Advances in Experimental Medicine and Biology 796
Marta Filizola Editor G Protein-Coupled Receptors - Modeling and Simulation Advances in Experimental Medicine and Biology
Volume 796
Editorial Board
Nathan Back, State University of New York at Buffalo, Buffalo, NY, USA Irun R. Cohen, The Weizmann Institute of Science, Rehovot, Israel N. S. Abel Lajtha, Kline Institute for Psychiatric Research, Orangeburg, NY, USA John D. Lambris, University of Pennsylvania, Philadelphia, PA, USA Rodolfo Paoletti, University of Milan, Milan, Italy
For further volumes: http://www.springer.com/series/5584
Marta Filizola Editor
G Protein-Coupled Receptors - Modeling and Simulation
123 Editor Marta Filizola Department of Structural and Chemical Biology Icahn School of Medicine at Mount Sinai New York, NY USA
ISSN 0065-2598 ISSN 2214-8019 (electronic) ISBN 978-94-007-7422-3 ISBN 978-94-007-7423-0 (eBook) DOI 10.1007/978-94-007-7423-0 Springer Dordrecht Heidelberg New York London
Library of Congress Control Number: 2013951640
© Springer Science+Business Media Dordrecht 2014 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com) Preface
G protein-coupled receptors (GPCRs) are membrane proteins of significant interest in pharmaceutical research owing to their involvement in several important biological processes, including those leading to some serious medical conditions. In spite of focused research, progress towards the discovery of effective therapeutics for GPCRs has long been hampered by the lack of high-resolution structural information about these receptors. Although the number of high-resolution crystal structures of GPCRs has grown significantly in the past few years, the information they provide is limited. Not only are we still far from a comprehensive structural coverage of the GPCR superfamily, but the available structures refer to static, heavily engineered, and generally inactive conformational states of receptor subtypes stripped out of their natural lipid environment. Furthermore, it has become increasingly clear that a full understanding of GPCR structure and function requires dynamic information at a level of detail that is likely to require integration of experimental and computational approaches. Significant progress has been made over the past decade in the devel- opment and application of computational approaches to the large family of GPCRs. A dedicated book that discusses in depth this important topic is lacking but strongly needed owing to: (a) the critical (but sometimes unappreciated) impact that these computational approaches have on under- standing the molecular mechanisms underlying the physiological function of GPCRs in support of rational drug discovery, (b) the recent advances in theory, hardware, and software, and (c) the potential for much-improved applications using newly available experimentally-derived structural and dynamic information on GPCRs. Thus, I sought the help of experts with an established reputation in the development and/or application of computational methods to GPCRs, and asked them to contribute their state-of-the-art views on modeling and simulation of this important family of membrane proteins. I am indebted to the highly distinguished authors of this book for agreeing to participate in this project and to provide chapters for four different sessions: a first one describing the impact of currently available GPCR crystal structures on structural modeling of other receptor subtypes, a second one reporting on critical insights from simulations, a third one focusing on recent progress in rational ligand discovery and mathematical modeling, and a fourth one providing an overview of bioinformatics tools and resources that are available for GPCRs.
v vi Preface
Heartfelt thanks also go to the several anonymous reviewers of the chapters, and to Thijs van Vlijmen from Springer for the opportunity he offered me to edit this volume. I believe this book adds a unique facet to the “Advances in Experimental Medicine and Biology” series, and I hope the reader will find it both fascinating and of enduring interest.
New York, NY, USA Marta Filizola, Ph.D. June 3, 2013 Contents
Part I Progress in Structural Modeling of GPCRs
1 The GPCR Crystallography Boom: Providing an Invaluable Source of Structural Information and Expanding the Scope of Homology Modeling ...... 3 Stefano Costanzi and Keyun Wang 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures: From Monomers to Signaling Complexes ...... 15 Angel Gonzalez, Arnau Cordomí, Minos Matsoukas, Julian Zachmann, and Leonardo Pardo
Part II GPCRs in Motion: Insights from Simulations
3 Structure and Dynamics of G-Protein Coupled Receptors .... 37 Nagarajan Vaidehi, Supriyo Bhattacharya, and Adrien B. Larsen 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated by Their Coupling to the Membrane Environment ...... 55 Sayan Mondal, George Khelashvili, Niklaus Johner, and Harel Weinstein 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids and Cholesterol with Rhodopsin ... 75 Joshua N. Horn, Ta-Chun Kao, and Alan Grossfield 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G Protein-Coupled Receptors with Enhanced Molecular Dynamics Methods ...... 95 Jennifer M. Johnston and Marta Filizola
Part III GPCR-Focused Rational Design and Mathematical Modeling
7 From Three-Dimensional GPCR Structure to Rational Ligand Discovery ...... 129 Albert J. Kooistra, Rob Leurs, Iwan J.P. de Esch, and Chris de Graaf
vii viii Contents
8 Mathematical Modeling of G Protein-Coupled Receptor Function: What Can We Learn from Empirical and Mechanistic Models? ...... 159 David Roche, Debora Gil, and Jesús Giraldo
Part IV Bioinformatics Tools and Resources for GPCRs
9 GPCR & Company: Databases and Servers for GPCRs and Interacting Partners ...... 185 Noga Kowalsman and Masha Y. Niv 10 Bioinformatics Tools for Predicting GPCR Gene Functions ... 205 Makiko Suwa
Index ...... 225 Part I Progress in Structural Modeling of GPCRs The GPCR Crystallography Boom: Providing an Invaluable Source 1 of Structural Information and Expanding the Scope of Homology Modeling
Stefano Costanzi and Keyun Wang
Abstract G protein-coupled receptors (GPCRs) are integral membrane proteins of high pharmaceutical interest. Until relatively recently, their structures have been particularly elusive, and rhodopsin has been for many years the only member of the superfamily with experimentally elucidated structures. However, a number of recent technical and scientific advancements made the determination of GPCR structures more feasible, thus leading to the solution of the structures of several receptors. Besides providing direct structural information, these experimental GPCR structures also provide templates for the construction of GPCR models. In depth studies have been performed to probe the accuracy of these models, in particular with respect to the interactions with their ligands, and to assess their applicability the rational discovery of GPCR modulators. Given the current state of the art and the pace of the field, the future of GPCR structural studies is likely to be characterized by a landscape populated by an increasingly higher number of experimental and theoretical structures.
Keywords Rhodopsin ¥ Homology ¥ Crystal structures ¥ Docking ¥ 3D Models
Lefkowitz and Brian Kobilka for their pioneering 1.1 Introduction and revolutionary studies that led to our current understanding of the structure and the For the scientific community involved in the functioning of these notable receptor proteins. study of G protein-coupled receptors (GPCRs), GPCRs form a superfamily that encompasses 2012 was a special year, since the Nobel a large number of receptors embedded on the Prize in Chemistry was awarded to Robert lipid bilayer of the plasma membrane of animal cells as well as other eukaryotic cells, including S. Costanzi ()¥K.Wang yeast. In light of their characteristic topology, Department of Chemistry and Center for Behavioral which features one polypeptide chain that spans Neuroscience, American University, Washington, DC ’ 20016, USA seven times the membrane with seven -helices, e-mail: [email protected] they are also known as seven transmembrane
M. Filizola (ed.), G Protein-Coupled Receptors - Modeling and Simulation, Advances 3 in Experimental Medicine and Biology 796, DOI 10.1007/978-94-007-7423-0__1, © Springer ScienceCBusiness Media Dordrecht 2014 4 S. Costanzi and K. Wang
(7TM) receptors (Pierce et al. 2002). These for drug discovery, as they provide a direct insight receptors function as cellular sensors for stimuli into the structure-function relationships of the provided by extracellular first messengers, targets and a platform for the discovery of novel which can be either endogenous or exogenous modulators of their activity (Congreve et al. 2011; substances. The first category encompasses a Congreve and Marshall 2009; Mason et al. 2012). wide variety of signaling molecules secreted by However, being integral membrane proteins, the organism, among which are neurotransmitters GPCRs are not easy targets for structural biology or hormones. Conversely, the second category studies. The first experimentally determined includes environmental substances that are three-dimensional (3D) GPCR structures, first detected by the organism, for instance molecules in the form of a projection map (Schertler et al. responsible for odor or taste. Beyond chemical 1993), and then of 3D X-ray crystallographic first messengers, some GPCRs are also structures (Palczewski et al. 2000), were stimulated by physical messengers such as light determined for rhodopsin, a photoreceptor photons. Upon activation, GPCRs couple with an activated by light photons. Rhodopsin is a rather array of direct intracellular partners, chiefly G peculiar GPCR, which features a covalently proteins and arrestins, and transduce the stimulus bound 11-cis-retinal molecule that acts as an into biochemical signals involving complex inverse agonist. Light photons isomerize the pathways (Pierce et al. 2002). ligand to its all-trans form and consequently As stated by Robert Lefkowitz in an insightful trigger activation of the receptor. However, personal retrospective on his career devoted to the despite these unique characteristics, rhodopsin study of seven transmembrane receptors, GPCR shares its main structural features with the signaling ultimately regulates virtually all known entire GPCR superfamily. Historically, the first physiological pathways (Lefkowitz 2007a). Thus, direct evidence that, beyond rhodopsin, a seven the superfamily of GPCRs provides a wealth of transmembrane domain topology was actually targets for pharmacological intervention. As a shared by all GPCRs emerged in the 1980s, result, the modulation of GPCR signaling is the when the laboratories of Lefkowitz and Strader mechanism of action of many of the currently cloned the “2 adrenergic receptor and detected a approved drugs (Overington et al. 2006). Some striking similarity with the hydropathicity profile of these drugs act directly on the receptors, ei- shown by rhodopsin (Dixon et al. 1986; Dohlman ther by stimulating their activity, as in the case et al. 1987). As Lefkowitz recalled in his of agonists of the opioid receptors used for personal retrospective, prior to these discoveries, the treatment of pain (Spetea and Schmidham- nobody had thought of any possible relation- mer 2012), or by preventing their stimulation ships between rhodopsin and other receptors by natural agonists, as in the case of blockers (Lefkowitz 2007a, b). of the P2Y12 receptor used for the prevention We now know that rhodopsin is just one of of platelet aggregation (Jacobson et al. 2011). the members of the very large superfamily of Conversely, other drugs alter GPCR signaling G protein-coupled receptors (Fredriksson et al. through indirect means, for instance by prevent- 2003) (Gloriam et al. 2007). In humans, the ing the reuptake of the natural agonists of the superfamily counts over 800 distinct members receptors, as in the case of the selective serotonin that, on the basis of sequence similarity, can be reuptake inhibitors (SSRIs) used for the treatment divided into five different families: the glutamate of depression (Donati and Rasenick 2003), or by family (G), also known as class C or family III; altering the activity of the enzymes responsible the rhodopsin family (R), which is by far the for the synthesis of these molecules, as in the case largest and is also known as class A or family I; of the inhibitors of the angiotensin-converting the adhesion family (A) and the secretin family enzyme used for the treatment of hypertension (S), which taken together are also known as (Rosskopf and Michel 2008). class B or family 2; and the frizzled (F)/taste2 Experimentally determined three-dimensional family (Fredriksson et al. 2003; Gloriam et al. structures of drug targets are very powerful tools 2007). 1 The GPCR Crystallography Boom: Providing an Invaluable Source of Structural Information... 5
As reviewed by Costanzi et al., rhodopsin magnetic resonance (NMR) spectroscopy (Park remained for several years the only GPCR with et al. 2012). Given the fast pace of the field, many experimentally elucidated three-dimensional more structures are very likely to be solved in the structures (Costanzi et al. 2009). Beyond pro- near future. viding direct information on the light-activated receptor, these structures also superseded those of rhodopsin as the template of choice for the 1.2 There Once Was Rhodopsin study of other members of the superfamily, in particular those belonging to class A (Audet and Since the 1970s, Rhodopsin has been the ob- Bouvier 2012; Costanzi et al. 2009). However, ject of biochemical investigations intended to in light of the vast pharmaceutical interest unveil its amino acid sequence, facilitated by the revolving around GPCRs, enormous efforts fact that the protein could be obtained in high were put toward the crystallization of other quantities from the retinas of cows (Hargrave members of the superfamily. In 2007, these 2001). The first complete amino acid sequence efforts culminated in a number of breakthroughs of rhodopsin was published in 1982 and 1983 that yielded the solution of the X-ray structure by the laboratories of Ovchinnikov and Har- of the “2-adrenergic receptors (Cherezov et al. grave (Hargrave et al. 1983; Ovchinnikov et al. 2007; Rasmussen et al. 2007; Rosenbaum et al. 1982). The hydropathicity profile inferred from 2007). Thanks to these advances, the field of the sequence indicated a seven transmembrane GPCR crystallography is now in full blossom topology. Moreover, Hargrave and coworkers had and, at the time of a recent survey study previously determined that the C-terminus of the conducted by Jacobson and Costanzi during the receptor was located in the cytoplasm. Hence, summer of 2012, the Protein Data Bank (www. taken together, these data led to the drawing rcsb.org) enlisted structures for a total of 73 of the first bi-dimensional model of rhodopsin structures for 15 distinct receptors (Jacobson as a serpentine originating in the extracellular and Costanzi 2012). At the time of this writing, milieu, spanning seven times the plasma mem- structures for two additional receptors, namely brane with seven ’-helical domains connected by the NTSR1 neurotensin receptor 1 (NTSR1) three extracellular and three intracellular loops, (White et al. 2012) and the protease-activated and terminating in the cytoplasm (Hargrave et al. receptor 1 (PAR1) (Zhang et al. 2012), have 1983)ÐseeFig.1.1. been published. Moreover, another milestone in The first direct experimental insights into the the field of GPCR structural studies has been three-dimensional structure of rhodopsin were reached with the determination of the structure of provided by Schertler and coworkers, who sub- the chemokine CXCR1 receptor through nuclear jected the receptor to electron crystallography
EL1 EL2 EL3 + 3HN Extracellular milieu
Plasma TM2 TM1 TM7 TM6 TM4 TM5 Fig. 1.1 Schematic membrane TM3 representation of Rhodopsin as a membrane spanning serpentine. Abbreviations: TM Cytosol transmembrane domain, IL COO intracellular loop, EL extracellular loop IL1 IL2 IL3 6 S. Costanzi and K. Wang
also part of the termini and almost the entirety of the intracellular and extracellular loops that interconnect the transmembrane domains were visible in the structure. In this regard, the structure unveiled that the second extracellular loop, which connects the fourth and fifth transmembrane domains, in rhodopsin assumes a “-hairpin conformation and obstructs the access to the helical bundle from the extracellular milieu like a closed lid, thus creating a closed binding pocket for retinal. As we discuss later in the article, subsequent X-ray crystallography studies highlighted that the extracellular domains in general, and the second extracellular loop in particular, are highly variable structures that adopt different conformations in different members of the GPCR family.
1.3 2007: The Year of Change
Twenty years after the discovery of the homology between rhodopsin and the “ adrenergic receptors (Dixon et al. 1986; Dohlman et al. 1987) and 7 years after the solution of the crystal structure of rhodopsin (Palczewski et al. 2000), crystal structures were solved for members of the “ Fig. 1.2 The three-dimensional structure of bovine rhodopsin (PDB ID: 1F88), showing the backbone of the adrenergic receptor family (Fig. 1.3) (Cherezov receptor schematically represented as a ribbon et al. 2007; Day et al. 2007; Rasmussen et al. 2007; Rosenbaum et al. 2007). This milestone and obtained a 2D projection map (Schertler et al. ultimately sanctioned the structural homology 1993). On the basis of this map, Baldwin subse- between rhodopsin and the “ adrenergic receptors quently built a molecular model of the receptor and, more in general, the entire superfamily of that allocated the individual TMs to the various GPCRs postulated 20 years earlier by Lefkowitz peaks (Baldwin 1993). and coworkers (Dixon et al. 1986; Dohlman et al. The year 2000 brought another leap forward 1987). Just as the experiments conducted in the for the structural studies of GPCRs, with the 1980s had suggested, the “ adrenergic receptors publication by Palczewski and coworkers of a indeed share a very high degree of structural 2.8 Å resolution X-ray crystal structure of bovine similarity with rhodopsin. In particular, the root rhodopsin in its inactive, dark-adapted state (PDB mean square deviations (RMSD) of the atomic ID: 1F88) (Palczewski et al. 2000). Palczewski’s coordinates between the C’ atoms of the amino structure, which is shown in Fig. 1.2, provided acid residues located in the membrane spanning the first three-dimensional representation of regions of the two receptors amounts to a mere rhodopsin experimentally elucidated at the about 2.7 Å (Cherezov et al. 2007; Rosenbaum atomic level. In particular, the structure revealed et al. 2007). Moreover, a significant overlap be- the fine features of the backbone of the protein as tween the ligands co-crystallized with the two well as the side chains of its amino acid residues. receptors is noticeable (Cherezov et al. 2007; Importantly, not only the ’-helical bundle, but Rosenbaum et al. 2007). 1 The GPCR Crystallography Boom: Providing an Invaluable Source of Structural Information... 7
Fig. 1.3 The three-dimensional structures of (a) the hu- the “2 adrenergic receptor for the first blind assessment man “2-adrenergic receptor (PDB ID: 2RH1) and (b)the of GPCR modeling and docking (Michino et al. 2009). adenosine A2A receptor (PDB ID: 3EML). (c) Our model The figure shows the backbone of the three receptors of the adenosine A2A receptor constructed on the basis of schematically represented as a ribbon
The solution of the structures of the adrenergic However, the various ligands co-crystallized receptors marked the beginning of a new era. with their respective receptors occupy different The technical expedient that made the solution regions of this cavity, more or less overlapping of these structures attainable, which include the with each other. Moreover, in some cases fusion of GPCRs with easily crystallizable pro- different ligands were found to bind to the same teins, the use of antibodies, and the introduc- receptor by exploiting different regions of the tion of stabilizing mutations, provided general binding cavity. This situation is particularly tools for the crystallization of other members evident when comparing the binding of a small of the superfamily as well (Hanson and Stevens molecule antagonist and a much larger cyclic 2009; Salon et al. 2011; Steyaert and Kobilka peptide antagonist to the CXCR4 chemokine 2011;Tate2012). As a result, in 2008 it be- receptor, which show a complete lack of overlap. came the turn of a member of the adenosine For pictorial views of the ligand binding cavities receptor family, namely the A2A receptor, to be of the crystallized GPCRs and a comparison of solved through X-ray crystallography (Fig. 1.3) ligand binding modes, see a recent review by (Jaakola et al. 2008). The solution of the struc- Jacobson and Costanzi (2012). ture of various other receptors belonging to the It is also now evident that the extracellular large rhodopsin family rapidly followed, with a regions assume different conformations in the continuous flow that keeps making the list of various receptor families. This high variability is experimentally solved GPCRs longer (Jacobson also found in the second extracellular loop, which and Costanzi 2012). in most receptors is the extracellular domain Taken together, the currently solved GPCR that shows more contacts with the bound ligand. structures indicate that the members of the Notably, however, all the structures of receptors superfamily, or at least those belonging to naturally activated by peptides that were solved the rhodopsin family, share a high degree to date, including chemokine, opioid, and noci- of structural similarity, which is particularly ceptin receptors, featured a very similar second pronounced within the helical bundle. However, extracellular loop, which adopted a particularly it also highlighted the fact that each receptor has solvent exposed “-hairpin conformation that kept its own peculiarity. Specifically, it is now evident the binding cavity wide open and ready for the that for all the crystallized receptors the ligand binding of large ligands (Granier et al. 2012; binding cavity lies within the heptahelical bundle, Manglik et al. 2012; Thompson et al. 2012;Wu near its opening toward the extracellular space. et al. 2010, 2012). 8 S. Costanzi and K. Wang
Most of the crystal structure of GPCRs have topology (Audet and Bouvier 2012; Okada and been solved solely in the inactive state, which is Palczewski 2001). Conversely, as we mentioned believed to be more rigid and, therefore, more in the introduction, the seminal work conducted amenable to crystallization. However, structures by Lefkowitz, Strader and coworkers in the that reflect activated states have been solved for 1980s revealed beyond doubts that rhodopsin some receptors, including rhodopsin, the “ adren- and GPCRs were indeed members of a single ergic receptors, the adenosine receptors (Choe superfamily of receptors (Dixon et al. 1986; et al. 2011;Parketal.2008; Rasmussen et al. Dohlman et al. 1987). Thus, in light of this proven 2011a, b; Scheerer et al. 2008; Standfuss et al. homology, the structures of rhodopsin published 2011;Xuetal.2011). As reviewed by Audet and by Baldwin on the basis of Schertler’s projection Bouvier, these data provide significant insights maps, the X-ray structure of rhodopsin published into the molecular mechanisms at the basis of the in 2000 and a number of additional subsequently activation of GPCRs (Audet and Bouvier 2012). solved X-ray structures of the same receptor, Most notable is the structure that was solved became the templates of choice for the modeling by Kobilka and coworkers for the “2 adrenergic of GPCRs and maintained this status for several receptor in its activated state, in complex with an years, until the structures of additional members agonist and an activated G protein heterotrimer of the superfamily became available (Audet and (Rasmussen et al. 2011b). This structure, which Bouvier 2012; Costanzi et al. 2009). captures a GPCR in the act of signaling, arguably Currently, due to the explosion of GPCR crys- represents the current pinnacle of the structural tallography, a variety of templates are available studies of GPCRs. for the construction of GPCR homology models. Ultimately, the success of any homology mod- eling effort rests in the use of a suitable tem- 1.4 Assessing the Accuracy plate and the accuracy of the sequence alignment. of Homology Models in Light The presence of high sequence similarity and of the Corresponding Crystal the detection of shared features such as com- Structures mon sequence motifs, putatively shared disulfide bridges and a similar distribution of proline and Besides unveiling the three-dimensional topology glycine residues, which notoriously affects the of the protein for which they have been shape of ’-helices, are criteria commonly used determined, experimental structures also provide for template selection. Beyond the selection of templates for the construction of homology single templates, the use of multiple templates to models of homologous proteins. This application model different regions of a target receptor have is particularly prominent in the field of been proposed as well (Costanzi 2012; Mobarec GPCRs, given the size of the superfamily, the et al. 2009; Worth et al. 2009). For a detailed pharmaceutical relevance of its members and the methodological article on the procedures that paucity of direct structural information. Prior typically followed by our group for the con- to the solution of the structures of rhodopsin, struction of homology modeling of GPCRs, the earlier GPCR modeling studies were based on the reader is referred to an article recently published structure of bacteriorhodopsin, the structure of in the journal “Methods in Molecular Biology” which was first obtained in the mid 1970s through (Costanzi 2012). electron microscopy (Ballesteros and Weinstein For many years a direct and straightforward 1992; Henderson and Unwin 1975; IJzerman analysis of the accuracy of GPCR models was et al. 1992, 1994; Oliveira et al. 1993; Pardo not possible, since experimental structures were et al. 1992). Bacteriorhodopsin, however is not a available only for rhodopsin. However, such GPCR but a proton pump. Moreover, despite comparisons became suddenly possible with the featuring seven transmembrane domains, it availability of multiple templates yielded by the shows a significantly different three-dimensional breakthroughs of 2007. In fact, it became possible 1 The GPCR Crystallography Boom: Providing an Invaluable Source of Structural Information... 9 to build a model of an experimentally-solved that the side chains of the residues lining the bind- receptor on the basis of a different template, to ing cavity be modeled with accuracy (Costanzi then compare the similarity between the model 2010; Mobarec et al. 2009). Thus, whenever the and the experimental structure. rotameric state of a residue thought to be in close In this context, our group studied the feasi- proximity to the ligand cannot be guessed from bility of the construction of molecular models the template, the identification and incorporation of GPCRs in complex with their ligands through of experimentally derived constraints that can homology modeling followed by molecular dock- guide the selection of the conformation of its side ing (Costanzi 2010). In particular, soon after chain may yield a markedly more accurate model. the unveiling of the crystal structure of the “2 Concerning the overall structure of the adrenergic receptor in complex with a potent in- receptor, at the level of the transmembrane verse agonist, a comparison of this experimental domains, which demonstrated a high degree structure with models of the same receptor-ligand of structural conservation, homology models complex obtained through rhodopsin-based ho- usually show a significant structural accuracy, mology modeling followed by molecular docking with average RMSD values below 3 Å. was published (Costanzi 2008). The encouraging Conversely, they are substantially less accurate results of these comparative studies were con- at the level of the extracellular and intracellular firmed by our subsequent success in the first domains, especially the long ones. This is not blind assessment of GPCR modeling and dock- surprising, in light of the low degree of sequence ing, which was conducted in concomitance with and structural conservation shown by these the solution of the crystal structure of the adeno- regions. As mentioned, the second extracellular sine A2A receptors (Fig. 1.3) (Jaakola et al. 2008; loop is often of particular importance for Michino et al. 2009). This controlled assessment ligand recognition, since it lines the interhelical was organized to gage how closely receptor- binding cavity for many GPCRs. The homology ligand complexes obtained through homology modeling of this domain, which assumes modeling and molecular docking would resemble different conformations in different receptor the experimental structure itself. Importantly, the families, is not always feasible. However, the assessment was conducted in a blind manner, i.e. modeling process is assisted by the fact that, the models were submitted to the evaluators be- in most GPCRs, this domain is endowed with fore the unveiling of the experimental structure. a characteristic disulfide bridge that links it to Taken together, our models of the “2 adrenergic TM3. The portion of the loop that shows the most receptor (Costanzi 2008) as well as the models contacts with the ligand is usually the segment of the adenosine A2A receptor submitted to the downstream of this bridge. This is typically a blind assessment (Michino et al. 2009) illustrated relatively short stretch of residues, and hence, is that reasonably accurate models of the complexes suited for de novo modeling, which typically can be constructed. The most accurate models yields more accurate results than homology resulted to be those that incorporated experimen- modeling. Particularly encouraging results in tal derived from site-directed mutagenesis, with this direction have been obtained by Friesner and average root mean square deviations (RMSD) coworkers, who demonstrated the possibility from the atomic coordinates of the experimental of reconstructing through de novo modeling structures ranging from 1.7 Å (for the “2 adren- loops previously expunged from GPCR structures ergic receptor) to 2.8 Å (for the adenosine A2A (Goldfeld et al. 2011, 2012). receptor) for the ligands and 2.7 Å (for the “2 A second blind assessment intended to probe adrenergic receptor) to 3.4 Å (for the adenosine the status of GPCR modeling and docking was A2A receptor) for the residues surrounding the organized in 2010, in coordination with the binding pocket. When the models are built for the experimental elucidation of the structures of study of receptor-ligand interactions of for drug- the dopamine D3 receptor and the chemokine discovery purposes, it is of particular importance CXCR4 receptor through X-ray crystallography 10 S. Costanzi and K. Wang
(Chien et al. 2010; Kufareva et al. 2011; Shoichet and coworkers revealed that a homology Wu et al. 2010). This exercise highlighted model of the dopamine D3 receptor constructed that the accuracy of the predictions depend on the basis of a close homologue, namely the “2 dramatically on the similarity of the target adrenergic receptor, was as effective in a prospec- receptor with the templates. In particular, while tive virtual screening as the crystal structure of the structures of receptors that share a substantial the same receptor (Carlsson et al. 2011). Con- sequence similarity with the template used for versely, a model of the CXCR4 receptor built on their modeling are generally endowed with the basis of a more distant homologue performed a significant level of accuracy, predicting the poorly in comparison to the crystal structure of structures of receptors that are more distant the same receptor (Mysinger et al. 2012). from the available templates is a particularly difficult problem, especially in the absence of experimentally derived constraints (Kufareva 1.5 Summary and Conclusions et al. 2011). The ability of experimental structures and ho- Due to the pharmaceutical interest that revolves mology models of GPCRs to serve as a platform around them, G protein-coupled receptors for virtual screening campaigns intended to iden- (GPCRs) have always been the object of intense tify active compounds out of large databases of efforts toward their structural determination. molecules has been probed through a number of However, until relatively recently their structures controlled experiments that measured the ability proved to be elusive. In this context, rhodopsin of these structures to retrieve a pool of known has been for years the only GPCR with an ligands dispersed within a larger set of inactive experimentally solved structure and has been compounds, a non-comprehensive list of which heavily used to construct models of the other is provided in the reference (Cavasotto 2011; members of the superfamily through a technique Cavasotto et al. 2008; Katritch et al. 2010; Phatak known as homology modeling. Thanks to a et al. 2010; Vilar et al. 2010, 2011)Ðfora number of technical and scientific advancements, review of the application of virtual screening to now the solution of GPCR structures through X- GPCRs, see Chapter 18 of the book “G Protein- ray crystallography and, more recently, NMR is Coupled Receptors: From Structure to Function” more feasible. Thus, structures of several GPCRs published by the Royal Society of Chemistry have been recently disclosed. These structures, (Costanzi 2011). These studies highlighted that besides providing direct structural information accurate models are indeed capable of priori- on the receptor in question, also provide a tizing binders versus non-binders, albeit not as larger basis of templates for the construction well as crystal structures. Not surprisingly, the of more accurate models of those receptors that models tended to perform according to their level have not yet been experimentally solved yet. of structural accuracy. For instance, a study con- The availability of multiple templates has also ducted by Katrich and coworkers on the basis offered the possibility of performing accurate of the A2A models submitted to the blind as- studies intended to probe the strengths and the sessment revealed very good performance levels limits of homology modeling and molecular for the three most structurally accurate models docking applied to homology models. Moreover, (Costanzi, Katritch/Abagyan and Lam/Abagyan) controlled virtual screening experiments have (Katritch et al. 2010). Conversely, less accurate been conducted to investigate the extent of the models performed very poorly. applicability of homology models to molecular Not surprisingly, virtual screening campaigns recognition campaigns and to compare it to based on homology models seem to perform that shown by experimental structures. The better when the models are built on the ba- current pace of structural biology and the sis of closely rather than distantly related tem- rapid advancements of computational chemistry plates. Accordingly, a recent study conducted by suggest a future landscape populated by 1 The GPCR Crystallography Boom: Providing an Invaluable Source of Structural Information... 11 increasingly more numerous experimental and Costanzi S (2010) Modeling G protein-coupled receptors: theoretical GPCR structures. The former and the a concrete possibility. Chim Oggi 28:26Ð31 latter, especially when supported by experimental Costanzi S (2011) Chapter 18. Structure-based virtual screening for ligands of G protein-coupled receptors. data and built on the basis of closely related In: Giraldo J, Pin J-P (eds) G protein-coupled recep- templates, will provide a solid basis for the tors: from structure to function. The Royal Society of rational discovery of novel modulators of GPCR Chemistry, London and Cambridge, pp 359Ð374 activity. Costanzi S (2012) Homology modeling of class a G protein-coupled receptors. Methods Mol Biol 857:259Ð279 Costanzi S, Siegel J, Tikhonova I, Jacobson K (2009) References Rhodopsin and the others: a historical perspective on structural studies of G protein-coupled receptors. Curr Pharm Des 15:3994Ð4002 Audet M, Bouvier M (2012) Restructuring G-protein- Day P, Rasmussen S, Parnot C, Fung J, Masood A, coupled receptor activation. Cell 151:14Ð23 Kobilka T, Yao X, Choi H, Weis W, Rohrer D et Baldwin J (1993) The probable arrangement of the helices al (2007) A monoclonal antibody for G protein- in G protein-coupled receptors. EMBO J 12:1693Ð coupled receptor crystallography. Nat Methods 4: 1703 927Ð929 Ballesteros JA, Weinstein H (1992) Analysis and refine- Dixon R, Kobilka B, Strader D, Benovic J, Dohlman H, ment of criteria for predicting the structure and relative Frielle T, Bolanowski M, Bennett C, Rands E, Diehl orientations of transmembranal helical domains. Bio- R et al (1986) Cloning of the gene and cDNA for phys J 62:107Ð109 mammalian beta-adrenergic receptor and homology Carlsson J, Coleman RG, Setola V, Irwin JJ, Fan H, with rhodopsin. Nature 321:75Ð79 Schlessinger A, Sali A, Roth BL, Shoichet BK (2011) Dohlman H, Bouvier M, Benovic J, Caron M, Lefkowitz Ligand discovery from a dopamine D(3) receptor ho- R (1987) The multiple membrane spanning topogra- mology model and crystal structure. Nat Chem Biol phy of the beta 2-adrenergic receptor. Localization 7:769Ð778 of the sites of binding, glycosylation, and regulatory Cavasotto CN (2011) Homology models in docking phosphorylation by limited proteolysis. J Biol Chem and high-throughput docking. Curr Top Med Chem 262:14282Ð14288 11:1528Ð1534 Donati RJ, Rasenick MM (2003) G protein signaling and Cavasotto CN, Orry AJ, Murgolo NJ, Czarniecki MF, the molecular basis of antidepressant action. Life Sci Kocsi SA, Hawes BE, O’Neill KA, Hine H, Burton 73:1Ð17 MS, Voigt JH et al (2008) Discovery of novel chemo- Fredriksson R, Lagerström M, Lundin L, Schiöth H types to a G-protein-coupled receptor through ligand- (2003) The G-protein-coupled receptors in the human steered homology modeling and structure-based vir- genome form five main families. Phylogenetic analy- tual screening. J Med Chem 51:581Ð588 sis, paralogon groups, and fingerprints. Mol Pharmacol Cherezov V, Rosenbaum D, Hanson M, Rasmussen S, 63:1256Ð1272 Thian F, Kobilka T, Choi H, Kuhn P, Weis W, Kobilka Gloriam D, Fredriksson R, Schiöth H (2007) The G B et al (2007) High-resolution crystal structure of an protein-coupled receptor subset of the rat genome. engineered human beta2-adrenergic G protein-coupled BMC Genomics 8:338 receptor. Science 318:1258Ð1265 Goldfeld DA, Zhu K, Beuming T, Friesner RA (2011) ChienEY,LiuW,ZhaoQ,KatritchV,HanGW,Hanson Successful prediction of the intra- and extracellular MA, Shi L, Newman AH, Javitch JA, Cherezov V et al loops of four G-protein-coupled receptors. Proc Natl (2010) Structure of the human dopamine D3 receptor Acad Sci U S A 108:8275Ð8280 in complex with a D2/D3 selective antagonist. Science Goldfeld DA, Zhu K, Beuming T, Friesner RA (2012) 330:1091Ð1095 Loop prediction for a GPCR homology model: algo- Choe HW, Kim YJ, Park JH, Morizumi T, Pai EF, Krauss rithms and results. Proteins 81:214Ð228 N, Hofmann KP, Scheerer P, Ernst OP (2011) Crystal Granier S, Manglik A, Kruse AC, Kobilka TS, Thian FS, structure of metarhodopsin II. Nature 471:651Ð655 Weis WI, Kobilka BK (2012) Structure of the delta- Congreve M, Marshall F (2009) The impact of GPCR opioid receptor bound to naltrindole. Nature 485:400Ð structures on pharmacology and structure-based drug 404 design. Br J Pharmacol 159:986Ð996 Hanson MA, Stevens RC (2009) Discovery of new GPCR Congreve M, Langmead CJ, Mason JS, Marshall FH biology: one receptor structure at a time. Structure (2011) Progress in structure based drug design for G 17:8Ð14 protein-coupled receptors. J Med Chem 54:4283Ð4311 Hargrave PA (2001) Rhodopsin structure, function, and to- Costanzi S (2008) On the applicability of GPCR ho- pography the Friedenwald lecture. Invest Ophthalmol mology models to computer-aided drug discovery: a Vis Sci 42:3Ð9 comparison between in silico and crystal structures Hargrave P, McDowell J, Curtis D, Wang J, Juszczak E, of the beta2-adrenergic receptor. J Med Chem 51: Fong S, Rao J, Argos P (1983) The structure of bovine 2907Ð2914 rhodopsin. Biophys Struct Mech 9:235Ð244 12 S. Costanzi and K. Wang
Henderson R, Unwin PN (1975) Three-dimensional Okada T, Palczewski K (2001) Crystal structure of model of purple membrane obtained by electron mi- rhodopsin: implications for vision and beyond. Curr croscopy. Nature 257:28Ð32 Opin Struct Biol 11:420Ð426 IJzerman A, Van Galen P, Jacobson K (1992) Molecular Oliveira L, Paiva ACM, Vriend G (1993) A common modeling of adenosine receptors. I. The ligand binding motif in G-protein-coupled 7 transmembrane helix site on the A1 receptor. Drug Des Discov 9:49Ð67 receptors. J Comput Aided Mol Des 7:649Ð658 IJzerman AP, van der Wenden EM, Van Galen PJ, Ja- Ovchinnikov Y, Abdulaev N, Feigina M, Artamonov I, cobson KA (1994) Molecular modeling of adenosine Zolotarev A, Kostina M, Bogachuk A, Miroshnikov A, receptors. The ligand binding site on the rat adenosine Martinov V, Kudelin A (1982) The complete amino- A2A receptor. Eur J Pharmacol 268:95Ð104 acid-sequence of visual rhodopsin. Bioorganicheskaya Jaakola V, Griffith M, Hanson M, Cherezov V, Chien Khimiya 8:1011Ð1014 E, Lane J, Ijzerman A, Stevens R (2008) The 2.6 Overington JP, Al-Lazikani B, Hopkins AL (2006) How angstrom crystal structure of a human A2A adeno- many drug targets are there? Nat Rev Drug Discov sine receptor bound to an antagonist. Science 322: 5:993Ð996 1211Ð1217 Palczewski K, Kumasaka T, Hori T, Behnke C, Motoshima Jacobson KA, Costanzi S (2012) New insights for drug H, Fox B, Le Trong I, Teller D, Okada T, Stenkamp R design from the x-ray crystallographic structures et al (2000) Crystal structure of rhodopsin: a G protein- of g-protein-coupled receptors. Mol Pharmacol 82: coupled receptor. Science 289:739Ð745 361Ð371 Pardo L, Ballesteros JA, Osman R, Weinstein H (1992) Jacobson KA, Deflorian F, Mishra S, Costanzi S (2011) On the use of the transmembrane domain of bacte- Pharmacochemistry of the platelet purinergic recep- riorhodopsin as a template for modeling the three- tors. Purinergic Signal 7:305Ð324 dimensional structure of guanine nucleotide-binding Katritch V, Rueda M, Lam P, Yeager M, Abagyan R regulatory protein-coupled receptors. Proc Natl Acad (2010) GPCR 3D homology models for ligand screen- Sci U S A 89:4009Ð4012 ing: lessons learned from blind predictions of adeno- Park JH, Scheerer P, Hofmann KP, Choe HW, Ernst OP sine A2a receptor complex. Proteins 78:197Ð211 (2008) Crystal structure of the ligand-free G-protein- Kufareva I, Rueda M, Katritch V, Stevens RC, Abagyan coupled receptor opsin. Nature 454:183Ð187 R (2011) Status of GPCR modeling and docking as Park SH, Das BB, Casagrande F, Tian Y, Nothnagel reflected by community-wide GPCR Dock 2010 as- HJ, Chu M, Kiefer H, Maier K, De Angelis AA, sessment. Structure 19:1108Ð1126 Marassi FM et al (2012) Structure of the chemokine Lefkowitz RJ (2007a) Seven transmembrane receptors: receptor CXCR1 in phospholipid bilayers. Nature 491: a brief personal retrospective. Biochim Biophys Acta 779Ð783 1768:748Ð755 Phatak SS, Gatica EA, Cavasotto CN (2010) Ligand- Lefkowitz RJ (2007b) Seven transmembrane receptors: steered modeling and docking: a benchmarking study something old, something new. Acta Physiol (Oxf) in class a G-protein-coupled receptors. J Chem Inf 190:9Ð19 Model 50:2119Ð2128 Manglik A, Kruse AC, Kobilka TS, Thian FS, Mathiesen Pierce K, Premont R, Lefkowitz R (2002) Seven- JM, Sunahara RK, Pardo L, Weis WI, Kobilka BK, transmembrane receptors. Nat Rev Mol Cell Biol Granier S (2012) Crystal structure of the mu-opioid 3:639Ð650 receptor bound to a morphinan antagonist. Nature Rasmussen S, Choi H, Rosenbaum D, Kobilka T, Thian F, 485:321Ð326 Edwards P, Burghammer M, Ratnala V, Sanishvili R, Mason JS, Bortolato A, Congreve M, Marshall FH (2012) Fischetti R et al (2007) Crystal structure of the human New insights from structural biology into the drugga- beta2 adrenergic G-protein-coupled receptor. Nature bility of G protein-coupled receptors. Trends Pharma- 450:383Ð387 col Sci 33:249Ð260 Rasmussen SG, Choi HJ, Fung JJ, Pardon E, Casarosa Michino M, Abola E, 2008 Participants G, Brooks Cr, P, Chae PS, Devree BT, Rosenbaum DM, Thian FS, Dixon J, Moult J, Stevens R (2009) Community-wide Kobilka TS et al (2011a) Structure of a nanobody- assessment of GPCR structure modelling and ligand stabilized active state of the beta(2) adrenoceptor. docking: GPCR Dock 2008. Nat Rev Drug Discov Nature 469:175Ð180 8:455Ð463 Rasmussen SG, DeVree BT, Zou Y, Kruse AC, Chung KY, Mobarec J, Sanchez R, Filizola M (2009) Modern homol- Kobilka TS, Thian FS, Chae PS, Pardon E, Calinski D ogy modeling of G-protein coupled receptors: which et al (2011b) Crystal structure of the beta2 adrenergic structural template to use? J Med Chem 52:5207Ð5216 receptor-Gs protein complex. Nature 477:549Ð555 Mysinger MM, Weiss DR, Ziarek JJ, Gravel S, Doak Rosenbaum D, Cherezov V, Hanson M, Rasmussen S, AK, Karpiak J, Heveker N, Shoichet BK, Volk- Thian F, Kobilka T, Choi H, Yao X, Weis W, Stevens R man BF (2012) Structure-based ligand discovery for et al (2007) GPCR engineering yields high-resolution the protein-protein interface of chemokine receptor structural insights into beta2-adrenergic receptor func- CXCR4. Proc Natl Acad Sci U S A 109:5517Ð5522 tion. Science 318:1266Ð1273 1 The GPCR Crystallography Boom: Providing an Invaluable Source of Structural Information... 13
Rosskopf D, Michel MC (2008) Pharmacogenomics of for ligands of G protein-coupled receptors: Not only G protein-coupled receptor ligands in cardiovascular crystal structures but also in silico models. J Mol medicine. Pharmacol Rev 60:513Ð535 Graph Model. doi:10.1016/j.jmgm.2010.1011.1005 Salon JA, Lodowski DT, Palczewski K (2011) The signif- Vilar S, Karpiak J, Berk B, Costanzi S (2011) In silico icance of G protein-coupled receptor crystallography analysis of the binding of agonists and blockers to for drug discovery. Pharmacol Rev 63:901Ð937 the beta2-adrenergic receptor. J Mol Graph Model Scheerer P, Park JH, Hildebrand PW, Kim YJ, Krauss 29:809Ð817 N, Choe HW, Hofmann KP, Ernst OP (2008) Crystal White JF, Noinaj N, Shibata Y, Love J, Kloss B, Xu F, structure of opsin in its G-protein-interacting confor- Gvozdenovic-Jeremic J, Shah P, Shiloach J, Tate CG mation. Nature 455:497Ð502 et al (2012) Structure of the agonist-bound neurotensin Schertler G, Villa C, Henderson R (1993) Projection receptor. Nature 490:508Ð513 structure of rhodopsin. Nature 362:770Ð772 Worth C, Kleinau G, Krause G (2009) Comparative se- Spetea M, Schmidhammer H (2012) Recent advances in quence and structural analyses of G-protein-coupled the development of 14-alkoxy substituted morphinans receptor crystal structures and implications for molec- as potent and safer opioid analgesics. Curr Med Chem ular models. PLoS One 4:e7011 19:2442Ð2457 Wu B, Chien EY, Mol CD, Fenalti G, Liu W, Katritch Standfuss J, Edwards PC, D’Antona A, Fransen M, Xie G, V, Abagyan R, Brooun A, Wells P, Bi FC et al Oprian DD, Schertler GF (2011) The structural basis (2010) Structures of the CXCR4 chemokine GPCR of agonist-induced activation in constitutively active with small-molecule and cyclic peptide antagonists. rhodopsin. Nature 471:656Ð660 Science 330:1066Ð1071 Steyaert J, Kobilka BK (2011) Nanobody stabilization Wu H, Wacker D, Mileni M, Katritch V, Han GW, Vardy of G protein-coupled receptor conformational states. E, Liu W, Thompson AA, Huang XP, Carroll FI et al Curr Opin Struct Biol 21:567Ð572 (2012) Structure of the human kappa-opioid receptor Tate CG (2012) A crystal clear solution for determining G- in complex with JDTic. Nature 485:327Ð332 protein-coupled receptor structures. Trends Biochem Xu F, Wu H, Katritch V, Han GW, Jacobson KA, Gao Sci 37:343Ð352 ZG, Cherezov V, Stevens RC (2011) Structure of an Thompson AA, Liu W, Chun E, Katritch V, Wu H, agonist-bound human A2A adenosine receptor. Sci- Vardy E, Huang XP, Trapella C, Guerrini R, Calo G ence 332:322Ð327 et al (2012) Structure of the nociceptin/orphanin FQ Zhang C, Srinivasan Y, Arlow DH, Fung JJ, Palmer receptor in complex with a peptide mimetic. Nature D, Zheng Y, Green HF, Pandey A, Dror RO, Shaw 485:395Ð399 DE et al (2012) High-resolution crystal structure Vilar S, Ferino G, Phatak SS, Berk B, Cavasotto CN, of human protease-activated receptor 1. Nature 492: Costanzi S (2010) Docking-based virtual screening 387Ð392 Modeling of G Protein-Coupled Receptors Using Crystal Structures: 2 From Monomers to Signaling Complexes
Angel Gonzalez, Arnau Cordomí, Minos Matsoukas, Julian Zachmann, and Leonardo Pardo
Abstract G proteinÐcoupled receptors constitute a large and functionally diverse family of transmembrane proteins. They are fundamental in the transfer of extracellular stimuli to intracellular signaling pathways and are among the most targeted proteins in drug discovery. Recent advances in crys- tallization methods have permitted to resolve the molecular structure of several members of the family. This chapter focuses on the impact of these structures in the use of homology modeling techniques for building three- dimensional models of homologous G proteinÐcoupled receptors, higher order oligomers, and their complexes with ligands and signaling proteins.
