G protein-coupled receptor hetero-dimerization: contribution to pharmacology and function

Authors

  • Graeme Milligan

    1. Molecular Pharmacology Group, Neuroscience and Molecular Pharmacology, Faculty of Biomedical and Life Sciences, University of Glasgow, Glasgow, UK
    Search for more papers by this author

Graeme Milligan, Molecular Pharmacology Group, Neuroscience and Molecular Pharmacology, Faculty of Biomedical and Life Sciences, University of Glasgow, Glasgow G12 8QQ, UK. E-mail: g.milligan@bio.gla.ac.uk

Abstract

The concept that G protein-coupled receptors (GPCRs) can form hetero-dimers or hetero-oligomers continues to gain experimental support. However, with the exception of the GABAB receptor and the sweet and umami taste receptors few reported examples meet all of the criteria suggested in a recent International Union of Basic and Clinical Pharmacology sponsored review (Pin et al., 2007) that should be required to define distinct and physiologically relevant receptor species. Despite this, there are many examples in which pairs of co-expressed GPCRs reciprocally modulate their function, trafficking and/or ligand pharmacology. Such data are at least consistent with physical interactions between the receptor pairs. In recent times, it has been suggested that specific GPCR hetero-dimer or hetero-oligomer pairs may represent key molecular targets of certain clinically effective, small molecule drugs and there is growing interest in efforts to identify ligands that may modulate hetero-dimer function selectively. The current review summarizes key recent developments in these topics.

Abbreviations:
DOP

δ-opioid peptide

GPCR

G protein-coupled receptor

MOP

µ-opioid peptide

RET

resonance energy transfer

Introduction

G protein-coupled receptors (GPCRs) are the largest family of transmembrane signalling proteins and have been the most tractable target class for the development of therapeutic, small molecule drugs. Until relatively recently GPCRs were considered to exist as monomeric polypeptides. At least in part, this view was based on the monophasic binding characteristics of most antagonist ligands selective for individual GPCRs, while the pharmacology of individual receptors was presumed to provide a detailed and definitive signature of an individual GPCR that would be invariant between tissues. Indeed, prior to the era of cDNA cloning and genome sequencing, subtle variations in ligand structure–activity relationships between tissues, and even between species, was a driver in efforts to subdivide and further characterize GPCRs that responded to the same or overlapping endogenous ligands. In the last decade a series of revolutions in our understanding of protein–protein interactions in cells, and the changes in ligand pharmacology that can be observed with minor sequence variation in individual GPCRs, has altered our view of many of these assumptions. It is now well established that the interaction of a GPCR with intracellular GPCR-interacting proteins can alter the pharmacology, function and/or the trafficking of GPCRs in cells (Milligan and White, 2001; Hall and Lefkowitz, 2002). Furthermore, appreciation that open reading frame polymorphisms of GPCRs can alter pharmacology, function or regulation of a receptor is of crucial importance to the development of concepts of personalized medicine and drug treatment (Tang and Insel, 2005; Insel et al., 2007; Kazius et al., 2008). Equally, appreciation of the variation in pharmacology of ligands at species orthologues of GPCRs (Link et al., 1992; Lim et al., 2008) has resulted in consideration of the most appropriate animal models to assess likely efficacy of drugs in man and has ensured that screening and characterization of potential novel therapeutic agents are performed on GPCRs of human origin. Finally, it is becoming increasingly apparent that the basic lexicon of ligand pharmacology: agonist, antagonist and inverse agonist, must be considered to be system- and context-dependent. In a substantial number of cases described in recent years the same ligand can act as an agonist, antagonist or inverse agonist at the same GPCR, dependent upon the assay end point that is measured. This may reflect differential binding modes and the stabilization of distinct conformations of the receptor (Lane et al., 2007; Bhattacharya et al., 2008; Galandrin et al., 2008). Equally the same GPCR may generate distinct signals in different cells. For example, it has recently been shown that the calcium-sensing receptor may cause stimulation or inhibition of secretion of parathyroid hormone-related protein depending on cellular context and coupling to Gs or Gi (Mamillapalli et al., 2008).

