dependent) stimulated voltage-transients. TTX, which is not voltage-dependent, does not exhibit as much use-dependence as the local an- esthetic drugs.

8.4.4 Potassium Channels

Potassium channels are deceptively challenging targets for which to develop useful fluorescence assays. The K+ channel family is relatively large and individual members are activated in a wide variety of voltage and nonvoltage mechanisms. While it is straightforward to add high K+ to cells and elicit a depolarization, it is much more challenging to develop an assay specific to a single target channel. The high and variable K+ channel background conductances make cell line and assay development difficult. The most common approach utilizes membrane potential dyes, although recently, the application of low affinity Ca2+ probes in combination with thallium as a K+ substitute has been used to develop HTS compatible fluorescence assays [24]. Membrane potential assays require the use of cell lines with minimal background conductances and relatively positive resting potentials. For blocker assays, expression of the target K+ channel will set the membrane potential at near its ac- + tivation potential or near the K reversal potential (Krev), so that block will result in a depolarization (if there are no other significant K+ conductances). The block can be monitored continuously after compound addition or by probing for the re- 202 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

Fig. 8.6 Activity of use-dependent NaV block- of the figure, are the highest concentrations ers in E-VIPR membrane potential assay. Con- tested. Each blocker was diluted 1 to 4 across centration–response traces for TTX, tetra- the plates, as illustrated by the arrow. Use-de- caine, amitriptyline, mexiletine, lidocaine and pendent block is evident for all compounds

lamotrigine block of hNaV1.3-dependent de- except TTX, and is manifested as greater polarization transients using a 5 Hz stimula- block at the end of the response train com- tion protocol. The concentrations shown pared to the first response. Control responses above the first ratio responses, at the left side without blockers are shown on the far right.

duction on high K+ induced depolarization. For activators or “openers”, the expressed channel is closed at less negative potentials and compound activation causes the cell to hyperpolarize toward the potassium equilibrium potential. This

strategy has been employed for identifying and profiling KATP channels openers using redistribution dyes such as DiBAC4 in the bladder [8, 45, 46], myocytes [47] and in cells heterologously expressing SUR1 and the inward rectifier Kir6.2 [48].

DiBAC4 has also been used to identify blockers, for example fluoxetine inhibition of Ca-activated K+ channels [49]. Fluorescent probes that directly measure K+ are difficult to apply for monitoring K+ channel-dependent [K+] changes because intracellular and extracellular [K+]do not change sufficiently under experimentally accessible conditions to elicit a large fluorescence change. Thus, K+ surrogates have been widely used for developing K+ channel assays. Rb86+ is routinely used as a radioactive substitute for K+ in flux assays (see Chapter 7). Recent developments in the application of atomic spectroscopy detection have led to nonradioactive Rb+ flux assays. Thallium (Tl+), another K+ surrogate, can readily enter cells via Na+/K+ ATPase and Na+/K+/Cl+ co-trans- 8.4 Fluorescence Assays for Ion Channels 203 port mechanisms [50]. Tl+ readily fluxes through K+ channels and is known to quench various water-soluble fluorophores. Its quenching properties have been exploited to probe the activity of acetylcholine receptors [51, 52]. Recently, Weaver and colleagues have identified a coumarin benzothiazole-based low affinity Ca2+ fluorescent probe BTC, structure shown in Fig. 8.3, that functions as an intracellular Tl+ sensor from a screen of fluorescent metal chelating probes [24]. Using this probe in combination with a Cl–-free assay buffer, which overcomes TlCl solubility limitations, they devised an HTS compatible fluorescent assays for measuring KCNQ2 and SK3 K+ channel modulator activities. Two-fold fluorescence changes were achieved with minimal interference from intracellular Ca2+ and fluorescent test compounds because of weak Ca2+ affinity and the visible excitation wavelengths. Researchers considering using Tl+ should be aware that it is readily absorbed and is known to be toxic [53].

8.4.5 Chloride Channels

Extracellular and intracellular ligand-gated Cl– channels regulate membrane potential in excitable and nonexcitable cells and have been identified as drug targets for a number of different conditions and diseases, including neurological disor- – ders and cystic fibrosis. For example, the GABAA receptor is a ligand-gated Cl channel that is the primary target of many drugs for the treatment of anxiety, epilepsy, muscle spasms, sleep related conditions and is the target for benzodiazepine drugs such as diazepam. Despite GABAA being a well validated drug target, historically it has been difficult to develop fluorescence assays because of desensitization, limitations of Cl– sensitive probes, background conductances, and the desire to identify compounds that modulate the effects of GABA. Recently, FRET membrane potential probes have been successfully applied to develop assays for characterizing GABAA pharmacological agents, including potentiators [54, 55]. This approach has also helped to identify new selective allosteric modulators that can differentiate between beta subunits [56] and to illuminate the structural basis of the selectivity of neurosteriods on the receptor [57]. Cystic fibrosis is due to mutations in the gene encoding the epithelial PKA- gated Cl– channel known as the cystic fibrosis transmembrane conductance regulator (CFTR) [58, 59]. The most common mutation is a deletion of phenylalanine at position 508 in the first nucleotide-binding domain of CFTR. This mutation leads to impaired trafficking and gating of CFTR to the apical membrane of re- spiratory epithelia, resulting in the defective anion transport that underlies the lung disease in patients [60]. Several fluorescent-based assays have been developed to identify and track the structure–activity relationship (SAR) of small molecules that directly target mutant CFTR to restore its defective gating or trafficking [61]. These include the use of fluorescent membrane potential and halide sensors, including genetically-targeted fluorescent proteins. Voltage-sensitive membrane potential probes offer a sensitive and convenient approach to the indirect monitoring of anion flux through chloride channels in a 204 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

variety of different cells types, including recombinant cells and primary cell cultures. The excellent throughput, sensitivity, and reproducibility of such assays make them especially useful for tracking SAR to support medicinal chemistry optimization. However, because the membrane potential response is nonlinear, reaching equilibrium at the reversal potential for Cl–, the assay window and ability to monitor potency (Fig. 8.7) and efficacy (Fig. 8.8) can be compromised. Two approaches can be used to circumvent this problem. First, the rate of membrane potential change, rather than the absolute magnitude, can be used to monitor Cl– flux. To do this, the temporal response of the dye must be sufficiently fast to accurately track the membrane potential. Second, decreasing channel density or redu- cing agonist concentrations can reduce the assay sensitivity and expand the linear range. Although this is particularly useful for the identification of highly effica-

Fig. 8.7 CFTR membrane potential assay de- 100% wild-type CFTR-expressing cells. In the monstrates efficacy limitations compared to fluorescence-based assay, the half maximal re- flux assays: Response to CFTR activation be- sponse was observed at ~3% wild-type CFTR tween fluorescent-based and electrophysiolo- and was nonlinear as the concentration of gical assay formats. To monitor the response wild-type CFTR expressing cells was in- to increasing amounts of CFTR activation creased. In contrast, the half-maximal re- cells expressing wild-type CFTR were mixed sponse in the using chamber assay was with parental cells in the indicated propor- reached at ~60% wild-type CFTR and was lin- tions (% wild-type CFTR). The response to ear. These results highlight the nonlinearity of CFTR activation using a maximal concentra- the fluorescence-based assays, which can tion of forskolin was monitored in both the limit the SAR evaluation of agonist efficacy fluorescence assay (black circles) and in an because the sensitive response saturates with electrophysiological assay (red circles) under low amounts of CFTR activation. (This figure voltage-clamp control. The response in both also appears with the color plates.) assays was normalized to the response using 8.5 Assays for Monitoring Channel Trafficking 205 cious compounds, it can potentially reduce the range of potency, limiting SAR evaluation. Another assay approach is to monitor Cl– flux directly using genetically-encoded anion-sensitive fluorescent proteins, such as mutants of yellow fluorescent protein

(YFP) [62, 63]. Anion binding shifts the YFP chromophore pKa values and results in quenched fluorescence. Cellular YFP sensor expression enables measurement of transmembrane halide fluxes. Similar to small molecule halide indicators, iodide and nitrate are more sensitively detected and consequently similar anion replacement strategies have been successfully applied to assaying CFTR activity, including high throughput screening [61, 64]. In another application, GABAA-dependent intracellular Cl– changes were measured in cultured hippocampal neurons using a Cl– sensitive YFP mutant fused to Cl– insensitive FRET acceptor protein [65]. The chimeric fluorescent probe provides emission ratio detection that reduces experimental artifacts compared to single intensity indicators. Despite successful applications of fluorescent protein probes for ion channel analysis, their broad use is limited by relatively small fluorescence changes and cellular expression.

8.5 Assays for Monitoring Channel Trafficking

In addition to using fluorescence-based assays to monitor changes in CFTR gating, they can be used to monitor increased channel density due to compound rescue of protein processing and/or trafficking. A few pharmacological agents have been identified that partially rescue the defective trafficking of mutant CFTR, leading to increased cell surface density. 4-Phenyl butyrate, for example, acts at millimolar concentrations to produce a modest increase in DF508-CFTR density in vitro [66], but had limited efficacy in vivo [67]. To enable the detection of compounds that alter protein trafficking, it is necessary to incubate the compounds for several hours prior to monitoring activity to allow for de novo synthesis and insertion into the membrane. This presents a potential issue in that incomplete wash-out of compounds that potentiate channel activity rather than increase its density may also be identified. To separate compounds that rescue the defective processing or trafficking from those that alter channel gating it is necessary to obtain independent confirmation using biochemical assays. For example, the correct processing and trafficking of mutant CFTR can be confirmed by monitoring its maturation via passage through the golgi network. Compounds may also act by increasing the gating activity, as well as trafficking to the cell’s surface, further complicating the assessments of channel density in the membrane. To directly monitor changes in the cell surface density and eliminate the impact of channel gating, fluorescence-based assays that monitor protein localization could be used. For example, GFP has been tagged to the cytoplasmic C terminal tail of CFTR and has been successfully used to monitor the cytoplasmic and cell surface protein localization [68]. Although the throughput of such assays is lim- 206 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes 8.6 Summary 207

ited, improvements in algorithms designed to separate cytoplasmic from nuclear staining could improve the utility of such assays. In addition to cystic fibrosis, protein trafficking and processing abnormalities of ion channels have been identified as the underlying cause of diseases, such as fa- milial hyperinsulinism (Kir6.2), Sjögren’s syndrome (aquaporin-5), nephrogenic diabetes insipidus (aquaporin-2), Brugada Syndrome (NaV1.5), inherited long QT syndrome (KCNQ1, hERG), and Andersen-Tawil syndrome (Kir2.1). In some cases, channel ligands (i.e, agonists and antagonists) have been demonstrated to partially rescue the defective trafficking and membrane density of mutant channels, including NaV1.5 (mexiletine)[69], hERG (astemizole, terfenadine) [70], and Kir6.2 (sulfonylureas) [71]. Although a few agents have been identified, fluorescence-based assays similar to those used for mutant CFTR could be used to identify more drug-like, potent, and efficacious agents to correct the defective folding and/or trafficking of Na+ and K+ channels.

8.6 Summary

Fluorescence approaches to studying ion channel function and the interactions of modulating molecules offer significant advantages in sensitivity, ease of use, and high throughput analysis. Existing fluorescent cellular probes provide access to all classes of ion channels. These approaches primarily use membrane potential and ion-sensitive probes, which respond to the net ion flux through the channel. Each has it own assay development challenges and in some cases there are significant limitations on the types of information that can be obtained. Through a combination of improved probes and instrumentation, increasingly higher information content analysis is now possible, for example, fluorescence assays for use-dependent blockade and channel trafficking. Looking forward, additional detail should be attainable through fluorescent proteins and assays that can monitor different channel conformational states and their interactions with other proteins.

3 Fig. 8.8 CFTR membrane potential assay de- genous K+ conductance. Only at very low CFTR

pendent effects on membrane potential were tions did the EC50 for forskolin approximate its calculated using the Goldman–Hodgkin–Katz Kd. These results illustrate that the potency of equation incorporating the open probability at ion channel agonists in a fluorescence-based each agonist concentration and CFTR densities assay is highly sensitive to the channel density. of 0.01 to 100 times the background permeabil- (This figure also appears with the color plates.) ity,which was assumed to be due to an endo- 208 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

References

1 J. E. Gonzalez, P. A. Negulescu, Intra- 11 J. E. Gonzalez, M. P. Maher, Cellular cellular detection assays for high- fluorescent indicators and voltage/ion throughput screening, Curr. Opin. Bio- probe reader (VIPR) tools for ion chan- technol. 1998, 9(6), 624–631. nel and receptor drug discovery, Recep- 2 L. M. Loew, Potentiometric membrane tors Channels 2002, 8(5–6), 283–295. dyes, in Fluorescent and Luminescent 12 J. E. Gonzalez, K. Oades,Y. Leychkis, Probes for Biological Activity,W.T. Mason A. Harootunian, P. A. Negulescu, (Ed.), Academic Press, San Diego, 1993, Cell-based assays and instrumentation pp. 150–160. for screening ion-channel targets, Drug 3 D. F. Baxter, et al., A novel membrane Discov. Today 1999, 4(9), 431–439. potential-sensitive fluorescent dye im- 13 M. Falconer, et al., High-throughput proves cell-based assays for ion chan- screening for ion channel modulators, nels, J. Biomol. Screen., 2002, 7(1), 79– J. Biomol. Screen. 2002, 7(5), 460–465. 85. 14 D. E. Epps, M. L. Wolfe,V. Groppi, 4 K. L. Whiteaker, S. M. Gopalakrish- Characterization of the steady-state and nan, D. Groebe, C. C. Shieh, U. War- dynamic fluorescence properties of the rior, D. J. Burns, M. J. Coghlan,V. E. potential-sensitive dye bis-(1,3-dibutyl- Scott, M. Gopalakrishnan,Validation barbituric acid)trimethine oxonol (Di- of FLIPR membrane potential dye for bac4(3)) in model systems and cells, high throughput screening of potas- Chem. Phys. Lipids. 1994, 69(2), 137– sium channel modulators, J. Biomol. 150. Screen. 2001, (5), 305–312. 15 T. Kenakin, Stimulus-Response Me- 5 J. E. Gonzalez, R.Y. Tsien,Voltage sen- chanisms, in Pharmacologic Analysis of sing by fluorescence resonance energy Drug-Receptor Interaction, Lippincott-Ra- transfer in single cells, Biophys. J. 1995, ven, Philadelphia, 1997, pp. 78–105. 69(4),1272–1280. 16 J. Koester, Membrane Potential, in 6 J. E. Gonzalez, R.Y. Tsien, Improved Principles of Neural Science, E. R. Kan- indicators of cell membrane potential del, J. H. Schwartz, T. M. Jessell (Eds.), that use fluorescence resonance energy Elsevier, New York, 1991, pp. 81–94. transfer, Chem. Biol. 1997, 4(4), 269– 17 C. Wolff, B. Fuks, P. Chatelain, Com- 277. parative study of membrane potential- 7 R. F. Flewelling,W. L. Hubbell,The sensitive fluorescent probes and their membrane dipole potential in a total use in ion channel screening assays, J. membrane potential model. Applica- Biomol. Screen. 2003, 8(5), 533–543. tions to hydrophobic ion interactions 18 W. Tang, J. Kang, X. Wu, D. Rampe, with membranes, Biophys. J.1986, L. Wang, H. Shen, Z. Li, D. Dunning- 49(2), 541–552. ton, T. Garyantes, Development and 8 K. O. Holevinsky, Z. Fan, M. Frame, evaluation of high throughput func- J. C. Makielski,V. Groppi, D. J. Nel- tional assay methods for HERG potas- son, ATP-sensitive K+ channel opener sium channel, J. Biomol. Screen. 2001, acts as a potent Cl- channel inhibitor in 6(5), 325–331. vascular smooth muscle cells, J. Membr. 19 K. R. Gee, K. A. Brown,W. N. Chen, Biol. 1994, 137(1), 59–70. J. Bishop-Stewart, D. Gray, I. John- 9 K. S. Schroeder, B. D. Neagle, FLIPR: son, Chemical and physiological char- A new instrument for accurate, high acterization of fluo-4 Ca(2+)-indicator throughput optical screening, J. Biomol. dyes, Cell Calcium 2000, 27(2), 97–106. Screen. 1996, 1, 75–80. 20 A. Minta, J. P. Kao, R.Y. Tsien, Fluor- 10 J. Farinas, A. W. Chow, H. G. Wada, A escent indicators for cytosolic calcium microfluidic device for measuring cellu- based on rhodamine and fluorescein lar membrane potential, Anal. Biochem. chromophores, J. Biol. Chem. 1989, 2001, 295(2), 138–142. 264(14), 8171–8178. References 209

21 G. Velicelebi, K. A.Stauderman, 32 M. Xia, J. P. Imredy, K. S. Koblan, M. A. Varney, M. Akong, S. D. Hess, P. Bennett, T. M. Connolly, State-de- E. C. Johnson, Fluorescence techni- pendent inhibition of L-type calcium ques for measuring ion channel activity, channels: cell-based assay in high- Methods Enzymol. 1999, 294, 20–47. throughput format, Anal. Biochem. 22 A. Minta, R. Y. Tsien, Fluorescent indi- 2004, 327(1), 74–81. cators for cytosolic sodium, J. Biol. 33 D. E. Clapham, L. W. Runnels, Chem. 1989, 264(32), 19449–19457. C. Strubing, The TRP ion channel fa- 23 G. P. Amorino, M. H. Fox, Intracellu- mily, Nat. Rev. Neurosci. 2001, 2(6), 387– lar Na+ measurements using sodium 396. green tetraacetate with flow cytometry, 34 R. Inoue, TRP channels as a newly Cytometry 1995, 21(3), 248–256. emerging non-voltage-gated CA2+ entry 24 C. D. Weaver, D. Harden, S. I. Dwo- channel superfamily, Curr. Pharm. Des. retzky, B. Robertson, R. J. Knox, A 2005, 11(15), 1899–1914. thallium-sensitive, fluorescence-based 35 C. Montell, The TRP superfamily of assay for detecting and characterizing cation channels, Sci. STKE 2005, (272): potassium channel modulators in re3. mammalian cells, J. Biomol. Screen. 36 N. Lee, et al., Expression and characteri- 2004, 9(8), 671–677. zation of human transient receptor po- 25 M. R. West, C. R Molloy, A microplate tential melastatin 3 (hTRPM3), J. Biol. assay measuring chloride ion channel Chem. 2003, 278(23), 20890–20897. activity, Anal. Biochem. 1996, 241(1), 37 L. Mere, T. Bennett, P. Coassin, 51–58. P. England, B. Hamman, T. Rink, 26 S. Jayaraman, L. Teitler, B. Skalski, S. Zimmerman, P. Negulescu, Minia- A. S. Verkman, Long-wavelength turized FRETassays and microfluidics: iodide-sensitive fluorescent indicators key components for ultra-high-through- for measurement of functional CFTR put screening, Drug Discov. Today 1999, expression in cells, Am. J. Physiol. 1999, 4(8), 363–369. 277(5 Pt. 1), C1008–1018. 38 H. K. Rami, et al., Discovery of small 27 F. Munkonge, et al., Measurement of molecule antagonists of TRPV1, Bioor. halide efflux from cultured and primary Med. Chem. Lett. 2004, 14(14), 3631– airway epithelial cells using fluores- 3634. cence indicators, J. Cystic Fibrosis 2004, 39 H. J. Behrendt, T. Germann, C. Gil- 3 (Suppl. 2), 171–176. len, H. Hatt, R. Jostock, Characteriza- 28 E. Sullivan, E. M. Tucker, I. L. Dale, tion of the mouse cold-menthol recep- Measurement of [Ca2+] using the tor TRPM8 and vanilloid receptor type- Fluorometric Imaging Plate Reader 1 VR1 using a fluorometric imaging (FLIPR), Methods Mol. Biol. 1999, 114, plate reader (FLIPR) assay, Br. J. Phar- 125–133. macol. 2004, 141(4), 737–745. 29 G. Grynkiewicz, M. Poenie, R. Y. 40 J. Ishikawa, K. Ohga, T. Yoshino, Tsien, A new generation of Ca2+ indi- R. Takezawa, A. Ichikawa, H. Kubota, cators with greatly improved fluores- T. Yamada, A pyrazole derivative,YM- cence properties, J. Biol. Chem. 1985, 58483, potently inhibits store-operated 260(6), 3440–3450. sustained Ca2+ influx and IL-2 produc- 30 G. P. Miljanich, Ziconotide: neuronal tion in T lymphocytes, J. Immunol. calcium channel blocker for treating 2003, 170(9), 4441–4449. severe chronic pain, Curr. Med. Chem. 41 G. Appendino, et al., Development of 2004, 11(23), 3029–3040. the first ultra-potent “capsaicinoid” ago- 31 K. L. Rogers,W. F. Fong, J. Redburn, nist at transient receptor potential vanil- L. R. Griffiths, Fluorescence detection loid type 1 (TRPV1) channels and its of plant extracts that affect neuronal vol- therapeutic potential, J. Pharmacol. Exp. tage-gated Ca2+ channels, Eur. J. Ther. 2005, 312(2), 561–570. Pharm. Sci. 2002, 15(4), 321–330. 210 8 Ion Channel Assays Based on Ion and Voltage-sensitive Fluorescent Probes