Keywords Transmembrane helices ¥ Homology modeling ¥ Conformation ¥ Ligand binding ¥ Activation mechanism
organism, including fungi and plants. They 2.1 Introduction are highly diversified in mammalian genomes with current estimates of about 1,000 genes Membrane receptors coupled to guanine (2Ð3 % of the human proteome) (Fredriksson and nucleotide-binding proteins (commonly known Schioth 2005). GPCRs transduce sensory signals as G protein-coupled receptors, GPCRs) of external origin such as odors, pheromones, comprise one of the widest and most adaptable or tastes; and endogenous signals such as families of cellular sensors, as they are able neurotransmitters, (neuro)peptides, proteases, to mediate a wide range of transmembrane glycoprotein hormones, purine ligands and signal transduction processes (Kristiansen 2004). ions, among others. The response is operated GPCRs are present in almost every eukaryotic through second messenger cascades controlled by different heterotrimeric guanine nucleotide- A. Gonzalez ¥ A. Cordomí ¥ M. Matsoukas ¥ J. Zachmann binding proteins (G-proteins) coupled at their ¥L.Pardo() intracellular regions (Oldham and Hamm 2008). Laboratori de Medicina Computacional, Unitat de Due to their relevance to cellular physiology Bioestadística, Facultat de Medicina, Universitat Autònoma de Barcelona, 08193 Bellaterra, Spain (Smit et al. 2007) and their accessibility from the e-mail: [email protected] extracellular environment, membrane proteins
M. Filizola (ed.), G Protein-Coupled Receptors - Modeling and Simulation, Advances 15 in Experimental Medicine and Biology 796, DOI 10.1007/978-94-007-7423-0__2, © Springer ScienceCBusiness Media Dordrecht 2014 16 A. Gonzalez et al. represent a significant portion of therapeutic drug set of membrane proteins maintains a strong targets (Arinaminpathy et al. 2009; Imming et al. conservation of the TM structure even at low 2006). sequence identity (<20 %) (Olivella et al. 2013). The GPCR family is not an exception. All crystal structures preserve analogous 2.2 The Structure of G secondary/tertiary structures at the seven-helical- Protein-Coupled Receptors bundle domain (Fig. 2.1) despite the percentage of sequence identity in the TM segments is very Significant advances in crystallization of GPCRs low (Mobarec et al. 2009; Gonzalez et al. 2012). (Day et al. 2007; Serrano-Vega et al. 2008)have Structure conservation in the GPCR family is permitted to elucidate the crystal structures of associated, in contrast to other proteins, to the many receptors (Table 2.1) (see (Katritch et al. presence of at least one highly conserved amino 2012, 2013) for recent reviews). All these struc- acid in each helix (Mirzadegan et al. 2003): N tures share the common architecture of seven in TM1 (present in 98 % of the sequences), D plasma membrane-spanning (or transmembrane) in TM2 (93 %), R in TM3 (95 %), W in TM4 domains (TMs, which also terms this family of (96%),PinTM5(76%),PinTM6(98%), proteins as 7TM receptors) connected to each and P in TM7 (93 %). This feature was used other with three extracellular (ECL) and three by Ballesteros and Weinstein (1995) to define intracellular loops (ICL), a disulphide bridge be- a general numbering scheme consisting of two tween ECL 2 and TM 3, and a cytoplasmic CÐ numbers: the first (1 through 7) corresponds to terminus containing an ’-helix (Hx8) parallel to the helix in which the amino acid of interest the cell membrane. In addition, GPCRs contain is located; the second indicates its position an extracellular N-terminal region (N-terminus) relative to the most conserved residue in the and an intracellular C-terminal tail (C-tail). helix, arbitrarily assigned to 50. Significantly, the position of these highly conserved amino acids in each helix is the same in the superimposition 2.3 Homology Modeling of G of the currently available crystal structures Protein-Coupled Receptors (Fig. 2.1). This finding validates the use of these amino acids as reference points in TM sequence Because of the limited high-resolution struc- alignments (instead of the common procedure tural information on GPCRs, computational of using substitution matrices and fast sequence techniques to predict their structure from the similarity search algorithms) (see red box in amino acid sequence are a valuable tool (Pieper Fig. 2.2), and in the construction of homology et al. 2013). Recently, de novo techniques using models of GPCRs with unknown structure (de la evolutionary constraints have been applied to Fuente et al. 2010; Blattermann et al. 2012). predict 3D structures of TM proteins (Hopf et al. 2012). However, homology models of proteins with unknown experimental structure can also 2.4 The Conformation be built from homologous proteins of known of Transmembrane Helices structure and similar sequence (templates). This in G Protein-Coupled method is based on the fact that in homologous Receptors proteins, structure is more conserved than sequence. Thus, in general, homologous proteins Figure 2.1 shows the superimposition of the with a sequence identity above 35 % have a TM domain of representative crystal structures. similar 3D structure (Krissinel and Henrick Clearly, the structure of the cytoplasmic part is 2004). Because membrane proteins contain highly conserved. This structural conservation only two types of folds in their TM domains, correlates with the fact that most conserved ’-helix bundles and “-barrels, a significant residues are clustered in the central and 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 17
Table 2.1 Crystal structures of G protein coupled receptors Receptor Ligand PDB Reference Rhodopsin Bovine Rhodopsin (bRho) 11-cis retinal 1F88, 1GZM Palczewski et al. (2000) and Li et al. (2004) Squid Rhodopsin (sRho) 11-cis retinal 2Z73 Murakami and Kouyama (2008) Opsin 3CAP Park et al. (2008) Opsin C transducin peptide 3DQB Scheerer et al. (2008) Constitutively active rhodopsin 2X72 Standfuss et al. (2011) Metarhodopsin II 11-trans retinal 3PXO Choe et al. (2011) Metarhodopsin II C transducin peptide 11-trans retinal 3PQR Choe et al. (2011) Biogenic amine receptors
“1-adrenergic (“1AR) Cyanopindolol 2VT4 Warne et al. (2008) “1AR Isoprenaline 2Y03 Warne et al. (2011) “1AR homo-oligomer 4GPO Huang et al. (2013) “2-adrenergic (“2AR) Carazolol 2RH1 Cherezov et al. (2007)and Rosenbaum et al. (2007)
“2AR C nanobody BI-167107 3POG Rasmussen et al. (2011a) “2AR C Gs BI-167107 3SN6 Rasmussen et al. (2011b) Dopamine D3 (D3R) Eticlopride 3PBL Chien et al. (2010)
Histamine H1 (H1R) Doxepin 3RZE Shimamura et al. (2011) Muscarinic M2 (M2R) 3-quinuclidinyl-benzilate 3UON Haga et al. (2012) Muscarinic M3 (M3R) Tiotropium 4DAJ Kruse et al. (2012) Serotonin 5HT1B (5HT1BR) Ergotamine 4IAR Wang et al. (2013a) Serotonin 5HT1B (5HT2BR) Ergotamine 4IB4 Wacker et al. (2013) Nucleotide
Adenosine A2A (A2AR) ZM241385 3EML Jaakola et al. (2008) A2AR UK-432097 3QAK Xu et al. (2011) C A2AR C Na ZM241385 4EIY Liu et al. (2012) Peptide receptors CXCR4 CVX15 3OE0 Wu et al. (2010) CXCR4 IT1t 3ODU Wu et al. (2010) -opioid ( -OR) “-funaltrexamine 4DKL Manglik et al. (2012) ›-opioid (›-OR) JDTic 4DJH Wu et al. (2012) •-opioid (•-OR) Naltrindole 4EJ4 Granier et al. (2012) Nociceptin/orphanin FQ C-24 4EA3 Thompson et al. (2012) Neurotensin1 (NTSR1) Neurotensin (8Ð13) 4GRV White et al. (2012) Protease-activated receptor 1 (PAR1) Vorapaxar 3VW7 Zhang et al. (2012) Lipid
Sphingosine S1P (S1P1R) ML056 3V2Y Hanson et al. (2012) Frizzled (class F) Smoothened (SMO) LY2940680 4JKV Wang et al. (2013b) intracellular regions of the receptor (Mirzadegan that GPCRs, during their evolution, have evolved et al. 2003). In contrast, there is a low degree of to adjust the structural characteristics of their sequence conservation among different GPCRs cognate ligands, by customizing a preserved at their extracellular domains. Accordingly, the scaffold (7TM receptors) through conformational structure of the extracellular part of TM helices plasticity (Deupi et al. 2007). We use this is more divergent. We have previously suggested term to describe the structural differences 18 A. Gonzalez et al.
Fig. 2.1 Comparison of the TM bundle of GPCRs. The structures of bRho (PDB code 1U19), “2R(2RH1),D3R(3PBL), H1R (3RZE), M2R, 5HT1bR(4IAR),A2AR (4EIY), CXCR4 (3ODU), OR (4DKL), NTSR1 (4GRV), PAR1 (3VW7), and S1P1R (3V2Y) are shown. The colour code of the helices is TMs 1 in white,2inyellow,3inred, 4ingrey,5ingreen,6in dark blue,and7inlight blue. The highly conserved N1.50 (in white), D2.50 (in yellow), R3.50 (in red), W4.50 (in grey), P5.50 (in green), P6.50 (in dark blue), and P7.50 (in light blue) are shown as spheres
Fig. 2.2 Sequence alignments of TMs 1–7 of GPCRs with known structures. The highly conserved amino acids in each helix, used as reference points in TM sequence alignments are boxed in red among different receptor subfamilies within the modulating the polytopic membrane protein extracellular side, near the binding site crevices, architecture (Riek et al. 2001). Deviations from responsible for recognition and selectivity of the regular ’-helical context have been associated diverse ligands. to prolines (Von Heijne 1991), glycines (Senes Moreover, comparison among the crystal et al. 2000), serines and threonines (Deupi et al. structures of GPCRs revealed backbone 2004, 2010), or to the insertion or deletions anomalies, in the form of kinks and bulges, (indels) of residues within the TMs (Deville in the majority of TM helices. These non- et al. 2009). Moreover, specific intra- and canonical elements are frequent in TM proteins, interhelical interactions involving polar side 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 19
Fig. 2.3 Comparison of individual TM helices in representative structures of the inactive state of GPCRs. The most conserved domain of the helices is superimposed. The crystal structures of GPCRs revealed backbone anomalies, in the form of kinks and bulges, in the majority of TM helices
chains, backbone carbonyls, disulphide bridges (Fig. 2.3). It appears shifted towards the central and, in some cases, structural water molecules axis of the receptor in Rho, A2AR, CXCR4, embedded in the TM bundle (Pardo et al. 2007) opioid receptors, NTSR1, and PAR1. The major also cause these distortions. Here we present a displacement of TM1 corresponds to CXCR4 due detailed analysis of these distortions and their to the formation of a disulphide bond between implication in modeling other GPCRs. Figure 2.3 the C28 in the N-terminal region and C2747.25 shows the superimposition of the more conserved in TM7 (Wu et al. 2010). In contrast, TM1 is part of individual TM helices in representative pointing outside of the bundle in biogenic amine structures of the inactive state of GPCRs. receptors. The highly conserved N1.50 (97 % in class A non-olfactory GPCRs) most probably influences the packing of the TM bundle (see 2.4.1 Transmembrane Helix 1 Fig. 2.3) since its N•2-H atoms act as hydrogen bond donors in the interaction with the backbone The extracellular region of TM1 displays a bend- carbonyl oxygen of residues at positions 1.46 and ing propensity in some of the crystal structures 7.46, linking TMs 1 and 7. Moreover, O•1 of 20 A. Gonzalez et al.
N1.50 interacts with the highly conserved D2.50 helix at the extracellular part ( 3.6 residues per (in 92 % of the sequences), via a conserved turn). This conformation of TM2 moves its extra- water molecule, linking TMs 1 and 2. Previous cellular part away from the TM bundle, relative to studies have shown that interactions involving the other structures, and modifies the orientation a polar Asn side chain provide a strong ther- of the side chains at the extracellular side. In modynamic driving force for membrane helix order to translate these structural observations association (Choma et al. 2000). into the sequence space, a two-residue gap in the sequences of S1P1R, CXCR4, opioid receptors and PAR1, or one-residue gap in the sequences 2.4.2 Transmembrane Helix 2 – of bRho, biogenic amine receptors and NTSR1, Extracellular Loop 1 relative to sRho, must be inserted (Gonzalez et al. 2012) (Fig. 2.2). The shape of TM2 at the extracellular part, which Importantly, the conserved Trp residue bends towards TM1 and leans away from TM3, in ECL1, part of the (W/F) (F/L)G motif is similar in all structures (Fig. 2.3); despite previously identified (Klco et al. 2006), points the amino acid sequence is strongly divergent toward the helical bundle, between TMs 2 and with, for instance, Pro residues at either po- 3, in the crystal structures with the exception of sition 2.58 (CXCR4, opioid receptors, PAR1), S1P1R (not shown). 2.59 (biogenic amine receptors, NTSR1) or 2.60 (sRho). The only exception is TM2 of A2AR, which contains Pro at position 2.59 but kinks 2.4.3 Transmembrane Helix 3 towards TM3 due to the Cys-bridge between ECL1 and ECL2 exclusive of this family (not TM3 is the longest and most tilted helix in the shown); and TM2 of S1P1R that lacks Pro in receptor structures. No major deviations among the helix (see below). Contrarily to S1P1R, the structures are observed with the exception of also Pro-lacking bRho and muscarinic receptors A2AR, due to the Cys-bridge between ECL1 and possess TM2 structurally similar to the other ECL2 exclusive of this family (see above). The Pro-containing receptors due to the presence of highly conserved C3.25 forms a disulphide bridge the GGxTT motif in bRho and N2.59 in mus- with a Cys residue located at various positions carinic receptors that hydrogen bonds the back- in ECL2. The cytoplasmic side of TM3 contains bone carbonyl at position 2.55 (Gonzalez et al. the highly conserved (D/E)R3.50(Y/W) motif in- 2012). Interestingly, the superimposition of struc- volved in receptor activation (see below). Impor- tures reveals that the highly D2.50 and the Pro tantly, the central location of TM3 within the residue located at position 2.58, 2.59 or 2.60 TM bundle allows the helix to interact with the are perfectly overlaid (Gonzalez et al. 2012). ligand at the extracellular part and with the G Thus, the backbone helical conformation of the protein at the intracellular part (Venkatakrishnan amino acids located between these two residues et al. 2013). must differ. In this region, TM2 of CXCR4, opioid receptors, and PAR1 adopts a 310 or tight turn ( 3.0 residues per turn), TM2 of biogenic 2.4.4 Transmembrane Helix 4 amine receptors and NTSR1 adopts a -bulge or wide turn ( 4.8 residues per turn), and TM2 of TM4, the shortest helix, is almost perpendicular sRho presents an extreme distortion ( 9 residues to the membrane. However, significant structural per turn) characterized by a cis P2.60 backbone divergences at the extracellular part of TM 4 conformation, which is stabilized by two water are found among structures, which may be re- molecules (Gonzalez et al. 2012). In contrast to lated to the structural requirements necessary these receptors, S1P1R contains a canonical ’- to accommodate the diverse ECL2 architectures 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 21
5.50 5.38 (see below). For instance, in contrast to TM4 relative to P )ofA2AR with F/Y of the of other biogenic amine receptors, TM4 of mus- P5.50-containing structures (i-12 relative to P5.50) carinic receptors bends towards outside of the and F/Y/W5.39 of the P5.50-lacking structures bundle, away from TM3, due to the hydrogen (Fig. 2.2). bond interactions between the side chain of Q4.65 and the backbone carbonyl oxygen at position 4.62. Significantly, the shape of TM4 at the ex- 2.4.6 Transmembrane Helix 6 tracellular part, in peptide receptors (in which ECL2 is formed by two “-strands, see below), TM6 presents the most pronounced kink in the bends towards TM3. In CXCR4, TM4 is longer TM bundle. This severe distortion is energetically and substantially deviate from the conformation stabilized through two structural and functional observed in other peptide receptors. elements. First, P6.50 of the highly conserved CWxP6.50(Y/F) motif introduces a flexible point in TM6 facilitating this extreme conformation. 2.4.5 Transmembrane Helix 5 Second, a structural water molecule located in a small cavity between TMs 6 and 7 help to main- P5.50 (conserved in 76 % of the rhodopsin-like tain the Pro induced distortion. This water acts as sequences) induces a local opening of TM5, at a hydrogen-bond acceptor in the interaction with the 5.43Ð5.48 turn (Pro-unwinding), in all crystal the backbone N-H amide at position 6.51, and as structures except S1P1R (see below), which has a hydrogen bond donor in the interactions with been proposed to be involved in the mechanism of the backbone carbonyl at position 6.47 and 7.38. ligand-induced receptor activation (Sansuk et al. Thus, in addition to stabilizing the kink of TM6, 2011; Rasmussen et al. 2011a). Thus, P5.50 trig- this water molecule links TMs 6 and 7. gers a -bulge or wide turn conformation ( 5 residues per turn). However, A2AR displays an extended opening of TM5 from positions 5.35Ð 2.4.7 Transmembrane Helix 7 5.48, in contrast to other P5.50-containing struc- tures in which the opening of the helix is re- TM7 start at different position among receptors. stricted to the 5.43Ð5.48 range of amino acids. TM7 in CXCR4 is two helical turns longer than Moreover, P5.50 is absent in melanocortin, gly- in other GPCRs. In this case, the longer TM7 coprotein hormone, lysosphingolipid, prostanoid, allows C7.25 to be placed at the tip of the helix and cannabinoid receptors. In these cases, in a favorable position to form a disulphide bond the similarly conserved Y5.58 (73 % of with Cys28 in the N-terminal region. TM7 is the sequences), functionally involved in the kinked at P7.50 of the highly conserved NPxxY stabilization of the active state of the receptor motif. This region of TM7, involved in key con- by interacting with R3.50 of the (D/E)RY motif formational changes associated with GPCR ac- in TM3, as revealed by the crystal structures tivation (Rosenbaum et al. 2009), is highly ir- of “2AR in complex with Gs (Rasmussen regular. A network of water molecules stabilizes et al. 2011b) and the ligand-free opsin (Park the helical deformation of TM7 and provides et al. 2008), is used as reference for sequence hydrogen-bonding partners to polar side chains. alignment of TM5 (Fig. 2.2). The absence of For instance, the unusual P7.50 deformation re- Pro in TM5 of S1P1R leads to a regular a- moves the intrahelical hydrogen bond between helical conformation ( 3.6 residues per turn). the carbonyl group and the N-H amide at po- Thus, the alignment of the S1P1R sequence to sitions 7.45 and 7.49, respectively. A conserved the other receptors requires two-residue gap water molecule is located between the backbone relative to A2AR and one-residue gap relative carbonyl at position 7.45 and the backbone N-H to all other structures, which overlays Y5.37 (i-13 amide at position 7.49. 22 A. Gonzalez et al.
covers the other half (Hanson et al. 2012). In 2.5 The Extracellular Surface these cases, retinal (Hildebrand et al. 2009; in Class A G Protein-Coupled Park et al. 2008) and sphingosine-1-phosphate Receptors (Hanson et al. 2012) may gain access to the binding pocket from the lipid bilayer (Martin- The extracellular surface of GPCRs is defined by Couce et al. 2012). In contrast, ECL2 in the conformation of the N-terminus region and biogenic amine receptors, adenosine and peptide ECLs1-3. Notably, the N-terminus and ECL2 in receptors adopt different spatial conformations particular are highly variable in sequence, length, that maintain the binding site rather accessible and structure (Peeters et al. 2011) (Fig. 2.4). from the extracellular environment (Fig. 2.4). In rhodopsin, the N-terminus (formed by two ECL2 of peptide receptors are formed by two “- “-strands) and ECL2 (two “-strands) block the strands, whereas a helical segment forms ECL2 access of the extracellular ligand to the core of of adrenergic receptors. This ’-helix between the receptor (Palczewski et al. 2000). Similarly, TM4 and the disulphide bridge is not conserved in S1P1R, the N-terminus (contains a short in the other members of the biogenic amine ’-helix) covers half the binding pocket and receptor family. Thus, each receptor subfamily ECL2 (formed by a family-specific disulphide has probably developed, during evolution, a bridge within ECL 2, but lacking the conserved specific N-terminus/ECL2 to adjust the structural disulphide bridge between TM3 and ECL 2) characteristics of its cognate ligands, and to
Fig. 2.4 Molecular surface of the extracellular domain in known crystal structures of GPCRs. The N-terminus domain is shown in red, ECL2 is shown in yellow, and the ligand in the binding site is shown as spheres 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 23 modulate the ligand binding/unbinding events (Hurst et al. 2010;Droretal.2011; Gonzalez 2.7 Intracellular Structural et al. 2011). Changes Associated with Activation of G Protein-Coupled Receptors 2.6 Ligand Binding to G Protein-Coupled Receptors The publication of the crystal structure of the ligand-free opsin (Park et al. 2008), which con- Analysis of the known crystal structures of tains several distinctive features of the active state GPCRs shows that ligand binding mostly occurs as it has been confirmed in the recent structure in a main cavity located between the extracellular of the “2-adrenergic receptor bound to Gs (Ras- segments of TMs 3, 5, 6, and 7 or in a minor mussen et al. 2011b), showed that during the binding cavity located between the extracellular process of receptor activation the intracellular segments of TMs 1, 2, 3, and 7 (Rosenkilde part of TM6 tilts outwards, TM5 nears TM6, and et al. 2010) (Fig. 2.5a). Despite these common R3.50 within the (D/E)RY motif in TM3 adopts pockets, different ligands penetrate to different an extended conformation pointing towards the depths within the TM bundle (Venkatakrishnan protein core, to interact with the highly conserved et al. 2013) (Fig. 2.5bÐf). A major issue in these Y5.58 in TM5 and Y7.53 of the (N/D)PxxY motif in common binding modes is the specificity of TM7 (Fig. 2.6). As shown in the original publica- ligands among subfamilies of receptors. tion of the opsin structure, these conformational
Fig. 2.5 Ligand binding to GPCRs.(a) Binding cav- (d) The binding of doxepin (white)toH1R(Shimamura ities in “2AR. (b) The binding of vorapaxar (white)to et al. 2011) and ergotamine (gray)to5HT1bR (Wang et al. PAR1 (Zhang et al. 2012), IT1t (gray) to CXCR4 (Wu 2013a). (e) The binding of ZM241385 (white)toA2AR et al. 2010), and morphinan (olive)to -OR (Manglik (Jaakola et al. 2008). (f) The binding of ML056 to S1P1R et al. 2012). (c) The binding of the CVX15 cyclic peptide (Hanson et al. 2012). The structures of retinal (orange (olive) to CXCR4 (Wu et al. 2010) and aminoacids 8Ð sticks)andC3.25 and W6.48 (green sticks) are shown in 13 of neurotensin (pink) to NTSR1 (White et al. 2012). panels B-F for comparison purposes 24 A. Gonzalez et al.
Fig. 2.6 Intracellular structural changes associated with receptor activation. Comparison of (a, c) the crystal structure of inactive rhodopsin (1GZM) with (b, d)the crystal structure of the ligand-free opsin (3CAP), which contains several distinctive features of the active state, in views parallel (c, d)and perpendicular (a, b)tothe membrane. Panel B shows the positions of TMs 3, 5Ð7 in rhodopsin (transparent cylinders) and opsin (opaque cylinders) for comparison purposes changes disrupt the ionic interaction between bulky hydrophobic side chain at position 3.40 R3.50 with negatively charged side chains at po- (Fig. 2.7a), highly conserved in the whole Class sitions 3.49 in TM3 and 6.30 in TM6 (Fig. 2.6a, A GPCR family (I:40 %, V:25 %, L:11 %). c) and facilitates the interaction between K5.66 Mutation of I3.40 to either Ala or Gly, i.e. re- in TM 5 and E6.30 in TM 6 (Fig. 2.6b, d). It moving the bulky side chain at this position, has been suggested that conserved hydrophobic abolishes the constitutive activity of the histamine amino acids in the environment of these key H1 receptor, the effect of constitutive-activity polar residues form hydrophobic cages, which increasing mutations, as well as the histamine- also restrain GPCRs in inactive conformations induced receptor activation (Sansuk et al. 2011). (Caltabiano et al. 2013). Thus, the inward movement of P5.50 upon ag- onist binding repositions I3.40 and F6.44, which contributes to a rotation and outward movement 2.8 Mechanism of of TM6 for receptor activation (Rasmussen et al. Ligand-Induced G 2011a). Protein-Coupled Receptor The structures of metarhodopsin II (Choe et al. Activation 2011), the constitutively active rhodopsin (Stand- fuss et al. 2011) and the A2A adenosine receptor The crystal structure of a nanobody-stabilized in complex with the agonist UK-432097 (Xu 6.48 active state of the “2-adrenergic receptor bound to et al. 2011) have shown that W moves toward the BI-167107 agonist (Rasmussen et al. 2011a) TM5 relative to the inactive structures (Fig. 2.7b), shows hydrogen bonding interactions with S5.42 facilitating the rotation and tilt of the intracellular and S5.46 (Fig. 2.7a). These interactions stabilize part of TM6. a receptor conformation that includes a 2.1 Å The role of the extracellular domain in re- inward movement of TM5 at position 5.46 and ceptor function still remains unclear. However, 5.50 1.4 Å inward movement of the conserved P NMR studies on the “2-adrenergic receptor have relative to the inactive, carazolol-bound structure shown ligand-specific conformational changes on (Rosenbaum et al. 2007). This key distortion is the extracellular domain (Bokoch et al. 2010). stabilized in the known crystal structures by a Similarly, it has recently been reported that a 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 25
Fig. 2.7 Mechanisms of ligand-induced receptor ac- carazolol-boundstructure. (b) The conformational change tivation.(a) Detailed view of the “2-adrenergic recep- of inactive 11-cis retinal (in white) to the active 11-trans tor bound to the full agonist BI-167107 (in orange). retinal (in orange) stabilizes a receptor conformation that The hydrogen bond interaction between full agonists includes an inward movement of TM5 together with a and S5.46 stabilizes a receptor conformation that in- movement of W6.48 toward TM5 relative to the inactive cludes an inward movement of TM5 relative to the structures (shown in white for comparison purposes) inactive (shown in white for comparison purposes), small cavity (vestibule) present at the entrance and a single G protein, is maximally activated of the ligand-binding cavity controls the extent by agonist binding to a single protomer. Inverse of receptor movement to govern a hierarchical agonist binding to the second protomer enhances order of G-protein coupling (Bock et al. 2012). signaling, whereas agonist binding to the second Finally, the N-terminal domain of melanocortin protomer blunts signaling. Moreover, binding of receptors plays a significant role in their consti- agonists or the G protein to “2- regulates receptor tutive, ligand-independent, activity (Ersoy et al. oligomerization (Fung et al. 2009). Cysteine 2012). cross-linking experiments have suggested that receptor oligomerization involves hydrophobic interactions via the surfaces of TMs1, 4, and/or 2.9 G Protein-Coupled Receptor 5 (Klco et al. 2003; Guo et al. 2005, 2008). Oligomerization Nevertheless, electrostatic interactions of the intracellular domains are key in the formation of GPCRs have been classically described as receptor heteromers (Navarro et al. 2010). monomeric TM receptors that form a ternary The recent release of the high-resolution crys- complex: a ligand, the GPCR, and its associated tal structures of OR (Manglik et al. 2012) and G protein. This is compatible with observations “1-AR (Huang et al. 2013) in the form of homo- that monomeric rhodopsin and “2-adrenergic oligomers (Fig. 2.8) facilitates the task of mod- receptor are capable of activating G proteins eling GPCR dimers and higher order oligomers. (Ernst et al. 2007; Whorton et al. 2007). The structure of OR shows receptor molecules Nevertheless, it is now well accepted that many associated into pairs through two different in- GPCRs have been observed to oligomerize in terfaces (Fig. 2.8a). The first interface is via cells (Pin et al. 2007; Ferre et al. 2009). It has TMs1 and 2 and Hx8, and the second interface been shown that receptor activation is modulated comprises TMs 5 and 6. The structure of “1-AR by allosteric communication between protomers contains a similar TMs1 and 2 and Hx8 interface of dopamine class A GPCR dimers (Han et al. but the other interface engages residues from 2009). The minimal signaling unit, two receptors TMs4 and 5 (Fig. 2.8b). 26 A. Gonzalez et al.
Fig. 2.8 GPCR oligomerization. The recent high- of homo-oligomers, and sRho (Murakami and Kouyama resolution crystal structures of (a) OR (Manglik et al. 2008)andH1R(Shimamuraetal.2011) in the form of 2012)and(b) “1-AR (Huang et al. 2013) in the form homo-dimers 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 27
Additional crystal structures with GPCR terminal ’5 helix of G’ binds to the intracellular dimers have been published. Interestingly, a TM1 cavity that is opened by the movement of interface, similar to the one observed for OR the cytoplasmic end of TM6 away from TM3 and “1-AR, is present in the structures of the ›OR and towards TM5 (see above). The C-terminal (Wu et al. 2012), opsin (Scheerer et al. 2008), and ’5 helix of the ’-subunit interacts with the metarhodopsin II (Choe et al. 2011). Moreover, extended conformation of R3.50, the short loop the TM4/5 interface of “1-AR resembles the connecting TM7 and Hx8, and the inner side of interface previously obtained for rhodopsin the cytoplasmic TMs 5 and 6 (Fig. 2.9). using atomic force microscopy (Fotiadis et al. 2003). The crystal structure of the histamine H1 receptor (Shimamura et al. 2011) contains a TM4 2.11 The Binding of the C-Tail of G interface (Fig. 2.8d), which is different from the Protein-Coupled Receptors TM4/5 interface of “1-AR due to the absence to Arrestin of TM5 contacts. Similarly, the structures of CXCR4 (Wu et al. 2010) and squid rhodopsin Phosphorylation of several residues of the C-tail (Murakami and Kouyama 2008) contain a TM5 of GPCRs, by Ser/Thr kinases called G protein- interface (Fig. 2.8c), which are different from the coupled receptor kinases (GRKs), promotes the TM4/5 interface of “1-AR or the TM5/6 interface interactions between the receptor and arrestin, of OR. leading to receptor desensitization (Lefkowitz and Shenoy 2005). GPCRs can bind four different arrestin proteins: arrestin-1 and arrestin-4 (known 2.10 The Binding of G as visual arrestins) bind to the phosphorylated Protein-Coupled Receptors form of active rhodopsin, whereas arrestin-2 to the G Protein and arrestin-3 interact and regulate the activity of non-visual GPCRs (Gurevich and Gurevich The formation of the complex between the 2006). active conformation of the receptor and the Arrestin comprises two domains (N- and heterotrimeric G protein triggers GDP release C- domains) of antiparallel “-sheets connected from the G’-subunit, GTP binding to the G’- through a hinge region (Granzin et al. 1998) subunit and dissociation of the G“”-subunits (Fig. 2.10). The binding region for phosphory- (Chung et al. 2011), which finally leads to a lated ligand-activated receptor is located at the N- cascade of signals depending on the G-protein terminal domain, which is occupied by the long type. Noteworthy, more than 800 known GPCRs C-terminal tail in the basal state (blue peptide can bind 17 different G’ subunits, which have in Fig. 2.10a). The crystal structure of arrestin-2 been grouped into four different classes (G’s, in complex with a phosphorylated 29-aminoacid G’i,G’q and G’12)(Simonetal.1991). To carboxy-terminal peptide derived from the human date, the crystal structures of the ligand-free V2 vasopressin receptor (V2Rpp) (Shukla et al. opsin (Scheerer et al. 2008), metarhodopsin 2013) has recently released. This structure shows II (Choe et al. 2011) and the constitutively that the phosphorylated C-tail region of GPCRs active rhodopsin mutant E3.28Q (Standfuss (yellow peptide in Fig. 2.10a) displaces the C- et al. 2011) in complex with a peptide derived tail of arrestin. Moreover, an active conformation from the carboxy terminus of the ’-subunit of arrestin-1, mimicked by C-tail truncation, of the G protein transducin, together with the has also been published (Kim et al. 2013). structure of the “2-adrenergic receptor bound Both structures show significant conformational to Gs (Rasmussen et al. 2011b) have been changes relative to inactive, basal, arrestin. released. These structures have shown that the C- These include rotation of the N- and C-terminal 28 A. Gonzalez et al.
Fig. 2.9 G-protein binding.(a) Crystal structure of complex depicted in panel B.(b) Detailed view of the the “2-adrenergic receptor in complex with the Gs het- interaction between the C-terminal ’5 helix of the ’- erotrimer (’-subunit in olive, “-subunit in white,and”- subunit (in orange) with the short loop connecting TM7 subunit in gray). The C-terminal ’5 helix of the ’-subunit and Hx8 (light blue), TM3 (red), and the inner side of the is shown in orange.Therectangle shows the part of the cytoplasmic TMs 5 (green)and6(blue) domains relative to each other, and major of solved structures that include several members reorientations of the lariat, middle, and finger of the GPCR family (bound to either agonists, loops (Fig. 2.10b). antagonists, or inverse agonists), in the form of monomers or homo-oligomers, in complex with the G protein, or the C-tail bound to 2.12 Conclusions arrestin. Thus, the used of these structures as templates allows molecular modelers to GPCRs are disordered allosteric proteins that simulate the process of signal transduction exhibit modulator behavior with a number of through the cell membrane. These tailor-made guests in both the extracellular (ligand) and models can study ligand binding, receptor intracellular (G protein, arrestin) spaces (Kenakin specificity, receptor activation, G protein and Miller 2010). This considers GPCRs as coupling, allosteric communication among monomeric TM receptors. Nevertheless, it is protomers, among others. However, we want now well accepted that many GPCRs form to emphasize that homology modeling of GPCRs homo- and hetero-oligomers (Khelashvili et al. is far from being a routine technique. Clearly, 2010). Since 2007, innovative crystallographic the inclusion of experimental results can improve techniques (Venkatakrishnan et al. 2013)have the reliability of the models, and their predictive resulted in an exponential growth in the number character. 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 29
Fig. 2.10 The binding of the C-tail of GPCRs to ar- of inactive arrestin (blue peptide). (b) Detailed view of restin.(a) The active conformation of arrestin-2 (PDB the finger, middle and lariat loops, in the presumably id 4JQI, shown in orange) is superimposed to inactive active conformation of arrestin-2, which interact with the arrestin-2 (1G4M, in gray). The phosphorylated C-tail phosphorylated C-tail of GPCRs region of GPCRs (yellow peptide) displaces the C-tail
Ballesteros JA, Weinstein H (1995) Integrated methods References for the construction of three dimensional models and computational probing of structure-function relations Arinaminpathy Y, Khurana E, Engelman DM, Gerstein in G-protein coupled receptors. Methods Neurosci MB (2009) Computational analysis of membrane pro- 25:366Ð428 teins: the largest class of drug targets. Drug Discov Blattermann S, Peters L, Ottersbach PA, Bock A, Konya Today 14(23Ð24):1130Ð1135 V, Weaver CD, Gonzalez A, Schroder R, Tyagi R, 30 A. Gonzalez et al.
Luschnig P, Gab J, Hennen S, Ulven T, Pardo L, Mohr Deupi X, Dolker N, Lopez-Rodriguez M, Campillo M, K, Gutschow M, Heinemann A, Kostenis E (2012) A Ballesteros J, Pardo L (2007) Structural models of biased ligand for OXE-R uncouples G alpha and G class a G protein-coupled receptors as a tool for drug beta gamma signaling within a heterotrimer. Nat Chem design: insights on transmembrane bundle plasticity. Biol 8(7):631Ð638 Curr Top Med Chem 7(10):999Ð1006 Bock A, Merten N, Schrage R, Dallanoce C, Batz J, Deupi X, Olivella M, Sanz A, Dolker N, Campillo M, Klockner J, Schmitz J, Matera C, Simon K, Kebig A, Pardo L (2010) Influence of the g- conformation of Peters L, Muller A, Schrobang-Ley J, Trankle C, Hoff- Ser and Thr on the structure of transmembrane helices. mann C, De Amici M, Holzgrabe U, Kostenis E, Mohr J Struct Biol 169(1):116Ð123 K (2012) The allosteric vestibule of a seven transmem- Deville J, Rey J, Chabbert M (2009) An indel in trans- brane helical receptor controls G-protein coupling. Nat membrane helix 2 helps to trace the molecular evo- Commun 3:1044 lution of class A G-protein-coupled receptors. J Mol Bokoch MP, Zou Y, Rasmussen SG, Liu CW, Nygaard Evol 68(5):475Ð489 R, Rosenbaum DM, Fung JJ, Choi HJ, Thian FS, Dror RO, Pan AC, Arlow DH, Borhani DW, Maragakis P, Kobilka TS, Puglisi JD, Weis WI, Pardo L, Prosser Shan Y, Xu H, Shaw DE (2011) Pathway and mecha- RS, Mueller L, Kobilka BK (2010) Ligand-specific nism of drug binding to G-protein-coupled receptors. regulation of the extracellular surface of a G-protein- Proc Natl Acad Sci U S A 108(32):13118Ð13123 coupled receptor. Nature 463(7277):108Ð112 Ernst OP, Gramse V, Kolbe M, Hofmann KP, Heck Caltabiano G, Gonzalez A, Cordomi A, Campillo M, M (2007) Monomeric G protein-coupled receptor Pardo L (2013) The role of hydrophobic amino acids rhodopsin in solution activates its G protein transducin in the structure and function of the rhodopsin family at the diffusion limit. Proc Natl Acad Sci U S A of G protein-coupled receptors. Methods Enzymol 104(26):10859Ð10864 520:99Ð115 Ersoy BA, Pardo L, Zhang S, Thompson DA, Millhauser Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen G, Govaerts C, Vaisse C (2012) Mechanism of N- SG, Thian FS, Kobilka TS, Choi HJ, Kuhn P, terminal modulation of activity at the melanocortin-4 Weis WI, Kobilka BK, Stevens RC (2007) High- receptor GPCR. Nat Chem Biol 8(8):725Ð730 resolution crystal structure of an engineered human Ferre S, Baler R, Bouvier M, Caron MG, Devi LA, beta2-adrenergic G protein-coupled receptor. Science Durroux T, Fuxe K, George SR, Javitch JA, Lohse 318(5854):1258Ð1265 MJ, Mackie K, Milligan G, Pfleger KD, Pin JP, ChienEY,LiuW,ZhaoQ,KatritchV,HanGW,Hanson Volkow ND, Waldhoer M, Woods AS, Franco R (2009) MA, Shi L, Newman AH, Javitch JA, Cherezov V, Building a new conceptual framework for receptor Stevens RC (2010) Structure of the human dopamine heteromers. Nat Chem Biol 5(3):131Ð134 D3 receptor in complex with a D2/D3 selective antag- Fotiadis D, Liang Y, Filipek S, Saperstein DA, Engel onist. Science 330(6007):1091Ð1095 A, Palczewski K (2003) Atomic-force microscopy: Choe HW, Kim YJ, Park JH, Morizumi T, Pai EF, rhodopsin dimers in native disc membranes. Nature Krauss N, Hofmann KP, Scheerer P, Ernst OP 421(6919):127Ð128 (2011) Crystal structure of metarhodopsin II. Nature Fredriksson R, Schioth HB (2005) The repertoire of G- 471(7340):651Ð655 protein-coupled receptors in fully sequenced genomes. Choma C, Gratkowski H, Lear JD, DeGrado WF (2000) Mol Pharmacol 67(5):1414Ð1425 Asparagine-mediated self-association of a model Fung JJ, Deupi X, Pardo L, Yao XJ, Velez-Ruiz GA, transmembrane helix. Nat Struct Biol 7(2):161Ð166 Devree BT, Sunahara RK, Kobilka BK (2009) Ligand- Chung KY, Rasmussen SG, Liu T, Li S, DeVree BT, Chae regulated oligomerization of beta(2)-adrenoceptors in PS, Calinski D, Kobilka BK, Woods VL Jr, Sunahara a model lipid bilayer. EMBO J 28(21):2384Ð2392 RK (2011) Conformational changes in the G protein Gonzalez A, Perez-Acle T, Pardo L, Deupi X (2011) Gs induced by the beta2 adrenergic receptor. Nature Molecular basis of ligand dissociation in beta- 477(7366):611Ð615 adrenergic receptors. PLoS One 6(9):e23815 Day PW, Rasmussen SG, Parnot C, Fung JJ, Masood Gonzalez A, Cordomí A, Caltabiano G, Campillo A, Kobilka TS, Yao XJ, Choi HJ, Weis WI, Rohrer M, Pardo L (2012) Impact of helix irregulari- DK, Kobilka BK (2007) A monoclonal antibody for ties on sequence alignment and homology mod- G protein-coupled receptor crystallography. Nat Meth- elling of G protein-coupled receptors. Chembiochem ods 4(11):927Ð929 13(10):1393Ð1399 de la Fuente T, Martin-Fontecha M, Sallander J, Benhamu Granier S, Manglik A, Kruse AC, Kobilka TS, Thian B, Campillo M, Medina RA, Pellissier LP, Claeysen FS, Weis WI, Kobilka BK (2012) Structure of the S, Dumuis A, Pardo L, Lopez-Rodriguez ML (2010) delta-opioid receptor bound to naltrindole. Nature Benzimidazole derivatives as new serotonin 5-HT(6) 485(7398):400Ð404 receptor antagonists. Molecular mechanisms of recep- Granzin J, Wilden U, Choe HW, Labahn J, Krafft tor inactivation. J Med Chem 53(3):1357Ð1369 B, Buldt G (1998) X-ray crystal structure of ar- Deupi X, Olivella M, Govaerts C, Ballesteros JA, restin from bovine rod outer segments. Nature Campillo M, Pardo L (2004) Ser and Thr residues 391(6670):918Ð921 modulate the conformation of pro-kinked transmem- Guo W, Shi L, Filizola M, Weinstein H, Javitch JA brane alpha-helices. Biophys J 86(1):105Ð115 (2005) Crosstalk in G protein-coupled receptors: 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 31
changes at the transmembrane homodimer interface SR, Javitch JA, Lohse MJ, Milligan G, Neubig RR, determine activation. Proc Natl Acad Sci U S A Palczewski K, Parmentier M, Pin JP, Vriend G, Cam- 102:17495Ð17500 pagne F, Filizola M (2010) GPCR-OKB: the G protein Guo W, Urizar E, Kralikova M, Mobarec JC, Shi L, coupled receptor oligomer knowledge base. Bioinfor- Filizola M, Javitch JA (2008) Dopamine D2 receptors matics 26(14):1804Ð1805 form higher order oligomers at physiological expres- Kim YJ, Hofmann KP, Ernst OP, Scheerer P, Choe HW, sion levels. EMBO J 27(17):2293Ð2304 Sommer ME (2013) Crystal structure of pre-activated Gurevich VV, Gurevich EV (2006) The structural basis arrestin p44. Nature 497(7447):142Ð146 of arrestin-mediated regulation of G-protein-coupled Klco JM, Lassere TB, Baranski TJ (2003) C5a receptors. Pharmacol Ther 110(3):465Ð502 Receptor oligomerization. I. Disulfide trapping Haga K, Kruse AC, Asada H, Yurugi-Kobayashi T, Shi- reveals oligomers and potential contact surfaces roishi M, Zhang C, Weis WI, Okada T, Kobilka BK, in a G protein-coupled receptor. J Biol Chem Haga T, Kobayashi T (2012) Structure of the human 278(37):35345Ð35353 M2 muscarinic acetylcholine receptor bound to an Klco JM, Nikiforovich GV, Baranski TJ (2006) antagonist. Nature 482(7386):547Ð551 Genetic analysis of the first and third extracellular Han Y, Moreira IS, Urizar E, Weinstein H, Javitch JA loops of the C5a receptor reveals an essential (2009) Allosteric communication between protomers WXFG motif in the first loop. J Biol Chem of dopamine class A GPCR dimers modulates activa- 281(17):12010Ð12019 tion. Nat Chem Biol 5(9):688Ð695 Krissinel E, Henrick K (2004) Secondary-structure match- Hanson MA, Roth CB, Jo E, Griffith MT, Scott FL, Rein- ing (SSM), a new tool for fast protein structure align- hart G, Desale H, Clemons B, Cahalan SM, Schuerer ment in three dimensions. Acta Crystallogr D Biol SC, Sanna MG, Han GW, Kuhn P, Rosen H, Stevens Crystallogr 60(Pt 12 Pt 1):2256Ð2268 RC (2012) Crystal structure of a lipid G protein- Kristiansen K (2004) Molecular mechanisms of ligand coupled receptor. Science 335(6070):851Ð855 binding, signaling, and regulation within the superfam- Hildebrand PW, Scheerer P, Park JH, Choe HW, Piechnick ily of G-protein-coupled receptors: molecular model- R, Ernst OP, Hofmann KP, Heck M (2009) A ligand ing and mutagenesis approaches to receptor structure channel through the G protein coupled receptor opsin. and function. Pharmacol Ther 103(1):21Ð80 PLoS One 4(2):e4382 Kruse AC, Hu J, Pan AC, Arlow DH, Rosenbaum DM, Hopf TA, Colwell LJ, Sheridan R, Rost B, Sander C, Rosemond E, Green HF, Liu T, Chae PS, Dror RO, Marks DS (2012) Three-dimensional structures of Shaw DE, Weis WI, Wess J, Kobilka BK (2012) Struc- membrane proteins from genomic sequencing. Cell ture and dynamics of the M3 muscarinic acetylcholine 149(7):1607Ð1621 receptor. Nature 482(7386):552Ð556 Huang J, Chen S, Zhang JJ, Huang XY (2013) Crystal Lefkowitz RJ, Shenoy SK (2005) Transduction of receptor structure of oligomeric beta1-adrenergic G protein- signals by beta-arrestins. Science 308(5721):512Ð517 coupled receptors in ligand-free basal state. Nat Struct Li J, Edwards PC, Burghammer M, Villa C, Schertler Mol Biol 20(4):419Ð425 GF (2004) Structure of bovine rhodopsin in a trigonal Hurst DP, Grossfield A, Lynch DL, Feller S, Romo crystal form. J Mol Biol 343(5):1409Ð1438 TD, Gawrisch K, Pitman MC, Reggio PH (2010) Liu W, Chun E, Thompson AA, Chubukov P, Xu F, A lipid pathway for ligand binding is necessary for a Katritch V, Han GW, Roth CB, Heitman LH, IJzerman cannabinoid G protein-coupled receptor. J Biol Chem AP, Cherezov V, Stevens RC (2012) Structural basis 285(23):17954Ð17964 for allosteric regulation of GPCRs by sodium ions. Imming P, Sinning C, Meyer A (2006) Drugs, their targets Science 337(6091):232Ð236 and the nature and number of drug targets. Nat Rev Manglik A, Kruse AC, Kobilka TS, Thian FS, Mathiesen Drug Discov 5(10):821Ð834 JM, Sunahara RK, Pardo L, Weis WI, Kobilka BK, Jaakola VP, Griffith MT, Hanson MA, Cherezov V, Granier S (2012) Crystal structure of the micro-opioid Chien EY, Lane JR, Ijzerman AP, Stevens RC (2008) receptor bound to a morphinan antagonist. Nature The 2.6 angstrom crystal structure of a human A2A 485(7398):321Ð326 adenosine receptor bound to an antagonist. Science Martin-Couce L, Martin-Fontecha M, Palomares O, 322(5905):1211Ð1217 Mestre L, Cordomi A, Hernangomez M, Palma S, Katritch V, Cherezov V, Stevens RC (2012) Diversity and Pardo L, Guaza C, Lopez-Rodriguez ML, Ortega- modularity of G protein-coupled receptor structures. Gutierrez S (2012) Chemical probes for the recog- Trends Pharmacol Sci 33(1):17Ð27 nition of cannabinoid receptors in native systems. Katritch V, Cherezov V, Stevens RC (2013) Structure- Angewandte Chemie 51(28):6896Ð6899 function of the G protein-coupled receptor superfam- Mirzadegan T, Benko G, Filipek S, Palczewski K ily. Annu Rev Pharmacol Toxicol 53:531Ð556 (2003) Sequence analyses of G-protein-coupled Kenakin T, Miller LJ (2010) Seven transmembrane recep- receptors: similarities to rhodopsin. Biochemistry tors as shapeshifting proteins: the impact of allosteric 42(10):2759Ð2767 modulation and functional selectivity on new drug Mobarec JC, Sanchez R, Filizola M (2009) Modern discovery. Pharmacol Rev 62(2):265Ð304 homology modeling of G-protein coupled receptors: Khelashvili G, Dorff K, Shan J, Camacho-Artacho M, which structural template to use? J Med Chem Skrabanek L, Vroling B, Bouvier M, Devi LA, George 52(16):5207Ð5216 32 A. Gonzalez et al.