As well as each of the issues noted above, the last decade has seen a substantial re-evaluation of the concept that GPCRs exist primarily as monomeric polypeptides, with increasing support for a model in which GPCRs can exist as homo- or hetero-dimers or as homo- or hetero-multimers. The implications of this for ligand pharmacology and its potential for novel drug design will be the focus of this review.

What is the evidence for GPCR homo- and hetero-multimerization?

Homo-dimerization

Protein dimerization is a remarkably common theme in biology (Marianayagam et al., 2004; Mei et al., 2005), and one reason for this is suggested to be the role of protein–protein interfaces in the control of protein folding during synthesis. Furthermore, there may be a propensity for individual GPCR monomers to dimerize or oligomerize in membranes as molecular dynamic simulations of individual molecules of rhodopsin in phospholipid bilayers demonstrate such self-assembly (Periole et al., 2007). Early studies inconsistent with the concept of GPCRs as monomeric species were scattered in the literature and have been reviewed by Salahpour et al. (2000). However, in concert with studies that showed that co-expression of two distinct forms of the angiotensin II AT1 receptor, each unable to bind ligand, resulted in the presence of a ligand binding-competent form of the receptor via protein trans-complementation (Monnot et al., 1996), it was the advent of a series of co-immunoprecipitation studies following the co-expression of pairs of differentially epitope-tagged GPCRs that provided initial biochemical evidence for the presence of multiple copies of each of the β2-adrenoceptor (Hebert et al., 1996), the dopamine D2 receptor (Ng et al., 1996) and the δ-opioid peptide (DOP) receptor (Cvejic and Devi, 1997) within a complex. Each of these studies indicated an indeterminate proportion of the GPCR was likely to exist as a dimer and provided impetus for a vast range of related experiments that continue to the present day. Interestingly, each of these early studies also explored aspects of the selectivity, functional consequences, ligand regulation and/or molecular basis of GPCR dimerization and generated data that remain controversial. For example, in the studies of DOP receptor dimerization, Cvejic and Devi (1997) provided evidence to indicate that the addition of certain agonists would inhibit or reverse receptor dimerization and, as a corollary, they concluded that the DOP receptor was likely internalized from the cell surface as a monomer in response to agonist challenge. McVey et al. (2001) subsequently used combinations of co-immunoprecipitation studies and two distinct forms of living cell-based resonance energy transfer (RET) techniques to confirm the ability of the DOP receptor to form a homo-multimer that was present at the cell surface. However, efficacious agonists were unable to dissociate the complex. Ng et al. (1996) provided evidence that the dopamine D2 receptor was not only able to homo-dimerize but also to hetero-dimerize with the 5-hydroxytryptamine 5-HT1B receptor, and that certain ligands were selective in binding either dopamine D2 receptor monomer or dimer. While the prospect of identifying ligands that selectively bind monomers or homo-multimers of the same GPCR remains attractive and would offer great potential for in situ identification of monomers and dimers, this has not been verified or extended convincingly. Hebert et al. (1996) reported that addition of a synthetic peptide corresponding to transmembrane domain VI of the β2-adrenoceptor was able to interfere with receptor dimerization and limit agonist activation of adenylyl cyclase, suggesting that the dimer was important for G protein activation and for the function of the receptor. However, recent studies in which monomers of the β2-adrenoceptor were incorporated into reconstituted high-density lipoprotein phospholipid bilayer particles, together with the stimulatory G protein Gs, demonstrated the capacity of the receptor monomer to produce G protein activation (Whorton et al., 2007). Similar approaches have also indicated the capacity of monomeric rhodopsin to activate transducin (Whorton et al., 2008). Despite this, studies have indicated that the β2-adrenoceptor (and a number of other receptors) forms dimers/oligomers during protein synthesis and maturation and prior to cell surface delivery (Salahpour et al., 2004) and that the β2-adrenoceptor is internalized from the cell surface as a homo-dimer in response to binding of a single molecule of agonist to either protomer (Sartania et al., 2007). Such observations question the physiological relevance of the capacity of a purified and reconstituted β2-adrenoceptor monomer to function in an artificial system (Whorton et al., 2007), despite the elegance of the approach used to demonstrate that a β2-adrenoceptor monomer is sufficient to cause G protein activation. Furthermore, although the early studies of Hebert et al. (1996) indicated a likely central role for transmembrane domain VI as a ‘dimer interface’ many recent studies, both experimental and theoretical, have implicated transmembrane domains IV and V as key elements in many GPCRs (Lee et al., 2003; Carrillo et al., 2004; Guo et al., 2005; Kim and Jacobson, 2006; Wang et al., 2006; Harikumar et al., 2007; Mancia et al., 2008). However, it should be noted that in various studies, almost every element of one or other GPCR, including the intracellular C-terminal tail and the extracellular N-terminal region (see Milligan, 2008 for review) and the third intracellular loop (Ciruela et al., 2004) has been suggested to be important for dimerization. There remain skeptics who are unconvinced by the data that support GPCR dimerization/multimerization (Chabre and le Maire, 2005), or who have explored the methods used, particularly those based on RET, and questioned the conclusions reached (James et al., 2006; Meyer et al., 2006). However, a substantial number of reports have attempted to appraise the strengths and weaknesses of the various approaches used (Milligan and Bouvier, 2005; Gandía et al., 2008a), to rebut the criticisms raised (Bouvier et al., 2007; Salahpour and Masri, 2007) or to expand the techniques employed to address this issue (Mesnier and Banères, 2004; Maurel et al., 2008) and support the concept of dimerization. Overall, the vast majority of reports have provided evidence in support of the concept of receptor homo-dimerization (see Milligan, 2004; 2007; 2008; Park et al., 2004; Maggio et al., 2007 for review), while certain studies also indicate the potential for higher-order complexity in GPCR structure (Klco et al., 2003; Carrillo et al., 2004; Park and Wells, 2004; Lopez-Gimenez et al., 2007; Gandía et al., 2008b; Guo et al., 2008) for a number of rhodopsin family receptors although others claim that dimers but not oligomers, of at least the family B secretin receptor (Harikumar et al., 2008) and the family C metabotropic glutamate receptors (Maurel et al., 2008), can be observed.