42 R. W. Fitch,Y. Xiao, K. J. Kellar, J. W. Cell Mol. Neurobiol. 1995, 15(3), 351– Daly, Membrane potential fluores- 360. cence: a rapid and highly sensitive assay 51 H. P. Moore, M.A Raftery, Direct for nicotinic receptor channel function, spectroscopic studies of cation translo- Proc. Natl. Acad. Sci. U S A 2003, cation by Torpedo acetylcholine receptor 100(8), 4909–4914. on a time scale of physiological rele- 43 R. G. Vickery, et al., Comparison of the vance, Proc. Natl. Acad. Sci. U S A 1980, pharmacological properties of rat 77(8), 4509–4513. Na(V)1.8 with rat Na(V)1.2a and hu- 52 W. C. Wu, H. P. Moore, M. A. Raftery, man Na(V)1.5 voltage-gated sodium Quantitation of cation transport by re- channel subtypes using a membrane constituted membrane vesicles contain- potential sensitive dye and FLIPR, Re- ing purified acetylcholine receptor, Proc. ceptors Channels 2004, 10(1), 11–23. Natl. Acad. Sci. U S A 1981, 78(2), 775– 44 J. P. Felix, et al., Functional assay of 779. voltage-gated sodium channels using 53 R. S. Hoffman, Thallium toxicity and membrane potential-sensitive dyes, As- the role of Prussian blue in therapy, Tox- say Drug Dev. Technol. 2004, 2(3), 260– icol. Rev. 2003, 22(1), 29–40. 268. 54 C. E. Adkins, et al., GABA(A) receptors 45 W. A. Carroll, et al., Synthesis and characterized by fluorescence resonance structure-activity relationships of a no- energy transfer-derived measurements vel series of tricyclic dihydropyridine- of membrane potential, J. Biol. Chem. based KATP openers that potently inhi- 2001, 276(42), 38934–38939. bit bladder contractions in vitro, J. Med. 55 A. J. Smith, P. B. Simpson, Methodolo- Chem. 2004, 47(12), 3180–3192. gical approaches for the study of GA- 46 M. Gopalakrishnan, et al., Characteri- BA(A) receptor pharmacology and func- zation of the ATP-sensitive potassium tional responses, Anal. Bioanal. Chem. channels (KATP) expressed in guinea 2003, 377(5), 843–851. pig bladder smooth muscle cells, J. 56 A. J. Smith, B. Oxley, S. Malpas, G. V. Pharmacol. Exp. Ther. 1999, 289(1), 551– Pillai, P. B. Simpson, Compounds ex- 558. hibiting selective efficacy for different 47 K. L. Whiteaker, R. Davis-Taber,V. E. beta subunits of human recombinant Scott, M. Gopalakrishnan, Fluores- gamma-aminobutyric acid A receptors, cence-based functional assay for sarco- J. Pharmacol. Exp. Ther. 2004, 311(2), lemmal ATP-sensitive potassium chan- 601–609. nel activation in cultured neonatal rat 57 G. V. Pillai, A. J. Smith, P. A. Hunt, ventricular myocytes, J. Pharmacol. Toxi- P. B. Simpson, Multiple structural fea- col. Methods 2001, 46(1), 45–50. tures of steroids mediate subtype-selec- 48 M. Gopalakrishnan, et al., Pharmacol- tive effects on human alpha4beta3delta ogy of human sulphonylurea receptor GABAA receptors, Biochem. Pharmacol. SUR1 and inward rectifier K(+) channel 2004, 68(5), 819–831. Kir6.2 combination expressed in HEK- 58 P. M. Quinton, Missing Cl conduc- 293 cells, Br. J. Pharmacol. 2000, 129(7), tance in cystic fibrosis, Am. J. Physiol. 1323–1332. 1986, 251(4 Pt. 1), C649–652. 49 G. C. Terstappen, A. Pellacani, 59 J. R. Riordan, et al., Identification of L. Aldegheri, F. Graziani, C. Carig- the cystic fibrosis gene: cloning and nani, G. Pula, C. Virginio, The anti- characterization of complementary depressant fluoxetine blocks the human DNA, Science 1989, 245(4922), 1066– small conductance calcium-activated 1073. potassium channels SK1, SK2 and SK3, 60 R. C. Boucher, New concepts of the Neurosci. Lett. 2003, 346(1–2), 85–88. pathogenesis of cystic fibrosis lung dis- 50 T. Brismar, S. Anderson,V. P. Col- ease, Eur. Respir. J. 2004, 23(1), 146–58. lins, Mechanism of high K+ and Tl+ 61 H. Yang, et al., Nanomolar affinity uptake in cultured human glioma cells, small molecule correctors of defective References 211

Delta F508-CFTR chloride channel gat- 67 R. C. Rubenstein, P. L. Zeitlin, A pi- ing, J. Biol. Chem. 2003, 278(37), lot clinical trial of oral sodium 4-phenyl- 35079–35085. butyrate (Buphenyl) in deltaF508-homo- 62 R. M. Wachter, S.J Remington, Sensi- zygous cystic fibrosis patients: partial tivity of the yellow variant of green restoration of nasal epithelial CFTR fluorescent protein to halides and ni- function, Am. J. Respir. Crit. Care Med. trate, Curr. Biol. 1999, 9(17), R628–629. 1998, 157(2), 484–490. 63 R. M. Wachter, D. Yarbrough, K. Kal- 68 B. D. Moyer, B.A Stanton, Analysis of lio, S. J. Remington, Crystallographic CFTR trafficking and polarization using and energetic analysis of binding of se- green fluorescent protein and confocal lected anions to the yellow variants of microscopy, Methods Mol. Med. 2002, green fluorescent protein, J. Mol. Biol. 70, 217–227. 2000, 301(1), 157–171. 69 C. R. Valdivia, et al., A trafficking de- 64 A. S. Verkman, S. Jayaraman, Fluores- fective, Brugada syndrome-causing cent indicator methods to assay func- SCN5A mutation rescued by drugs, Car- tional CFTR expression in cells, Meth- diovasc. Res. 2004, 62(1), 53–62. ods Mol. Med. 2002, 70, 187–196. 70 E. Ficker, C. A. Obejero-Paz, S. Zhao, 65 T. Kuner, G. J. Augustine, A geneti- A. M. Brown, The binding site for cally encoded ratiometric indicator for channel blockers that rescue mispro- chloride: capturing chloride transients cessed human long QT syndrome type in cultured hippocampal neurons, Neu- 2 ether-a-gogo-related gene (HERG) ron 2000, 27(3), 447–459. mutations, J. Biol. Chem. 2002, 277(7), 66 R. C. Rubenstein, M. E. Egan, P. L. 4989–4998. Zeitlin, In vitro pharmacologic restora- 71 F. Yan, C. W. Lin, E. Weisiger, E. A. tion of CFTR-mediated chloride trans- Cartier, G. Taschenberger, S. L. port with sodium 4-phenylbutyrate in Shyng, Sulfonylureas correct traffick- cystic fibrosis epithelial cells containing ing defects of ATP-sensitive potassium delta F508-CFTR, J. Clin. Invest. 1997, channels caused by mutations in the 100(10), 2457–65. sulfonylurea receptor, J. Biol. Chem. 2004, 279(12), 11096–11105. 213

9 Approaches for Ion Channel Structural Studies Randal B. Bass and Robert H. Spencer

9.1 Introduction

Biological membranes are fascinating and essential components for all living cells. Whether they are actively transporting molecules into or out of the cell, transducing a signal, or allowing solutes to cascade down a concentration gradient, decades of work in numerous laboratories has illustrated the remarkable properties of these proteins that operate within what is essentially a thin layer of very complicated grease. Since the first atomic resolution structure of a membrane protein was reported over 20 years ago (the bacterial photosynthetic reaction center from Rhodopseudomonas viridis [1]), a wealth of information about how membrane proteins function has been gained from the determination of high-resolution structures from diverse classes of membrane proteins. The ability to visualize the structure of these proteins, as has been done for their soluble counter- parts, has allowed keener understanding of their folding and functions. However, it has only been in the last decade that significant progress on the structure of ion channel proteins has been seen. Many advances in the field of membrane protein structural biology have con- tributed to the growing number of these proteins in the databank. For ion channels, breakthroughs on solving their structures have primarily come from the identification of bacterial homologs which could be expressed in sufficient quantities for labor-intensive crystallization trials. Although general structural information on mammalian ion channels has been obtained using electron microscopy, nearly all of the X-ray crystal structures of ion channels have come from bacterial sources – primarily due to the difficulties associated with their expression in heterologous systems and establishing biochemical conditions to stabilize them outside the lipid membrane. However, very recently the crystal structure of the first mammalian K+ channel has been reported and will hopefully be a harbinger of advances yet to come [2]. In this chapter, we hope to provide a broad overview of the issues and advances made in pursuing membrane proteins such as ion channels for structural studies, and touch on topics that are specifically relevant to their expression, biochemistry and structural analysis. In summarizing these advances,

we also aim to highlight those researchers who are at the cutting edge and changing the way in which the sometimes daunting problem of obtaining high-resolution structures is being approached. So just how many membrane proteins are out there, and is it really worth pursuing more? Most estimates that have come from sequencing the genomes from everything from humans to bacteria suggest that roughly 30% of all proteins in the genomes are membrane proteins. For humans, that means around 10000 of the ~30 000 protein-coding genes can be found in the lipid bilayer. Several authors have noted that the growth of structures for membrane proteins follows a similar trajectory to those of soluble proteins. In the 20 years following the first high-resolution structure of myoglobin (determined by Kendrew et al. in 1958 [3]), 132 stable protein structures had been solved whereas by 2003, in a similar period since the first membrane protein structure, 75 unique membrane protein structures had been solved [4]. Progress on the structures of membrane proteins is clearly slower than for soluble proteins, but it is possible that the end of 2005 will see the high-resolution determination of the 100th unique membrane protein. Many of these are structures of homologous proteins from other organisms, thus the number of unique folds is somewhat smaller. In stark contrast, the Protein Data Bank (www.rcsb.pdb.org) currently has nearly 30 000 deposited protein structures in total. One of the great hopes of membrane protein structural biology that drives the desire for more structures is that the knowledge gained can be applied to the treatment of disease and help refine the development of better, targeted drugs. Determining the structure of membrane proteins will not only continue to en- lighten the molecular and structural basis for their functions, but is also likely to have a broader impact by facilitating lead optimization for the development of novel therapies. Due to the critical role that ion channels play in neuronal and cell signaling and ion homeostasis, they are, and will continue to be, valuable molecular targets for the treatment of human disease. Approximately one third of currently marketed drugs produce their pharmacological actions via ion channels; this also includes around 15% of the top 100 best-selling medicines [5]. However, the development of novel, selective ion channel modulators has been difficult with only a single new ion channel drug being approved since 1997; the exception  being ziconotide (Prialt ), a recently approved Ca2+-channel blocking peptide for the treatment of severe chronic pain. Given current estimates of over 400 ion channel genes encoded within the human genome, there is an abundance of ion channel targets [6], and yet we have no structural information for any of these proteins. For that matter, there currently is not a single X-ray structure of a human transmembrane protein. Fortunately, we have the structure of a handful of bacterial and archaeal ion channels at atomic resolution which have been extremely en- lightening and valuable for molecular modeling purposes (see Chapter 10), and now the first structure of the first mammalian K+ channel [2]. However, high-resolution structures of the many other mammalian ion channels, especially human, remain a significant need. A major breakthrough for structural biology, as for all branches of biology, was the molecular biology revolution that allowed the recombinant expression of vir- 9.1 Introduction 215 tually any protein. Reviewing the methods used to obtain protein samples used in crystallization attempts, one can quickly understand that the ability to clone, express and purify recombinant proteins is nearly universal in its application to obtaining structures. This technique has been widely applied in membrane proteins with great success. However, unlike the myriad of soluble bacterial and eukaryotic proteins successfully expressed in recombinant systems, there has been only limited success for the expression, purification and solution of a eukaryotic membrane protein crystal structure produced from a recombinant system. Certainly there are many cases of expression of membrane proteins in these systems, some of which have been useful for structural studies using electron microscopy (EM) [7–9]. The recent structure of the rat Kv1.2-b2 solved from protein expressed in the yeast Pichia pastoris is a tremendous advance and represents a significant breakthrough in the field that may open the door for many other mammalian membrane proteins [2, 10]. Why are the membrane proteins so recalcitrant to expression in these systems when the soluble proteins are not? One potential issue may be the vastly different lipids that make up eukaryotic bilayers, more specifically mammalian, compared to those that are found in prokaryotes. Whereas human cell membranes are composed largely of phosphatidylcholine, phosphatidylethanolamine, sphingomyelin and cholesterol, the E. coli inner membrane is mainly composed of phosphatidylethanolamine (75%) [11]. The remainder of the lipid content in E. coli is largely cardiolipin and phosphatidylglycerol, neither of which are prevalent in human bilayers [12]. Even with a fantastic expression system, once a eukaryotic membrane protein enters the inner membrane of E. coli, it finds a lipid environment far different from the one in which it would normally reside. Thus, it is not surprising that many eukaryotic membrane proteins fail to express in E. coli, and if they do express they typically end up in inclusion bodies. This presents a new problem. Un- like soluble proteins, the successful refolding from inclusion bodies of a complex membrane protein containing multiple transmembrane helices has not yet been widely reported, although there is a proprietary technology reported by m·phasys GmbH (Tübingen, Germany) which suggests this approach may be feasible. An- other promising advance was recently reported using a membrane-integrating protein called Mistic (membrane integrating sequence for translation of integral membrane protein constructs) identified from Bacillus subtilis that, when linked to the N-terminus of a eukaryotic membrane protein, produced significant quantities of several target proteins, including a voltage-gated K+ channel [13]. Although relatively hydrophilic, Mistic folds autonomously within the bacterial membrane and appears to facilitate the integration and folding of attached eukaryotic ‘cargo’ membrane proteins that have been extremely difficult to express in prokaryotes. This and other nascent technologies should be extremely valuable in moving forward the entire field of membrane protein structural biology. Certainly it is possible for mammalian membrane proteins to be expressed and correctly folded in bacterial cells, as several examples have been cited in the literature [8, 9, 14]. However, it is likely that many will not. This fact underscores the 216 9 Approaches for Ion Channel Structural Studies

need to broaden the search for suitable candidates for expression in these systems. One means to address this is using a multiple orthologue approach such as that applied in the Rees [15–17] and Chang laboratories [18, 19]. In all of these cases, orthologous bacterial proteins were expressed in E. coli to find combinations that could produce sufficient quantities of protein that were properly folded, and would produce crystals of sufficient diffraction quality. If these examples of bacterial membrane proteins are any indicator, the successful expression of eukaryotic membrane proteins is also likely to require this same approach.

9.2 Expression of Membrane Proteins for Structural Studies

9.2.1 Mammalian Expression

For mammalian membrane proteins, including ion channels, it seems intuitive to seek to isolate these from native tissues or to express them heterologously in cell culture. However, since ion channels are so efficient (with flux rates of ~107 ions s–1) mammalian cells natively express relatively low levels of these proteins, typically 10–1000 channels of any one type per cell, and thus are not good sources for structural studies. Nevertheless , there are a few examples (see Table 9.2, below) where sufficient quantities have been obtained for single-particle electron microscopy (EM) including the TRPC3 cation channel [20], L-type Ca2+ channel [21],

type 1 IP3 receptor [22–24], and the ryanodine receptor [25–27]. There are also a few nonmammalian examples of abundant natural sources of ion channels which have been exploited for EM structures, including the nicotinic acetylcholine receptor (nAChR) from Torpedo ray [28] and the voltage-gated Na+ channel from electric eel [29]. Stably transfected recombinant cell lines are also relatively poor sources of ion channel protein as expression levels at the cell membrane rarely exceed a few thousand channels per cell. Additionally, there can often be a large pool of imma- ture, incorrectly processed or folded protein remaining within the cell (see Chap- ter 4) resulting in a heterogeneous protein sample upon purification. One way to resolve this is by purifying only the mature, fully glycosylated channels. For exam- 2+ ple, in EM studies of both the L-type (dihydropyridine) and IP3 receptor-type Ca channels, wheat germ lectin affinity chromatography was commonly used during their purification [30–32], and was also employed for the 20 Å resolution structure of human CFTR [33]. Another approach is to simply remove oligosaccharide altogether. In fact, given the heterogeneous nature of glycosylation the removal of oligosaccharide is likely to be beneficial, perhaps even essential, for high-resolution structural studies. This was found to be critical for obtaining well diffracting 2D crystals of human aquaporin-1 [34, 35] and, in the cases of Shaker, aquaporin-1, Kv1.3, and SkM1 the effects of deglycosylation on the function of these channels is minimal 9.2 Expression of Membrane Proteins for Structural Studies 217

[36–39]. Additionally, synchrotron radiation circular dichroism spectroscopy and thermal denaturation studies on the E. electricus Na+ channel show no net change in protein secondary structure resulting from the deglycosylation procedure [40]. In summary, given the current limitations on protein processing and expression level, the development of improved mammalian expression systems is likely to be required before this becomes a realistic option for the production of membrane proteins for structural studies. Viral expression systems, such as Semliki Forest virus (SFV) or BacMam (see Chapter 4), may provide a solution in the future.

9.2.2 Insect Expression

Employing insect cells for the expression of eukaryotic channels has shown some potential and has several advantages compared to mammalian systems since the cell culture is simpler and easier to scale up. The Drosophila Shaker K+ channel was one of the first ion channels successfully expressed heterologously in insect cells for the purpose of biochemical and structural studies [41]. The human Kv1.1 channel has been functionally expressed using baculovirus in Sf9 cells and proven useful for biochemical studies [42]. Based on whole-cell currents, the cell membrane contains no more than a few thousand channels per cell, but it is possible that a much larger intracellular pool of protein may be recoverable for structural studies. It has also been applied for biochemical and biophysical studies of CFTR, the glycine receptor, the KATP channel, and the heteromeric amiloride Na+ channel complex [43–46]. However, in spite of these advances, the general utility of insect cells for producing reasonable quantities of intact mammalian ion channel protein for structural studies is still in question. Although, the first images of the tetrameric structure of a voltage-gated K+ channel came from protein expressed in Sf9 cells, there has been little progress for other mammalian ion channels [47]. In the case of CFTR, no structural information has been gained for this channel expressed in insect cells. However, recently, 2D crystals of this channel were reported using heterologous expression in mammalian cells (BHK cells) that provided structural information at a resolution of 20 Å [33]. In spite of these issues, it should be recognized that the crystal structure of many soluble proteins has been obtained using protein derived from heterologous expression in insect cells and thus may yet prove similarly valuable for membrane proteins.

9.2.3 Yeast Expression

Although the possibility of using yeast for the production of mammalian membrane proteins has been pursued for many years, it was not until very recently that this was successfully applied for material suitable for structural studies. In 2003, Parcej and Eckhardt-Strelau reported the expression, purification and struc- 218 9 Approaches for Ion Channel Structural Studies

ture of the rat Kv1.2 channel in complex with the Kvb2 subunit at 21 Å resolution using single-particle electron microscopy [10]. The Kv1.2-b2 protein was expressed in the methylotrophic yeast Pichia pastoris and found to be in a relatively homogeneous form following the removal of a consensus N-glycosylation site and several putative phosphorylation sites. This is certainly one of the first breakthrough methods for the production of milligram quantities of eukaryotic membrane protein that appears to be both properly folded and functionally active based on ligand-binding analysis. This method was recently adapted for the determination of the crystal structure of this channel at 2.9 Å resolution [2] – a landmark achieve- ment. It is certainly hoped that this system can be broadly used to explore the structure of many other mammalian membrane proteins in addition to ion channels.

9.2.4 Bacterial Expression

The heterologous expression of proteins in E. coli cells is certainly one of the most widely used systems applied to soluble proteins. However, in the case of membrane proteins, it is only in the past decade that significant progress has been seen on the successful application of this system for X-ray and EM structural studies. It was not until 1998 that the first successful X-ray crystal structures of ion channel proteins were determined (KcsA and MscL), and these paved the way for many other ion channel structures that were determined using similar methods [15, 48]. As clearly summarized in Table 9.1, with a single exception [2], all of the known ion channel structures determined using X-ray crystallography were derived from protein expressed and purified using E. coli as the host. The primary advantage of using a prokaryotic expression system is the ability to easily scale-up the process and the tremendously lower costs of production. In general, relatively traditional methods have been successfully employed for the expression of bacterial membrane proteins in E. coli. Plasmids such as the pET (Novagen, Madison, WI) or pQE (Qiagen,Valencia, CA) vectors are the most commonly utilized for expression in suitable E. coli strains [BL-21 (DE3), for pET vectors; XL-1 or SG13009 (pREP4) for pQE vectors]. Although the process appears relatively straightforward, the over expression of ion channels, or other membrane proteins, is often toxic to E. coli, requiring optimization of media, temperature, cell density and induction conditions for each protein of interest. Additionally, we have observed that over expression of these proteins will often result in cellysis at later stages of induction. Thus, it is necessary to manage this process carefully and consistently by harvesting the cells before this occurs. For bacterial membrane proteins, the use of E. coli will continue to be an important means for bringing forward new structures. However, the expression of eukaryotic membrane proteins in E. coli is still a significant challenge. There are scattered reports of limited success for a few mammalian membrane proteins, but none of these has yet been successful at providing any structural information using either EM or X-ray methods. The recent report on the expression of several 9.3 The Detergent Factor 219 eukaryotic membrane proteins in E. coli, including a GPCR and an ion channel, using the autonomously folding integral membrane protein, Mistic, is very en- couraging [13]. It is hoped that this and similar advances will pave the way for the structure determination of the many eukaryotic membrane proteins that still evade our grasp.

