Main

G-protein-coupled receptors (GPCRs) are the largest superfamily of cell-surface receptors encoded by the human genome. They are pre-eminent environmental sensors and mediate signalling of a diverse range of extracellular activating ligands (agonists) including photons, ions, sensory stimuli, lipids, hormones, neurotransmitters, metabolites, peptides and proteins. GPCRs have evolved to transmit external environmental signals from one part of the protein to another. Therefore, they are intrinsically allosteric because GPCR signal transduction involves communication between spatially distinct yet conformationally linked binding sites1. In response to agonist binding to the cognate ‘orthosteric’ site, GPCRs undergo a conformational change to interact with intracellular transducers such as heterotrimeric G proteins and arrestins, ultimately resulting in integrated cellular responses. A small fraction of GPCRs already account for ~30% of US Food and Drug Administration (FDA)-approved medicines, highlighting their vital role in health and disease, high degree of tractability to drug discovery, and scope for further exploitation for leveraging understanding of GPCR biology1,2. This situation has developed rapidly within the last decade owing to breakthroughs enabled by the application of structural and biochemical studies of GPCRs, particularly advances in protein purification and engineering, lipid-based crystallography, X-ray diffraction, cryo-electron microscopy (cryo-EM) and computational biology3. All GPCRs share a common seven-transmembrane domain (7TMD) architecture and more variable extracellular domains (ECDs). There are now more than 200 reported GPCR structures (PDB entries) from more than 50 unique receptors, including full-length structures from members of classes A, B and F, and incomplete (7TMD or ECD) structures from class B, C and F GPCRs3 (Supplementary Table 1). Collectively, these structural studies have provided new insights into the allosteric nature of GPCRs, including activation, modulation by exogenous and endogenous molecules, and dimerization (Box 1). This Review highlights some of the key insights and challenges for the field arising from this new information for understanding allosteric mechanisms in GPCR biology.

Allosteric transitions govern GPCR activation

The simplest GPCR-activation mechanism is one of conformational selection comprising two states, with an ‘allosteric transition’ mediating the change between inactive and active states. Agonists preferentially stabilize the active state, inverse agonists stabilize the inactive state, whereas ‘neutral’ antagonists show similar affinities for both states but can block the actions of both agonists and inverse agonists if they interact via an overlapping site. Rhodopsin is a prototypical receptor that exemplifies this mechanism, and numerous structures have been reported that capture key states of rhodopsin (Supplementary Table 1). Rhodopsin is unusual among GPCRs, however, in that it is covalently linked to an inverse agonist, 11-cis-retinal, which ensures homogeneity of the inactive state. In response to light, 11-cis-retinal isomerizes to the agonist, all-trans-retinal, resulting in an allosteric transition that culminates in a 6–8 Å movement of transmembrane helix 6 (TM6)away from the transmembrane bundle, thereby creating an intracellular transducer-binding site. Other general structural changes that accompany GPCR activation include a contraction of the top of the receptor, an inward movement of TM5 and TM7, and a rotation of TM34. These changes are likely to be facilitated by rearrangements in a conserved network of ‘microswitches’ (Box 1, top) and water molecules5, common to most class A GPCRs, and are key components of allosteric communication between the orthosteric and intracellular transducer sites, which are typically located ~40 Å or more apart. A second unique feature of rhodopsin is that its active conformation is essentially fully attained even in the absence of co-bound G protein6. By contrast, most other known GPCR structures indicate that GPCRs require binding of a high efficacy agonist and an intracellular state-selective binding partner, such as a G protein, G-protein mimetic or arrestin to attain their fully active state. Notably, these structural observations are in agreement with pharmacological studies that were carried out nearly 40 years ago, which first identified a ternary complex between agonist, receptor and G protein as the molecular basis for drug efficacy at GPCRs7. Collectively, these studies highlight that signal transduction by GPCRs is an allosteric process that involves distal communication between conformationally linked sites.