Hetero-dimerization

If the existence of receptor homo-dimers remains uncertain then clarity over the expression and importance of receptor hetero-dimers is yet more clouded. Although genetic as well as biochemical and pharmacological studies define that the class C GABAB receptor (Jones et al., 1998; White et al., 1998) and the sweet and umami taste receptors (Zhao et al., 2003) represent constitutively formed hetero-dimers of distinct (but closely related) GPCR polypeptides, the significance of a wide range of other reported hetero-dimers remains unclear. In large part, this reflects that key studies have often been limited to experiments involving the co-transfection of pairs of GPCR cDNAs into heterologous cell lines that are easy to manipulate, or that important observations from native cells and tissues have not always been verified independently. For example, although AbdAlla et al. have published a series of interesting papers on angiotensin AT1 receptor hetero-dimers in pre-eclampsia and in experimental hypertension (AbdAlla et al., 2001; 2005), these studies have not been replicated independently. Equally, fascinating observations of the ability of both angiotensin AT1 receptor blockers and β-blockers to inhibit downstream signalling via both receptors have been interpreted as reflecting their in vivo hetero-dimerization (Barki-Harrington et al., 2003) but, currently, no follow-up of these pharmacologically surprising results has been reported. Many of the early studies on GPCR hetero-dimerization have been reviewed and appraised widely (Milligan, 2006; Szidonya et al., 2008). The International Union of Basic and Clinical Pharmacology has recently suggested guidelines for the range of evidence that might be accumulated prior to claiming evidence of novel GPCR hetero-multimers (Pin et al., 2007). However, in only a limited number of examples has full concordance with these guidelines been achieved (Pin et al., 2007). Although there are many appealing aspects of GPCR hetero-dimers as novel therapeutic targets (Milligan, 2006), until recently the pharmaceutical industry has been unsure of the viability of exploring this avenue (Kent et al., 2007). Not least, this reflects the requirement for more clear-cut validation of the expression of pathophysiologically relevant hetero-dimers, a better understanding of their tissue distribution and the challenges inherent in attempting to screen for and identify small molecule ligands with specificity, or at least significant selectivity, for GPCR hetero-dimers (Milligan, 2006; Eglen et al., 2007; Dalrymple et al., 2008).

Questions relating to the selectivity of GPCR hetero-multimerization, the existence of such complexes in native cells and tissues and their contribution to pathophysiology, the effects of ligands on hetero-dimer behaviour, the molecular basis of dimerization and multimerization and the potential to identify ligands that specifically bind to and regulate GPCR hetero-multimers are currently the most actively researched themes in this area.

