GPCR structure, function, drug discovery and crystallography: report from Academia-Industry International Conference (UK Royal Society) Chicheley Hall, 1–2 September 2014
- 3.5k Downloads
G-protein coupled receptors (GPCRs) are the targets of over half of all prescribed drugs today. The UniProt database has records for about 800 proteins classified as GPCRs, but drugs have only been developed against 50 of these. Thus, there is huge potential in terms of the number of targets for new therapies to be designed. Several breakthroughs in GPCRs biased pharmacology, structural biology, modelling and scoring have resulted in a resurgence of interest in GPCRs as drug targets. Therefore, an international conference, sponsored by the Royal Society, with world-renowned researchers from industry and academia was recently held to discuss recent progress and highlight key areas of future research needed to accelerate GPCR drug discovery. Several key points emerged. Firstly, structures for all three major classes of GPCRs have now been solved and there is increasing coverage across the GPCR phylogenetic tree. This is likely to be substantially enhanced with data from x-ray free electron sources as they move beyond proof of concept. Secondly, the concept of biased signalling or functional selectivity is likely to be prevalent in many GPCRs, and this presents exciting new opportunities for selectivity and the control of side effects, especially when combined with increasing data regarding allosteric modulation. Thirdly, there will almost certainly be some GPCRs that will remain difficult targets because they exhibit complex ligand dependencies and have many metastable states rendering them difficult to resolve by crystallographic methods. Subtle effects within the packing of the transmembrane helices are likely to mask and contribute to this aspect, which may play a role in species dependent behaviour. This is particularly important because it has ramifications for how we interpret pre-clinical data. In summary, collaborative efforts between industry and academia have delivered significant progress in terms of structure and understanding of GPCRs and will be essential for resolving problems associated with the more difficult targets in the future.
KeywordsGPCRs, G-protein coupled receptors β2AR, β2-adrenergic receptor GLP-1, Glucagon-like peptide-1 receptor CCK2R, cholecystokinin receptor-2 δ-OR, delta-opioid receptor CRF1, corticotropin releasing factor receptor 1 CXCR1, CXCR2, CCR4 and CCR5, chemokine receptors 5-HT2B and 5-HT2C, human 5-hydroxytryptamine receptors 2B and 2C, respectively H1, histamine receptor 1 hM3R, human muscarinic M3 receptor Dopamine D2 receptor α1B Adrenergic receptor T4L, T4-lysozyme BRIL, apocytochrome b562RIL XFELs, x-ray free electron lasers SDM, site-directed mutagenesis MD, molecular dynamic simulations 3D, three-dimensional 7TM, seven-transmembrane domain TM, trans-membrane helix ECL, extracellular loop HGMP, hierarchical GPCR modelling protocol GLAS, GPCR-likeness assessment score ProS, pairwise protein similarity method PDB, Protein Data Bank
GPCRs are a large family of integral membrane proteins that have enormous physiological and biomedical importance. Since GPCRs are involved in mediating cell signalling processes, they are implicated in many diseases and are the targets of numerous therapeutic drugs. This renders GPCRs one of the most important classes of current pharmacological targets. This is borne out by the fact that 60 % of all prescription drugs today target GPCRs (Schöneberg et al. 2004), developed for just 50 established GPCR targets out of the 800 known members of the gene family. The importance of GPCRs was recently highlighted with the Nobel Prize for Chemistry 2012 being awarded to two eminent GPCR researchers, Prof. Brian Kobilka (Stanford University, CA, USA) and Prof. Robert Lefkowitz (Duke University, NC, USA). There remains an ongoing need to better understand the interplay between structure and function of these receptors to advance our scientific knowledge and capacity to more effectively harness therapeutic capabilities. As a result, GPCR crystallography and modelling are rapidly expanding. However, there is still a sizeable gap between ongoing academic research and the needs of the pharmaceutical industry. A major reason for this is that academic and industrial scientists have too few productive opportunities to meet and interact, particularly to establish cross-discipline collaborations. The intimate atmosphere of such a small conference provided a unique opportunity to stimulate the generation of new networks and partnerships between academia and industry, and to promote current GPCR research and its applications to drug discovery. The invitation of scientists representing structural biology, protein engineering, pharmacology as well as computational and medicinal chemistry provided an interdisciplinary core to enable fruitful discussion and debate.
Many of the presentations stressed the importance of collaboration. For example, Prof. Schertler pointed out that during 18 years at the MRC Lab in Cambridge, he had a continuous string of collaborations with industry partners from small and large Pharma that created a very valuable network of complementary expertise. This collaborative network led to breakthroughs in the expression and purification of difficult membrane protein targets, and later this network was an important ingredient in the formation of the drug discovery spin out company Heptares Therapeutics. Several of Prof. Schertler’s earlier industry partners became the drivers of this GPCR-oriented company with their expertise on target selection, business models and intellectual property (IP) complemented perfectly the expertise of the MRC academic partners. The spin out company has in the meantime grown from three post docs to about 80 people, and it is able to tackle the most difficult GPCR targets with resources that would otherwise not have been accessible in Europe from any funding agency.
The basis of good industry academia collaboration is a clear agreement about the goals of the collaboration. Very often selecting a pre-competitive goal allows the academic partner to fully publish the results and gives the industry partner a significant edge in accessing emerging technologies. The example of Heptares Therapeutics illustrates that a spin out company can generate resources that would neither be available inside a company or from public funding agencies. If the spin out is able to reach milestones and refinance, then it can become a powerful tool to drive the application of new technology and lead to a change in the research culture in industry.
Most ventures between academia and industry are dependent on track record and trust. Individuals have to commit to a longer-term perspective, which is aimed at changing the scientific landscape. This scientific environment is essential for large companies that exist and are established to be able to recruit an excellent workforce and for small companies to pick up competitive innovation. In multinational companies, the academic and technology environment can start to dictate the location of departments inside a large organisation. From this follows that for a stable business development, the academic environment is as essential to the company as the in-house research activities.