GPCRs adopt multiple conformational states

High-resolution structures of GPCRs are snapshots of a dynamic system, and a two-state formalism is unlikely to accommodate their full conformational repertoire. In addition, currently solved structures represent the most stable conformations under specific experimental conditions; since GPCR stabilization is often achieved using a combination of detergents, fusion proteins, antibodies or mutations, this may limit the conformations explored by the receptor, as supported by biophysical studies8,9. These factors highlight an important caveat for interpreting structural studies and the need for complementary approaches. Accordingly, breakthroughs in NMR spectroscopy and molecular dynamics simulations have added to our understanding of the conformational plasticity associated with GPCR signalling, with studies indicating the existence of multiple inactive- and active-state conformations9,10,11. Figure 1 summarizes some of these findings in terms of a ‘continuum of conformations’.

Fig. 1: A continuum of GPCR conformational states.
figure 1

Representative examples of apo, ligand-, transducer- and/or nanobody (Nb)-bound GPCR structures, highlighting the potential diversity of conformational states. Inactive states (R′′, R′) exemplified by β2AR bound to the inverse agonist carazolol (Cz, red spheres; PDB: 2RH1) alone or together with the inverse agonist Nb60 (orange; PDB: 5JQH). For Rapo, the structure of rhodopsin in an active conformation that includes a molecule of detergent (not shown) in the orthosteric binding site is shown (PDB: 4J4Q). There are several structures of GPCRs in intermediate-active conformations (R′*), as exemplified by the A2AR bound to the agonist 5-N-ethyl-carboxamidoadenosine (NECA) (green spheres; PDB: 2YDV). Structures of multiple GPCRs have also been determined in active-state complexes with either nanobodies (R*) or transducers (R*T), as exemplified by the β2AR determined in complex with Nb80 (cyan, PDB: 3P0G) or the heterotrimeric Gs (blue, PDB: 3SN6).

Full size image

Most GPCR structures solved to date are of inactive states. Unsurprisingly, the largest differences between the inactive-state structures are found in the extracellular regions, from the orthosteric pocket extending to the extracellular loops (ECLs), and are reflective of diversity in ligand binding across the GPCR superfamily. Otherwise, inactive-state structures are broadly similar, particularly with respect to intracellular regions. Of note, 19F-NMR studies on the β2 adrenergic receptor (β2AR) and the adenosine A2A receptor (A2AR) describe two rapidly converting inactive conformations representing a stabilized and a broken ‘ionic lock’, respectively, between TM3 and TM610,11. This lock can be further stabilized by the co-binding of inactive-state-specific nanobodies or antibodies12,13, indicating that even so-called ‘inactive states’ can themselves display conformational heterogeneity. Although there are more than 50 agonist-bound structures, many of these more closely resemble an inactive rather than active conformation because they lack an additional binding partner, such as a G protein or arrestin, to allosterically stabilize the fully active conformation (R*). However, some of these agonist-bound states may represent intermediate conformations (R′*) between the transition from inactive (R′) to active (R*) conformations. General features of intermediate conformations include rearrangement of residues in the orthosteric site, microswitch regions, and possible partial outward movement of TM6. There are also multiple structures of agonist-bound GPCR–G-protein ternary complexes, and a structure of rhodopsin bound to arrestin (Supplementary Table 2). These structures unambiguously reveal the full extent of intracellular movement that is required for transducer binding.

By contrast, there are few structures of ligand-free GPCRs. Rhodopsin was the first to be reported, and multiple active-state opsin structures have also been determined, although a detergent molecule occupies the orthosteric site in several of these structures14. Several other receptors have had their structures determined in apparently ligand-free inactive conformations, however these finding remain controversial15,16,17. These studies highlight the challenges currently facing the field in obtaining true apo GPCR structures; however, such structures, if solved, could reveal novel GPCR conformations that facilitate ‘state-selective’ structure-based drug discovery that targets allosteric sites, rather than typical orthosteric-focused programs that (by default) require a ligand-bound structure. Nevertheless, the existing repertoire of active, inactive, intermediate and possibly apo GPCR structures supports the hypothesis that GPCRs adopt multiple conformational states that can be differentially stabilized by molecules binding to topographically disparate sites, either alone or in combination.