Murakami M, Kouyama T (2008) Crystal structure of Riek RP, Rigoutsos I, Novotny J, Graham RM (2001) squid rhodopsin. Nature 453(7193):363Ð367 Non-alpha-helical elements modulate polytopic mem- Navarro G, Ferre S, Cordomi A, Moreno E, Mallol J, brane protein architecture. J Mol Biol 306(2):349Ð362 Casado V, Cortes A, Hoffmann H, Ortiz J, Canela EI, Rosenbaum DM, Cherezov V, Hanson MA, Ras- Lluis C, Pardo L, Franco R, Woods AS (2010) Inter- mussen SG, Thian FS, Kobilka TS, Choi HJ, Yao actions between intracellular domains as key determi- XJ, Weis WI, Stevens RC, Kobilka BK (2007) nants of the quaternary structure and function of recep- GPCR engineering yields high-resolution structural in- tor heteromers. J Biol Chem 285(35):27346Ð27359 sights into beta2-adrenergic receptor function. Science Oldham WM, Hamm HE (2008) Heterotrimeric G protein 318(5854):1266Ð1273 activation by G-protein-coupled receptors. Nat Rev Rosenbaum DM, Rasmussen SG, Kobilka BK (2009) The Mol Cell Biol 9(1):60Ð71 structure and function of G-protein-coupled receptors. Olivella M, Gonzalez A, Pardo L, Deupi X (2013) Nature 459(7245):356Ð363 Relation between sequence and structure in mem- Rosenkilde MM, Benned-Jensen T, Frimurer TM, brane proteins. Bioinformatics 29(13):1589Ð1592. Schwartz TW (2010) The minor binding pocket: a doi:10.1093/bioinformatics/btt249 major player in 7TM receptor activation. Trends Phar- Palczewski K, Kumasaka T, Hori T, Behnke CA, Mo- macol Sci 31(12):567Ð574 toshima H, Fox BA, Trong IL, Teller DC, Okada Sansuk K, Deupi X, Torrecillas IR, Jongejan A, Nijmeijer T, Stenkamp RE, Yamamoto M, Miyano M (2000) S, Bakker RA, Pardo L, Leurs R (2011) A structural in- Crystal structure of rhodopsin: a G protein-coupled sight into the reorientation of transmembrane domains receptor. Science 289(5480):739Ð745 3 and 5 during family a G protein-coupled receptor Pardo L, Deupi X, Dolker N, Lopez-Rodriguez ML, activation. Mol Pharmacol 79(2):262Ð269 Campillo M (2007) The role of internal water Scheerer P, Park JH, Hildebrand PW, Kim YJ, Krauss molecules in the structure and function of the N, Choe HW, Hofmann KP, Ernst OP (2008) Crystal rhodopsin family of G protein-coupled receptors. structure of opsin in its G-protein-interacting confor- Chembiochem 8(1):19Ð24 mation. Nature 455(7212):497Ð502 Park JH, Scheerer P, Hofmann KP, Choe HW, Ernst OP Senes A, Gerstein M, Engelman DM (2000) Statistical (2008) Crystal structure of the ligand-free G-protein- analysis of amino acid patterns in transmembrane coupled receptor opsin. Nature 454(7201):183Ð187 helices: the GxxxG motif occurs frequently and in Peeters MC, van Westen GJ, Li Q, Ijzerman AP (2011) association with “-branched residues at neighboring Importance of the extracellular loops in G protein- positions. J Mol Biol 296:921Ð936 coupled receptors for ligand recognition and receptor Serrano-Vega MJ, Magnani F, Shibata Y, Tate CG (2008) activation. Trends Pharmacol Sci 32(1):35Ð42 Conformational thermostabilization of the beta1- Pieper U, Schlessinger A, Kloppmann E, Chang GA, adrenergic receptor in a detergent-resistant form. Proc Chou JJ, Dumont ME, Fox BG, Fromme P, Hendrick- Natl Acad Sci U S A 105(3):877Ð882 son WA, Malkowski MG, Rees DC, Stokes DL, Stow- Shimamura T, Shiroishi M, Weyand S, Tsujimoto H, ell MH, Wiener MC, Rost B, Stroud RM, Stevens RC, Winter G, Katritch V, Abagyan R, Cherezov V, Liu W, Sali A (2013) Coordinating the impact of structural Han GW, Kobayashi T, Stevens RC, Iwata S (2011) genomics on the human alpha-helical transmembrane Structure of the human histamine H1 receptor complex proteome. Nat Struct Mol Biol 20(2):135Ð138 with doxepin. Nature 475(7354):65Ð70 Pin JP, Neubig R, Bouvier M, Devi L, Filizola M, Javitch Shukla AK, Manglik A, Kruse AC, Xiao K, Reis RI, JA, Lohse MJ, Milligan G, Palczewski K, Parmen- Tseng WC, Staus DP, Hilger D, Uysal S, Huang tier M, Spedding M (2007) International union of LY, Paduch M, Tripathi-Shukla P, Koide A, Koide basic and clinical pharmacology. LXVII. Recommen- S, Weis WI, Kossiakoff AA, Kobilka BK, Lefkowitz dations for the recognition and nomenclature of G RJ (2013) Structure of active beta-arrestin-1 bound to protein-coupled receptor heteromultimers. Pharmacol a G-protein-coupled receptor phosphopeptide. Nature Rev 59(1):5Ð13 497(7447):137Ð141 Rasmussen SG, Choi HJ, Fung JJ, Pardon E, Casarosa Simon MI, Strathmann MP, Gautam N (1991) Diver- P, Chae PS, Devree BT, Rosenbaum DM, Thian sity of G proteins in signal transduction. Science FS, Kobilka TS, Schnapp A, Konetzki I, Suna- 252(5007):802Ð808 hara RK, Gellman SH, Pautsch A, Steyaert J, Weis Smit MJ, Vischer HF, Bakker RA, Jongejan A, Tim- WI, Kobilka BK (2011a) Structure of a nanobody- merman H, Pardo L, Leurs R (2007) Pharma- stabilized active state of the beta(2) adrenoceptor. cogenomic and structural analysis of constitutive G Nature 469(7329):175Ð180 protein-coupled receptor activity. Annu Rev Pharma- Rasmussen SG, DeVree BT, Zou Y, Kruse AC, Chung KY, col Toxicol 47:53Ð87 Kobilka TS, Thian FS, Chae PS, Pardon E, Calinski Standfuss J, Edwards PC, D’Antona A, Fransen M, Xie G, D, Mathiesen JM, Shah ST, Lyons JA, Caffrey M, Oprian DD, Schertler GF (2011) The structural basis Gellman SH, Steyaert J, Skiniotis G, Weis WI, Suna- of agonist-induced activation in constitutively active hara RK, Kobilka BK (2011b) Crystal structure of the rhodopsin. Nature 471(7340):656Ð660 beta2 adrenergic receptor-Gs protein complex. Nature Thompson AA, Liu W, Chun E, Katritch V, Wu H, Vardy 477(7366):549Ð555 E, Huang XP, Trapella C, Guerrini R, Calo G, Roth 2 Modeling of G Protein-Coupled Receptors Using Crystal Structures... 33
BL, Cherezov V, Stevens RC (2012) Structure of the agonist action on a beta(1)-adrenergic receptor. Nature nociceptin/orphanin FQ receptor in complex with a 469(7329):241Ð244 peptide mimetic. Nature 485(7398):395Ð399 White JF, Noinaj N, Shibata Y, Love J, Kloss B, Xu F, Venkatakrishnan AJ, Deupi X, Lebon G, Tate CG, Gvozdenovic-Jeremic J, Shah P, Shiloach J, Tate CG, Schertler GF, Babu MM (2013) Molecular Grisshammer R (2012) Structure of the agonist-bound signatures of G-protein-coupled receptors. Nature neurotensin receptor. Nature 490(7421):508Ð513 494(7436):185Ð194 Whorton MR, Bokoch MP, Rasmussen SG, Huang B, Zare Von Heijne G (1991) Proline kinks in transmembrane RN, Kobilka B, Sunahara RK (2007) A monomeric alpha-helices. J Mol Biol 218(3):499Ð503 G protein-coupled receptor isolated in a high-density Wacker D, Wang C, Katritch V, Han GW, Huang XP, lipoprotein particle efficiently activates its G protein. Vardy E, McCorvy JD, Jiang Y, Chu M, Siu FY, Proc Natl Acad Sci U S A 104(18):7682Ð7687 Liu W, Xu HE, Cherezov V, Roth BL, Stevens RC Wu B, Chien EY, Mol CD, Fenalti G, Liu W, Katritch (2013) Structural features for functional selectivity at V, Abagyan R, Brooun A, Wells P, Bi FC, Hamel serotonin receptors. Science 340(6132):615Ð619 DJ, Kuhn P, Handel TM, Cherezov V, Stevens RC Wang C, Jiang Y, Ma J, Wu H, Wacker D, Katritch (2010) Structures of the CXCR4 chemokine GPCR V, Han GW, Liu W, Huang XP, Vardy E, McCorvy with small-molecule and cyclic peptide antagonists. JD, Gao X, Zhou EX, Melcher K, Zhang C, Bai Science 330(6007):1066Ð1071 F, Yang H, Yang L, Jiang H, Roth BL, Cherezov Wu H, Wacker D, Mileni M, Katritch V, Han GW, Vardy V, Stevens RC, Xu HE (2013a) Structural basis for E, Liu W, Thompson AA, Huang XP, Carroll FI, molecular recognition at serotonin receptors. Science Mascarella SW, Westkaemper RB, Mosier PD, Roth 340(6132):610Ð614 BL, Cherezov V, Stevens RC (2012) Structure of the Wang C, Wu H, Katritch V, Han GW, Huang XP, Liu W, human kappa-opioid receptor in complex with JDTic. Siu FY, Roth BL, Cherezov V, Stevens RC (2013b) Nature 485(7398):327Ð332 Structure of the human smoothened receptor bound to Xu F, Wu H, Katritch V, Han GW, Jacobson KA, Gao an antitumour agent. Nature 497(7449):338Ð343 ZG, Cherezov V, Stevens RC (2011) Structure of an Warne T, Serrano-Vega MJ, Baker JG, Moukhametzianov agonist-bound human A2A adenosine receptor. Sci- R, Edwards PC, Henderson R, Leslie AG, Tate CG, ence 332(6027):322Ð327 Schertler GF (2008) Structure of a beta(1)-adrenergic Zhang C, Srinivasan Y, Arlow DH, Fung JJ, Palmer G-protein-coupled receptor. Nature 454:486Ð491 D, Zheng Y, Green HF, Pandey A, Dror RO, Shaw Warne T, Moukhametzianov R, Baker JG, Nehme R, DE, Weis WI, Coughlin SR, Kobilka BK (2012) Edwards PC, Leslie AG, Schertler GF, Tate CG High-resolution crystal structure of human protease- (2011) The structural basis for agonist and partial activated receptor 1. Nature 492(7429):387Ð392 Part II GPCRs in Motion: Insights from Simulations Structure and Dynamics of G-Protein Coupled Receptors 3
Nagarajan Vaidehi, Supriyo Bhattacharya, and Adrien B. Larsen
Abstract G-protein coupled receptors (GPCRs) are seven helical transmembrane proteins that mediate cell-to-cell communication. They also form the largest superfamily of drug targets. Hence detailed studies of the three dimensional structure and dynamics are critical to understanding the functional role of GPCRs in signal transduction pathways, and for drug design. In this chapter we compare the features of the crystal structures of various biogenic amine receptors, such as “1 and “2 adrenergic receptors, dopamine D3 receptor, M2 and M3 muscarinic acetylcholine receptors. This analysis revealed that conserved residues are located facing the inside of the transmembrane domain in these GPCRs improving the efficiency of packing of these structures. The NMR structure of the chemokine receptor CXCR1 without any ligand bound, shows significant dynamics of the transmembrane domain, especially the helical kink angle on the transmembrane helix6. The activation mechanism of the “2-adrenergic receptor has been studied using multiscale computational methods. The results of these studies showed that the receptor without any ligand bound, samples conformations that resemble some of the structural characteristics of the active state of the receptor. Ligand binding stabilizes some of the conformations already sampled by the apo receptor. This was later observed in the NMR study of the dynamics of human “2-adrenergic receptor. The dynamic nature of GPCRs leads to a challenge in obtaining purified receptors for biophysical studies. Deriving thermostable mutants of GPCRs has been a successful strategy to reduce the conformational heterogeneity and stabilize the receptors. This has lead to several crystal
N. Vaidehi () ¥ S. Bhattacharya ¥ A.B. Larsen Division of Immunology, Beckman Research Institute of the City of Hope, 1500, E. Duarte Road, Duarte, CA 91010, USA e-mail: [email protected]
M. Filizola (ed.), G Protein-Coupled Receptors - Modeling and Simulation, Advances 37 in Experimental Medicine and Biology 796, DOI 10.1007/978-94-007-7423-0__3, © Springer ScienceCBusiness Media Dordrecht 2014 38 N. Vaidehi et al.
structures of GPCRs. However, the cause of how these mutations lead to thermostability is not clear. Computational studies are beginning to shed some insight into the possible structural basis for the thermostability. Molecular Dynamics simulations studying the conformational ensemble of thermostable mutants have shown that the stability could arise from both enthalpic and entropic factors. There are regions of high stress in the wild type GPCR that gets relieved upon mutation conferring thermostability.
Keywords Thermostable mutants ¥ Predictions ¥ Conformational states ¥ Drug design ¥ Biogenic amines ¥ Multiscale method ¥ LItiCon
Although the class A GPCRs have a common 3.1 Introduction seven helical transmembrane topology, their cel- lular functions are divergent. Further, the same 3.1.1 Introduction to Class A GPCRs GPCR can bind to different ligands and thereby elicit different responses depending on the nature G-protein coupled receptors (GPCRs), also of the ligand that binds to the GPCR (Audet and known as the seven transmembrane (TM) Bouvier 2012). They do so by varying the three receptors form a superfamily of membrane dimensional structural conformation of the recep- bound proteins that mediate cell signaling tor. Different ligands also elicit different types of and are responsible for many physiological response at a receptor level (Vaidehi and Kenakin functions of the cell. GPCRs belong to one of 2010). A full agonist elicits a maximal response the largest superfamily of receptors that serve while a partial agonist elicits only a fraction of the as important drug targets. GPCRs have been maximal response at saturating concentrations of broadly classified into three classes based on the the ligand. A very low efficacy ligand that may lengths of their amino termini. In this chapter not produce any measurable response but could we discuss only the class A GPCRs (denoted as prevent further response from a full or a partial just GPCRs hereafter) that have short, about agonist is known as neutral antagonist. Ligands 10Ð75 amino acids in their amino terminus. that reduce the basal response exhibited in a GPCRs have a structural topology consisting of ligand-independent fashion in many GPCRs are seven transmembrane (TM) helices connected by known as an inverse agonist. GPCRs exist in intracellular (ICL) and extracellular (ECL) loops. multiple “active” and “inactive” conformational Ligands ranging from small molecules to pep- states that are in equilibrium even in the absence tides and proteins activate GPCRs by binding ei- of any ligand (Niesen et al. 2011; Vaidehi and ther in the TM domain or in the extracellular loop Bhattacharya 2011). This leads to the ligand in- region. Ligand binding leads to conformational dependent basal activation of the receptor and changes in the receptor that results in coupling the level of the basal activation is different for with cytosolic proteins such as trimeric G-protein each GPCR. In the ligand independent state, the or “-arrestin to elicit cellular response (Reiter population of the receptor in the inactive state of et al. 2012). Understanding the mechanism of the receptor is still higher than the active state of how ligand binding in the extracellular region the receptor (Niesen et al. 2011). The “inactive of the receptor, causing an allosteric effect of state” is defined as a conformation of the recep- coupling to cytosolic proteins, triggering the sig- tor that does not activate the G-protein. Upon nal transduction cascade is an important step in binding of an agonist (full or partial) and the modulating various cellular processes initiated by G-protein, the GPCR undergoes conformational GPCRs. changes that lead to the “active state” of the 3 Structure and Dynamics of G-Protein Coupled Receptors 39 receptor (Rasmussen et al. 2011). Upon bind- receptors (Cherezov et al. 2007; Rasmussen et al. ing of an agonist or inverse agonist the relative 2011), muscarinic acetyl choline receptors M2 population of the active and inactive states shift and M3 (Haga et al. 2012; Kruse et al. 2012), depending on the efficacy of the ligand (Yao dopamine D3DR receptor (Chien et al. 2010), his- et al. 2006, 2009; Vaidehi and Kenakin 2010). tamine H1HR receptor (Shimamura et al. 2011), However binding of the agonist alone may not be adenosine A2A receptors (Lebon et al. 2011b, sufficient to activate the receptor, which is evident Xu et al. 2011) chemokine receptor CXCR4 (Wu from a few agonist bound crystal structures of et al. 2010), peptide receptor neurotensin receptor GPCRs (e.g. “2-adrenergic receptor bound to BI- 1 (White et al. 2012) and protease activated 167107, “1-adrenergic receptor bound to isopro- receptor 1 (Zhang et al. 2012), and opioid recep- terenol). These structures are very close to the tors (Granier and Kobilka 2012; Manglik et al. inactive conformations of various GPCRs. It has 2012; Thompson et al. 2012) have been crys- been suggested that binding of both the ago- tallized using these two strategies for stabilizing nist and G-protein are possibly required to shift specific conformations. With the publication of the conformational equilibrium of the receptor all these GPCR structures, one can understand to the fully active state. However, the agonist the structural differences between closely related bound adenosine receptor A2A shows active-like receptors as well as those that are far removed state movement (Lebon et al. 2011b). Exactly in sequence space. Additionally “2-adrenergic how do the receptor conformations change upon receptor has been crystallized both in the “in- binding of various ligands of different efficacies active” and “fully active” conformational states. and how the relative population of the various Adenosine A2A receptor has been crystallized conformational states are modulated by ligand in the inactive and partially active state (Lebon binding is a current topic in GPCR research et al. 2011b;Xuetal.2011). Many of these (Granier and Kobilka 2012). An understanding of GPCRs have also been crystallized with only the the structural and dynamic features of the various agonist bound. In the next section we compare active and inactive conformations of GPCRs, various GPCR crystal structures and analyze the and the molecular mechanism of the conforma- similarities and differences. tional changes from the inactive to the active states is important in designing functionally spe- cific drugs (Galandrin et al. 2007; Mailman and 3.1.2 Comparison of Various GPCR Murthy 2010; Kenakin et al. 2012). In fact this Structures dynamic flexibility of GPCRs poses one of the major bottlenecks in obtaining pure protein for GPCRs have a seven helical topology and the biophysical structural studies of GPCRs (Tate structural differences between two structures and Schertler 2009;Tate2012). These bottle- arise from the differences in the helical tilts necks have been overcome partially, resulting and kinks, helical rotational and translational in many recently published three dimensional orientations, packing of the helices, and the loop crystal structures of GPCRs. There are two major structures. Two GPCR structures are compared strategies that have been used to stabilize GPCRs by aligning the C’ atoms, and the root mean for crystallization: (1) Attachment of an antibody squared deviation (RMSD) in coordinates of a set (Steyaert and Kobilka 2011) or a protein such as of chosen atoms that is common between the two T4Lysozyme (denoted as T4L hereafter) to the compared structures is reported. RMSD is a crude ICL or ECL to facilitate crystallization (Chere- measure of the differences between structures zov et al. 2007; Katritch et al. 2012) and (2) and does not describe the nuances in structural screening for thermostable mutations that stabi- similarities and differences especially for two lize a particular receptor conformation to enable GPCR structures with divergent sequences. crystallization. In fact many GPCRs such as “1- Therefore we developed a computational adrenergic (Warne et al. 2008) and “2-adrenergic method (GPCRCompare) that provides structural 40 N. Vaidehi et al.
Fig. 3.1 Comparison of rigid body structural differences Helices are represented by cylinders; block arrows repre- (helical translation, tilt and axial rotation) of CXCR4 (pdb sent center of mass translations, curved arrows represent ID: 3ODU) and D3DR (pdb ID: 3PBL) crystal structures tilts and rotations. Figures are schematic, and not to actual to that of “2AR (pdb ID: 2RH1); (a) CXCR4; (b)D3DR; scale differences in terms of the relative helical (10ı), while TM6 shows a minor translation tilt, rotational and translational orientations (0.6 Å), tilt (5ı), and rotation (8ı) compared to (Bhattacharya et al. 2013). Such a comparison the “2-adrenergic receptor. provides a more intuitive understanding of the The crystal structures of several class A structural similarities and differences. Figure 3.1 GPCRs have been compared in previous pub- shows the comparison of two crystal structures lications (Katritch et al. 2012; Deupi 2012; Bhat- of similar receptors such as the dopamine D3DR tacharya et al. 2013; Filizola and Devi 2013). In receptor and “2 adrenergic receptor and also the next section we demonstrate the advantages of comparison of two different receptors such as the comparing GPCR structures using helical struc- chemokine receptor, CXCR4 and the dopamine tural parameters with GPCRCompare program receptor D3DR that have very different amino to compare the crystal structures of the inactive acid sequences. The structural comparisons here states of the biogenic amine receptors (“1 and “2 have been made based on relative translational, adrenergic receptors, muscarinic acetyl choline rotational and tilt orientations of the seven receptors M2 and M3, dopamine receptor D3DR helices. Such an analysis provides insight into and histamine receptor H1HR) crystallized so far. the type of helical movements that occur when comparing GPCR structures. The structural differences in the helical movements between 3.1.3 Comparison of the Inactive CXCR4 and “2 adrenergic receptor are more States of the Biogenic Amine predominant as anticipated, compared to the Receptors differences between D3DR and “2 adrenergic receptor. Significant structural differences Here we have compared the adrenergic receptor between CXCR4 and “2 adrenergic receptor structures to the dopamine, histamine and are in the rotational and tilt orientations of muscarinic acetylcholine receptors. Since the TM2, TM4 and TM5 as shown in Fig. 3.1a. structures of “1 and “2-adrenergic receptors The difference in the tilt angle of TM1 is also are similar, we report a comparison of the significant, but this difference could stem from adrenergic receptor structures with those of the truncated amino terminus of each of these the other biogenic amine family receptors. The receptors. In D3DR, TM4 shows a small rotation major differences between the dopamine D3 3 Structure and Dynamics of G-Protein Coupled Receptors 41 receptor and the “1 and “2 adrenergic receptors in all these receptors. TM3 has more conserved are tilts of TM4, TM5 and TM6 (10ı,9ı,8ı residues than any other TM helix. The extracel- respectively) and rotations of TM4 and TM5 (9ı lular loops show less conserved residues than and 10ı). The histamine H1 receptor shows a the intracellular loops. ICL1 is highly conserved 7ı tilt of TM4 and major rotation of TM6, and in all the receptors. We have omitted histamine TM7 (13ı,8ı). The muscarinic acetylcholine H1HR in this analysis since it had only 20 % se- M2 receptor shows a 1.2 Å translation of TM5, quence identity to other subtypes of the histamine tilts of TM4 and TM5 (7ı,6ı), and rotations receptors and skewed this analysis. of TM1, TM5, TM6 (11ı,6ı,13ı). The M3 muscarinic receptor shows significant differences in the rotational orientations of TM5 and TM6 3.1.4 Analysis of the NMR (9ı,13ı respectively). Thus in comparison to the Structures of the Chemokine “1 and “2 adrenergic receptors, TM4, TM5 and Receptor CXCR1 TM6 are the most diverse in conformation among the biogenic amine family of receptors. In these We have also compared the extent of variability receptors, TM3, TM5 and TM6 contain a large and the dynamics observed in the recently pub- proportion of the residues that line the ligand lished three dimensional NMR structures of the binding pocket of the orthosteric ligands. While chemokine receptor CXCR1 (Park et al. 2012). TM3 consists of a conserved aspartate residue This structure is the first, class A GPCR structure that forms a salt bridge with the amine group to be solved without any ligand bound to it. of the endogenous biogenic amines, the confor- The NMR structures of CXCR1 in phospholipids mations of TM5 and TM6 give selectivity to the show significant dynamics. We have analyzed specific ligands each of these receptors binds. The the extent of the dynamics of these structures diverse conformation of TM4 could be related to using GPCRCompare program in this section. selectivity of dimerization with other GPCRs, The RMSD in coordinates among these ten NMR since TM4 has been shown to form the dimer models ranges from 1.8 to 2.2 Å. The ensemble interface in several GPCRs (Johnston et al. 2011). of NMR conformations is shown in Fig. 3.3. With the availability of representative struc- The RMSD in coordinates among these ten NMR tures of the biogenic amine receptors an in depth models range from 1.8 to 2.2 Å. The kink an- analysis of the location of the conserved residues gle on TM6 ranges from 21ı to 34ı showing and their role in structure packing would be significant dynamics in the ten NMR structures useful for structure prediction protocols (Bhat- as shown in Figs. 3.3 and 3.4a. To quantify the tacharya et al. 2008a, b, 2013; Bhattacharya and dynamic fluctuations of different regions of the Vaidehi 2010;Halletal.2009; Nedjai et al. receptor, we computed the average root mean 2011). We have analyzed the extent to which square fluctuations (RMSF) of the C’ coordinate the residues are conserved in each one of these of each residue from the average NMR structure. receptors within their subfamilies and compared Figure 3.4b shows a representative NMR struc- the location of the conserved residues in the re- ture with the residues colored according to their spective receptor structures as shown in Fig. 3.2. mean RMSF, with red being large fluctuations The conserved residues are shown in blue and and blue being less than 1 Å change in RMSD. the non-conserved ones are shown in red. It can TM6 and helix 8 are the most dynamic segments be seen from Fig. 3.2, that the residues in the in the CXCR1 structure, followed by TM3 and intracellular half of the transmembrane domains TM5. Among the loops, ECL2 is packed within of the receptors are more conserved than the the TM bundle and shows very little fluctuations, extracellular half of the receptors. As shown in whereas ICL1 and ICL3 are the most dynamic. Fig. 3.2c, d, the residues that point inwards into In order to further “dissect” the nature of helical the transmembrane barrel are more conserved motions observed in CXCR1, we compared the than the residues that point to the membrane rigid body orientations of the seven TM helices of 42 N. Vaidehi et al.
Fig. 3.2 Comparison of the location of conserved the receptors. (b) The extracellular loops shows a lot of residues in the crystal structures of “1 and “2 adrenergic variability that could confer specificity to ligand binding; receptors, M2 and M3 muscarinic acetylcholine recep- (c) The view from the extracellular side of the receptor tors, D3 dopamine receptor. The residues that are highly showing that the conserved residues mostly face inside the conserved within their respective subtypes are shown in TM bundle, and (d) The view from the intracellular side of blue and the non-conserved residues are shown in red. the receptors showing more conserved residues pointing (a) The side view of the seven TM domains showing that inwards towards the TM bundle more conserved residues are in the intracellular region of the ten NMR conformations against one another deformations (kinks) of the helices. For CXCR1, using the program GPCRCompare. In the course the aligned RMSDs of the TM helices range of this analysis, we measured the C’ RMSD of from 0.4 to 0.8 Å with TM7 showing the highest individual helices, first by aligning the overall aligned RMSD. Next we compared the rigid body structures (referred to as “in place” in Fig. 3.4), orientations of the TM helices. TM5, TM6, and and then by aligning the TM helix of one struc- TM7 show significant variations in the tilt and ture to that of the other (referred to as “aligned” in rotational motion, whereas TM3 shows less tilt Fig. 3.4). Figure 3.4c compares the mean RMSD and larger rotation. A schematic of the dynamic of the TM domains both before and after align- motions of CXCR1 is shown in Fig. 3.4d. The ta- ment. Since the alignment removes the rigid body ble in Fig. 3.4e lists the magnitudes of rigid body orientational differences between the helices, the movements of the different helices, as calculated “aligned” RMSD gives a measure of the internal by GPCRCompare. 3 Structure and Dynamics of G-Protein Coupled Receptors 43
Fig. 3.3 Superposition of the 10 NMR derived structures from the pdb file 2LNL. The backbone of each structure is shown in a different color. The kink on helix6 is dynamic in the NMR time scale
ligands of varied efficacies on the dynamics of the 3.2 Flexibility of GPCR receptor has also been well studied using several Structures and Relevance biophysical techniques. to Drug Design
GPCRs exhibit multiple functions in the cell, that 3.2.1 NMR and Fluorescence is achieved by adopting various receptor confor- Spectroscopy Shows mations influenced by binding of the ligand, G- Flexibility in GPCR protein, other proteins coupling to the receptor, Conformations and by the lipid bilayer environment. Hence an understanding of the receptor flexibility and con- The NMR structure of the chemokine receptor formational dynamics would lead to more selec- CXCR1 in phospholipid bilayers has been pub- tive design of drugs. The fact that the GPCR con- lished recently (Park et al. 2012). As discussed formations are dynamic has been substantiated by in the previous section, comparison of the ten many experimental and computational techniques NMR models shows significant level of dynam- (Swaminath et al. 2004, 2005; Bhattacharya et al. ics in the system. The structural variations in 2008a; Ahuja and Smith 2009;Droretal.2009; the ten NMR structures are in the intracellular Khelashvili et al. 2009; Bhattacharya and Vaidehi part of TM3, kink modulation on TM6 and he- 2010; Vaidehi and Kenakin 2010; Vaidehi 2010; lix8 as shown in Figs. 3.3 and 3.4. Interestingly, Niesen et al. 2011; Provasi et al. 2011;Parketal. GPCR crystal structures show the TM helices 2012; Nygaard et al. 2013). Further the effect of to be less flexible (as evident by low b-factors), 44 N. Vaidehi et al.
Fig. 3.4 Variability among the 10 NMR derived confor- (d) Schematic representation showing movements exhib- mations of CXCR1 (pdb ID: 2LNL); (a) Cartoon represen- ited in TM5, TM6 and TM7 deduced from the ensemble tation of the 10 NMR conformations; (b) Two representa- of NMR structures; (e) Table showing the average value tive NMR structures with residues colored according to of the translation, tilt and rotational orientations of seven the mean RMSD fluctuations in coordinates; (c)RMSDs helices for the ten NMR derived structures calculated by of each transmembrane helix before and after alignment; the program GPCRCompare (Bhattacharya et al. 2013) whereas the loops are more mobile with higher dynamic of all the TMs. The kink angle on b-factors. However in the apoprotein structure of transmembrane helix 6 is also dynamic and varies CXCR1 the TM helices show high mobility. This from 21ı to 34ı. Based on earlier studies on is a marked distinction from crystal structures, activation of GPCRs, the side chain conformation which are cryogenic conformations, whereas the of the Trp6.48 of the highly conserved motif WxP NMR structures represent the dynamic nature of in TM6, was expected to toggle upon activation GPCRs at the physiological temperature. Also (Yao et al. 2006). This toggle was observed in the use of phospholipid bilayers in the NMR the NMR studies of the activation of rhodopsin measurements could contribute to the increased (Ahuja and Smith 2009) but was not seen in mobility of the TM regions. However the high the crystal structure of the active state of “2- mobility of the TM helices is in direct agreement adrenergic receptor (Rasmussen et al. 2011), or with the established notion on GPCR activation, the partial active state of adenosine A2A receptor which suggests that the helical domains (most im- (Lebon et al. 2011a, b;Xuetal.2011). The portantly TM6) undergo conformational change side chain conformation of Trp2556.48 in CXCR1 to trigger activation. It is noteworthy that in the shows significant variation among the ten NMR NMR conformations of CXCR1, TM6 is the most structures, denoting the dynamic nature of this 3 Structure and Dynamics of G-Protein Coupled Receptors 45 residue in the absence of any ligand. Here we to the human “2-adrenergic receptor lead to have used the Ballesteros and Weinstein num- stabilization of ligand specific conformations bering for residues that were defined for class of the receptor (Swaminath et al. 2004, 2005; A GPCRs (Ballesteros and Weinstein 1995). The Yao et al. 2006). Bioluminescence Resonance first number denotes the TM helix in which the Energy Transfer (BRET) studies pioneered by residue is located and the second number is the Bouvier and coworkers, on the kinetics of G position of the residue from the most conserved protein coupling to “2-adrenergic receptor and residue position denoted as x.50 in that helix. “1-adrenergic receptor in living cells, showed In the CXCR1 NMR structures, the distance be- differential activation kinetics by full, partial tween the intracellular regions of TM3 and TM6 and inverse agonists (Vilardaga et al. 2003; (for example, the distance between Arg135 and Galés et al. 2005; Galandrin et al. 2007). Met241) varies between 9.1 and 12.6 Å in the Fluorescence experiments of reconstituted human ten NMR structures. This analysis shows that “2-adrenergic receptor in lipid nanoparticles the GPCR without any ligand samples a wide coupled with the G-protein, showed that ligands range of conformations. Future studies on NMR modulate the binding affinity of the receptor structures with ligand on CXCR1 would reveal to the G-protein and similarly binding of the if conformational selection from the subset of G-protein also modulates the efficacy of the conformations sampled by the apoprotein could receptor to the ligand (Yao et al. 2009). Since be one of the predominant mechanisms of ligand the experimental results on activation of GPCRs binding in GPCRs. are fragmented, computational methods play an Biophysical studies on the visual receptor important role in combining these observations to rhodopsin, using spin-labeling techniques, provide atomic level insight into the mechanism solid-state NMR, fluorescent spectroscopy, of the conformational transitions and activation computational methods, and crystal structure in GPCRs. of the partially active state opsin show that the intracellular region of the TM3 and TM6 move apart upon activation (Park et al. 3.2.2 Multiscale Computational 2008). Additionally, there are several inter- Methods for Understanding residue contacts also known as “conformational the Mechanism switches” made or broken upon activation of of Conformational Transitions rhodopsin (Krishna et al. 2002; Hubbell et al. and Activation of GPCRs 2003; Schertler 2005;Parketal.2008; Ahuja and Smith 2009; Zaitseva et al. 2010). Briefly, Computational methods have been used to comparison of the rhodopsin and opsin crystal delineate the atomic level mechanisms of structures showed considerable conformational activation of GPCRs. All-atom molecular changes in the intracellular region of TM5 dynamics (MD) simulations combined with and TM6. TM5 also shows elongation of the inter-residue distances measured by NMR as helix in this region. Comparatively smaller restraints provided insight into the movement changes were observed in the extracellular of ECL2 in the activation of rhodopsin (Ahuja loop 2 (ECL2) (Ahuja and Smith 2009). There and Smith 2009). Starting from the inactive are substantial rearrangements in all the three state crystal structure of rhodopsin, several MD intracellular loops in the opsin structure. Thus simulation studies have provided insight into the it is evident that activation is associated with early events of activation of rhodopsin (Saam considerable conformational changes especially et al. 2002; Crozier et al. 2007; Grossfield et al. in the intracellular region of the receptor. 2008; Khelashvili et al. 2009). The activation Fluorescence spectroscopic life time mea- mechanism of rhodopsin from MD trajectories surements with purified receptors, have shown has been described in terms of structural that binding of ligands of varied efficacies changes in inter-residue contacts also known 46 N. Vaidehi et al.
Fig. 3.5 A scheme of the conformational sampling in formational transitions. All-atom Cartesian simulations multiscale simulation method. The coarse grained sim- are then performed starting from the low energy and well ulation methods in this figure represents methods that populated structures from the coarse grained simulations. sample wider range of conformations with accurate all- This would lead to more accurate sampling of the low atom forcefield. The coarse grained methods capture con- energy states as microdomain switches. These microdomain all-atom MD simulations. This is because the switches are also modulated by interaction with system gets trapped in the potential energy min- cholesterol (Khelashvili et al. 2009). imum of the inactive state with high free energy Dror et al. performed microseconds of MD barriers to transition to the active state, and the simulations on the inactive state of “2AR crystal dynamic movement to the active state has to structure with and without T4L lysozyme (T4L) occur through a series of rare events as the system bound to show that there are two inactive states moves from one potential energy basin to another. of the receptor with and without the ionic lock Thus advanced multiscale dynamics methods are formed between Arg3.50 and Glu6.30 (Dror et al. required to study the activation mechanism of 2009). More recently, long time scale MD sim- GPCRs. ulations on the pathway of ligand entry into “2- adrenergic receptor showed that the ligand ini- tially binds to the extracellular loop region where 3.2.3 Multiscale Method they shed the water molecules and hence show Simulations of Activation a finite lifetime in this region. Following this of GPCRs event the ligand trickles down into the orthosteric binding site. The pathway is the same for agonists Multiscale simulation methods comprise of using and antagonists tested in these simulations (Dror (1) a coarse grained MD technique that samples et al. 2011). Although all-atom MD simulations multiple conformational states by overcoming provide valuable insights into atomic level mech- energy barriers, (2) followed by finer scale all- anism of the early events of GPCR activation, atom MD simulations starting from conforma- there still exists a bottleneck in simulating long tions extracted from the coarse grained simula- time scale biological processes such as GPCR tions (see Fig. 3.5). In here, by “coarse grained” activation using all atom MD simulations. Start- we mean a wider and enhanced sampling dy- ing from the inactive state of the receptor, it namics method than all-atom MD simulations is not feasible to sample the active state using and not necessarily the coarse grained forcefield. 3 Structure and Dynamics of G-Protein Coupled Receptors 47
Steered MD (Isralewitz et al. 2001), Acceler- inactive state (Niesen et al. 2011). Subsequent ated MD (Hamelberg et al. 2007), metadynamics NMR experiments studying the ensemble MD (Provasi et al. 2011), methods are examples of conformations without the ligand present of methods that sample multiple conformational have actually shown that the “2-adrenergic states of proteins. Both Steered MD (Gouldson receptor without any ligand is lot more flexible et al. 2004) and metadynamics simulations have and samples multiple states, than when a ligand been used to study the ligand specific confor- is bound (Nygaard et al. 2013). This is an mational states of “2-adrenergic receptor. This example, where the computational methods method provides insight into how ligands of var- provided predictions and insights into the ied efficacies modulate the free energy surface conformational sampling of the receptor without of the “2-adrenergic receptor. The method how- any ligand prior to the experiments. Binding ever requires prior knowledge of the active and of agonist norepinephrine or partial agonist inactive states of the receptor as well as pre- salbutamol leads to selection of a subset of determined activation pathways (Provasi et al. these conformations sampled by the receptor, 2011). while inverse agonist carazolol selects only For cases where the active state of the inactive state conformations. This demonstrates receptor is not known, we developed a multiscale the usefulness of the computational methods dynamics approach by combining a coarse- to provide insights that are difficult to probe grained discrete conformational sampling experimentally. Thus we conclude that ligand method, called LITICon (Bhattacharya et al. binding in GPCRs lead to conformational 2008a, b) with fine-grained molecular dynamics selection to a greater extent than the induced investigating the effect of various ligand binding conformational changes. However the role of on the ensemble of conformations sampled by ligand induced receptor movements cannot be human “2-adrenergic receptor (Bhattacharya eliminated completely. and Vaidehi 2010; Niesen et al. 2011). The discrete conformational sampling method allows traversing over energy barriers and 3.3 Insights into the Structural samples multiple conformations of the receptor Basis of Thermostability (Fig. 3.5). We showed that the receptor, in the in GPCRs absence of any ligand samples an extensive conformational space that includes a shear 3.3.1 Thermostabilization of GPCR motion of the intracellular regions of TM5 Conformations and TM6 and a breathing motion of the ligand binding site. This shear motion is similar to The dynamic nature of GPCRs confers its adapt- the reorganization of the TM helices observed ability and versatility to get activated by multiple in the crystal structure of the active state ligands and trigger multiple signaling pathways of “2-adrenergic receptor (Rasmussen et al. in the cell. At the same time the dynamic nature 2011). However the extent of the movement has been a great impediment and challenge to in the intracellular regions of TM5 and TM6 purify GPCRs in detergent solution and amenable with respect to TM3, is considerably less to crystallization. Although membrane proteins when no ligand is present compared to when are extremely stable in bilayer environment they the agonist and G-protein are bound to the are often unstable when solubilized in deter- receptor. Thus, the computational data prior gents (Bowie 2001). Two major strategies have to experiments lead to the prediction that the been used to circumvent this challenge. One is receptor samples “active-like” conformational to attach T4L or antibody to the loop regions states even in the absence of any ligand, and (Hanson and Stevens 2009; Steyaert and Kobilka the population of these “active-like” states is 2011) and the other is to derive thermostable mu- much less compared to the population of the tants that remain stable in detergents (Tate 2012). 48 N. Vaidehi et al.