One of the most significant advances in previous years has been the structural advances, both in terms of stabilising protein conformational states to make them more amenable to x-ray crystallography, but also in new technology such as x-ray free electron lasers which have the potential to accelerate structural biology not just for GPCRs but for many integral membrane protein drug targets. A significant portion of the meeting discussed these advances and also how they had been used in recent drug-discovery programs within both industry and academia. Another focus in recent times has been the concept of ‘ligand bias’. It has become clear that individual GPCRs can exist in multiple receptor conformations and can elicit numerous functional responses, both G protein- and non-G protein-mediated. This has led to the discovery that different ligands can stabilise distinct subsets of receptor conformations that can ‘traffic’ stimulus to diverse functional outputs with varying prominence, a concept referred to as biased signalling (also known as functional selectivity, stimulus bias or ligand-directed signalling). In principle, biased signalling can result in the development of more efficacious and safer drugs, but there are some unresolved questions regarding the best system in which to assess these aspects. The structural information alongside the realisation of biased signalling has also been explored with in silico modelling, and it was demonstrated at the meeting that this can give very useful insight into underlying properties of signalling control. The final section of the meeting focussed on how to best resolve problems in the future, including modelling processes at a higher level. In the following sections, we expand on these discussions in more detail.
Developments in GPCR crystallography
A significant advantage of the thermostabilised receptors is their use in drug discovery, which for the first time opens up the opportunities for structure-based drug design (Congreve et al. 2014). The thermostabilised receptors are readily purified and crystallised on a routine basis, which facilitates co-crystallisation with fragments and lead compounds. This was recently demonstrated for the β1AR where a fragment screen performed by surface plasmon resonance (SPR), followed by minimal hit optimisation, produced nM-affinity antagonists with novel scaffolds that were readily co-crystallised with the receptor (Christopher et al. 2013). The thermostabilisation of GPCRs is a central platform in Heptares Therapeutics, resulting in numerous crystal structures.
Another structural approach gaining momentum is the application of x-ray free-electron lasers (XFEL) to GPCRs. Structural studies of GPCRs, and other biomedically relevant membrane proteins and complexes, are hampered by challenges related to growing sufficiently large crystals capable of withstanding radiation damage and yielding high-resolution data at synchrotron sources. The recent introduction of a new generation of x-ray sources, x-ray free electron lasers (XFELs), producing ultra-bright pulses of coherent x-rays with an ultrashort duration, holds the promise to advance our understanding of structure and function of these challenging targets.
Results demonstrating the great utility of this approach were highlighted at the meeting. They included the structure of the human delta-opioid receptor bound to a bi-functional peptide ligand (Fenalti et al. 2014), the structure of a major GPCR signalling complex (unpublished data) and the first novel GPCR structure solved entirely by the LCP-SFX approach (unpublished data). In the future, this method could lead to the development of an efficient GPCR structure-based drug design pipeline by removing the major obstacles, such as the difficulties in preparation of large amounts of homogeneous and stable protein and growing sufficiently large crystals for a large number of different protein-ligand complexes. Another advantage of XFELs is the ability to freeze protein motion in time and obtain structures of unstable intermediates, advancing our knowledge about the signal transduction mechanism in GPCRs.
It was also noted that there are many signalling complexes in the cell that have component parts that have some flexibility as part of the structure necessary for function. Where these structures are difficult to stabilise, crystallisation is often problematic. Quite often crystallography can stumble because only very small crystals are attainable, with very weak diffraction, and this can be due, in part, to some partial disorder in a biologically relevant part of the molecule. The crystallographic solution up to now has been to delete or modify the flexible regions of the protein in order to create constructs that are more ordered so that crystals of significant size could be grown. With diffraction from microcrystals now possible, the structures of these more challenging biological assemblies are within reach.
Improvements in the understanding of GPCR function
In terms of function, the human β2-adrenergic receptor (β2AR) is probably the best-studied GPCR at the molecular, cellular and physiological level (Seifert 2013). The β2AR was of critical importance for the development of current models of receptor activation including biased signalling. Recent research has also shown for this system that ligand bias can depend on the system studied (native versus recombinant) (Seifert 2013) and also on the (patho)physiological state (healthy versus diseased), a feature sometimes also referred to as dynamic bias (Michel et al. 2014). Accordingly, the analysis of ligands at receptors such as the β2AR has become much more complex, requiring multidimensional approaches.
The β2AR constitutes an important drug target; agonists for this receptor are used for treatment of bronchial asthma and chronic-obstructive lung disease. However, safety and efficacy of β2AR agonists are not optimal. Most strikingly, the use of β2AR agonists alone in asthma is associated with increased mortality. Receptor desensitisation and activation of deleterious non-canonical signalling pathways, i.e. pathways different from the canonical Gs pathway, could contribute to this situation. Moreover, specific β2AR polymorphisms may be associated with decreased responsiveness to certain ligands, and the use of racemic β2AR ligands may be problematic. Specifically, the distomeric (not therapeutically active) ligands may contribute to drug toxicity (Seifert and Dove 2009).
In aggregate, these studies revealed that most reported ligands are biased towards canonical Gs signalling (Reinartz et al. 2015). This is particularly evident for ligands with large N-alkyl substituents, suggesting that these ligand domains, through constrained mobility of transmembrane helices, impede with coupling to Gi proteins and β-arrestin. In principle, this pharmacological pattern of fenoterol stereoisomers should be favourable for asthma treatment. However, the Seifert group did not identify any ligand with bias towards Gi or β-arrestin.