9.3 The Detergent Factor

Successful crystallization is only possible if the membrane protein, once removed from the bilayer, is properly folded in a detergent. The list of detergents used in successful crystallizations is both lengthy and diverse. A table of successfully used detergents is shown in Fig. 9.1. The a-helical type membrane proteins have seen the greatest success n-octyl-b-d-glucopyranoside (OG), N,N-dimethyldodecyla- mine-N-oxide (LDAO) and n-dodecyl-b-d-maltopyranoside (DDM). Octylglucoside stands out as the detergent that has been used in the most successful crystallizations. However, it is critical to note that the success rates partly reflect the length of time that each of these have been available. Thus, it is not clear if there some- thing universal about OG and LDAO that makes them so useful, or simply if they have they seen the greatest success because they have been utilized in membrane protein biochemistry for much longer. Similarly, chain length trends for the detergents, where multiple chains lengths are available, have been frequently speculated about. Again, it is important to note that the many variations within a given detergent family, such as the glucosides, maltosides and so forth, have come about only recently, precluding clear conclusions. While it is possible to observe some general correlations, for example a-helical-type proteins are more apt to crystallize in glucopyranosides then they are in polyoxyethylenes, few other trends beyond such cursory conclusions can be defined at present. It is no surprise that removal of membrane proteins from their native lipid environment poses one of the most serious challenges to maintaining a native fold conducive for crystallization trials. Certainly one of the most significant advances for membrane protein crystallography is the availability of a myriad of high quality, diverse detergents. In this respect, the work of the scientists at Anatrace (Mau- mee, OH) is of particular note. The development of several new detergent series, CYMAL, FOS-CHOLINE, FOS-MEA, CYFOS and C-HEGA was aided through a NIH Small Business Innovative Research grant. Recognition by the funding agen- cies of the need for better reagents for novel research is to be congratulated and encouraged. The effects of detergent on membrane proteins cannot be overstated. Proper choice of detergent will greatly impact on whether crystallization is possible. The choice of detergent can be broken into two critical components. The first step is to identify a detergent that can successfully extract the protein from the bilayer. Membrane proteins often contain multiple subunits so the detergent must be able to extract the protein while maintaining both its tertiary and its quaternary 220 9 Approaches for Ion Channel Structural Studies

Fig. 9.1 Graph of detergents successfully with transmembrane spanning beta strands. used for membrane protein structure determi- Alpha type proteins, found in both prokaryotic nation. Data from the compilation of known and eukaryotic membrane proteins, span the membrane protein structures at http:// bilayer with alpha helices. The x-axis denotes www.mpibp-frankfurt.mpg.de/michel/public/ the number of successful structure determi- memprotstruct.html. Beta-type proteins refer nations in each of the listed detergents. to outer membrane proteins of prokaryotes

structure. As illustrated in Fig. 9.2A, one can design screens to identify optimal expression constructs using multiple homologs, expression vectors, varying types or location of affinity tags, using a few key detergents. Once an optimized expression system has been identified, one can then broadly screen for detergent solubility, relative purity and integrity using a 96-well format as illustrated in Fig. 9.2B. Further analysis of the structural integrity of the protein must then be carried out. While spectroscopic techniques such as circular dichroism are helpful for determining the state of tertiary structure, assessing quaternary structure can be more challenging. Size-exclusion chromatography, although low resolution, works quite 9.3 The Detergent Factor 221

Fig. 9.2a Expression and detergent screening. developed for protein expression. Following (A) Schematic diagram for expression con- induction of protein expression (typically struct optimization and detergent screening. using a range of conditions), culture samples Multiple constructs, homologs, and deter- are individually aliquoted, lysed and solubi- gents can be simultaneously screened to ra- lized with a panel of unique detergents. pidly identify the optimal conditions for ex- Cleared lysates are analyzed by Western blot pression and solubilization. Ideally, each using an antibody to the affinity tag to deter- open reading frame (ORF) is cloned into sev- mine relative expression for each construct eral expression vectors that place an affinity expressed in each bacterial strain, as well as tag at either the N- or C-terminus. These are their relative detergent solubility. transfected into a series of bacterial strains 222 9 Approaches for Ion Channel Structural Studies

Fig. 9.2b Expression and detergent screening. quently eluted into a collection plate (avail- (B) Schematic diagram of a high throughput able from several vendors including Novagen, detergent screening procedure. To determine San Diego, CA; Qiagen,Valencia, CA; Vi- the relative solubility and yield for a mem- vascience, Inc., Edgewood, NY). Protein re- brane protein using a broad series of deter- covery can be evaluated by UV spectrometry gents, a 96-well extraction and purification or other protein assay, and purity can be vi- procedure can be performed using a robotic sualized by high-throughput SDS-PAGE (Invi- liquid-handling system or manually. Briefly, a trogen, Carlsbad, CA). Protein identity can be panel of unique detergents is added to each confirmed by Western blot analysis and mass well of a deep 96-well block containing the spectrometry, and functional and/or struc- cell lysate. After shaking gently, affinity resin tural integrity can be evaluated by size exclu- is added across the wells to bind the target sion chromatography (SEC) or ligand-bind- protein. The resin is then transferred to a 96- ing. well filter plate where it is washed, and subse-

well for this, in spite of the presence of detergent micelles. Macromolecules and complexes of over 200 kDa can be routinely resolved with size-exclusion chromatography and perhaps more importantly, aggregated protein in the sample can be identified. The presence of aggregated protein is a significant obstacle to successful crystallization. Every attempt should be made to eliminate it, or significantly reduce it, either by reengineering constructs, altering growth and production of the protein or by chromatography, prior to setting up crystallization trials. To assess whether the protein of interest is maintained as a higher order, oligomeric structure in any detergent one can crudely, but quickly, address this using ultra filters of an appropriate molecular weight cutoff and simply look for the oligomeric protein in the retentate versus the filtrate (i.e. monomeric protein will 9.4 Purification 223

Fig. 9.3 Crosslinking to determine proper oligomeric state. Polyacrylamide gel showing crosslinking of two homologs of MscL from M. tuberculosis (denoted Tb) and E. coli (Eco) with disuccinimidyl suberate (DSS). Shown on the left side of the gel are the positions of the molecular weight markers.. The two left hand lanes are monomers of the homologs without addition of the bifunctional cross-linking reagents DSS. The two right lanes are with the addition of DSS, showing a clear ladder of monomer, dimer, trimer, tetramer and finally pentamer that is the full oligomeric structure of MscL, as seen in the crystal structure. Reprinted with permission from Ref [15].

pass through the filter). A more visual approach uses bifunctional cross linking reagents to covalently link subunits together and subsequently visualize them on SDS-PAGE gels, as shown in Fig. 9.3. This technique has been used to determine the oligomeric state of the ion channel MscL [15]. An enhancement of this technique is cross linking in conjunction with capillary electrophoresis (CE) employing a sieving gel. The advantage of this latter technique is the ability to easily quantify peaks in an electropherogram and deduce the stoichiometry of a complex macro- molecule. Although densitometry of Coomassie-stained gels can also accomplish this, CE offers the advantage of high resolution, ease of quantification, and excellent separation, even between subunits of similar molecular weight. On a practical note, although it may be advantageous to make membrane preparations for large-scale purification of membrane proteins, this technique may be unnecessary and misleading if used in conjunction with detergent screens. Care must be taken whenever sonication is used for detergent screening to avoid false positive results. Heavy sonication can produce very small (~5 mM diameter) vesicles that can remain suspended while attempting to spin down a detergent screen trial. If membrane preparation by sonication is employed, it is necessary to apply ultra- centrifugation to pellet any small membrane vesicles, prior to gel and analysis.

9.4 Purification

The first membrane protein structures were all derived from naturally abundant proteins. For example Deisenhofer’s original membrane protein structure, the photosynthetic reaction center from Rhodopseudomonas virdis, is found in purple patches in the membrane in a highly concentrated “pre-purified” form [1]. Addi- tionally, the first GPCR, bovine rhodposin, occurs naturally at high concentration in the outer rod segments and is relatively pure in that membrane. Although determining the structure of both these proteins certainly presented numerous challenges, considerable additional challenges are presented by proteins found in eu- 224 9 Approaches for Ion Channel Structural Studies

Table 9.1 Ion channel structures determined by X-ray crystallography. Examples of crystal structures determined for ion channels using X-ray crystallography.

Name Ref. Year Ion PDB ID Resolution Subunits a) TM a-helices Species Selectivity (A˚ ) (per subunit)

KcsA 48 1998 K+ 1BL8 3.2 4 2 S. lividans MscL 15 1998 Non-selective 1MSL 3.5 5 2 M. tuberculosis KcsA b) 69 2001 K+ 1K4C 2.0 4 2 S. lividans EcCIC 70 2002 Cl– 1KPK 3.5 2 17 E. coli MscS 17 2002 Non-selective 1MXM 3.9 6 3 E. coli Mthk 71 2002 K+ 1LNQ 3.3 4 2 M. thermoautotrophicum StClC 70 2002 Cl– 1KPL 3.0 2 17 S. typhimurium EcClC 72 2003 Cl– 1OTS 2.5 2 17 E. coli KirBac1.1 73 2003 K+ 1P7B 3.6 4 2 B. pseudomallei

KvAP 58 2003 K+ 1ORQ 3.2 4 6 A. pemix Kv1.2-b-2 2 2005 K+ 2A79 2.9 4 6 R. norvegicus

Table 9.2 Ion channel structures determined by electron microscopy. Examples of ion channel structures determined using either single particle electron microscopy or electron crystallography from two-dimensional protein crystals.

Year Name Ref. Ion Resolution Subunits a) TM a-helices Selectivity (A˚ ) (per subunit)

1994 Shaker 47 K+ n/a 4 6

1995 5-HT3 receptor 74 Cation n/a 5 4 1997 Kv1.3 38 K+ n/a 4 6 1998 KcsA 75 K+ 642 1999 Cx43 76 Non-selective 7.5 12 4 2001 EcCIC 77 Cl– 6.5 2 17 2001 Voltage-sensitive 29 Na+ 19 1 24 Na+ channel 2001 Shaker 78 K+ 25 4 6 2003 L-type Ca2+channel 21 Ca2+ 23 1 24 2003 nAChR 28 Cation 4.0 5 4 2003 Kv1.2-b210K+ 21 4 6 2004 CFTR 33 Cl– 20 1 12 2+ 2004 IP3 receptor 79 Ca 15 4 6 2004 Kv4.2 / KChlP2 80 K+ 21 4 6 2004 KvAP 81 K+ 10.5 4 6 2005 AMPA receptor 82 Na+ 42 5 3

2005 Ryanodine receptor 83 Ca2+ 10 4 6–8 2005 TRPC3 20 Cation 30 4 6 a) Stoichiomety for pore-forming domain. b) Complex with Fab fragment. 9.4 Purification 225

Table 9.1 (continued)

Host Affinity Tag Tag Detergent Detergent Removal (Extraction) (Crystallization)

E. coli His6, C-term Yes n-Decyl-b-D-maltoside LDAO E. coli His10, N-term No n-Dodecyl-b-D-maltoside n-Dodecyl-b-D-maltoside E. coli His6, C-term Yes n-Decyl-b-D-maltoside n-Decyl-b-D-maltoside E. coli His6, C-term Yes n-Decyl-b-D-maltoside n-Octyl-b-D-maltoside E. coli His6, N-term No Fos-Choline-14 Fos-Choline-14 E. coli His6, C-term Yes n-Decyl-b-D-maltoside LDAO E. coli His6, C-term Yes n-Decyl-b-D-maltoside n-Octyl-b-D-maltoside E. coli His6, C-term Yes n-Decyl-b-D-maltoside n-Decyl-b-D-maltoside E. coli His6, C-term No n-Decyl-b-D-maltoside + Cymal-4, HEGA-10 n-Tridecyl-b-D-maltoside

E. coli His6, C-term Yes n-Decyl-b-D-maltoside n-Octyl-b-D-glucoside P. pastoris His6, N-term No n-Dodecyl-b-D-maltoside n-Decyl-b-D-maltoside

Table 9.2 (continued)

Species Host Affinity Tag Tag Detergent Removal

D. melanogaster Sf9 (Baculovirus) – – CHAPS M. musculus NG 108-15 – – n-Dodecyl-b-D-maltoside

H. sapiens CV-1 (Vaccinia) His6, N-term N CHAPS S.lividans E. coli His6, N-term N n-Dodecyl-b-D-maltoside R. norvegicus BHK – – Tween-20

E. coli E. coli His10, C-term Y n-Dodecyl-b-D-maltoside E. electricus Native – – Lubrol- PX

D. melanogaster COS 1D4, C-term N CHAPS O. cuniculus Native – – Digitonin T. mamorata Native – – –

R. norvegicus P. pastoris His9, N-term N Tween-80 H. sapiens BHK His10, C-term N n-Dodecyl-b-D-maltoside M. musculus Native – – CHAPS H. sapiens COS7 1D4, C-term N CHAPS

A. pernix E. coli His6, C-term Y n-Decyl-b-D-maltoside R. norvegicus Native – – CHAPS or n-Decyl-b-D-maltoside O. cuniculus Native – – CHAPS M. musculus HEK293 FLAG, C-term N n-Decyl-b-D-maltoside 226 9 Approaches for Ion Channel Structural Studies

karyotic plasma membrane where many proteins of interest to both the academic and commercial structural biologist reside, such as transporters, channels and receptors. These proteins are found typically in low abundance and in a membrane that is host to hundreds of other proteins. It would be difficult, if not impossible, to obtain structural information for these proteins without the use of recombinant technology in conjunction with affinity purification tags. Many affinity tag systems are commercially available, such as FLAG, maltose binding protein, streptavidin, thioredoxin, and many more. A near universal tool for this approach is the use of polyhistidine tags and immobilized metal chelate chromatography (IMAC). So pervasive is its application, that it has been a significant technology for the high throughput world of structural genomics as well as a core technique for many membrane protein crystallography laboratories. The use of these tags is now so routine that they need not be thoroughly discussed here. Indeed, as shown in Tables 9.1 and 9.2, nearly all of the recombinant expressed ion channel proteins for which structures have been determined were prepared using some type of polyhistidine tag. However, a few specific points regarding the use of polyhistidine tags for membrane proteins should be made.

Although hexahistidine (His6) tags were originally used, His8,His10 and even longer tags are now more common although, in our hands, the utility of tags

longer than His10 is not clear. A critical advantage of polyhistidine tags is their ability to efficiently bind metal columns in the presence of detergents. Although other affinity tags can also be used in the presence of detergent in virtually all cases, our experience supports the use of polyhistidine tags in conjunction with the >60 detergents that are routinely used in our screens. In short, the presence of detergents above their critical micellar concentration (CMC) does not preclude the use of this type of tag, though it should be noted that the efficiency with which the tag binds the column is often diminished. It is this reduction in affinity that has led to the use of vectors encoding the elongated tags listed above. In the light of lowered affinity in the presence of detergent, very high levels of imidazole should be carefully tested when washing and eluting the column, in order to obtain maximum purity while balancing the need for high yield. These opposing forces can easily be monitored, even at very small scale, by employing Wes- tern blots with anti-hexahistidine antibodies which readily bind the longer polyhistidine tags, and which are commercially available and inexpensive. Regardless of the affinity tag used, the ability to manipulate the detergent in this capture step is often of great importance. The detergent used for extraction from the membrane may not be one in which the protein remains stable for long periods of time, or it may not prove successful in crystallization. At the affinity purification step, which is usually the first step after extraction, detergent concentration is often high (e.g. 0.1% or more), depending on the CMC of the detergent. This step is therefore an opportunity to lower the detergent concentration to a level at which, though still above the CMC, the protein remains stable and which allows for at least the possibility of crystal formation [19–23]. The organization of protein molecules into a three-dimensional crystal lattice requires the formation of intermolecular crystal contacts which stabilize its archi- 9.5 Crystallization 227 tecture. However, due to associated detergent, there is significantly less available surface area on purified membrane proteins for making these necessary contacts. The crystallization and X-ray structure of soluble proteins fused to large affinity tags, such as the maltose-binding protein (MBP), thioredoxin (TRX), or glutathione-S-transferase (GST), has recently been reported, and it has been suggested that these tags may aid membrane protein crystallization by increasing the available hydrophilic surface area [49]. This technique would be analogous to the approach taken by Kaback and coworkers using cytochrome b562 fused to lac permease [50], or Iwata and coworkers attaching protein Z to cytochrome bo3 to facilitate crystallization [51]. This approach has not yet been proven successful for obtaining high-resolution structures of any membrane protein, and moreover, in the two examples mentioned, their structures were finally solved as His-tag fusions [52, 53]. However, since MPB is targeted to the periplasm, some have speculated that an N-terminal fusion with MBP could potentially “drag” the first transmembrane spanning region through the bilayer, providing a critical foothold to proper folding and further insertions into the membrane. In addition another possible benefit is the “chaperone-like effect” provided by MBP, enhancing the likelihood of a proper fold. The successful expression of the human Na+/glucose co-transporter illustrates both the use of an alternative affinity tag, as well as the expression of functional human membrane protein in E. coli [9]. The authors used a FLAG-tagged protein expressed in E. coli to generate 0.3 mg of pure protein per liter of culture. The use of a lac promoter system, in simple LB medium, at reduced temperature (20 8C), and a bacterial strain defective in the outer membrane protein OmpT, illustrates the successful expression and purification of a human membrane protein in a bacterial system. Perhaps more importantly, the protein produced was functional when reconstituted into liposomes. This level of expression of a functional protein illustrates the utility of bacterial expression when one considers that a scale up to a fermentor would easily achieve the protein production required for crystallization attempts. Moreover, a functional assay, such as sugar transport in this example, adds to the likelihood that this method can produce material that has the ability to be crystallized.

9.5 Crystallization

In general, once successfully purified in a detergent-containing buffer system in which it is stable, the crystallization of a membrane protein, though challenging, is not that much different from that of soluble proteins. Crystals for membrane proteins are obtained either in batch mode or from vapor diffusion, with both hanging and sitting drop formats being successfully used. Sitting drops have the advantage of easier use with 96-well plates, enabling faster screening of more conditions, and smaller drop sizes, thus preserving valuable supplies of painstakingly purified protein as far as possible. An online list of crystal screening kits and the 228 9 Approaches for Ion Channel Structural Studies

composition of each tube or well is maintained at the Structural Biology La- boratories at the University of Uppsala (http://xray.bmc.uu.se/markh/php/xtal- screens.php). Membrane protein specific screens have been developed, most nota-  bly the MembFac screen originally developed by Michael Stowell and commer- cialized by Hampton Research, which was the first attempt at compiling the conditions used to successfully crystallize membrane proteins into a single screen [54]. As the number of successful structures continues to increase so does the breadth of conditions used for successful crystal formation. Making use of the commercially available screens is an excellent place to begin. When attempting to determine suitable conditions, if availability of material allows, it is worth trying them all, since they are relatively inexpensive and cover a wide range of precipi- tants, salts, buffers and pH. Commercial screens in 96-well format from Hampton Research, Nextal Biotechnologies, Emerald Biostructures and others are available, facilitating simple application of this higher throughput format. Microfluidics has the potential to vastly extend the amount of crystallization screening possible but, as of now, these systems are still in their infancy. The purification and crystallization of the light-harvesting complex of photosys- tem II (LHC-II) illustrates several techniques useful for the extraction and stabilization of membrane proteins [55]. In this case detergent (n-nonyl-b-d-glucoside), lipid [digalactosyldiacylglycerol (DGDG)], and chlorophyll were used to solubilize the complex. The use of a combination of detergent, a critical plant lipid present in the thylakoid membrane where the complex is found, as well as chlorophyll, is readily validated from the structure. Adjacent trimers of the LHC-II complex are largely mediated by two pairs of DGDG and two pairs of chlorophyll molecules. The crystallization solution contained a second detergent, N, N-bis-(3-d-glucona- midopropyl) deoxycholamide (BigCHAP). This illustrates a highly complicated extraction and stabilization of a membrane protein but perhaps most striking is the resulting crystalline lattice. The packing of the complexes resembles that of the highly symmetric (in this case icosahedral particles) packing of viruses. In a ‘Type-II’ membrane protein crystal the detergent micelle forms a torus around the membrane-spanning region and crystal contacts are mediated by polar regions of the molecule, as described above. In the case of LHC-II, the complex exists in a proteoliposome and contacts between adjacent molecules are made through hydrophobic contacts at the lipid. The resultant ‘Type III’ membrane protein crystal is clearly formed by the judicious addition of lipid, cofactor and detergents. This may prove especially important in maintaining the correct fold of eukaryotic membrane proteins expressed in organisms that do not have bilayers with a similar lipid composition. The recent Kv1.2-b2 structure also highlights the potential importance of exogenous lipid, including phosphatidylcholine, phosphatidylethanolamine and phosphatidylglycerol, which was reported to be essential during the purification and crystallization process [2]. Thus, the use of lipids in membrane protein crystallography has proved useful and should be explored further. 9.6 Use of Antibody Fragment 229

9.6 Use of Antibody Fragments

Crystal formation by membrane proteins, even assuming a stable homogenous protein is purified, can be hampered if limited polar surface area is present. The hydrophobic surface of membrane spanning regions is coated by the detergent micelle, allowing solubilization of the protein. The polar head groups of the detergent are of course identical and lack a regular structure that can promote crystal formation and growth. Polar protein surfaces are responsible for mediating crystal contacts which can be minimal for many integral membrane proteins. One way of enlarging these regions is to use antibody fragments, either Fab domains, or Fv fragments specific for the membrane protein. These antibody fragments bind specifically, and therefore regularly, to the membrane protein and provide additional polar surface area. The fragments are also rigid in structure, allowing a higher probability of producing a regular three-dimensional lattice. The application of antibody fragments, both Fab and Fv, has been successful with membrane proteins [56–58]. One route to produce these is to simply use a well established hybridoma protocol to generate suitable monoclonal antibodies and then create Fab fragments using proteolytic cleavage [58]. This technique works well, as long as the proteolytic enzymes cut uniformly and produce a homogenous Fab sample. Alternatively, variable region fragments (Fv), can be generated, either as independent light and heavy chain fragments or as a single chain with a linker sequence. Both forms of variable region fragments are readily expressed in E. coli. This latter approach can be initiated from the more well established hybridoma/Fab approach described above, by cloning of the variable region of the antibodies using RT-PCR amplification of the mRNA from the hybridoma cell line. Several well characterized examples are in the literature [59]. Regardless of the route used to create antibody fragments, one feature that likely contributes to crystal formation is the ability of the fragments to lock a single conformation of the target protein. Of course any perturbation of one protein by another, although sometimes required for a successful structure determination, leads to speculation as to the usefulness of the resulting structure. Trans- membrane-spanning regions are particularly at risk of structural perturbation from antibody fragments since these regions are often more mobile, partly due to the lack of constraints that normally exist in the membrane. However, some membrane proteins, notably channels and transporters, are inherently flexible in these regions due to their function. The ability to “lock” the structure in a particular conformation using an antibody fragment is an extra benefit, in addition to increasing polar surface area, which can promote a successful crystallization. How- ever, it should be noted that, for one of the better known structures determined in this manner (KvAP), significant skepticism remained about the relevance of the resulting model [60–62]. Another technique to generate antibodies is to use filamentous phage. Phage display offers several advantages in the creation of antibody fragments, most notably speed and recombinant expression of the resulting fragments. Fab or scFv 230 9 Approaches for Ion Channel Structural Studies

fragments can be generated in as little as one month. Recently Röthlisberger and coworkers created a phage display library that produces Fab fragments that are stable in detergent [63]. Screening of the library was carried out in the presence of 0.1% dodecylmaltoside, with subsequent binding analysis using Biacore performed in buffer containing the same concentration of detergent. The authors also provide guidance as to which affinity tags are most useful in the screening process. The resulting antibody fragments, Fab in this case, were shown to bind to their target in the low nanomolar range in a conformationally dependent manner.

