Oligomerization of G protein-coupled receptors: Still doubted?

Abstract

Experimental data accumulated during the last ten years strongly support the existence and functional significance of oligomerization of G protein-coupled receptors (GPCRs). In this essay, we discuss the unique biochemical properties of GPCR oligomers in the frame of ‘allosterism’ and how these properties can be used to identify GPCR oligomers in artificial systems and in native tissues. We also address how the significant advances in biophysical, computational and crystallization techniques have provided significant structural insights about the mechanisms behind allosterism in GPCRs, giving distinct clues about the mechanisms of allosteric interactions in GPCR homomers and heteromers. Finally, we elaborate on the emerging picture of the role of GPCR oligomers as components of pre-coupled macromolecular complexes that include different G proteins and effectors, such as adenylyl cyclase. Allosteric properties of GPCR oligomers can therefore extend to ligand interactions through the different components of the macromolecular complexes.

1. Introduction

Ten years have passed since a group of experts established by consensus a series of definitions and recommendation related to receptor oligomerization (Ferré et al., 2009). ‘Receptor heteromer’ was defined as “a macromolecular complex composed of at least two (functional) receptor units (protomers) with biochemical properties that are demonstrably different from those of its individual components”. This definition would be distinguished from a ‘heteromeric receptor’, defined as a dimeric or oligomeric receptor for which the minimal functional unit is composed of two or more different subunits that are not functional on their own. ‘Receptor homomer’ and ‘homomeric receptor’ would then be respectively defined the same as receptor heteromer and heteromeric receptor but composed of two or more identical protomers or identical non-functional subunits. Additional definitions were proposed, such as ‘allosteric interaction in the receptor heteromer’ and ‘biochemical fingerprint’ (see below), as part of recommendations for the identification of receptors oligomers in native tissues (Ferré et al., 2009).

These recommendations were meant to facilitate the research in the field, particularly, of G protein-coupled receptor (GPCR) oligomerization, homo and heteromerization, which was held with skepticism by several leaders in the GPCR research community. More precisely, although the evidence of GPCR oligomerization in artificial systems was generally accepted, their presence and functional significance in native tissues was clearly challenged (see open discussions by Bouvier and Hébert, 2014, and Lambert and Javitch, 2014). This would not be the case for family C GPCRs (which include glutamate and GABA metabotropic receptors), which are generally accepted to be obligatory homodimers, stabilized by disulfide bonds in their large extracellular domains or coiled-coil α helices in their C termini (Kniazeff et al., 2011). During the last ten years, with the development of new biophysical and computational methods, a series of studies led to the unequivocal demonstration of oligomerization in artificial systems of not only family C but also family A and family B GPCRs, as well as to the identification of their unique biochemical properties, which would allow the demonstration of GPCR oligomers in native tissues.

In this essay, we discuss the unique biochemical properties of GPCR oligomers in the frame of ‘allosterism’, which we believe it provides sufficient evidence for the demonstration of GPCR homo and heteromers in native tissues. We will also address how the significant advances in biophysical, computational and crystallization techniques, with the boost of crystal structures of GPCRs in different conformational states, including those bound to G proteins and to orthosteric and allosteric ligands, have provided significant structural insights about the mechanisms behind allosterism in GPCR. These insights provided distinct clues about the mechanisms of allosteric interactions in GPCR homomers and heteromers. Furthermore, we discuss how recent experiments challenge some classical pharmacological concepts of interactions between GPCR ligands believed to occur downstream in the signaling pathway, also known as interactions “via intracellular loops” or interactions at the “second-messenger level” (Zoli et al., 1993; Agnati et al., 1995; Ferré et al., 1997). These studies indicate that some of these interactions should be reevaluated as allosteric, mediated by macromolecular complexes of GPCR heteromers, G proteins and effectors.

2. Classical GPCR allosterism

Allosterism has been defined as the process by which the interaction of a chemical or protein at one location on a protein or macromolecular complex (the allosteric site) influences the binding or function of the same or another chemical or protein at a topographically distinct site (Smith and Milligan, 2010). The prefix ‘allo’ has a Greek origin meaning ‘other’ and the term allosterism was introduced by Jacques Monod and Francois Jacob (1961), to account for an ‘allosteric inhibition’. This implied a situation where an enzyme inhibitor (the allosteric ligand) was not a steric analog of the substrate. Their assumptions were largely based on results obtained by their graduate student Jean-Pierre Changeux with the enzyme L-threonine deaminase, its substrate L-threonine and the regulatory inhibitor L-isoleucine (Changeux, 1961). Allosterism would provide a new mechanism of regulatory proteins different to covalent modulation, such as phosphorylation. The introduction of these non-covalent modulations would initiate a revolution in Biochemistry, as Monod would anticipate when describing allosterism as “the second secret of life”, after the genetic code (Monod, 1971).

Interestingly enough, the concept of allosterism developed initially in the frame of what it seemed a common oligomeric structure of regulatory enzymes, with the Monod-Wyman-Changeux (MWC) model (Monod et al, 1965). Structurally, the MWC model implied that allosteric proteins are oligomers resulting from the assembly of ‘protomers’ associated in such a way that the molecule possesses at least one axis of symmetry. Functionally, it represented the first elaborated model of the two-state mechanism of pre-existing conformational states, a consequence of a selective, rather than an instructive, effect of ligands (Changeux, 2012). Oligomeric allosteric proteins would have access to at least two reversible states, which would be dependent on the conformational constraints imposed by the protomers. The affinity of both ligands, the substrate and the allosteric ligand, are altered with the transition of the protein from one to the other state and the molecular symmetry is always conserved. Therefore, the presence of the allosteric ligand is associated with a modification of the apparent affinity of the protein for the substrate, and conversely. As mentioned by Fenton (2008), allosterism can be more strictly defined as a comparison of how one ligand binds in the absence versus the presence of the second ligand.