These two strategies have proved successful in in purifying these proteins. There are multiple obtaining GPCR crystal structures. The experi- challenges in obtaining purified GPCRs for mental procedure for deriving the thermostable structural studies. Some of these challenges mutant GPCRs is tedious and therefore an under- are: (1) Low level of expression of mammalian standing of how do these thermostable mutations GPCRs, (2) optimization of detergents for GPCR affect the stability of GPCRs is unknown. In this solubilization. When solubilized in detergents section we will focus on the thermostable mu- for purification, GPCRs tend to aggregate (Tate tants and demonstrate how computational meth- 2012; Grisshammer 2009;Chiuetal.2008) and, ods have been used to understand the effect of (3) The flexibility of GPCR conformations poses thermostable mutations on the stability of the the greatest challenge for the receptor purification receptor structures. due to the conformational heterogeneity. Rational Membrane proteins show mutations that en- single point mutations to engineer stabilizing hance the protein stability occurring with a high disulfide bridges (Standfuss et al. 2007), or frequency, thus showing that these wild type mutating certain non-conserved residues that proteins are probably not optimized for high sta- would strengthen helical interfaces were tested bility (Bowie 2001). Thermostable mutants have (Roth et al. 2008) for producing thermostable been engineered for GPCRs by time consuming mutants of rhodopsin and “2-adrenergic receptor. processes as described in detail in the next sec- Although these strategies were successful, tion. Various types of assays have been used to they could not be generalized and applied to measure thermostability for membrane proteins any GPCR. Tate and coworkers developed a (Kawate and Gouaux 2006; Alexandrov et al. systematic procedure for deriving thermostable 2008). Among these techniques, the radiolabelled GPCR mutants (Serrano-Vega et al. 2008; ligand binding is the most used technique for Magnani et al. 2008; Shibata et al. 2009;Tate GPCRs (Tate 2012). Thermostable mutants have 2012; Cooke et al. 2013). This procedure consists lead to crystal structures for turkey “1-adrenergic of several steps: receptor (Warne et al. 2008), human adenosine ¥ Ala/Leu scanning mutagenesis where every A2A receptor (Doré et al. 2011, Lebon et al. residue was mutated to Ala (except for Ala 2011a, b) and neurotensin receptor 1 (White residues that were mutated to Leu), and each et al. 2012). Analysis of the amino acid positions mutant was expressed and its thermostability that lead to thermostability shows that there is was determined relative to the wild type re- no sequence consensus in the positions among ceptor. The thermostability was determined by various receptors, nor there is clustering of the measuring the antagonist binding affinity at thermostable mutant positions. Thus the process high temperatures. of engineering thermostable GPCR mutants is te- ¥ Once thermostabilizing single point mutations dious and time consuming. Therefore a molecular were identified, these positions were mutated level understanding of the effect of thermostable to other amino acid residues to identify im- mutations would throw light into rational engi- provements in thermostability. neering of mutants for other GPCRs. ¥ The best thermostabilizing mutations were then combined to give the optimally stable mutant. 3.3.2 Structural Basis This is a systematic but time-consuming pro- of Thermostability cess, and most of the mutation positions are from Computational Methods not transferable from one GPCR to another and this fact makes it more tedious (Serrano-Vega The fact that GPCRs are dynamic and exhibit and Tate 2009). Insights into why certain mu- multiple active and inactive states have been tations stabilize the receptor, as well as why demonstrated amply with experimental and certain mutations specifically stabilize agonist computational data. This poses a great challenge binding versus antagonist binding (Lebon et al. 3 Structure and Dynamics of G-Protein Coupled Receptors 49
Fig. 3.6 Schematic representation of the receptor population distribution in wt-“1AR (blue) and thermostable mutant m23-“1AR (red). Note that in the mutant m23-“1AR the receptor conformational population is shifted towards the inactive state of the receptor
1.59 2011b) states would greatly aid the design pro- “1AR has six point mutations namely R68S , cess (Balaraman et al. 2010). M90V2.53, Y227A5.58, A282LICL3, F327A7.37 and The structural basis of thermostability of sol- F338M7.48. Using long time scale all-atom MD uble proteins has been studied extensively from simulations in explicit lipid bilayer and water, the perspectives of structure, biophysics and ther- we have studied the effect of the mutations on modynamics (Sterpone and Melchionna 2012). In the stability and dynamics of the m23-“1AR as contrast, studies on membrane protein stability compared to the wild type “1AR referred to as are less common, which is probably because wt-“1AR hereafter (Niesen et al. 2013). they are far less tractable experimentally (White We found that thermostabilization results in an and Wimley 1999;Bowie2011). Measurements increase in the number of accessible microscopic of reversible folding and unfolding processes to conformational substates arising from side chain determine the thermodynamic and kinetic prop- fluctuations within the inactive state ensemble, erties of folding have been done only for a few effectively increasing the entropy of the inactive membrane proteins (Booth and Curnow 2009). state. Figure 3.6 shows the schematic represen- These measurements have not yet been possible tation of the energy landscape of the inactive to study the role of thermostabilizing mutations in state of the receptor for both wild type and m23 GPCRs. Therefore computational methods offer a mutant. The number of states shown in the figure complementary and feasible solution to study the comes from computational data (Niesen et al. effect of thermostable mutations on the structure 2013). It is seen that the number of microscopic and dynamics of GPCRs. states in the mutant m23-“1AR has increased at room temperature. The thermostable mutant m23-“1AR showed an increase in side chain 3.3.3 Insights into the Structural movement within the inactive state (multiple mi- Basis of Thermostability croscopic states within the inactive state). This from Computational Studies increase in side chain flexibility in the m23-“1AR plays a critical role in absorbing the extra ther- Serrano-Vega et al. derived a thermally stable mu- mal energy at higher temperatures, thus retain- tant of the inactive state of turkey “1-adrenergic ing the native fold even at higher temperatures receptor (“1AR) and showed that a combina- thus preventing unfolding and aggregation. Such tion of six single point mutations resulted in a entropic stabilization has been observed in sol- ı 20 C increase in thermal stability of the inactive uble thermophilic proteins (Colombo and Merz state of “1AR (Serrano-Vega et al. 2008). This 1999; Hernandez et al. 2000; Wintrode et al. inactive state mutant of “1AR known as m23- 2003; Meinhold et al. 2008). The dynamics of the 50 N. Vaidehi et al.
mutant m23-“1AR also shows loss of correlated leads not only stress relieving stabilization but motion between residues that are in the ligand also loss of activation of the receptor. P3056.50 binding site to the G-protein coupling site and on TM6 showed a high level of stress in the this reduces the ability of the receptor to go into wild type receptor. P3056.50 forms the hinge for the active state. The thermostabilizing mutations, the kink in TM6 and modulation of the proline Y227A5.58 and F338M7.48 stabilize the inactive kink could be crucial for the outward motion of state by removing long range correlated move- TM6 leading to activation (Hornak et al. 2010; ment between residues in the G-protein coupling Bhattacharya et al. 2008a). Thus reduced stress site to the extracellular region thus inhibiting ac- at P3056.50 could lead to increased stability of the 5.58 tivation (Niesen et al. 2013). The Y227A mu- inactive state in m23-“1AR. tation also stabilized the receptor by increasing Thus the basis of thermostabilization of the the propensity for salt-bridge formation between mutants are multifold. They stem from both en- R1393.50 on TM3 and E2856.30 on TM6. The salt- thalpic and entropic contributions. They not only bridge between TM3 and TM6, a putative sig- affect the neighborhood residues of the location nature of the inactive state (Hubbell et al. 2003; of the mutation sites, but also show long range Yao et al. 2006; Palczewski et al. 2000), is sta- allosteric effects upon mutation. Thus studying bilized by these mutations and hence contribute the effect of the mutations on the overall structure to increase in stability. The mutation F338M7.48 and the conformational ensemble is important on TM7 alters the interaction in the conserved rather than looking for specific changes in mi- motif NPxxY(x)5,6F and hence modulates the crodomains of protein structures. interaction of TM2-TM7-helix8 micro-domain (Balaraman et al. 2010). The R68S mutation that is present in the intracellular loop 1, leads to 3.4 Conclusions formation of a hydrogen bond with R355 on helix 8 to gain stabilization. We observed the formation GPCRs are seven helical transmembrane proteins of the hydrogen bond in the simulations and that play a critical physiological role and this was also found subsequently in the crystal are implicated in many serious diseases. The structure of the mutant (Warne et al. 2012). dynamics of GPCR structure plays an important The distribution of the net intra and inter- role in their activation and signal transduction molecular forces, intramolecular repulsive and function. Here we have described a method, attractive forces within a protein structure gov- GPCRCompare to compare the three dimensional erns how the protein reacts to thermal stress, structures of GPCRs in terms of the relative since non-optimal distribution of internal stress rotational, tilt and translational orientations of the can lead to structural destabilization at elevated seven helices. This software is freely available for temperatures (Stadler et al. 2012). However, in- users by contacting the corresponding author. ternal stress also leads to movement of those Comparison of the ten NMR structures of regions of the receptor that lead to the active state. the chemokine receptor shows a significantly Thus a balance of these internal forces is required dynamic range of rotation and tilt angles for to maintain receptor stability and activity. The transmembrane helices 5, 6 and 7. The kink distribution of the net force on each residue angle on TM6 ranges from 21ı to 34ı showing calculated from the snapshots from the MD sim- significant dynamics in the ten NMR structures. ulations of both m23-“1AR and wt-“1AR showed Comparison of the crystal structures of the that Y2275.58, F3277.37 and F3387.48 are all in po- biogenic amine receptors available to date, sitions of high stress in wt-“1AR that get relieved showed that the conserved amino acids face upon mutation, leading to thermostabilization. the interior of the transmembrane barrel. We Y2275.58 is also a position that shows long range have also elaborated the structural dynamics and torsional correlation to the extracellular regions activation mechanism of “2-adrenergic receptor of the receptor. Therefore mutating this residue as derived from multiscale MD simulation meth- 3 Structure and Dynamics of G-Protein Coupled Receptors 51 ods. We observed that the “2-adrenergic receptor Bowie JU (2001) Stabilizing membrane proteins. Curr samples conformations with active state like Opin Struct Biol 11:397Ð402 characteristics and the population of such states Bowie JU (2011) Membrane protein folding: how im- portant are hydrogen bonds? Curr Opin Struct Biol are much lower than those of the inactive state. 21:42Ð49 Ligand binding leads to conformational selection Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen of substates (Niesen et al. 2011). Finally, study SG, Thian FS, Kobilka TS, Choi HJ, Kuhn P, Weis WI, Kobilka BK, Stevens RC (2007) High-resolution of the dynamics of the conformational ensemble crystal structure of an engineered human beta2- of thermostable mutants showed that the stability adrenergic G-protein coupled receptor. Science 318: could arise from both enthalpic and entropic 1258Ð1265 factors. There are regions of high stress in the ChienEY,LiuW,ZhaoQ,KatritchV,HanGW,Hanson MA, Shi L, Newman AH, Javitch JA, Cherezov V, wild type GPCR that get relieved upon mutation Stevens RC (2010) Structure of the human dopamine conferring thermostability. D3 receptor in complex with a D2/D3 selective antag- onist. Science 330:1091Ð1095 Acknowledgements We acknowledge the helpful dis- Chiu ML, Tsang C, Grihalde N, MacWilliams MP (2008) cussions with Dr. Christopher Tate and Dr. Reinhard Over-expression, solubilization and purification of G- Grisshammer. We thank NIH-R01097261 for providing protein coupled receptors for structural biology. Comb funding for our study on GPCR thermostable mutants. Chem High Throughput Screen 11:439Ð462 Colombo G, Merz KM (1999) Stability and activity of mesophilic subtilisin E and its thermophilic homolog: insights from molecular dynamics simulations. J Am References Chem Soc 121:6895Ð6903 Cooke RM, Koglin M, Errey JC, Marshall FH (2013) Ahuja S, Smith SO (2009) Multiple switches in G protein- Preparation of purified GPCRs for structural studies. coupled receptor activation. Trends Pharmacol Sci Biochem Soc Trans 41(1):185Ð190 30:494Ð502 Crozier PS, Stevens MJ, Woolf TB (2007) How a small Alexandrov AI, Mileni M, Chien EY, Hanson MA, change in retinal leads to G-protein activation: initial Stevens RC (2008) Microscale fluorescent thermal events suggested by molecular dynamics calculations. stability assay for membrane proteins. Structure Proteins 66:559Ð574 16:351Ð359 Deupi X (2012) Quantification of structural distortions in Audet M, Bouvier M (2012) Restructuring G-protein the transmembrane helices of GPCRs. Methods Mol coupled receptor activation. Cell 151:14Ð23 Biol 914:219Ð235 Balaraman GS, Bhattacharya S, Vaidehi N (2010) Struc- Doré AS, Robertson N, Errey JC, Ng I, Hollenstein K, tural insights into conformational stability of wild Tehan B, Hurrell E, Bennett K, Congreve M, Mag- type and mutant beta1-adrenergic receptor. Biophys J nani F, Tate CG, Weir M, Marshall FH (2011) Struc- 99:568Ð577 ture of the adenosine A2A receptor in complex with Ballesteros J, Weinstein J (1995) Integrated methods for ZM241385 and xanthines XAC and caffeine. Structure modeling G-protein coupled receptors. Methods Neu- 19(9):1283Ð1293 rosci 25:366Ð428 Dror RO, Arlow DH, Borhani DW, Jensen M¯, Piana S, Bhattacharya S, Vaidehi N (2010) Computational map- Shaw DE (2009) Identification of two distinct inactive ping of the conformational transitions in agonist selec- conformations of the beta2-adrenergic receptor rec- tive pathways of a G-protein coupled receptor. J Am onciles structural and biochemical observations. Proc Chem Soc 132(14):5205Ð5214 Natl Acad Sci U S A 106:4689Ð4694 Bhattacharya S, Hall SE, Vaidehi N (2008a) Agonist- Dror RO, Pan AC, Arlow DH, Borhani DW, Maragakis P, induced conformational changes in bovine rhodopsin: Shan Y, Xu H, Shaw DE (2011) Pathway and mecha- insight into activation of G-protein-coupled receptors. nism of drug binding to G-protein coupled receptors. J Mol Biol 382:539Ð555 Proc Natl Acad Sci U S A 108:13118Ð13123 Bhattacharya S, Hall SE, Li H, Vaidehi N (2008b) Ligand- Filizola M, Devi LA (2013) Grand opening of structure- stabilized conformational states of human beta(2) guided design for novel opioids. Trends Pharmacol Sci adrenergic receptor: insight into G-protein-coupled 34(1):6Ð12 receptor activation. Biophys J 94(6):2027Ð2042 Galandrin S, Oligny-Longpré G, Bouvier M (2007) The Bhattacharya S, Lam AR, Li H, Balaraman G, Niesen MJ, evasive nature of drug efficacy: implications for drug Vaidehi N (2013) Critical analysis of the successes and discovery. Trends Pharmacol Sci 2007(28):423Ð430 failures of homology models of G protein-coupled re- Galés C, Rebols RV, Hogue M, Trieu P, Brelt A, Hébert ceptors. Proteins 81:729Ð739. doi:10.1002/prot.24195 TE, Bouvier M (2005) Real-time monitoring of re- Booth PJ, Curnow P (2009) Folding scene investigation: ceptor and G-protein interactions in living cells. Nat membrane proteins. Curr Opin Struct Biol 19(1):8Ð13 Methods 2:177Ð184 52 N. Vaidehi et al.
Gouldson PR, Kidley NJ, Bywater RP, Psaroudakis G, Kawate T, Gouaux E (2006) Fluorescence-detection Brooks HD, Diaz C, Shire D, Reynolds CA (2004) size exclusion chromatography for precrystallization Toward the active conformations of rhodopsin and the screening of integral membrane proteins. Structure beta2-adrenergic receptor. Proteins 56(1):67Ð84 14:673Ð681 Granier S, Kobilka B (2012) A new era of GPCR structural Kenakin T, Watson C, Muniz-Medina V, Christopoulos and chemical biology. Nat Chem Biol 8:670Ð673 A, Novick S (2012) A simple method for quantifying Granier S, Manglik A, Kruse AC, Kobilka TS, Thian functional selectivity and agonist bias. ACS Chem FS, Weis WI, Kobilka BK (2012) Structure of Neurosci 3:193Ð203 the •-opioid receptor bound to naltrindole. Nature Khelashvili G, Grossfield A, Feller SE, Pitman MC, 485(7398):400Ð404 Weinstein H (2009) Structural and dynamic effects Grisshammer R (2009) Purification of recombinant of cholesterol at preferred sites of interaction with G-protein coupled receptors. Methods Enzymol rhodopsin identified from microsecond length molec- 463:631Ð645 ular dynamics simulations. Proteins 76:403Ð417 Grossfield A, Pittman MC, Feller SE, Soubias O, Krishna A, Menon ST, Terry TJ, Sakmar TP (2002) Gawrisch K (2008) Internal hydration increases dur- Evidence that helix 8 of rhodopsin acts as a ing activation of the G-protein-coupled receptor membrane-dependent conformational switch. Bio- rhodopsin. J Mol Biol 381:478Ð486 chemistry 41:8298Ð8309 Haga K, Kruse AC, Asada H, Yurugi-Kobayashi T, Shi- Kruse AC, Hu J, Pan AC, Arlow DH, Rosenbaum DM, roishi M, Zhang C, Weis WI, Okada T, Kobilka BK, Rosemond E, Green HF, Liu T, Chae PS, Dror RO, Haga T, Kobayashi T (2012) Structure of the human Shaw DE, Weis WI, Wess J, Kobilka BK (2012) Struc- M2 muscarinic acetylcholine receptor bound to an ture and dynamics of the M3 muscarinic acetylcholine antagonist. Nature 482:547Ð551 receptor. Nature 482:552Ð556 Hall SE, Mao A, Nicolaidou V, Finelli M, Wise EL, Nedjai Lebon G, Bennett K, Jazayeri A, Tate CG (2011a) Ther- B, Kanjanapangka J, Harirchian P, Chen D, Selchau V, mostabilization of an agonist bound conformation Ribeiro S, Schyler S, Pease JE, Horuk R, Vaidehi N of the human adenosine A2A receptor. J Mol Biol (2009) Elucidation of binding sites of dual antagonists 409:298Ð310 in the human chemokine receptors CCR2 and CCR5. Lebon G, Warne T, Edwards PC, Bennett K, Langmead Mol Pharm 75:1325Ð1336 CJ, Leslie AG, Tate CG (2011b) Agonist bound adeno- Hamelberg D, de Oliveira CA, McCammon JA (2007) sine A2A receptor structures reveal common features Sampling of slow diffusive conformational transitions of GPCR activation. Nature 474:521Ð525 with accelerated molecular dynamics. J Chem Phys Magnani F, Shibata Y, Serrano-Vega MJ, Tate CG (2008) 127(15):155102Ð155110 Co-evolving stability and conformational homogene- Hanson MA, Stevens RC (2009) Discovery of new GPCR ity of the human adenosine A2a receptor. Proc Natl biology: one receptor structure at a time. Structure Acad Sci U S A 105(31):10744Ð10749 17(1):8Ð14 Mailman RB, Murthy V (2010) Ligand functional se- Hernandez G, Jenney FE, Adams MWW, LeMaster DM lectivity advances our understanding of drug mecha- (2000) Millisecond time scale conformational flexibil- nisms and drug discovery. Neuropsychopharmacology ity in a hyperthermophile protein at ambient tempera- 35:345Ð346 ture. Proc Natl Acad Sci U S A 97:3166Ð3170 Manglik A, Kruse AC, Kobilka TS, Thian FS, Mathiesen Hornak V, Ahuja S, Eilers M, Goncalves JA, Sheves JM, Sunahara RK, Pardo L, Weis WI, Kobilka BK, M et al (2010) Light activation of rhodopsin: in- Granier S (2012) Crystal structure of the -opioid sights from molecular dynamics simulations guided by receptor bound to a morphinan antagonist. Nature solid-state NMR distance restraints. J Mol Biol 396: 485(7398):321Ð326 510Ð527 Meinhold L, Clement D, Tehei M, Daniel R, Finney Hubbell WL, Altenbach C, Hubbell CM, Khorana HG JL et al (2008) Protein dynamics and stability: the (2003) Rhodopsin structure, dynamics, and activation: distribution of atomic fluctuations in thermophilic a perspective from crystallography, site directed spin and mesophilic dihydrofolate reductase derived us- labeling, sulfhydryl reactivity, and disulfide cross- ing elastic incoherent neutron scattering. Biophys linking. Adv Protein Chem 63:243Ð290 J 94:4812Ð4818 Isralewitz B, Gao M, Schulten K (2001) Steered molecular Nedjai B, Li H, Stroke IL, Wise EL, Webb ML, Merritt dynamics and mechanical functions of proteins. Curr JR, Henderson I, Klon AE, Cole AG, Horuk R, Vaidehi Opin Struct Biol 11:224Ð230 N, Pease JE (2011) Small-molecule chemokine mimet- Johnston JM, Aburi M, Provasi D, Bortolato A, Urizar ics suggest a molecular basis for the observation E, Lambert NA, Javitch JA, Filizola M (2011) that CXCL10 and CXCL11 are allosteric ligands of Making structural sense of dimerization interfaces CXCR3. Br J Pharmacol 166(3):912Ð923 of delta opioid receptor homodimers. Biochemistry Niesen M, Bhattacharya S, Vaidehi N (2011) Conforma- 50(10):1682Ð1690 tional selection upon ligand binding in G-protein cou- Katritch V, Cherezov V, Stevens RC (2012) Diversity and pled receptors. J Am Chem Soc 133(33):13197Ð13204 modularity of G-protein coupled receptor structures. Niesen M, Bhattacharya S, Tate CG, Grisshammer R, Trends Pharmacol Sci 33:17Ð27 Vaidehi N (2013) Thermostabilization of the “1- 3 Structure and Dynamics of G-Protein Coupled Receptors 53
adrenergic receptor Correlates with Increased Entropy Shimamura T, Shiroishi M, Weyand S, Tsujimoto H, of the Inactive State. J Phys Chem 117(24):7283Ð7291 Winter G, Katritch V, Abagyan R, Cherezov V, Liu W, Nygaard R, Zou Y, Dror RO, Mildorf TJ, Arlow DH, Han GW, Kobayashi T, Stevens RC, Iwata S (2011) Manglik A, Pan AC, Liu CW, Fung JJ, Bokoch MP, Structure of the human histamine H1 receptor complex Thian FS, Kobilka TS, Shaw DE, Mueller L, Prosser with doxepin. Nature 475:65Ð70 RS, Kobilka BK (2013) The dynamic process of “2- Stadler AM, Garvey CJ, Bocahut A, Sacquin-Mora adrenergic receptor activation. Cell 152(3):532Ð542 S, Digel I et al (2012) Thermal fluctuations of Palczewski K, Kumasaka T, Hori T, Behnke CA, haemoglobin from different species: adaptation to Motoshima H, Fox BA, Le Trong I, Teller DC, Okada temperature via conformational dynamics. J R Soc T, Stenkamp RE, Yamamoto M, Miyano M (2000) Interface 9(76):2845Ð2855 Crystal structure of rhodopsin: a G protein-coupled Standfuss J, Xie G, Edwards PC, Burghammer M, receptor. Science 289:739Ð745 Oprian DD, Schertler GF (2007) Crystal structure Park JH, Scheerer P, Hoffman KP, Choe H-W, Ernst OP of a thermally stable rhodopsin mutant. J Mol Biol (2008) Crystal structure of the ligand-free G-protein 372(5):1179Ð1188 coupled receptor opsin. Nature 454:183Ð187 Sterpone F, Melchionna S (2012) Thermophilic proteins: Park SH, Das BB, Casagrande F, Tian Y, Nothnagel HJ, insight and perspective from in silico experiments. Chu M, Kiefer H, Maier K, De Angelis AA, Marassi Chem Soc Rev 41(5):1665Ð1676 FM, Opella SJ (2012) Structure of the chemokine Steyaert J, Kobilka BK (2011) Nanobody stabilization receptor CXCR1 in phospholipid bilayers. Nature of G-protein coupled receptor conformational states. 491(7426):779Ð783 Curr Opin Struct Biol 21:567Ð572 Provasi D, Artacho MC, Negri A, Mobarec JC, Filizola M Swaminath G, Xiang Y, Lee TW, Steenhuis J, Parnot (2011) Ligand induced modulation of the free energy C, Kobilka BK (2004) Sequential binding of ag- landscape of the G-protein coupled receptors explored onists to the beta2 adrenoceptor-kinetic evidence by adaptive biasing techniques. PLoS Comput Biol for intermediate conformational states. J Biol Chem 7(10):e1002193 279:686Ð691 Rasmussen SG, DeVree BT, Zou Y, Kruse AC, Chung Swaminath G, Deupi X, Lee TW, Zhu W, Thian FS, KY, Kobilka TS, Thian FS, Chae PS, Pardon E, Calin- Kobilka TS, Kobilka BK (2005) Probing the beta2 ski D, Mathiesen JM, Shah ST, Lyons JA, Caffrey adrenoceptor binding site with catechol reveals differ- M, Gellman SH, Steyaert J, Skiniotis G, Weis WI, ences in binding and activation by agonists and partial Sunahara RK, Kobilka BK (2011) Crystal structure of agonists. J Biol Chem 280:22165Ð22171 the “2-adrenergic receptor-Gs protein complex. Nature Tate CG (2012) A crystal clear solution for determining G- 477:549Ð555 protein coupled receptor structures. Trends Biochem Reiter E, Ahn S, Shukla AK, Lefkowitz RJ (2012) Molec- Sci 37:343Ð352 ular mechanism of “-arrestin biased agonism at seven Tate CG, Schertler GF (2009) Engineering G protein- transmembrane receptors. Ann Rev Pharmacol Toxicol coupled receptors to facilitate their structure determi- 52:179Ð197 nation. Curr Opin Struct Biol 19:386Ð395 Roth CB, Hanson MA, Stevens RC (2008) Stabilization Thompson AA, Liu W, Chun E, Katritch V, Wu H, Vardy of the human beta2-adrenergic receptor TM4-TM3- E, Huang XP, Trapella C, Guerrini R, Calo G, Roth TM5 helix interface by mutagenesis of Glu122(3.41), BL, Cherezov V, Stevens RC (2012) Structure of the a critical residue in GPCR structure. J Mol Biol nociceptin/orphanin FQ receptor in complex with a 376(5):1305Ð1319 peptide mimetic. Nature 485(7398):395Ð399 Saam J, Tajkhorshid E, Havashi S, Schulten K (2002) Vaidehi N (2010) Dynamics and flexibility of G-protein Molecular dynamics investigation of primary photoin- coupled receptor conformations and their relevance to duced events in the activation of rhodopsin. Biophys drug design. Drug Discov Today 15:951Ð957 J 83:3097Ð3112 Vaidehi N, Bhattacharya S (2011) Multiscale computa- Schertler GF (2005) Structure of rhodopsin and the tional methods for mapping conformational ensembles metarhodopsin I photointermediate. Curr Opin Struct of G-protein coupled receptors. Adv Protein Chem Biol 15:408Ð15 Struct Biol 85:253Ð280 Serrano-Vega MJ, Tate CG (2009) Transferability of Vaidehi N, Kenakin T (2010) The role of confor- thermostabilizing mutations between beta adrenergic mational ensembles of seven transmembrane recep- receptors. Mol Membr Biol 26:385Ð396 tors in functional selectivity. Curr Opin Pharmacol Serrano-Vega MJ, Magnani F, Shibata Y, Tate CG (2008) 10:775Ð781 Conformational thermostabilization of the beta1- Vilardaga JP, Bunemann M, Krasel C, Castro M, Lohse adrenergic receptor in a detergent resistant form. Proc M (2003) Measurement of the millisecond activation Natl Acad Sci U S A 105(3):877Ð882 switch of G protein-coupled receptors in living cells. Shibata Y, White JF, Serrano-Vega MJ, Magnani F, Aloia Nat Biotechnol 21:807Ð812 AL, Grisshammer R, Tate CG (2009) Thermostabi- Warne T, Serrano-Vega MJ, Baker JG, lization of the neurotensin receptor NTS1. J Mol Biol Moukhametzianov R, Edwards PC, Henderson 2390(2):262Ð277 R, Leslie AG, Tate CG, Schertler GF (2008) Structure 54 N. Vaidehi et al.
of a beta1-adrenergic G-protein-coupled receptor. Xu F, Wu H, Katritch V, Han GW, Jacobson KA, Gao ZG, Nature 454(7203):486Ð491 Cherezov V, Stevens RC (2011) Structure of an ag- Warne T, Edwards PC, Leslie AGW, Tate CG (2012) Crys- onist bound human A2A adenosine receptor. Science tal structures of a stabilized “1-adrenoceptor bound to 332(6027):322Ð327 the biased agonists bucindolol and carvedilol. Struc- Yao X, Parnot C, Deupi X, Ratnala VRP, Swami- ture 20:841Ð849 nath G, Farrens D, Kobilka BK (2006) Coupling White SH, Wimley WC (1999) Membrane protein folding ligand structure to specific conformational switches and stability: physical principles. Annu Rev Biophys in the beta2-adrenoceptor. Nat Chem Biol 2006(2): Biomol Struct 28:319Ð365 417Ð422 White JF, Noinaj N, Shibata Y, Love J, Kloss B, Xu F, Yao XJ, Ruiz G, Whorton MR, Rasmussen SG, De Vree Gvozdenovic-Jeremic J, Shah P, Shiloach J, Tate CG, BT, Deupi X, Sunahara RK, Kobilka B (2009) The Grisshammer R (2012) Structure of the agonist bound effect of ligand efficacy on the formation and stability neurotensin receptor. Nature 490:508Ð513 of a GPCR-G-protein complex. Proc Natl Acad Sci U Wintrode PL, Zhang DQ, Vaidehi N, Arnold FH, God- S A 106:9501Ð9506 dard WA (2003) Protein dynamics in a family of Zaitseva E, Brown MF, Vogel R (2010) Sequential rear- laboratory evolved thermophilic enzymes. J Mol Biol rangement of interhelical networks upon rhodopsin ac- 327:745Ð757 tivation in membranes: the Meta II(a) conformational Wu B, Chien EY, Mol CD, Fenalti G, Liu W, Katritch substate. J Am Chem Soc 132:4815Ð4821 V, Abagyan R, Brooun A, Wells P, Bi FC, Hamel Zhang C, Srinivasan Y, Arlow DH, Fung JJ, Palmer DJ, Kuhn P, Handel TM, Cherezov V, Stevens RC D, Zheng Y, Green HF, Pandey A, Dror RO, Shaw (2010) Structures of the CXCR4 chemokine GPCR DE, Weis WI, Coughlin SR, Kobilka BK (2012) with small-molecule and cyclic peptide antagonists. High-resolution crystal structure of human protease- Science 330(6007):1066Ð1071 activated receptor 1. Nature 492(7429):387Ð392 How the Dynamic Properties and Functional Mechanisms of GPCRs 4 Are Modulated by Their Coupling to the Membrane Environment
Sayan Mondal, George Khelashvili, Niklaus Johner, and Harel Weinstein
Abstract Experimental observations of the dependence of function and organization of G protein-coupled receptors (GPCRs) on their lipid environment have stimulated new quantitative studies of the coupling between the proteins and the membrane. It is important to develop such a quantitative understanding at the molecular level because the effects of the coupling are seen to be physiologically and clinically significant. Here we review findings that offer insight into how membrane-GPCR coupling is connected to the structural characteristics of the GPCR, from sequence to 3D structural detail, and how this coupling is involved in the actions of ligands on the receptor. The application of a recently developed computational approach designed for quantitative evaluation of membrane remodeling and the energetics of membrane-protein interactions brings to light the importance of the radial asymmetry of the membrane-facing surface of GPCRs in their interaction with the surrounding membrane. As the radial asymmetry creates adjacencies of hydrophobic and polar residues at specific sites of the GPCR, the ability of membrane remodeling to achieve complete hydrophobic matching is limited, and the residual mismatch carries a significant energy cost. The adjacencies are shown to be affected by ligand-induced conformational changes. Thus, functionally important organization of GPCRs in the cell membrane can depend both on ligand-determined properties
S. Mondal ¥ G. Khelashvili ¥ N. Johner Department of Physiology and Biophysics, Weill Cornell Medical College, Cornell University, Room E-509, 1300 York Avenue, 10065 New York City, NY, USA The HRH Prince Alwaleed Bin Talal Bin Abdulaziz H. Weinstein () Alsaud Institute for Computational Biomedicine, Weill Department of Physiology and Biophysics, Weill Cornell Cornell Medical College, Cornell University, New York, Medical College, Cornell University, Room E-509, 1300 NY 10065, USA York Avenue, 10065 New York City, NY, USA e-mail: [email protected]
M. Filizola (ed.), G Protein-Coupled Receptors - Modeling and Simulation, Advances 55 in Experimental Medicine and Biology 796, DOI 10.1007/978-94-007-7423-0__4, © Springer ScienceCBusiness Media Dordrecht 2014 56 S. Mondal et al.
and on the lipid composition of various membrane regions with different remodeling capacities. That this functionally important reorganization can be driven by oligomerization patterns that reduce the energy cost of the residual mismatch, suggests a new perspective on GPCR dimerization and ligand-GPCR interactions. The relation between the modulatory effects on GPCRs from the binding of specific cell-membrane components, e.g., cholesterol, and those produced by the non-local energetics of hydropho- bic mismatch are discussed in this context.
Keywords GPCR ¥ Transmembrane proteins ¥ Structure and remodeling of lipid membranes ¥ Hydrophobic mismatch ¥ Lipid/protein interactions ¥ CTMD ¥ Continuum elastic theory ¥ Molecular dynamics simulations ¥ Residual exposure ¥ Residual mismatch ¥ Membrane deformation en- ergy ¥ Oligomerization ¥ Serotonin receptor ¥ Dopamine receptor ¥ Rhodopsin ¥ Cholesterol/GPCR binding ¥ DHA ¥ Structure-function relations in GPCRs ¥ Functional microdomains in GPCRs
Abbreviations Warne et al. 2008; Chien et al. 2010;Wu et al. 2010; Choe et al. 2011; Lebon et al. 5HT 5-hydroxytryptamine 2011; Rosenbaum et al. 2011;Xuetal.2011; CTMD Continuum-Molecular Dynamics Granier et al. 2012;Hagaetal.2012; Hanson hybrid approach et al. 2012; Liu et al. 2012;Wuetal.2012; DHA ¨-3 Docosahexaenoic Acid Manglik et al. 2012). The literature abounds GPCR G protein-coupled receptor in such information, obtained from decades of MD Molecular Dynamics detailed examination of the molecular structural SASA Solvent Accessible Surface Area determinants for the pharmacologically and SDPC 1-stearoyl-2-docosahexaenoyl-sn- physiologically measured ligand-recognition Glycero-3-phosphocholine properties and mechanisms of receptor activation. SDPE 1-stearoyl-2-docosahexaenoyl-sn- Such structure-based details have emerged from Glycero-3-phosphoethanolamine experiments in which the receptors were probed SM/FM Sequence Motif/ Functional with structurally defined ligands and mutations Microdomain to reveal mechanistic details in activation and TM Transmembrane segment dimerization (Ballesteros et al. 2001; Han et al. 2009; Strader et al. 1987; Suryanarayana et al. 1992; Luttrell et al. 1999; Liapakis et al. 2000; 4.1 Introduction Barak et al. 1995; Chelikani et al. 2007; Fritze et al. 2003; Green and Liggett 1994; Guo et al. A rich body of knowledge and information 2005; Javitch et al. 1995a, b, 1998; Kahsai et al. concerning the molecular pharmacology of 2011; Kofuku et al. 2012;O’Dowdetal.1988; the receptors, and their coupling to signal Prioleau et al. 2002; Shi and Javitch 2004; Shi transduction cascades in the cell, is offering a et al. 2002). Together, computational modeling, functional context for the recently determined mutagenesis, and crystallography have shown structures of GPCRs (Palczewski et al. 2000; that each of the five GPCR families share highly Ruprecht et al. 2004; Cherezov et al. 2007; conserved sequence motifs (SMs) that constitute Rasmussen et al. 2007, 2011; Jaakola et al. functional microdomains (FMs) mediating 2008;Parketal.2008; Scheerer et al. 2008; receptor activation (Ballesteros et al. 1998; 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 57
ple the response of specific membrane regions to the transition among such conformations. This coupling is becoming recognized as a determi- nant factor in (i)-the specificity of ligand binding, and (ii)-the responses elicited from the various activated states. Importantly, the effects of the membrane environment are strongly dependent on the lipid composition and hence can vary between different cell types, and also within dif- ferent regions of the plasma membrane of the same cell. For example, cholesterol, a compo- nent of biological membranes with a relative abundance that varies among cell types and even within different regions of the same plasma mem- brane, has been shown to affect GPCR signaling. Functionally important receptor properties such as thermal stability, ligand binding, and local- ization in the cell membrane have been shown to be cholesterol-dependent for beta2-adrenergic, Fig. 4.1 The position of Structural Motifs recognized 5HT1A, oxytocin, and metabotropic glutamate as Functional Microdomains (SM/FMs) in GPCRs, il- receptors (Xiang et al. 2002; Pucadyil and Chat- lustrated in the molecular model of 5HT2AR (Reprinted topadhyay 2004; Gimpl and Fahrenholz 2002; with permission from Shan et al. 2012) Eroglu et al. 2003). Rhodopsin has served as a prototype for elucidating the role of the lipid Visiers et al. 2002; Prioleau et al. 2002; Barak environment. In systematic studies, the activa- et al. 1995; Fritze et al. 2003; Lagerstrom and tion and the photochemical function of rhodopsin Schioth 2008; Rosenbaum et al. 2009; Palczewski were found to be affected by the cholesterol et al. 2000; Warne et al. 2008; Cherezov et al. concentration, as well as by molecular-level char- 2007; Weinstein 2005). Prominent examples of acteristics of the lipids such as their acyl chain these SM/FMs from Class-A GPCRs include the length, unsaturation of the acyl chain, headgroup “arginine cage” around the E/DRY motif in TM3 charge, and headgroup size (Brown 1994;Brown (Ballesteros et al. 1998), and the NPxxY motif in et al. 2002; Botelho et al. 2002, 2006;Gibson TM7 (Ballesteros et al. 1998; Barak et al. 1995; and Brown 1991, 1993). To take the example Fritze et al. 2003; Prioleau et al. 2002). Figure 4.1 of acyl chain unsaturation, the equilibrium be- illustrates some of these SM/FMs in a structural tween the inactive Metarhodopsin I and the active model of the serotonin receptor 5HT2AR. Metarhodopsin II was shown to shift in favor With the focus on the conformational changes of Metarhodopsin II with increasing molar frac- of structural elements of the GPCR proteins, the tion of lipids containing the polyunsaturated ¨- mechanistic models have largely ignored the in- 3 docosahexaenoic (DHA) tail, as compared to volvement of the membrane environment in the monounsaturated lipid tails (Brown 1994). Such discriminant properties of the GPCRs, and the modulation is physiologically and clinically sig- differential ligand-determined effects. This situa- nificant, as dietary deficiency in ¨-3 fatty acids tion is now changing rapidly, because the mem- is well-known to be implicated in a wide vari- brane environment has emerged as an essential ety of diseases (Innis 2008; Lavie et al. 2009; part of functional mechanisms including the ac- Neuringer et al. 1984; Stahl et al. 2008), at tivation, coupling, and cell-surface organization least some of which involve the modulation of of the GPCRs. In particular, the conformations GPCR function by the lipid environment. In rat of the GPCRs in the various states of activation models, for example, deficiency of dietary ¨-3 Ð both ligand-determined and constitutive Ð cou- fatty acids impaired photochemical function, with 58 S. Mondal et al. the Metarhodopsin II formation being both lower coupling. The focus is on the manner in and slower in response to dim light stimulus (Niu which such coupling is (a)-connected to et al. 2002). the structural/sequence characteristics of the As the structures of GPCRs become known receptor, and (b)-involved in the modes and from crystallography at increasingly high consequences of receptor-ligand interactions. resolution, and various biophysical methods of investigation offer increasingly detailed quantitative information about mechanistic 4.2.1 How GPCRs Couple elements of GPCR function, it becomes possible to the Surrounding to focus on the structural context of the mech- Membrane anisms. In parallel, this development requires a corresponding evaluation of the demonstrated Results from experimental investigations and a role of the lipid environment using appropriate variety of theoretical and computational studies quantitative methods to assess the membrane- suggest that two major aspects of the coupling protein coupling components. Achieving this between the membrane and the embedded GPCR goal requires the development and application have the greatest impact on the functional prop- of quantitative approaches, capable of relating erties of the receptors: (1) the interaction of the lipid-protein interactions to the structural individual lipid molecules with specific sites of characteristics of the receptor in its various the receptor molecule, and (2) the mismatch be- states, and especially in the functional context of tween the hydrophobic/hydrophilic character of ligand-induced conformational rearrangements. the membrane and the membrane-facing surface Significant progress in this direction has been of the receptor. As detailed below, a significant achieved with biophysical methods anchored in part of this mismatch persists even after the fundamental experimental data, and formulated membrane is remodeled around the embedded in the framework of physics-based computational protein, giving rise to an energy cost that affects modeling approaches ranging from molecular the functional properties and spatial organiza- dynamics simulations to continuum-based and tion of the GPCRs. This is due to the inherent mean field methods. We review, in Sect. 4.2, asymmetry of the membrane-facing surface of the new types of insights attained from the the embedded protein, caused by the differences application of such approaches. In Sect. 4.3,we among the TM segments. Consequently, it is an present the essential methodological advances important common characteristic of the energet- that enabled quantitative investigation of the ics of membrane coupling for multi-TM segment relation between lipid-protein interactions and proteins. the GPCR structure. To enable clear references to structural elements in different GPCRs, 4.2.1.1 Specific Interactions of Lipid residues and motifs are identified throughout the Molecules chapter with the Ballesteros-Weinstein notation with Membrane-Embedded (Ballesteros and Weinstein 1995). GPCRs With the details obtained from atomistic Molecular Dynamics (MD) simulations it has 4.2 Relation Between Receptor become possible to discern the dynamics of Structure and Its Interaction membrane components surrounding the receptor with the Membrane molecules, such as cholesterol, DHA, and various in a Functional Context other lipids. The simulations revealed both the nature of the interactions of the various lipid This section reviews findings that demonstrate molecules with specific sites of the GPCR, how the membrane interacts with the embedded and the effect on the membrane and the GPCR, i.e., the nature of the membrane-GPCR proteins as a whole. For example, in a 1.6 s 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 59
Fig. 4.2 Cholesterol dynamics around rhodopsin time. Cholesterol is shown in space-fill, and Pro7.38 in from atomistic MD simulation. Cholesterol interacting green. The TM6-TM7 bundle of rhodopsin is shown as with Pro7.38 in TM7. The time-sequence in panel (a) ribbons (a), or as cylinders (b). (c) The angle between is for the z-directional distance between the cholesterol TM7 and H8 over the course of the trajectory. (Reprinted oxygen and the C’ atom of Pro7.38. Panel (b)showsthe with permission from Khelashvili et al. 2009. Copyright angle between the cholesterol ring and the axis of the by Wiley Periodicals Inc.) extracellular segment of TM7 as a function of trajectory all-atom MD simulation of the prototypical (W,Y)] Ð [ 4.46 (I,V,L)] Ð [2.41 (F,Y)] was GPCR rhodopsin, three molecules of cholesterol calculated to occur in 21 % of human class A were found to be localized preferentially at GPCRs (Hanson et al. 2008), suggesting that three distinct sites of the protein (Khelashvili such specific interactions are a general feature of et al. 2009). Similar preferential localization GPCR-membrane coupling. of cholesterol molecules to specific sites was Results from the MD simulations suggested observed in s-scale MD simulations of ligand- as well a mechanistic role for the specific bind- bound serotonin receptor 5-HT2AR (Shan et al. ing of cholesterol. This was based on observed 2012). Such observations of specific cholesterol- correlations between the dynamics of choles- receptor interactions at identified sites in MD terol molecules and the conformations adopted simulations are consistent with structural data by the SM/FM regions of the receptor struc- for several GPCRs (Khelashvili et al. 2009). ture that had been identified for their involve- Indeed, electron microscopy of the inactive ment in activation. For example, the binding photo-intermediate Metarhodopsin I (Ruprecht mode of a cholesterol molecule in the TM6- et al. 2004), and several X-ray structures of TM7 region of rhodopsin was found to cor- GPCRs, show individual cholesterol molecules relate with the angle between TM7 and H8, associated with specific receptor sites (Hanson which involves the NPxxY SM/FM (see Fig. 4.2) et al. 2012; Liu et al. 2012; Cherezov et al. 2007). (Khelashvili et al. 2009). Similarly, in simula- A proposed cholesterol binding motif identified tions of the agonist-bound 5HT2AR (Shan et al. by groups of residues [4.39Ð4.43 (R,K)] Ð [4.50 2012), the movement of cholesterol away from 60 S. Mondal et al.