9.7 Generation of First Diffraction Datasets

One striking thing that has been observed for a variety of membrane protein structures is the variability in the quality of diffraction from crystal to crystal, as well as the sensitivity of the crystals to cryoprotection. This latter complication may be due to the high solvent content of membrane protein crystals which can typically reach 80% [15]. First diffraction images can produce less than inspiring data, when compared with a typical soluble protein. However, any diffraction is a significant step towards a structure and even the weakest of spots, often clustered just around the beamstop, are cause for celebration. Once any diffraction is obtained, a critical foothold in the process is reached, since this represents a milestone against which further modifications can be assessed. Now it is possible to assess any changes and ask, is this better or not? A long stream of blank images will likely precede this milestone, from which it is difficult to assess the impact of any changes made. As shown in Fig. 9.4, it can be seen that the first diffraction image may only produce a few spots around the beamstop, and cannot be successfully indexed with any confidence. However, the process of crystal screening and cryoprotection optimization is pursued to improve their diffraction qualities (Fig. 9.4B). Finally, in the case of MscL, heavy metal soaks of the crystals produced higher resolution reflections that produced the final structure (Fig. 9.4C and D). A similar approach has been successfully applied for the structures of the bacterial transporters MsbA and EmrE, bacterial homologs to the ABC transporter family [18, 19]. Advances in automation are having a major impact on several aspects of membrane protein crystallography, particularly for screening of crystals. 96 and 384- well format crystallization trays now allow researchers to rapidly set up thousands of crystallization trials using a variety of robotic liquid-handling platforms such as the Mosquito (TTP Labtech, UK) or Hydra II (Matrix Tech. Corp., NH), but these need to be complemented with crystal imaging technology. The limit in the number of trials is often determined not by the chemical matrices that can be thought up for crystallization, but rather by restrictions on the amount of purified protein that can be produced. Since it is likely that many tens or hundreds of crystals will need to be produced and screened, obtaining sufficient protein preparations from the outset is critical. 9.7 Generation of First Diffraction Datasets 231

Robots are also finding good use in the screening of crystals at the synchrotron. Anyone who has screened at the synchrotron will tell you that most of the time spent screening is taken up by mounting crystals and closing up the experimental hutch. With current state-of-the-art beam lines, the time taken to obtain the test diffraction image can be as short as 10 s. In contrast, getting the hutch open, mounting and centering the crystal and closing up the hutch takes many minutes. Robots such as the ACTOR system (Rigaku/MSC, TX) mounted at the beam line are equipped with cryo-pucks that hold dozens of crystals and are now in use at several synchrotrons. Due to the efficiency of not having to manipulate the crystals manually, one can now screen crystals at the beam line remotely. A Dewar flask loaded with pucks and dozens of crystals is shipped to the synchrotron facil- ity and a technician can load up the robot. Automatic mounting and centering of the crystals at a preset exposure time will facilitate the screening of hundreds of crystals within a few hours. Once screened, each crystal is safely recovered and re- turned to the Dewar by the robot for future data collection. This is without doubt the way of the future since it massively increases the “open shutter” time at the beam lines, enhancing the efficiency of a seriously taxed resource.

9.8 Selenomethionine Phasing of Membrane Proteins

Congratulations! You’ve just successfully crystallized a membrane protein and got a dataset. Now all you have to do is obtain your phases and start model building. Membrane proteins involved in respiration and light harvesting have metal-containing cofactors that can be useful for phase determination. However, many membrane proteins, channels, receptors and transporters, typically do not. Although traditional multiple isomorphous replacement methods for obtaining phase information on membrane proteins and channels have been used, an increasingly useful technique widely used in soluble proteins, and just as valuable for membrane proteins, is selenomethionine multiple anomalous diffraction phasing (SeMet MAD). In spite of the often modest resolution of membrane proteins, this technique has seen spectacular successes and should not be discounted as a useful route for phase determination. A possible complication to using this technique is the large size of many membrane proteins. The larger the protein the higher the likelihood that a large number of methionine residues, and thus se- lenomethionines, will be present. Successful phasing using selenomethionine and MAD phasing have been reported for membrane proteins with up to 42 methionines per asymmetric unit, all at a resolution of 4 Å, underscoring the utility of this approach [17]. Incorporation of selenomethionine has been shown in E. coli (reviewed in Ref. [64]), in yeast [65], baculovirus [66], and mammalian tissue culture (BioXtal, Swit- zerland). In the Rees Lab at Caltech, bacterial expression of membrane proteins with incorporation of selenomethionine started with an M9 glucose minimal media with subsequent media augmentation. A modified M9 medium, containing 9.9 MAD Phasing and Edge Scanning 233

50 mg L–1 of selenomethionine and the other 19 amino acids at 40 mgml–1 (or 80 mg, for d,l-amino acids) was sufficient for nearly complete incorporation of selenomethionine. As with any change in growth medium, optimization of induction reagents, induction time and growth temperature should be undertaken. The addition of naturally derived media components, such as tryptone and yeast extract, should be avoided since these reagents and others contain methionine, and will likely reduce the level of selenomethionine incorporation. As one would expect, the yield of recombinant protein in minimal media is considerably lower than that for rich media. Sufficient cell mass for purification and crystallization can require scaling up, in our case up to a 60 L fermentor. Similarly, for yeast expression, both extensive strain selection as well as media screening produced high incorporation rates [65]. In some cases, cells can be grown in methionine-containing media with subsequent transfer to selenomethionine-containing media, as was done for the first selenomethionine crystal structure that utilized insect cells and baculovirus infection. In this latter case, cells were grown and infected in methionine-rich media, then grown in methionine-depleted media and finally grown in selenomethionine-containing media. This approach is also readily applied to bacterial expression if cell growth in selenomethionine minimal media is insufficient. Current high throughput efforts in structural genomics have applied selenomethionine labeling right from the beginning. In fact, as with soluble proteins, some have suggested that it is advantageous to simply move immediately to expression of membrane proteins in selenomethionine incorporating systems. While this surely would speed the process to structure solution, it seems highly unlikely that the limitations imposed in media used for selenomethionine bioin- corporation would lead to the successful membrane protein expression. A far more fruitful approach would seem to rely on a combination of multiple targets and expression systems, as mentioned throughout this chapter and a return to selenomethionine once diffraction quality crystals have been obtained.

9.9 MAD Phasing and Edge Scanning

A fluorescent scan of the crystal should be obtained before data are collected. Mul- tiple wavelength anomalous diffraction (MAD) beam lines are equipped with -ray fluorescence detectors and obtaining the values of f ' and f '' for the anomalous scattering atom(s) aids in finding the precise wavelength you should use for collecting the peak and inflection datasets. Although it is possible to look up these values, the dataset will be better for choosing the wavelengths from the scan. The Blu-Ice program (developed at SSRL) can perform an edge scan and calculate these values for the user [67]. A minimum of two wavelengths are required to do MAD phasing (hence “multiple” wavelength anomalous diffraction); however, in practice, the peak wavelength, inflection, and a high-energy remote wavelength are usually collected. The peak wavelength is, as it sounds, the peak of the f '' plot. 234 9 Approaches for Ion Channel Structural Studies

A second wavelength, at the inflection point, is the wavelength at which the f ' value inflects at a minimum (very near the half maximal point on the f '' plot. The final wavelength is a high energy remote, usually collected at a lower wavelength distant from the edge of X-ray absorption of the anomalous scattering atom. Fi- nally, a critical factor is to obtain highly redundant data. Anomalous scattering only produces relatively small differences between pairs of reflections (Bijvoet pairs). By obtaining data redundancy of 4-fold (each reflection measured 4 times), 5-fold or even higher, each of these pairs is measured multiple times, so the small differences in the intensity of each Bijvoet pair are statistically significant. Even high symmetry space groups, where a small angular wedge of data can produce a complete dataset, benefit, in terms of phasing, by collecting as much data as possible. Of course, the maximum would be to collect 3608 of data at each of the wavelengths chosen. If radiation damage is not a problem, as determined by integrating and scaling the data and obtaining the Rmerge of each dataset, then this approach should be used. Since the anomalous signal from selenium is small compared to heavy atoms used in multiple isomorphous replacement, such as gold, mercury and platinum, the data from a selenomethionine MAD experiment need special treatment. Searching for the anomalous scattering atoms, selenium in this case, which have relatively small differences in the intensities of Bijvoet pairs is enhanced by allowing the search programs (Solve, for example) to use local scaling. Local scaling of the data, scaling using a reciprocal space sphere of neighboring reflections, allows the best signal to noise ratio when obtaining these slight differences in intensities. In practice, what this means is stronger anomalous signals accurately measured to the highest resolution for that dataset, which will increase the chances of solving the crystal structure.

9.10 Negative B- factor Application (Structure Factor Sharpening)

Anisotropic decay in the intensity of high resolution reflections has been observed frequently for membrane proteins. Since many membrane protein crystals dif- fract to only modest resolution (3–4 Å) obtaining the most from these weak reflections in the highest resolution shell is crucial. A successful approach that has been used with great success is structure factor sharpening (the application of a

negative B-factor to the observed structure factor, Fobs). The CCP4 program can be run to apply an overall B-factor and scale to the data that will increase the Fobs of the high-resolution reflection in a disproportional manner relative to the low-resolution reflections [68]. However, a disadvantage is that the application of negative B-factors increases the noise in electron density maps. Nevertheless, this technique is most valuable if noncrystallographic symmetry is present, since this excess noise is averaged out leaving maps that can be of excellent quality, thus producing a better model. References 235

9.11 Conclusions

In spite of the somewhat daunting task of crystallizing a membrane protein such as an ion channel, there are emerging tools and techniques that will, without doubt, make structures of these intriguing proteins more common. Ion channels and membrane proteins in general have such tremendous therapeutic potential to be unlocked by the availability of high-resolution structures; one can expect the sheer number and speed of their solutions to only increase in the coming years.

References

1 J. Deisenhofer, O. Epp, K. Miki, 9 M. Quick, E. M. Wright, Employing R. Huber, H. Michel, X-ray structure Escherichia coli to functionally express, analysis of a membrane protein com- purify, and characterize a human trans- plex. Electron density map at 3 A reso- porter, Proc. Natl. Acad. Sci.U S A 2002, lution and a model of the chromo- 99, 8597–8601. phores of the photosynthetic reaction 10 D. N. Parcej, L. Eckhardt-Strelau, center from Rhodopseudomonas viri- Structural characterisation of neuronal dis, J. Mol. Biol. 1984, 180, 385–398. voltage-sensitive K+ channels heterolo- 2 S. B. Long, E. B. Campbell, R. Mackin- gously expressed in Pichia pastoris, J. non, Crystal Structure of a Mammalian Mol. Biol. 2003, 333, 103–116. Voltage-Dependent Shaker Family K+ 11 M. A. Lesney, HTS: Seeking Signals Channel, Science 2005, 309, 897–903. Through the Noise, Modern Drug Discov. 3 J. C. Kendrew, G. Bodo, H. M. Dintzis, 2002, 5. R. G. Parrish, H. Wyckoff, D. C. Phil- 12 C. Tanford (ed.), The Hydrophobic Ef- lips. A three-dimensional model of the fect: Formation of Micelles and Biological myoglobin molecule obtained by x-ray Membranes, John Wiley & Sons, New analysis, Nature 1958, 181, 662–666. York, 1980. 4 S. H. White, The progress of mem- 13 T. P. Roosild, J. Greenwald, M. Vega, brane protein structure determination, S. Castronovo, R. Riek, S. Choe, Protein Sci. 2004, 13, 1948–1949. NMR structure of Mistic, a membrane- 5 M. Wortman, The Re-Emergence of integrating protein for membrane pro- Ion Channel Drug Discovery, in Start- tein expression, Science 2005, 307, Up: Windhover’s Review of Emerging 1317–1321. Medical Ventures, 2003, p. 9. 14 J. F.White, L. B. Trinh, J. Shiloach, 6 J. C. Venter, et al., The sequence of the R. Grisshammer, Automated large- human genome, Science 2001, 291, scale purification of a G protein-coupled 1304–1351. receptor for neurotensin, FEBS Lett. 7 R. Grisshammer, J. Tucker, Quantita- 2004, 564, 289–293. tive evaluation of neurotensin receptor 15 G. Chang, R. H. Spencer, A. T. Lee, purification by immobilized metal affi- M. T. Barclay, D. C. Rees, Structure of nity chromatography, Protein Expr. the MscL homolog from Mycobacter- Purif. 1997, 11, 53–60. ium tuberculosis: a gated mechanosen- 8 G. Fiermonte,V. Dolce, F. Palmieri, sitive ion channel, Science 1998, 282, Expression in Escherichia coli, func- 2220–2226. tional characterization, and tissue distri- 16 K. P. Locher, A. T. Lee, D. C. Rees, The bution of isoforms A and B of the phos- E. coli BtuCD structure: a framework phate carrier from bovine mitochon- for ABC transporter architecture and dria, J. Biol. Chem. 1998, 273, 22782– mechanism, Science 2002, 296, 1091– 22787. 1098. 236 9 Approaches for Ion Channel Structural Studies

17 R. B. Bass, P. Strop, M. Barclay, D. C. 27 Serysheva II, M. Schatz, M. van Rees, Crystal structure of Escherichia Heel,W. Chiu, S. L. Hamilton, Struc- coli MscS, a voltage-modulated and me- ture of the skeletal muscle calcium re- chanosensitive channel, Science 2002, lease channel activated with Ca2+ and 298, 1582–1587. AMP-PCP. Biophys. J. 1999, 77, 1936– 18 G. Chang, C. B. Roth, Structure of 1944. MsbA from E. coli: a homolog of the 28 A. Miyazawa,Y. Fujiyoshi, N. Unwin, multidrug resistance ATP binding cas- Structure and gating mechanism of the sette (ABC) transporters, Science 2001, acetylcholine receptor pore, Nature 293, 1793–1800. 2003, 423, 949–955. 19 C. Ma, G. Chang, Crystallography of 29 C. Sato,Y. Ueno, K. Asai, K. Taka- the integral membrane protein EmrE hashi, M. Sato, A. Engel,Y. Fuji- from Escherichia coli, Acta Crystallogr. yoshi, The voltage-sensitive sodium D 2004, 60, 2399–2402. channel is a bell-shaped molecule with 20 K. Mio, T. Ogura,Y. Hara,Y. Mori, several cavities. Nature 2001, 409, 1047– C. Sato, The non-selective cation- 1051. permeable channel TRPC3 is a tetrahe- 30 Serysheva II, S. J. Ludtke, M. R. dron with a cap on the large cytoplas- Baker,W. Chiu, S. L. Hamilton, Struc- mic end, Biochem Biophys Res Commun ture of the voltage-gated L-type Ca2+ 2005, 333, 768–777. channel by electron cryomicroscopy, 21 M. Wolf, A. Eberhart, H. Gloss- Proc. Natl. Acad. Sci.U S A 2002, 99, mann, J. Striessnig, N. Grigorieff, 10370– 10375. Visualization of the domain structure of 31 M. C. Wang, G. Velarde, R. C. Ford, an L-type Ca2+ channel using electron N. S. Berrow, A. C. Dolphin, A. Kit- cryo-microscopy, J. Mol. Biol. 2003, 332, mitto, 3D structure of the skeletal mus- 171–182. cle dihydropyridine receptor, J. Mol. 22 P. C. da Fonseca, S. A. Morris, E. P. Biol. 2002, 323, 85–98. Nerou, C.W. Taylor, E. P. Morris, Do- 32 C. C. Chadwick, A. Saito, S. Flei- main organization of the type 1 inositol scher, Isolation and characterization of 1,4,5-trisphosphate receptor as revealed the inositol trisphosphate receptor from by single-particle analysis, Proc. Natl. smooth muscle, Proc. Natl. Acad. Sci.U Acad. Sci.U S A 2003, 100, 3936–3941. SA1990, 87, 2132–2136. 23 Q. X. Jiang, E. C. Thrower, D. W. 33 M. F. Rosenberg, A. B. Kamis, L. A. Chester, B. E. Ehrlich, F. J. Sig- Aleksandrov, R. C. Ford, J. R. Rior- worth, Three-dimensional structure of dan, Purification and crystallization of the type 1 inositol 1,4,5-trisphosphate the cystic fibrosis transmembrane con- receptor at 24 A resolution, EMBO J. ductance regulator (CFTR), J. Biol. 2002, 21, 3575–3581. Chem. 2004, 279, 39051–39057. 24 Serysheva II, D. J. Bare, S. J. Ludtke, 34 A. K. Mitra, A. N. van Hoek, M. C. C. S. Kettlun,W. Chiu, G. A. Mig- Wiener, A. S. Verkman, M. Yeager, nery, Structure of the type 1 inositol The CHIP28 water channel visualized 1,4,5-trisphosphate receptor revealed by in ice by electron crystallography, Nat. electron cryomicroscopy, J. Biol. Chem. Struct. Biol. 1995, 2, 726–729. 2003, 278, 21319–21322. 35 G. Ren,V. S. Reddy, A. Cheng, P. Mel- 25 Serysheva II, S. L. Hamilton, nyk, A. K. Mitra,Visualization of a W. Chiu, S. J. Ludtke, Structure of water-selective pore by electron crystal- Ca2+ release channel at 14 A resolution, lography in vitreous ice, Proc. Natl. J. Mol. Biol. 2005, 345, 427–431. Acad. Sci.U S A 2001, 98, 1398–1403. 26 E. V. Orlova, Serysheva II, M. van 36 L. Santacruz-Toloza,Y. Huang, S. A. Heel, S. L. Hamilton,W. Chiu, Two John, D. M. Papazian, Glycosylation of structural configurations of the skeletal shaker potassium channel protein in in- muscle calcium release channel. Nat. sect cell culture and in Xenopus oo- Struct. Biol. 1996, 3, 547–552. cytes, Biochemistry 1994, 33, 5607–5613. References 237