With regard to polymorphisms, 60 healthy volunteers were studied (see Fig. 4b). From these volunteers, blood was drawn and neutrophils were isolated. The pharmacological profile of the β2AR with respect to inhibition of superoxide radical formation was assessed. Moreover, the β2AR gene of the individuals was sequenced. Overall, substantial variability in the pharmacological profile of the β2AR in neutrophils was noted, but no association of the pharmacological β2AR profile with a specific polymorphism emerged. Thus, at the time being, there is no evidence for the notion that β2AR polymorphisms can be used to optimise asthma therapy.
To summarise this aspect, our increase in knowledge of the β2AR has resulted in a situation that renders future research more complicated. Most importantly, it is not anymore sufficient to determine a single parameter for a receptor such as Gs-mediated adenylyl cyclase activation. Rather, multiple parameters have to be determined including non-canonical Gi- and β-arrestin signalling. It is important to analyse the β2AR not only in recombinant but also in native systems. It will also be very informative to resolve crystal structures of the β2AR in complex with various ligands and coupling partners to understand the molecular basis of functional selectivity (Seifert and Dove 2009).
The GLP-1 receptor represents a good model system for studying class B receptor function (Koole et al. 2013a, b; Wootten et al. 2013a). GLP-1 is a key incretin peptide that promotes insulin secretion in response to nutrient ingestion, but also has a range of other actions including preservation of B-cell mass, reduction in gastric emptying and reduction in appetite that make it a desirable target for treatment of type II diabetes. Class B secretin-like receptors, like many other GPCRs, are pleiotropically coupled to a spectrum of both G-protein-dependent and -independent signalling pathways, and while cAMP production is the best characterised signalling endpoint for these receptors, physiological and therapeutic responses are the product of the integrated signalling response from all activated pathways. Moreover, the different contacts that are made between distinct ligands and their respective receptor can engender unique receptor conformations that give rise to distinct signalling profiles. This behaviour can be observed through differences in activation of second messengers, but also through changes to how receptors are desensitised and down-regulated. Biased signalling is further complicated when allosteric drugs are considered, as conformational preferences of the receptor when allosteric and orthosteric (endogenous ligand) sites are co-occupied may be different than when either site is individually occupied. Biased signalling is particularly relevant to receptor systems that have multiple endogenous ligands, and where exogenous mimetics are used clinically, as is often seen with class B receptors.
Clear (ligand-directed) bias for both peptides and small molecule agonists of the GLP-1 was demonstrated (Koole et al. 2013a, b), and this is consistent with earlier studies of pituitary adenylyl cyclase-activating polypeptide type 1 receptor (PACAPR) receptors and unclassified class B members like hCTRs, suggesting that this is likely to be a common feature of agonist activation of class B receptors. Transmembrane helical packing and conformational transition involved in receptor activation are assumed to involve key hydrogen bond networks formed around polar residues in the transmembrane helices.
The importance of conserved polar residues in class B receptors at the GLP-1 receptor has been recently studied (Koole et al. 2013a; Wootten et al. 2013b). This work has revealed networks of interaction that differentially contribute to global receptor activation and biased signalling. In particular, there appear to be two key interaction networks, one at the base of the receptor that may serve a similar role to the D[E]RY motif in class A receptors to maintain an inactive state; the second is located in the mid-region of the helical core and plays a critical role in pathway specific signalling, in a ligand-dependent manner. Mutation of residues within the central network has identified differences in how ligands propagate activation transition for individual signalling pathways and that distinct ligands utilise only subsets of the network for signal propagation providing initial insight into molecular mechanisms for biased signalling.
Several ligands stimulating phospholipase-C with efficacies reaching up to 50 % of that of CCK but which were inefficient to stimulate β-arrestin1/2 recruitment and receptor internalisation have been identified (Magnan et al. 2011). These ligands, which were initially considered as full antagonists at the CCK2R, are therefore more accurately described as antagonists on the β-arrestin-dependent signalling pathway of this receptor, but are partial agonists of the G-protein-dependent signalling pathway.
Extending these studies to ligands that are antagonists of G-protein-dependent pathways, it was discovered that the antagonist termed GV150013X (N-(+)-[1-(Adamant-1-ylmethyl)-2,4-dioxo-5-phenyl-2,3,4,5-tetrahydro-1H-1,5-benzodiazepin-3-yl]-N′-phenylurea) (see Fig. 5) could not inhibit β-arrestin1/2 recruitment and CCK-induced CCK2R internalisation (Magnan et al. 2013). Schild plot analysis of antagonist activity of GV150013X on CCK-induced phospholipase-C activation indicated that this molecule competitively inhibited the CCK effect. This information led the team to dock GV150013X in the orthosteric binding site of the modelled CCK2R, which had been previously validated by site-directed mutagenesis. The docking study predicted that the absence of effect of GV150013X on CCK2R recruiting β-arrestins was due to the presence of a bulky adamantane moiety in the ligand. GV150013X competitively inhibited CCK-induced G-protein-dependent pathway whereas it was inefficient on the β-arrestin-dependent pathway. These data suggested that the CCK2R could adopt two distinct conformations upon CCK activation and these two conformations are distinguishable at the binding site level. Fourmy’s lab therefore performed a site-directed mutagenesis study of the CCK2R orthosteric binding site with a focus on amino acids presumably in contact with the adamantane moiety of GV150013X. In parallel, the team synthesised an analogue of GV150013X in which the adamantane moiety was substituted by a methyl group: 1-(2,3,4,5-tetrahydro-1-methyl-2,4-dioxo-5-phenyl-1H-benzo[b][1,4]diazepin-3-yl)-3-phenylurea, termed GV-CH3. Pharmacological studies with CCK2R mutants and with GV-CH3 consistently demonstrated that the prediction of modelling and docking study was correct: the absence of effect of GV150013X on recruitment of β-arrestins to the CCK2R was due to a steric hindrance within the binding site which impedes binding of the antagonist (Magnan et al. 2013).
