Review pubs.acs.org/CR
Structurally Enabled Discovery of Adenosine A2A Receptor Antagonists Ali Jazayeri,* Stephen P. Andrews,† and Fiona H. Marshall Heptares Therapeutics Limited, BioPark, Broadwater Road, Welwyn Garden City, Hertfordshire AL7 3AX, United Kingdom ABSTRACT: Over the past decade there has been a revolution in the field of G protein-coupled receptor (GPCR) structural biology. Many years of innovative research from different areas have come together to fuel this significant change in the fortunes of this field, which for many years was characterized by the paucity of high-resolution structures. The determination to succeed has been in part due to the recognized importance of these proteins as drug targets, and although the pharmaceutical industry has been focusing on these receptors, it can be justifiably argued and demonstrated that many of the approved and commercially successful GPCR drugs can be significantly improved to increase efficacy and/or reduce undesired side effects. In addition, many validated targets in this class remain to be drugged. It is widely recognized that application of structure-based drug design approaches can help medicinal chemists a long way toward discovering better drugs. The achievement of structural biologists in providing high-resolution insight is beginning to transform drug discovery efforts, and there are a number of GPCR drugs that have been discovered by use of structural information that are in clinical development. This review aims to highlight the key developments that have brought success to GPCR structure resolution efforts and exemplify the practical application of structural information for the discovery of adenosine A2A receptor antagonists that have potential to treat multiple conditions.
CONTENTS 1. Introduction 2. Methods in G Protein-Coupled Receptor Drug Discovery 3. Advances Enabling G Protein-Coupled Receptor Structural Biology 3.1. Protein Engineering Solutions 3.2. Technical Solutions 4. Role of Adenosine A2A Receptor in Different Diseases 4.1. Adenosine A2A Receptor in Inflammation and Respiratory Diseases 4.2. Adenosine A2A Receptor in Central Nervous System Diseases 4.3. Adenosine A2A Receptor in Cancer 5. Applying Structure-Based Drug Design Approaches for Adenosine A2A Receptor Antagonist Discovery 5.1. Virtual Screening at the Adenosine A2A Receptor 5.2. Structure-based Lead Optimization of Adenosine A2A Receptor Antagonists 6. Concluding Remarks Author Information Corresponding Author Present Address Notes Biographies References © XXXX American Chemical Society
1. INTRODUCTION G protein-coupled receptors (GPCRs) have long been established as a major target class for drug action, with over 30% of prescription drugs directed at this target class.1 This position has been maintained even in recent times with the progression of many alternative therapies such as protein and antibody therapeutics. GPCRs are expressed on all cells in the body and play a pivotal role in coordinating communication between cells as well as transmitting signals from hormones, neurotransmitters, and metabolites that maintain the body’s homeostasis. GPCR drugs are used across many therapeutic areas including neuroscience,2 cardiovascular disease,3 respiratory disease,4 inflammation,5 and gastrointestinal disorders.6 There are over 350 nonolfactory GPCRs in humans; however, only 25% of these have been drugged with either small molecules or protein therapeutics. Advances in genomics and genetics have revealed a wealth of new biology linking novel GPCR targets to disease.2,7,8 In the last 5 years, nine new GPCRs have been drugged with first-in-class (FIC) molecules reaching the market directed at orexin receptors (suvorexant),9 PAR1 (vorapaxar),10 β3 (mirabegron),11 P2Y12 (ticagrelor),12 GLP2 (teduglutide),13 SMO (vismodegib),14 5HT2c agonist (locarserin),15 S1P1 (fingolimod),16 and GHRH (tesamorelin).17 In order to maximize the richness of biology around the GPCR target class for therapeutic benefit, improvements are needed in our ability to discover safe
A B B C D F F G H
H H K M M M M M M M
Special Issue: G-Protein Coupled Receptors Received: February 16, 2016
A
DOI: 10.1021/acs.chemrev.6b00119 Chem. Rev. XXXX, XXX, XXX−XXX
Chemical Reviews
Review
Ticagrelor can be considered a FIC drug since it is the first orally available antagonist of the platelet P2Y12 receptor. This drug is used to treat thrombosis associated with acute coronary syndrome and other vascular disorders.12 The target was clinically validated with older drugs such as clopidogrel, discovered prior to identification and cloning of the P2Y12 receptor. The starting point for ticagrelor, like many GPCR drugs, was the endogenous ligand of the receptor, in this case adenosine diphosphate (ADP) and the natural antagonist adenosine triphosphate (ATP).31 ATP itself has low potency for the receptor and poor stability. Efforts were taken to make stable ATP analogues with increased affinity for the P2Y12 receptor. Based on understanding the structure of ATP, analogues were designed that eliminated the phosphates and altered the core purine and sugar moiety resulting in the identification of AR-C109318X, the first non-phosphate P2Y12 antagonist. Further chemistry to improve the PK properties led to ticagrelor. Many approaches have been used to identify drugs directed at GPCRs; however, the process has been hindered by the lack of available X-ray structures of the receptors. The recent availability of X-ray structures and the ability to carry out structure-based drug design at GPCRs is set to revolutionize how we carry out GPCR drug discovery in the future.