Fig. 4.3 Cholesterol dynamics correlates with the V7.52, and L7.55 residues on TM7 moves towards TM6 structural transitions in agonist-bound 5-HT2AR.(a) and engages in interactions with the residues K6.35, I6.39, Evolution of the minimum distances between the choles- F6.42, and V6.46 on TM6. (b) Snapshots at 10 and terol at the extracellular (EC) end of TM6 and selected 30 ns showing the cholesterol from the top panel of (a) TM6 residues in the 5-HT simulation (top panel). Time interacting with EC end of TM6. (c) Snapshots at 50, traces of the minimum distances between the cholesterol 167.6 and 250 ns showing the cholesterol from the bottom at the intracellular (IC) ends of TM6Ð7 and selected panels of (a) interacting with either IC end of TM6 or of residues on TM6 and 7 (middle and bottom panels). The TM7 (Reprinted with permission from Shan et al. 2012) cholesterol initially in contact with the L7.44, V7.48, its association with TM7 and towards TM6 (see components in the first shell (Soubias et al. 2006). Fig. 4.3), coincided with the intracellular end of Statistical analyses of the MD trajectories further TM6 beginning to bend away from TM3. Overall, suggested that DHA preferentially solvates the cholesterol-TM6 distance was found to be rhodopsin due to the lower entropic cost for correlated with agonist-induced conformational its association with rhodopsin, compared to that changes in 5-HT2AR (Shan et al. 2012). incurred by the stearoyl chain (Grossfield et al. Similarly, DHA lipid chains were found to 2006a). localize preferentially at a small number of specific sites of the GPCR from a statistical 4.2.1.2 Hydrophobic Mismatch analysis of 26 independent 100 ns atomistic and Membrane Remodeling MD simulations of rhodopsin embedded in a Alongside the specific sites of interactions membrane composed of a 2:2:1 molar mixture between lipid molecules and the GPCR of SDPC (1-stearoyl-2-docosahexaenoyl-sn- molecules, the remodeling of the membrane Glycero-3-phosphocholine)/SDPE (1-stearoyl-2- surrounding the protein constitutes another docosahexaenoyl-sn-Glycero-3-phosphoethanol- important form of coupling between membranes amine)/cholesterol (Grossfield et al. 2006b). and integral proteins. This remodeling is The suggestion of specific sites for DHA, an driven by the mismatch of the hydrophobic- important component of the cell membrane hydrophilic interfaces between the membrane environment, in binding to rhodopsin (Grossfield and the protein. Deformation of the membrane et al. 2006b), is consistent with spectroscopic to match the hydrophobic length of an embedded data from NMR experiments measuring peptide has been demonstrated in studies magnetization transfer from rhodopsin to lipid employing single-helical peptides as models to 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 61
Fig. 4.4 Membrane-deformation profiles u(x,y)for posed of (a)diC14:1PC, (b)diC20:1PC, and (c) 7:7:6 rhodopsin immersed in lipid bilayers of differ- SDPC/POPC/cholesterol membranes (Reprinted with per- ent bulk thicknesses. u(x,y) was calculated directly mission from Mondal et al. 2011. Copyright by Elsevier) from MD trajectories for rhodopsin in bilayers com- investigate membrane-protein coupling (Harroun assumptions, the calculations also revealed that et al. 1999a). Furthermore, the treatment of a the membrane deforms differently near different membrane-inserted peptide as a cylinder enabled parts of the 7-TM GPCR, in a specific pattern the assumption of radially symmetric bilayer dependent on the properties of the embedded deformations and perfect hydrophobic matching GPCR. This is illustrated in Fig. 4.4 by results between the peptide and the membrane so that the obtained for the systems of rhodopsin embed- energy cost of bilayer remodeling could be eval- ded in a thinner di(C14:1)PC membrane and a uated using the continuum theory of membrane thicker di(C20:1)PC bilayer that had been ex- elasticity (Huang 1986; Andersen and Koeppe amined earlier by the Gawrisch lab (Soubias 2007; Lundbæk and Andersen 1999; Lundbaek et al. 2008). Indeed, the di(C14:1)PC membrane et al. 2003; Nielsen et al. 1998; Goforth et al. is seen to become thicker on average around 2003). This energy cost has been shown to rhodopsin, and the di(C20:1)PC membrane to affect the function of membrane-embedded become thinner. The calculated extents of av- molecules (Andersen and Koeppe 2007; erage membrane deformation around the GPCR Lundbaek et al. 2010). agreed with the findings from solid state NMR However, the membrane remodeling pattern measurements on these systems (Soubias et al. is much more complex for GPCRs. This was 2008). But in addition, our CTMD calculations shown by results from calculations with a novel brought to light the radial asymmetry of the method that did not require the assumption of membrane deformations that exhibit both local radially symmetric membrane deformations and thickening and local thinning near different re- cylindrical proteins. We developed the theoret- gions of the protein (Mondal et al. 2011). For ical framework of the hybrid Continuum-MD example, the di(C14:1)PC lipid bilayer thickened (CTMD) method, described in detail in Sect. 4.3, by 2 Å on average near the protein, similar to to make possible the quantitative evaluation of results from NMR experiments, but it thickened radially asymmetric bilayer deformations around locally by 5 Å near TM4 (Mondal et al. 2011). membrane-inserted proteins (Mondal et al. 2011). The theoretical framework allowed the conclu- The CTMD calculations confirmed the expec- sion that the radial asymmetry of the membrane tation that the average remodeling of a thick deformation is due to the radially asymmetric membrane will make it thinner around an em- hydrophobic surface of the GPCR, which has bedded GPCR, and a thin membrane will become TMs of different hydrophobic lengths (Mondal thicker on average. But, freed of the symmetry et al. 2011). 62 S. Mondal et al.
ab c Gul5.36 Gul5.36 C2 atoms
E5.36 (polar)
Q5.60 F5.63 (polar) Phe5.63 C2 atoms Phe5.63 (hydrophobic)
d V2.66, V2.67 N1.33
Fig. 4.5 Illustration of hydrophobic adaptation and penalty at TM5 for rhodopsin in diC14:1PC (Panels A and residual exposure at the molecular level.(a) A snap- B are reprinted with permission from Mondal et al. 2011. shot from the MD trajectory of rhodopsin in the thin Copyright by Elsevier). (c) In Rhodopsin, adjacency of diC14:1PC bilayer. TM5 is shown in blue, and two residues the hydrophobic Phe5.63 to the polar Gln5.60. The left (Glu5.36 and Phe5.63) are highlighted. (b) A snapshot of panel shows rhodopsin in VdW representation in order rhodopsin in the thick diC20:1PC bilayer. The diC20:1PC to depict the irregular surface with which the membrane bilayer thins near Glu5.36, which substantially reduces interacts. The right panel shows the protein in the cartoon its exposure to the hydrophobic core of the bilayer, thus representation to show the helices. (d) In the dopamine D2 showing hydrophobic adaptation. Phe5.63 remains un- receptor model, adjacency of polar N1.33 and hydropho- favorably exposed to the polar environment in the thin bic V.266, V2.67 in the membrane-facing surface, with diC14:1PC bilayer but not in the thick diC20:1PC bilayer. the protein shown in both, VdW (left panel) and cartoon Thus, Phe5.63 contributes to residual exposure energy representations (right panel)
4.2.1.3 The Residual Hydrophobic more, a number of computational studies of the Mismatch and the Energy Cost coupling between the membrane and embedded of Protein-Membrane Coupling multi-TM proteins identified specific residues of Although hydrophobic mismatch is energetically the GPCRs where the hydrophobic matching re- costly, membrane deformations do not achieve a mains incomplete in a membrane of specific com- complete hydrophobic matching at all TM seg- position, thus highlighting specific regions of the ments (Mondal et al. 2011), and residues in the GPCR where unfavorable hydrophobic-polar in- TM-bundle can remain exposed to unfavorable teractions carry a significant energy cost (Mondal hydrophobic-polar interactions. For example, in et al. 2011, 2012; Shan et al. 2012; Khelashvili MD simulations of rhodopsin in di(C14:1)PC, et al. 2012). TM5 maintains a residual hydrophobic mismatch Results from CTMD calculations and coarse- on its extracellular side where a hydrophobic grained simulations suggest that some of this phenylalanine is exposed to the solvent in spite of energy cost contributes to the modulation of re- the overall membrane remodeling around the pro- ceptor oligomerization by the membrane. GPCRs tein (see Fig. 4.5a). Such incomplete adaptation have been shown to self-associate spontaneously of the membrane to the hydrophobic surface has upon reconstitution into liposomes, without the also been inferred from EPR experiments mea- need for any cellular machinery (Mansoor et al. suring lipid-protein associations for a number of 2006; Harding et al. 2009), with oligomeriza- transmembrane proteins (Marsh 2008). Further- tion occurring preferentially at specific interfaces 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 63
(Fung et al. 2009). The oligomerization is a That adjacencies responsible for the resid- reversible process (Dorsch et al. 2009;Hernetal. ual exposure of multi-TM proteins are a com- 2010), and is driven by hydrophobic mismatch mon feature of GPCRs, is evident from the re- (Botelho et al. 2006). Coarse-grained MD sim- cently determined crystal structures and validated ulations of the interaction of GPCRs in the mem- structural models of a variety of rhodopsin-like brane have provided results consistent with the GPCRs, including B2AR, KOR, DOR, 5-HT2AR, experimental observations, and offered important and D2R in the literature (Mondal et al. 2011, insights into the energetics of this oligomeriza- 2012; Shan et al. 2012). Indeed, computational tion (Periole et al. 2007, 2012). Such simula- modeling with the CTMD method showed that tions, taken together with the quantitative re- residual exposure occurs at these adjacencies in sults from CTMD calculations, suggest that the different GPCRs and in a number of different residual exposure contributes to the organiza- membranes (see Tables 4.1 and 4.2) (Mondal tional role of the hydrophobic mismatch (Mondal et al. 2011, 2012; Shan et al. 2012). This is et al. 2011, 2012). In particular, these calcu- illustrated in Fig. 4.5aÐc showing the residual lations suggest that there is a drive for GPCR exposure in TM5 of rhodopsin at a site where oligomerization at specific interfaces that remove Phe and Gln are juxtaposed. Figure 4.5dshows the structural context for the residual hydropho- another example of the adjacency, with the polar bic mismatch, thereby reducing the hydrophobic N1.33 next to the hydrophobic V2.66 and V2.67 mismatch and its energy cost (Mondal et al. 2011, in a structural model of the dopamine D2 recep- 2012). tor. Here too, the calculations show a residual The detailed quantitative analysis of the exposure, with an energy cost of 2 kT at N1.33 incomplete hydrophobic matching associated in a raft-like membrane. with the protein-membrane coupling of multi- These results from computational modeling TM-segment proteins, revealed a persistent local highlight the two key elements of the coupling exposure of residues to unmatched hydrophobic energy between the membrane and the GPCR, environments, which could not be eliminated viz. (1)-the radially asymmetric membrane de- by the overall membrane remodeling. This formation, and (2)-the residual exposure remain- was termed “residual exposure” (Mondal et al. ing after the membrane deforms to alleviate the 2011). Calculations in several different GPCR- hydrophobic mismatch. The nature of these two membrane systems (Mondal et al. 2011, 2012) identifiable elements demonstrates that the cou- showed that this residual exposure is due to the pling between the membrane and GPCR proteins radial asymmetry of the GPCR’s hydrophobic is not a simple result of the difference between surface. Specifically, if polar and hydrophobic the average hydrophobic length of the protein and residues, e.g., in two different TMs, are adjacent that of the pure membrane as previously postu- in space, membrane deformation would need to lated for model peptides (Harroun et al. 1999a, produce a match to both polar and hydrophobic b; Lundbæk and Andersen 1999;Marsh2008). residues in a small neighborhood, and this may be Rather, the important consequences of the cou- energetically unfavorable. Therefore, adjacencies pling emerge from specific features of the protein of hydrophobically disparate residues mark surface, both with respect to residue identity possible sites of residual exposure and thus and to local structure. Therefore, very different energetically costly membrane-protein interac- consequences can be expected from even small tions. The CTMD approach that determines the differences between highly homologous proteins energetic cost of the residual exposure caused (Mondal et al. 2012), from point mutations, or by such adjacencies (Mondal et al. 2011), is from structural changes due to ligand binding presented in Sect. 4.3 below. (see next Section). 64 S. Mondal et al.
2 Table 4.1 Residual SAres values (in Å ) and the corresponding Gres values (in kBT) calculated for TMs of rhodopsin in diC14:1PC, diC16:1PC, diC18:1PC, diC20:1PC, and 7:7:6 SDPC/POPC/cholesterol membranes
diC14:1PC diC16:1PC diC18:1PC diC20:1PC SDPC/POPC/cholesterol
TM SAres Gres SAres Gres SAres Gres SAres Gres SAres Gres 1 210 9.9 80 3.8 75 3.5 70 3.3 63 3.0 2 74 3.5 29 1.4 0 0 32 1.5 27 1.3 3N.D.a N.D. N.D. N.D. N.D. N.D. N.D. N.D. N.D. N.D. 4 0 0 77 3.6 90 4.3 79 3.7 134 6.3 5 115 5.4 1 0 9 0.4 7 0.3 17 0.8 6 36 1.7 31 1.5 56 2.7 38 1.8 51 2.4 7 119 5.6 44 2.1 22 1.1 0 0 0 0 Reprinted with permission from Mondal et al. (2011). Copyright by Elsevier aN.D. Not Determined
2 Table 4.2 Residual SAres values (in Å ) and the corresponding Gres values (in kBT) calculated for TMs of 5-HT2AR in complex with 5-HT, LSD, and KET, respectively, in an SDPC/POPC/cholesterol lipid bilayer
5-HT2AR with 5-HT 5-HT2AR with LSD 5-HT2ARwithKET
TM SAres Gres SAres Gres SAres Gres 1 0 0 0 0 107 5.1 2301.400 00 3N.D.a N.D. N.D. N.D. N.D. N.D. 4 25 1.2 81 3.8 103 4.9 5 94 4.4 81 3.8 69 3.3 6401.900 00 71 0 0 0 0 0 Reprinted with permission from Mondal et al. (2011). Copyright by Elsevier aN.D. Not Determined
4.2.2 Properties profile reflects the different conformational of GPCR-Membrane rearrangements of the receptor, e.g., the different Interactions Determined tilt angles of TM6 (Shan et al. 2012), induced by Ligand Binding by the three ligands. The patterns of residual exposure calculated for the receptor complexes Of particular significance to GPCR signaling with the three ligands and the host membrane is the observation that both the membrane are shown in Table 4.2 (Mondal et al. 2011). deformation profile and the residual exposure Incomplete hydrophobic matching with an pattern, can be specific to the ligand that energy penalty substantially larger than 1 kT acts on the receptor. For example, analysis of is predicted for TMs 1, 4, and 5 for the ligand-dependent conformational states of a 5- Ketanserin complex, but only for TMs 4 and HT2AR receptor model showed that membrane 5 for the LSD-bound receptor. These calculations remodeling around the receptor bound to the illustrate how ligand binding to GPCRs can affect agonist 5-HT is different from that observed membrane-protein interactions, and consequently around 5-HT2AR in complex with either the the patterns of oligomerization (both extent partial agonist LSD, or the inverse agonist and interface). Indeed, in addition to being Ketanserin (see Fig. 4.6) (Shan et al. 2012). membrane-driven, GPCR oligomerization has The membrane is thinner near TM4 and TM6 been reported to be ligand-sensitive for the when either the agonist 5-HT, or the partial dopamine D2, serotonin 5HT2C, and B2AR agonist LSD, is bound, compared to the structure receptors (Guo et al. 2005; Mancia et al. 2008; stabilized by the inverse agonist Ketanserin. Fung et al. 2009). The findings reviewed here The difference in membrane deformation suggest that the modulation of oligomerization 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 65
Fig. 4.6 Hydrophobic thickness profiles of the mem- colored fields represent distances (in Å) between lipid branes around 5-HT2AR in complex with 5-HT, LSD, backbone C2 atoms from the opposing leaflets. For this or KET. The structures of the various ligand-bound recep- analysis, the membrane plane was divided into square tor models averaged over the last 100 ns of MD simula- 2Å 2 Å bins, and the average C2ÐC2 distances in each tions are shown in cartoon, with only the helices depicted bin were collected by scanning the last 100 ns of trajectory (in different colors) with corresponding TM numbers. The (Reprinted with permission from Shan et al. 2012) in the membrane by a specific ligand involves a radial asymmetry of the hydrophobic surface contribution from its effect on the hydrophobic plays for GPCRs (Sect. 4.2), CTMD appears mismatch. The CTMD approach presented next, particularly suited for, and was in fact first in Sect. 4.3, can be used to obtain a quantitative developed for, the analysis of GPCR-membrane prediction regarding the effects of binding of interactions. We note, however, that the CTMD different ligands to a given GPCR, specifically formulation is general enough to be applied to whether the different effects might result in any transmembrane protein. a substantial difference in the energetics of membrane-protein interactions. 4.3.1 Membrane Deformations
4.3 Continuum-Molecular CTMD employs the well-established continuum, Dynamics (CTMD) Approach elastic theory of membranes (Nielsen et al. 1998; Huang 1986; Lundbaek et al. 2010; The hybrid Continuum-Molecular Dynamics Andersen and Koeppe 2007) to evaluate the (CTMD) approach we developed evaluates the energy cost of membrane deformation (Gdef ). energy cost of hydrophobic mismatch for multi- Formulations with elastic terms representing the TM proteins, such as GPCRs (Mondal et al. compression-extension and the splay-distortion 2011). Taking into consideration the energy of the membrane have explained the functional cost of membrane deformations (Sect. 4.3.1) effects of hydrophobic mismatch in model lipid- and the energy penalty due to residual exposure protein systems at a quantitative level (Huang at specific residues (Sect. 4.3.2), the CTMD is, 1986; Nielsen et al. 1998; Goforth et al. 2003). to our knowledge, the only method currently To date, typical quantitative applications of the available that takes into account the radial continuum, elastic theory of membranes have asymmetry of the protein hydrophobic surface been limited to single transmembrane helical in evaluating the energy cost of hydrophobic proteins because the calculations for such model mismatch. In view of the central role that the peptides could be simplified by assuming the 66 S. Mondal et al. peptide to be a cylinder and assuming perfect cylindrical symmetry. The method (Mondal et al. hydrophobic matching between the protein 2011) is implemented in the CTMDapp (http:// and the membrane. In the CTMD approach memprotein.org/resources/servers-and-software) the formalism has enabled the quantitative and is described below. application of this theory to multi-helical In CTMD, the membrane shape is defined by transmembrane proteins by taking into account the local deformation u(x,y) as the radial asymmetry of the hydrophobic surface of such proteins. Thus CTMD considers (1)-the 1 u .x;y/ D .d .x;y/ d0/ ; (4.1) radially asymmetric membrane deformations, 2 and (2)-the possibility of incomplete alleviation of the hydrophobic mismatch between the protein where d(x,y) is the local bilayer thickness and and the membrane. To this end, the membrane d0 is the bulk thickness of the bilayer (i.e., the deformations at the membrane-protein interface equilibrium thickness away from the protein). are determined from cognate MD trajectories, Gdef is defined as the sum of contributions and the membrane deformations away from from compression-extension, splay-distortion, the protein are solved at the continuum level and surface tension (Nielsen et al. 1998; Huang without the typical simplifying assumption of 1986):
Z ( !) 1 .2 /2 @2 @2 2 @ 2 @ 2 G K u K u u C ˛ u u d; def D a 2 C c 2 C 2 o C C (4.2) 2 d0 @x @y @x @y
where Ka and Kc are the elastic constants evaluate the membrane deformations and the for compression-extension and splay-distortion corresponding energy costs without invoking the respectively, C0 is the monolayer spontaneous typical assumption of cylindrical protein, we next curvature, and ˛ the coefficient of surface perform the following two steps: tension. The phenomenological constants Ka, 1. The membrane-protein interface is determined Kc, C0, and d0 characterize the membrane from cognate MD simulations, with in and property at the continuum level. Experimental u0(x,y) being obtained from the time-averaged measurements for these constants are available position of the P-atoms around the protein. for typical lipid types, see (Rawicz et al. 2000)for Specifically, the trajectory is first centered example. and aligned on the protein, keeping the Z- The local membrane deformations u(x,y)at axis along the membrane normal. Then, the continuum level are calculated by solving the time-averaged z-position of the P-atoms the Euler-Lagrange equations corresponding to is calculated on a rectangular grid with a Eq. 4.2. Following the Euler-Lagrange formula- 2 Å by 2 Å mesh to obtain the membrane- tion, the Gdef definition in Eq. 4.2 leads to the protein interface as well as the membrane boundary value problem, thickness on this interface. This allows for 4K ˇ a non-cylindrical in corresponding to the K 4 ˛ 2 a 0; ˇ .x;y/ ; c r u r u C 2 u D u in D uo membrane-protein interface. Additionally, the d0 ˇ ˇ membrane deformations u0(x,y)isnot simply 2 ˇ 2 ˇ uj D 0; r u in D vo .x;y/ ; r u D 0 out out the difference between the hydrophobic length (4.3) of the protein and that of the unperturbed membrane, thus allowing for residual where in and out represent the boundary at exposure. Note that the pattern of local the GPCR-membrane interface and the outer membrane deformations at the membrane- boundary of the simulation box respectively. To protein interface from MD accounts for 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 67
various interfacial interactions at the atomistic where is the polar angle corresponding to the level, including interactions between single coordinates (x,y), and an and bn are the Fourier lipid molecules and specific sites of the coefficients. This step is required to make the GPCRs and any tilting of the TMs within method feasible for multi-TM proteins with long the steric constraint of the whole TM-bundle. membrane-protein interface, e.g., the 7 TM- 2. The resulting boundary value problem bundle of GPCRs has a diameter on the order in Eq. 4.3 is numerically solved without of 50 Å, thus sharing a long circumferential simplifying the fourth order partial dif- boundary with the membrane. A given pair of ferential equation in u(x,y)toanordinary fan, bng in Eq. 4.5 defines a particular v0(x,y) differential equation by assuming radial vector, for which Eq. 4.4 yields the membrane symmetry. This is achieved by first expressing deformations u(x,y) and Eq. 4.2 the Gdef . the partial differential equation in u as In the CTMDapp, the non-linear optimization coupled partial differential equations in problem is solved with the objective function u and v, Gdef D ffan, bng using the BFGS optimization algorithm, which is a standard global, quasi- 4K K 2 ˛ a 0; Newtonian optimization procedure. cr v v C 2 u D d0 We note that the procedure actually involves ˇ 2 ˇ two nested minimization procedures. The inner r u D v; uˇ D uo .x;y/ ; in minimization occurs via the Euler-Lagrange for- ˇ ˇ malism to give the membrane deformation shape ˇ ˇ uˇ D u1 .x;y/ ; vˇ D v0 .x;y/ ; that minimizes the energy cost, given u0 and v0. out in ˇ The outer minimization obtains the v0 that min- ˇ imizes the energy cost from the Euler-Lagrange vˇ D v1 .x;y/ (4.4) out formalism, thus obtaining the curvature at the membrane-protein interface for which the mem- This set of simultaneous equations can be brane deformations are least costly. To verify that solved on a rectangular grid using standard finite membrane shapes calculated from CTMD and difference schemes for Poisson equations. As MD are similar, the macroscopic u(x,y) obtained implemented in the CTMDapp, Eq. 4.4 is dis- by the nested optimization procedure can be com- cretized using the central 5-point approximation pared to u(x,y) calculated directly from the cog- of the Laplacian operator, and the system is then nate microscopic MD simulation. The calcula- solved using the iterative Gauss-Siedel algorithm. tions for a number of GPCR-membrane systems As a guideline, we mention that in our experience have shown that the macroscopic u(x,y) indeed the procedure converged within 130 iterations for converges to a profile similar (within 1Å) GPCR monomers in a 100 Å by 100 Å box. to the u(x,y) obtained from the cognate MD While , and u (x,y) are obtained from in out 0 trajectories. This is illustrated for rhodopsin in MD trajectories, the boundary condition on the di(C14:1)PC in Fig. 4.7. Importantly, the CTMD curvature is obtained self-consistently using an calculation evaluates the G using the well- optimization procedure that searches for v (x,y) def 0 tested elastic continuum theory of membrane to minimize G globally. The complexity of def deformations. this optimization procedure is made numerically tractable by reducing the size of the v0(x,y) vector by expressing it as a truncated Fourier series 4.3.2 Residual Exposure X7 vo .x;y/ W v0 ./ fan cos .n/ The residual exposure of each membrane-facing nD0 residue (SAres,i) is quantified in terms of the sur- face area involved in unfavorable hydrophobic- C bn sin .n/g ; (4.5) polar interactions. SAres,i is obtained in terms of 68 S. Mondal et al.
Fig. 4.7 Protocol for the CTMD approach, illustrated in panel A and a random curvature boundary condition schematically for rhodopsin in a diC14:1PC lipid bi- to produce the starting point for the free-energy-based layer.(a) The membrane-deformation profile u(x,y)is optimization (Eqs. 4.4 and 4.5). The right panel is the calculated directly from the MD simulations. The left membrane-deformation profile calculated with the natu- panel is the deformation profile shown as a color map ral boundary condition, which minimizes the membrane- projected onto the surface defined by fitting a grid (spacing deformation energy penalty. Note the agreement between 2 Å) to the positions of the phosphate atoms in the two the profiles in A, calculated using a microscopic theory, leaflets during the trajectory, followed by time averaging and B, calculated using the continuum theory (they are and spatial smoothing. The right panel shows the same de- within 0.5 Å RMSD of each other). The CTMD approach formation color map on the x-y plane. (b) The membrane- also allows evaluation of the protein-induced membrane- deformation profile u(x,y) on a 100 100 Å patch, calcu- deformation energy penalty, which is 4.7 kBT in this case lated with CTMD. The left panel represents the membrane (Reprinted with permission from Mondal et al. 2011. shape calculated with the deformation boundary condition Copyright by Elsevier) at the membrane-protein interface from the MD profile the Solvent Accessible Surface Areas (SASAs) the solute taken to be the protein and the hy- calculated from cognate MD trajectories, as fol- drophobic part of the membrane. lows: The corresponding energy penalty (Gres,i) For hydrophobic residues, SAres,i is calculated can be approximated as linearly proportional to as the SASA with the solute taken to be the SAres,i, protein and hydrophobic core of the bilayer. This gives the surface area of the hydrophobic residues Gres;i D res SAres;i (4.6) exposed to the membrane headgroup or water. For hydrophilic residues, SAres,i quantifies the with the constant of proportionality res taken to surface area of the residue embedded in the hy- be 0.028 kcal/(mol. Å2) (Choe et al. 2008; Ben- drophobic core of the membrane. It is calculated Tal et al. 1996). as the difference of the SASA with the solute In these calculations, Lys, Arg, Trp, Tyr, Ser, taken to be the protein only and the SASA with and Thr need to be treated with special attention 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 69 to the structural context. If the polar Lys or Arg has been taken to suggest that the lipid-mediated residues reside within the hydrophobic part of effect on rhodopsin function is not due to lipids the membrane but near the water-membrane in- of a specific chemical nature (Botelho et al. terface, their long side chains can extend towards 2006). Such “non-specific interactions” have the phosphate headgroups so that the polar part been interpreted to involve energetically costly of their side-chains interacts with the polar phos- hydrophobic mismatch. phate headgroups (Strandberg and Killian 2003; The computational results described herein Sankararamakrishnan and Weinstein 2002). This identify the nature of both “specific” and “non- conformational preference is known as “snorke- specific” interactions and suggest that the ling”, and it has been estimated that as a result, “specific interactions” are local and the “non- only a small energy penalty is associated with specific” hydrophobic mismatch interactions hydrophobic matching by means of snorkeling are non-local in nature. Notably, however, (Strandberg and Killian 2003). Therefore, Lys individual lipid molecules that bind to specific and Arg residues near the water-membrane in- sites of the GPCR can also participate in terface are not considered in the calculations of hydrophobic matching. For example, in the residual exposure if they are found in the MD 1.6 s simulation of rhodopsin (see Sect. 4.2.1), simulations to snorkel to the lipid headgroup re- the cholesterol that binds specifically at the gion. Interfacial Trp and Tyr are also not consid- extracellular end of TM2-TM3 partly reduces ered as candidates for residual exposure, as their solvent accessibility at the hydrophobic Leu3.27 commonly observed interfacial location is con- (Khelashvili et al. 2009). Therefore, what is sidered favorable (Yau et al. 1998). Finally, Ser fundamental to this “specific” interaction is its and Thr residues are polar, but when they reside local nature, involving electrostatic/H-bonding/ within the hydrophobic part of the membrane, VdW interactions between particular protein their polar parts can form H-bonds with the helix residues and the particular cholesterol molecule. backbone (Gray and Matthews 1984). In such On the other hand, the “non-specific” nature of situations, the Ser and Thr are interacting directly the hydrophobic mismatch is mitigated by the with the protein and not the membrane so they evidence that the membrane deforms differently would not be candidates for residual exposure. near TMs of different hydrophobic lengths, and the residual exposure occurs at specific residues of the receptor depending on the spatial 4.4 Perspective on “Specific” vs. organization of polar and hydrophobic residues. “Non-specific” However, the deformation profile involves the GPCR-Membrane membrane around the GPCR as a whole thus Interactions substantiating the non-local character of the effect. Consequently, membrane deformation Functionally relevant interaction of lipids with profiles could indeed be obtained in the CTMD GPCRs has historically been viewed as being of approach by treating the entire membrane as two different types: (1)-specific lipid-protein an elastic continuum and minimizing the total interactions, e.g., the binding of individual energetic cost of the membrane deformations cholesterol molecules to specific residues of (provided the thickness boundary conditions the GPCR; and (2)-non-specific interactions, were known from atomistic MD simulations). such as hydrophobic mismatch (Botelho et al. We showed for a number of membrane-GPCR 2002, 2006;Brown1994; Gibson and Brown systems that the membrane deformation profile 1991, 1993). For example, in affecting rhodopsin thus obtained was consistent with results from activation, lipids with PE headgroups can microscopic MD calculations that treat the compensate for the absence of lipids with PS membrane at the level of lipid molecules (Mondal headgroups and lipids with DHA tails, which et al. 2011). 70 S. Mondal et al.
for Computational Biomedicine at Weill Medical College 4.5 Conclusion of Cornell University, (2) the New York Blue Gene Com- putational Science facility housed at Brookhaven National The multifaceted role of the membrane in GPCR Lab, and (3) NSF Teragrid allocation MCB090022. function has, of late, come to the forefront of GPCR research. This has been driven by the experimental data suggesting that the membrane References environment affects oligomerization, stability, and activity of GPCRs and the observation Andersen OS, Koeppe RE (2007) Bilayer thickness and of lipid molecules co-crystallized with some membrane protein function: an energetic perspective. Annu Rev Biophys Biomol Struct 36:107Ð130 GPCR crystals. A central aspect of GPCR- Ballesteros JA, Weinstein H (1995) Integrated methods membrane interactions is how the receptor for the construction of three-dimensional models and and the membrane couple to each other at the computational probing of structure-function relations molecular level. Recent computational modeling in G protein-coupled receptors. In: Sealfon SC (ed) Methods in neurosciences, vol 25. Academic, San has provided significant insight into connecting Diego, pp 366Ð428 GPCR-membrane interactions with the receptor Ballesteros J, Kitanovic S, Guarnieri F, Davies P, Fromme structure. In particular, this review brings to light BJ, Konvicka K, Chi L, Millar RP, Davidson JS, Wein- stein H (1998) Functional microdomains in G-protein- the importance of the radial asymmetry of the coupled receptors. J Biol Chem 273(17):10445Ð10453 membrane-facing surface of GPCRs in their Ballesteros JA, Jensen AD, Liapakis G, Rasmussen SGF, interaction with the surrounding membrane. The Shi L, Gether U, Javitch JA (2001) Activation of the radial asymmetry creates a context of adjacent B2-adrenergic receptor involves disruption of an ionic lock between the cytoplasmic ends of transmembrane hydrophobic and polar residues at specific sites segments 3 and 6. J Biol Chem 276(31):29171Ð29177 of the GPCR where hydrophobic matching may Barak LS, Menard L, Ferguson SSG, Colapietro remain incomplete. Moreover, it is due to this AM, Caron MG (1995) The conserved seven- radial asymmetry that there exist a few specific transmembrane sequence NP (X) 2, 3Y of the G- protein-coupled receptor superfamily regulates multi- sites where specific lipid molecules such as ple properties of the beta 2-adrenergic receptor. Bio- cholesterol bind. These interactions have been chemistry 34(47):15407Ð15414 shown to be involved in various aspects of the Ben-Tal N, Ben-Shaul A, Nicholls A, Honig B (1996) function and organization of GPCRs. Free-energy determinants of alpha-helix insertion into lipid bilayers. Biophys J 70(4):1803Ð1812 We have shown how to take into account Botelho AV, Gibson NJ, Thurmond RL, Wang Y, Brown the radial asymmetry of the hydrophobic surface MF (2002) Conformational energetics of rhodopsin of GPCRs in providing quantitative information modulated by nonlamellar-forming lipids. Biochem- about their interaction with the membrane. With istry 41(20):6354Ð6368 Botelho AV, Huber T, Sakmar TP, Brown MF (2006) the new understanding gained from the findings Curvature and hydrophobic forces drive oligomeriza- and methods presented here, it becomes possible tion and modulate activity of rhodopsin in membranes. to investigate how GPCR mechanisms, includ- Biophys J 91(12):4464Ð4477 ing ligand-determined states and their functional Brown MF (1994) Modulation of rhodopsin function by properties of the membrane bilayer. Chem Phys Lipids properties, are affected in particular lipid envi- 73(1Ð2):159Ð180 ronments. In particular, it becomes possible to Brown MF, Thurmond RL, Dodd SW, Otten D, Beyer quantify the contribution of the lipid-protein in- K (2002) Elastic deformation of membrane bilayers teractions with respect to (1)-the individual lipid probed by deuterium NMR relaxation. J Am Chem Soc 124(28):8471Ð8484 molecules interacting locally with specific sites Chelikani P, Hornak V, Eilers M, Reeves PJ, Smith of the GPCR, (2)-the deformation/remodeling SO, RajBhandary UL, Khorana HG (2007) Role of of the membrane, and (3)-the residual exposure group-conserved residues in the helical core of beta2- profile and its energy cost. adrenergic receptor. Proc Natl Acad Sci U S A 104(17):7027Ð7032 Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen Acknowledgement This work was supported by the Na- SGF, Thian FS, Kobilka TS, Choi HJ, Kuhn P, Weis tional Institutes of Health grants DA012923-09 and U54- WI, Kobilka BK (2007) High-resolution crystal struc- GM087519 to H.W. We also gratefully acknowledge the ture of an engineered human beta2-adrenergic G pro- allocations of computational resources at (1) the Institute tein coupled receptor. Science 318(5854):1258Ð1265 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 71
Chien EYT, Liu W, Zhao Q, Katritch V, Won Han G, Han- Grossfield A, Feller SE, Pitman MC (2006b) A role for son MA, Shi L, Newman AH, Javitch JA, Cherezov V, direct interactions in the modulation of rhodopsin by Stevens RC (2010) Structure of the human dopamine omega-3 polyunsaturated lipids. Proc Natl Acad Sci U D3 receptor in complex with a D2/D3 selective antag- S A 103(13):4888Ð4893 onist. Science 330(6007):1091Ð1095 Guo W, Shi L, Filizola M, Weinstein H, Javitch JA (2005) Choe S, Hecht KA, Grabe M (2008) A continuum method Crosstalk in G protein-coupled receptors: changes at for determining membrane protein insertion energies the transmembrane homodimer interface determine and the problem of charged residues. J Gen Physiol activation. Proc Natl Acad Sci U S A 102(48):17495Ð 131(6):563Ð573 17500 Choe HW, Kim YJ, Park JH, Morizumi T, Pai EF, Krauß Haga K, Kruse AC, Asada H, Yurugi-Kobayashi T, Shi- N, Hofmann KP, Scheerer P, Ernst OP (2011) Crystal roishi M, Zhang C, Weis WI, Okada T, Kobilka BK, structure of metarhodopsin II. Nature 471(7340):651Ð Haga T, Kobayashi T (2012) Structure of the human 655 M2 muscarinic acetylcholine receptor bound to an Dorsch S, Klotz KN, Engelhardt S, Lohse MJ, Bunemann antagonist. Nature 482:547Ð551 M (2009) Analysis of receptor oligomerization by Han Y, Moreira IS, Urizar E, Weinstein H, Javitch JA FRAP microscopy. Nat Methods 6(3):225Ð230 (2009) Allosteric communication between protomers Eroglu Ç, Brügger B, Wieland F, Sinning I of dopamine class A GPCR dimers modulates activa- (2003) Glutamate-binding affinity of Drosophila tion. Nat Chem Biol 5(9):688Ð695 metabotropic glutamate receptor is modulated by Hanson MA, Cherezov V, Griffith MT, Roth CB, Jaakola association with lipid rafts. Proc Natl Acad Sci U S A VP, Chien EYT, Velasquez J, Kuhn P, Stevens RC 100(18):10219 (2008) A specific cholesterol binding site is estab- Fritze O, Filipek S, Kuksa V, Palczewski K, Hofmann KP, lished by the 2.8 Å structure of the human beta2- Ernst OP (2003) Role of the conserved NPxxY (x) adrenergic receptor. Structure 16(6):897Ð905 5, 6F motif in the rhodopsin ground state and during Hanson MA, Roth CB, Jo E, Griffith MT, Scott activation. Proc Natl Acad Sci U S A 100(5):2290Ð FL, Reinhart G, Desale H, Clemons B, Cahalan 2295 SM, Schuerer SC, Sanna MG, Han GW, Kuhn P, Fung JJ, Deupi X, Pardo L, Yao XJ, Velez-Ruiz GA, Rosen H, Stevens RC (2012) Crystal structure of a DeVree BT, Sunahara RK, Kobilka BK (2009) Ligand- lipid G protein-coupled receptor. Science 335(6070): regulated oligomerization of 2-adrenoceptors in a 851Ð855 model lipid bilayer. EMBO J 28(21):3315Ð3328 Harding PJ, Attrill H, Boehringer J, Ross S, Wadhams Gibson NJ, Brown MF (1991) Role of phosphatidylserine GH, Smith E, Armitage JP, Watts A (2009) Consti- in the MI-MIII equilibrium of rhodopsin*. Biochem tutive dimerization of the G-protein coupled receptor, Biophys Res Commun 176(2):915Ð921 neurotensin receptor 1, reconstituted into phospholipid Gibson NJ, Brown MF (1993) Lipid headgroup and acyl bilayers. Biophys J 96(3):964Ð973 chain composition modulate the MI-MII equilibrium Harroun TA, Heller WT, Weiss TM, Yang L, Huang of rhodopsin in recombinant membranes. Biochem- HW (1999a) Experimental evidence for hydropho- istry 32(9):2438Ð2454 bic matching and membrane-mediated interactions in Gimpl G, Fahrenholz F (2002) Cholesterol as stabilizer lipid bilayers containing gramicidin. Biophys J 76(2): of the oxytocin receptor. Biochimi Biophys Acta 937Ð945 Biomembr 1564(2):384Ð392 Harroun TA, Heller WT, Weiss TM, Yang L, Huang Goforth RL, Chi AK, Greathouse DV, Providence LL, HW (1999b) Theoretical analysis of hydrophobic Koeppe RE, Andersen OS (2003) Hydrophobic cou- matching and membrane-mediated interactions in lipid pling of lipid bilayer energetics to channel function. J bilayers containing gramicidin. Biophys J 76(6): Gen Physiol 121(5):477Ð493 3176Ð3185 Granier S, Manglik A, Kruse AC, Kobilka TS, Thian Hern JA, Baig AH, Mashanov GI, Birdsall B, Corrie FS, Weis WI, Kobilka BK (2012) Structure of the JET, Lazareno S, Molloy JE, Birdsall NJM (2010) delta-opioid receptor bound to naltrindole. Nature Formation and dissociation of M1 muscarinic receptor 485(7398):400Ð404 dimers seen by total internal reflection fluorescence Gray TM, Matthews BW (1984) Intrahelical hydrogen imaging of single molecules. Proc Natl Acad Sci U S A bonding of serine, threonine and cysteine residues 107(6):2693Ð2698 within [alpha]-helices and its relevance to membrane- Huang HW (1986) Deformation free energy of bilayer bound proteins. J Mol Biol 175(1):75Ð81 membrane and its effect on gramicidin channel life- Green SA, Liggett SB (1994) A proline-rich region of time. Biophys J 50(6):1061Ð1070 the third intracellular loop imparts phenotypic beta 1- Innis SM (2008) Dietary omega 3 fatty acids and the versus beta 2-adrenergic receptor coupling and seques- developing brain. Brain Res 1237:35Ð43 tration. J Biol Chem 269(42):26215Ð26219 Jaakola VP, Griffith MT, Hanson MA, Cherezov V, Chien Grossfield A, Feller SE, Pitman MC (2006a) Contribution EYT, Lane JR, Ijzerman AP, Stevens RC (2008) of omega-3 fatty acids to the thermodynamics of mem- The 2.6 angstrom crystal structure of a human A2A brane protein solvation. J Phys Chem B 110(18):8907Ð adenosine receptor bound to an antagonist. Science 8909. doi:10.1021/jp060405r 322(5905):1211Ð1217 72 S. Mondal et al.