A logical generalization would follow from allosterism of regulatory enzymes to allosterism of membrane receptors, which would be equivalent to regulatory proteins which activated state would not imply a modification of the substrate (enzymes) but a signal transduction (Changeux and Edelstein, 2005). An obvious parallelism could be established between substrate and orthosteric ligand (which binds to the same receptor site as the endogenous transmitter). The allosteric ligand, by binding to a non-orthosteric site, can then modify either of the two independent properties of the orthosteric ligand, affinity (the avidity to bind to the receptor) and intrinsic efficacy (the power of the bound ligand to induce a physiological response). As pointed out by Changeux (2012), “the notion that ligands selectively stabilize the state to which they preferentially bind and thereby mediate signal transduction via selection of conformational states arises directly from the formal description of the MWC model”.

3. Allosterism in GPCR monomers

The seminal studies by Nobel laureate Brian Kobilka and coworkers and other research groups (for recent review, see ref (Weis and Kobilka, 2018) provided, first, an example of a well-characterized allosteric mechanism within family A GPCRs, more specifically, of a process by which the interaction of a ligand at one location of the GPCR (the orthosteric site, but in this case, equivalent to the allosteric site in the definition of allosterism; see above) influences the binding of a protein at a topographically distinct site. With the more recent application of nuclear magnetic resonance (NMR) and double electron-electron resonance spectroscopy and single-molecule fluorescence resonance energy transfer (FRET) analysis, their studies also provided a strong support for the existence of pre-existing conformational states of GPCRs and the selection and stabilization of several different states (conformational selection), from more inactive to more active, from inverse agonists to neutral antagonists, partial agonists and full agonists, as well as their dependence on G protein and nucleotide binding (Liu et al., 2012; Ye et al., 2016; Manglik et al., 2015; Gregorio et al., 2017).

Thus, GPCRs are dynamic proteins that permit rapid small-scale structural fluctuations and pass through an energy landscape (Deupi and Kobilka, 2010) to adopt a number of conformations ranging from inactive to active (Figure 1A). The transition probability among conformations depends on their energy difference and the energy barrier between them. Ligand binding to the orthosteric site of a monomer (Figure 1A) changes the shape of the energy landscape relative to the unliganded landscape (black line), in such a manner that inverse agonists and antagonists (in red) stabilize inactive conformations (red line), agonists (in green) stabilize intermediate conformations, and the final formation of the agonist-receptor-active G protein complex stabilizes active conformations (green line) (adapted from Nygaard et al., 2013). The inactive and active conformations, at the molecular level, have been elucidated by X-ray crystallography. An agonist binding at the extracellular side, apparently, triggers small local structural changes near the binding site that are translated into larger-scale helix movements at the intracellular side (Rasmussen et al., 2011). We then have a significant knowledge about the basic molecular mechanisms of this GPCR allosterism, by which an agonist transfers a “signal” from the orthosteric ligand-binding pocket to a G protein binding to a specific portion of the GPCR intracellular domain, the Gα subunit-binding “intracellular cavity”.

Figure 1.

Dynamic properties of GPCR monomers and heteromers. (A) Energy landscape of a GPCR monomer in the unliganded form (black line), after the action of an inverse agonists/antagonists (in red) or an agonist (in green). (B) Energy landscape of a GPCR monomer in the presence of a negative allosteric modulator (NAM, in brown) and a positive allosteric modulator (PAM, in yellow). (C) Energy landscape of a GPCR heteromer in the unliganded form (black line), in the presence of an antagonist (in red), an agonist (in green), one antagonist and one agonist (in blue), and two agonists (in light green).

In a nutshell, the binding of the orthosteric agonist to the active conformation of the receptor is dependent on a very subtle rearrangement of some key residues of GPCRs (more specifically family A GPCRs), localized in three of the seven transmembrane domains (TMs) of the GPCR, TM 3, TM 5, and TM 6, in the orthosteric ligand-binding pocket, which leads to a mechanically amplified larger movement of particularly TM 6 in its intracellular site (Weis and Kobilka, 2018). This movement of TM 6 is simply mediated by a lever-arm effect, due to the presence of a kink (caused by an interruption in the intrahelical hydrogen bonds at conserved proline residues) close to its extracellular end. In this way, a small approximation of the extracellular side of TM 6 to TM 3 (which is a straight TM that forms the core of the GPCR) leads to a large separation of the intracellular side of TM 6 from TM 3 and also TM 5 and TM 7, creating an intracellular cavity for the helix 5 of the Ras-GTPase domain of the Gα subunit. Beautifully, with this TM 6 movement, some residues from the intracellular site of TM 3 and TM 5 that interact with TM 6 in the inactive state of the receptor are freed to form direct interactions with the C-terminal α5 helix of the G protein in the active state (Weis and Kobilka, 2018).

Considering GPCRs as monomeric entities, allosteric modulators are ligands that bind to sites that are topographically distinct from the orthosteric site and also alter the energy landscape of the receptor (Christopoulos, 2014; Figure 1B). Negative allosteric modulators (NAMs) (in brown) decrease the action (potency and/or efficacy) of the orthosteric agonist (in green) by destabilizing intermediate and active conformations. In contrast, positive allosteric modulators (PAMs) (in yellow) increase the action (potency and/or efficacy) of the orthosteric agonist by stabilizing intermediate and active conformations. Ago-PAMs are allosteric modulators than in addition to increase the potency and/or efficacy of orthosteric agonists, can induce receptor activation by themselves. There are numerous allosteric binding sites as elucidated by the increasing number of published crystal structures of classes A, B, C, and F of GPCRs in complex with small-molecule allosteric modulators (for a recent review, see Lu and Zhang, 2019). In addition to the prevailing site at the TMs near the extracellular side (classes A, C, and F), other binding sites are also observed at the TMs near the intracellular side (classes A and B), or at the extra-helical domains facing the membrane bilayer (classes A and B), among others.