37 A. N. van Hoek, M. C. Wiener, J. M. channels in insect cells using baculo- Verbavatz, D. Brown, P. H. Lipniu- virus, FEBS Lett. 1998, 429, 390–394. nas, R. R. Townsend, A. S. Verkman, 46 U. S. Rao, A. Mehdi, R. E. Steimle, Ex- Purification and structure-function ana- pression of amiloride-sensitive sodium lysis of native, PNGase F-treated, and channel: a strategy for the coexpression endo-beta-galactosidase-treated CHIP28 of multimeric membrane protein in Sf9 water channels, Biochemistry 1995, 34, insect cells, Anal. Biochem. 2000, 286, 2212–2219. 206–213. 38 R. H. Spencer, et al., Purification, vi- 47 M. Li, N. Unwin, K. A. Stauffer,Y. N. sualization, and biophysical characteri- Jan, L. Y. Jan, Images of purified Sha- zation of Kv1.3 tetramers, J. Biol. Chem. ker potassium channels, Curr. Biol. 1997, 272, 2389–2395. 1994, 4, 110–115. 39 E. Bennett, M. S. Urcan, S. S. Tinkle, 48 D. A. Doyle, et al., The structure of the A. G. Koszowski, S. R. Levinson, Con- potassium channel: molecular basis of tribution of sialic acid to the voltage de- K+ conduction and selectivity, Science pendence of sodium channel gating. A 1998, 280, 69–77. possible electrostatic mechanism, J. 49 D. R. Smyth, M. K. Mrozkiewicz,W. J. Gen. Physiol. 1997, 109, 327–343. McGrath, P. Listwan, B. Kobe, Crystal 40 N. B. Cronin, A. O’Reilly, H. Duclo- structures of fusion proteins with large- hier, B. A. Wallace, Effects of deglyco- affinity tags, Protein Sci. 2003, 12, 1313– sylation of sodium channels on their 1322. structure and function, Biochemistry 50 J. Zhuang, G. G. Prive, G. E. Werner, 2005, 44, 441–449. P. Ringler, H. R. Kaback, A. Engel, 41 K. Klaiber, N. Williams, T. M. Ro- Two-dimensional crystallization of Es- berts, D. M. Papazian, L. Y. Jan, cherichia coli lactose permease, J. C. Miller, Functional expression of Struct. Biol. 1999, 125, 63–75. Shaker K+ channels in a baculovirus-in- 51 B. Byrne, J. Abramson, M. Jansson, fected insect cell line, Neuron 1990, 5, E. Holmgren, S. Iwata, Fusion protein 221–226. approach to improve the crystal quality 42 R. A. Gubitosi-Klug, D. J. Mancuso, of cytochrome bo(3) ubiquinol oxidase R. W. Gross, The human Kv1.1 channel from Escherichia coli, Biochim. Biophys. is palmitoylated, modulating voltage Acta 2000, 1459, 449–455. sensing: Identification of a palmitoyla- 52 J. Abramson, I. Smirnova,V. Kasho, tion consensus sequence, Proc. Natl. G. Verner, H. R. Kaback, S. Iwata, Acad. Sci.U S A 2005, 102, 5964–5968. Structure and mechanism of the lactose 43 C. E. Bear, C. H. Li, N. Kartner, R. J. permease of Escherichia coli, Science Bridges, T. J. Jensen, M. Ramjee- 2003, 301, 610–615. singh, J. R. Riordan, Purification and 53 J. Abramson, et al., The structure of the functional reconstitution of the cystic ubiquinol oxidase from Escherichia coli fibrosis transmembrane conductance and its ubiquinone binding site, Nat. regulator (CFTR), Cell 1992, 68, 809– Struct. Biol. 2000, 7, 910–917. 818. 54 M. H. Stowell, in UCLA Crystallization 44 M. Cascio, N. E. Schoppa, R. L. Grod- Workshop 1993. zicki, F. J. Sigworth, R.O. Fox, Func- 55 Z. Liu, et al., Crystal structure of spi- tional expression and purification of a nach major light-harvesting complex at homomeric human alpha 1 glycine re- 2.72 A resolution, Nature 2004, 428, ceptor in baculovirus-infected insect 287–292. cells, J. Biol. Chem. 1993, 268, 22135– 56 S. Iwata, C. Ostermeier, B. Ludwig, 22142. H. Michel, Structure at 2.8 A resolu- 45 M. V. Mikhailov, P. Proks, F. M. Ash- tion of cytochrome c oxidase from Para- croft, S. J. Ashcroft, Expression of coccus denitrificans, Nature 1995, 376, functionally active ATP-sensitive K- 660–669. 238 9 Approaches for Ion Channel Structural Studies

57 C. Hunte, J. Koepke, C. Lange,T. Ross- 67 T. M. McPhillips, et al., Blu-Ice and manith, H. Michel, Structure at 2.3 A the Distributed Control System: soft- resolution of the cytochrome bc(1) com- ware for data acquisition and instru- plex from the yeast Saccharomyces cere- ment control at macromolecular crystal- visiae co-crystallized with an antibody lography beamlines, J. Synchrotron Ra- Fv fragment, Struct. Fold. Des. 2000, 8, diat. 2002, 9, 401–406. 669–684. 68 Collaborative Computational Project, N. 58 Y. Jiang, A. Lee, J. Chen,V. Ruta, The CCP4 suite: programs for protein M. Cadene, B. T. Chait, R. MacKin- crystallography, Acta Crystallogr. D 1994, non, X-ray structure of a voltage-depen- 50, 760–763. dent K+ channel, Nature 2003, 423, 33– 69 Y. Zhou, J. H. Morais-Cabral, 41. A. Kaufman, R. MacKinnon, Chemis- 59 G. Kleymann, C. Ostermeier, B. Lud- try of ion coordination and hydration re- wig, A. Skerra, H. Michel, Engi- vealed by a K+ channel-Fab complex at neered Fv fragments as a tool for the 2.0 A resolution, Nature 2001, 414, 43– one-step purification of integral multi- 48. subunit membrane protein complexes, 70 R. Dutzler, E. B. Campbell, Biotechnology (NY) 1995, 13, 155–160. M. Cadene, B. T. Chait, R. MacKin- 60 L. G. Cuello, D. M. Cortes, E. Perozo, non, X-ray structure of a ClC chloride Molecular architecture of the KvAP vol- channel at 3.0 A reveals the molecular tage-dependent K+ channel in a lipid basis of anion selectivity, Nature 2002, bilayer, Science 2004, 306, 491–495. 415, 287–294. 61 F. Bezanilla, The voltage-sensor struc- 71 Y. Jiang, A. Lee, J. Chen, M. Cadene, ture in a voltage-gated channel, Trends B. T. Chait, R. MacKinnon, Crystal Biochem. Sci. 2005, 30, 166–168. structure and mechanism of a calcium- 62 M. Laine, D. M. Papazian, B. Roux, gated potassium channel, Nature 2002, Critical assessment of a proposed model 417, 515–522. of Shaker, FEBS Lett. 2004, 564, 257– 72 R. Dutzler, E. B. Campbell, 263. R. MacKinnon, Gating the selectivity 63 D. Rothlisberger, K. M. Pos, filter in ClC chloride channels, Science A. Pluckthun, An antibody library for 2003, 300, 108–112. stabilizing and crystallizing membrane 73 A. Kuo, et al., Crystal structure of the proteins – selecting binders to the potassium channel KirBac1.1 in the citrate carrier CitS, FEBS Lett. 2004, closed state, Science 2003, 300, 1922– 564, 340–348. 1926. 64 W. A. Hendrickson, J. R. Horton, 74 F. G. Boess, R. Beroukhim, I. L. Mar- D. M. LeMaster, Selenomethionyl tin, Ultrastructure of the 5-hydroxy- proteins produced for analysis by multi- tryptamine3 receptor, J Neurochem wavelength anomalous diffraction 1995, 64, 1401– 1405. (MAD): a vehicle for direct determina- 75 H. L. Li, et al., Two-dimensional crystal- tion of three-dimensional structure, lization and projection structure of EMBO J. 1990, 9, 1665–1672. KcsA potassium channel, J. Mol. Biol. 65 D. A. Bushnell, P. Cramer, R. D. 1998, 282, 211–216. Kornberg, Selenomethionine incor- 76 V. M. Unger, N. M. Kumar, N. B. poration in Saccharomyces cerevisiae Gilula, M. Yeager, Three-dimensional RNA polymerase II, Structure (Camb.) structure of a recombinant gap junction 2001, 9, R11–R14. membrane channel, Science 1999, 283, 66 J. J. Bellizzi, J. Widom, C. W. Kemp, 1176–1180. J. Clardy, Producing selenomethio- 77 J. A. Mindell, M. Maduke, C. Miller, nine-labeled proteins with a baculovirus N. Grigorieff, Projection structure of expression vector system, Struct. Fold. a ClC-type chloride channel at 6.5 A re- Des. 1999, 7, R263–R267. solution, Nature 2001, 409, 219–223. References 239

78 O. Sokolova, L. Kolmakova-Par- 81 Q. X. Jiang, D. N. Wang, R. MacKin- tensky, N. Grigorieff, Three-dimen- non, Electron microscopic analysis of sional structure of a voltage-gated potas- KvAP voltage-dependent K+ channels sium channel at 2.5 nm resolution, in an open conformation, Nature 2004, Structure (Camb.) 2001, 9, 215–220. 430, 806–810. 79 C. Sato, et al. Inositol 1,4,5-trispho- 82 T. Nakagawa,Y. Cheng, E. Ramm, sphate receptor contains multiple cav- M. Sheng, T. Walz, Structure and dif- ities and L-shaped ligand-binding do- ferent conformational states of native mains, J. Mol. Biol. 2004, 336, 155–164. AMPA receptor complexes, Nature 80 L. A. Kim, J. Furst, D. Gutierrez, 2005, 433, 545–549. M. H. Butler, S. Xu, S. A. Goldstein, 83 M. Samso, T. Wagenknecht, P. D. N. Grigorieff, Three-dimensional Allen. Internal structure and visualiza- structure of I(to); Kv4.2-KChIP2 ion tion of transmembrane domains of the channels by electron microscopy at 21 RyR1 calcium release channel by cryo- Angstrom resolution, Neuron 2004, 41, EM, Nat. Struct. Mol. Biol. 2005, 12, 513–519. 539–544. 241

10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels Daniele Bemporad, Alessandro Grottesi, Shozeb Haider, Zara A. Sands, and Mark S.P. Sansom

10.1 Introduction

The past decade has seen substantial advances in the structural biology of ion channels and of membrane proteins in general (see Refs. [1, 2] and Chapter 9). This has been primarily from X-ray diffraction methods, but electron microscopy (both from two-dimensional crystals [3] and from single particles [4]) is increasingly important, especially for defining the conformations of ion channels within a membrane environment. However, determining the structure of a channel protein is not an end in itself. Rather, one wishes to use structures to understand channel physiology and pharmacology. Understanding channel mechanisms at atomic resolution is both an intellectual challenge, and also an important step in the design of novel channel-perturbing ligands. The aim of this chapter is to describe how computational methods can be used to aid exploration of the relationship between ion channel structure and function. To achieve this we will focus on a particular class of ion channels, namely potassium channels [5]. K+ channels are of importance from a physiological and pharmacological perspective, for example in the central nervous system, and in the regulation of cardiac activity. They are also the best understood class of channels in terms of X-ray structures, and of experimental and computational studies of channel permeation and gating mechanisms. It is useful to review the structural biology of K+ channels. All of the structures that have been determined by X-ray diffraction are of bacterial homologues of mammalian K+ channels (Fig. 10.1). The structures include: (i) KcsA (PDB code 1K4C [6, 7]), a ‘minimalist’ K+ channel activated by low intracellular pH; (ii) Kir- Bac1.1 and 3.1 (PDB codes 1P7B and 1XL4 [8, 9]), homologs of mammalian inward rectifier channels; (iii) KvAP (PDB code 1ORQ [10]), a voltage activated K+ channel; and (iv) MthK (PDB code 1LNQ [11]), a K+ channel activated by Ca2+ binding to an intracellular domain. Each of these structures shares a common architecture for the pore domain of K+ channels. K+ channels are composed of four subunits arranged symmetrically around a central pore. The pore-forming domain

Fig. 10.1 X-ray structures of K+ channels. In each case, for clarity, only two of the four subunits are present. Channels with the pore domain in a closed conformation (A) are represented by KcsA and KirBac3.1; channels with the pore domain in an open conformation (B) are represented by KvAP and MthK (for which only the transmembrane domain is shown). The dotted horizontal lines represent the approximate extent of the lipid bilayer.

is formed of a M1-P-F-M2 motif, where M1 and M2 are transmembrane (TM) a-helices, and the short P-helix and extended filter (F) region form a re-entrant loop between the two TM helices. (Note that in Kv channels the S5 and S6 helices are the equivalents of M1 and M2). The filter contains a sequence motif (TVGYG) that is characteristic of K+ channels and is the structural element responsible for the selective permeability to K+ ions. The four subunits of the pore domain thus form an eight TM helix bundle, with the M2 (or S6) helices lining the central pore in an inverted truncated conical geometry, whilst the P loop is inserted at the extracellular mouth of the M2 bundle where it forms the selectivity filter. The M2 (or S6) helices cross at the intracellular mouth of the pore, which corresponds to the activation gate of a K+ channel. In KcsA and KirBac the channel is crystallised in a closed state, and the M2 helices pack together to form a narrow hydrophobic region in the pore which corresponds to the (closed) gate. In MthK the protein was crystallised in the presence of Ca2+ and hence the channel was captured in an open state in which the M2 helices are kinked so as to open the intracellular gate. In KvAP the pore domain also appears to be in an open state with the S6 helices kinked.

10.2 Computational Methods

Given advances in the structural biology of K+ channels, what roles can computational methods play? The main uses of computational methods can be divided into two areas: (i) modeling; and (ii) simulation. Modeling is important because the available structures are all of bacterial K+ channels (as these are more readily 10.2 Computational Methods 243 expressed and crystallised), whereas from a physiological and pharmacological perspective we wish to understand structure/function relationships in eukaryotic (e.g. human) K+ channels. Simulations are important because an X-ray structure provides only a static snapshot of a protein, whereas a more dynamic description is needed in order to understand the functional behavior of channels. More explicitly, an X-ray structure is a spatial and temporal average structure of a channel protein in a crystal environment (i. e. with detergents and/or antibody fragments) at a temperature of ~100 K. We wish to characterise the dynamic behavior of a channel protein in a membrane (i.e. lipid bilayer) environment at a more physiologically relevant temperature of ~300 K. Modeling of mammalian ion channels on the basis of the structure of bacterial homologs is relatively well developed as a method [12]. Starting from a sequence alignment, homology modeling (Table 10.1) will yield a predicted structure for a mammalian channel based on a bacterial channel template. The quality of the resultant model depends to a large extent on that of the sequence alignment. Thus it is often necessary to adjust manually the initial sequence alignment to take account of: (i) experimental data identifying functionally important residues; (ii) experimental topology data; and (iii) preferential TM helix locations within the channel sequence. The latter may be predicted with ~80% accuracy using a number of

Table 10.1 Computational methods.

Method Description Code/Websites homology generation of a mamma- Modeller [107, 108] http://salilab.org/modeller/ modeling lian channel model using SwissModel [109] http://swissmodel.expasy.org/ a bacterial homolog structure as a template. molecular atomistic simulations of Gromacs [110] http://www.gromacs.org/ dynamics protein motions (on a Charmm [111] http://www.charmm.org/ simulations timescale of 10–50 ns). NAMD [112] http://www.ks.uiuc.edu/Research/namd/ network coarse-grained descrip- GNM & ANM [51, 52] models tions of large scale pro- http://ribosome.bb.iastate.edu/software.html tein motions Poisson– calculation of the electro- APBS [113] http://agave.wustl.edu/apbs/ Boltzmann static field around and UHBD [114] http://adrik.bchs.uh.edu/uhbd/ electrostatics on the surface of a chan- GRASP [115] http://honiglab.cpmc.columbia.edu/ nel protein. ligand prediction of ligand bind- Autodock [116] docking ing sites and interactions http://www.scripps.edu/mb/olson/doc/autodock/ on the surface of a channel protein.

This list of methods is representative of some readily available codes that have been used in channel modeling and simulation studies. 244 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

TM helix prediction algorithms. However, rather than relying on a single algo- rithm, it is better to run multiple methods in order to generate a consensus prediction [13]. The major limitation of homology modeling and related approaches is that large (>15 residues) insertions (often corresponding to surface loops) in a mammalian channel sequence relative to that of the bacterial homolog are difficult to model with any certainty. In such cases it is preferable to omit the inserted region from the model. If the model with such omissions is to be used in simulations, a simple distance restraint can be used to mimic the effect of the missing region on the remaining structure. Molecular dynamics (MD) allows one to perform atomistic simulations of membrane protein motions on a timescale of 10 to 50 ns, through the application of an appropriate molecular mechanics forcefield (see Table 10.1). Such methods have been applied to a range of membrane proteins [14] including ion channels [15– 19], aquaporins [20–23], and bacterial outer membrane proteins [24–26]. The out- come of such a simulation is a trajectory, i.e. a description of the position and energetics of a membrane protein and its environment as a function of time, which can subsequently be analysed to yield insights into the relationship between the dynamic behavior and its biological function. A typical simulation system is illustrated in Fig. 10.2B. It consists of a K+ channel (here just the transmembrane domain) embedded in a phosphatidylcholine (PC) bilayer, with water molecules and ions on either side of the membrane. Typi- cally, the size of such a simulation system is between 50 000 and 100 000 atoms. For such a system size, using current simulation codes (see Table 10.1) and relatively modest computational facilities (e. g. a small linux cluster) simulation times of the order of 10 ns can be achieved within a few days to weeks. This enables one to run multiple simulations to compare e.g. the dependence of the simulation behavior on the initial configuration of ions within the filter of a channel [27] or the behavior of various channel mutants [28]. More detailed calculations, such as the free energy landscape for ions within the filter of a channel [29], take a little longer but are still achievable with the same resources and simulation codes. Over the past few years, MD simulations have been widely exploited as a method of studying K+ channels [12, 17, 30–37]. It is important to note a major limitation of MD simulations, namely that of timescale. Even with substantial computational resources, for a channel simulation it is challenging to achieve a simulation time in excess of 100 ns (0.1 ms). To put this in context, the mean time for a single ion to pass through a channel is ~10 ns, and protein conformational changes underlying channel gating occur on a timescale of ~1 ms (106 ns). Thus, to use computational methods to obtain a complete description of single channel physiological function on the basis of molecular structure will require a hierarchy of simulation methods (see Fig. 10.2). As yet this hierarchy of methods is incompletely developed and integrated. Although the focus in the current article is on MD simulations, it is useful to consider other methods and their applications, real or potential, to ion channels. With respect to ion permeation a longer timescale can be addressed by the use of Brownian dynamics (BD) simulations. These are based on solution of the Pois- 10.2 Computational Methods 245

Fig. 10.2 Molecular modeling and simulations. The overall aim (A) is to relate static X-ray structures of channels to the dynamic single channel physiological properties ((C) illustrated with single channel recordings from Kir6.2 channels, courtesy of F.M. Ashcroft [62]). Multi- level simulations are required to address a wide range of timescales (from ns to ms). In this chapter the focus is on atomistic modeling and simulations (B). son–Boltzmann equation to describe the electrostatic field around a channel protein, followed by simulations in which the protein is treated as a static entity, the water as a continuum solvent, and ion diffusion in the electrostatic field created by the protein and membrane is simulated. This method has been used with some success in simulation of ion flow through high conductance, low ion selectivity channels such as bacterial outer membrane porins [38]. A detailed comparison of the results of MD and BD simulations has been performed for this system [39]. The application of these methods to lower conductance, high selectivity channels such as K+ channels is more challenging, and has been explored in detail by Chung and coworkers [40–43]. An even more challenging limitation of MD timescales occurs when trying to address the conformational changes underlying channel gating. One approach is to use ‘steered’ MD simulations in which a conventional MD simulation is modified by the inclusion of a biasing force that drives a conformational change such as channel gating [44]. Another MD-based approach is to identify the dominant low frequency protein motions from an MD simulation by way of principal components analysis (PCA; see Box 1). This method is also referred to as ‘essential dynamics’ [45–47], and has the advantage that it allows direct extrapolation from MD simulations to predict possible motions on longer timescales. However, it is well recognised that ~10 ns MD simulations incompletely capture the motions of membrane proteins [48], and thus biologically important long timescale motions may not be observed. There is consequently a considerable need to develop and apply coarse- grained (CG) models of ion channel motions. CG models of proteins have been explored in general for simulations of protein conformational changes and folding [49]. One promising class of CG methods is network models (Gaussian network models, GNM, and anisotopic network models, ANM) which represent each amino acid in a protein as a single ‘particle‘, coupled to other particles via a harmonic potential. Such methods have been shown to reproduce the temperature factors of proteins (a crystallographic measure of thermal fluctuations) [50] and have been used to explore the large scale dynamic behavior of various proteins [51, 52], includ- 246 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

Box 1 Principal component analysis.

Principal components analysis (PCA) of positional fluctuations is a valuable tool to extract the dominant low frequency protein motions from a molecular dynamics (MD) simulation. This enables one to extrapolate to the dynamics of the protein on a more extended timescale. The method is based on the definition of collective coordinates, C, that represent the major contribution to atomic fluctuations. These are obtained by diagonalizing the positional fluctuation covariance matrix A, whose elements are defined as:

aij : {(xi – [xi])(xj – [xj])}

where […] denotes an average over time. A set of eigenvectors and eigenvalues are obtained solving the following equation:

AC = Cf

where z = diag(zn) is a diagonal matrix whose nth element corresponds to the eigenvalues zn. Thus C is the eigenvector matrix. Generally 70–80% of the total protein fluctuations are accounted for by the first 10 eigenvectors (that are accordingly referred to as ‘essential space’, [46]). Therefore, biological relevant motion can be captured by filtering the initial MD coordinate trajectory onto one of the essential degrees of freedom and then analyzing the resultant dynamics.

ing ion channel domains (see below, [53]). Further development and application studies are needed to explore fully the application of CG methods to channel gating, and their integration with MD simulations. Another computational approach that is widely used in studies of channels and receptors is computational docking of small molecules (i.e. ligands) to proteins. This is an important and active field of research in computational structural biology, and interested readers are referred to a recent review as a starting point [54]. An overview of a range of theoretical methods as applied to ion channels in general is provided in a special issue of IEEE Transactions on Nanobioscience [55]. In this chapter, we will focus on studies from the authors’ laboratory which are con- cerned with MD and related approaches to two classes of K+ channels, namely: inward rectifiers, and voltage-gated K+ channels.

10.3 Kir Channels

10.3.1 Structures

K+ channels of the inward rectifier (Kir) class have two main physiological roles: they regulate cell excitability by stabilizing the membrane potential close to the [K+] equilibrium potential, and they are involved in K+ transport across membranes [56, 57]. For example, Kir3.1/Kir3.4 channels modulate cardiac electrical 10.3 Kir Channels 247 activity, and Kir6.2 is involved in insulin release from pancreatic b-cells. Kir6.2 is the pore-forming component of the ATP-sensitive potassium (KATP) channel which couples cell metabolism to electrical activity by regulating K+ flux across the membrane. Channel closure is mediated by ATP, which binds to the intracellular domain of Kir6.2. Kir channels have two TM helices per subunit (similar to KcsA and MthK). Kir channels also have a large intracellular (IC) domain, composed of ~50 residues from the N-terminal tail of the protein plus a C-terminal domain of ~200 residues. This domain plays an important functional role via binding cytosolic regulators of

Kir activity, such as ATP and PIP2. Three recent structures, of the isolated IC domain of a mammalian Kir (Kir3.1 = GIRK1 [58]) and of the two intact bacterial Kir homologues (KirBac1.1 [8] and KirBac3.1 [9]), offer a detailed understanding of structure/function relationships in this family of K+ channels (Fig. 10.3).