Exogenous allosteric modulators

One of the most exciting areas in modern drug discovery is the pursuit of molecules that bind to spatially distinct allosteric sites to modulate the actions of orthosteric ligands. Positive allosteric modulators (PAMs) enhance activity and negative allosteric modulators (NAMs) inhibit activity; whereas neutral allosteric ligands (NALs) have no net effect on the activity of orthosteric ligands, but competitively block the actions of PAMs or NAMs that bind to the same allosteric site18. Allosteric ligands may mediate agonism or inverse agonism in their own right, with or without PAM, NAM or NAL properties, suggesting novel pharmacological opportunities for drug development1 (Box 2). ‘Bitopic’ molecules have also been described, which bridge both orthosteric and allosteric sites on a single GPCR; such an interaction was structurally validated by a recent bitopic ligand-bound structure of the leukotriene B4 receptor19. Allosteric drug discovery targeting GPCRs has largely been driven by classic pharmacological or biochemical approaches, but the last five years have witnessed an explosion in structural understanding of the phenomenon (Fig. 2, Table 1). Although a general classification of GPCR allosteric sites can be complicated because such sites must, by necessity, be considered in relation to the widely divergent orthosteric sites between GPCRs, it is clear that druggable allosteric sites span the entire receptor surface, encompassing at least five key regions (Fig. 2, Box 1, middle):

Fig. 2: Diversity in the binding sites of synthetic allosteric modulators across GPCR classes.
figure 2

Chemical structures of allosteric modulators (coloured spheres) mapped onto representative family members of class A (M2 mAChR, PDB: 4MQT), class B (GCGR, PDB: 5EE7) and class C (mGlu5, PDB: 4OO9) GPCRs. Dashed lines indicate the boundary of the lipid bilayer. Further information is listed in Table 1.

Full size image
Table 1 Reported structures of GPCRs in complex with synthetic allosteric modulators
Full size table

1. The N-terminal ECD

This region is particularly relevant to the ECD of class B, C and F GPCRs. To our knowledge, there are currently no solved structures with synthetic modulators in this region, but there are structures bound to endogenous allosteric modulators (discussed below).

2. The extracellular vestibule

This region encompasses the ECLs and the classic (class A) biogenic amine orthosteric site in the 7TMD. A prototypical example is the M2 muscarinic acetylcholine receptor20 (mAChR; Box 1, middle; Fig. 2). This structure highlights a mechanism whereby activation promotes a closure of the extracellular vestibule, which is preferentially stabilized by PAMs, a contraction of the orthosteric site, and the opening of the intracellular region for transducer binding. Recent studies on the β2AR suggest that G-protein binding to the inside of the GPCR allosterically stabilizes a similar closed conformation at the extracellular end of the receptor21. Therefore, stabilization of a closed extracellular vestibule may be a common mechanism by which PAMs can increase agonist activity via this region. Conversely, molecular dynamics simulations of inactive-state M2 mAChRs suggest that NAMs binding in the vestibule have the opposite effect; these tend to be larger molecules that favour a more open vestibule22. A similar NAM-bound structure has been reported for proteinase-activated receptor 2 (PAR2), although there is a lack of structural and pharmacological comparisons with its agonist23 (Fig. 2, Table 1). In addition, the HIV drug maraviroc has been described as a NAM for the chemokine receptor CCR5 because it binds deeper into the 7TMD, away from the primary extracellular chemokine recognition site24. However, recent structure-function studies and a CCR5–chemokine antagonist structure suggest that this site may overlap with part of the orthosteric site, bringing into question claims of an allosteric mode of action of maraviroc25,26.