Why are receptors dimers?

Facilitation of G protein activation?

It has been suggested, based on the molecular dimensions of the atomic level structures of bovine rhodopsin and of hetero-trimeric G proteins, that a GPCR dimer might provide the most appropriate footprint to bind a G protein (Fotiadis et al., 2006). However, despite evidence that the leukotriene B4 BLT1 receptor–Gi G protein complex is a pentamer consisting of one copy of each of the α, β and γ G protein subunits and two copies of the receptor polypeptide (Banères and Parello, 2003), it is now clear that isolated GPCR monomers can bind and activate G proteins (Whorton et al., 2007; 2008). Indeed, monomers of the neurotensin NTS1 receptor are reported to activate G protein more effectively than dimers (White et al., 2007). It thus appears unlikely that a requirement for signal generation underlies GPCR dimerization (see Gurevich and Gurevich, 2008 for review).

Control of cell surface delivery?

A number of studies have suggested that GPCR dimerization may be important in cell surface delivery and that dimerization is initiated early in protein synthesis. Taking advantage of clear understanding of the role of the C-terminal endoplasmic reticulum (ER) retention motif of GABAB1 in the production and cell surface delivery of the functional GABAB receptor hetero-dimer (Margeta-Mitrovic et al., 2000), Salahpour et al. (2004) replaced the C-terminal tail of the β2-adrenoceptor with the C-terminal tail of the GABAB1 and demonstrated intracellular ER retention of this construct when expressed in human embryonic kidney (HEK)293 cells. When co-expressed with the retained mutant, wild-type β2-adrenoceptor was also retained intracellularly (Salahpour et al., 2004). Such a ‘dominant negative’ effect of the chimeric β2-adrenoceptor–GABAB1 construct supports the idea that protein–protein interactions between the engineered ER-retained and wild-type forms of the receptor occurred during receptor synthesis and maturation. Addition of a distinct, 14 amino acid ER retention motif from the α2C-adrenoceptor to the C-terminal tail of the chemokine CXCR1 receptor resulted in intracellular retention of this chimeric construct in HEK293 cells. This ER-retained construct also limited cell surface transport of co-expressed wild-type forms of both CXCR1 and CXCR2 (Wilson et al., 2005), supporting other evidence of both CXCR1 receptor homo-dimerization and CXCR1–CXCR2 receptor hetero-dimerization (Milligan et al., 2005). The selectivity of such interactions was indicated because the presence of the trapped CXCR1 receptor was without effect on cell surface delivery of a co-expressed α1A-adrenoceptor (Wilson et al., 2005). Equally, by employing a form of the α1B-adrenoceptor containing mutations in both transmembrane domains I and IV that is retained in the ER and fails to mature properly (Canals et al., 2009) and adding distinct bimolecular fluorescence complementation-competent forms of enhanced yellow fluorescent protein to this receptor, ‘dimerization’ could be shown within the ER (Lopez-Gimenez et al., 2007). The transmembrane domains I and IV mutant of the α1B-adrenoceptor is, however, compromised in dimerization/oligomerization because it displays a reduced effectiveness to generate FRET signals when appropriately tagged forms are co-expressed (Lopez-Gimenez et al., 2007). This mutant was also able to interact with co-expressed wild-type α1B-adrenoceptor and, by acting as a ‘dominant negative’, retain the wild-type receptor within the ER. As with a number of ER-retained mutant GPCRs, the transmembrane domains I and IV mutant of the α1B-adrenoceptor could be trafficked to the surface of cells by the maintained presence of a ‘pharmacological chaperone’ (see Conn et al., 2007 for review) in the form of the α1-adrenoceptor antagonist prazosin (Canals et al., 2009). Cell surface delivery was proceeded by maturation of the N-glycosylation status of the mutant receptor and improved dimerization/oligomerization of the mutant receptor as measured by enhanced FRET signals (Canals et al., 2009). Interestingly, when the transmembrane domains I and IV mutant of the α1B-adrenoceptor was co-expressed with, and caused ER-trapping of, a form of the α1B-adrenoceptor that was mutated to prevent the binding of prazosin but was otherwise wild-type, the maintained presence of prazosin caused cell surface delivery of both the transmembrane domains I and IV mutant and the ligand binding-deficient form of the receptor (Canals et al., 2009). These data are best explained by the generation of dimer/oligomer interface interactions within the ER and the cell surface trafficking of a dimer/oligomer containing both forms of the α1B-adrenoceptor (Figure 1). These studies appear to provide strong support for the concept that GPCRs traffic to the cell surface as dimeric or oligomeric complexes and only after passing ER/Golgi export quality control. Along with real-time FRET-based studies on the location and transport of 5-hydroxytryptamine 5-HT2C receptor dimers/oligomers (Herrick-Davis et al., 2006), such studies help to define ER-to-cell surface trafficking of mammalian class A GPCR homo-dimers or homo-multimers. There have also been a number of studies in which co-expression of pairs of GPCRs promotes surface localization that has been used to support the concept of hetero-dimerization (see Minneman, 2007 for review). In part, such studies have built on the recognition that co-expression of the GABAB2 polypeptide was required to interact with the GABAB1 polypeptide to allow trafficking of the hetero-dimer complex to the cell surface and functional expression of the GABAB receptor (Margeta-Mitrovic et al., 2000). This reflects the capacity of the GABAB2 subunit to mask an ER retention motif within the C-terminal tail of the GABAB1 subunit. In class A receptors, interactions between the α1B-adrenoceptor and the α1D-adrenoceptor promoted cell surface delivery of the α1D-adrenoceptor (Hague et al., 2004b), and a similar effect was produced by co-expression of the α1D-adrenoceptor with the β2-adrenoceptor (Uberti et al., 2005). Hague et al. (2004a) have also demonstrated the capacity of a co-expressed β2-adrenoceptor to allow cell surface delivery of certain olfactory receptors. Although the physiological significance of this remains to be established, the lack of effectiveness of other adrenoceptors to promote cell surface delivery of the olfactory receptors suggested that it should be straightforward to define the structural determinants of such interactions. Indeed, a follow-up study has suggested a key role for transmembrane domain II of the β2-adrenoceptor (Bush et al., 2007).