We are also starting to understand the mechanistic aspects of these allosteric modulators. For instance, the crystal structure of the muscarinic M₂ receptor (M₂R) simultaneously bound to an orthosteric agonist and a PAM shows that the allosteric modulator binds at the extracellular vestibule inducing a slight contraction of the binding pocket, enhancing the potency without modifying the efficacy of the orthosteric agonist (Kruse et al., 2013). In contrast, a crystal structure of the β₂-adrenergic receptor (β₂R) also in complex with an orthosteric agonist and a PAM shows that the allosteric modulator binds on the receptor’s inner surface in a pocket created by intracellular loop 2 and TMs 3 and 4, stabilizing the loop in an α-helical conformation required to engage the G protein (Liu et al., 2019). Thus, the mechanisms by which PAMs elicit their function is, in contrast to orthosteric agonists, highly diverse. Going back to the definition of allosterism, these represent mechanistic explanations for a process by which the interaction of a ligand at one location of the GPCR (the allosteric site) influences the binding of another ligand at a topographically distinct site (the orthosteric site).

4. Allosterism in GPCR homomers

From radioligand binding experiments, it has been long recognized that agonists show ‘binding heterogeneity’, a dispersion of affinities for GPCRs. This phenomenon has been reproduced in experiments from membrane extracts from multiple native tissues and mammalian transfected cells as well as from artificial systems, including GPCR reconstituted in detergent micelles, liposomes (phospholipid vesicles with or without different proportions of cholesterol and integral membrane proteins) or nanodisks (phospholipid bilayer preparations held together by membrane scaffold proteins) (for review, see Ferré et al., 2014). The typical illustration is a competition experiment between a moderate concentration of a radio-labelled antagonist and increasing concentrations of the agonist, which results in a non-steep or even biphasic curve, apparently indicating the existence of two populations of GPCRs, with high and low affinities for the agonist (Figure 2A; solid curve). In the presence of a G protein, guanylyl nucleotides promote an ‘apparent’ inter-conversion from the high to the low affinity sites, with steepening and shifting to the right of the competition curve (Figure 2A; dotted curve), which led to the general assumption that the agonist binding heterogeneity was just a G protein-dependent phenomenon, counteracted by guanylyl nucleotides.

Figure 2.

Theoretical [³H]antagonist/agonist competitive inhibition curves from radioligand binding experiments in membrane preparations from a native tissue (A), or from preparations of GPCRs reconstituted as monomers in nanodisks (B) or oligomers in liposomes (C), in excess (relative to the GPCR) or absence of G proteins. In A, the dotted curve represents the effect of guanylyl nucleotides. In B, the red curve represents the effect of an intermediate concentration of G proteins. In C, the orange curve represents the effect of guanylyl nucleotides in the presence of an excess of G proteins (see text and Redka et al., 2013, 2014)

In line with the classical view of GPCR signal transduction, a still common view of GPCR signaling holds that agonists promote the formation of a transient complex between the monomeric receptor and the G protein, with conformational changes in both molecules that lead to an increased affinity of the receptor for the agonist and for the heterotrimeric Gαβγ protein, which leads to G protein activation. The deep-rooted ‘ternary complex’ model of ligand-GPCR binding (De Lean et al., 1980) postulated that the receptor-G protein complex accounts for the high affinity site for the agonist, while the G protein-uncoupled receptor accounts for the low affinity site. Guanylyl nucleotides would promote uncoupling of the receptor-G protein complex, converting all receptors into low affinity sites. According to the model, a non-steep or biphasic antagonist-agonist competitive-inhibition curve implied the existence of two populations of receptors, coupled and non-coupled to the G protein. Implicitly, this interpretation assumed there is a limited pool of G proteins, an assumption difficult to reconcile with the fact that the expression of G proteins in native cell systems outnumbers that of GPCRs (Neubig, 1994).

However, a large amount of experimental data points to the existence of GPCR oligomerization and indicate that, in fact, oligomerization is a main property of GPCRs. This was initially suggested from a study by Limbird, Meyts and Robert Lefkowitz (who shared the Nobel prize with Brian Kobilka), that demonstrated negative cooperativity of the binding of a radio-labelled β₂R agonist in membrane preparations from frog erythrocytes (Limbird et al., 1975). The existence of such a negative cooperativity was confirmed by kinetic experiments, which showed a significant increase in the rate of dissociation of the labelled β₂R agonist by the non-labelled β₂R, indicating the presence of a non-competitive negative interaction between labelled and non-labelled β₂R agonist (Limbird et al., 1975). Thus, ligands that would compete for the same site on a monomeric GPCR should not influence one another’s dissociation kinetics. In contrast, allosteric modulations between two simultaneously bound and interacting sites, either within a receptor monomer or across a receptor homodimer or oligomer, should alter ligand dissociation (May et al., 2007). Many more experiments would follow which would be completely in line with the existence of allosteric modulations between different molecules of the same orthosteric agonist in oligomeric and most commonly homodimeric GPCRs, a ‘homodimer-cooperativity model’ (reviewed in Ferré et al. 2014).