3. Within the 7TMD

The best example of allosteric modulators binding deeper within the 7TMD is from 7TMD structures of the class C metabotropic glutamate receptors mGlu127 and mGlu528 in complex with the NAMs FITM and mavoglurant, respectively, and further validated by newer structures29,30. These NAMs stabilize an inactive state through a cavity within the 7TMD (Fig. 2, Table 1), which may generically possess more than one allosteric site, as indicated by the structure of the class F Smoothened receptor (SMO). The NAM, SANT-1, adopts a similar pose to that seen in the mGlu1–FITM structure31. The deepest allosteric pocket within the 7TMD identified thus far is exemplified by CP-376395, a NAM of the class B corticotropin-releasing factor 1 receptor (CRF1R), which acts by sterically preventing formation of a kink in the conserved Pro6.47b-X-X-Gly6.50b (superscript refers to the Wootten numbering system for class B GPCRs32) motif in TM6, which is vital for the activation of class B GPCRs33.

4. Outside the 7TMD extending to the bilayer

Several extrahelical allosteric sites have been structurally confirmed (Table 1). A common pocket was identified outside TM5–7 near the intracellular surface for NAMs that target the glucagon receptor (GCGR)34 and glucagon-like peptide-1 receptor (GLP-1R)35,36 (Fig. 2). These NAMs prevent the outward movement of transmembrane helices that are required for activation. BPTU, a NAM of the purinergic receptor P2Y1, binds in a shallow pocket defined by residues in TM1–3 and ECL137. Similarly, AZ3451, a NAM of the PAR2 receptor, binds in a pocket created by residues in TM2–423. Despite binding to different GPCRs, both NAMs are likely to share a common mechanism that prevents rotation and translation of TM3 (which is usually required for activation of class A GPCRs). Multiple allosteric agonist co-structures have been reported for the free fatty acid receptor GPR40, including with the drugs TAK-87538 and MK-866639, which bind within the 7TMD but extend into the lipid bilayer through TM3–4, and a co-structure with both MK-8666 and the PAM AP8; AP8 binds in a second allosteric pocket between TM3 and TM5 above ICL239. Recently, a structure of the complement C5a receptor bound to the NAM NDT9513727 was determined, revealing an allosteric site that closely resembles the GPR40–AP8 PAM site (Fig. 2). NDT9513727 is likely to function as a NAM because it acts as a steric ‘wedge’, preventing the allosteric transition of TM5 that is required for activation40.

5. The intracellular surface

Structures of the chemokine receptors CCR2 and CCR9 in complex with their respective NAMs, CCR2-RA-[R]41 and vercirnon42 revealed a druggable intracellular allosteric site below the middle of the 7TMD that spatially overlaps with the G-protein-binding site. These NAMs probably stabilize an inactive conformation of TM6 while sterically impeding receptor interactions with a G protein (and possibly arrestin). Most recently, a structure of the β2AR in complex with a novel beta-blocker NAM, Cmpd-15-PA, also confirmed the existence of an intracellular allosteric site43. Of note, CCR2-RA-[R], vercirnon and Cmpd-15-PA all bind in very similar locations on the respective receptors, albeit with distinctly different chemical structures and interactions (Fig. 2, Table 1). The common binding site may therefore encompass a conserved intracellular allosteric site for small drug-like molecules. Although speculative, it may be possible to chemically modify such molecules to also promote PAM activity, although this may necessitate molecules that are better at opening this intracellular vestibule, which is the opposite mechanism to that of PAMs, which close the extracellular vestibule. An alternative mechanism has been proposed for the actions of the GLP-1R PAM BETP, which modifies a cysteine residue at the ICL3–TM6 junction to facilitate the activating outward movement of TM644.