Figure 1.

The α1B-adrenoceptor traffics to the cell surface as a dimer/oligomer. A form of the α1B-adrenoceptor containing mutations in transmembrane domains (TM) I and IV (α1BTMITMIV, blue) (Lopez-Gimenez et al., 2007) is retained in the endoplasmic reticulum (ER) when expressed in HEK293 (human embryonic kidney) cells and study employing bimolecular fluorescence complementation indicate that it exists as a dimer/oligomer (Lopez-Gimenez et al., 2007). Sustained treatment of such cells with the α1-adrenoceptor antagonist prazosin results in maturation of the receptor and its movement to the plasma membrane (PM). A form of the α1B-adrenoceptor that is wild-type except that it contains an Asp125Ala mutation that eliminates its ability to bind prazosin (α1BD125A, yellow) is delivered successfully to the PM when expressed. When α1BD125A is co-expressed with α1BTMITMIV, α1BD125A becomes ER-retained because α1BTMITMIV interacts with α1BD125A and functions as a ‘dominant negative’. Treatment of these cells with prazosin results in movement of both α1BTMITMIV and α1BD125A to the cell surface. As α1BD125A cannot bind prazosin, these observations indicate that the two forms of the α1B-adrenoceptor move to the PM as a dimer/oligomer (see Canals et al., 2009 for further details).