effective drugs for emerging targets. X-ray crystallography and structure-based design has revolutionized our ability to design drugs for soluble enzyme targets such as proteases and kinases; however, until recently, drug discovery at GPCRs has relied mainly on empirical or ligand-based methods in medicinal chemistry. In the past few years, X-ray structures of 35 GPCRs have been reported and these structures are now actively being used for structure-based drug design.18 In this review we focus on the adenosine A2A receptor as an example study where X-ray structures have been used to design and optimize novel antagonists with greatly improved properties compared to previous compounds identified via more conventional approaches. Adenosine A2A antagonists have potential utility in a variety of different disease areas,19,20 including Parkinson’s disease,21,22 attention deficit hyperactivity disorder (ADHD),23and more recently in the immune-oncology area.24,25
2. METHODS IN G PROTEIN-COUPLED RECEPTOR DRUG DISCOVERY Many drugs currently on the market directed at GPCRs are derived from natural products, such as opioids, or are derivatives of natural ligands of the receptor, such as β-blockers or antihistamines. Once receptors were cloned and recombinant cell-based assays were developed, high-throughput screening (HTS) of large compound libraries became the method of choice for initiation of GPCR drug discovery projects. Despite the extensive effort, HTS has not proved very effective in GPCR drug discovery, with only two of nine new FIC GPCR drugs in the last 5 years being derived from HTS (suvorexant and vorapaxar). Although aminergic receptors are highly tractable, many other GPCR target classes, including those with large peptide or protein ligands or lipid ligands, have proved less tractable, with HTS screens failing to identify progressible hits. When hits are identified, these may have a high molecular weight and high lipophilicity, making them difficult to optimize1,26 to compounds with good druglike properties.27 Compounds with high molecular weight and lipophilicity have an increased chance of offtarget toxicity, as was seen with the calcitonin gene-related peptide (CGRP) antagonist talcagepant and the orexin antagonist almorexant. Surprisingly, two of the nine FIC drugs were originally derived from phenotypic screening and only later were shown to mediate their effects through GPCRs (vismodegib and fingolimod). While this shows the power of phenotypic screening, these compounds were not ideally optimized for potency or selectivity for their relevant target.28 Not surprisingly, further work has been directed at the specific receptor, resulting in improved compounds that may replace the first-in-class molecules. For example, RPC1063 is an optimized S1P1 agonist compared to fingolimod. RPC1063 has increased potency and selectivity for the target receptor S1P1, while fingolimod is relatively nonselective across the S1P family.29 In addition, RPC1063 and other “second-generation” S1P1 agonists have an altered pharmacokinetic (PK) profile with a slower time to peak plasma concentration. These modifications are expected to reduce cardiovascular and other side effects seen with fingolimod. Two of the nine FIC drugs approved were peptides directed at class B GPCRs: teduglutide, a GLP2 agonist, and tesamorelin, a GHRH agonist. In addition to these drugs, many further peptide GLP1 agonists have been approved for the treatment of diabetes.30 So far, despite extensive high-throughput screening, no small-molecule agonist drugs directed at class B receptors have been identified. New approaches are clearly needed to obtain oral drugs at these receptors, given the strong therapeutic role of many of the secretin family of receptors.