Endogenous allosteric modulators

Given the diverse nature of GPCR binding pockets, it is possible that some of these may serve as sites for endogenous allosteric ligands. There is certainly pharmacological and biochemical evidence of such a role for various ions, amino acids, peptides, proteins, lipids and autoantibodies in both health and disease45, but a general mechanistic understanding of how these putative modulators alter GPCR function is largely lacking and beyond the scope of this Review. Instead, we highlight three specific examples of endogenous ligands that have benefited from structural breakthroughs. The first is in the general field of ‘accessory proteins’, which we broadly define as endogenous proteins that interact with specific GPCRs to regulate their binding, signalling or trafficking properties in a ligand-, pathway- or cell-specific manner. One of the best examples is a family of three single membrane-spanning proteins known as ‘receptor activity-modifying proteins’ (RAMPs), which associate with various GPCRs to modulate their physiological and pharmacological phenotype46. Although no structures of RAMPs in complex with full-length GPCRs have been solved, there are structures of RAMP1 or RAMP2 in complex with the ECD of the calcitonin receptor-like receptor, which unambiguously reveal a RAMP–GPCR interface that forms a drug binding site46. A second important example of endogenous allosteric modulators is that of ions. Indeed, sodium (Na+) was shown to negatively modulate agonist binding at opioid receptors more than 40 years ago47, and its allosteric effects have since been observed with other class A GPCRs45,48. A recent high-resolution structure of the A2AR49 revealed an intramembranous Na+ binding site and associated water cluster. Similar Na+ sites have been observed in other GPCRs (Supplementary Table 3). The functional role of the intramembranous Na+ site is unclear, although it is possible that Na+ may be acting as a cofactor, stabilizing the ligand-free state of some GPCRs and balancing the active-state–inactive-state allosteric transition50. Importantly, the drug amiloride and its derivatives bind to the Na+ site, albeit with low affinity, suggesting that this site may be targetable by drugs19,48.

Finally, there is a growing interest in the role of lipids in GPCR function, given that they are membrane proteins. To date, most structural studies have focused on the role of cholesterol as a putative allosteric modulator of GPCRs. Biochemically, cholesterol increases the thermal stability of rhodopsin and the β2AR51, and the binding affinities of some agonists at other GPCRs52. Crystal structures of the β2AR reveal three cholesterol binding sites; one of these, located in a cleft between TM2 and TM4, has been termed a ‘cholesterol consensus motif’ (CCM) owing to the high sequence prevalence (44%) across class A GPCRs51. Molecular dynamics simulations suggest that cholesterol allosterically modulates the β2AR by limiting its conformational flexibility53; this is supported by the necessity for cholesterol for crystallization of the receptor51,54. Indeed, cholesterol is present in over 40 PDB entries, occupying over 20 distinct sites (Fig. 3a, Supplementary Table 3). However, the only other structures with cholesterol bound near the region proposed to contain the CCM are those of the cannabinoid receptor CB155 and the purinergic receptor P2Y1256, both of which lack an actual CCM. Therefore, at this stage, there is a lack of crystallographic evidence for generalization of cholesterol binding to distinct sites across all GPCRs, other than to note that cholesterol is attracted to transmembrane grooves and hydrophobic patches52. By contrast, several reported structures show cholesterol binding to the extracellular cysteine-rich domain (CRD) of SMO57,58, which allosterically activates the 7TMD and mediates downstream signalling57,59 (Fig. 3b). It is therefore likely that the potential role and druggability of cholesterol-binding sites across various GPCRs will remain a fertile area of research in future. Irrespective of the type of endogenous substance, it should be noted that if endogenous allosteric modulators can be proven to mediate disease, then entirely new classes of medicines based on NALs may be pursued in such cases to specifically target the endogenous allosteric ligand site while sparing the actions of the cognate orthosteric agonist45. This is because NALs will act as competitive antagonists of endogenous disease-mediating substances that bind to the same allosteric site, while having no effect on normal signalling mediated by the orthosteric agonist.