GPCR hetero-dimerization and ligand pharmacology

The ability of a selective ligand at one GPCR to modulate the function of a second, co-expressed GPCR does not inherently imply hetero-dimerization between the two GPCRs. Indeed, such effects often reflect either heterologous sensitization or desensitization or input to common signalling pathways downstream of the receptors in question. For example, Lopez-Gimenez et al. (2008) have recently shown that co-activation of a 5-hydroxytryptamine 5-HT2A receptor co-expressed with the µ-opioid peptide (MOP) receptor in HEK293 cells results in the MOP receptor agonist morphine being able to induce each of internalization, desensitization and down-regulation of the MOP receptor. By contrast, morphine was unable to produce any of these effects in the same cells lacking the 5-HT2A receptor, even when 5-hydroxytryptamine was added along with morphine (Lopez-Gimenez et al., 2008). Tagging of the two GPCRs with distinct auto-fluorescent proteins demonstrated that in untreated cells the MOP receptor was present almost exclusively at the cell surface whereas the bulk of the 5-HT2A receptor was present in punctate, intracellular vesicles and could not, therefore, be complexed within a hetero-dimer with the MOP receptor (Lopez-Gimenez et al., 2008). This can be contrasted with functional interactions between the cannabinoid CB1 receptor and the orexin OX1 receptor. When expressed alone in HEK293 cells the OX1 receptor was present at the cell surface, whereas when the cannabinoid CB1 receptor was expressed alone in such cells it displayed a distribution pattern consistent with rapid, ligand-independent recycling between the cell surface and endosomes (Ellis et al., 2006). Following co-expression, the OX1 receptor adopted the distribution pattern of the cannabinoid CB1 receptor and the addition of RET-competent tags to the C-terminal tail of each receptor demonstrated hetero-interactions between the two co-expressed receptors (Ellis et al., 2006). The presence of the cannabinoid CB1 receptor antagonist/inverse agonist SR-141716A (N-(piperidino-1-yl)-5-(4-chlorophenyl)-1-(2,4-dichlorophenyl)-4-methyl-pyrazole-3-carboxamide), also known as rimonabant, resulted in redistribution of the CB1 receptor to the cell surface, whether expressed in isolation or co-expressed with the OX1 receptor (Ellis et al., 2006). Importantly, in cells co-expressing these two receptors treatment with SR-141716A also caused redistribution of the OX1 receptor to the cell surface, despite SR-141716A having no significant affinity to bind directly to the OX1 receptor (Ellis et al., 2006). In an equivalent manner, the selective OX1 receptor antagonist SB-674042 (1-(5-(2-fluorophenyl)-2-methylthiazol-4-yl)-1-((S)-2-(5-phenyl-[1,3,4]oxadiazol-2-ylmethyl)-pyrrolidin-1-yl)-meth-anone), which displayed no significant affinity to bind directly, caused redistribution of the cannabinoid CB1 receptor to the surface of cells, but only when the OX1 receptor was co-expressed (Ellis et al., 2006). The most obvious interpretation of these observations is that when co-expressed, the OX1 receptor and cannabinoid CB1 receptor form a stable hetero-dimer complex that is able to bind and be regulated by both cannabinoid CB1 and OX1 receptor ligands. A substantial number of studies have shown the ability of selective, usually agonist, ligands to produce co-trafficking of a GPCR that has been co-expressed along with the receptor that binds the selective ligand or to limit internalization of the partner receptor (Jordan et al., 2001; So et al., 2005; Roumy et al., 2007; Ecke et al., 2008; Fiorentini et al., 2008). Although it has been suggested that such ligand-induced ‘co-internalization’ may reflect other factors (Janoshazi et al., 2007), it is certainly consistent with the concept of ‘hetero-dimerization’ and internalization of the intact complex from the cell surface.

Selective ligands may also have direct and selective effects on the conformation of the GPCR protomers that comprise the hetero-dimer. In the case of the GABAB receptor, binding of the neurotransmitter GABA in the N-terminal region of the GABAB1 element of the hetero-dimer is transmitted to activation of G protein via the GABAB2 protomer. This implies communication between the two elements of the hetero-dimer and a conformational alteration of GABAB2 in response to the ligand, although it does not bind directly to this element. Communication between the promoters of a leukotriene B4 BLT1 receptor ‘homo-dimer’, consisting of a wild-type protomer and a modified protomer able to bind the agonist only weakly, was indicated by alteration in response to an agonist ligand of the fluorescence properties of a 5-hydroxytryptophan introduced into the mutant protomer (Mesnier and Banères, 2004). Equally, based on co-expression of the MOP receptor with a form of the α2A-adrenoceptor able to detect conformational change within this receptor because of the introduction of a pair of FRET-competent reporters, Vilardaga et al. (2008) reported that morphine was able to produce a conformational alteration in the α2A-adrenoceptor and that this was associated with inhibition of α2A-adrenoceptor-mediated regulation of extracellular signal-regulated kinase (ERK) mitogen-activated protein (MAP) kinase phosphorylation. Such detailed studies employing biophysical and chemical biology approaches clearly demonstrate that certain GPCR–GPCR interactions can alter receptor structure and, potentially, function. Growing evidence also indicates that GPCR hetero-dimerization can alter receptor pharmacology. If so, it should be possible to detect such interactions in either ligand binding or functional screens (Franco et al., 2008a).