Javitch JA, Fu D, Chen J (1995a) Residues in the fifth Lundbaek JA, Collingwood SA, Ingólfsson HI, Kapoor membrane-spanning segment of the dopamine D2 re- R, Andersen OS (2010) Lipid bilayer regulation of ceptor exposed in the binding-site crevice. Biochem- membrane protein function: gramicidin channels as istry 34(50):16433Ð16439 molecular force probes. J R Soc Interface 7(44): Javitch JA, Fu D, Chen J, Karlin A (1995b) Mapping the 373Ð395 binding-site crevice of the dopamine D2 receptor by Luttrell LM, Ferguson SSG, Daaka Y, Miller WE, Maud- the substituted-cysteine accessibility method. Neuron sley S, Della Rocca GJ, Lin FT, Kawakatsu H, Owada 14(4):825Ð831 K, Luttrell DK, Caron MG, Lefkowitz RJ (1999) Javitch JA, Ballesteros JA, Weinstein H, Chen J (1998) Beta2-arrestin-dependent formation of beta2 adrener- A cluster of aromatic residues in the sixth membrane- gic receptor-Src protein kinase complexes. Science spanning segment of the dopamine D2 receptor is 283(5402):655Ð661 accessible in the binding-site crevice. Biochemistry Mancia F, Assur Z, Herman AG, Siegel R, Hendrick- 37(4):998Ð1006 son WA (2008) Ligand sensitivity in dimeric associ- Kahsai AW, Xiao K, Rajagopal S, Ahn S, Shukla AK, ations of the serotonin 5HT2c receptor. EMBO Rep Sun J, Oas TG, Lefkowitz RJ (2011) Multiple ligand- 9(4):363Ð369 specific conformations of the B2-adrenergic receptor. Manglik A, Kruse AC, Kobilka TS, Thian FS, Mathiesen Nat Chem Biol 7(10):692Ð700 JM, Sunahara RK, Pardo L, Weis WI, Kobilka BK, Khelashvili G, Grossfield A, Feller SE, Pitman MC, Granier S (2012) Crystal structure of the mu-opioid Weinstein H (2009) Structural and dynamic effects receptor bound to a morphinan antagonist. Nature of cholesterol at preferred sites of interaction with 485(7398):321Ð327 rhodopsin identified from microsecond length molecu- Mansoor SE, Palczewski K, Farrens DL (2006) Rhodopsin lar dynamics simulations. Prot Struct Funct Bioinform self-associates in asolectin liposomes. Proc Natl Acad 76(2):403Ð417 Sci U S A 103(9):3060 Khelashvili G, Albornoz PBC, Johner N, Mondal S, Marsh D (2008) Energetics of hydrophobic match- Caffrey M, Weinstein H (2012) Why GPCRs behave ing in lipid-protein interactions. Biophys J 94(10): differently in cubic and lamellar lipidic mesophases. 3996Ð4013 J Am Chem Soc. doi:10.1021/ja3056485 Mondal S, Khelashvili G, Shan J, Andersen OS, Wein- Kofuku Y, Ueda T, Okude J, Shiraishi Y, Kondo K, Maeda stein H (2011) Quantitative modeling of membrane M, Tsujishita H, Shimada I (2012) Efficacy of the B2- deformations by multihelical membrane proteins: ap- adrenergic receptor is determined by conformational plication to G-protein coupled receptors. Biophys J equilibrium in the transmembrane region. Nat Com- 101(9):2092Ð2101 mun 3:1045 Mondal S, Khelashvili G, Wang H, Provasi D, Andersen Lagerstrom MC, Schioth HB (2008) Structural diversity of OS, Filizola M, Weinstein H (2012) Interaction with G protein-coupled receptors and significance for drug the membrane uncovers essential differences between discovery. Nat Rev Drug Discov 7(4):339Ð357 highly homologous GPCRs. Biophys J 102(3):514a Lavie CJ, Milani RV, Mehra MR, Ventura HO (2009) Neuringer M, Connor WE, Van Petten C, Barstad L (1984) Omega-3 polyunsaturated fatty acids and cardiovascu- Dietary omega-3 fatty acid deficiency and visual loss lar diseases. J Am Coll Cardiol 54(7):585Ð594 in infant rhesus monkeys. J Clin Invest 73(1):272 Lebon G, Warne T, Edwards PC, Bennett K, Lang- Nielsen C, Goulian M, Andersen OS (1998) Energetics mead CJ, Leslie AGW, Tate CG (2011) Agonist- of inclusion-induced bilayer deformations. Biophys J bound adenosine A2A receptor structures reveal com- 74(4):1966Ð1983 mon features of GPCR activation. Nature 474(7352): Niu S-L, Mitchell DC, Litman BJ (2002) Manipulation of 521Ð525 cholesterol levels in rod disk membranes by methyl- Liapakis G, Ballesteros JA, Papachristou S, Chan WC, beta-cyclodextrin. J Biol Chem 277(23):20139Ð20145. Chen X, Javitch JA (2000) The forgotten serine: a doi:10.1074/jbc.M200594200 critical role for Ser-2035.42 in ligand binding to and O’Dowd BF, Hnatowich M, Regan JW, Leader WM, activation of the B2-adrenergic receptor. J Biol Chem Caron MG, Lefkowitz RJ (1988) Site-directed mu- 275(48):37779Ð37788 tagenesis of the cytoplasmic domains of the human Liu W, Chun E, Thompson AA, Chubukov P, Xu F, beta 2-adrenergic receptor. Localization of regions Katritch V, Han GW, Roth CB, Heitman LH, Ijzerman involved in G protein-receptor coupling. J Biol Chem AP, Cherezov V, Stevens RC (2012) Structural basis 263(31):15985Ð15992 for allosteric regulation of GPCRs by sodium ions. Palczewski K, Kumasaka T, Hori T, Behnke CA, Science 337(6091):232Ð236 Motoshima H, Fox BA, Trong IL, Teller DC, Lundbæk JA, Andersen OS (1999) Spring constants Okada T, Stenkamp RE (2000) Crystal structure for channel-induced lipid bilayer deformations esti- of rhodopsin: a G protein-coupled receptor. Science mates using gramicidin channels. Biophys J 76(2): 289(5480):739Ð745 889Ð895 Park JH, Scheerer P, Hofmann KP, Choe HW, Ernst Lundbaek JA, Andersen OS, Werge T, Nielsen C (2003) OP (2008) Crystal structure of the ligand-free G- Cholesterol-induced protein sorting: an analysis of protein-coupled receptor opsin. Nature 454(7201): energetic feasibility. Biophys J 84(3):2080Ð2089 183Ð187 4 How the Dynamic Properties and Functional Mechanisms of GPCRs Are Modulated... 73
Periole X, Huber T, Marrink SJ, Sakmar TP (2007) Shi L, Javitch JA (2004) The second extracellular loop G protein-coupled receptors self-assemble in dynam- of the dopamine D2 receptor lines the binding- ics simulations of model bilayers. J Am Chem Soc site crevice. Proc Natl Acad Sci U S A 101(2): 129(33):10126Ð10132 440Ð445 Periole X, Knepp AM, Sakmar TP, Marrink SJ, Huber T Shi L, Liapakis G, Xu R, Guarnieri F, Ballesteros JA, (2012) Structural determinants of the supramolecular Javitch JA (2002) B2 Adrenergic receptor activation: organization of G protein-coupled receptors in bilay- modulation of the proline kink in transmembrane 6 ers. J Am Chem Soc 134(26):10959Ð10965 by a rotamer toggle switch. J Biol Chem 277(43): Prioleau C, Visiers I, Ebersole BJ, Weinstein H, Sealfon 40989Ð40996 SC (2002) Conserved helix 7 tyrosine acts as a mul- Soubias O, Teague WE, Gawrisch K (2006) Evidence for tistate conformational switch in the 5HT2C receptor. specificity in lipid-rhodopsin interactions. J Biol Chem J Biol Chem 277(39):36577Ð36584 281(44):33233Ð33241 Pucadyil TJ, Chattopadhyay A (2004) Cholesterol mod- Soubias O, Niu SL, Mitchell DC, Gawrisch K ulates ligand binding and G-protein coupling to (2008) Lipid- rhodopsin hydrophobic mismatch serotonin 1A receptors from bovine hippocam- alters rhodopsin helical content. J Am Chem Soc pus. Biochimi Biophys Acta Biomembr 1663(1Ð2): 130(37):12465Ð12471 188Ð200 Stahl LA, Begg DP, Weisinger RS, Sinclair AJ (2008) The Rasmussen SGF, Choi HJ, Rosenbaum DM, Kobilka TS, role of omega-3 fatty acids in mood disorders. Curr Thian FS, Edwards PC, Burghammer M, Ratnala VRP, Opin Investig Drugs 9(1):57Ð64 Sanishvili R, Fischetti RF, Schertler GFX, Weis WI, Strader CD, Sigal IS, Register RB, Candelore MR, Rands Kobilka BK (2007) Crystal structure of the human E, Dixon RA (1987) Identification of residues required beta2 adrenergic G-protein-coupled receptor. Nature for ligand binding to the beta-adrenergic receptor. Proc 450(7168):383Ð387 Natl Acad Sci U S A 84(13):4384Ð4388 Rasmussen SGF, Choi HJ, Fung JJ, Pardon E, Casarosa Strandberg E, Killian JA (2003) Snorkeling of lysine side P, Chae PS, DeVree BT, Rosenbaum DM, Thian chains in transmembrane helices: how easy can it get? FS, Kobilka TS (2011) Structure of a nanobody- FEBS Lett 544(1Ð3):69Ð73 stabilized active state of the beta2 adrenoceptor. Na- Suryanarayana S, Von Zastrow M, Kobilka BK (1992) ture 469(7329):175Ð180 Identification of intramolecular interactions in Rawicz W, Olbrich KC, McIntosh T, Needham D, Evans adrenergic receptors. J Biol Chem 267(31): E (2000) Effect of chain length and unsaturation 21991Ð21994 on elasticity of lipid bilayers. Biophys J 79(1): Visiers I, Ebersole BJ, Dracheva S, Ballesteros J, Sealfon 328Ð339 SC, Weinstein H (2002) Structural motifs as functional Rosenbaum DM, Rasmussen SG, Kobilka BK (2009) The microdomains in G-protein coupled receptors: ener- structure and function of G-protein-coupled receptors. getic considerations in the mechanism of activation Nature 459(7245):356Ð363 of the serotonin 5HT2A receptor by disruption of the Rosenbaum DM, Zhang C, Lyons JA, Holl R, Aragao ionic lock of the arginine cage. Int J Quantum Chem D, Arlow DH, Rasmussen SGF, Choi HJ, DeVree 88(1):65Ð75 BT, Sunahara RK, Chae PS, Gellman SH, Dror RO, Warne T, Serrano-Vega MJ, Baker JG, Moukhametzianov Shaw DE, Weis WI, Caffrey M, Gmeiner P, Ko- R, Edwards PC, Henderson R, Leslie AGW, Tate CG, bilka BK (2011) Structure and function of an irre- Schertler GFX (2008) Structure of a beta1-adrenergic versible agonist-beta 2 adrenoceptor complex. Nature G-protein-coupled receptor. Nature 454(7203): 469(7329):236Ð240 486Ð491 Ruprecht JJ, Mielke T, Vogel R, Villa C, Schertler GFX Weinstein H (2005) Hallucinogen actions on 5-HT re- (2004) Electron crystallography reveals the structure ceptors reveal distinct mechanisms of activation and of metarhodopsin I. EMBO J 23(18):3609Ð3620 signaling by G protein-coupled receptors. AAPS J Sankararamakrishnan R, Weinstein H (2002) Positioning 7(4):871Ð884 and stabilization of dynorphin peptides in membrane Wu B, Chien EYT, Mol CD, Fenalti G, Liu W, bilayers: the mechanistic role of aromatic and basic Katritch V, Abagyan R, Brooun A, Wells P, Bi residues revealed from comparative MD simulations. FC, Hamel DJ, Kuhn P, Handel TM, Chere- J Phys Chem B 106(1):209Ð218 zov V, Stevens RC (2010) Structures of the Scheerer P, Park JH, Hildebrand PW, Kim YJ, Krauß CXCR4 chemokine GPCR with small-molecule N, Choe HW, Hofmann KP, Ernst OP (2008) Crystal and cyclic peptide antagonists. Science 330(6007): structure of opsin in its G-protein-interacting confor- 1066Ð1071 mation. Nature 455(7212):497Ð502 Wu H, Wacker D, Mileni M, Katritch V, Han GW, Vardy Shan J, Khelashvili G, Mondal S, Mehler EL, Weinstein E, Liu W, Thompson AA, Huang XP, Carroll FI, H (2012) Ligand-dependent conformations and dy- Mascarella SW, Westkaemper B, Mosier PD, Roth namics of the serotonin 5-HT2A receptor determine BL, Cherezov V, Stevens RC (2012) Structure of the its activation and membrane-driven oligomerization human kappa-opioid receptor in complex with JDTic. properties. PLoS Comput Biol 8(4):e1002473 Nature 485(7398):327Ð332 74 S. Mondal et al.
Xiang Y, Rybin VO, Steinberg SF, Kobilka B (2002) agonist-bound human A2A adenosine receptor. Sci- Caveolar localization dictates physiologic signaling ence 332(6027):322Ð327 of beta2-adrenoceptors in neonatal cardiac myocytes. Yau WM, Wimley WC, Gawrisch K, White J Biol Chem 277(37):34280Ð34286 SH (1998) The preference of tryptophan for Xu F, Wu H, Katritch V, Han GW, Jacobson KA, Gao membrane interfaces. Biochemistry 37(42):14713Ð ZG, Cherezov V, Stevens RC (2011) Structure of an 14718 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions 5 of Lipids and Cholesterol with Rhodopsin
Joshua N. Horn, Ta-Chun Kao, and Alan Grossfield
Abstract Protein function is a complicated interplay between structure and dy- namics, which can be heavily influenced by environmental factors and conditions. This is particularly true in the case of membrane proteins, such as the visual receptor rhodopsin. It has been well documented that lipid headgroups, polyunsaturated tails, and the concentration of cholesterol in membranes all play a role in the function of rhodopsin. Recently, we used all-atom simulations to demonstrate that different lipid species have preferential interactions and possible binding sites on rhodopsin’s surface, consistent with experiment. However, the limited timescales of the simulations meant that the statistical uncertainty of these results was substantial. Accordingly, we present here 32 independent 1.6 s coarse-grained simulations exploring lipids and cholesterols surrounding rhodopsin and opsin, in lipid bilayers mimicking those found naturally. Our results agree with those found experimentally and in previous simu- lations, but with far better statistical certainty. The results demonstrate the value of combining all-atom and coarse-grained models with experiment to provide a well-rounded view of lipid-protein interactions.
Keywords Rhodopsin ¥ All-atom simulations ¥ Coarse-graining ¥ Lipid-protein interactions ¥ Protein function
5.1 Lipid-Protein Interactions
Protein structure and dynamics play a major role J.N. Horn ¥ T.-C. Kao ¥ A. Grossfield () in function. It is not surprising then that mem- Department of Biochemistry and Biophysics, brane protein function would be strongly in- University of Rochester Medical Center, 601 Elmwood fluenced by interactions at the protein-lipid in- Ave, Box 712, Rochester, NY 14642, USA terface (Marsh, 2008). There are many reasons e-mail: alan_grossfi[email protected] that the lipid environment plays a major part
M. Filizola (ed.), G Protein-Coupled Receptors - Modeling and Simulation, Advances 75 in Experimental Medicine and Biology 796, DOI 10.1007/978-94-007-7423-0__5, © Springer ScienceCBusiness Media Dordrecht 2014 76 J.N. Horn et al. in the structure and thus the function of inte- properties, protein structure and flexibility, and gral membrane proteins. First, lipid composi- protein function (Haswell et al., 2011). One such tion of membranes is incredibly diverse, mean- channel, MscL (mechanosensitive channel of ing that there is no single property that defines large conductance), responds to turgor pressure a membrane; rather, membranes from different changes and hypotonic shock by opening a cells (or different parts of the same cell) can 30 Å pore to release osmolytes and solutes differ wildly from each other (Brügger et al., (Martinac et al., 1987; Cruickshank et al., 1997). Moreover, the membrane-water interface 1997). The mechanism by which MscL “senses” is a “region of tumultuous chemical heterogene- these conditions is by responding structurally ity” (Wiener and White, 1992), giving membrane to tension changes in the lipid bilayer. Turgor proteins a highly diverse surrounding environ- pressure, as well as cellular processes like cell ment (Engelman, 2005). division, stretches or compresses the bilayer. A number of interactions are at play in the Thorough reviews of this interplay between membrane environment, including hydrophobic membrane structure and protein function for mismatch (Mouritsen and Bloom, 1984), these channels exist elsewhere (Kung et al., 2010; lipid fluidity, membrane tension, hydrocarbon Martinac, 2011; Haswell et al., 2011). chain packing (Fattal and Ben-Shaul, 1993), The bacterial potassium channel, KcsA, bilayer free volume (Mitchell et al., 1990), the is another prototypical example. It has been intrinsic curvature of lipids (Gruner, 1985) and demonstrated that KcsA requires a lipid bilayer elastic strain resulting from bilayer curvature for proper folding, despite being stable and frustration (Lee, 2004). Furthermore, lipid- active in experiments using detergent micelles, protein interactions can be subdivided into bulk and requires the binding of a single negatively interactions, or those that result from bilayer charged lipid on each monomer for activation properties, and specific interactions, or those (Valiyaveetil et al., 2002). Simulation studies that involve an association with individual lipids confirmed a Arg64ÐArg89 binding motif (Valiyaveetil et al., 2002). Specific interactions for acidic lipids that was discovered in the can be further subdivided into those with annular experimental studies (Sansom et al., 2005). lipids, or the boundary lipids forming the first Linear gramicidins are a family of bilayer- shell around a protein, and non-annular lipids, spanning antibacterial cation channels that which can be described as “co-factor” lipids increase the permeability of target membranes with unique binding sites (Simmonds et al., (Harold and Baarda, 1967). The natural folding 1982; Lee, 2003). In some cases, lipids can even preference is to form intertwined (double- be structural elements of membrane proteins stranded) dimers (Burkhart et al., 1998). In the (Lee, 2003). A more complete survey of the presence of lipid bilayers they refold into the suggested lipid-protein interaction models can functional end-to-end (single-stranded) dimers be found in a number of thorough reviews (Andersen et al., 1999), with all four Trp residues (Mouritsen and Bloom, 1993; Andersen and in each subunit hydrogen bonding with the Koeppe, 2007). bilayer at the membrane surface (O’Connell et al., 1990), effectively anchoring the structure in a bilayer spanning configuration. The preference 5.1.1 Model Systems for for this conformation in the presence of a Lipid-Protein Interactions lipid bilayer is driven by these Trp residues at the bilayer-solvent interface, which would A number of model proteins exist that depend create a penalty if they were buried in the heavily on lipid interaction for function. One hydrophobic core. These effects have been such class of proteins is that of mechanosensitive noted for other proteins that show a similar channels. The function of these channels Trp anchoring pattern (Killian and von Heijne, depends on the relationship between membrane 2000). 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 77
provide near-atomic resolution into the possible 5.2 Rhodopsin general and specific mechanisms by which the lipid bilayer interacts with rhodopsin. Rhodopsin, a G protein-coupled receptor (GPCR) responsible for dim-light vision, is a well- characterized transmembrane protein activated 5.2.1 Rhodopsin-Lipid Interactions by the isomerization of 11-cis-retinal to the all- trans configuration via light absorption. The A number of studies by the Brown lab, start- ligand’s isomerization initiates a cascade of ing in the late 1980s and early 1990s (Wied- thermal relaxations in the protein, ending with mann et al., 1988; Gibson and Brown, 1993; metarhodopsin I (MI). MI exists in equilibrium Brown, 1994), focused on MII production as with metarhodopsin II (MII), the tranducin- a result of the photoisomerization of rhodopsin binding (or “active”) form of the protein reconstituted in membranes with a variety of (Liebman et al., 1987). The MI-MII transition phospholipid and fatty acid combinations. They features significant structural motions that lead showed that the population of MII depended to the activation of the G protein and ultimately on the lipid headgroup composition as well as results in signal transduction. the concentration of polyunsaturated acyl chains. Rhodopsin is found in large concentrations Native ROS membranes, as expected, showed the in the rod outer segment disks (ROS) of rod greatest quantities of MII. In membranes of PC cells, making up the vast majority of the with short, saturated acyl chains, for instance protein component of each disk’s membrane di(14:0)PC, rhodopsin is essentially inactive. Us- and occupying about a third of the total area ing di(22:6)PC, they demonstrated that polyun- (Molday, 1998; Buzhynskyy et al., 2011). The saturation increased the degree of activity, but disk membrane phospholipid distribution is not to native levels. The addition of PE lipids about 44 % phosphatidylcholine (PC), 41 % also increased activity, though this increase was phosphatidylethanolamine (PE), 13 % phospha- minor. The presence of polyunsaturation or PE tidylserine (PS), and 2 % phosphatidylinositol lipids alone does not recreate native activity. In- (Boesze-Battaglia and Albert, 1992), with a high stead, a mixture of phospholipids containing both concentration polyunsaturated docosahexaenoic polyunsaturated chains (22:6!3) and PE head acid (DHA) tails (Boesze-Battaglia and Albert, groups had the highest activity among non-native 1989). The concentration of cholesterol in systems (Wiedmann et al., 1988). new disks is high (30 %) and decreases as Exploring the role of chain length, and in turn the disk ages (Boesze-Battaglia et al., 1989). the hydrophobic mismatch between the bilayer Given that rhodopsin is an integral membrane and rhodopsin, has shown that the MII popu- protein with a cascade of structural changes lation is maximized with chain lengths around implicated in its function, it is not surprising 18 carbons, with the equilibrium shifting back that this unique membrane environment has towards MI with chain lengths between 16 and been shown to affect the behavior of rhodopsin, 20 carbons. This is coupled to local bilayer com- particularly the equilibrium between the MI pression and stretching effects (Botelho et al., and MII states; recent reviews of these effects 2006). This mismatch and the resulting bilayer are available (Soubias and Gawrisch, 2012). deformations affect rhodopsin activation by al- Given the biomedical importance of GPCRs tering the helical content of the protein (Soubias (Drews, 2000) and studies of polyunsaturated et al., 2008). All of this suggests that mechanical fatty acids in dietary intake (Neuringer, 2000), and physical properties of the bilayer, including the implications for bilayer regulation of GPCRs the bilayer thickness and area per lipid, likely to human diseases are clear. Here, we intend to modulate the MI-MII transition. However, these not only highlight experimental and simulation bulk effects do not exclude the possibility of work that explores these effects, but also present localized lipid binding sites on the surface of the long time-scale coarse-grained simulations that protein. 78 J.N. Horn et al.
Some preliminary tests were included 5.2.2 Role of Cholesterol exploring the role of headgroups in rhodopsin activation, by comparing the effects of PE and PS A major component of the rod disk membranes, headgroups (Gibson and Brown, 1993). Later, the cholesterol has been shown to have an effect role of the membrane potential at the membrane- on rhodopsin activation as well. The presence water interface due to lipid headgroups was ex- of cholesterol in membranes drives the MI- plored more thoroughly. It was demonstrated that MII equilibrium towards MI, reducing signaling lipids with PS headgroups have two contradictory (Mitchell et al., 1990). However, it is unclear effects: they alter the bilayer’s properties in ways from these experiments whether this is caused by that oppose MII formation, but their net charge cholesterol’s effects on bilayer properties, direct creates an electrostatic environment rich with interactions between it and the protein, or some C H3O ions that promotes MII formation (Brown, combination of these effects. 1997). This is in agreement with other results Cholesterol’s effects on liquid crystalline that showed that MII formation is enhanced in membrane structure are well documented. The acidic conditions (Delange et al., 1997). Finally, presence of cholesterol causes tighter packing they concluded that both PS headgroups and of lipid hydrocarbons (Niu et al., 2002) and the combination of PE headgroups and DHA increases the thickness of the bilayer, leading to chains are needed to maximize the population changes in the lateral compressibility (Needham of MII. The theory suggests that membranes et al., 1988). These bilayer effects may create with only PC and PS headgroups would favor an environment that inhibits the conformational MII based only on electrostatics, but this would transition from MI to MII, though cholesterol be counteracted by the structural unfavorability also promotes negative curvature elastic stress, a of a charged bilayer surface. As a result, ROS property of PE headgroups that tends to promote membranes have high concentrations of lipids MII formation (Chen and Rand, 1997). with highly negative spontaneous curvature (PE) The alternative means by which cholesterol and polyunsaturated chains to counteract this by may affect rhodopsin activation is through di- providing curvature stress and thus promote MII rect and possibly specific interactions. Studies of (Wang et al., 2002). These results led Brown cholesterol, cholestatrienol and ergosterol in ROS and coworkers to a general model, known as the membranes have suggested that there is at least flexible surface model (FSM), where composition one cholesterol binding site on rhodopsin (Albert of the lipid matrix actively regulates rhodopsin et al., 1996). Sites of preferential cholesterol function (Botelho et al., 2002). interaction have been identified via molecular The effects of headgroups on the MI-MII simulation as well (Grossfield et al., 2006b), equilibrium are not limited to electrostatics and although these and other simulations suggested membrane elasticity. In fact, in work intended to that cholesterol is, on the whole, depleted at the discern the energetic contribution of membrane protein surface (Pitman et al., 2005). elasticity to rhodopsin function, Gawrisch and coworkers noted that PE headgroups also induce a shift toward MII that correlates with their 5.2.3 Simulation Applied hydrogen-bonding ability (Soubias et al., 2010). to Rhodopsin Furthermore, saturation transfer NMR studies of rhodopsin in mixed PC/PE bilayers showed In 2000, the first crystal structure of a GPCR greater magnetization transfer to PE lipids when was solved in the form of bovine rhodopsin at compared to PC lipids (Soubias et al., 2006), 2.8 Å resolution (Palczewski et al., 2000), open- suggesting that in addition to their effects on bulk ing the way for the use of molecular dynamics bilayer properties, there may be a specific role simulation to probe the atomic-level interactions for PE headgroups at the surface of rhodopsin. between rhodopsin and its environment. An early 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 79 simulation of rhodopsin, performed by Huber alyzed to explore the interactions between the et al. (2002), was compared to available NMR palmitoyl moieties attached to a pair of cysteines data to quantify membrane deformation in the in the cytoplasmic helix H8 and the helical bundle presence of protein and compute cross-sectional of rhodopsin (Olausson et al., 2012). There was protein areas. They found that rhodopsin imposed a high degree of contact between both chains curvature in the bilayer, which could facilitate and the protein, but also a significant difference selection for polyunsaturated lipids at the surface between the two, even though they are attached of rhodopsin. This was seen as further support for to consecutive residues. The high level of contact the flexible surface model (Huber et al., 2004). between the palmitate on Cys322 and the protein As the available computer power improved, helices suggests that it may play an important role simulations began to explore the interactions beyond that of a nonspecific lipid anchor. between rhodopsin and bilayer constituents via Recently, some groups have begun using longer molecular dynamics trajectories of the coarse-grained simulations to explore rhodopsin- protein embedded in a more realistic bilayer. lipid interactions as well. For example, These systems featured cholesterol, as well as a Periole et al. (2007) explored the effects of mixture of lipids with two different headgroups, varying the bilayer hydrophobic thickness on each of which was linked to one polyunsaturated the oligomerization of rhodopsin, using the tail and one saturated tail. Early results showed a MARTINI force field (Marrink et al., 2007; preference for the polyunsaturated tails at the Monticelli et al., 2008). The results demonstrated surface of rhodopsin and little indication of that oligomerization is driven by frustration specific binding sites for cholesterol (Feller et al., of lipid-protein interactions, something that 2003; Pitman et al., 2005). had already been seen experimentally (Botelho Later, a series of 26 independent 100 ns et al., 2006). Rhodopsin significantly altered simulations of rhodopsin in a realistic lipid the local membrane thickness, encouraging composition was used to address these questions, oligomerization as a way to reduce unfavorable with better statistical sampling (Grossfield protein-lipid interactions. While useful for et al., 2006b). Polyunsaturated tails were again exploring bilayer adaptations attributed to enriched at the surface of the protein. It was hydrophobic mismatch, each simulation featured also now possible to identify residues that homogeneous bilayers, so no new evidence preferentially interacted with cholesterol and was gleaned about the role for cholesterol each of the lipid components. Although this and different lipid headgroups in rhodopsin work required a heroic effort at the time, activation. the sampling was not sufficient to give high In this work, we employed simulations confidence in the predictions about specific featuring membranes with native-like compo- residues, particularly for cholesterol, since any sitions to more thoroughly explore protein-lipid given interaction was only seen in a small interactions. Our focus in these systems was fraction the trajectories (Grossfield et al., the role played by the saturation state of the 2006b). Nonetheless, the results demonstrated lipids, the headgroups of the lipids, and the that computational methods could confirm the presence of cholesterol in modulating lipid- experimentally suggested trend of preferential protein interactions. interactions with polyunsaturated tails, allowing some to speculate on the roles these flexible lipids could play in the activation of rhodopsin 5.3 Methods (Feller and Gawrisch, 2005). Preferential sites of interaction between rhodopsin and cholesterol 5.3.1 Simulation Systems were also identified (Khelashvili et al., 2009). More recently, the same data (supplemented The goal of this research was to explore lipid- by an additional 1.6 s simulation) was rean- protein interactions in a system with rhodopsin 80 J.N. Horn et al. and a biologically relevant model membrane. We retinal-free opsin was also modeled (PDB ID: sought to extend previously published work that 3CAP Park et al. 2008). The MARTINI coarse- featured 26 separate 100 ns all-atom simulations grained force field was used (Marrink et al., of rhodopsin in a membrane (Grossfield et al., 2007) with the extension for proteins (Monticelli 2006b). These systems featured a 2:2:1 ratio of 1- et al., 2008). Recently, the MARTINI forcefield stearoyl-2-docosahexaenoyl-phosphatidylcholine was applied to study the aggregation properties (SDPC), 1-stearoyl-2-docosahexaenoyl-phosph- of rhodopsin (Periole et al., 2012; Knepp et al., atidylethanolamine (SDPE), and cholesterol, and 2012). To create the coarse-grained model for were intended to mimic those membranes found each protein, all molecules present other than biologically and used in experimental studies rhodopsin/opsin were removed from the crystal with model membranes (Boesze-Battaglia and structure. We then mapped the coarse amino Albert, 1989; Wiedmann et al., 1988; Albert acids onto this structure using the martinize.py et al., 1996). The systems described here mimic script available online at the MARTINI website these all-atom systems in an attempt to bring (MARTINI, 2011). The long C-terminal tail together a great deal of experimental work that found in the rhodopsin structure, which was has been done investigating the role of various not resolved in the opsin structure, was not membrane properties on rhodopsin activation, removed. It makes minimal contact with the taking advantage of the ability of molecular bilayer and thus should not affect our results dynamics simulation to explore interactions on in this context. The acyl chains on the cysteine the molecular level. residues at positions 322 and 323 were manually While being a remarkable effort at the time, added with parameters used for palmitoyl chains previous all-atom simulations had some limita- in lipids. The resulting system was energy tions that faster computers and more efficient minimized. Retinal was not explicitly represented simulation methods have allowed us to over- in the rhodopsin model; in the MARTINI protein come. By employing a coarse-grained model, we model, protein fluctuations are dominated by the simulated larger systems for longer timescales, network of restraints required to stabilize the allowing for better sampling of long time-scale tertiary structure, and since the retinal itself does processes critical for bilayer-protein interactions, not interact with the membrane environment in such as the lateral reorganization of the lipid any significant way, we believe it is sufficient to bilayer. The work described here features systems model the retinal-stabilized dark-state rhodopsin nearly three times larger than the all-atom work, structure of the protein in its absence. with timescales an order of magnitude longer. First, we built an SDPC bilayer via self- Furthermore, simulations of “dark” and “active” assembly with a box of randomly placed SDPC rhodopsin were done to make comparisons about lipids and water. For the rhodopsin system, we the lipid and cholesterol binding surfaces in these inserted the protein into this SDPC lipid bilayer differing structures. Overall, the new results agree using a bilayer expansion and compression with the earlier all-atom work, while provid- technique (Kandt et al., 2007). In this method, ing more robust statistics and allowing us to the lipids are translated in the plane of the bilayer make some conclusions about specific and gen- by a large scaling factor, creating space for the eral interactions between the lipid environment protein insertion without clashes. Then, the lipids and rhodopsin. are scaled back to the ideal area-per-lipid using a number of cycles of translation and minimization. This reduces lipid-lipid and lipid-protein clashes. 5.3.2 Construction Water and ions were added to solvate and neutralize the system; additional NaCl was added Rhodopsin was modeled after the same crystal to bring the concentration to 100 mM. After- structure used in the previous work (PDB ID: wards, we held the protein position fixed and 1U19 Okada et al. 2004). For comparison, performed a 10 ns simulation to allow the lipid 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 81 and solvent environment to relax. To generate simulations with only the parameters necessary unique starting states for the bilayer, randomly to maintain the secondary structure. However, selected SDPC lipids were converted to SDPE under these conditions the rhodopsin structure lipids by simply changing the head group beads, moved rapidly away from the crystal structure, while others were swapped with cholesterol. The reaching a transmembrane alpha carbon RMSD result was a set of 16 unique rhodopsin systems as high as 6 Å (data not shown). This problem with 2:2:1 SDPC:SDPE:cholesterol bilayers. We in the MARTINI force field with maintaining minimized the resulting systems and performed protein tertiary structure has been noted another round of equilibration, again holding the before and protocols have been developed for protein fixed. overcoming these limitations utilizing distance- To make the equivalent opsin model, based restraints (Periole et al., 2009). To maintain rhodopsin was replaced with opsin after the integrity of our proteins, we included a backbone alignment for each of the rhodopsin similar network model, restraining the distances systems. These were then equilibrated with between backbone beads between 2 and 10 Å a short simulation (100 ps) to ensure that the apart. We tested multiple force constants using starting configurations for the opsin systems and short trajectories to try to match the amplitude the rhodopsin systems were similar. of fluctuations for the transmembrane helices The final systems included 1 protein molecule to previous all-atom simulations; with a force (either rhodopsin or opsin), 180 SDPC, 180 constant of 800 kJ/mol nm2, the rhodopsin SDPE, 90 cholesterol, about 19,000 water beads transmembrane helix RMSD fluctuated between (each representing 4 water molecules), and about 2.0 and 2.5 Å, consistent with previous all-atom 140 each of NaC and Cl . This brings the system results (Grossfield et al., 2008), and the opsin size to about 26,000 CG beads, which is roughly RMSD fluctuated between 2.5 and 3.0 Å. equivalent to an all-atom system with 105,000 We ran 32 independent simulations, 16 each atoms (double the total system size and more for rhodopsin and opsin. Each simulation was than triple the bilayer size of the previous work). 1.6 s, for a total of 51.2 s of simulation time (effective time of about 205 s if we apply a 4X scaling to the time, as suggested by previous 5.3.3 Simulation Protocol authors to account for the enhanced kinetics of the coarse-grained model Marrink et al. 2007). Simulations were performed with version 4.5.4 All times we report here are the actual simulation of the GROMACS molecular dynamics package times, without the 4X scaling. (van der Spoel et al., 2005; Hess et al., 2008)on a Linux cluster. We used a time step of 10 fs, as suggested for accurate integration (Winger et al., 5.3.4 Analysis 2009; Marrink et al., 2010), with the neighbor list updated every 5 steps. We held the temperature at All simulation analyses were performed using 300 K using Nosé-Hoover temperature coupling tools developed using the LOOS library. LOOS (Nosé and Klein, 1983; Hoover, 1985) and treated is an object-oriented library implemented in C++ the pressure semi-isotropically with a reference and Boost for rapidly creating new tools for an- of 1 bar using the Parrinello-Rahman barostat alyzing molecular dynamics simulations (Romo (Parrinello and Rahman, 1981). A shift function and Grossfield, 2009, 2012). All analysis was per- was employed for electrostatics with a coulomb formed on trajectories with 1 ns time resolution. cutoff of 12 Å. The Lennard-Jones potential was shifted between 9 and 12 Å. 5.3.4.1 Bound Lipid Lifetimes It is well established that external restraints In order to measure the time required for lipids are required when simulating native proteins to exchange off the surface of the protein, we using MARTINI. Initially, we performed our calculated a conditional survival probability. 82 J.N. Horn et al.
Specifically, we identified those lipids within 6 Å Where N is the number of atoms in residue of any protein bead as “bound”, and computed n, M is the sum of all atoms for all molecules the probability that if a lipid is bound at time of lipid component m in the system, rij is the t it will also be bound at time t C t. Error distance between atoms i and j . The normalized bars are estimated by treating each simulation as residue score is then simply the residue score an independent measure of the decay curve and divided by the average residue score for all trans- computing the standard error in the means. This membrane alpha helix residues: calculation was implemented using LOOS. R Rı nm nm D 1 PN (5.2) tm Rkm 5.3.4.2 Lateral Radial Distribution Ntm k Function We computed the lateral radial distribution func- 5.3.4.5 Statistical Significance tion (RDF) of various bilayer species relative to Because we have multiple independently con- the center of mass of the protein structure in the structed trajectories of each system, we have membrane plane using the xy_rdf tool distributed attempted to assess the statistical significance of with LOOS tools. Each molecule was treated as a our results. This was done using a Welch’s T- single unit, located at its centroid. test, treating the average result (e.g. the radial distribution density at a particular distance) from 5.3.4.3 Density Maps each trajectory as a single data point. We typically We created 2D density maps to show average plot the p-values on the same axes as the results density of each lipid component in the plane themselves. To compute these p-values, we used of the bilayer. We aligned the protein structure statistical tools available in SciPy (Jones et al., of each frame of the simulation to the initial 2001Ð). structure. Then we binned the centroid of each component in a grid on the plane parallel to the membrane, with 1 Å2 bins. The resulting density 5.4 Results and Discussion histogram is displayed as a heat map. To probe the 3-dimensional distribution of Before beginning any kind of analysis of lipid lipid components about the membrane, we first binding, it is critical that we show thorough aligned our trajectories using transmembrane sampling of the exchange between surface 3 C˛’s. Then we used a 1 Å grid superimposed bound lipids and bulk lipids; otherwise, the over the protein’s bounding box, padded by results would be dominated by the starting 30 Å. Each atom is then placed in the nearest configuration of the lipids. To accomplish this bin and the resulting histogram is convolved we calculated the probability if a lipid is found with a gaussian for a smoother visualization. at the protein surface at time t it will also The gaussian smoothing is only applied to the be there at time t C t. The resulting decay 3D visualizations, not the 2D density maps. curves for SDPE and SDPC are shown in Fig. 5.1. These curves can be fit to a double 5.3.4.4 Residue-Based Binding Scans exponential, resulting in two characteristic decay To highlight interactions between lipid compo- times. The fast time, accounting for roughly nents and individual residues, we computed a half the amplitude, is below 1 ns. Given that residue binding score for each residue of the the analysis was performed using 1 ns samples, protein to each lipid component. The residue this is most likely the result of flickering at score R for residue n and lipid component m can the edge of the “bound” state. The slow decay be expressed as: times for SDPE and SDPC in the rhodopsin system are 78.3 ns (˙2.88) and 63.2 ns (˙1.31), N M 1 X X 1 respectively. Rnm D (5.1) N r6 i j ij 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 83
Fig. 5.1 Survival 1 probability for lipids at the SDPE surface of rhodopsin. Error 0.9 SDPC bars are the standard error of the means of the 0.8 16 trajectories 0.7
0.6
0.5
Probability 0.4
0.3
0.2
0.1
0 0 100 200 300 400 500 Δt (ns)
5.4.1 RDFs Demonstrate Surface D and E show these p-value plots for choles- Preferences for DHA terol, DHA and stearoyl, comparing the means between rhodopsin and opsin. Statistically signif- To begin our analysis, we used a two-dimensional icant (P < 0:01) differences appear between all radial distribution function (RDF) in the plane of analyzed bilayer constituents, but some for only the membrane, as was done in previous work brief stretches of the RDF curves. (Feller et al., 2003), to assess the packing of The short region of cholesterol significance the different members of the bilayer against coincides with a peak in the opsin curve that is rhodopsin. In Fig. 5.2, we show RDFs for the not present in rhodopsin, indicating a region of two lipid tail types and cholesterol. Here we can bulk density at the very surface of opsin where see a drastic enrichment of docosahexaenoic acid the DHA density is the highest. The significant (DHA) between 15 and 20 Å from the center regions are much more substantial for the lipids of the protein, with the peaks for cholesterol tails. For stearoyl, large stretches of the curve and stearoyl beyond 20 Å. Interestingly, while show very significant differences (P < 0:001). cholesterol is not enriched at the surface, it has Visual inspection of the RDFs shows greater significant density deeper into the protein than penetration of the lipid tails between 10 and 20 Å. stearoyl, and nearly as deep as DHA. Given its This can be explained to some degree by the smaller size, this may be indicative of regions greater flexibility and more “open” structure for accessible only to cholesterol. the opsin system; there is a greater area accessible To accurately assess the significance of the to the lipid tails between the helices and in the difference between RDF curves for opsin and protein interior. rhodopsin, we calculated p-values for each point. Given that each point is the mean of a set of 16 independent samples, we have a fairly large 5.4.2 Density Maps Show DHA set of data from which to do this assessment Preference (unusual in the simulation community). The re- sulting p-values are plotted below the RDF curves The above results are consistent with previous with the same x-axis, with confidence levels of simulation and experimental results. However, 0.01 and 0.05 shown for reference. Panels C, simple lateral radial distribution functions 84 J.N. Horn et al.