Fig. 3: Cholesterol-binding sites across GPCRs.
figure 3

Cholesterol molecules have been observed bound to GPCRs in more than 40 PDB entries for 12 different receptors (Supplementary Table 2). a, Schematic of cholesterol-binding sites shown as ovals, coloured according to the GPCR subtype, with the red dot indicating the hydroxyl group of cholesterol. The arrow indicates a proposed cholesterol consensus motif (CCM). Positions of cholesterol are mapped onto their approximate location on the 7TMD cartoon. b, Cholesterol is an endogenous allosteric modulator of Smoothened (PDB: 5L7D), and binds to the cysteine-rich domain (CRD). Dashed lines indicate the boundary of the lipid bilayer.

Full size image

Biological agents as allosteric modulators

One of the major methodological breakthroughs for ensuring conformational homogeneity in GPCR structural biology has been the development of biological agents, such as antibodies, antibody fragments (Fabs) and nanobodies, to stabilize distinct active or inactive states in the presence of co-bound ligand60. Such biologic tools are intrinsically allosteric. For example, nanobody 80 (Nb80), which facilitated the solution of the first active-state β2AR structure (Supplementary Table 2) binds to the intracellular surface of the β2AR in a manner analogous to the C-terminal α-helix of Gs and acts as a PAM of agonist binding. Allosteric nanobodies have facilitated the crystallization of active conformations of other GPCRs, including the M2 mAChR, the μ-opiod receptor (μOR), viral chemokine receptor US28 and the κ-opioid receptor (Supplementary Table 2). Biological agents have also been used to stabilize heterotrimeric G proteins61 and β-arrestins62. This breakthrough enabled the first structure of a GPCR–G protein complex, and has since been extended to assist in solution of the first cryo-EM structures of GPCR–G protein complexes63,64. Although it may be argued that biological agents that recognize intracellular epitopes or proteins may not be directly translatable as therapeutics, their utility is not solely limited to structural studies60. For example, lipidated peptide fragments from GPCRs, termed ‘pepducins’, interact as intracellular allosteric modulators and are currently in clinical trials65. Moreover, newer biological agents have emerged as allosteric modulators, such as RNA aptamers that stabilize ligand-specific conformations66. Finally, the potential for extracellular allosteric biological agents should not be overlooked as a promising new area of research, as evidenced by the ongoing and successful clinical development of therapeutic antibodies67.

Biased agonism: cytosol-directed allostery

Key aspects of the allosteric nature of GPCRs are required for both a ligand and transducer to ensure stimulus transfer and reciprocity in communication between ligand and transducer sites. Although these mechanisms have been appreciated for decades7,68,69,70, diversity of cellular outcomes imparted by drugs (for example, full versus partial agonists) acting at the same GPCR was traditionally ascribed to post-receptor mechanisms; however, this concept is outdated. Following activation, GPCRs are phosphorylated by GPCR kinases (GRKs) in a GPCR-region-specific manner, leading to recruitment of arrestins to terminate G-protein signalling and initiate alternative waves of signalling71. Convergent data indicating that GPCRs adopt multiple ligand-stabilized active states, each linked to distinct physiological outcomes through promiscuity in their preferential choice of transducer to the relative exclusion of others, have led to the phenomenon termed ‘biased agonism’ or ‘functional selectivity71, and is transforming drug discovery. However, these phenomena are natural consequences of allostery. The mechanistic trigger underlying biased agonism is the tripartite communication between ligand, GPCR and transducer; this link has been explicitly referred to as ‘biased allostery’72 (Fig. 4a). It is therefore timely to consider the extent to which structural studies have addressed the molecular basis of GPCR-biased agonism.

Fig. 4: Conformational mechanisms of biased agonism.
figure 4

a, Biased agonism as a consequence of conformational selection by ligand occurring at the level of GPCR conformation. Two different agonists (red and green) preferentially select different active states (R*T1 and R*T2), which are characterized by different strengths of coupling (indicated by arrow size) to transducer proteins T1 and T2. b, Biased agonism as a consequence of conformational selection occurring at the level of the transducer protein. Two different agonists (red and blue) promote a similar active state (R*) with a given preference for transducer T1, but bias arises through allosteric communication between the receptor, ligand and transducer as the latter also undergoes conformational selection (T1*, T1**).