According to the homodimer-cooperativity model, binding heterogeneity would be G protein-independent, at least under conditions of receptors in excess of G proteins. There are no high and low affinity states of a GPCR monomer for the agonist, but an affinity of the agonist for the binding to the first protomer in the homodimer, which allosterically modulates the affinity of the same ligand for the second protomer. In fact, the ‘monomer-ternary complex model’ could never explain some radioligand binding data, such as an apparent modulation of antagonist binding by guanine nucleotides (Burgisser et al., 1982; De Lean et al., 1982; Klotz et al., 1990), biphasic antagonist/antagonist competition curves (see below and Orrú et al., 2011), bell-shaped agonist/antagonist competitive curves (Casadó-Anguera et al., 2019) or a positive cooperativity (Albizu et al., 2006). Unfortunately, the initial data indicating cooperativity of the same orthosteric ligand and, therefore, indicating GPCR oligomerization, was overshadowed by the discovery of the powerful allosteric modulation by G proteins and guanylyl nucleotides and by the success of the pragmatic monomer-ternary complex model.

In apparent support for the monomer-ternary complex model, biphasic and guanylyl nucleotide-sensitive antagonist/agonist competitive inhibition curves were obtained with GPCR monomers (of β₂R, μ opioid receptors [MOR], rhodopsin and M₂R) either in solution or reconstituted in nanodisks, where they could also bind and activate G proteins (Bayburt et al. 2007; Ernst et al., 2007; Whorton et al., 2008; Kuszak et al., 2009; Redka et al., 2013, 2014). In these studies, the proportion of receptor with high affinity for agonists increased from zero to 100% by increasing the G protein pool, indicating that the low-affinity sites were receptors in excess of G proteins, that the agonist binding heterogeneity was totally G protein-dependent, implying a G protein-mediated allosteric modulation of an orthosteric ligand (Figure 2B). According to our definition of allosterism, this would be an example of the process by which the interaction of a protein (the α subunit of a G protein, which would act as the allosteric modulator), at one location of another protein (GPCR) influences the binding of a chemical (the orthosteric ligand) at a topographically distinct site. In these preparations the effect of G proteins on the binding of the orthosteric agonist was reversed by GTP and GTP analogues. Therefore, GPCR monomers in the presence of G proteins and guanylyl nucleotides are indistinguishable from GPCR monomers in the absence of G proteins (Figure 2B; blue curve). Guanylyl nucleotides, and GTP under physiological conditions, act as additional negative allosteric modulators, as potent inhibitors of the ability of helix 5 of the Ras-GTPase domain of the Gα subunit to interact with the intracellular cavity of the GPCR active conformation.

The results of isolated GPCR monomers were initially interpreted as a demonstration of their prevalence in native tissues, as main functional units, and as a demonstration of the main G protein dependence of agonist binding heterogeneity. But recent seminal studies by James Wells’s group demonstrated that, in the presence of enough G protein molecules, without receptors in excess of G proteins, GPCR oligomers show a significant G protein-independent binding heterogeneity, cooperativity (Redka et al., 2013, 2014) (Figure 2C; green curve). These studies compared the effect of G proteins and guanylyl nucleotides in radioligand binding experiments in purified M₂R, M₂R monomers reconstituted in nanodisks and M₂R oligomers (mostly tetramers as the predominant species) reconstituted in liposomes (Figure 2C). Importantly, the experiments were also performed in cardiac muscle membrane preparations, which revealed that M₂R in native tissues have the same biochemical characteristics than when reconstituted in liposomes as oligomers (Redka et al., 2013, 2014), demonstrating that a predominant proportion of M₂R in cardiac muscle is forming oligomers. We are then closing the loop and going back to the initial evidence for negative cooperativity of β₂R described by Limbird et al. (1975) to explain GPCR agonist binding heterogeneity.

The observation of GPCR agonist binding cooperativity therefore reveals that at least two GPCR protomers form part of a minimal GPCR functional unit. An important additional concept that then arises from the field of GPCR oligomerization is that the pentameric structure constituted by one GPCR homodimer and one heterotrimeric G protein provides a main functional unit and oligomeric entities can be viewed as multiples of dimers (Ferré et al., 2014). Thus, the relatively large size of the heterotrimeric G protein would make it unlikely to sustain a simultaneous binding of more than one G protein with a GPCR dimer. Additional evidence has been provided from a large number of studies by application of resonance energy transfer (RET), bimolecular complementation of fluorescent of bioluminescent molecules and other biophysical techniques in living cells, such as fluorescence correlation and cross-correlation spectroscopy (FCS and FCCS) and analysis of single fluorescence-labeled receptor molecules by total internal reflection fluorescence microscopy (TIRF) (Calebiro et al., 2013; Herrick-Davis et al., 2013; Teichmann et al., 2014). These methods differ in sensitivity and the population of receptors that they analyze. RET (which includes fluorescence and bioluminescence resonance energy transfer [FRET and BRET], fluorescence lifetime imaging [FLIM] and fluorescence recovery after photobleaching [FRAP]) and bimolecular complementation are proximity-based assays that monitor the distance between the fluorescent and bioluminescent probes, whereas TIRF, FCS and FCCS have near single molecule sensitivity.

Some of these studies have provided evidence for a receptor-dependent differential degree of oligomerization (oligomers versus monomers) and of the average stoichiometry of the oligomer species (number of monomers per oligomer), as recently supported by experiments in liposomes also using fluorescence-microscopy-based techniques (Walsh et al., 2018) and computational methods that characterize the energetics of oligomer interfaces, based on large-scale coarse-grained molecular dynamics simulations (MDS) (Periole et al., 2012; Meral et al., 2018). Those studies indicate a low degree of oligomerization for B₁R and MOR, intermediate for β₂R and predominant for cannabinoid CB₁ receptor (CB₁R) and rhodopsin. Nevertheless, apart from the fact that MDS only consider TM interfaces (see below for the role of intracellular domain interactions in GPCR heteromerization), the results of these studies performed in artificial systems most probably represent an underestimation of what occurs in native tissues. As also demonstrated in artificial systems, just the lipidic composition and membrane curvature can influence the degree of GPCR oligomerization (Botleho et al., 2006; Prasanna et al., 2014; Guixa-Gonzalez et al., 2016; Walsh et al., 2018). In addition, as discussed below, there is increasing evidence for an elaborated formation of protein complexes that include GPCR oligomers (homo and heteromers), signaling-related proteins and effectors. The emerging specific quaternary structure of these macromolecular complexes points to the existence of multiple simultaneous points of contact between their elements, calling for a significant stabilization of GPCR oligomers (Navarro et al., 2018).