Fig. 10.3 The bacterial inward rectifier homolog KirBac. (A) Ca trace of KirBac1.1 showing the transmembrane (TM) and intracellular (IC) domains. The dotted horizontal lines represent the approximate extent of the lipid bilayer. (B), (C) TM pore-forming domains of Kir- Bac1.1 and KirBac3.1, showing two of the four subunits as Ca traces, and the pore-lining surface (calculated using HOLE [117]).

10.3.2 Molecular Modeling

Molecular modeling studies have focussed on modeling mammalian Kir structures using bacterial K+ channel structures as templates. Earlier studies [28, 59, 60] focussed on the pore domain of Kir6.2. MD simulations were used to aid in evaluation of homology models of the Kir6.2 pore domain, derived using the KcsA pore domain structure as a template, and in particular to compare the relative con- 248 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

formational stability of two alternative models of Kir6.2 which differed slightly in the sequence alignment of the M2 helices [59]. More recent studies have combined two templates in modeling the structure of Kir6.2 and related mammalian Kir channels. The IC domain of Kir6.2 has been modeled on the X-ray structure of the Kir3.1 C-terminal domain [58]. This has also enabled us to explore the conformational stability of the homology model of the Kir6.2 C-terminal domain on a 10 ns timescale. A homology model of the intact Kir6.2 molecule has been constructed [61] based on the crystal structures of KirBac1.1 and of the IC domain of Kir3.1. KirBac1.1 was used as a template for the TM domains of Kir6.2, whereas the IC domain was modeled on that of Kir3.1. This was because, over the region modeled, the IC domains of Kir3.1 and Kir6.2 exhibit a greater sequence identity (48%) than those of Kir6.2 and KirBac1.1 (27%). Further, the crystal structure of the IC domain of Kir3.1 was determined at a resolution higher (2 Å) than that of KirBac1.1 (3.6 Å). Each of the three segments of the model (TMs, N and C domains) were constructed separately, and then joined together. The spatial orientation of the IC domain, with respect to the TMs, was determined from the location of conserved residues in the IC domain of KirBac1.1. This model of Kir6.2 has been used as the basis of ligand docking and mutational studies (see below).

10.3.3 Simulations

Simulations of Kir channels have focussed on conformational changes and their relationship to channel gating. At a physiological level, Kir channel gating behavior may be (albeit somewhat crudely) divided into ligand gating, which occurs on a slow (multi-ms) timescale, and fast (‘filter’) gating which is seen as flickering transitions between open and closed within bursts of channel activity (see Fig. 10.2) [62]. However, even such fast gating is on a ms timescale. Thus, as discussed above, gating presents a considerable challenge to simulation studies, which cannot yet address ms timescales directly. However, by careful design of simulation ‘experiments’ and by suitable analyses, we may extrapolate from results on a multi-ns timescale and so formulate atomic resolution hypotheses relating to both aspects of channel gating.

10.3.4 Filter Flexibility

A number of simulations of KcsA have indicated that the original model of K+ channel function, which stressed a relatively rigid structure for the selectivity filter [6], requires modification to allow for filter flexibility [17, 27, 29, 63–65]. These simulation results correlate well with crystallographic studies of changes in the conformation of the selectivity filter of KcsA as a function of ion occupancy [7, 66] (Fig. 10.4), with crystallographic studies of conformational changes during block by tetraethyl ammonium [67], and with more indirect structural studies [68]. More 10.3 Kir Channels 249

Fig. 10.4 Distortion of the K+ channel selectivity filter. The X-ray structure of the KcsA filter in the presence of (A) high and (B) low concentrations of K+ ions. Note the distortion to the filter in the presence of low [K+] due to mutual repulsion of the carbonyl oxygen atoms [7]. Distor- tions of the selectivity filter seen in simulations of inward rectifier channels. (C) The KirBac filter from a simulation in the absence of K+ ions [69]. (D) A Kir6.2 model filter for a mutation close to the filter (V127T) that changes open channel permeation and kinetic properties [28].

recently, simulations have revealed the filter of Kir channels to undergo distor- tions comparable to those seen in the low [K+] structure of KcsA. Thus in simulations of KirBac1.1 in which K+ ions were omitted from the filter (i.e. were replaced with water molecules) [69] the filter distorted via a ‘flip’ of one or more of the constituent polypeptide backbones resulting in carbonyl oxygen atoms being directed away from the pore. The close resemblance between the low [K+] X-ray structure and the zero-[K+] MD structure (Fig. 10.4) provides strong evidence for a model of K+ channel permeation in which the occupancy of the filter, by multiple (2 or 3) K+ ions simultaneously, is responsible for induced fit. In such a model there is a coupling of ion binding to a protein conformational change which underlies the high conduction rates of the channel [66]. Interestingly, a similar conformational change has been observed in simulations of models of mutants of Kir6.2 [28]. It had been noted that the single-channel conductance of Kir channels varies significantly between different members of the family. On this basis a mutation (V127T) was made close to the filter of Kir6.2 which was been shown to produce channels with reduced (40% of wild- type) single-channel conductance [62]. Homology modeling (based on a KcsA template) combined with MD simulations was used to explore whether changes in structural dynamics of the filter were induced by the V127T mutations. Relative to the wild-type simulation, the V127T mutant showed significant distortion of the filter, such that ~50% of the simulation time was spent in a distorted ‘filter-closed’ conformation (Fig. 10.4D). While in this conformation, translocation of K+ ions between adjacent binding sites within the filter was blocked. Significantly, the distorted filter conformation resembled that of KcsA crystallized in the presence of a low [K+]. Taken together, these studies suggest that local distortion of the selectivity filter may be a general model for determining the conductance of K+ channels and/or may be related to ‘fast’ gating of Kir channels. 250 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

10.3.5 M2 Helices and Hinge Motion

KirBac1.1 and 3.1 have similar, but nonidentical, pore domain structures (Fig. 10.3B), differing in the extent to which the intracellular gate is closed. Multiple extended MD simulations (each of >20 ns duration) of the isolated TM domain of both KirBac channels in two membrane environments (a PC bilayer and a membrane-mimetic octane slab) have been performed in order to explore possible M2 helix motions underlying Kir channel gating [70]. In these simulations just the TM domain was used, in an attempt to uncouple the M2 helix gate from the IC ‘gate-keeper’ domain, thus enhancing the likelihood of observing conformational changes in the gate on a 20 ns timescale. Analysis of these simulations focussed on the conformational dynamics of the pore-lining M2 helices. Principal components analysis (see Box 1) was used to analyse bending of the M2 helix. Such bending occurs at a molecular hinge formed by a conserved glycine residue (Gly134 in KirBac1.1, Gly120 in KirBac3.1). This glycine has also been suggested to form a channel gating hinge on the basis of comparison of the structures of KcsA (i. e. a closed K+ channel) and MthK (an open K+ channel). A more detailed analysis of the M2 helix bundle dynamics was suggestive of a dimer-of-dimers motion in which opposing pairs of helices moved together. This may be related to a similar pattern of motions observed for the isolated IC domain (see below). The first two eigenvectors describing the motions of M2 in these simulations correspond to helix kink and helix swivel motions. The conformational flexibility of M2 seen in these simulations correlates well with differences in M2 conformation captured in the X-ray structures, as can be seen if one compares closed channels (KcsA and KirBac) in which the helix is undistorted, with open channels (e.g. MthK) in which the M2 helix is kinked (Fig. 10.5). Thus the simulations, albeit on

Fig. 10.5 M2 helices, comparing structure from molecular dynamics (MD) simulations with those from X-ray structures. (A) M2 helices from three crystal structures (KcsA – closed; KirBac1.1 – closed; and MthK – open). (B) M2 helices from an MD simulation of KirBac1.1. (C) M2 helices from an MD simulation of KirBac3.1. In each case the M2 helices are superimposed on the Ca atoms N-terminal to the conserved glycine hinge residue. 10.3 Kir Channels 251 a timescale substantially shorter than that of channel gating, support a gating model in which the intrinsic flexibility of M2 about a molecular hinge is used to modify the pore dimension at the intracellular mouth, enabling switching between a closed conformation (undistorted M2 helix, narrow hydrophobic pore mouth) and an open conformation (kinked M2 helix, wide pore mouth). Simula- tions of the closed structure of KcsA in comparison with simulations of a model of the KcsA open state (based on MthK) also support this model [71], as do steered MD simulations [72], and normal mode analysis [73] calculations based on the closed KcsA structure.

10.3.6 Intracellular Domain Dynamics

As noted above, the intracellular C-terminal (IC) domain of Kir channels regulates channel gating in response to changes in concentration of various ligands. MD simulations (~10 ns) have been used to probe the dynamics of two Kir C-terminal domain tetramers, namely Kir3.1 (a crystal structure) and Kir6.2 (a homology model). The Kir6.2 simulations were performed with and without bound ATP (introduced by docking). The results of these simulations revealed comparable conformational stability and dynamics for the crystal structure and for the homology model. However, principal components analysis (PCA) of the simulations did reveal a conserved pattern of motion of relevance to channel gating. Thus, in both the Kir3.1 and Kir6.2 tetramers PCA revealed loss of symmetry, consistent with a dimer-of-dimers motion of subunits in the IC domains of the corresponding Kir channels. Of course, the timescale on which this was observed is considerably shorter than that associated with channel gating. To test this hypothesis further, coarse-grained (anisotropic network model [52]) calculations were performed on the IC domain. These also revealed a dimer-of-dimers motion of the IC domain tetramer, essentially identical to that seen in the MD simulations (Fig. 10.6). This suggests that extrapolation from the MD simulation timescale to a channel gating timescale may be valid. Of course, to fully understand gating of Kir channels we need not only to understand the intrinsic flexibility of the TM and IC domains, but also the nature of the interactions of regulatory ligands with the IC domain. Unfortunately, on a 10 ns timescale, the flexibility of the Kir6.2 tetramer was not changed greatly by the presence of docked ATP, other than in two loop regions.

10.3.7 Interactions with Ligands

Ligand docking studies with a homology model of Kir6.2 have been used to help identify the ATP-binding site [61, 74]. The resultant intact model is consistent with a substantial body of functional data and has been tested by mutagenesis. Li- gand binding occurs at the interface between two subunits. The phosphate tail of ATP interacts with two basic sidechains (R201 and K185) in the C-terminus of one subunit, and with a further basic sidechain (R50) in the N-terminus of another; 252 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

Fig. 10.6 Kir intracellular C-terminal (IC) do- mulation of a model of the IC domain tetra- main dynamics, with the protein shown as a mer from Kir6.2. (B) The first eigenvector Ca trace, and with arrows attached to each Ca (i.e. the lowest frequency mode) from an ani- atom to indicate the direction of an eigenvec- sotropic network model [52] analysis of the IC tor and the magnitude of the corresponding domain tetramer of Kir3.1. (C) A schematic eigenvalue. (A) The motion corresponding to diagram summarising the motions of the four the first eigenvector from PCA of an MD si- subunits as revealed by the eigenvectors.

the N6 atom of the adenine ring interacts with E179 and R301 in the same subunit. Significantly, mutation of residues lining the binding pocket reduced ATP- dependent channel inhibition, lending support to the model. This provides a clear example of how a homology model can be used to rationalise existing mutation data, and to aid design of novel mutational experiments.

Kir6.2 channels are modulated by ligands other than ATP, including PIP2.Si- mulations of PIP2 may be used to explore inositol headgroup conformations relative to the PC bilayer environment of the PIP2 molecule (Haider and Sansom, manuscript in preparation). In parallel, continuum electrostatics (Poisson–Boltz- mann) calculations may be used to identify regions on the surface of the Kir6.2 channel model which have a highly positive potential and thus are likely to inter-

act with the polyanionic headgroup of PIP2. By combining such calculations with docking studies of an IP3-fragment from the PIP2 simulations (Fig. 10.7) a model of the interactions of Kir6.2 with PIP2 may be generated and subsequently used in simulations to further explore how such ligands modulate the behavior of Kir channels (Haider and Sansom, unpublished work). 10.3 Kir Channels 253

Fig. 10.7 Interactions of PIP2 with Kir6.2. sible binding site for PIP2. (B) Snapshot from (A) A molecular surface representation of a a simulation of three PIP2 molecules within a model of the Kir6.2 channel calculated using POPC bilayer. The PIP2 molecules are shown GRASP [115], color on electrostatic potential in space-filling format whilst the phosphorus (from –6.6 to 5.4 kT, red to blue). The region atoms of the POPC headgroups are shown as of positive electrostatic potential (blue) near green spheres. (This figure also appears with the intracellular membrane/water interface the color plates.) (indicated by the circle) corresponds to a pos-

10.3.8 Towards an Integrated Gating Model

It can be seen how simulation and modeling studies may provide insights into the intrinsic flexibility of the two major domains (TM and IC) of the Kir channel, and into the nature of the interactions of the IC domain with various ligands. One may attempt to integrate this information to formulate a gating model for Kir channels. In such a model (Fig. 10.8) a transition between exact tetrameric symmetry and dimer-of-dimers symmetry of the IC domains is associated with a change in TM helix packing coupled to gating of the channel. In this model, the open state of the channel has four kinked M2 helices which interact with a tetrameric IC domain, thus holding the helices in their open conformation. In contrast, in the closed channel the helices are no longer kinked, and pack together in a dimer-of-dimers conformation along with the IC domain. This model receives some support from the X-ray structures in that both KirBac1.1 and KirBac3.1 (which are in closed conformations) exhibit a degree of dimer-of-dimers symmetry, i.e. the crystallographic asymmetric unit is a dimer rather than a monomer. Of course, this model is speculative. However, it does provide a framework for further exploration, both computational (via more coarse-grained protein simula- 254 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

Fig. 10.8 Proposed gating mechanism for Kir terminal domains (pale and dark gray circles), channels. The M2 helices of the TM domain as indicated by the arrows, result in closure of of two opposite subunits are shown in rib- the channel via adoption of a dimer-of-dimers bons format with the glycine hinge in dark packing by the M2 helices, which also lose gray. In the open state these helices are their kink upon channel closure. kinked. Asymmetric movements of the C-

tions) and experimental, of possible changes in symmetry associated with Kir channel gating. An even more challenging task will be to understand the way in which interactions with specific ligands stabilise either the open or closed form of the channel.

10.4 Kv Channels

10.4.1 Structures

The other major class of K+ channel for which structural data are available is the voltage-gated (Kv) channels. Kv channels open and close in response to changes in transmembrane voltage, and play a key role in electrical signalling by excitable cells. Kv channel subunits contain two distinct but functionally coupled TM domains (Fig. 10.9). The pore domain has a similar fold to that of other K+ channels, and shares significant sequence homology in the extracellular pore region with KcsA. The voltage sensor (VS) domain is responsible for triggering a change in conformation of the pore domain in response to changes in TM voltage so as to open the channel. Each Kv channel subunit consists of six a-helices (S1-S6). The first four TM helices form the VS domain (Fig. 10.9B), whereas the last two TM helices (S5-S6), along with an intervening re-entrant P loop, form the pore domain (Fig. 10.9C). Crystallographic studies of a bacterial Kv homologue, KvAP [10], have confirmed that the pore domain of KvAP has an architecture similar to 10.4 Kv Channels 255

Figure 10.9 (A) Schematic diagram of the transmembrane topology of a Kv channel subunit, showing the voltage sensor (S1 to S4) and pore (S5 to S6) domains. The intact channel is made up of four such subunits. The intracellular (IC) and extracellular (EC) faces of the membrane are labelled. (B) Structure of the (isolated) voltage sensor domain of the KvAP, with the S4 helix in dark gray. (C) Structure of the pore domain taken from the X-ray structure of the intact KvAP channel.

that of other K+ channels. The main activation gate therefore lies at the intracellular mouth of the channel, at the crossing point of the S6 helices. The revelation of the structure of KvAP has resulted in some controversy concerning the VS domain. This domain is made of helices S1 to S4, with the positively charged S4 helix (which has every third residue positively charged in its N- terminal half) responding to a change in TM voltage in order to initiate the conformational change that results in channel activation gating. On the basis of the X- ray structure of the intact KvAP channel it has been suggested that the S3 and S4 helices form a ‘paddle’ that lies close to the intracellular surface of the membrane in the resting state of the channel, and which is induced to cross the bilayer upon membrane depolarisation, thus triggering activation gating of the channel [75]. However, the exact conformation and orientation of the VS domain remains un- certain. In particular, the structure of the VS domain within the crystal of the full length channel construct is apparently at odds with various physiological and biophysical data [76, 77]. Interestingly, the orientation of the VS domain in the X-ray structure of the intact channel, and in a single particle EM structure [78] seem to be somewhat different, suggestive of a conformational transition associated with channel gating. Significantly, in addition to the crystal structure of the full length KvAP channel, the structure of an isolated VS domain fragment has also been solved (PDB code 1ORS) [10] (Fig. 10.9B). The isolated VS domain structure is perhaps more consistent with a range of biophysical and physiological data, and thus may be more representative of the conformation of the VS domain under physiological conditions. Combined with the relatively high resolution (1.9 Å) of the VS structure, this domain thus provides an attractive candidate for simulation studies of its conformational dynamics in the context of possible channel gating mechanisms. Further- more, the structure of the isolated KvAP VS domain closely resembles that of the 256 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

equivalent domain in a recent X-ray structure of an intact mammalian Kv1.2 channel [79]. However, let us first reconsider the pore domain and the gate per se.

10.4.2 S6 Helices, Hinges and Gating

Comparison of crystal structures of potassium channels and simulations of e.g. KirBac (see above) suggest that the gating mechanism of Kv channels may involve a key role for the pore-lining S6 helix. Within the S6 helix sequence there is a conserved PVP sequence motif. Mutations of this motif have been shown to alter the gating of Kv channels in a manner consistent with the formation of a molecular gating hinge in this region [80, 81]. Molecular dynamics simulations of isolated S6 helices in membrane-like environments [82] and of models of the pore-forming domain of Kv channels [83] have been used to explore the conformational dynamics of the S6 helix hinge. The results of these simulations indicate that the PVP motif may form a molecular hinge. In the isolated helices there is considerable (albeit anisotropic) motion about the PVP hinge (Fig. 10.10A). Substantial hinge-bending motion remains, even when the S6 helix forms part of a more complete pore domain model. Thus the intrinsic conformational dynamics of S6 are modulated by the remainder of protein but it remains flexible. These simulation results are suggestive of a channel gating model in which S6 bends in the vicinity of the PVP motif (Fig. 10.10B) in addition to the region around the conserved glycine that is N-terminal to the PVP motif. Evi- dence in favor of a helix distortion of S6 in the region of the PVP motif has also been obtained by cysteine modification studies [84, 85], and a PVP-hinge is also supported by the recent X-ray structure of a Kv1.2 channel [79, 86]. Thus, K+ channel gating may depend on a complex switch in conformation with three rigid helical sections linked by two molecular hinges, one at the conserved glycine and one at the PVP motif.

Fig. 10.10 (A) S6 helix structures taken from a simulation of the isolated helix in a membrane mimetic octane slab [82]. The N- terminus of the helix is at the top of the diagram, and the location of the PVP hinge motif is indicated. (B) Schematic model of the inner helix bundle of Kv channel pore as formed by four kinked S6 helices. 10.4 Kv Channels 257

10.4.3 The Barrier at the Gate

Homology modeling has been used to generate models of two states of the pore domain of the Shaker Kv channel: one, based on the X-ray structure of KcsA, represents a closed state of the channel, whilst the other, based on KvAP, represents an open state pore domain (Fig. 10.11A and B) [87]. It is evident that in the Kv- closed model the intracellular mouth of the channel (i.e. the activation gated) is much narrower (radius ~1.4 Å) than in the Kv-open model (radius >10 Å). A simple calculation of the energetic barrier presented by the hydrophobic gate region in the Kv-closed model may be obtained via a Poisson–Boltzmann calculation [16, 88], which provides a first approximation to the barrier height presented by a hydrophobic gate. The pore radius profile of the Kv-closed model has an average hydrophobic gate radius of 1.4 Å, extending over a length of 2 Å. This yields a broad

Fig. 10.11 Comparison of Kv-closed and Kv- shown by the horizontal broken lines. The sur- open models. Homology models of the Sha- face of the pore is shown as calculated using ker Kv pore domain, based on (A) KcsA (for Hole [117]. (C) Poisson–Boltzmann energy as the Kv-closed model) and (B) KvAP (for the a function of position along the pore axis for Kv-open model). In both cases the model is a singly charged cation for the two models restricted to the pore-forming domain and shown in (A). The black line corresponds to only two of the four subunits are shown. The Kv-closed, and the gray line to Kv-open. approximate location of the lipid bilayer is 258 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

barrier of maximum height ~80 kT in the centre of the gate (Fig. 10.11C). A comparison of Poisson–Boltzmann energies and atomistic simulation-based free energies for simple nanopore models [88] suggests that Poisson–Boltzmann calculations may overestimate the barrier height by a factor of about two. Correcting for this still yields an estimated barrier height of ~40 kT for the Kv-closed channel. Thus the model clearly corresponds to a fully closed conformation of the channel. In contrast, there is no barrier at all in the Kv-open model. Thus, S6 helix bending motions are sufficient to switch the pore fully from a functionally closed to a functionally open state.