Fig. 5.2 Lateral radial a distribution functions of 1.4 each of the lipid tails and cholesterol for the (a) 1.2 rhodopsin and (b) opsin 1 systems. Comparison of means tests between 0.8 rhodopsin and opsin for (c) 0.6 cholesterol, (d) DHA, and (e) stearoyl 0.4 CHOL 0.2 DHA Radial Distribution Function STEA b 0 1.4 1.2 1 0.8 0.6 0.4 CHOL 0.2 DHA Radial Distribution Function STEA 0 c 1 .05 .01 .001
P−value .0001 d 1 .05 .01 .001
P−value .0001 e 1 .05 .01 .001
P−value .0001
5 10 15 20 25 30 35
X(Å) contain limited information, because they treat dark region in the center of each frame repre- rhodopsin as a featureless cylinder, integrating sents the excluded volume of the helix bundle, out the distinctions between different portions as we look down from extracellular side. Opsin of the protein surface. Moreover, in these and rhodopsin were aligned by their backbones plots, both leaflets were treated together, again so that they are oriented the same way in all averaging away potentially valuable information. of the heat maps for comparison. These images Accordingly, we instead project our results along represent the average of all 16 simulations for 2 dimensions, using lateral density heat maps. each system. Figure 5.3 shows density maps for the dif- In the plots of DHA density, we see a ferent lipid components for both rhodopsin and bright, thin ring tracing the protein space. In opsin, in both the upper and lower leaflets. The contrast with the low densities for stearoyl 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 85
Rhodopsin Opsin Upper Lower Upper Lower DHA 30 20 10 )
Å 0 Y ( −10 −20 −30 STEA 30 20 10 )
Å 0 Y ( −10 −20 −30 CHOL 30 20 10 )
Å 0 Y ( −10 −20 −30
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30
X (Å) X (Å) X (Å) X (Å) 1
0 0.5 1 1.5 2 2.5 3 3.5 2 7 Density
3 6
4 5
Fig. 5.3 Density maps of each bilayer component, for origin and aligned against a reference structure so that each leaflet, in each system (rhodopsin or opsin). Density the maps can be directly compared. The orientation of is reported as lipid components per Å2. All images are the helices is shown in the small panel in the bottom viewed from the extracellular side of the protein. The right corner, also viewed from the extracellular side of upper leaflet refers to the leaflet on the extracellular side. the protein For every map, rhodopsin or opsin was centered at the and cholesterol, this indicates that DHA is stearoyl rings are dimmer and, in general, more preferentially packed against the surface of diffuse. the protein, with the exception of a bright The lateral radial distribution functions, cholesterol spot next to helices H1 and H7. The coupled with the density heat maps, suggest a corresponding stearoyl densities show rings as strong preference for DHA at the surface of both well, immediately outside the DHA ring. The opsin and rhodopsin, in agreement with previous 86 J.N. Horn et al. experimental and computational results. Previous comparison, p-values were computed for every work suggests that this preference is entropically point in the maps and plotted. Panel C shows driven (Grossfield et al., 2006a). It has been the p-value plot comparing SDPE (panel A) and demonstrated that DHA is extremely flexible SDPC (panel B) for the upper leaflet. There (Feller et al., 2002) and rapidly interconverts is a large region of statistical significance that between conformations (Soubias and Gawrisch, indicates differences between the densities of the 2007), making it ideal for packing against two headgroups along helices H3, H4 and H5 the relatively rigid but uneven surface of the on the extracellular side of the protein. No such protein. regions of interest are seen in the other leaflet, and The region just beyond the first shell of DHA it is less pronounced for the opsin system (data chains is enriched in stearate. This result is not not shown). surprising, because the lipids used in these sim- ulations each have one DHA and one stearoyl. For every lipid with a DHA tail packed against 5.4.4 Mapping Density to Structure rhodopsin, there is also a stearoyl facing away Probes Cholesterol Binding from the protein, accounting for the inner DHA Sites ring and the outer stearoyl ring. This outer ring is not as bright in the heat maps as the DHA ring The present simulations, by accessing the mi- because the accessible surface area in this ring is crosecond timescale, are much more effective at far greater, so the motion of these tails is more allowing for sampling of the lateral motion of diffuse. cholesterol and lipids in the bilayer than previous work. Considering the enhanced rate of lipid diffusion in the MARTINI forcefield, each of 5.4.3 SDPE is Preferred at Protein our simulations arguably samples nearly 6.5 s Surface of lateral reorganization. With this in mind, we have the ability to probe for the appearance of In Fig. 5.4, we compare preferences between cholesterol binding sites in all of our simulations SDPE and SDPC within each system to explore and hopefully converge on representative sites, a possible preference for one headgroup over whereas previous work could only note sites another. Visual inspection suggests a slight where some fraction of the simulations had seen preference for SDPE over SDPC at the surface of strong contacts. the protein, which is confirmed to be statistically In the lateral radial distribution plots, there significant; the lower panels show the p-value is evidence that cholesterol can pack inside the for the difference between SDPC and SDPE and helical bundle (deeper than either lipid tail), albeit demonstrates significance at the 0.01 level for not with high abundance. Indeed, integrating the the entire first “solvation” shell of rhodopsin, RDF suggests the presence of roughly 1 choles- and most of that region for opsin. Panel E shows terol immediately at the protein surface (data not that the differences in the SDPC RDF between shown). The heat maps clarify this result; visual rhodopsin and opsin are marginally significant at examination shows one site of clear cholesterol best, but the SDPE RDF (Panel F) does show a preference, next to helices H1 and H7, approxi- significant difference in the location of this initial mately where the palmitoyls attach to helix H8. rise at the protein surface. DHA lipids are excluded from this location, as This is in agreement with experimental results seen by a patch of low density. Whether this that suggest that PE is a preferred partner for site represents a true competitive binding site, rhodopsin (Soubias et al., 2006, 2010). To explore where binding to cholesterol is preferred over the possibility of specific PE interaction sites, other bilayer constituents, or simply a deep pro- we generated density maps of SDPE and SDPC tein pocket or crevice that is only accessible to for rhodopsin, found in Fig. 5.5. For quantitative cholesterol, is unclear. 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 87
1.4 a b 1.2
1
0.8
0.6
0.4 Radial Distribution Function 0.2 SDPE SDPE SDPC SDPC 0 1 .05 .01 .001 P−value .0001 c d 5 10 15 20 25 30 35 5 10 15 20 25 30 35 X (Å) X (Å) 1 .05 .01 .001 P−value .0001 e 1 .05 .01 .001 P−value .0001 f 5 10 15 20 25 30 35
X (Å)
Fig. 5.4 Lateral radial distribution functions of SDPE systems. P-values comparing differences in mean between and SDPC for the (a) rhodopsin and (b) opsin systems. rhodopsin and opsin were plotted for (e)SDPCand(f) Comparison of means tests were performed comparing SDPE SDPE to SDPC within the (c) rhodopsin and (d) opsin
In order to better understand the details of spot” is found in both the opsin and rhodopsin the cholesterol binding, we further expanded the simulations, as is a second region found on the data to look at the full 3-dimensional distribu- opposite side of the protein, between the cyto- tion. Figure 5.6 shows contours of regions with plasmic end of helix H3 and helices H4 and H5. high average cholesterol density superimposed This corresponds with a cholesterol interaction onto the structures of rhodopsin and opsin. The site predicted from a pair of 800 ns simulations brightest cholesterol spot in the 2D maps, which of the adenosine A2A receptor (Lee and Lyman, as mentioned excludes other lipids, corresponds 2012). with the region of high density beside H8, the Lastly, at the density contour level chosen, a intracellular helix lying parallel to the bilayer, third high density cholesterol region is present and packed against H1 and H7 of the protein. in the opsin system packed against helices H5 The density surface packs neatly behind the pair and H6. At lower contour thresholds, this region of post-translational palmitates on Cys322 and appears in the rhodopsin systems as well, but Cys323. In these simulations, H8 is embedded in its presence here indicates a possible difference the bilayer interface, creating a pocket in the hy- in cholesterol packing interfaces between the drophobic core that is too small for lipids to com- two structures. Given that the greatest structural fortably diffuse, leading us to conclude that in this changes between the chosen opsin and rhodopsin case, cholesterol is able to pack where other lipids structures are the orientations of helices H5 cannot penetrate efficiently. This cholesterol “hot and H6, as well as the elongation of helix H5 88 J.N. Horn et al.
ab c 30
20
10
0 Y (Å) −10
−20
−30
cd f 30
20
10
0 Y (Å) −10
−20
−30
−30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30 −30 −20 −10 0 10 20 30 X (Å) X (Å) X(Å)
0 0.5 1 1.5 2 2.5 3 3.5 0 0.002 0.004 0.006 0.008 0.01 Density P−value
Fig. 5.5 Density maps of (a)SDPEintheupper leaflet lipid components per Å2. Maps of p-values comparing the of rhodopsin, (b)SDPCintheupper leaflet rhodopsin, means for SDPE and SDPC for (c)theupper leaflet and (f) (d)SDPEinthelower leaflet of rhodopsin and (e)SDPC the lower leaflet are also shown. The protein orientation is in the lower leaflet of rhodopsin. Density is reported as identical to that in Fig. 5.3 in opsin, these changes in cholesterol packing et al., 2006b), as discussed in the methods sec- suggest a role for helix-helix interactions and tion. Unlike that work, here we account for the arrangement in cholesterol preferences. The size of the residues, and report the ratio of the presence of the extra cholesterol binding spot packing score to that of the average score of all is consistent with Fig. 5.2, which shows greater residues in the transmembrane region (Eq. 5.1). cholesterol binding at the surface of the opsin Figure 5.7 shows the protein structures colored system. This is indicative of a general trend by residue score. The residues clustered into three that the opsin structure is more amenable to well-defined groups based on physical location; cholesterol binding, again either because the Table 5.1 lists the residues with values at least surface has greater preference for cholesterol three times greater than the average. These group- or the more “open” opsin structure provides a ings correspond nicely with the three high density greater number of cholesterol accessible pockets. regions for cholesterol as seen previously for the The above visualizations are convenient, but opsin structure. Overall, the high-scoring clusters do not in themselves tell us precisely which pro- bear a striking resemblance to the groups of tein residues are involved in the “binding sites.” residues identified in the previous all-atom work Accordingly, we decided to track specific lipid- (Grossfield et al., 2006b). residue contacts, using a variant of the pack- The most dominant cluster is a collection ing score applied in previous work (Grossfield of residues that pack between helices H1 and 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 89
H8, with the greatest contribution to the cluster score coming from the palmitoyl chains attached to H8. In recent ˇ2-adrenergic receptor crystal structures, cholesterol and palmitic acids have been resolved in the H1/H8 interface, close to where the palmitoyl post-translational modifica- tions in rhodopsin are located (Cherezov et al., 2007). However, it has been suggested that these cholesterols may exist as an artifact of crystal packing or protein dimerization. Our simulations suggest a real effect exists, showing preferential interactions between cholesterol and rhodopsin at helix H8, despite the presence of only one rhodopsin molecule. Despite success correlating our results with previous simulation and crystallographic data, we have been unable to detect significant con- tacts between cholesterol and the groove on the intracellular ends of helices H1, H2, H3 and Fig. 5.6 3D images of regions of high cholesterol density H4, which had been detected in previous all- (gray) for both rhodopsin and opsin. Bottom panels are top down views, as seen from the extracellular side. atom simulations (Khelashvili et al., 2009) and in Rhodopsin and opsin are colored with a spectrum from structures of the ˇ2-adrenergic GPCR (Cherezov the N-terminus (blue) to the C-terminus (red) et al., 2007; Hanson et al., 2008). We were also unable to locate cholesterol binding sites that correspond with three sites identified in a recent high-resolution structure of the adenosine A2A receptor (Liu et al., 2012).Oneofthesesites, between the intracellular ends of helices H1 and H2, was noted in all-atom simulations (Lyman et al., 2009) and crystal structures of the ˇ2- adrenergic structure (Hanson et al., 2008), and absent in other simulations of A2A (Lee and Lyman, 2012).
5.5 Conclusions
The membrane environment around rhodopsin contains a diverse set of constituents that impact receptor activation, from general bilayer struc- tural properties to specific binding interactions. Utilizing coarse-grained simulation and a com- Fig. 5.7 3D images of rhodopsin and cholesterol inter- bination of radial distribution functions, density actions, color coded from low contact in yellow (binding representations and quantitative binding scans, score less than 3) to high contact in red (binding score we explored and identified a number of bilayer- greater than 3) rhodopsin interactions. 90 J.N. Horn et al.
Table 5.1 Clusters of key cholesterol-binding residues identified via binding scan Structure Group Residues Average score 1 43, 46, 50, 53, 56, 294, 301, 321, 322, 323 6.0683 Rhodopsin 2 126, 159, 206, 209, 210, 213, 214, 220, 221 3.6042 3 263 3.3597 1 50, 53, 54, 56, 318, 321, 322, 323 6.7907 Opsin 2 126, 159, 162, 206, 209, 210, 213, 214, 220 3.9135 3 256, 259, 263 3.4568
DHA chains were found in higher concentra- uniquely suited to interactions with cholesterol tions at the protein surface, with stearoyl chains are then unable to interact. The use of an all-atom excluded to a second solvation shell in the bilayer. model would overcome these limitation to some There was enrichment of PE headgroups over degree, but the loss of sufficient sampling remains PC headgroups at the surface of the protein. A a barrier to utilizing these models to explore a region of significant difference was discovered, process as slow as bilayer lateral reorganization. suggesting a possible specific binding site for In the future, a number of other variables could lipids with PE headgroups. Possible cholesterol- be explored. First, it would be worth exploring binding sites were also identified, with the pre- variations in concentration of lipid components in dominant one at the helix H1 and helix H8 inter- the bilayer. Cholesterol concentrations vary with face, behind the palmitoyl chains attached to the the age of the disc membrane and can result in protein. a drastic change in rhodopsin activity. Also, an We also found differences between the improved MARTINI model that accurately main- rhodopsin and opsin systems for these lipid con- tains tertiary structure would remove the need for stituents. In the opsin system, the concentrations restraints and allow us to explore the effect of of cholesterol, stearoyl and DHA reach bulk the lipid bilayer and specific lipid species on the levels deeper in the protein and the stearoyl and structure of rhodopsin. DHA peaks were much higher at the surface of In designing a simulation, there are always two the protein, suggesting a more open structure questions to answer: Does the proposed model with greater available surface area. represent the system with sufficient accuracy that The use of restraints to maintain protein sta- one can draw conclusions from the results? Can bility limited the motions available to rhodopsin the problem be solved using available resources? and opsin, likely preventing any major structural The former questions is answered by evaluating changes that would result from bilayer-protein the force field (and perhaps the system size and interactions. We do not feel that this is an issue contents), while the latter has to do with the for our particular simulations, as we are probing computational cost of running the simulations preferential interaction sites along the surface of long enough to obtain valid statistics. Satisfying the protein, not the effects of these interactions both criteria at the same time—finding a model on protein structure. Our chosen rhodopsin and that is accurate and readily solvable—is often opsin structures represent distant endpoints along a challenge, in that more detailed models are the activation path of rhodopsin, allowing us usually more expensive to implement. Actually, to explore the effect of these major structural an analogous phenomenon exists experimentally, changes on the surface available for interaction. for instance when choosing between more real- Coarse-grained models can limit the ability istic in vivo experiments and simpler (but more of the simulations to capture specific binding interpretable) in vitro ones. interactions. For example, coarse-grained repre- We believe that there is often a certain synergy sentations of cholesterol maintain the molecule’s to combining all-atom and coarse-grained simu- hydrophobicity, but cannot capture the chem- lations, as we have done here. The all-atom cal- ically distinct faces. Interaction sites that are culations can serve to validate surprising results 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 91 from the coarse-grained ones; usually, this role is Botelho AV, Huber T, Sakmar TP, Brown MF (2006) played by comparison to experiment, but some- Curvature and hydrophobic forces drive oligomeriza- times there is no readily available experimental tion and modulate activity of rhodopsin in membranes. Biophys J 91:4464Ð4477 comparison. Alternatively, the all-atom simula- Brown MF (1994) Modulation of rhodopsin function by tions may suggest interesting phenomena, but properties of the membrane bilayer. Chem Phys Lipids lack sufficient duration to draw strong conclu- 73(1Ð2):159Ð180 sions. In this case (as happened here), coarse- Brown MF (1997) Influence of non-lamellar-forming lipids on rhodopsin. Curr Top Membr 44:285Ð356 grained simulations can be used to flesh out the Brügger B, Erben G, Sandhoff R, Wieland FT, Lehmann case. WD (1997) Quantitative analysis of biological mem- In a broader sense, if one asks the question brane lipids at the low picomole level by nano- electrospray ionization tandem mass spectrometry. “Do I want to use all-atom or coarse-grained Proc Natl Acad Sci 94(6):2339Ð2344 simulation or experiment?”, the best answer is Burkhart BM, Li N, Langs DA, Pangborn WA, Duax simply “Yes”. All approaches have their own WL (1998) The conducting form of gramicidin a is a strengths and weaknesses, and the best way to right-handed double-stranded double helix. Proc Natl Acad Sci 95(22):12950Ð12955. doi:10.1073/pnas.95. answer the scientific question is to attack from as 22.12950 many directions as possible. Buzhynskyy N, Salesse C, Scheuring S (2011) Rhodopsin is spatially heterogeneously distributed in rod outer Acknowledgements We would like to thank Nick segment disk membranes. J Mol Recognit 24(3):483Ð Leioatts, Dejun Lin and Tod Romo for critical reviews 489. doi:10.1002/jmr.1086 of this manuscript. We would also like to gratefully Chen Z, Rand R (1997) The influence of choles- acknowledge financial support from the U.S. National terol on phospholipid membrane curvature and bend- Institutes of Health (1R01GM095496). We also thank the ing elasticity. Biophys J 73(1):267Ð276. doi:10.1016/ University of Rochester’s Center for Integrated Research S0006-3495(97)78067-6 Computing for the computing resources necessary to Cherezov V, Rosenbaum DM, Hanson MA, Rasmussen support this work. SGF, Thian FS, Kobilka TS, Choi HJ, Kuhn P, Weis WI, Kobilka BK, Stevens RC (2007) High- resolution crystal structure of an engineered human beta2-adrenergic G protein-coupled receptor. Science References 318(5854):1258Ð1265. doi:10.1126/science.1150577 Cruickshank C, Minchin R, Dain AL, Martinac B Albert AD, Young JE, Yeagle P (1996) Rhodopsin- (1997) Estimation of the pore size of the large- cholesterol interactions in bovine rod out segment disk conductance mechanosensitive ion channel of Es- membranes. Biochim Biophys Acta 1285:47Ð55 cherichia coli. Biophys J 73(4):1925Ð1931. doi:10. Andersen OS, Koeppe RE (2007) Bilayer thickness and 1016/S0006-3495(97)78223-7 membrane protein function: an energetic perspective. Delange F, Merkx M, Bovee-Geurts PHM, Ann Rev Biophys Biomol Struct 36:107Ð130. doi:10. Pistorius AMA, Degrip WJ (1997) Modulation 1146/annurev.biophys.36.040306.132643 of the metarhodopsin I/metarhodopsin II Andersen O,Apell HJ, Bamberg E, Busath D, Koeppe equilibrium of bovine rhodopsin by ionic R, Sigworth F, Szabo G, Urry D, Woolley A (1999) strength. Eur J Biochem 243(1Ð2):174Ð180. doi: Gramicidin channel controversy Ð the structure in a 10.1111/j.1432-1033.1997.0174a.x lipid environment. Nat Struct Mol Biol 6(7):609Ð609 Drews J (2000) Drug discovery: a historical perspective. Boesze-Battaglia K, Albert AD (1989) Fatty acid compo- Science 287(5460):1960Ð1964 sition of bovine rod outer segment plasma membrane. Engelman DM (2005) Membranes are more mosaic Exp Eye Res 49(4):699Ð701 than fluid. Nature 438(7068):578Ð580. doi:10.1038/ Boesze-Battaglia K, Albert AD (1992) Phospholipid dis- nature04394 tribution among bovine rod outer segment plasma Fattal DR, Ben-Shaul A (1993) A molecular model for membrane and disk membranes. Exp Eye Res lipid-protein interactions in membranes: the role of 54(5):821Ð823 hydrophobic mismatch. Biophys J 65:1795Ð1809 Boesze-Battaglia K, Hennessey T, Albert AD (1989) Feller SE, Gawrisch K (2005) Properties of docosahex- Cholesterol heterogeneity in bovine rod outer segment aenoic acid-containing lipids and their influence on disk membranes. J Biol Chem 264(14):8151Ð8155 the function of the GPCR rhodopsin. Curr Opin Struct Botelho AV, Gibson NJ, Thurmond RJ, Wang Y, Brown Biol 15:416Ð422 MF (2002) Conformational energetics of rhodopsin Feller SE, Gawrisch K, MacKerell AD Jr (2002) Polyun- modulated by nonlamellar-forming lipids. Biochem- saturated fatty acids in lipid bilayers: intrinsic and istry 41:6354Ð6368 environmental contributions to their unique physical properties. J Am Chem Soc 124(2):318Ð326 92 J.N. Horn et al.
Feller SE, Gawrisch K, Woolf TB (2003) Rhodopsin Khelashvili G, Grossfield A, Feller SE, Pitman MC, exhibits a preference for solvation by polyunsaturated Weinstein H (2009) Structural and dynamic effects docosohexaenoic acid. J Am Chem Soc 125(15):4434Ð of cholesterol at preferred sites of interaction with 4435. doi:10.1021/ja0345874 rhodopsin identified from microsecond length molec- Gibson NJ, Brown MF (1993) Lipid headgroup and acyl ular dynamics simulations. Proteins 76(2):403Ð417. chain composition modulate the MI-MII equilibrium doi:10.1002/prot.22355 of rhodopsin in recombinant membranes. Biochem- Killian J, von Heijne G (2000) How proteins adapt istry 32:2438Ð2454 to a membrane-water interface. Trends Biochem Sci Grossfield A, Feller SE, Pitman MC (2006a) Contribution 25(9):429Ð434. doi:10.1016/S0968-0004(00)01626-1 of omega-3 fatty acids to the thermodynamics of mem- Knepp AM, Periole X, Marrink SJ, Sakmar TP, Huber brane protein solvation. J Phys Chem B 110(18):8907Ð T (2012) Rhodopsin forms a dimer with cytoplasmic 8909. doi:10.1021/jp060405r helix 8 contacts in native membranes. Biochemistry Grossfield A, Feller SE, Pitman MC (2006b) A 51(9):1819Ð1821. doi:10.1021/bi3001598 role for direct interactions in the modulation of Kung C, Martinac B, Sukharev S (2010) rhodopsin by omega-3 polyunsaturated lipids. Proc Mechanosensitive channels in microbes. Natl Acad Sci USA 103(13):4888Ð4893. doi:10.1073/ Ann Rev Microbiol 64(1):313Ð329. doi: pnas.0508352103 10.1146/annurev.micro.112408.134106 Grossfield A, Pitman MC, Feller SE, Soubias O, Gawrisch Lee AG (2003) Lipid-protein interactions in biological K (2008) Internal hydration increases during activation membranes: a structural perspective. Biochimica et of the G-protein-coupled receptor rhodopsin. J Mol Biophysica Acta (BBA) Ð Biomembranes 1612(1):1Ð Biol 381(2):478Ð486. doi:10.1016/j.jmb.2008.05.036 40. doi:10.1016/S0005-2736(03)00056-7 Gruner SM (1985) Intrinsic curvature hypothesis for Lee AG (2004) How lipids affect the activities of in- biomembrane lipid composition: a role for nonbilayer tegral membrane proteins. Biochimica et Biophysica lipids. Proc Natl Acad Sci 82(11):3665Ð3669 Acta (BBA) Ð Biomembranes 1666(1Ð2):62Ð87. doi: Hanson MA, Cherezov V, Griffith MT, Roth CB, Jaakola 10.1016/j.bbamem.2004.05.012 VP, Chien EY, Velasquez J, Kuhn P, Stevens RC (2008) Lee JY, Lyman E (2012) Predictions for cholesterol in- A specific cholesterol binding site is established by the teraction sites on the A(2A) adenosine receptor. J 2.8 angstrom structure of the human beta2-adrenergic Am Chem Soc 134(40):16512Ð16515. doi:10.1021/ receptor. Structure 16(6):897Ð905. doi:10.1016/j.str. ja307532d 2008.05.001 Liebman PA, Parker KR, Dratz EA (1987) The molecular Harold FM, Baarda JR (1967) Gramicidin, valinomycin, mechanism of visual excitation and its relation to and cation permeability of Streptococcus faecalis. J the structure and composition of the rod outer seg- Bacteriol 94(1):53Ð60 ment. Ann Rev Physiol 49(1):765Ð791. doi:10.1146/ Haswell E, Phillips R, Rees D (2011) Mechanosensi- annurev.ph.49.030187.004001 tive channels: what can they do and how do they Liu W, Chun E, Thompson AA, Chubukov P, Xu F, do it? Structure 19(10):1356Ð1369. doi:10.1016/j.str. Katritch V, Han GW, Roth CB, Heitman LH, IJz- 2011.09.005 erman AP, Cherezov V, Stevens RC (2012) Struc- Hess B, Kutzner C, van der Spoel D, Lindahl E (2008) tural basis for allosteric regulation of GPCRs by GROMACS 4: algorithms for highly efficient, load- sodium ions. Science 337(6091):232Ð236. doi:10. balanced, and scalable molecular simulation. J Chem 1126/science.1219218 Theory Comput 4(3):435Ð447. doi:10.1021/ct700301q Lyman E, Higgs C, Kim B, Lupyan D, Shelley JC, Farid Hoover WG (1985) Canonical dynamics: equilibrium R, Voth GA (2009) A role for a specific choles- phase-space distributions. Phys Rev A 31(3):1695Ð terol interaction in stabilizing the apo configuration 1697. doi:10.1103/PhysRevA.31.1695 of the human A(2A) adenosine receptor. Structure Huber T, Rajamoorthi K, Kurze VF, Beyer K, Brown MF 17(12):1660Ð1668. doi:10.1016/j.str.2009.10.010 (2002) Structure of docosahexaenoic acid-containing Marrink SJ, Risselada HJ, Yefimov S, Tieleman DP, phospholipid bilayers as studied by 2H NMR and de Vries AH (2007) The MARTINI force field: coarse molecular dynamics simulation. J Am Chem Soc grained model for biomolecular simulations. J Phys 124:298Ð309 Chem B 111(27):7812Ð7824. doi:10.1021/jp071097f Huber T, Botelho AV, Beyer K, Brown MF (2004) Marrink SJ, Periole X, Tieleman DP, de Vries AH (2010) Membrane model for the G-protein-coupled receptor Comment on “on using a too large integration time rhodopsin: hydrophobic interface and dynamical struc- step in molecular dynamics simulations of coarse- ture. Biophys J 86:2078Ð2100 grained molecular models” by M. Winger, D. Trzes- Jones E,Oliphant T, Peterson P, et al (2001Ð) SciPy: open niak, R. Baron and W. F. van Gunsteren, Phys. Chem. source scientific tools for Python. http://www.scipy. Chem. Phys., 2009, 11, 1934. Phys Chem Chem org/ Phys 12(9):2254Ð2256; author reply 2257Ð2258. doi: Kandt C, Ash WL, Tieleman DP (2007) Setting up and 10.1039/b915293h running molecular dynamics simulations of mem- Marsh D (2008) Protein modulation of lipids, and vice- brane proteins. Methods 41(4):475Ð488. doi:10.1016/ versa, in membranes. Biochimica et Biophysica Acta j.ymeth.2006.08.006 5 Coarse-Grained Molecular Dynamics Provides Insight into the Interactions of Lipids... 93
(BBA) Ð Biomembranes 1778:1545Ð1575. doi:10. Crystal structure of rhodopsin: a G protein-coupled 1016/j.bbamem.2008.01.015 receptor. Science 289:739Ð745 Martinac B (2011) Bacterial mechanosensitive channels Park JH, Scheerer P, Hofmann KP, Choe HW, Ernst OP as a paradigm for mechanosensory transduction. Cell (2008) Crystal structure of the ligand-free G-protein- Physiol Biochem 28(6):1051Ð1060 coupled receptor opsin. Nature 454(7201):183Ð187. Martinac B, Buechner M, Delcour AH, Adler J, Kung C doi:10.1038/nature07063 (1987) Pressure-sensitive ion channel in Escherichia Parrinello M, Rahman A (1981) Polymorphic transitions coli. Proc Natl Acad Sci 84(8):2297Ð2301 in single crystals: a new molecular dynamics method. MARTINI (2011) http://mdchemrugnl/cgmartini/ J Appl Phys 52(12):7182Ð7190. doi:10.1063/1.328693 Mitchell DC, Straume M, Miller JL, Litman BJ Periole X, Huber T, Marrink SJ, Sakmar TP (2007) (1990) Modulation of metarhodopsin formation G protein-coupled receptors self-assemble in dynam- by cholesterol-induced ordering of bilayer ics simulations of model bilayers. J Am Chem Soc lipids. Biochemistry 29(39):9143Ð9149. doi: 129(33):10126Ð10132. doi:10.1021/ja0706246 10.1021/bi00491a007 Periole X, Cavalli M, Marrink SJ, Ceruso MA (2009) Molday RS (1998) Photoreceptor membrane proteins, Combining an elastic network with a coarse-grained phototransduction, and retinal degenerative diseases. molecular force field: structure, dynamics, and in- The Friedenwald Lecture. Invest Ophthalmol Vis Sci termolecular recognition. J Chem Theory Comput 39(13):2491Ð2513 5(9):2531Ð2543. doi:10.1021/ct9002114 Monticelli L, Kandasamy S, Periole X, Larson R, Tiele- Periole X, Knepp AM, Sakmar TP, Marrink SJ, Huber man D, Marrink S (2008) The MARTINI coarse T (2012) Structural determinants of the supramolec- grained forcefield: extension to proteins. J Chem The- ular organization of G protein-coupled receptors in ory Comput 4:819Ð839 bilayers. J Am Chem Soc 134(26):10959Ð10965. doi: Mouritsen OG, Bloom M (1984) Mattress model of lipid- 10.1021/ja303286e protein interactions in membranes. Biophys J 46:141Ð Pitman MC, Grossfield A, Suits F, Feller SE (2005) Role 153 of cholesterol and polyunsaturated chains in lipid- Mouritsen OG, Bloom M (1993) Models of lipid-protein protein interactions: molecular dynamics simulation interactions in membranes. Ann Rev Biophys Biomol of rhodopsin in a realistic membrane environment. Struct 22:145Ð171 J Am Chem Soc 127(13):4576Ð4577. doi:10.1021/ Needham D, McIntosh TJ, Evans E (1988) ja042715y Thermomechanical and transition properties Romo TD, Grossfield A (2009) LOOS: an extensible of dimyristoylphosphatidylcholine/cholesterol platform for the structural analysis of simulations. bilayers. Biochemistry 27(13):4668Ð4673. doi: Conf Proc IEEE Eng Med Biol Soc 2009:2332Ð2335. 10.1021/bi00413a013 doi:10.1109/IEMBS.2009.5335065 Neuringer M (2000) Infant vision and retinal function Romo TD, Grossfield A (2012) LOOS: a lightweight in studies of dietary long-chain polyunsaturated fatty object-oriented software library. LOOS: Lightweight acids: methods, results, and implications. Am J Clin object oriented structure analysis, Grossfield Lab. Nutr 71(1 Suppl):256SÐ267S http://loos.sourceforge.net Niu SL, Mitchell DC, Litman BJ (2002) Manipulation of Sansom MS, Bond PJ, Deol SS, Grottesi A, Haider S, cholesterol levels in rod disk membranes by methyl-ˇ- Sands ZA (2005) Molecular simulations and lipid- cyclodextrin. J Bio Chem 277:20139Ð20145 protein interactions: potassium channels and other Nosé S, Klein ML (1983) Constant pressure molecular membrane proteins. Biochem Soc Trans 33(Pt 5):916Ð dynamics for molecular systems. Mol Phys 50:1055Ð 920. doi:10.1042/BST20050916 1076 Simmonds A, East J, Jones O, Rooney E, McWhirter O’Connell A, Koeppe R, Andersen O (1990) Kinet- J, Lee A (1982) Annular and non-annular binding ics of gramicidin channel formation in lipid bilay- sites on the (Ca2CC Mg2C)-ATPase. Biochimica et ers: transmembrane monomer association. Science Biophysica Acta (BBA) Ð Biomembranes 693(2):398Ð 250(4985):1256Ð1259. doi:10.1126/science.1700867 406. doi:10.1016/0005-2736(82)90447-3 Okada T, Sugihara M, Bondar AN, Elstner M, Entel P, Soubias O, Gawrisch K (2007) Docosahexaenoyl chains Buss V (2004) The retinal conformation and its envi- isomerize on the sub-nanosecond time scale. J ronment in rhodopsin in light of a new 2.2 angstrom Am Chem Soc 129(21):6678Ð6679. doi:10.1021/ crystal structure. J Mol Biol 342:571Ð583 ja068856c Olausson BES, Grossfield A, Pitman MC, Brown MF, Soubias O, Gawrisch K (2012) The role of the lipid Feller SE, Vogel A (2012) Molecular dynamics simu- matrix for structure and function of the GPCR lations reveal specific interactions of post-translational rhodopsin. Biochim Biophys Acta 1818(2):234Ð240. palmitoyl modifications with rhodopsin in membranes. doi:10.1016/j.bbamem.2011.08.034 J Am Chem Soc 134(9):4324Ð4331. doi:10.1021/ Soubias O, Teague WE, Gawrisch K (2006) Evidence for ja2108382 specificity in lipid-rhodopsin interactions. J Biol Chem Palczewski K, Kumasaka T, Hori T, Behnke CA, Mo- 281(44):33233Ð33241. doi:10.1074/jbc.M603059200 toshima H, Fox BJ, Le Trong I, Teller DC, Okada Soubias O, Niu SL, Mitchell DC, Gawrisch K T, Stenkamp RE, Yamamoto M, Miyano M (2000) (2008) Lipid-rhodopsin hydrophobic mismatch 94 J.N. Horn et al.
alters rhodopsin helical content. J Am Chem Soc metarhodopsin II formation in visual transduction. J 130(37):12465Ð12471. doi:10.1021/ja803599x Am Chem Soc 124(26):7690Ð7701 Soubias O, Teague WE, Hines KG, Mitchell DC, Wiedmann TS, Pates RD, Beach JM, Salmon A, Brown Gawrisch K (2010) Contribution of membrane elastic MF (1988) Lipid-protein interactions mediate the energy to rhodopsin function. Biophys J 99(3):817Ð photochemical function of rhodopsin. Biochemistry 824. doi:10.1016/j.bpj.2010.04.068 27:6469Ð6474 van der Spoel D, Lindahl E, Hess B, Groenhof G, Mark Wiener MC, White SH (1992) Structure of a fluid AE, Berendsen HJC (2005) GROMACS: fast, flexible, dioloeoylphosphatidylcholine bilayer determined by and free. J Comput Chem 26(16):1701Ð1718. doi:10. joint refinement of x-ray and neutron diffraction data: 1002/jcc.20291 III. Complete structure. Biophys J 61:434Ð447 Valiyaveetil FI, Zhou Y, MacKinnon R (2002) Lipids Winger M, Trzesniak D, Baron R, van Gunsteren in the structure, folding, and function of the KcsA WF (2009) On using a too large integration time K+ channel. Biochemistry 41(35):10771Ð10777. doi: step in molecular dynamics simulations of coarse- 10.1021/bi026215y grained molecular models. Phys Chem Chem Phys Wang Y, Botelho AV, Martinez GV, Brown MF (2002) 11(12):1934Ð1941. doi:10.1039/b818713d Electrostatic properties of membrane lipids coupled to Beyond Standard Molecular Dynamics: Investigating 6 the Molecular Mechanisms of G Protein-Coupled Receptors with Enhanced Molecular Dynamics Methods
Jennifer M. Johnston and Marta Filizola
Abstract The majority of biological processes mediated by G Protein-Coupled Receptors (GPCRs) take place on timescales that are not conveniently ac- cessible to standard molecular dynamics (MD) approaches, notwithstand- ing the current availability of specialized parallel computer architectures, and efficient simulation algorithms. Enhanced MD-based methods have started to assume an important role in the study of the rugged energy landscape of GPCRs by providing mechanistic details of complex receptor processes such as ligand recognition, activation, and oligomerization. We provide here an overview of these methods in their most recent application to the field.