5. Allosterism in GPCR heteromers

The first evidence for possible allosteric interactions in GPCR heteromers was labelled as “intramembrane receptor-receptor interactions” by Luigi Agnati and Kell Fuxe, based on observations from early radio-ligand binding experiments in the 1980’s, where a selective ligand for a GPCR changed the binding characteristics of another selective ligand for a different GPCR (reviewed in Agnati et al., 2003). Although, initially, allosteric mechanisms were implicitly assumed to be involved, it was not until recently than an operational definition of ‘allosteric interaction in the receptor heteromer’ was advanced: “an intermolecular interaction by which binding of a ligand to one of the receptor unit in the receptor heteromer changes the binding properties of another receptor unit” (Ferré et al., 2009). More recently, it has been explicitly formulated that GPCR heteromerization implies the possibility of multiple allosteric interactions between orthosteric ligands of different GPCRs (Ferré et al., 2014) (Figure 3A).

Figure 3.

Allosteric interactions in GPCR heteromers: In A, allosteric interaction between ligands, where a ligand of one of the protomers changes the affinity or efficacy of a ligand for the other protomer. In B, ligand-independent allosteric interaction, where the binding of one protomer changes the properties of a ligand for the other protomer. In C, allosteric interaction through the effector. Homomeric interactions are not included and only one protomer of each different GPCR is shown. AC, adenylyl cyclase.

However, if there are still doubts about the existence of GPCR homomers, even more skepticism exists around the concept of GPCR heteromerization. Nevertheless, the number of putative functional and pharmacologically significant GPCR heteromers keeps increasing. It must however be acknowledged that many candidates do not follow the consensus criteria recently established for their identification in native tissues (Ferré et al., 2009, 2014; Gomes et al., 2016). The same as for GPCR homomers, a long list of biophysical techniques has been used to study GPCR heteromerization and, the same as for GPCR homomers, the most important challenge is their demonstration in native tissues (Ferré et al., 2009, 2014; Gomes et al., 2016). Although some attempts have been made, application of those or other techniques that imply a “direct” visualization of GPCR oligomers in native tissues remains problematic (reviewed in Ferré et al., 2014). An indirect but nevertheless straightforward demonstration depends on the identification of a ‘biochemical fingerprint of the GPCR heteromer’, which is currently defined as “a biochemical characteristic of a receptor heteromer, which can be used for its identification in a native tissue” (Ferré et al., 2009). To ascertain a biochemical fingerprint of the receptor heteromer, the putative biochemical property should be disrupted with molecular or chemical tools that significantly destabilize the quaternary structure of the heteromer (such as introducing mutations of key determinant residues at the GPCR heteromer interface or using competing peptides that harbor the same amino acid sequence as that of the heteromer interface).

Allosteric interactions in GPCR heteromers seem to constitute a universal property of GPCR heteromers which can then be commonly used as a biochemical fingerprint to identify the heteromer in native tissues. As an example, in this essay, we will address the allosteric interactions in the heteromer constituted by adenosine A_2A receptors (A_2AR) and dopamine D₂ receptors (D₂R), which is one of the few examples of a generally accepted GPCR heteromer (Gomes et al., 2016). Since the first report in 1991 (Ferré et al., 1991), A_2AR ligands have been shown to modulate the binding properties of D₂R ligands in membrane preparations from transfected cells and native tissues. Although initially thought to be restricted to a negative crosstalk between agonists, where an A_2AR agonist reduced the affinity of dopamine and D2R agonists, more recent findings extended the interactions to any orthosteric ligand. This included the ability of selective A_2AR antagonists or caffeine, a non-selective A₁R-A_2AR antagonist, to reduce the affinity of D₂R ligands (Bonaventura et al., 2015; Ferré et al., 2016). That these interactions represent allosteric modulations in the A_2AR-D₂R heteromer was demonstrated by the ability of specific mutations or synthetic interfering peptides that selectively disrupt or destabilize the quaternary structure of the heteromer to selectively disrupt the interactions between A_2AR and D₂R ligands (for review, see Ferré et al., 2016). Importantly, the selective effect of interfering peptides could also be demonstrated in native tissue, in sheep brain (striatum) membrane preparations, demonstrating the existence of the A_2AR-D₂R heteromer in the brain and the dependence on the right quaternary structure of the heteromer to disclose allosteric interactions between orthosteric A_2AR and D₂R ligands (Bonaventura et al., 2015; Ferré et al., 2016).