Overall, this area of research shows how pharmacological analysis of GPCR signalling combined with molecular modelling of GPCRs and chemistry of ligands can be used to analyse the origin of biased signalling. Such a strategy together with forthcoming determination of GPCR structures in complex with various signalling proteins (G proteins, β-arrestins) opens the possibility of rational drug design of biased ligands.
One of the biased-agonism pathways involves arrestins, and they themselves present an alternative line of investigation. Arrestins specifically bind active phosphorylated GPCRs, precluding further G protein activation and channelling the signalling to G-protein-independent pathways (Gurevich and Gurevich 2006). Based on the elucidation of an arrestin structure and key functional elements, special arrestins to channel cell signalling in a desired direction were constructed (Gurevich and Gurevich 2012). Enhanced phosphorylation-independent arrestin mutants were designed by disrupting key stabilising intra-molecular interactions that hold arrestins in a basal conformation (Gurevich and Gurevich 2012). Enhanced arrestin-1 was shown to compensate for the defects of rhodopsin phosphorylation in vivo, prolonging the survival of mutant rod photoreceptors, improving their light sensitivity and speeding up photoresponse recovery (Song et al. 2009). While the compensation with the first-generation enhanced mutant was only partial, new more powerful phosphorylation-independent forms of arrestin-1 hold promise for a better compensation (Vishnivetskiy et al. 2013). In rod photoreceptors, rhodopsin-specific arrestin-1 is the prevalent form, so it is clear that one needs to target this subtype to compensate for disease-causing defects of rhodopsin phosphorylation.
However, activating mutations in many GPCRs underlie various human disorders (Schöneberg et al. 2004). Since the two non-visual arrestins are fairly promiscuous, interacting with hundreds of GPCR subtypes, and most cells express 5–25 different GPCRs, only one of which is a mutant, to use a compensational approach, one needs receptor-specific non-visual arrestins. To this end, the elements of non-visual arrestins that determine their receptor preference were identified (Vishnivetskiy et al. 2011), and on the backbone of the most promiscuous non-visual subtype, arrestin-3, mutants with high (>50-fold) receptor specificity were created (Gimenez et al. 2012). This finding showed that targeting individual receptors with engineered non-visual arrestins is feasible. Arrestins interact with numerous partners, organising multi-protein complexes and recruiting them to particular sub-cellular compartments (Gurevich and Gurevich 2014a, b). This creates the potential of constructing signalling-biased arrestins that activate or inhibit certain pathways without affecting others. Recently, Gurevich and colleagues designed an arrestin-3 mutant that acts as a silent scaffold: it binds all kinases in the c-Jun N-terminal kinase (JNK) activation cascade, but does not promote its phosphorylation (Breitman et al. 2012). This mutant was shown to act in a dominant-negative fashion, suppressing JNK activation in the cell (Breitman et al. 2012). Multi-functionality of arrestins suggests that parts acting on particular pathways can be separated and used to modify cell signalling. Indeed, a small element of arrestin-3 that acts as a mini-scaffold, promoting JNK activation in vitro and in cells, has been identified (Zhan et al. 2014). Anti-proliferative activity of this element can be used for therapeutic purposes. Since arrestins play a role in numerous signalling pathways, targeted mutagenesis can yield arrestin-based molecular tools to tell the cell what to do in a language it cannot disobey (Gurevich and Gurevich 2014a, b).
Challenges and solutions for GPCR drug discovery
Despite many examples of successful marketed drugs that modulate the function of GPCRs, there remain a large number of potential therapeutically relevant GPCRs that are regarded as difficult to drug effectively. Methods used to analyse the GPCR’s ligand binding sites with a view to designing ligands were reviewed and included a discussion on the importance of water molecules for mediating interactions between ligands and receptor.
Antihistamines are one of the most well-studied drugs (Simons and Simons 2011). Indeed, it is now more than a century since the discovery of histamine as an important biogenic substance and more than 50 years since the production of the first antihistamine drugs (e.g. chlorpheniramine, diphenhydramine, hydroxyzine). H1R antagonists (antihistamines) are widely used in the treatment of a broad range of allergic diseases like rhinitis, conjunctivitis, urticaria and non-allergic disorders like pruritus and insomnia (Simons and Simons 2011). The first generation of H1R antagonists that had been introduced in the period of 1942–1980 demonstrated considerable side effects. Poor selectivity for the H1R and the ability to cross the blood–brain barrier (BBB) interfering with the histaminergic transmission (sedation) were among the most unwanted side effects. Recent studies have shown an increase in the number of allergic diseases, currently affecting more than 30 % of the world population (Qin 2007). Hence, there is an urgent need for more effective and safe anti-allergic drugs. The second-generation H1R antagonists, introduced in the early 1980s, had notable advantages such as being significantly more selective and non-sedating due to the lack of the ability to cross the BBB. These second-generation, non-sedating H1R antagonists have been widely used in the treatment of allergic conditions but still demonstrated some cardiotoxic side effects, e.g. induction of torsades-de-pointes arrhythmias. This has recently led to the development of third-generation H1R antagonists where both sedative and cardiovascular side effects were addressed (Oppenheimer and Casale 2002; Canonica and Blaiss 2011).
The crystal structure of the first-generation H1R antagonist doxepine bound to H1R was solved in 2011 (Shimamura et al. 2011). Doxepine has been associated with a large number of different side effects that can be rationalised by its lack of H1R selectivity and being a potent binder of H2R, some members of muscarinic, serotonic and α-adrenergic GPCR subfamilies and also of some protein kinases. Due to the lack of doxepine selectivity, the need for additional crystal structures of H1R bound to second and third generation of antihistamines that will rationalise the selectivity cannot be overestimated.