As noted above, co-expression of the cannabinoid CB1 receptor and the OX1 receptor results in alterations in the cellular trafficking of both receptors in response to antagonists highly selective for either GPCR (Ellis et al., 2006). The motivation to explore potential hetero-dimerization of these receptors stemmed from earlier studies that showed a marked enhancement of the potency of orexin A to activate the ERK1/2 MAP kinases when the OX1 receptor was co-expressed with the cannabinoid CB1 receptor (Hilairet et al., 2003). Furthermore, this effect was blocked by SR-141716. Although different, in that Ellis et al. (2006) noted little effect of simply expressing the cannabinoid CB1 receptor on the potency of orexin A to activate ERK1/2 MAP kinase activity, they also noted that SR-141716 reduced the potency of orexin A in cells co-expressing the two receptors. Given that rimonabant (Acomplia™) is a drug that has been employed clinically to treat obesity, and the cell studies demonstrated its effect to reduce the potency of a signal anticipated to be strongly pro-orexegenic, it is clearly an interesting speculation that rimonabant exerts at least part of its action via a cannabinoid CB1 receptor–OX1 receptor hetero-dimer. However, the existence of such a complex in vivo remains to be confirmed. It would be of great interest to know the effectiveness of rimonabant in OX1 receptor knockout mouse models.

As noted earlier, the identification of small molecule ligands specific or highly selective for particular GPCR hetero-dimers would be of great value. As well as being of use as ‘proof of concept’ agents to promote efforts in screening by the pharmaceutical industry, they would allow the analysis of hetero-dimer expression and function in primary cells, tissues and animal models. Currently, the best-described hetero-dimer-selective ligand is 6′-guanidinonaltrindole (Waldhoer et al., 2005). This simple derivative of a κ-opioid peptide (KOP) receptor ligand is reported to act as a selective agonist in cells co-expressing the KOP and DOP opioid receptors, via the DOP–KOP receptor hetero-dimer. It has also been reported to function as a spinally selective analgesic, despite it being only some sixfold more potent in cells co-expressing the KOP and DOP receptors than in cells expressing only the KOP receptor (Waldhoer et al., 2005). However, as in many initial ligand screening campaigns, the potency data were derived from studies employing the channelling of signals to the elevation of [Ca2+] via use of a promiscuous, chimeric G protein (Milligan and Rees, 1999) rather than via measurement of an end point, such as regulation of ion channel function, which might be directly relevant to opioid function in vivo. Despite significant interest in these studies (Park and Palczewski, 2005), independent confirmation of the observations is still awaited. It is difficult to envisage 6′-guanidinonaltrindole spanning the two ligand binding pockets of two GPCR protomers that form a ‘contact’ DOP–KOP hetero-dimer. However, although they are likely to be less favourable from an energetic standpoint and, therefore, if they exist, may represent only a small proportion of any GPCR dimer (Bakker et al., 2004), ‘domain swap’ hetero-dimers might be anticipated to generate unique ligand binding pockets that would bind ligands with distinct affinity and produce unique ligand structure–activity relationships. There is, moreover, a substantial literature on alterations in ligand potency with co-expression of pairs of opioid receptors and on ligand pharmacology in native tissues that is not reproduced by the expression of individual opioid receptor subtypes in heterologous cell lines (Levac et al., 2002). Furthermore, the generation of synthetic opioid ligands containing a backbone linker separating two distinct opioid pharmacophores has provided evidence for the organization of co-expressed opioid receptor subtypes as hetero-dimers (Xie et al., 2005).