The fact that both A_2AR agonists and antagonists produced the same effect on D₂R agonist binding and also signaling (Bonaventura et al., 2015) represented a challenge to our initial assumption of allosteric interactions in the A_2AR-D₂R heteromer being involved in the opposite and counteractive effects of A_2AR agonists and antagonists observed in behavioral and functional studies. Those effects would include the psychomotor depression induced by A_2AR agonists and the psychomotor activation induced by A_2AR antagonists, including caffeine, as well as their competitive counteraction upon co-administration, which would be expected from their competition for the A_2AR orthosteric site. The conundrum could be solved by the realization of the tetrameric structure of the A_2AR-D₂R heteromer, composed of A_2AR and D₂R homodimers, and by the existence of simultaneous heteromeric and homomeric allosteric interactions in the heterotetramer (Bonaventura et al., 2015). We could assume that, in the A_2AR-D₂R heterotetramer, occupancy of the A_2AR homodimer with either an orthosteric agonist or an orthosteric antagonist is associated with a conformational state that determines the allosteric modulation of D₂R ligands, while simultaneous occupancy of the A_2AR homodimer by an orthosteric agonist and an orthosteric antagonist would not be possible with this conformational state. What could be demonstrated is that an A_2AR agonist, but not an antagonist, allosterically modifies the dissociation kinetics of a radioactive A_2AR antagonist ligand in striatal membrane preparations (Bonaventura et al., 2015). The functional significance of these simultaneous homomeric and heteromeric interactions in the brain was demonstrated with electrophysiological and locomotor activity experiments. We could show that, under specific experimental conditions, A_2AR antagonists behaved as A_2AR agonists, decreasing D₂R function, and these effects were counteracted upon co-administration of both A_2AR agonists and antagonists. For instance, at high doses, caffeine and A_2AR antagonists produce locomotor depression, which can be counteracted by locomotor depressant doses of A_2AR agonists (Bonaventura et al., 2015). Since under physiological conditions there is a tone of adenosine, this could in fact be the main mechanism by which caffeine and A_2AR antagonists produce psychomotor stimulant activation, by counteracting the functional effects that depend on D₂R signaling by the A_2AR-D₂R heterotetramer.

We are also beginning to explore the mechanisms behind the allosteric interactions through the heteromeric interface in GPCR heteromers. For instance, we have shown that in the heteromer of CB₁R and serotonin 5-HT_2A receptors (5HT_2AR) a CB₁R antagonist blocks the signalling triggered by a 5HT_2AR agonist and that, conversely, a 5HT_2AR antagonist blocks the signalling triggered by a CB₁R agonist (Viñals et al., 2015). This cross-antagonism, the ability of an orthosteric antagonist of one of the protomers in a receptor heteromer to antagonize the signaling of an agonist of the other molecularly different protomer, has been described frequently, representing a common allosteric property of GPCR heteromers (Ferré et al., 2014). Thus, one of the protomers allosterically modulates the energy landscape of the interacting receptor (Figure 1C). We proposed, and demonstrated using interfering peptides, that this allosteric modulation is determined by the ability of CB₁R and 5HT_2AR to establish a symmetrical molecular interaction via TM 5 and TM 6 (Viñals et al., 2015). Relative to the unliganded landscape (black line), in which both TM 5 and TM 6 of both protomers are in a dynamic equilibrium of conformations, antagonist binding to either protomer stabilizes the closed, inactive, conformations of TM 5 and TM 6 (see above), promoting the interaction with TM 5 and TM 6 of the other protomer via the formation of a stable four-helix bundle, as described for the dimeric crystal of the MOR (Manglik et al., 2012) (red line). Importantly, this energy minimum of the antagonist-bound receptor is more stable in heteromers than in monomers, because of the additional formation of the four-helix bundle. These low free-energy states of the antagonist-bound heteromer (inactive states) impede the possibility of an agonist binding to the other protomer, which would require the energy minima of the active states (blue line), leading to cross antagonism (Figure 1C). The same as for the A_2AR-D₂R heteromer, a negative crosstalk could also be observed in the CB₁R-5HT_2AR heteromer, implying that an agonist binding to one of the protomers leads to a decrease in the signaling promoted by an agonist binding to the other molecularly different protomer. In this case, agonist binding to one protomer stabilizes the active conformations of TM 5 and TM 6 (open cavity) and the binding of the G-protein to the active protomer (dark green line). But simultaneous activation of both protomers by their corresponding agonists leads to energy minima less stable than activation of a single protomer (light green line), most probably because of a steric clash of the active conformations of TM 5 and TM 6 in both protomers, leading to negative crosstalk (Figure 1C). These putative mechanisms for cross-antagonism and negative crosstalk would apply irrespective of a dimeric or heterotetrameric structure of a GPCR heteromer, since these allosteric modulations would, after all, depend on the key role of the TM5 and TM6 on heteromerization and on receptor activation. Thus, restrictions of active conformations of the GPCR protomers involved in heteromerization would be directly imposed by molecular arrangements within the heteromeric interface.

In addition to providing the frame for allosteric interactions between ligands binding to their orthosteric sites, heteromerization can lead to changes in the properties of specific ligands for one of the protomers, independent of other ligands binding to the other protomer (Figure 3B). Indeed, the proof of concept came from experiments in transfected mammalian cells, where the potencies of different selective A_2AR antagonists at binding to the A_2AR alone or when co-expressed either with D₂R or A₁R were compared (Orrú et al., 2011). Interestingly, the most dramatic finding was that the A_2AR antagonist SCH442416 showed a selective low affinity for the A_2AR when co-expressed with D₂R compared with its affinity for the A_2AR alone or co-expressed with A₁R. More specifically, SCH442416 showed a pronounced negative cooperativity of its binding to the A_2AR homodimer within the A_2AR-D₂R heterotetramer, a biphasic antagonist/antagonist competition curve (Orrú et al., 2011). These results were in fact the first to reveal that the quaternary structure of the A_2AR-D₂R heteromer included A_2AR homomers. The specific behavior of SCH442416 in the A_2AR-D₂R heteromer could then be used as a biochemical fingerprint of the heteromer in native tissues. Thus, SCH442416 negative cooperativity could be demonstrated in mouse striatal preparations, but not from mice with conditional striatal D₂R-KO (Ferré et al., 2018).