3. ADVANCES ENABLING G PROTEIN-COUPLED RECEPTOR STRUCTURAL BIOLOGY It is widely recognized that GPCRs and indeed other membrane proteins are difficult proteins for structural studies because these proteins are hydrophobic in nature, which makes their purification challenging as they aggregate in aqueous solution. To overcome this problem, membrane proteins need to be prepared in detergent to make them miscible in water, which in turn results in occlusion of important contact surfaces required for nucleation and crystal formation. To further complicate matters, membrane proteins and particularly GPCRs exhibit significant structural and conformational flexibility. Part of this flexibility is only manifested once the receptors are taken out of membrane bilayer and into detergent micelles and is caused by removal of the lateral pressure that the membrane exerts on the receptors. Also, receptors have long and unstructured termini and loops that are flexible and interfere with crystallization. In addition, conformational flexibility is an inherent feature of these proteins is needed to translate the ligand binding event on the extracellular surface to G protein activation on the intracellular surface. Most recent experiments indicate that an unliganded receptor can exist in different inactive conformations that exchange within hundreds of microseconds. Upon agonist binding, the equilibrium shifts toward a conformation capable of engaging cytoplasmic G proteins. Critically, careful analysis of this equilibrium shift indicates that even after agonist binding there remains significant conformational heterogeneity and the coexistence of inactive and active states in the same population.32 These data demonstrated the degree of heterogeneity that normally is present in GPCRs, which presents a major hurdle to formation of well-diffracting crystals. Finally, GPCRs and other membrane proteins are difficult to express at high levels due to the engagement of the secretory machinery that is required for transportation of fully folded proteins to the cell surface. Over the past decade, a number of technical advances have resulted in a significant increase in the success rate of GPCR structure resolution efforts (Figure 1). These solutions can be categorized as either protein engineering solutions, which aim to improve the characteristics of the receptors, or technical solutions, B
DOI: 10.1021/acs.chemrev.6b00119 Chem. Rev. XXXX, XXX, XXX−XXX
Chemical Reviews
Review
In addition to scanning mutagenesis, stabilized receptors can be generated by random mutagenesis.45−48 There are two key requirements for generating stabilized receptors by random mutagenesis: first, a high-throughput selection strategy is needed to identify the population containing the stabilized variants, and second, the selection method must maintain a link between DNA and the protein it encodes, so the single stable variants can be identified from the enriched population. This has been achieved by expressing receptors in the membrane of Escherichia coli or Saccharomyces cerevisiae and selecting the correctly folded receptors by use of a fluorescently labeled ligand. The population of cells expressing the correctly folded receptor can then be selected and iteratively enriched through fluorescence-activated cell sorting (FACS).45,46,49 However, such a method allows selection only for increases in expression, and although the variants with increased expression do exhibit some increased stability, there is no direct relationship between stability and expression levels per se.36 It is therefore better to select for increased stability in detergent directly. The key development that has allowed this approach to be adapted to directly select receptors on the basis of stability in detergent has been the application of an encapsulation technique that allows bacterial membrane to be solubilized without dispersion of the solubilized receptors and the plasmid that encodes for these receptors.47,48 Importantly, the encapsulated particles can be combined with a fluorescently labeled ligand and the encapsulated cells can be sorted by FACS to select for populations harboring receptors with increased stability. Following rounds of iteration, the DNA encoding these variants can be isolated and sequenced. This method clearly allows access to a significantly larger mutational space than scanning mutagenesis and in principle could be faster and less labor-intensive. However, the main drawback of this approach is that the encapsulation method is currently applicable only to bacterial cells, and many GPCRs cannot be expressed as functional proteins in the inner membrane as bacteria cannot support their expression and proper folding. This is particularly a problem for receptors that require complex post-translational modifications for proper folding and trafficking.50 In addition, the dependence on fluorescent ligands further limits the scope of this technique, as generation of fluorescent ligands that retain activity is challenging. Aside from increasing stability that will lead to increased homogeneity of the purified protein, stabilized receptors have the additional advantage that they exhibit higher affinity for compounds that preferentially bind to the conformation of the stabilized receptors. As a result, agonist-stabilized receptors tend to exhibit increased affinity for agonist molecules and reduced affinity for inverse agonist molecules. Conversely, inverse agoniststabilized receptors lose binding affinity to agonists while showing increased affinity for inverse agonists. This property has a number of consequences; first, conformationally stabilized receptors provide a more sensitive template for compound screening that offers the opportunity of identifying hits that may not have enough affinity to be identified on the wild-type receptor. This is particularly pertinent to fragment screening, as fragments generally have low affinity; however, they represent excellent starting points for lead generation. Therefore, the ability to readily generate such hits can be very enabling at the start of a discovery project. In addition, the increased ligand affinity for a particular conformation means that the hits are more likely to represent molecules that have the pharmacological activity matching the conformation of the stabilized receptor when tested on wild-type receptor. In other words, the stabilized receptor conformation can be used as a filter to infer information about ligand activity purely based on binding affinity. This concept of “reverse pharmacology” has been
Figure 1. Critical advances that have enabled GPCR structural studies. Protein engineering solutions include conformational stabilization and the application of fusion partners or antibodies. The primary aim of these techniques is to improve the properties of the receptors and make them more suitable for crystallization. Technical solutions include novel detergents and new crystallography techniques that provide better conditions for receptor purification and allow crystallization to take place in a more native environment, respectively.