These new structures provide insights into potency and selectivity, the key challenges for the design of new generation of H1R antagonists and provide a significant contribution to potentially aid computational guided structure-based drug discovery of new antihistamine drugs targeting H2, H3 and H4 receptors where crystal structures are still absent.
Structural information in conjunction with careful modelling can provide greater insight into the location and functional relevance of druggable binding locations including both orthosteric and allosteric sites. Several recent structural publications have provided greater clarity on the binding modes and kinetics of existing drugs in both orthosteric sites, such as bronchodilator tiotropium binding to muscarinic M3 receptor (Tautermann et al. 2013), and allosteric binding sites, such as anti-viral maraviroc which acts as a negative allosteric modulator of chemokine receptor CCR5 (Kruse et al. 2012; Tan et al. 2013). Provision of knowledge of this type should begin to assist in design of more subtype selective ligands, especially when combined with leading edge computational techniques such as homology modelling (Storer et al. 2014) and molecular dynamic simulations.
Agonism of the 5-HT2C receptor has therapeutic implications in a number of important disease areas including CNS disorders and obesity. However, the design of ligands that are selective over agonism of the closely related 5-HT2B receptor, which has been linked with irreversible heart valvulopathy, has proved highly challenging (Storer et al. 2014). Furthermore, ligand-protein interaction features that lead to receptor agonism versus antagonism versus inverse agonism are not well understood despite numerous examples of subtle ligand structural changes driving pronounced differences in functional efficacy of GPCRs (Storer et al. 2014).
This is further complicated in instances where the same ligand exerts a different functional effect in different isoforms of the same receptor. This has been demonstrated by differential effects in the human isoform compared to the equivalent preclinical species receptors, hampering both in vivo efficacy and safety studies. An illustrative example is the histamine H4 receptor antagonist program where an inverse agonist of the human H4-receptor was a partial agonist of rat and an antagonist of dog H4 receptor, complicating the interpretation of preclinical in vivo studies (Mowbray et al. 2011). Therefore, a greater appreciation for the structural basis for these effects could ultimately assist in prediction and systematic avoidance of similar issues in future programs.
Biased signalling is of current interest to both academia and the pharmaceutical industry based on the hypothesis that it could provide improved therapeutic benefit whilst avoiding undesirable activities that unbiased signalling of certain receptors has historically caused (Tautermann 2014). This has provided motivation to revisit some GPCR targets that were previously either poorly drugged or deemed undruggable due to lack of therapeutic index over adverse events (Correll and McKittrick 2014). However, despite advances in biology and chemistry providing assay methods to measure bias and clear examples of biased ligands emerging, a deeper structural understanding of protein conformational changes and interactions that lead to differential receptor signalling is still in its infancy but clearly of keen interest to the medicinal chemistry community (Violin et al. 2014).
This complex picture of CCR4 resulted from extensive chemistry and pharmacological studies, and in hindsight it is clear why ‘traditional’ screening approaches such as binding or whole cell studies did not clearly identify the range of different ligands and sites. In one such screening approach, it was rationalised that the intracellular ligand binding site (helix 8) in CCR4 was likely modified owing the proximal chemical modification required in the commonly used FLIPR format.
Thus, structurally enabled approaches to ‘difficult’ GPRC targets should improve the tractability for chemistry by the direct identification of novel multiple ligand binding sites and subsequent chemical optimisation of small molecule ligands to drugs. This is an opportunity for drug discovery in a class of GPCR targets not easily accessible by previous methods.
Whilst many GPCR targets are chemically tractable resulting in multiple ligands and successful drugs, there are still many GPCR targets which are not amenable to ‘traditional’ approaches, such as focussed or high throughput screening, or where the natural ligands are unsuitable starting points for the design of oral drugs owing to their properties (e.g. large, lipophilic or peptidic).
In silico driven GPCR drug discovery
Today, GPCR modelling is widely used in the structure-based drug discovery process. The availability of structural information on the binding site of a targeted GPCR plays a key role in rationalisation, efficiency and cost-effectiveness of drug design. X-ray crystallography, a traditional source of structural information, is not currently feasible for every GPCR or GPCR-ligand complex. This situation significantly limits the ability of crystallography to impact drug discovery for GPCR targets in ‘real-time’ and hence there is an urgent need for other practical alternatives. GPCR modelling is widely used as a practical alternative in the absence of crystallographic data, but can also provide much more useful information. Today, it can address such key issues like GPCR flexibility and dynamics, ligand kinetics (k on/k off rates), prediction of water positions and their role in ligand binding, calculation of the free energy of binding (affinity) and prediction of the effects of mutations on ligand binding, etc. Some of these modelling approaches were reviewed in this meeting.
The role of flexibility and dynamics in the development of antipsychotic drugs and how this can be addressed by computational approaches was discussed. Recent pharmacological studies revealed that clinically effective antipsychotic agents act by binding to several bioamine receptors (Roth et al. 2004). In particular, the interaction with the serotonin (5-HT2A and 5-HT6) and dopamine (D2 and D3) receptors (group 1) induces cognition-enhancing effects, while the histamine (H1), 5-HT2C and 5-HT2B receptors (group 2) modulation causes unwanted side effects (Selvam et al. 2013). Due to the complex pharmacological profile of CNS disorders, the attempts to develop drugs based on the one-target-one-disease paradigm have been limited (Allen and Roth 2011). As a result, there remains an urgent need for innovative approaches to develop new effective multi-target agents that improve patients’ health while reducing care costs. Ideal candidates should selectively target disease-active members of the family (group 1) while not binding to members responsible for undesired side effects (group 2). This is a very challenging goal to achieve, as the residues forming the orthosteric binding pocket, i.e. the binding site of endogenous ligands, are conserved within the receptor family, thus causing the recognition of a drug by many members. A strategy to overcome this issue is to design allosteric drugs targeting a less conserved allosteric site, which modulates the orthosteric site, or to design bitopic drugs, which bind to both allosteric and orthosteric sites.