Keywords Enhanced methods ¥ Biased simulations ¥ Metadynamics ¥ Umbrella sampling ¥ Free-energy calculations ¥ Coarse-graining ¥ Ligand recog- nition ¥ GPCR activation ¥ Molecular mechanisms ¥ Oligomerization ¥ Dimers ¥ GPCR structure ¥ GPCR dynamics ¥ GPCR function
termed GPCR signalosomes (Huber and Sakmar 6.1 Introduction 2011), are suggested to operate through a multitude of dynamic steps and allosteric GPCRs are components of complex macromolec- signaling conduits whose properties may not ular machines with multiple ligand-induced depend necessarily on their individual elements. ‘active’ states that can engender different Thus, it appears evident that a conceptual signaling outputs through association with framework strongly relying on a combination specific accessory proteins. These functionally of high-content experimental platforms and versatile macromolecular complexes, recently computational approaches is necessary to account for the complexity of these signalosomes and J.M. Johnston ¥ M. Filizola () Department of Structural and Chemical Biology, Icahn ultimately understand the relationships between School of Medicine at Mount Sinai, One Gustave L. Levy their structure, dynamics, and function (Huber Place, New York, NY 10029, USA and Sakmar 2011). e-mail: marta.fi[email protected]
M. Filizola (ed.), G Protein-Coupled Receptors - Modeling and Simulation, Advances 95 in Experimental Medicine and Biology 796, DOI 10.1007/978-94-007-7423-0__6, © Springer ScienceCBusiness Media Dordrecht 2014 96 J.M. Johnston and M. Filizola
One way to interpret the functional versatility receptor (B2AR) (Dror et al. 2011)), these of GPCRs is in terms of their structural plasticity, hardware/software configurations are presently through system representation as energy land- the prerogative of a small group of investigators. scapes (Deupi and Kobilka 2010; Choe et al. Undoubtedly, atomistic descriptions provide 2011; Vaidehi and Kenakin 2010). Building upon the most comprehensive and complete repre- a concept put forth in 1991 by Frauenfelder sentations of a GPCR, but it is arguable that and colleagues (1991), the energy landscape of multiple, lengthy stochastic simulations of such a protein is described as a hyper-surface in a systems are not best suited to access so-called conformational space of very high dimension, “rare events” in the lives of GPCRs. Since the with a very large number of valleys (confor- number of calculations required per step of a mational sub-states) and peaks (energy barriers). MD simulation of a GPCR system scales with The valleys and peaks of such a rugged landscape the square of the number of particles, reducing cannot be classified individually, but must be the system size considerably increases the speed described by distributions. Specific energy levels with which simulations can be performed. This are no longer relevant, but are more accurately size reduction can be achieved by eliminating described in statistical terms. At any given time, the explicit representation of a component of an individual protein molecule can jump be- the system (e.g. the solvent or the membrane, as tween conformational sub-states as it navigates in Generalized Born models), or by grouping its own energy landscape, and this exploration individual atoms into interaction sites (e.g., is critically dependent on temperature (Frauen- coarse-grained bead models). Of course, such felder et al. 1991). The available X-ray crystal decisions must be taken with care and are structures of GPCRs (for a review see Katritch inevitably dependent on the nature of the question 2012) represent single, static sub-states of these that the simulation is designed to answer. In the proteins. Sections of missing density in many of cases where the fastest degrees of freedom can the structures (e.g. Wu et al. 2012) highlight the be neglected, these approaches have helped to inherent flexibility, even at low temperatures, of smooth the energy landscape of GPCRs and some regions of the receptor, particularly parts of thereby extend the range of accessible time and the intra and extracellular loops and the N- and length scales. C-termini, and remind us that the dynamic and The reduced representation of a GPCR sys- adaptable nature of the GPCR structure is a key tem may, however, still be insufficient to bridge part of their functionality. the gap between the timescales accessible to Ligands with different efficacies can modulate standard MD simulations and average experi- the GPCR energy landscape by shifting the mental timescales. We will illustrate here how conformational equilibrium towards active or both simplified physical representations and en- inactive conformations, thus eliciting different hanced MD-based methods have recently proven physiological responses (Deupi and Kobilka useful in the study of the rugged energy land- 2010; Vaidehi and Kenakin 2010). Examining scape of GPCRs by providing otherwise inac- the molecular basis of this ligand-induced cessible details of important events in the life conformational shift is not an easy task using cycle of these receptors, such as recognition of standard MD approaches due to the limited ligands, ligand-specific conformational changes, timescales they can access. Although much and oligomerization. However, we acknowledge has been accomplished by standard MD of that both the rate at which advances are being atomistically-represented GPCRs using some made in this field, and the space constraints of of the largest multiprocessor computing clusters, this chapter do not permit an entirely compre- running the most efficient MD codes currently hensive review here, but rather our goal is to available (e.g., see the work recently done to provide an overview by way of a few selected simulate ligand binding to the “2-adrenergic examples. 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 97
than rhodopsin. However, for drug discovery 6.2 Ligand Recognition in GPCRs purposes, a limitation of GPCR homology models has been that the binding pockets are not A thorough understanding of the mechanisms always big enough to accommodate the variety of by which GPCRs recognize their ligands is fun- ligands that have been demonstrated to bind to a damental to successful drug discovery. While particular receptor through other experimental multiple long timescale atomistic standard MD assays. Indeed, retinal is a small, covalently simulations (of the order of hundreds of s) bound ligand, so ligands of interest that are have recently permitted researchers to visualize, different in shape or size to retinal, and most for instance, the binding of different ligands to importantly, not covalently bound, might not be the B2AR (Dror et al. 2011), biased MD tech- expected to be a good fit to receptor models niques have been successfully employed to en- generated from a rhodopsin backbone template. hance the probability of observing such events Kimura and colleagues (2008) have developed a during shorter simulation timescales. Three stud- method for increasing the available space for ies we discuss in this section focus on the premise ligand binding within the orthosteric pocket that from knowledge of the pathways by which a of a model of a GPCR, which may be readily ligand can exit a receptor, one can infer specific transferable to other types of binding pockets for details of the binding mechanism (Gonzalez et al. other proteins. Once the binding pocket has been 2011; Selvam et al. 2012; Wang and Duan 2009). expanded using their algorithm, the endogenous The probability of spontaneous ligand exit from ligand can be docked into the pocket, and induced a binding cavity, on the timescale accessible to fit docking methods, sidechain rotamer sampling MD simulations, can be enhanced by addition of or MD methods can be used to refine the pocket an external force imposed upon the ligand, as is before virtual screening or docking of novel the case for steered MD and random acceleration ligands is performed. MD. We compare the results of these studies Conceptually, the method proposed by Kimura with those obtained by the very long timescale, and colleagues is similar to inflating a balloon unbiased simulations from Dror and colleagues inside the pocket. By slowly increasing the pres- (2011). A fourth study we discuss below focuses sure on the cavity walls during a MD simulation, on a binding event in which the ligand travels the pocket can be expanded. To achieve this, along a predetermined pathway from the bulk the authors placed several (>20) Lennard-Jones solvent, towards a bound state within the receptor (LJ) beads of small radius (0.25 Å) within the (Provasi et al. 2009). We also comment on a cavity, and these particles collided with the cavity novel method of altering the orthosteric binding walls with velocities consistent with a Maxwell pocket of receptor homology models, to permit distribution according to the temperature of the successful docking of ligands of differing shape simulation. A slow increase in pressure was im- and size to that in the template structure (Kimura plemented by increasing the radii of the particles. et al. 2008). The particles filled the pocket, and were tethered by weak (k D 1 kcal/(mol Å2)) harmonic forces to their four immediate neighbors, but these bonds 6.2.1 Exploring the Binding Site were reassigned with every increase of the par- in Homology Models ticle radius. The radius was increased in steps of 0.05Ð0.1 Å every 300Ð500 ps. The simula- Prior to the application of receptor engineering tions were performed using NAMD, with ex- innovations that have given rise to the recent plicitly represented solvent and palmitoyl-oleoyl- influx of solved crystal structures of GPCRs, phosphatidyl-choline (POPC) lipid bilayer. The homology modeling was the only tool in system was minimized and equilibrated for sev- the arsenal of the researcher wishing to eral hundreds of picoseconds prior to the pocket computationally investigate any receptor other expansion stages. During the expansion stages, 98 J.M. Johnston and M. Filizola the backbone dihedral angles of the protein were r D 0.3 state satisfied available experimental data weakly restrained (k D 2.0 kcal/(mol rad2)) to of agonist ligand binding, although required the prevent gross deformations and unwinding of the re-orientation of some side chains (Strader et al. TM helices. 1988, 1989). It seems that key advantage pro- Kimura and colleagues tested their protocol vided by the balloon expansion methodology is on three receptors; firstly, in their 2008 paper that it generates an ensemble of models, unbiased (Kimura et al. 2008), they generated an apo by the inherent dependence on the template in- state of the rhodopsin structure (PDB ID: 1LH9 troduced during the homology modeling process, (Okada et al. 2002)). By removing the retinal which may then be evaluated in the context of the ligand and simulating the receptor for a period available mutagenesis, SAR and small molecule of 4 ns, the sidechains were rearranged such pharmacophoric data. However, these expanded that the pocket was too small to accommodate models require further refinement before the re- retinal without modification. After the pocket sults of any docking studies may be considered expansion simulations, docking studies were per- any more than qualitatively accurate. formed with GLIDE (Friesner et al. 2004, 2006; Halgren et al. 2004) and a comparison of the crystal structure, the MD relaxed structure and 6.2.2 Exploring Ligand Egress the structure after expansion with the balloon Pathways with Random potential indicate that re-docking the retinal is Acceleration MD significantly improved between the relaxed struc- ture and the balloon expansion structure, but the Random acceleration MD (Wang and Duan 2007, crystal structure pose is not recovered exactly, 2009), also called random expulsion MD (Lude- with an RMSD of 7.2 Å between the re-docked mann et al. 2000) uses a randomly directed, exter- pose and the crystallographic pose. nally applied force to encourage unbinding of a Secondly, they built a homology model of ligand from the pocket of a receptor. This ensures the chemokine receptor type 2 (CCR2) based thorough exploration of the pocket, and good on a rhodopsin template, Three potent antago- sampling of possible modes of ligand egress. nists, RS-504393 (MacKerell et al. 1998), TAK- According to the formulation, the direction of the 779 (Baba et al. 1999), and a Teijin lead com- force applied to the center of mass (COM) of the pound (Shiota 1999) were docked into the recep- ligand is chosen at random, and it is maintained tor at different stages of the pocket expansion. for N steps. During these steps, the COM of the At r D 0.35 Å, a cluster of poses correlating with ligand is expected to move a minimum distance the available mutagenesis data was observed. rmin, i.e. the average velocity,
Fig. 6.1 Exit/entry pathways (red transparent arrows, labeled 1Ð4) for the B1AR (left)andthe B2AR (right). Top (top panels)andside (bottom panels) views. Receptor is shown in cartoon representation, colored blue (TM1) to red (H8). The crystallized antagonists cyanopindolol (in B1AR) and carazolol (in B2AR) are shown in stick representation
between TM2, TM3 and TM7 at the extracellu- Atomic Force Microscopy (AFM), by applying lar opening. Subsequently, the ligand interacted a time-dependent external force to the system with F193 as it passed through the TM7-EL2 through the atomistically fine tip, and measuring junction on the way to the bottom of the binding the mechanical resistance properties of the pocket, where it was oriented and stabilized by biomolecule. Such a process can be mimicked the polar interactions with TM3, TM5 and TM7. virtually, using steered MD. Steered MD uses The difference in results between the B2AR and the application of an external force to drive rhodopsin studies described here, alongside spec- the system towards a desired state, be it via a ulative reports of different lipid phase ligand conformational change, or a ligand-unbinding entry pathways for e.g. the cannabinoid receptor event. Steered MD is a non-equilibrium (Hurst et al. 2010) and for the PAR1 (Zhang et al. technique, during which the MD pathway 2012) indicate that ligand binding pathways may is irreversible. Analysis of the position and not be general across receptor types. interactions of the dissociating ligand, and the evolution of the applied forces during a forced ligand unbinding by steered MD can 6.2.3 Forced Ligand Unbinding provide reliable qualitative insights into the by Steered MD irreversible work required for the unbinding process (Isralewitz et al. 1997; Jensen et al. Under experimental conditions, unbinding 2002; Fishelovitch et al. 2009; Yang et al. 2009). of a ligand can be monitored by means of Such information may enable researchers to 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 101 determine structural features of these receptor- 3 ns each. The force profile, as a function of ligand complexes that contribute crucially to simulation time, revealed the ease with which the ligand binding. Unlike random acceleration MD, ligand can be extracted from the receptor along the direction of the force must be predetermined, the different pathways (Gonzalez et al. 2011). To though usually this is unknown. The egress address the shortcomings of an arbitrary choice of pathway can therefore often be arbitrary, and exit pathway from the receptors, CAVER (Petrek as such the sampling of different possible modes et al. 2006) was used to establish feasible egress of ligand exit may be less comprehensive than pathways connecting the orthosteric binding with random acceleration MD. pocket with the surface of the receptor. This In steered MD, the elastic force is proportional showed two possible pathways, depicted in to the change in the spring extension, relative to Fig. 6.1, by red arrows labeled 1 and 2. Two its equilibrium length: additional, lipid phase exit pathways between TM1/7 and TM5/6 inter-helical clefts were also ! ! ! F.t/ D k vt r.t/ r 0 n (6.2) tested (red arrows 3 and 4 in Fig. 6.1). Though not found by the CAVER exploration, they have where k is the spring constant; v is the constant been implicated as entry/exit pathways for retinal velocity of pulling, mimicking the retracting can- in rhodopsin (Hildebrand et al. 2009), and an tilever; r0 and r(t) are the ligand center of mass exit pathway between TM5/6 was found to be position at initial and current time t respectively significant in the random acceleration MD study and n is the direction of the pulling vector. of rhodopsin, from Wang and Duan (2007) The potential of mean force (PMF) along the (neither of these pathways was found to be reaction coordinate was calculated by the second- significant in their investigation of B2AR (Wang order cumulant expansion of the irreversible work and Duan 2009)). In Gonzalez’s investigation, measurements (Park et al. 2003) according to: the initial force peaks for extracting ligands Z t through pathways 3 and 4 were found to be twice W.t/ D v F.t/dt (6.3) that for extraction via pathway 1 or pathway 0 2, so the lipid phase extraction channels were considered to be unfavorable for the adrenergic 1 2 2 PMF DhW i hW i hW i (6.4) receptors. For B2AR, extraction of carazolol 2kB T via pathway 1 (bounded by TM5, 6 and 7) was where
6.2.4 Exploring Ligand Binding DOP receptor was simulated for 500 ns, in Pathways Using a hydrated DPPC-cholesterol lipid bilayer, Well-Tempered Metadynamics using ten walkers. All simulations were performed using GROMACS 4.0.5 (Van der Well-tempered metadynamics (Barducci et al. Spoel et al. 2005) with PLUMED (Bonomi et al. 2008) is an enhanced sampling technique 2009). that enables more efficient exploration of To represent the binding event, two CVs were the multidimensional free energy surfaces of chosen to describe (1) the distance between the biological systems by adding Gaussian bias to a center of mass of the heavy atoms of NLX and the standard MD simulation (Laio and Parrinello center of mass of the heavy atoms of the alkaloid- 2002; Leone et al. 2010). The dynamics is binding pocket of opioid receptor and (2) the biased by a non-Markovian (history-dependent) distance between the center of mass of the DOP potential, constructed as a sum of Gaussian receptor alkaloid binding pocket and the center of “hills” localized along the trajectory in the mass of the heavy atoms of the middle residues of direction of the reaction coordinate (or collective the EL2. This second CV was selected to enable variable, CV) of interest. The accumulation enhanced conformational sampling of the EL2 re- of this biasing potential enables the simulated gion of the DOP receptor, given its uncertain pre- system to overcome high energy barriers in dicted structure (this work was performed before order to explore efficiently its free energy the crystal structure of the DOP receptor became landscape as defined by the CVs. The success of available (Granier et al. 2012)). Using these two a metadynamics study is predicated on a careful CVs, we reconstructed the free-energy surface of apriorichoice of a set of CVs to provide a the NLX binding event, determining that the non- satisfactory description of the process of interest. selective antagonist NLX exhibits a molecular In well-tempered metadynamics, the height of recognition site on the DOP receptor surface at the added Gaussian hills depends both on a a cleft formed by EL2 and EL3, and ends in a temperature scaling factor and the underlying preferred orientation into the receptor alkaloid bias, and decreases to zero once a given energy binding pocket, after visiting some less stable threshold is reached. As a result, convergence of states in between. The most stable NLX-bound the algorithm to the correct free energy profile state of DOP corresponded to a conformation can be proven rigorously, and exploration of in which NLX interacted directly with D1283.32 physically relevant regions of the conformational via a salt bridge, and with Y3087.43, W2746.48, space is ensured for complex systems. H2786.52,in broad agreement with experimental In 2009, we performed well-tempered data from mutagenesis and competition binding metadynamics to elucidate the mechanistic assays (Befort et al. 1996a, b;Lietal.1999; details of flexible-ligand, flexible-protein Mansour et al. 1997; Bot et al. 1998; Spivak et docking of naloxone (NLX), a non-selective al. 1997; Surratt et al. 1994). The ligand was antagonist for opioid receptors, from the bulk stabilized in a specific orientation through inter- water environment into the orthosteric binding actions with a number of residues on TM3, TM5, pocket of a B2AR-based homology model of TM6 and TM7, in particular: M1323.35 as well as • opioid (DOP) receptor (Provasi et al. 2009). F2185.43 and F2225.47. The recent crystal struc- Using the multiple walker method (Raiteri et ture of the DOP bound to naltrindole confirmed al. 2006), several well-tempered metadynamics some key interactions in the bound state, in par- simulations can be simultaneously performed, ticular, the interactions with Y3087.43, H2746.52 each contributing to the same history-dependent M1323.36 and D1283.32 were observed. On the bias potential; this is a key advantage of pathway to this final binding mode, NLX visited a the method, enabling the most efficient use number of different less stable alternative pockets of cluster computing resources. Thus, the and in particular, two metastable states charac- 104 J.M. Johnston and M. Filizola terized by different degrees of opening of EL2. reduced system representations and steered MD, The existence of these metastable states before as well as combinations of adiabatic biased MD binding to the orthosteric pocket is reminiscent and metadynamics techniques. We also discuss of the pathways observed for the B2AR using ligand-specific conformations obtained depend- steered MD (Gonzalez et al. 2011) and unbiased ing on the physiological response of the bound MD (Dror et al. 2011), though the intermediate ligand. In particular, we discuss applications to poses visited by NLX along the binding pathway two GPCRs, the B2AR and the 5HT2A serotonin are, as expected, not the same as those observed receptor, and their perceived success as judged by for ligand binding to B2AR. Some residue posi- virtual ligand screening and, in the case of B2AR, tions are, however, implicated to interact along when compared with the recently published crys- the entry pathway for both receptors, in particular tal structures. Y/L7.35, H/W6.58 and D300/D290 in the B2AR and DOP receptor respectively. Sampling of the NLX ligand in the bulk sol- 6.3.1 Using Guided MD to Build vent was corrected for, using a methodology put Activated GPCRs forward by the Roux lab (Roux 1999; Allen et al. 2004). Using the free energy surface from There are very few currently available X-ray the metadynamics simulation, with the collec- crystal structures of GPCRs displaying the char- tive variables described above, we obtained a acteristics of an active state, and these are: the restrained free-energy profile, following the pro- ligand free opsin crystal structures at either low tocol and equations described in (Provasi et al. pH (PDB ID: 3CAP (Park et al. 2008)) or in 2009). Our calculated equilibrium constant, of complex with a synthetic peptide derived from Keq D 80 ˙ 13 nM, for the final bound state of carboxy terminus of the alpha-subunit of the het- NLX at DOP was remarkably close to the ma- erotrimeric G protein (PDB ID: 3DQB (Scheerer jority of reported experimental values (e.g. Toll et al. 2008)); the crystal structure of the proposed et al. 1998), and thus, this methodology offers active state of the A2A adenosine receptor (PDB great potential for describing, quantitatively, the ID: 3QAK, 2YDO (Xu et al. 2011; Lebon et al. binding events of other GPCRs. 2011)); the crystal structures of metarhodopsin II (PDB ID: 3PXO (Choe et al. 2011)) and a consti- tutively active metarhodopsin II (PDB ID: 4A4M 6.3 Activation Mechanisms (Deupi et al. 2012)); and finally the crystal struc- in GPCRs tures of B2AR in complex with the Gs protein (PDB ID: 3SN6 (Rasmussen et al. 2011b)), and Activation processes of GPCRs are known to a camelid nanobody (PDB ID: 3P0G (Rasmussen occur on timescales that are inaccessible to cur- et al. 2011a)). rent simulations using standard MD. From recent In 2011, two groups combined experimentally crystal structures, the most pronounced, common derived restraints and ligand binding data from conformational rearrangements that mark the ac- extensive literature searches with steered MD to tivation of a receptor include: breaking of the so- guide the generation of the active states of the called “ionic lock” between the E/DRY motif in B2AR (Simpson et al. 2011) and the 5HT2A re- TM3 and acidic residues in TM6, upon a large ceptor (Isberg et al. 2011), and have assessed the (i.e. 11 Å in the case of B2AR (Rasmussen et al. success of their approaches by virtual screening 2011a)) outward movement and slight rotation of of test sets of known agonists, antagonists and the intracellular end of TM6; smaller (i.e. 6Å inverse agonists, diluted among non-binders. for B2AR (Rasmussen et al. 2011a)) outward To build an initial model for the active state movement of TM5; and some smaller slightly of the B2AR, Simpson and colleagues have com- inward movements of TM3 and TM7. Here we bined the intracellular part of the opsin crystal describe activated GPCR models obtained using PDB ID: 3DQB (Scheerer et al. 2008) with the 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 105 extracellular portion of the inactive B2AR crystal colleagues modeled the whole heterotrimeric structure at 2.1 Å (PDB ID: 2RH1 (Cherezov assembly, using structural details from a number et al. 2007)) to generate a template, and used of different crystallographic structures (namely an alignment based on common motifs in the PDB IDs: 1AZS (Tesmer et al. 1997), 1GOT transmembrane regions as input for MODELLER (Lambright et al. 1996) and 1GP2 (Wall version 9v1 (Fiser and Sali 2003). Two hundred et al. 1995)). initial models were narrowed down to a single The initial model of the B2AR from Simpson model that scored well according to the MOD- and colleagues was energy minimized using ELLER objective function (measuring spatial re- the CHARMM (Brooks et al. 1983) force field straint violation) and had a low backbone RMSD and an implicitly represented membrane/solvent from the template. environment described by the Generalized Born Isberg and colleagues generated a single ho- algorithm with the simple switching function mology model of the 5HT2A based on the B2AR (GBSW) (Im et al. 2004). In the GBSW model, inactive structure, PDB ID: 2RH1 (Cherezov et the influence of the membrane is included as al. 2007). They modified this structure, to match a solvent-inaccessible, infinite planar slab of the active characteristics found in the opsin struc- low dielectric constant. A simple smoothing ture, PDB ID: 3DQB (Scheerer et al. 2008). function is included to approximate the dielectric Specifically, (1) the lower parts of TM5 and TM6 boundary between the “bulk water” and the were tilted outwards, manually; (2) the missing “membrane” (Im et al. 2003). For this study, density of IL3 in the B2AR structure was re- the membrane thickness was set at 35 Å with placed by the opsin IL3 structure, and (3) the a smoothing length of 5 Å. The use of the G’i peptide was inserted into the structure and implicit membrane can minimize problems mutated to G’s and then subsequently to the associated with complete sampling of phase 3.50 G’q subtype. Finally, (4), the rotamer of R131 space during short standard MD simulations, was set to their G protein interacting confor- since the number of explicit particles for mation, and W2866.48 was set to its presumed which pair-wise interactions are required to be active conformation, as predicted by some earlier calculated is dramatically reduced. In particular investigations (see e.g. Holst et al. 2010; Shi et for this application, this meant that the entire al. 2002, among others), but not yet observed in G protein assembly could be included without any active receptor crystals to date (see supple- the prohibitive contribution to the number of mentary figure 7a of Taddese et al. 2012). This particles from explicitly represented solvent. modified structure was then used as the input to The simulation length total was approximately MODELLER version 9v6 (Eswar et al. 2007). 175 ns. The choice to use an implicit solvent In agreement with subsequent observations model could be controversial, given previous of the activated B2AR during long timescale reports of a structural and functional role for MD simulations (Rosenbaum et al. 2011), H-bonding interactions of waters entering the which revealed that interaction with either the rhodopsin structure upon activation (Grossfield G protein or a nano-body G protein surrogate is et al. 2008), and the internal crystallographic essential to maintaining an activated receptor waters noted in many X-ray derived structures of conformation, both groups have included GPCRs. Simpson and colleagues have indirectly interactions with the G protein to complete incorporated the effects of such solvation, their activated conformations. The 3DQB through specific experimentally derived restraints opsin structure represents the apo-receptor, that would open out parts of the structure stabilized by a peptide derived from the main sufficient to permit entry of water molecules upon binding site of the heterotrimeric G protein, activation under explicit solvation conditions. the C-terminus of the ’ subunit (G’CT). Isberg and colleagues embedded their Isberg and colleagues included the G protein model of the 5HT2A receptor in an explicitly peptide, mutated to G’q while Simpson and represented DPPC membrane, with the TIP4P 106 J.M. Johnston and M. Filizola explicit water model and counterions under derived from site-directed spin labeling, disulfide the OPLS-AA 2001 (Kaminski et al. 2001) cross-linking, engineered zinc binding and site- forcefield. Their simulations were carried out directed mutagenesis experiments. The restraints using the Schrödinger implementation of the were introduced slowly, as the system was heated DESMOND software package, version 2.4 to simulation temperature, and all but the toggle originally developed by DE Shaw and associates switch restraints were maintained throughout the (Chow et al. 2006). Perhaps reflecting the 150 ns production simulation. increased number of degrees of freedom, these The success of the models, in both cases, was simulations are shorter in duration, four stages probed by virtual screening during which it was of 5 ns simulations with restraints for a total demonstrated that the predicted active models of 20 ns. The choice of DPPC may not be successfully discriminated known receptor- the best lipid for activation studies of GPCRs, selective agonists from antagonists. Simpson since Vogel and colleagues presented evidence and colleagues refined their active structure that, at least for rhodopsin, the fully saturated by flexibly docking full agonists epinephrine, chains of DPPC can inhibit the activation isoprotenerol, TA-2005 and salmeterol into transition to the receptor activated meta II state the binding pocket, and screened these refined (Vogel et al. 2004). Furthermore, the melting receptors using GLIDE (Friesner et al. 2004, temperature of DPPC is 41 ıC, and so, under 2006; Halgren et al. 2004). They were able physiological conditions, the membrane might to check all binding poses, because they be expected to be gel-like, casting some doubt used a fairly small set of ligands, comprising on the choice of this lipid as a valid membrane 29 full and partial B1/B2AR agonists, 38 mimetic. Nevertheless, Isberg and colleagues B1/B2AR antagonists and inverse agonists, ensured a fluid membrane environment for their and 56 non-peptide ligands selective for 5HT2A model, by maintaining their simulation opioid, muscarinic, neurokinin, neurotensin, conditions at T D 325 K ( 51 ıC). bradykinin and serotonin receptors. Docking Both groups derived sets of restraints from scores demonstrated that the active model previously published experimental information to showed a preference for B1/B2AR agonists define their activated states, and in both cases, over antagonists, inverse agonists and other during the earliest parts of the simulations, har- ligands. Active enrichment curves showed that monic restraints were applied to the backbone of the enrichment over the first 10 % of ligands was transmembrane regions of the receptor, to prevent significantly greater than expected at random. distortions of the helical structure. Isberg and Furthermore, enrichment for B2AR-selective colleagues applied pairwise harmonic distance agonists was higher than for B1AR-selective constraints, collated from mutagenesis and X- agonists, and enrichment for antagonists was ray crystallographic data, to inter-helical, ligand- only observed once all the agonists had been receptor and G protein-receptor distances. These found. Isberg and colleagues performed a larger constraints were used to drive the conforma- scale virtual screen, also using GLIDE. They tion of the 5HT2A towards a presumed activated selected 182 known 5HT2A ligands, including structure, achieved once they are satisfied. Simp- 139 agonists (112 phenylethylamine agonists son and colleagues applied six sets of harmonic and 27 others), 39 antagonists and 4 inverse distance restraints to the B2AR receptor model agonists. These were diluted with 9257 diverse during MD simulations of 150 ns in production compounds from the ZINC database (Irwin et al. length, to drive the conformation towards an 2012). GLIDE docking scores demonstrated that activated state. The basis of these restraints was the 29 highest ranked hits were phenylethylamine from experimental evidence sourced from the agonists, and these were the best scoring class of available literature on both B2AR and other class molecules by a large margin, with a mean G-score A GPCRs (in particular, the muscarinic recep- 1.5 lower than other agonists. Phenylethylamine tor and rhodopsin). The sets of restraints were antagonists also scored better than other agonists, 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 107 other antagonists, and other ligands. The virtual can be thought of as a direct consequence of screenings seem to demonstrate the success the simplification or averaging over interactions of modeling the active state of both receptors, collated from different structural/biophysical using restraints derived from experimental data. sources introduced by the homology modeling However, the subsequent solution of the active process, but can also be attributed to the static state crystal structures of B2AR (Rasmussen nature of a crystallographic structure versus et al. 2011a, b; Rosenbaum et al. 2011) has the dynamic manner in which the model was enabled direct comparison between the B2AR constructed. crystal and the active model, and Taddese and colleagues (2012) have discussed this comparison in a brief paper in 2012. Overall, 6.3.2 Combined Adiabatic Biased they report an average RMSD of 1.3 Å in the MD with Metadynamics invariant TM region (i.e. residues that have to Explore Activation the same environment when compared over a Pathways number of structures) between the active model from Simpson and colleagues (2011) and the As mentioned earlier in this chapter, and as active B2AR structure. The overall RMSD is observed in experimental studies, the value of perhaps less explanatory than the residue-by- GPCRs as pharmaceutical targets is enhanced residue variation between the B2AR model by the concept of “functional selectivity”, or and the B2AR crystal structure, which ranges “biased agonism”, a concept wherein a single from 1.0 Å in the TM region, to 7.2 Å in the ligand might display different efficacies at a loops. The larger value of RMSD for the loop single receptor isoform depending on the effector regions is not unexpected owing to their high pathway coupled to that receptor (Rives et al. mobility. The active B2AR model was based 2012). A simple mechanistic explanation for this on an active, G protein bound state of opsin, phenomenon can be derived from the dynamic and the difference between the opsin template equilibrium of GPCRs between different and the active B2AR crystal structure is 1.7 Å conformational sub-states. Different ligands or over the invariant TM region. The active model changes in protein-protein interactions (such shows good agreement with the overall position as may be observed in receptor dimerization) of the helices in the structure PDB ID: 3SN6 can shift the equilibrium population of these for TM1, TM2 and TM5, but seems more like sub-states toward conformations of the receptor the opsin template for TM3, TM4, TM6 and that are kinetically and structurally distinct. TM7. The orthosteric binding pocket appears to Evidence of different sub-states can be inferred have been successfully modeled, with an RMSD from crystallographic structures: not all of the of 1.3 Å between the residues in the active attributes usually associated with an ‘active- model and the crystal structure, and this may state’ receptor are present for active-state crystal be the reason for the strong enrichment results structures solved in the presence of different for the virtual screening. The key receptor-G agonists (Provasi et al. 2011). We have recently protein interactions were correctly predicted developed a general computational strategy from the opsin-G protein complex template, that combines adiabatic biased MD with path with 27/38 residues on the receptor and 20/34 CV metadynamics, which we have used to residues on the G protein surface correctly characterize possible metastable active states of predicted. Overall the modeling moderately rhodopsin (Provasi and Filizola 2010); to discern successfully reproduced the nature of the the effects of ligands with variable efficacies on activated characteristics observed in the crystal the conformation of the B2AR receptor (Provasi structure, but the atomic resolution specificity of et al. 2011) and, finally, to offer a mechanistic the interactions seen in the single conformation explanation of allosteric effects of ligand binding of the X-ray structure was not reproduced. This in a GPCR dimer of the 5HT2A serotonin receptor 108 J.M. Johnston and M. Filizola
Fig. 6.2 (a) Multidimensional scaling representation of inactive and active states. The receptor is shown in grey multiple adiabatic biased MD trajectories clustered ac- cartoon representation, in (b)and(c), except for TM5 and cording to RMSD. Representative structures selected from TM6, which show the progress from inactive to active, in each cluster (shown in (b): side view and (c): intracellu- colors corresponding to the clusters in (a) lar view) to define a homogenous pathway between the and the glutamate receptor (mGluR2) observed is bigger than the minimum value previously by our experimental colleagues (Fribourg et al. achieved during the trajectory. The adiabatic bi- 2011). ased MD trajectories were short (e.g. 1Ð5 ns), In the first stage of the method, a mono- with small elastic constants (0.01 kJ/nm2) and in- dimensional pathway defining activation of the dependently drawn Maxwellian initial velocities. receptor was derived. Frames depicting the re- All of the independent adiabatic biased MD tra- ceptor in intermediate states between ‘inactive’ jectories were then pooled, and all of the sampled and ‘active’ X-ray-derived crystal structures (or conformations clustered based on the RMSD of models, in the absence of crystal structures) de- the TM domain of the receptor. Representative fined the pathway. The frames were derived from structures from these clusters were then chosen multiple (e.g. 5 or 10) statistically independent as reference frames to define the path CVs for adiabatic biased MD simulations (Marchi and further metadynamics simulations. The number Ballone 1999). Like targeted MD (Schlitter et of clusters, and hence frames, was between 4 and al. 1994), adiabatic biased MD drives the system 10, depending on the nature of application. See towards the desired final state, by enforcing a Fig. 6.2 for an example of this. reduction in RMSD between the initial and target To obtain information about the relative stabil- conformations, to enable observations of a con- ity of the states populated by the receptor during formational transition during a single trajectory. transition from inactive to active conformations, In contrast to TMD, the adiabatic biased MD metadynamics simulations with path collective algorithm (Marchi and Ballone 1999) ensures variables (Branduardi et al. 2007) can be per- exploration of low-energy pathways by keeping formed. The CVs defining the phase space are the the total potential energy of the system constant position (s) along and the distance (z)fromthe during the MD run through application of a time- pre-determined activation pathway. The reference dependent biasing potential. The harmonic bias- frames, k, define the position of the progression ing potential is only applied when the RMSD of the system from inactive (s 0) to active 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 109
(s 1) during the pre-determined transition (CV and Filizola 2010) and NAMD 2.7 (Phillips et al. 1), and its distance (z) from it (CV 2), according 2005) with the Charmm27 force-field (MacKerell to the following equation: et al. 1998)forthe5HT2A, mGluR2 monomers and dimer (Fribourg et al. 2011). Xk 1 i 1 œd .i/ This combinatorial approach was initially val- s .R.t// D Z e TM R.t/; R k 1 idated on rhodopsin (Provasi and Filizola 2010), iD1 where the activation pathway was defined be- (6.5) tween the crystal structure of a photoactivated, deprotonated intermediate of rhodopsin (PDB ID: where • Dhd (R(i),R(i ˙ 1))i is the average dis- TM 2I37) and the low pH crystal structure of opsin tance between two adjacent frames in the transi- (PDB ID: 3CAP), the only crystal structure dis- tion pathway, the constant exponent, œ,mustbe playing characteristics of an active state avail- chosen in such a way that œ • Š 1, and Z is a able at the time of this study. Simulations were normalization factor defined by: performed for wild-type rhodopsin, embedded in a POPC membrane, with either a charged or an Xk œd .i/ 3.49 Z e TM R.t/; R (6.6) uncharged residue E134 within the E(D)RY iD1 motif, to simulate proton uptake from the bulk oc- curring in the late stage of rhodopsin activation. while Reconstruction of the rhodopsin free energy landscape indicated three common metastable 1 z .R.t// D œ log Z (6.7) states between 2I37 and 3CAP along the adia- batic biased MD pre-calculated transition path. The free-energy of the receptor as a function Two of the identified minima are comparable to of s and z can then be reconstructed from con- active intermediates of bovine rhodopsin, char- verged well-tempered metadynamics simulations acterized by different amplitude of the outward of varying length ( 80Ð300 ns) (see Provasi et al. movement of TM6. This was revealed by com- 2011; Provasi and Filizola 2010; Fribourg et al. paring intra-molecular distance analysis of these 2011 for details of the simulation protocols). states with results from double electron-electron Convergence of the reconstructed free-energy can resonance spectroscopy (Altenbach et al. 2008). be assessed in two ways: firstly, by ensuring The largest predicted separation between TM3 minimal variation in the free-energy difference and TM6 was in line with data obtained for Meta between the minima around the experimental IIb from spectroscopy (Knierim et al. 2007), and endpoints with time, and secondly, in a converged for opsin from crystallography (Park et al. 2008). simulation, exploration of the FES represented by Key residues, thought to influence the activa- the CVs faces no further energy barriers, there- tion pathway, were mutated and additional sim- fore one can expect frequent re-crossing of the ulations were carried out on these receptors for values of the CVs after convergence is attained. comparison to the WT. An interaction between All simulations in each of the applications K2315.66 and E2476.30 appeared to stabilize the of this method to date have been carried out predicted active conformation with the largest using an explicit atomistic representation of all separation between TM3 and TM6, as also seen components, and the receptors were embedded in the opsin crystal structure (Park et al. 2008). in solvated POPC/10 % cholesterol membranes. Furthermore, we carried out metadynamics sim- The versatility of the method lies in the im- ulations of the K231A5.66 mutant rhodopsin with plementation using Plumed plugin (Bonomi et either a charged or an uncharged residue E1343.49 al. 2009), meaning it has been performed using within the E(D)RY motif. We hypothesized that GROMACS (Van der Spoel et al. 2005) and the removal of a polar interaction between K2315.66 OPLS-AA force-field (Kaminski et al. 2001)for and E2476.30 by replacing K2315.66 with ala- rhodopsin and B2AR (Provasi et al. 2011;Provasi nine would switch the equilibrium toward the 110 J.M. Johnston and M. Filizola predicted conformation with a smaller separation Comparison of the free energy surfaces of between TM3 and TM6. Our simulations showed the receptor activation when complexed with that the putative active conformation of rhodopsin different ligands demonstrated selective inactive with the largest separation between TM3 and or active conformations depending on their TM6 was destabilized in the presence of a neu- known elicited functional responses. tral E1343.49 mimicking the activation-dependent In the absence of ligand, the receptor was proton uptake from the bulk. found to have two low energy conformations, one In the most comprehensive study using with s 0.2, i.e. close to the inactive state, and a this methodology (Provasi et al. 2011), we second, closer to an active state, with s 0.6, sep- were able to present compelling evidence for arated by a low energy barrier. This corresponds the concept that ligands of differing efficacy well with the known basal activity of the B2AR, access structurally distinct conformations of and the inherent conformational flexibility of the the B2AR. Path CV metadynamics simulations unliganded B2AR may help to explain the diffi- were conducted for six ligands, representing a culty in obtaining crystallographic structures of full agonist (epinephrine), weak partial agonists the B2AR in the absence of ligand. (dopamine and catechol) inverse agonists (cara- The free energy surface for B2AR complexed zolol and ICI-118,551) and a neutral antagonist with alprenolol, ostensibly a neutral antagonist, (alprenolol). These different ligands were also i.e. a ligand that competitively binds in the or- compared to the unliganded receptor. Parameters thosteric binding pocket of a receptor and blocks for the ligands were obtained manually, by the binding of (and thus cellular responses to) analogy with existing molecules in the OPLS- agonists or inverse agonists, also displayed two AA force-field, while Coulomb point charges stable states. One, close to the inactive, s 0.2, were obtained from quantum chemical geometry was marginally ( 1 kcal/mol) more stable than optimization using Gaussian 03 and restricted the second state, at s 0.6. This type of free en- Hartree-Fock calculations with the 6-31G* basis ergy profile illustrates possible reasons for some set. The ligands were either docked into the studies having found alprenolol to behave in dif- receptor using Autodock 4.0 (Morris et al. 1996) ferent ways in different assays: most notably, as (for dopamine, catechol and epinephrine) or an inverse agonist (Elster et al. 2007; Hopkinson positioned according to their crystallographically et al. 2000) or as a partial agonist (Callaerts-Vegh determined binding modes. The activation et al. 2004). pathway was homogenously defined by ten The profiles for the inverse agonists (ligands frames, between the inactive state of the B2AR, that preferentially stabilize a receptor in an inac- represented by PDB ID: 2RH1 (Cherezov et tive conformational state), ICI-118,551 and cara- al. 2007) and the active state, represented by zolol both displayed single conformational states the nanobody stabilized structure, PDB ID: with low values of s (s < 0.2), in deep free energy 3P0G (Rasmussen et al. 2011a). Metadynamics wells. The free energy minimum, in each case, simulations for each receptor/ligand complex was approximately 4 kcal/mol lower in energy were run for approximately 300 ns until the than the next lowest minima. This corresponds difference between the free energies of the with the solution of many crystal structures in metastable states observed was converged. The inverse agonist bound conformations. activation of the receptor was also measured in Results for the agonists were less clear-cut. terms of other structurally pertinent variables, The free energy profiles for both partial (catechol by reweighting the free energy surface. These and dopamine) and full (epinephrine) agonists variables were: the distance between the each displayed a deepest free energy minimum at sidechains of the R1313.50 and E2686.30 residues the expected value of s, i.e. s 0.6 for the partial of the ionic lock; the dihedral angle of the agonists, and s > 0.9 for the full agonist. How- sidechain of W2866.48, i.e. the rotamer toggle ever, other minima, only marginally less stable switch; and the outward displacement of TM6. than these were found in all cases, at lower values 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 111 of s. A feasible explanation for this arises from methysergide, and an inverse agonist, clozapine. the absence of G protein: as demonstrated by For the monomeric 5HT2A, clozapine and Rosenbaum and colleagues (2011), during very methysergide stabilized two different inactive- long timescale MD simulations, an active confor- like conformations, while DOI stabilized an mation of the B2AR will quickly lose many of active like conformation. These experiments the active characteristics in the absence of the G were repeated for dimeric arrangements of protein or a stabilizing nanobody surrogate. the receptors, with the lowest energy ligand- The results from this study provide an atomic stabilized conformations of 5HT2A complexed resolution view of structurally distinct receptor with a ligand-free mGluR2. When complexed conformations stabilized by different ligands, and with the clozapine-stabilized 5HT2A conforma- demonstrate that ligands with varying efficacies tion, the mGluR2 conformation is shifted toward might be used to control population shifts toward an activated state. In contrast, when bound to desirable GPCR conformations. Such knowledge either the DOI-stabilized active-like 5HT2A,or might be useful to aid rational drug design: biased the methysergideÐstabilized inactive-like 5HT2A ligands may represent novel, more efficacious conformation, the conformational equilibrium of therapeutics. mGluR2 is shifted toward an inactive-like state. Finally, in conjunction with tissue and animal These results (in combination with the tissue behavioral studies, we have recently applied and animal studies) suggest that the formation of this methodology to determine some of the the heteromeric complex enables modulation of mechanistic details driving the antipsychotic the mGluR2 response to its endogenous ligand properties of drugs targeting a heteromeric by binding of different ligands to the 5HT2A. arrangement of the serotonin 5HT2A and mGluR2 Therefore, the strong 5HT2A agonist, DOI, can (Fribourg et al. 2011), in particular, why some greatly stimulate Gq signaling, while decreasing 5HT2A inhibitors (e.g. clozapine) may exhibit Gi signaling, but the inverse agonist, clozapine, neuropsychological effects while others (e.g. abolishes Gq signaling while simultaneously methysergide) do not. Oligomeric receptor stimulating Gi signaling. These results go some complexes have been shown to have distinct way to explaining the complexities surrounding signaling properties when compared to their the disparity between the neuropsychological monomeric components, and these receptors effects of 5HT2A inhibitors, some of which have been shown to form a specific hetero- demonstrate antipsychotic effects, while others oligomeric complex in mammalian brain tissue do not. A drug bound to one receptor of the that is implicated in schizophrenia. The complex heteromer was shown to influence the signaling integrates the responses from the Gq-coupled response of the partner to its endogenous ligand mGluR2 and that of the Gi-coupled 5HT2A and such mechanistic insights are invaluable to modulate signal transduction and influence to therapeutic strategy design for disorders in downstream signaling events. To investigate this which oligomeric arrangements of receptors are computationally, and to provide a structural implicated. context for the experimental observations, we built homology models of both receptors: mGluR2 was rhodopsin based, using PDB 6.4 Oligomerization ID: 1U19 for the inactive state and 3DQB Mechanisms in GPCRs for the active-like state, while 5HT2A was based on B2AR, using 2RH1 as the inactive In the previous section, we have mentioned the template and 3P0G as the active-like template. distinct properties of oligomeric arrangements To these models, we applied the methodology of some GPCRs as compared to those of as in the previous cases, comparing the free their monomeric components (Milligan 2009). energy profiles of activation for 5HT2A bound However, the ability of GPCRs to assemble into to a strong agonist, DOI, a neutral antagonist, stable, heteromeric or homomeric, oligomeric 112 J.M. Johnston and M. Filizola arrangements remains a subject of much debate in residue-level detail is ensured by the extensive (Fonseca and Lambert 2009; Lambert 2010). calibration of a large library of chemical building From a computational perspective, understanding blocks against thermodynamic data (in particular, the thermodynamics and kinetics of GPCR oil/water partition coefficients). These versatile oligomerization is not a trivial problem owing building blocks permit the force field to flexibly to (1) the large size of the complex system, represent a large range of biomolecules without and (2) the stochastic nature of the process. the need to re-parameterize the model in each In this section we survey the enhanced MD- case. The MARTINI force field is compatible based methods that have recently been used in with GROMACS (Van der Spoel et al. 2005) and combination with reduced representations of the its CG mapping of heavy atoms (including water GPCR systems to study the dynamic behavior molecules) to beads is a 4:1 ratio, except in the of dimeric/oligomeric arrangements of these case of molecules containing rings, where a map- receptors. ping of 2:1 is used. There are four main types of beads, P D polar, N D non-polar, C D apolar and charged D Q, with subtypes depending on hy- 6.4.1 Spontaneous Self-Assembly drogen bonding capacity and polarity. Each CG of Coarse-Grained bead has a mass of 72 atomic mass units (amu), Representations of GPCRs equivalent to the mass of four water molecules, except for ring structure beads, which have a The number of calculations required by a bi- mass of 45 amu. This improves efficiency and en- ological molecule scales as the square of the ables a simulation timestep of 20Ð40 fs, approx- number of particles included in its representa- imately ten times that possible with an all-atom tion, and consequently, a lengthy simulation of GROMACS simulation. Based on comparison of a large, explicitly represented GPCR, particu- diffusion rates for atomistic and MARTINI CG larly in a dimeric or oligomeric state, becomes models, the CG model affords a scaling factor particularly difficult at atomistic resolution. This of 4 for the effective time sampled, though there issue has prompted several recent efforts focused has been some recent debate surrounding the best on reducing the number of particles included in time step (Winger et al. 2009; Marrink et al. a simulation of a protein and its environment, 2009). and a number of strategies have been proposed. The earliest application of the MARTINI CG We focus here on the MARTINI coarse-grained model to GPCRs was the self-aggregation of (CG) model (Marrink et al. 2008), one of several rhodopsin monomers in four different explicit, “bead models” that have been devised to coarsely CG, lipid membrane environments (Periole et al. represent biomolecules (see Tozzini 2005, 2010 2007). Systems with 16 CG rhodopsin monomers for recent reviews). (based on the 1L9H structure (Okada et al. 2002)) The MARTINI forcefield has proven to be a at a protein:lipid ratio of 1:100, were simulated successful means of studying oligomerization for for up to 8 s. Such time scales would be very GPCRs and has been used in extensive studies difficult to reach in an atomistic representation. of oligomerization for rhodopsin (Periole et al. The results of these simulations indicate a local- 2007, 2012) opioid receptors (Johnston et al. ized membrane adaptation to the presence of the 2011; Provasi et al. 2010) and the B1AR and receptor, supposedly driven by the motivation to B2AR (Johnston et al. 2012). Developed by Mar- overcome the hydrophobic mismatch between the rink and co-workers, the MARTINI force-field length of the hydrophobic part of the monomeric (Marrink et al. 2004, 2007, 2008; Monticelli et receptor and the equilibrium hydrophobic bi- al. 2008; Lopez et al. 2009) offers coarse-grained layer thickness. Hydrophobic mismatch has been (CG) representations of proteins, several lipid shown to promote self-assembly of rhodopsin types and cholesterol. Close comparability to reconstituted in membrane bilayers (Kusumi and both experimental and atomistic MD approaches Hyde 1982; Ryba and Marsh 1992). The details 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 113 of such phenomena could not be modeled in GPCRs. Two studies of dimerization of a simulation using an implicit representation of the DOP receptor were performed using a membrane, like the generalized Born method established methodologies, firstly umbrella (Im et al. 2004) described earlier in this chapter. sampling, from which we derived the PMF of Periole and colleagues simulated rhodopsin a dimerization event (Provasi et al. 2010), and in different phosphocholine (PC) lipid environ- secondly, a metadynamics study in which we ments under non-biased conditions, and they ob- established the most favorable orientation of served a clear dependence on chain length. Bi- the individual protomers involved in different layer adaptation was manifest as local thickening dimeric arrangements (Johnston et al. 2011), i.e. near helices TM2, TM4, and TM7 in (C12:0)2PC, comprised of different contact interfaces, of the (C16:1)2PC, and (C20:1)2PC bilayers and as lo- DOP receptor. The computational results of this cal thinning near helices TM1/H8 and TM5/TM6 study compared favorably with inferences from in (C20:1)2PC and (C20:0)2PC. The results in- cysteine crosslinking experiments, supporting the dicated a higher propensity for protein:protein role of specific residues at the interfaces. contact interactions in (C16:1)2PC. Clear reorga- To characterize dimerization for the DOP re- nization and increase in scope of the rhodopsin ceptor, we performed biased MD simulations dimer interfaces was observed during the simu- of a CG representation of a homo-dimeric ar- lations, indicating a search for a shape comple- rangement of this receptor in an explicitly CG mentarity that maximized the buried accessible represented, POPC:10 % cholesterol-water en- surface area. The number of contact interfaces vironment. Since these studies were conducted was higher in (C12:0)2PC and (C16:1)2PC, where prior to the crystallographic solution of the opioid strong forces were introduced by the hydrophobic receptor structures in 2012 (Wu et al. 2012; mismatch, than in (C20:1)2PC and (C20:0)2PC, Granier et al. 2012; Thompson et al. 2012; Man- where the forces were more balanced. Three glik et al. 2012), an all-atom structural model of contact interfaces were clearly visible on the 6Ð the DOP receptor protomer from Mus musculus, 8- s time scale in (C20:1)2PC; these included was generated by homology modeling, using the previously suggested homo- and heterodimeriza- crystal structure of the B2AR at 2.4 Å resolu- tion interfaces in rhodopsin and other GPCRs, tion (PDB ID: 2RH1) (Cherezov et al. 2007)in involving the exposed surfaces of the helices MODELLER 9v3 (Eswar et al. 2007), using the TM1/TM2/H8, TM4/TM5, and TM6/TM7. The same strategy we recently reported in the liter- authors concluded that hydrophobic mismatch ature for the human DOP receptor (Provasi et al. drives self-assembly of rhodopsin into liquid- 2009). The loop regions were built ab initio using like structures with short-range order, and that ROSETTA 2.2 (Wang et al. 2007). A pair of the the interactions may not be dictated by specific resultant DOP receptor models was placed facing residue-residue contacts at the contact surfaces. one another at a putative symmetrical interface, involving residue 4.58, inferred from cysteine cross-linking data on this and other GPCRs (see 6.4.2 Combining CG Representation e.g. Guo et al. 2005, 2008). and Biased MD to Investigate In an effort to improve the stability of the GPCR Dimerization secondary structure of the CG representation of the receptor, we combined an elastic By combining the MARTINI reduced representa- network model (ENM) with the MARTINI CG tion of the system with biased MD techniques, representation, using a method developed and it has been possible to make predictions about termed ELNEDIN by Periole and colleagues the relative strength of dimers formed at different (2009). ENMs are ideally suited to preserve the interfaces in an explicit membrane environment. secondary or tertiary structure of biomolecules, We pioneered the use of a free energy since they are structure derived, and therefore approach to characterize oligomerization of introduce an intrinsic bias toward the structure 114 J.M. Johnston and M. Filizola
Fig. 6.3 (a) Definition of collective variables (CVs) used 121 and Gly51 (shown as red spheres) on each protomer, in studies of dimers from the Filizola lab (Johnston et al. thus the variables are consistent across different interface 2011, 2012; Provasi et al. 2010). CV1 is the distance r arrangements. They describe the distance between the between the COMs of the protomers (CA)and(CB); CV2 receptors, d; the tilt of long axis of each receptor relative is the rotational angle A, defined by the projection on to the receptor-receptor direction ™1and™2; the rotation to the plane of the membrane of the COM of the TMi of the receptors around their long axis (parallel to the 0 0 of A,theCA,andtheCB, and CV3 corresponded to the membrane normal), ¥1, ¥3or™1 and ™2 , the relative 0 equivalent rotational angle B. Shown for the B2AR at a orientation of the receptor’s long axis, ¥2. d is the TM1/H8 interface, TM1 and H8 are colored black.The interfacial receptor distance and is defined as the distance, CVs are defined according to the interface of interest. d, between the receptors from which the distance at the (b) Definitions of the virtual bond algorithm used by minimum of the PMFs was subtracted, see supplementary Periole and colleagues in their recent publication (Periole material from (Periole et al. 2012). Once again, TM1 and et al. 2012). The axes (black dotted lines in Fig. 6.3b) H8 are colored black are anchored by the backbone CG beads of Cys 187, Gly upon which they are established. In a novel secondary structure (e.g. coil, bend, or turn), 2 extension to the ELNEDIN method, we applied a force constant of KSPRING D 250 kJ/mol/nm a secondary-structure-dependent construct to was applied. A comparison of the RMSF for models of the DOP receptor dimer (Provasi the C’ beads for the CG representation with et al. 2010). The strength of the force constant an equivalent atomistic simulation indicated a for the harmonic restraint, KSPRING,was qualitative agreement, permitting the secondary determined by the secondary structure of each structure of the helices to be maintained, without of the residues. If the residue was determined compromising the flexibility of the loop regions. to have a defined secondary structure (by We used the CVs illustrated in Fig. 6.3ato DSSP (Kabsch and Sander 1983)), e.g. ’- describe the relative position of interacting DOP helix as in the case of the TM regions of receptor protomers A and B during the simula- the DOP receptor, then a force constant of tions: CV1 represented the distance r between 2 KSPRING D 1,000 kJ/mol/nm was applied. For the COM of protomers A (CA) and B (CB); CV2 a sequence of >2 residues with undefined described the rotational angle A, defined by the 6 Beyond Standard Molecular Dynamics: Investigating the Molecular Mechanisms of G... 115 projection on to the plane of the membrane of potential. The advantage of umbrella sampling, the COM of the TM4 of A,theCA, and the considered to be key from a computational per- CB, and CV3 corresponded to the equivalent spective, is that the relative independence of the rotational angle B. To allow exploration of an windows from one another (though this should be experimentally-supported TM4 interface of DOP considered and confirmed when recombining the receptor homodimers involving position 4.58 in windows) allows the biased simulations along the a reasonable timescale, we limited the sampling length of — to be conducted in parallel on large ı of the two rotational angles A and B to a 25 computer clusters. interval, using steep repulsive potentials. To obtain the PMF, (W( )) for the range of The system setup was the same for both that is of interest, the simulations from each of the of the studies (Johnston et al. 2011;Provasi window must be unbiased and combined. Thus, et al. 2010). All simulations were performed it is crucial that the histogrammed variations in using GROMACS (Van der Spoel et al. 2005) the measured variable obtained from each of the incorporating the Plumed plugin (Bonomi et al. windows overlap sufficiently such that the whole 2009).Inthefirstofthesetwo,weusedumbrella of the coordinate space is covered. sampling to define the PMF of dimerization as a Roux offers a thorough description of the function of the separation between the centers of derivation in (Roux 1995), but briefly, the biased mass of the two protomers. distribution function, < ( )>(i) as obtained from Umbrella sampling was introduced in 1977 the ith biased ensemble, is: by Torrie and Valleau (1974), and has been con- 1 wi . /=kB T wi . /=kB T sidered the prototypic biased MD technique for h . /i.i/ D e h . /ihe i improving sampling of a PMF as defined by (6.10) the Kirkwood equation (Kirkwood 1935)(6.8), from the average distribution, < ( )>, along a where kBT is the Boltzmann factor and the unbi- predefined reaction coordinate, : ased PMF from the ith biased ensemble is: h . /i h . /i.i/ W . / D W . / kB T ln (6.8) Wi . / D W . / kB T ln h . /i h . /i