Full size image

Much research on GPCR–transducer interactions has focused on G proteins, although much remains to be learned with regards to the molecular determinants of G protein-mediated efficacy and selectivity73. By contrast, there is a relative lack of detailed structural information regarding direct interactions between GPCRs and GRKs, despite the central role that GRKs play in orchestrating interactions between GPCRs, G proteins and arrestin74; recent studies suggest the involvement of the ‘regulator of G-protein signalling homology domain’ of GRKs as a recognition site for GPCRs75,76. The interaction of arrestins with the GPCR is typically described by a two-step process77,78 involving binding of the C-tail of arrestin with the phosphorylated C terminus of the GPCR (the ‘hanging conformation’62,79,80), followed by rotation of the N- and C-terminal domains of arrestin to engage with the intracellular core of the GPCR (the ‘core conformation’)80,81. These events have significant implications for arrestin-mediated biased agonism. First, the strength of the initial interaction may be differentially encoded in a ligand-dependent manner via a ‘phosphorylation barcode’ on the GPCR81,82. Second, although binding of arrestin and G-protein binding to GPCRs have been considered to be mutually exclusive, the hanging conformation of arrestin may allow a GPCR to simultaneously engage with arrestin and a G protein, as suggested by a recently discovered macromolecular ‘megaplex’ with unique signalling properties83. Finally, the fully engaged arrestin–receptor core conformation, which is not compatible with G-protein binding, can nevertheless participate in β-arrestin-specific biased agonism.

A complete structural understanding of biased agonism requires the ability to decipher molecular interactions formed between biased ligands, GPCR and transducer. Although our understanding of how this occurs is in its infancy, some advances have been made. For example, structures of the 5-hydroxytryptamine receptors 5-HT1B and 5-HT2B bound to ergotamine (ERG) have been determined; the 5-HT1B complex is weakly biased towards β-arrestin, whereas the 5-HT2B complex is strongly biased towards β-arrestin84,85, relative to the complex with the endogenous agonist 5-hydroxytryptamine. The structure of 5-HT2B in complex with LSD, another strongly β-arrestin biased ligand, has also been determined86. The 5-HT2B structures share common features that are not observed in the 5-HT1B structure. Differential ligand contacts84 may explain the different GPCR conformations. Differential ligand-receptor contacts were also proposed to underlie the mechanism of the biased ligand carvedilol in complex with β1AR87.

Perhaps the most important conclusion of such structural studies to date is that they provide insights into the most proximal triggers of biased agonism. However, a challenge for the future is the need to incorporate the role of the transducer, which remains a non-trivial task. Nonetheless, recent intramolecular FRET and BRET studies focusing on arrestin loop movements88,89 have suggested that arrestin adopts multiple states upon receptor binding. Moreover, different agonists acting on the same GPCR produce distinct signatures, indicating that arrestin conformations can be influenced by GPCR ligands. Since activated arrestin conformations persist for some time after dissociation from receptors88, this may lead to differential engagement of downstream effectors, resulting in biased agonism, consistent with recent studies highlighting a role of kinetics in the phenomenon90,91. Perhaps more provocatively, similar arguments can be made for G proteins beyond their role as mediators of nucleotide exchange in a manner that is dictated solely by their affinity for a given active state. For example, salmon and human calcitonin promote distinct conformations of Gαs in complex with the calcitonin GPCR, each with differential GTP sensitivities, distinct receptor–G protein complex lifetimes and consequent differences in rates of cAMP formation92. Likewise, a comprehensive study of different β2AR ligands found that they can differentially affect movement of TM6, engagement by Gαs, and differential efficiencies of nucleotide exchange93. Furthermore, a recent cryo-EM structure of GLP-1R in complex with the biased peptide agonist, exendin-P5, revealed a different ECL3 conformation along with a 6° difference in the angle of G-protein engagement with the receptor94. Together, these findings suggest that to fully appreciate the molecular nature of GPCR efficacy and biased agonism, it is necessary to extend the concept of conformational selection by ligands from the level of ligand–GPCR to incorporate selection of transducer conformations through allosteric interactions (Fig. 4b).