The bioamine receptors of groups 1 and 2 have been subjected to molecular dynamics simulations in a realistic environment. The results of simulations show substantial flexibility and its variability across the members of the receptor family. Using molecular probe mapping technique combined with the volume-based Tanimoto similarity measurements, similar and different geometrical and physicochemical properties were shown across the conformational space of the receptor family and the unique pharmacophoric features of disease-active receptors (group 1) were highlighted. The unique features are then linked with mutational and ligand structure-activity relationship data and tested in retrospective screening. The combination of techniques used gives an efficient method to identify unique properties of the disease-related proteins on the reduced diverse conformational space and represents a novel application of existing computational methods for the investigation of structural reasons for selective polypharmacology (Selvam et al. 2013). This protocol can be now exploited by industry for other protein families, involving in cancer and infectious diseases, which require a multi-target approach.
The performance of HGMP in GPCR drug discovery projects such as MCH-1R for obesity treatment (Heifetz et al. 2013a), the Orexin-1 and -2 receptors for insomnia (Heifetz et al. 2012, 2013b) and the 5-HT2C for the treatment of metabolic disorders (Tye et al. 2011; Storer et al. 2014) (see Fig. 12b) were highlighted. Additionally, the HGMP technology was used in the solving of the two H1R crystals structures bound to the second- and third-generation antihistamines: Cetirizine (Gillard et al. 2002) and Fexofenadine (Sharma et al. 2014), respectively, as described by Dr. Moraes (above).
Another area that has received increasing attention over recent years has been the role of water networks and their elucidation by MD simulations (Tautermann et al. 2015). With the availability of more very high-resolution GPCR structures [e.g. the δ-OR (Fenalti et al. 2014) or the A2AR (Liu et al. 2012) with resolutions of 1.8 Å], it has become clear that GPCRs often exhibit conserved water networks, which extend from the extracellular side to the intracellular surface. This conserved solvent network has been implied to be crucial for signalling (Nygaard et al. 2009) and a highly solvated conserved allosteric sodium binding site close to the conserved NPxxY motifs of class A GPCRs is postulated to be involved in β-arrestin signalling (Fenalti et al. 2014; Tautermann 2014). Beyond the functional effects, the consideration of water molecules in GPCR ligand design has been shown to be crucial (Bortolato et al. 2013) because several crystal structures show water-mediated ligand-protein interactions (Congreve et al. 2014). The displacement of binding site water upon ligand binding is energetically disadvantageous. Therefore, a ligand always has to gain more free energy from binding to the receptor than the removal of water actually costs. Several methods are available which enable the very crude and quick estimation of the energy penalty for water displacement, and most efficiently they are used for growing a bound ligand (Bortolato et al. 2013). They can help medicinal chemists to decide if a certain sub-pocket of the receptor can be explored by hydrophobic moieties or if a displaced water has to be substituted by an entity which has to mimic the hydrogen bond network. Obviously, these quick methods are not thorough in a sense that the estimates of the free energy are physically sound, but often they are sufficiently good for a go/no-go decision. In order to get a (formally) correct estimate for the change in the free energy of binding upon ligand modification, methods like thermodynamic integration have to be applied (Christ and Fox 2014). These more accurate methods all rely on long MD simulations of the receptor, and therefore they can also capture reorganisations in the solvent network and induced effects in the binding pocket. One step further beyond assessing the free energy of ligand binding is the estimation of binding kinetics of ligands. Water networks also play a crucial role here as well, as the breaking of ligand-receptor hydrogen bonds usually ends up with solvated hydrogen bond acceptor and donor functional groups. As the energy barrier of the reaction determines the rate constant, the direct interaction of water molecules with the ligand-protein hydrogen bonds is decisive for the dissociation rate. If the hydrogen bond is solvent accessible, the breaking of the bond does not require high energy because the re-hydration occurs simultaneously. If the hydrogen bond is buried, a breaking leads to a high energy penalty because acceptor and donor do not find immediate new interaction partners and are in unfavourable solvation states (Schmidtke et al. 2011). Currently, the in silico prediction of association/dissociation rates is only possible for very small fragments rather than drug-like molecules.
An essential part of computational modelling is to re-evaluate predictions often in light of new structural data. Researchers at Novartis Horsham had performed a large amount of work exploring the binding sites of allosteric inhibitors of the chemokine receptor CXCR2, and several interesting things emerged after re-analysing this work in the light of crystallographic data (Salchow et al. 2010). It had been determined at Novartis Horsham that antagonists were acting at an allosteric binding site, and the patent literature around the time of the work suggested that the CXC chemokines could have a binding site close to the intracellular face of the receptor. One of the key residues in this binding site was proposed to be K320 as the CXCR2 antagonists commonly had an acidic functionality of some description within them and this residue was asparagine in the related CXCR1 receptor over which many of the CXCR2 antagonists had selectivity.
The NMR structures for the related CXCR1 receptor (as reproduced in Fig. 14b) consistently show that the critical N320 residue is not close to the other influential residues, and it would have been hard to rationalise the mutagenesis from this template. The CCR5 crystal structure shows distortion in the TM7/helix 8 region and the critical K320 residue was mutated to a GLU, but overall the helical alignment compared to Rhodopsin was very similar and so would not have provided any benefit compared to the Rhodopsin template used originally. The β2AR-Gs-protein structure was a marvellous achievement (Rasmussen et al. 2011), but for the purposes of acting as a template for the intracellular binding area of the antagonists, it is likely to be too different to CXCR2 to be useful. The most relevant structure could have been the CXCR4, but here again the TM7/helix 8 region is distorted in the crystal and a CXCR2 homology model was created by using a chimeric template of this receptor and the TM7/helix 8 region from an earlier β2AR structure (Rasmussen et al. 2011). Interestingly, this model had a related CXCR2 antagonist bound in the TM domain and the major learning point from the current GPCR symposium was just how much influence changes within the TM domain affected the intracellular coupling and vice versa. A conclusion drawn from the symposium with respect to the work completed at NIBR Horsham would be that more mutagenesis information from across the whole of the CXCR2 protein would have been needed in order to fully understand the binding area of these allosteric antagonists.