10.4.4 The Nature of the Voltage Sensor

As discussed above, there is some uncertainty as to the conformation and/or localisation of the VS domain within Kv channels. Comparison of the X-ray structures of the VS in the intact KvAP channel (PDB code 1ORQ), and of the isolated VS domain (PDB code 1ORS) reveals a number of differences in their conformation. In particular, the isolated VS more closely resembles a conventional membrane protein. It consists of a bundle of four TM a-helices which are capable of spanning a lipid bilayer. The isolated VS domain forms crystals capable of diffracting to a high resolution [10]. It also forms a stable folded domain in detergent micelles [10, 75] and in lipid bilayers [89]. Given all this, the isolated VS domain structure thus may represent a conformation of the VS domain within the intact channel under physiological conditions. It is therefore of interest to explore the localisation of the VS domain in a membrane (and in the KvAP channel) and to examine its conformational dynamics. One approach to help localise the VS in a membrane is via a number of toxins that interact with the voltage sensor of Kv channels [76, 90, 91]. VSTX1 and related toxins from spider venoms bind to the voltage sensor of Kv channels. These toxins are thought to access the VS via the lipid bilayer phase. Consequently,VSTX1 provides a probe to help localise the VS of Kv channels relative to the centre of the bilayer. However, this requires knowledge of the location of the toxin within the bilayer. This may be approached via MD simulations. In a recent study [92] a series of simulations was used to explore the net drift of VSTX1 relative to the centre of a PC bilayer, starting from different locations of the toxin. The preferred location of the toxin appears to be at the membrane/water interface, ~15 Å from the centre of the bilayer. This interfacial location allows the toxin to maximise its H-bonding to lipid headgroups and to water molecules, whilst retaining extensive hydrophobic interactions with the hydrophobic core of the bilayer. Thus, the toxin partitions to the water/bilayer interface (in itself an extensive region, about 10 Å thick [93]), with its hydrophobic residues pointing down towards the hydrophobic core of the bilayer, and its hydrophilic sidechains interacting with the lipid headgroups and with interfacial water molecules. This localisation in turn allows one to develop a model of VSTX1 interactions with the voltage sensor of Kv channels. 10.4 Kv Channels 259

Mutational studies on voltage sensor-binding toxins combined with mutational studies of Kv channels point to an interaction between polar/charged residues on the toxin surface with residues towards the C-terminus of the S3b helix of the voltage sensor [91, 94, 95]. Based on this information we can model the positions of VSTX1 and the KvAP voltage sensor relative to a POPC lipid bilayer. As seen in Fig. 10.12,VSTX1 is located such that a band of positively charged residues on its surface (which are likely to form the interactions between the toxin and the voltage sensor) is located exactly at the membrane interface. The VS crystal structure may be inserted into a POPC bilayer such that (i) it spans the bilayer symmetrically; and (ii) a surface-exposed tyrosine (Y46 of S1) is located at the interface. Note that tyrosine and tryptophan residues tend to be located in bands on the surfaces of membrane proteins corresponding to the lipid/water interface [96]. In this orientation, residues in the S3b helix (which forms part of the crucial gating paddle [75, 76, 97] of the voltage sensor) are aligned at the correct depth to interact with the basic residues of VSTX1. Thus, it would seem that VSTX1 binds to the tip of the gating paddle (as defined by S3b and the N-terminal region of S4), which is in turn located at the extracellular interfacial region of the membrane. A number of recent studies have focussed on the manner in which the VS and pore domains of Kv channels are packed together [98], and on exploring the changes in conformation and/or orientation of the VS in response to membrane depolarisation [76]. For example, site-directed spin label data [99] has been used to develop an alternative model of how the VS and pore domains of KvAP may pack together within a lipid bilayer environment. A key feature of this model is that the basic sidechains of S4 are exposed to the surrounding lipid environment. Several models of the conformational change underlying Kv channel gating have been

Fig. 10.12 A model for VSTX inhibition of Kv tal structure of the isolated KvAP voltage sen- channels. VSTX1 is shown partitioned to the sor domain is shown such that a cluster of re- water/membrane interface, at which location sidues in the S3b helix that may interact with it binds to the voltage sensor (VS). The toxin VSTX1 is shown in space filling format. The is shown in space filling format. The approxi- tyrosine residue of S1 which defines the extra- mate location of the interface is shown by the cellular interfacial region (Y46) is also shown phosphorus atoms of a POPC bilayer (small in space fill. The S3b and S4 helices together gray spheres), the other atoms of the lipid form the ‘paddle’ of the voltage sensor do- molecules being omitted for clarity. The crys- main (circled). 260 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

proposed, namely: the S4 helical screw [100], transporter [101], paddle [75], and twisted S4 models [99]. Thus, it is of some interest to explore the intrinsic flexibility of the VS domain in order to explore further the plausibility of these models. MD simulations have been performed on the isolated VS from KvAP located in a detergent micelle environment (Sands et al., manuscript submitted). Simula- tions at two different temperatures (300 K and 368 K) were used to probe the intrinsic flexibility of this domain on a ~10 ns timescale (Fig. 10.13A). The VS contains a positively charged (S4) helix which is packed against a more hydrophobic S3 helix. The simulations at elevated temperature reveal an intrinsic flexibility of the S3a region (i. e. the N-terminus of the S3 helix; Fig. 10.13B). It is also evident that the S4 helix undergoes hinge bending and swivelling about its central I130 residue (Fig. 10.13C). The conformational instability of the S3a region may facilitate the motion of the N-terminal segment of S4 (i.e. S4a). These simulations thus support a modified model of an S3b-S4a paddle that can swivel relative to the rest of the VS domain. This movement may form the underlying basis of voltage sensing by Kv channels.

10.4.5 A Possible Gating Model

On the basis of these results it is possible to formulate a possible gating model for Kv channels (Fig. 10.14). In such a model the VS domain is envisaged as a ‘spring and swivel‘. The positively charged S4 helix responds to a change in transbilayer electrostatic field (i.e. membrane depolarisation) such that its S4a region swivels relative to the C-terminal half of the S4 helix and the remainder of the voltage sensor. The conformational flexibility of S3a is suggested to provide a restoring 10.5 Summary and Future Directions 261

Fig. 10.14 A possible gating model for Kv channels. The VS domain is shown, with the S4 helix in black and the S3a segment in dark gray. The positively charged S4 helix is suggested to respond to membrane depolarisation (DV) by the S4a region swivelling relative to the S4b region which remains fixed relative to the remainder of the VS. The conformational flexibility of S3a may provide a restoring ‘spring’ able to return the voltage-sensing paddle (formed by S3b- S4a) to its initial conformation once the resting membrane potential is restored.

‘spring’ connected to the voltage-sensing paddle (formed by S3b-S4a) which resets the conformation upon return to the resting membrane potential. This model is therefore related to both the paddle model of MacKinnon and colleagues [75] and to models (reviewed in e.g. Ref. [97]) in which S4 is proposed to undergo a rotational motion, as suggested by e.g. fluorescence energy transfer experiments [102, 103]. It remains rather difficult to formulate detailed models of how a proposed motion of the VS upon depolarisation may be linked to the conformational change, involving bending of the S6 helices, that opens the Kv channel. The construction of such a model requires a more definitive structure and/or model of the way in which the VS and pore domains of Kv channels are packed together in the intact channel protein under physiological conditions. This remains elusive, although a number of methods can reveal important clues [99].

10.5 Summary and Future Directions

In summary, we have seen how homology modeling and simulation may be combined in order to extrapolate from static structures of bacterial ion channels to a dynamic description of the structural basis of function of their mammalian homologs. However, our discussion has been limited to K+ channels, for which substantial structural data are available. In contrast, for neurotransmitter-gated ion channels (e. g. the nicotinic acetylcholine receptor), more limited structural data are available [3, 104] which in turn restricts modeling [105] and simulation [106] studies. In brief, as structures for more channels are determined, computational studies relating structure to function will expand. Current simulation methodologies are able to address mechanisms of ion permeation through channels although further refinements are needed, especially with respect to understanding ion selectivity. However, modeling and simulation of gating remains challenging, as described above. We need improved methods that can address longer timescales and larger scale conformational transitions. 262 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

Network models are promising, but are likely to be too approximate to describe fully, for example, voltage- or ligand-induced conformational changes leading to gating. It is therefore important to explore the application to ion channels of a variety of coarse-grained models of proteins [49]. It will also be important to understand the role of protein dynamics in the interactions of channels with their ligands. To date, computational studies have largely focussed on single channel proteins and their component domains. There will be a need in the future to address more complex assemblies, i.e. channel proteins interacting with other protein within and adjacent to the cell membrane. From a computational perspective, this will require the development of multi-scale simulation methods, that are able to model both large-scale events (e.g. protein–protein interactions) at the same time as more fine-grained processes (e. g. ion permeation). Thus, ion channels will continue to provide a focus for the development and exploitation of new computational tools to bridge the gap between structure and physiology.

Acknowledgements

Research in MSPS’s laboratory is supported by grants from The Wellcome Trust, the EPSRC, and the BBSRC. Our thanks to our many colleagues for their ongoing interest in this work.

References

1 J. M. Gulbis, D. A. Doyle, Potassium nel: molecular basis of K+ conduction channel structures: do they conform? and selectivity, Science 1998, 280, 69–77. Curr. Opin. Struct. Biol. 2004, 14, 440– 7 Y. Zhou, J. H. Morais-Cabral, 446. A. Kaufman, R. MacKinnon, Chemis- 2 S. H. White, The progress of mem- try of ion coordination and hydration brane protein structure determination, revealed by a K+ channel-Fab complex at Prot. Sci. 2004, 13, 1948–1949. 2.0 Å resolution, Nature 2001, 414, 43– 3 N. Unwin, Refined structure of the 48. nicotinic acetylcholine receptor at 4 Å 8 A. Kuo, J. M. Gulbis, J. F. Antcliff, resolution, J. Mol. Biol. 2005, 346, 967– T. Rahman, E. D. Lowe, J. Zimmer, 989. J. Cuthbertson, F. M. Ashcroft, 4 O. Sokolova, L. Kolmakova-Par- T. Ezaki, D. A. Doyle, Crystal structure tensky, N. Grigorieff, Three-dimen- of the potassium channel KirBac1.1 in sional structure of a voltage-gated potas- the closed state, Science 2003, 300, sium channel at 2.5 nm resolution, 1922–1926. Structure 2001, 9, 215–220. 9 J. M. Gulbis, A. Kuo, B. Smith, D. A. 5 G. Yellen,The voltage-gated potassium Doyle, A. Edwards, C. Arrowsmith, channels and their relatives, Nature M. Sundstrom, Two intermediate gat- 2002, 419, 35–42. ing state crystal structures of the Kir- 6 D. A. Doyle, J. M. Cabral, R. A. Bac3.1 K+ channel, to be published. Pfuetzner, A. Kuo, J. M. Gulbis, S. L. 10 Y. Jiang, A. Lee, J. Chen,V. Ruta, Cohen, B. T. Cahit, R. MacKinnon, M. Cadene, B. T. Chait, R. Mackin- The structure of the potassium chan- non, X-ray structure of a voltage-depen- References 263

dent K+ channel, Nature 2003, 423, 33– ientational tuning, Science 2002, 296, 41. 525–530. 11 Y. Jiang, A. Lee, J. Chen, M. Cadene, 22 B. Ilan, E. Tajkhorshid, K. Schulten, B. T. Chait, R. MacKinnon, Crystal G. A. Voth, The mechanism of proton structure and mechanism of a calcium- exclusion in aquaporin channels, Pro- gated potassium channel, Nature 2002, teins: Struct. Funct. Bioinf. 2004, 55, 417, 515–522. 223–228. 12 A. Giorgetti, P. Carloni, Molecular 23 B. L. de Groot, H. Grubmuller,Water modeling of ion channels: structural permeation across biological mem- predictions, Curr. Opin. Chem. Biol. branes: Mechanism and dynamics of 2003, 7, 150–156. aquaporin-1 and GlpF, Science 2001, 13 J. M. Cuthbertson, D. A. Doyle, M. S. 294, 2353–2357. P. Sansom, Transmembrane helix pre- 24 D. P. Tieleman, H. J. C. Berendsen, A diction: a comparative evaluation and molecular dynamics study of the pores analysis, Proteins Eng. Des. Sel. 2005, 18, formed by Escherichia coli OmpF porin 295–308. in a fully hydrated palmitoyloleoylpho- 14 W. L. Ash, M. R. Zlomislic, E. O. sphatidylcholine bilayer, Biophys. J. Oloo, D. P. Tieleman, Computer simu- 1998, 74, 2786–2801. lations of membrane proteins, Biochim. 25 W. Im, B. Roux, Ions and counterions Biophys. Acta 2004, 1666, 158–189. in a biological channel: A molecular dy- 15 M. S. P. Sansom, I. H. Shrivastava, namics simulation of OmpF porin from K. M. Ranatunga, G. R. Smith, Simu- Escherichia coli in an explicit membrane lations of ion channels – watching ions with 1 M KCl aqueous salt solution, and water move, Trends Biochem. Sci. J. Mol. Biol. 2002, 319, 1177–1197. 2000, 25, 368–374. 26 P. J. Bond, M. S. P. Sansom, The simu- 16 D. P. Tieleman, P. C. Biggin, G. R. lation approach to bacterial outer mem- Smith, M. S. P. Sansom, Simulation brane proteins, Mol. Membr. Biol. 2004, approaches to ion channel structure- 21, 151–162. function relationships, Quart. Rev. Bio- 27 I. H. Shrivastava, M. S. P. Sansom, phys. 2001, 34, 473–561. Simulations of ion permeation through 17 M. S. P. Sansom, I. H. Shrivastava, a potassium channel: molecular dy- J. N. Bright, J. Tate, C. E. Capener, namics of KcsA in a phospholipid bi- P. C. Biggin, Potassium channels: layer, Biophys. J. 2000, 78, 557–570. structures, models, simulations, Bio- 28 C. E. Capener, P. Proks, F. M. Ash- chim. Biophys. Acta 2002, 1565, 294– croft, M. S. P. Sansom, Filter flexibility 307. in a mammalian K channel: models 18 B. Roux, T. Allen, S. Berneche,W. Im, and simulations of Kir6.2 mutants, Theoretical and computational models Biophys. J. 2003, 84, 2345–2356. of biological ion channels, Quart. Rev. 29 S. Bernèche, B. Roux, Energetics of + Biophys. 2004, 37, 15–103. ion conduction through the K 19 B. Roux, K. Schulten, Computational channel, Nature 2001, 414, 73–77. studies of membrane channels, Struc- 30 B. Roux, S. Bernèche,W. Im, Ion chan- ture 2004, 12, 1343–1351. nels, permeation and electrostatics: In- 20 M. O. Jensen, S. Park, E. Tajkhorshid, sight into the function of KcsA, Bio- K. Schulten, Energetics of glycerol chemistry 2000, 39, 13295–13306. conduction through aquaglyceroporin 31 B. Roux, Theoretical and computational GlpF, Proc. Nat. Acad. Sci. USA 2002, models of ion channels, Curr. Opin. 99, 6731–6736. Struct. Biol. 2002, 12, 182–189. 21 E. Tajkhorshid, P. Nollert, M. O. 32 L. Guidoni,V. Torre, P. Carloni, Po- Jensen, L. J. W. Miercke, J. O’Con- tassium and sodium binding in the nell, R. M. Stroud, K. Schulten, outer mouth of the K+ channel, Bio- Control of the selectivity of the aqua- chemistry 1999, 38, 8599–8604. porin water channel family by global or- 264 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

33 L. Guidoni,V. Torre, P. Carloni, 45 A. E. Garcia, Large-amplitude non- Water and potassium dynamics in the linear motions in proteins, Phys. Rev. KcsA K+ channel, FEBS Lett. 2000, 477, Lett. 1992, 68, 2696–2699. 37–42. 46 A. Amadei, A. B. M. Linssen, H. J. C. 34 M. Compoint, P. Carloni, C. Ram- Berendsen, Essential dynamics of pro- seyer, C. Girardet, Molecular dy- teins, Proteins: Struct. Funct. Genet. namics study of the KcsA channel at 1993, 17, 412–425. 2.0-Å resolution: stability and concerted 47 A. Amadei, M. A. Ceruso, A. Di Nola, motions within the pore, Biochim. Bio- On the convergence of the conforma- phys. Acta 2004, 1661, 26–39. tional coordinates basis set obtained by 35 J. Åqvist,V. Luzhkov, Ion permeation the essential dynamics analysis of pro- mechanism of the potassium channel, teins’ molecular dynamics simulations, Nature 2000, 404, 881–884. Proteins Struct. Funct. Genet. 1999, 36, 36 V. B. Luzhkov, J. Åqvist, A computa- 419–424. tional study of ion binding and protona- 48 J. D. Faraldo-Gómez, L. R. Forrest, tion states in the KcsA potassium chan- M. Baaden, P. J. Bond, C. Domene, nel, Biochim. Biophys. Acta 2000, 1481, G. Patargias, J. Cuthbertson, M. S. 360–370. P. Sansom, Conformational sampling 37 V. B. Luzhkov, J. Åqvist, Mechanisms and dynamics of membrane proteins of tetraethylammonium ion block in from 10-nanosecond computer simula- the KcsA potassium channel, FEBS Lett. tions, Proteins: Struct. Funct. Bioinf. 2001, 495, 191–196. 2004, 57, 783–791. 38 T. Schirmer, P. S. Phale, Brownian dy- 49 V. Tozzini, Coarse-grained models for namics simulation of ion flow through proteins, Curr. Opin. Struct. Biol. 2005, porin channels, J. Mol. Biol. 1999, 294, 15, 144–150. 1159–1167. 50 I. Bahar, A. R. Atilgan, B. Erman, 39 W. Im, B. Roux, Ion permeation and Direct evaluation of thermal fluctua- selectivity of OmpF porin: A theoretical tions in proteins using a single-para- study based on molecular dynamics, meter harmonic potential, Fold. Des. Brownian dynamics, and continuum 1997, 2, 173–181. electrodiffusion theory, J. Mol. Biol. 51 O. Keskin, R. L. Jernigan, I. Bahar, 2002, 322, 851–869. Proteins with similar architecture exhi- 40 S. H. Chung, T. W. Allen, M. Hoyles, bit similar large-scale dynamic behavior, S. Kuyucak, Permeation of ions across Biophys. J. 2000, 78, 2093–2106. the potassium channel: Brownian dy- 52 A. R. Atilgan, S. R. Durell, R. L. Jer- namics studies, Biophys. J. 1999, 77, nigan, M. C. Demirel, O. Keskin, 2517–2533. I. Bahar, Anisotropy of fluctuation dy- 41 S. H. Chung, T. W. Allen, S. Kuyucak, namics of proteins with an elastic net- Conducting-state properties of the KcsA work model, Biophys. J. 2001, 80, 505– potassium channel from molecular and 515. Brownian dynamics simulations, Bio- 53 S. Haider, A. Grottesi, F. M. Ash- phys. J. 2002, 82, 628–645. croft, M. S. P. Sansom, Conforma- 42 S. H. Chung, T. W. Allen, S. Kuyucak, tional dynamics of the ligand-binding Modeling diverse range of potassium domain of inward rectifier K channels channels with Brownian dynamics, as revealed by MD simulations: towards Biophys. J. 2002, 83, 263–277. an understanding of Kir channel gating, 43 S. H. Chung, S. Kuyucak, Ion chan- Biophys. J. 2005, 88, 3310–3320. nels: recent progress and prospects, 54 S. J. Campbell, N. D. Gold, R. M. Jack- Eur. Biophys. J. 2002, 31, 283–293. son, D. R. Westhead, Ligand binding: 44 J. Gullingsrud, K. Schulten, Gating functional site location, similarity and of MscL studied by steered molecular docking, Curr. Opin. Struct. Biol. 2003, dynamics, Biophys. J. 2003, 85, 2087– 13, 389–395. 2099. References 265

55 V. Krishnamurthy, S. H. Chung, properties of carbonyl ligands, Nature G. Dumont, Special issue on ion chan- 2004, 431, 830 – 834. nels – bionanotubes, IEEE Trans. Nano- 66 Y. Zhou, R. MacKinnon, The occu- biosci. 2005, 4, 1–2. pancy of ions in the K+ selectivity filter: 56 C. G. Nichols, A. N. Lopatin, Inward charge balance and coupling of ion rectifier potassium channels, Annu. Rev. binding to a protein conformational Physiol. 1997, 59, 171–191. change underlie high conduction rates, 57 F. Reimann, F. M. Ashcroft, Inwardly J. Mol. Biol. 2003, 333, 965–975. rectifying potassium channels, Curr. 67 M. J. Lenaeus, M. Vamvouka, P. J. Fo- Opin. Cell Biol. 1999, 11, 503–508. cia, A. Gross, Structural basis of TEA 58 M. Nishida, R. MacKinnon, Struc- blockade in a model potassium channel, tural basis of inward rectification: cyto- Nature Struct. Mol. Biol. 2005, 12, 454– plasmic pore of the G protein-gated in- 459. ward rectifier GIRK1 at 1.8 Å resolu- 68 L. G. Cuello, E. Perozo, Fast gating tion, Cell 2002, 111, 957–65. transitions determined by dynamic rear- 59 C. E. Capener, I. H. Shrivastava, rangements in the selectivity filter of K. M. Ranatunga, L. R. Forrest, G. R. KcsA, Biophys. J. 2002, 82, 853. Smith, M. S. P. Sansom, Homology 69 C. Domene, A. Grottesi, M. S. P. San- modeling and molecular dynamics si- som, Filter flexibility and distortion in a mulation studies of an inward rectifier bacterial inward rectifier K+ channel: si- potassium channel, Biophys. J. 2000, 78, mulation studies of KirBac1.1, Biophys. 2929–2942. J. 2004, 87, 256–267. 60 C. E. Capener, M. S. P. Sansom, MD Si- 70 A. Grottesi, C. Domene, M. S. P. San- mulations of a K channel model – sensi- som, Conformational dynamics of M2 tivity to changes in ions, waters and Helices in KirBac channels: helix flex- membrane environment, J. Phys. Chem. ibility in relation to gating via molecular B 2002, 106, 4543–4551. dynamics simulations, Biochemistry 61 J. F. Antcliffe, S. Haider, P. Proks, 2005, submitted. M. S. P. Sansom, F. M. Ashcroft, 71 J. Holyoake, C. Domene, J. N. Bright, Functional analysis of a structural M. S. P. Sansom, KcsA closed and model of the ATP-binding site of the open: modeling and simulation studies,