The same type of ligand-independent allosteric interaction has recently been reported for the selective significant decrease in potency of the MOR agonist methadone, compared with morphine and fentanyl, within the MOR-galanin Gal₁ receptor heteromer (Gal₁R) (Cai et al., 2019). Since these heteromers are selectively localized in the ventral tegmental area (localization of cell bodies of dopaminergic cells involved in the processing of natural rewards), this could explain a selective dissociation of the clinical therapeutic versus euphoric/rewarding effects of methadone compared with the other opioids (Cai et al., 2019). We are obtaining data from computational modeling indicating that binding of the Gal₁R to the MOR promotes a little displacement of TM6 which can selectively alter the stability of methadone in the MOR binding pocket (in preparation). Therefore, changes in the pharmacological properties of ligands induced by receptor heteromerization should constitute a very important strategy in the search for new therapeutic agents with further selectivity for their therapeutic versus side, unwanted effects.

6. Molecular architecture of GPCR heteromers

If the pentameric structure constituted by a GPCR homodimer and one heterotrimeric G protein provides a main functional unit, oligomeric entities should be expected to often represent multiples of dimers and GPCR heteromers would be often constituted by heteromers of homodimers (Ferré et al., 2014; Ferré, 2015). This seems to apply particularly to heteromers that include a GPCR homodimer preferentially coupling to Gs/olf (Gs for short) proteins and another GPCR homodimer preferentially coupled to Gi/o (Gi for short) proteins. It was postulated (Ferré, 2015) that such a ‘GPCR heterotetramer’ would sustain a functional pre-coupled macromolecular complex that includes two molecularly different GPCRs, their cognate G proteins and adenylyl cyclase (AC) and would provide the frame for a canonical interaction at the AC level. This canonical interaction implies the ability of an activated Gi-coupled receptor to inhibit a Gs-coupled receptor-mediated activation of certain subtypes of AC. More specifically, those AC isoforms than can be inhibited by the α subunits of Gi proteins, including AC1, AC5, and AC6 (Sadana et al., 2009a), and requires the simultaneous respective interaction of the Ras-GTPase domains of the α subunits of the Gs and Gi proteins with the C2 and C1 catalytic domains of AC (Dessauer et al., 1998).

The structural details of TM interfaces of GPCR heteromers, in the absence of crystal structures, cannot be predicted. Hence, in order to suggest the molecular interfaces of GPCR heteromers, we applied an indirect approach consisting on the application of synthetic peptides with the amino acid sequence of TMs of the corresponding GPCRs, to alter inter-protomer interactions in the heteromeric and homomeric interfaces in the A_2AR-D₂R heterotetramer. The disruptive peptides were fused to the HIV transactivator of transcription (TAT) peptide, which facilitates their insertion in the plasma membrane and determines its correct orientation (He et al., 2011). The TAT-TM peptides were then tested in bimolecular fluorescence complementation (BiFC) experiments, for their specific ability to disrupt BiFC of complementary halves of a fluorescent molecule separately fused to two different A_2AR protomers, two different D₂R protomers or to an A_2AR and a D₂R protomer. From the fourteen possible TAT-TM peptides corresponding to all TM domains of the A_2AR and the D₂R (abbreviated as TM1, TM2… of the respective GPCR), only TM6 of the A_2AR and TM6 of the D₂R disrupted BiFC of A_2AR-A_2AR and D₂R-D₂R homodimers, respectively (Navarro et al., 2018). These results therefore indicate a symmetrical TM 6/TM 6 interface for both homomers. The different apparent interface of D2R homomers reported in previous studies (which suggested the involvement of TM 4 and TM 5 (Guo et al., 2005) could be due to the different experimental approaches and, most likely, due to the presence of heteromeric partner receptors that influence the homomeric interfaces (Kofalvi et al., submitted). The involvement of TM 6 in D₂R oligomerization was further substantiated by the selective ability of TM6 of D₂R to completely counteract the decrease in binding of a bivalent ligand with selective picomolar affinity for the D₂R homodimer (Pulido et al., 2018). Since the binding experiments were performed in membrane preparations of mammalian striatum, the bivalent ligand allowed to identify D₂R homomers (homodimers or oligomers) in the native tissue. An important corollary of these findings is that they imply that D₂R homomers are the predominant species in relation to D₂R monomers in the brain, since a significant proportion of monomers would have masked the identification of the D₂R homomers.

The fact that rearrangement of TM 6 constitutes a main ligand-induced conformational change that determines G protein activation and modulation of ligand affinity (see above), provides a frame for the understanding of allosteric communications, negative cooperativity, through the protomers in A_2AR and D₂R homomers. In fact, when constructing computational models of the structures for A_2AR and D₂R homodimers with a symmetrical TM 6/TM 6 interface, TM 6 in the inactive closed conformation of the unliganded protomer can only interact with TM 6 in the active open conformation of the G protein-bound protomer, but simultaneous outward movements of TM 6 in the homodimer is not feasible due to a steric clash between active open conformation of both TM 6 (Navarro et al., 2018). That is to say, an active conformation of one of the protomers makes very unlikely the active conformation of the other protomer. These interpretations are in agreement with those of a previous study by Jonathan Javitch’s group (Han et al., 2009), who used several constructs of wild-type and mutant D₂R (binding- or G protein activation-deficient D₂R) fused and non-fused to a reporter G protein. The potential ability of a non-fused protomer to activate the G protein fused to another protomer allowed evaluating the key role of each protomer in the D₂R homodimer. The results of the study supported, first, that the minimal signaling D₂R unit is composed of two protomers and a single G protein. Second, the study showed that the homodimer is more strongly activated by an agonist binding to a single protomer, whereas simultaneous binding to the second protomer blunts signaling (Han et al., 2009). But also, a constitutive activity of the second protomer decreases agonist-induced activation of the first protomer, since maximal activation of the homodimer is achieved upon agonist and inverse agonist binding to the first and second protomer, respectively (Han et al., 2009). In other words, the inactive conformation of one of the protomers makes very likely the active conformation of the other protomer. An additional, mathematical, interpretation was provided by Rovira et al. (2010), who explained the results by Han et al. (2009) by considering two active states of the D₂R homodimer, which would be associated to different signaling pathways and whose corresponding protomers are singly or doubly activated.