Finally, network-based approaches, evolutionary algorithms and predictive modelling, all areas that will become more prevalent in the future, were discussed. The advent of the ‘omics’ age has brought with it huge quantities of data around diseases, targets, compounds and their effects. Networks of interactions and disease ‘interactomes’ can be built with the ultimate goal to understand disease networks and how they are influenced by the changes in small molecules and their properties.
It was also shown how GPCR targets and their interacting partners could be identified from Genome Wide Association Studies (GWAS) using network analysis followed by the analysis of compound gene expression data to complement target disease gene expression as a strategy for network-based drug design. Multiple algorithms have been developed that in combination with omics data and GPCR structure-based design make a powerful arsenal for today’s drug designer.
All participants agreed that GPCR research and drug discovery can benefit greatly from the collaboration between academia and industry. The effectiveness of such collaborations for GPCR research and drug discovery was widely described and exemplified during this conference: the non-profit GPCR Consortium (http://gpcrconsortium.org) described by Vadim Cherezov; the collaboration between MRC Laboratory and Heptares mentioned by Gebhard Schertler and Chris Tate; the consortium between Evotec Ltd, Oxford University UK and the Membrane Protein Laboratory at Diamond Light Source, UK described by Alexander Heifetz; and many other successful networks like the Adhesion-GPCR Consortium (AGC) or GLISTEN (Gutierrez-de-Teran 2014) (GPCR-Ligand Interactions, Structures, and Transmembrane Signalling: a European Research Network).
There was broad agreement with Schertler’s and Ceska’s comments that most ventures between academia and industry are dependent on track record and trust and that individuals have to commit to a longer-term perspective which is aimed at changing the scientific landscape. There was also agreement that meetings between academia and industry such as this conference are very useful to learn about developments within each other’s areas of expertise and to share the challenges whilst forging new links and networks. For future meetings of this type, it was proposed to also include a broader mix of younger-generation scientists early in their career in order to maximise benefit to a broader community.
We are grateful to the Royal Society for sponsoring this event and in particular to Mr. Ryan Ahmed, Mrs. Jennifer Howes and Judy Wilkins from DeVere Venues for their great help in organising this conference. We are also grateful to the Royal Society for Industry award granted to A.H. and in particular to scheme manager Mr. Ben Cunningham for all his help and support. AH also thanks the BBSRC for support (BB/L026287/1). We are also grateful to our Evotec colleagues Dr. Gabriele Hansen and Dr. Florence Eustache for all their assistance, guidance and for the design of the conference material.
The authors declare that they have no conflict of interest. A.H. research is sponsored by Royal Society UK, grant number IF100104. The conference was also sponsored by Royal Society UK, grant number IS130026. This article does not contain any studies with animals performed by any of the authors. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards.
- Bennett KA, Dore AS, Christopher JA, Weiss DR, Marshall FH (2014) Structures of mGluRs shed light on the challenges of drug development of allosteric modulators. Curr Opin Pharmacol 20c:1–7Google Scholar
- Christopher JA, Brown J, Dore AS, Errey JC, Koglin M, Marshall FH, Myszka DG, Rich RL, Tate CG, Tehan B, Warne T, Congreve M (2013) Biophysical fragment screening of the beta1-adrenergic receptor: identification of high affinity arylpiperazine leads using structure-based drug design. J Med Chem 56(9):3446–3455PubMedCentralPubMedCrossRefGoogle Scholar
- Heifetz A, Barker O, Verquin G, Wimmer N, Meutermans W, Pal S, Law RJ, Whittaker M (2013b) Fighting obesity with a sugar-based library: discovery of novel MCH-1R antagonists by a new computational-VAST approach for exploration of GPCR binding sites. J Chem Inf Model 53(5):1084–1099PubMedCrossRefGoogle Scholar
- Heifetz A, Morris GB, Biggin PC, Barker O, Fryatt T, Bentley J, Hallett D, Manikowski D, Pal S, Reifegerste R, Slack M, Law R (2012) Study of human Orexin-1 and -2G-protein-coupled receptors with novel and published antagonists by modeling, molecular dynamics simulations, and site-directed mutagenesis. Biochemistry 51(15):3178–3197PubMedCrossRefGoogle Scholar
- Keiser MJ, Setola V, Irwin JJ, Laggner C, Abbas AI, Hufeisen SJ, Jensen NH, Kuijer MB, Matos RC, Tran TB, Whaley R, Glennon RA, Hert J, Thomas KL, Edwards DD, Shoichet BK, Roth BL (2009) Predicting new molecular targets for known drugs. Nature 462(7270):175–181PubMedCentralPubMedCrossRefGoogle Scholar