KATP channel Kir6.2 subunit, EMBO J. Eur. Biophys. J. 2003, 33, 238–246. 2005, 24, 229–239. 72 P. C. Biggin, M. S. P. Sansom, Open- 62 P. Proks, C. E. Capener, P. Jones, state models of a potassium channel, F. Ashcroft, Mutations within the Biophys. J. 2002, 83, 1867–1876. P-loop of Kir6.2 modulate the intraburst 73 Y. F. Shen,Y. F. Kong, J. P. Ma, Intrin- kinetics of the ATP-sensitive potassium sic flexibility and gating mechanism of channel, J. Gen. Physiol. 2001, 118, 341– the potassium channel KcsA, Proc. Nat. 353. Acad. Sci. USA 2002, 99, 1949–1953. 63 I. H. Shrivastava, D. P. Tieleman, 74 S. Trapp, S. Haider, M. S. P. Sansom, P. C. Biggin, M. S. P. Sansom, K+ vs. F. M. Ashcroft, P. Jones, Identifica- Na+ ions in a K channel selectivity filter: tion of residues contributing to the ATP a simulation study, Biophys. J. 2002, 83, binding site of Kir6.2, EMBO J. 2003, 633–645. 22, 2903–2912. 64 C. Domene, M. S. P. Sansom, A potas- 75 Y. Jiang,V. Ruta, J. Chen, A. G. Lee, sium channel, ions and water: simula- R. Mackinnon, The principle of gating tion studies based on the high resolu- charge movement in a voltage-depen- tion X-ray structure of KcsA, Biophys. J. dent K+ channel, Nature 2003, 423, 42– 2003, 85, 2787–2800. 48. 65 S. Y. Noskov, S. Bernèche, B. Roux, 76 K. J. Swartz, Towards a structural view Control of ion selectivity in potassium of gating in potassium channels, Nature channels by electrostatic and dynamic Rev. Neurosci. 2004, 5, 905–916. 266 10 Molecular Modeling and Simulations of Ion Channels: Applications to Potassium Channels

77 D. J. S. Elliott, E. J. Neale, Q. Aziz, gical function, IEEE Trans. Nanobiosci. J. P. Dunham, T. S. Munsey, M. Hun- 2004, 4, 112–120. ter, A. Sivaprasadarao, Molecular me- 88 O. Beckstein, K. Tai, M. S. P. Sansom, chanism of voltage sensor movements Not ions alone: barriers to ion permea- in a potassium channel, EMBO J. 2004, tion in nanopores and channels, J. Am. 23, 4717–4726. Chem. Soc. 2004, 126, 14694–14695. 78 Q. X. Jiang, D. N. Wang, R. MacKin- 89 L. Cuello, D. M. Cortes, E. Perozo, non, Electron microscopic analysis of Structure and dynamics of the KvAP KvAP voltage-dependent K+ channels in voltage sensor domain reconstituted in an open conformation, Nature 2004, lipid bilayers, Biophys. J. 2005, 86, 458A. 430, 806–810. 90 H. Takahashi, J. I. Kim, H. J. Min, 79 S. B. Long, E. B. Campbell, K. Sato, K. J. Swartz, I. Shimada, So- R. MacKinnon, Crystal structure of a lution structure of hanatoxin1, a gating mammalian voltage-dependent Shaker modifier of voltage- dependent K+ chan- family K+ channel, Science 2005,in nels: Common surface features of gat- press. ing modifier toxins, J. Mol. Biol. 2000, 80 D. H. Hackos, K. J. Swartz, Mutations 297, 771–780. of a conserved proline in the inner helix 91 V. Ruta, R. MacKinnon, Localization of the pore domain of the Shaker K+ of the voltage-sensor toxin receptor on channel with altered gating properties, KvAP, Biochemistry 2004, 43, 10071– Biophys. J. 2000, 78, 398A. 10079. 81 A. J. Labro, A. L. Raes, I. Bellens, 92 D. Bemporad, C. L. Wee, Z. Sands, N. Ottschytsch, D. J. Snyders, Gating F. Blaney, A. Grottesi, M. S. P. San- of Shaker-type channels requires the som, Localising a Kv channel toxin in a flexibility of S6 caused by prolines, membrane with atomistic simulations, J. Biol. Chem. 2003, 278, 50724–50731. Biophys. J. 2005, submitted. 82 J. N. Bright, I. H. Shrivastava, F. S. 93 M. C. Wiener, S. H. White, Structure Cordes, M. S. P. Sansom, Conforma- of a fluid dioleoylphosphatidylcholine tional dynamics of helix S6 from Shaker bilayer determined by joint refinement potassium channel: simulation studies, of X-ray and neutron diffraction data. Biopolymers 2002, 64, 303–313. III. Complete structure, Biophys. J. 83 J. N. Bright, M. S. P. Sansom, The Kv 1992, 61, 434–447. channel S6 helix as a molecular switch: 94 S. Y. Lee, R. MacKinnon, A mem- simulation studies, IEE Proc. -Nanobio- brane-access mechanism of ion channel technol. 2004, 151, 17–27. inhibition by voltage sensor toxins from 84 D. D. Camino, M. Holmgren,Y. Liu, spider venom, Nature 2004, 430, 232– G. Yellen, Blocker protection in the 235. pore of a voltage-gated K+ channel and 95 H. C. Lee, J. M. Wang, K. J. Swartz, In- its structural implications, Nature 2000, teraction between extracellular hana- 403, 321–325. toxin and the resting conformation of 85 S. M. Webster, D. del Camino, J. P. the voltage-sensor paddle in Kv chan- Dekker, G. Yellen, Intracellular gate nels, Neuron 2004, 40, 527–536. opening in Shaker K+ channels defined 96 J. A. Killian, G. von Heijne, How pro- by high-affinity metal bridges, Nature teins adapt to a membrane-water inter- 2004, 428, 864–868. face, Trends Biochem. Sci. 2000, 25, 429– 86 S. B. Long, E. B. Campbell, 434. R. MacKinnon,Voltage sensor of Kv1.2: 97 Z. Sands, A. Grottesi, M. S. P. San- structural basis of electromechanical som,Voltage-gated ion channels, Curr. coupling, Science 2005, 309, 903–908 Biol. 2005, 15, R44–R47. 87 A. Grottesi, C. Domene, S. Haider, 98 M. Laine, D. M. Papazian, B. Roux, M. S. P. Sansom, Molecular dynamics Critical assessment of a proposed model simulation approaches to K channels: of Shaker, FEBS Lett. 2004, 564, 257– conformational flexibility and physiolo- 263. References 267

99 L. G. Cuello, D. M. Cortes, E. Perozo, mated protein homology-modeling ser- Molecular architecture of the KvAP ver, Nucleic Acids Res. 2003, 31, 3381– voltage-dependent K+ channel in a lipid 3385. bilayer, Science 2004, 306, 491–495. 110 E. Lindahl, B. Hess, D. van der Spoel, 100 C. C. Gandhi, E. Y. Isacoff, Molecular GROMACS 3.0: a package for molecu- models of voltage sensing, J. Gen. Phy- lar simulation and trajectory analysis, J. siol. 2002, 120, 455–463. Mol. Model. 2001, 7, 306–317. 101 D. M. Starace, F. Bezanilla, A proton 111 B. R. Brooks, R. E. Bruccoleri, B. D. pore in a potassium channel voltage Olafson, D. J. States, S. Swami- sensor reveals a focused electric field, nathan, M. Karplus, CHARMM: A Nature 2004, 427, 548–553. program for macromolecular energy, 102 A. Cha, G. E. Snyder, P. R. Selvin, minimisation, and dynamics calcula- F. Bezanilla, Atomic scale movement tions, J. Comput. Chem. 1983, 4, 187– of the voltage-sensing region in a potas- 217. sium channel measured via spectro- 112 L. Kalé, R. Skeel, M. Bhandarkar, scopy, Nature 1999, 402, 809–813. R. Brunner, A. Gursoy, N. Krawetz, 103 K. S. Glauner, L. M. Mannuzzau, J. Phillips, A. Shinozaki, K. Varada- C. S. Gandhi, E. Y. Isacoff, Spectro- rajan, K. Schulten, NAMD2: Greater scopic mapping of voltage sensor move- scalability for parallel molecular dy- ment in the Shaker potassium channel, namics, J. Comput. Phys. 1999, 151, Nature 1999, 402, 813–817. 283–312. 104 K. Brejc,W. J. van Dijk, R. V. Klaas- 113 N. A. Baker, D. Sept, S. Joseph, M. J. sen, M. Schuurmans, J. van der Oost, Holst, J. A. McCammon, Electrostatics A. B. Smit, T. K. Sixma, Crystal struc- of nanosystems: application to microtu- ture of an ACh-binding protein reveals bules and the ribosome, Proc. Nat. the ligand-binding domain of nicotinic Acad. Sci. USA 2001, 98, 10037–10041. receptors, Nature 2001, 411, 269–276. 114 M. E. Davis, J. D. Madura, B. A. Luty, 105 S. Amiri, K. Tai, O. Beckstein, P. C. J. A. McCammon, Electrostatics and dif- Biggin, M. S. P. Sansom, The a7 nicoti- fusion of molecules in solution: simula- nic acetylcholine receptor: molecular tions with the University of Houston modeling, electrostatics, and energetics, Brownian dynamics program, Comput. Mol. Membr. Biol. 2005, 22, 151–162. Phys. Commun. 1991, 62, 187–197. 106 R. J. Law, R. H. Henchman, J. A. 115 A. Nicholls, R. Bharadwaj, B. Honig, McCammon, A gating mechanism pro- GRASP: graphical representation and posed from a simulation of a human a7 analysis of surface properties, Biophys. nicotinic acetylcholine receptor, Proc. J. 1993, 64, 166–170. Natl. Acad. Sci. USA 2005, 102, 6813– 116 D. S. Goodsell, G. M. Morris, A. J. 6818. Olson, Automated docking of flexible 107 A. Sali, T. L. Blundell, Comparative ligands: applications of AutoDock, J. protein modeling by satisfaction of spa- Mol. Recog. 1996, 9, 1–5. tial restraints, J. Mol. Biol. 1993, 234, 117 O. S. Smart, J. G. Neduvelil, X. Wang, 779–815. B. A. Wallace, M. S. P. Sansom, Hole: 108 A. Fiser, R. Kinh Gian Do, A. Sali, A program for the analysis of the pore Modeling of loops in protein structures, dimensions of ion channel structural Prot. Sci. 2000, 9, 1753–1773. models, J. Mol. Graph. 1996, 14, 354– 109 T. Schwede, J. Kopp, N. Guex, M. C. 360. Peitsch, SWISS-MODEL: an auto- 269

Subject Index a – fluorescence assays 199 atomic absorption spectrometry (AAS) – high throughput screening 196–199 166–175, 202 calcium channel, voltage-gated (CaV) – BK channel assay 170–171 – assays 85–86, 196–197 – detection 169 – blockers 197 – SK channel assay 171–172 – cDNA instability 80 – analysis 169 – mutagenesis 47–48 – cell loading 169 caged amino acids 72 –Na+ channel assay 174 capacitance compensation 7, 117, 147, – ligand-gated channel assay 174 157 – voltage-gated K+ channel assay 172–173 capsaicin 198 –Cl– assay 174 capsazepine 198 access resistance 116 cDNA instability 80 accessory (auxiliary) factors 81, 88 cDNA sub-cloning 28–36 affinity tags – see expression tags cell attached recording 9 alanine scanning – see mutagenesis Cerenkov counting 167 scanning channel activation analysis 127 allosteric modulators 203 channel block 132 anion indicators 195 channel inactivation analysis 127 antibody fragments for crystallisation 229 channel run-down 134, 151 apamin 172, 180 chaperones 87–88, 95 automated electrophysiology 13–17, 82, Cheng-Prussof equation 175 97–99, 145–162 chimeric channels 36–43, 50 chloride channels 203 b – assay formats 174, 203–205 BacMam 87, 93–95 –Ca2+-activated, oocytes 18, 114 batrachatoxin 200 – CFTR 195, 203 benzodiazepine drugs 203 – halide sensors 195, 203–205 bicistronic – see IRES concatamers 52–53 biophysics simulation 142 coumarin dyes 190–192 BK channel – see K+ channels crystallisation, automation 230–232 Boltzmann analysis 124–125, 127 crystallisation screens 227–228, 230 Brownian dynamics 244 current clamp 112 Brugada syndrome 20 current-voltage relationship 119, BSC1 channel 17–18 122–127 BTC 195, 203 cysteine scanning – see mutagenesis, scanning c cystic fibrosis transmembrane regulator calcium (Ca2+) channels (CFTR) 80, 88, 195, 203–205 – toxicity 90 cytotoxicity 85–87, 90, 95, 101, 218

Expression and Analysis of Recombinant Ion Channels. Edited by Jeffrey J. Clare and Derek J. Trezise Copyright # 2006 WILEY-VCH Verlag GmbH & Co. KGaA,Weinheim ISBN: 3-527-31209-9 270 Subject Index

d FMP kit 190, 200–201 defective trafficking 205 FRET probes 188–192 Del-Castillo-Katz model 131–132 deltamethrin 201 g desensitisation 129 GABAA receptor 203 detergent solubilisation 219–223 gating 18–19, 122–135 dihydropyridines 197 – activation 18–19 dissociation constant 128 – charge 135 domain swaps – see mutagenesis, chimeras – currents 135–136 dwell time analysis 140–141 – deactivation 120, 130 dynaflow 154 – inactivation 18–19 – Kir channels 250–251, 253–254 e – Kv channels 130, 256–258, 260–261 electrochromic probes 188 – modifiers 132 electrophysiology, automated 145–162 – valence 124 electrophysiology, analysis 111–142 gene fusions 50–52, 227 electrophysiology, oocytes 5–12 Giga Ohm seals 150 electropositioning 150 Goldman-Hodgkin-Katz equation 119, endogenous channels – mammalian cells 123–125 81 endogenous channels – oocytes 3, 79, h 114 halide sensors 195, 203 envelope protocols 130 hERG – see K+ channels episomal vectors 91–92 high throughput screening E-VIPR 201–202 –K+ channel assays 175–180, 201–203 excised patch 9–12 – membrane potential probes 190–193 expression systems – scintillation proximity assay 178–180 – for structural studies 216–219 – membrane potential probes 193 – inducible 101–103 – state dependence 197 – oocyte 1–3, 113–115 – automated electrophysiology 159–161 – stable 96–103, 115, 157, 216 –Na+ channel assays 155, 200–201 – transient 91–92, 115, 148 – radioligand binding assays 175–180 – viral 87, 92–95, 217 Hill equation 128, 132 expression tags 29–30, 50–52, 226 Hodgkin and Huxley 112, 117 homology modelling 243 f host cell lines 81–85 flame photometry 168 – accessory factors 81, 83–85 FLIPR 189, 196 – adherence 83 FLIPR membrane potential kit 190 – endogenous channels 81 fluorescence assays 187–200 – neuroblastoma 83 –Ca2+ dyes 194 – resting potential 85 – FLIPR 189–190, 196 – FRET 188–192 i – ion indicators 194–205 iberiotoxin 170 – TRP channels 197–198 ICRAC 199 –Ca2+ channels 196–197 inclusion bodies 215 –Na+ channels 200 inducible expression – see expression – polarisation 180–181 systems fluorescence microscopy 18–19 intracellular solution 7, 114 fluorescent ion indicators 194 ion channel block 121 fluorescent polarisation 180–81 ion conductance 118 fluorimetric imaging plate reader (FLIPR) ion flux assays 166–175 189 ion selectivity 118–121 fluoro-ligands 180 ion-sensitive fluorescent probes 194–199 Subject Index 271

IonWorks HT 82, 147, 150 n IonWorks Quattro 97, 159–161 neurotoxins 175 IRES (internal ribosome entry site) expres- nicotinic acetylchloine receptor 65–66 sion vectors 82, 96–101 nimodipine 197 isochronous tail protocol 125–126 noise analysis 133–134 nonsense suppression 59-60 k kinetic analysis 129–132, 156 o oocytes 21 l – dominant genotype 115 Lambert-Beer-Bouger Law 168 – voltage-clamp 5–11 leak correction 116–117, 151 – biochemical analysis 12–13 ligand-gated channels 127–129, 154, 174, – nuclear injection 1 197–199 – recording configurations 114–115 – calcium assays 198–199 – surface expression 115 – radioligands 177 – vitelline membrane 10, 12, 79, 115 – Cl-channel 203 – endogenous currents 3, 79, 114 – desensitisation 152 – intracellular chloride 114 – electrophysiology analysis 126–129, – isolation & injection 2–4, 6, 11, 114 131 – for mutagenesis studies 46–48, 63, 64, liquid junction potentials 119 68–69 – single channel analysis 11 m open probability 117, 122 macropatch, 9–11 optical recording 192 macroscopic kinetics 129–136 Opus Xpress 16–17 macroscopic recordings 117–128 oxonol dyes 188–192 Markov model 139–141 Markov process 112, 129, 139 p membrane capacitance 7, 10, 116, 147, patch clamp electrophysiology 112–115, 157 145 membrane potential probes 188–193 PCR 31–33, 39–43 – advantages & limitations 192, 204 perforated patch clamp 150 – DIBAC 188–192 permeability analysis 118–121 – FRET 188 phase determination 232–234 – high throughput screening 190–193 planar array electrophysiology 145–161 – oxonol dyes 188–192 – capacitance compensation 151 microfluidics 153 – cell preparation 148–149 microspheres 179 – cell sealing & recording 149–150 minK 157 – drug application 152–154 molecular dynamic simulation – Kir – quality control 156 channels 249-254 – series resistance compensation 151 molecular dynamic simulation – Kv – equipment 147 channels 256, 258–260 – experimental methods & design molecular dynamics 244 147–148 MQAE 195 – population patch clamp 159–161 multiplexing 161 – theory 146–147 multi-subunit channels 52–53, 91, 93, 98, population patch clamp – see planar array 100–101 electrophysiology mutagenesis pore analysis 118–121 – chimeras 36–43, 50 post-translational processing 85–90 – gene fusions 50–52 – folding 85, 87–88, 215–216 – scanning 18, 45–49 – proteolytic cleavage 85 – site directed 20, 43–49 potassium (K+) channels – unnatural amino acid 60–65 – AAS assays 169–173 272 Subject Index

– automated electrophysiology 148–150, series resistance correction 113–114, 147, 154, 157 157 – BK channel 170–171 shaker – see K+ channels – hEAG2 channel 34 signal filtering 137-138 – hERG channel 67–72, 82, 89, 103, 153, single channel 173, 175, 208 – analysis 136–137 – high throughput screening 201–203 – oocytes 11–12 – KCNK2 (TREK-1) channel 85 – recording 136 – Kir channels 246–254 – conductance 123 – KATP channels 202 – kinetics 138 – radioligand binding – missed events 141 – SK channel 171–172 site-directed mutagenesis (SDM) – see muta- – structure 215, 218, 241–242, 247 genesis – Kv (shaker) channels SK channel – see K+ channels – – dynamic simulation 256, 258–260 sodium channels, voltage gated (NaV) – – gating 130, 256–258, 260–261 – activators & gating modifiers 200 – – structure 215, 218, 254–256 – automated electrophysiology 160 – – mutagenesis 18–19, 38, 39–42, 52 – cDNA instability 80 – – voltage sensor 258–260 – expression in mamalian cells 83–85 – – Kv7.1 (KCNQ1) 98–99, 157 – expression in oocytes 8 potassium (K+) depolarisation 202 – high throughput screening 174, 200–201 potassium (K+) indicators 195 – indicators 195 power spectrum 133 – mutations 19–21 principle component analysis (PCA) 246 sodium green 195 purification 223–227 spin labelling 72 purification, affinity 226, 227 SPQ 195 purification tags – see expression tags stable cell lines – see expression systems state transitions 112, 125–130 r state-dependence 197, 132 radioactive ion flux assays 167–168 structure factor sharpening 234 radioligand binding 175–181 SUR1 202 – assay interference 179 – filtration assays 177–178 t – oocytes 13 tags – see expression tags – scintillation proximity assay tail-current analysis 119–121, 125 178–180 taqman 81 – SK channel assay 178–179 TEA 136 receptor desensitisation 152 thallium 201–203 recovery from inactivation 130 trafficking 88 rectification 123 trafficking, assays 205–207 redistribution probes 188–190 trafficking abnormalities 207 reporter genes 85, 95 transient receptor channels 197 reversal potential 118–120, 123 transient transfection – see expression RNA injection 2–4, 113 systems Robocyte 14–15 TRP channels 197–198 two-electrode-voltage-clamp 5, 113 s SBFI, 195 u scanning mutagenesis – see mutagenesis use-dependence 158, 201-202 scintillation proximity assay (SPA) 167, 178–180 v selection marker 90, 100 variance-mean analysis 134–135 selenomethionine labeling 232–233 venfalerate 201 Semliki forest virus (SFV) 92 veratridine 200 Subject Index 273 viral expression – see expression systems, viral w vitelline membrane 10–12, 79, 115 whole-cell recording 115, 146, 150 voltage ion probe reader (VIPR) 196, 201– Woodhull model 121–123, 132 202 voltage-clamp 112, 115–116 x – oocytes 5–11 Xenopus laevis – see oocytes – control 114 – errors 113, 116 y voltage-dependence 158, 197 yellow fluorescent protein (YFP) 95, 205 voltage-dependent block 121–126, 132, 197 voltage-independent channels 197 z voltage-sensor 18–19, 136, 258–260 Z' value 156