which aim to facilitate protein purification, crystallization, or data collection. 3.1. Protein Engineering Solutions
Conformational thermal stabilization has become an established protein engineering method that has been applied to different GPCRs and has facilitated many successful structure resolution efforts.33−48 To this end, point mutations are introduced in the receptor that aim to achieve two goals: first, to increase the general stability of the receptor in detergent, and second, to lock the receptor in a single conformation. The overall result is the generation of a receptor that predominantly exists in a single conformation and can be readily purified in different detergents. In its most basic form, such a stabilized receptor is generated through screening for point mutations that increase the thermal stability of receptor in detergent, but importantly, this is done in complex with a ligand and the ligand binding is used as the readout of stability. The important consequence of using ligand binding to screen for stabilizing mutations is that any beneficial mutation identified will stabilize the receptor in the conformation that is imposed by the ligand, and therefore the ligand pharmacology will in effect be imprinted on the receptor. To identify stabilizing mutations, the receptor is submitted to scanning mutagenesis, followed by assessment of thermal stability of each mutation in detergent via ligand binding. Following identification of the single mutations, they are combined in a rational manner to identify the combination with maximum stability.33−38 Proper combination of mutations is an absolutely critical step that should ensure maximum stability is extracted from the mutation set and every mutation included should result in a significant increase in stability. This process of stabilization can be applied iteratively to continue to increase the stability.38 Initially, this methodology employed alanine scanning mutagenesis, primarily due to its enhanced helix-forming propensity. However, significant stabilization can also be achieved with different amino acid substitutions, and we frequently generate stabilized receptors through a combination of different scanning mutagenesis campaigns that target different parts of receptors with different amino acid substitutions. C
DOI: 10.1021/acs.chemrev.6b00119 Chem. Rev. XXXX, XXX, XXX−XXX
Chemical Reviews
Review
demonstrated very elegantly for the A2A receptor,51 where differential binding affinity to the inverse agonist-stabilized receptor compared to the wild-type receptor correlated with the functional inverse agonist activity of a range of antagonists. A further benefit of increased ligand affinity is that the structure of stabilized receptors can be solved in complex with ligands whose affinity on the wild-type receptor is too low to be used for this purpose. In a typical drug discovery project, early hits tend to have low affinity and the ability to get structural information about the binding of such compounds can have a significant positive impact on hit series development. It is worth mentioning that, even in cases where the ligand affinity has not increased significantly, the increased stability of stabilized receptors alone can allow structure resolution with low-affinity ligands. The main reason for this is that normally highaffinity ligands are desirable for purification and crystallization because, as a general rule, there is a direct relationship between ligand affinity and the extent to which the ligand imparts stabilization onto the receptor. In the case of stabilized receptors, the stability has been engineered in the construct and therefore is not necessarily required from ligand. This concept has been demonstrated by successful structure resolution of the stabilized A2A receptor in complex with the low-affinity antagonist caffeine.44 Moreover, stabilization can lead to significant improvements in the quality of the structures; for example, the stabilized form of the glucagon receptor resulted in a structure with significantly improved resolution, and importantly, as a result, the position of antagonist binding that was unassigned in the wild-type structure became fully resolved in the new structure.52,53 It is notable that in recent years conformational stabilization has been applied to non-GPCR membrane proteins and has led to structures of a number of these proteins being determined successfully.54−60 In addition to conformational stabilization, other protein engineering techniques such as utilization of stabilizing antibody fragments, conformation-inducing nanobodies, or fusion partners have all facilitated GPCR structure determination. Similarly to the conformational stabilization technique, these approaches also seek to reduce conformational flexibility and heterogeneity. One main source of problem in many GPCRs stems from the long and unstructured third intracellular loop (ICL3). Using an antibody that targets this part of the receptor or replacing it with a well-folded protein domain can reduce flexibility and additionally serve as an important hydrophilic surface to facilitate crystallization.61 A number of different fusion proteins such as T4 lysozyme, cytochrome b562, rubredoxin, and the catalytic domain of glycogen synthase have been used successfully.39,62−64 The common features that make these proteins suitable candidates for this purpose include small size (