- Liu W, Wacker D, Gati C, Han GW, James D, Wang D, Nelson G, Weierstall U, Katritch V, Barty A, Zatsepin NA, Li D, Messerschmidt M, Boutet S, Williams GJ, Koglin JE, Seibert MM, Wang C, Shah ST, Basu S, Fromme R, Kupitz C, Rendek KN, Grotjohann I, Fromme P, Kirian RA, Beyerlein KR, White TA, Chapman HN, Caffrey M, Spence JC, Stevens RC, Cherezov V (2013) Serial femtosecond crystallography of G protein-coupled receptors. Science 342(6165):1521–1524PubMedCentralPubMedCrossRefGoogle Scholar
- Magnan R, Escrieut C, Gigoux V, De K, Clerc P, Niu F, Azema J, Masri B, Cordomi A, Baltas M, Tikhonova IG, Fourmy D (2013) Distinct CCK-2 receptor conformations associated with beta-arrestin-2 recruitment or phospholipase-C activation revealed by a biased antagonist. J Am Chem Soc 135(7):2560–2573PubMedCrossRefGoogle Scholar
- Miller-Gallacher JL, Nehme R, Warne T, Edwards PC, Schertler GF, Leslie AG, Tate CG (2014) The 2.1 A resolution structure of cyanopindolol-bound beta1-adrenoceptor identifies an intramembrane Na+ ion that stabilises the ligand-free receptor. PLoS One 9(3):e92727PubMedCentralPubMedCrossRefGoogle Scholar
- Mowbray CE, Bell AS, Clarke NP, Collins M, Jones RM, Lane CA, Liu WL, Newman SD, Paradowski M, Schenck EJ, Selby MD, Swain NA, Williams DH (2011) Challenges of drug discovery in novel target space. The discovery and evaluation of PF-3893787: a novel histamine H4 receptor antagonist. Bioorg Med Chem Lett 21(21):6596–6602PubMedCrossRefGoogle Scholar
- Procopiou PA, Barrett JW, Barton NP, Begg M, Clapham D, Copley RC, Ford AJ, Graves RH, Hall DA, Hancock AP, Hill AP, Hobbs H, Hodgson ST, Jumeaux C, Lacroix YM, Miah AH, Morriss KM, Needham D, Sheriff EB, Slack RJ, Smith CE, Sollis SL, Staton H (2013) Synthesis and structure-activity relationships of indazole arylsulfonamides as allosteric CC-chemokine receptor 4 (CCR4) antagonists. J Med Chem 56(5):1946–1960PubMedCrossRefGoogle Scholar
- Procopiou PA, Browning C, Buckley JM, Clark KL, Fechner L, Gore PM, Hancock AP, Hodgson ST, Holmes DS, Kranz M, Looker BE, Morriss KM, Parton DL, Russell LJ, Slack RJ, Sollis SL, Vile S, Watts CJ (2011) The discovery of phthalazinone-based human H1 and H3 single-ligand antagonists suitable for intranasal administration for the treatment of allergic rhinitis. J Med Chem 54(7):2183–2195PubMedCrossRefGoogle Scholar
- Rasmussen SG, DeVree BT, Zou Y, Kruse AC, Chung KY, Kobilka TS, Thian FS, Chae PS, Pardon E, Calinski D, Mathiesen JM, Shah ST, Lyons JA, Caffrey M, Gellman SH, Steyaert J, Skiniotis G, Weis WI, Sunahara RK, Kobilka BK (2011) Crystal structure of the beta2 adrenergic receptor-Gs protein complex. Nature 477(7366):549–555PubMedCentralPubMedCrossRefGoogle Scholar
- Reinartz MT, Kalble S, Littmann T, Ozawa T, Dove S, Kaever V, Wainer IW, Seifert R (2015). Structure-bias relationships for fenoterol stereoisomers in six molecular and cellular assays at the beta-adrenoceptor. Naunyn Schmiedebergs Arch Pharmacol 388(1):51–65Google Scholar
- Simons FE, Simons KJ (2011) Histamine and H1-antihistamines: celebrating a century of progress. J Allergy Clin Immunol 128(6): 1139–1150.e1134Google Scholar
- Slack RJ, Russell LJ, Barton NP, Weston C, Nalesso G, Thompson SA, Allen M, Chen YH, Barnes A, Hodgson ST, Hall DA (2013) Antagonism of human CC-chemokine receptor 4 can be achieved through three distinct binding sites on the receptor. Pharmacol Res Perspect 1(2):e00019PubMedCentralPubMedCrossRefGoogle Scholar
- Storer RI, Brennan PE, Brown AD, Bungay PJ, Conlon KM, Corbett MS, DePianta RP, Fish PV, Heifetz A, Ho DK, Jessiman AS, McMurray G, de Oliveira CA, Roberts LR, Root JA, Shanmugasundaram V, Shapiro MJ, Skerten M, Westbrook D, Wheeler S, Whitlock GA, Wright J (2014) Multiparameter optimization in CNS drug discovery: design of pyrimido[4,5-d]azepines as potent 5-hydroxytryptamine 2C (5-HT(2)C) receptor agonists with exquisite functional selectivity over 5-HT(2)A and 5-HT(2)B receptors. J Med Chem 57(12):5258–5269PubMedCrossRefGoogle Scholar
- Tye H, Mueller SG, Prestle J, Scheuerer S, Schindler M, Nosse B, Prevost N, Brown CJ, Heifetz A, Moeller C, Pedret-Dunn A, Whittaker M (2011) Novel 6,7,8,9-tetrahydro-5H-1,4,7,10a-tetraaza-cyclohepta[f]indene analogues as potent and selective 5-HT(2C) agonists for the treatment of metabolic disorders. Bioorg Med Chem Lett 21(1):34–37PubMedCrossRefGoogle Scholar
- Weierstall U, James D, Wang C, White TA, Wang D, Liu W, Spence JC, Bruce Doak R, Nelson G, Fromme P, Fromme R, Grotjohann I, Kupitz C, Zatsepin NA, Liu H, Basu S, Wacker D, Han GW, Katritch V, Boutet S, Messerschmidt M, Williams GJ, Koglin JE, Marvin Seibert M, Klinker M, Gati C, Shoeman RL, Barty A, Chapman HN, Kirian RA, Beyerlein KR, Stevens RC, Li D, Shah ST, Howe N, Caffrey M, Cherezov V (2014) Lipidic cubic phase injector facilitates membrane protein serial femtosecond crystallography. Nat Commun 5:3309PubMedCentralPubMedCrossRefGoogle Scholar
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.