From the fourteen possible TAT-TM peptides corresponding to all TM domains of the A_2AR and the D₂R, specifically, TM4 and TM5 of both the A_2AR and the D₂R disrupted BiFC of A_2AR-D₂R heteromers (Navarro et al., 2018). These results therefore indicated a symmetrical TM 4–5/TM 5–4 heteromeric interface. When taking also into account the results of TAT-TM peptides mentioned above on the A_2AR and D₂R homomeric interfaces, computational modeling resulted in one minimal solution that accommodates the TMs 4/5 interface for A_2AR-D₂R heterodimerization and the TM 6 interface for both A_2AR-A_2AR and D₂R-D₂R homodimerization (Navarro et al., 2018). The existence of these interfaces implies two internal interacting A_2AR and D₂R protomers and two external A_2AR and D₂R protomers in which the α-subunits of Gi and Gs bind to the corresponding external protomers of the D₂R or A_2AR homodimers, respectively (Navarro et al., 2018). This would be the only feasible configuration to avoid any steric clash between the two G proteins simultaneously bound to the complex. Finally, the model also predicts a large distance between both βγ subunits (Navarro et al., 2018).

Apart from the two large catalytic domains, C1 and C2, the topology of AC includes a long cytoplasmic N terminus and two membrane-spanning domains, M1 and M2, each with six TMs. Carmen Dessauer’s group has provided evidence for a pre-coupling of AC5 and G proteins involving the βγ subunits and the N terminus of AC (Sadana et al., 2009b). Using BRET and BiFC techniques with interfering synthetic peptides (with the amino acid sequence of the twelve domains of AC5) we were able to demonstrate that specific TMs of AC5 establish molecular interactions with TMs of A_2AR and D₂R in the A_2AR-D₂R heterotetramer (Navarro et al., 2018). We therefore found for the first time a functional role for the TMs of AC, their coupling to GPCRs. These interactions, which involve TMs of the receptors that do not form part of the homomeric and heteromeric interfaces, change upon the presence of A_2AR or D₂R ligands. Computational modeling supports that the ligands induce a rearrangement of the A_2AR-D₂R-AC5 complex which depends on the rearrangement of the βγ subunits. This movement of the βγ subunits uncovers the Ras-GTPase domain of the Gα subunit that interacts with the corresponding catalytic domain of AC (Cabrera-Vera et al., 2003) and direct the movement of AC to allowing this Gα-AC interaction (Navarro et al., 2018).

The utilization of TAT-TM peptides in striatal primary neuronal cultures allowed demonstrating that the canonical interaction at the AC level required the right quaternary structure of the A_2AR-D₂R heteromer in the A_2AR-D₂R-AC5 complex. Application of TM peptides that specifically interfere with the heteromeric interface or with the receptor-AC5 interface specifically counteracted the ability of a D₂R agonist to inhibit AC5 activation by an A_2AR agonist (Navarro et al., 2018). More generally, these results indicate that, although Gi-coupled receptors can artificially inhibit forskolin-induced AC activation, to inhibit a Gs-coupled receptor-mediated AC activation, they need to be part of a GPCR heterotetramer. This has been so far shown for the A_2AR-D₂R, A₁R-D₁R, and D₁R-D₃R heterotetramers, where destabilization of their heteromeric interface leads to the disruption of the canonical Gs-Gi antagonistic interaction at the AC level (Navarro et al., 2018, Rivera-Oliver et al., 2019, Guitart et al., 2019). When again considering the definition of allosterism in the frame of GPCRs, an important corollary to the concept of pre-coupled macromolecular complexes of heterotetramers, Gs and Gi proteins and the effector molecule AC is that the canonical Gs-Gi antagonistic interaction at the AC level also represents an allosteric interaction. In this case, the process by which the interaction of a chemical at one location (an orthosteric ligand of the Gi-coupled receptor) of a macromolecular complex (formed by heterotetramers, Gs and Gi proteins and the effector molecule AC) influences the function of another chemical at a topographically distinct site (an orthosteric ligand of the Gs-coupled receptor). It also becomes obvious that we should modify the definition of ‘allosteric interaction in the receptor heteromer’ [Ferré et al., 2009] (see above). Otherwise, it would probably be more appropriate to avoid a general definition and to describe the specific type of allosteric interaction in the GPCR heteromer: Heteromeric or homomeric allosteric interactions between ligands, ligand-independent allosteric interactions or allosteric interactions through the effector (Figure 3).

It is important to underline that the GPCR heterotetramer represents only a very plausible common architecture of GPCR heteromers, but growing experimental evidence suggests the existence of heterodimeric structures or of heteromeric complexes larger than GPCR heterotetramers. Those complexes should be also determined by their immediate intramembrane context, by their role as elements of specific GPCR signaling complexes.

7. Final remarks

There seems to be no doubt that GPCR oligomers are not only demonstrable in artificial systems, but also in native tissues, where their prevalence (versus monomeric entities) should also be out of question, at least for some GPCRs. Experimental tools are now available to determine the oligomeric interfaces and quaternary structure of the oligomers, as well as to dissect the specific biochemical properties of GPCR heteromers. In fact, the same techniques are allowing to disclose a much more elaborated picture, with GPCR oligomers just being the tip of the iceberg, constituting components of pre-coupled macromolecular complexes that include G proteins and effectors, such as AC. Allosteric properties can therefore extend to ligand interactions through the different components of the macromolecular complexes. Identification of specific heteromers in individual cells of native tissues, as well as their specific biochemical properties, should lead to new therapeutic approaches (as reviewed elsewhere; see for instance Ferré et al., 2014; Gomes et al., 2016).