GPCR dimerization

Although the formation of dimers and higher-order oligomers is well established as a universal paradigm mediating binding, signalling and allostery in other receptor superfamilies1, the role of oligomerization of most GPCRs is not as clearly defined, as it is known that many have the potential to mediate signalling as monomers95,96. However, GPCR dimerization remains an intensively studied phenomenon and has been proposed to modulate trafficking, ligand binding cooperativity, and signalling efficacy97. Of particular relevance to the current Review, the degree to which structural studies unambiguously support GPCR dimerization as a molecular mechanism underlying ligand recognition, cooperativity, or allosteric transitions tends to vary with the receptor subfamily. For instance, multiple parallel dimeric interfaces have been observed in class A GPCR structures97 (Box 1, bottom). The different interfaces identified thus far in many of these structures may reflect artefacts of the crystallization process, and caution should therefore be used when linking these observations to functional studies. For class B GPCRs, a functional role for dimerization, particularly involving TM4, has been proposed from mutagenesis, resonance energy transfer and pharmacological experiments98.

Structurally, it is perhaps most instructive to consider the class C GPCRs, such as the metabotropic glutamate receptors (mGluRs), type B γ-aminobutyric acid receptor (GABAB), Ca2+-sensing receptor (CaSR) and taste receptors (T1Rs), because these receptors unambiguously function as obligate dimers99. Although there are currently no full-length class C GPCR structures, various inactive and active conformations of isolated venus flytrap domains from mGluRs, GABAB, CaSR100,101 and T1Rs102 have been solved99. It is particularly noteworthy that these obligate dimeric GPCRs are all key environmental nutrient sensors that tightly regulate the activity of their cognate agonists. Thus, by extension, one may speculate that a key role for dimerization of GPCRs (even those known to function as monomers) is to constrain or bias the range of possible receptor activity in either a location-, tone- and/or context-dependent manner when required.

Outlook

Despite rapid recent advances in structural biology of GPCRs, there remains a great deal to be discovered to facilitate fundamental understanding of the role of allostery, and to fully realize the potential for allosteric and biased medicines as new therapeutic classes. For instance, a mechanistic understanding of the allosteric transition leading to differential G-protein and/or arrestin recruitment (and thus pathway-biased signalling or modulation) requires structures of multiple GPCR–transducer–effector complexes—ideally with minimal perturbations associated with the process of protein engineering or crystallization artefacts. This principle extends to the solution of full-length structures of larger GPCRs (for example, class C GPCRs) and associated complexes in order to better understand dimerization. The solution of multiple structures of the same receptor bound to different ligand classes and/or transducers remains largely elusive yet can prove enormously beneficial for structure-based design. Developments in alternative approaches, particularly NMR, computational biology and single-molecule fluorescence, as well as ongoing rigorous application of mutagenesis, chemical biology and analytical pharmacology, are likely to make substantial contributions to understanding the rich repertoire of GPCR allosteric effects. Structural validation of the enormous diversity of allosteric modulator-binding sites, as well as common motifs for such sites among disparate GPCRs, also raises two important questions for the field. First, what are the mechanistic drivers of the selectivity associated with many allosteric modulators if the sites are more prevalent or conserved than previously appreciated? Second, is there a role for endogenous modulators in health and disease, and how can this be exploited? Encouragingly, it is worth appreciating that many of the mechanisms associated with allostery of GPCRs are likely to be universal, and thus have either been better studied in some respects, or remain to be exploited in others, in different receptor superfamilies1,103. This suggests that there is very fertile ground for cross-disciplinary breakthroughs in understanding the ongoing role of allostery in the life sciences.