Another indication of selectivity of ligands for receptor hetero-dimers comes from studies on interactions between co-expressed dopamine D1 and D2 receptors. Although the dopamine D1 receptor is commonly associated with elevation of cAMP levels via activation of Gs and the dopamine D2 receptor with inhibition of cAMP via activation of Gi-like G proteins, in cells co-expressing these two receptors Rashid et al. (2007) demonstrated that either the endogenous agonist dopamine or a combination of selective D1 and D2 receptor agonists resulted in elevation of Ca2+ via YM254890-sensitive G proteins of the Gq/G11 family. Similar results were observed in striatal tissue from wild-type but in neither D1 nor D2 knockout mice (Rashid et al., 2007), and SKF83959 (3-methyl-6-chloro-7,8-hydroxy-1-[3-methylphenyl]-2,3,4,5-tetrahydro- 1H-3-benzazepine) appeared to function as a hetero-dimer-selective agonist. Interestingly, there are a number of ways in which ligands that show little or no direct affinity for a specific GPCR can regulate the function of that GPCR if it forms a hetero-dimer with a receptor for which the ligand does have affinity. This concept has been discussed in terms of allosteric interactions within hetero-dimers (Milligan and Smith, 2007; Springael et al., 2007). For example, following co-expression of forms of the chemokine CXCR2 receptor and the DOP receptor that allow signal generation only if the two receptor constructs form a hetero-dimer, agonist function at the DOP receptor was enhanced by the presence of the CXCR2 blocker SB-225002 ((N-(2-hydroxy-4-nitrophenyl)-N'-(2-bromophenyl)urea) although this compound has no effect on DOP receptor function in the absence of the CXCR2 receptor (Parenty et al., 2008). Although it was assumed that the CXCR2 blocker functioned as an orthosteric antagonist at the CXCR2 receptor, this is clearly not the only means by which communication between the protomers of a hetero-dimer can be detected (Milligan and Smith, 2007) and it is, therefore, of considerable interest that certain CXCR1 and CXCR2 blockers appear to bind at an intracellular site on the receptor (Nicholls et al., 2008). It is thus possible that a range of compounds can function as selective allosteric modulators of GPCR hetero-dimers, and molecules of this nature may be uncovered in a relatively straightforward manner if appropriate hetero-dimer screens and homo-dimer counter-screens are established. The concept that knockout animal models may provide an excellent test bed to assess the chemical ‘fingerprint’ of GPCR hetero-dimers in native tissues has been raised (Franco et al., 2008a,b), and particularly tissue-specific or conditional knockouts may be ideally suited for such studies.

GPCR hetero-dimerization and pathophysiology

A contribution of glutamatergic dysregulation to the aetiology of schizophrenia has long been proposed and has led to approaches other than the standard of antagonism of mono-aminergic receptors being championed. In recent times the effectiveness of LY404039 [(-)-(1R,4S,5S,6S)-4-amino-2-sulfonylbicyclo[3.1.0]hexane-4,6-dicarboxylic acid], a selective agonist of mGlu2/3 receptors, in the treatment of both positive and negative aspects of schizophrenia (Patil et al., 2007) has garnered considerable attention as a potential novel approach. Hallucinogenic drug models of psychosis have a number of similarities to aspects of schizophrenia, and the contribution of signalling of the 5-hydroxytryptamine 5-HT2A receptor via a pertussis toxin-sensitive, Gi family-initiated pathway is central to the discrimination between hallucinogenic and non-hallucinogenic agonists of this receptor (González-Maeso et al., 2007). Interestingly, González-Maeso et al. (2008) have recently shown direct interactions between the 5-HT2A receptor and the mGlu2 but not mGlu3 receptor by combinations of co-immunoprecipitation studies employing human brain tissue and various RET-based studies performed in transfected cell lines. Furthermore, they were able to demonstrate that the affinity of hallucinogenic 5-HT2A receptor agonists to compete with the antagonist [3H]ketanserin for binding to the 5-HT2A receptor in mouse somatosensory cortex was higher in the presence of the mGlu2/3 receptor agonist LY379268 [(1R,4R,5S,6R)-4-Amino-2-oxabicyclo[3.1.0]hexane-4,6-dicarboxylic acid], while, by contrast, the affinity of mGlu2/3 agonists was reduced in the presence of the hallucinogen 1-(2,5)-dimethoxy-4-indophenyl)-2-aminopropane (DOI). Equally, the potency of DOI to promote binding of [35S]GTPγS to pertussis toxin-sensitive G proteins was greatly enhanced by the presence of the mGlu2 but not mGlu3, while this effect of mGlu2 was not evident when LY379268 was also present. LY379268 also caused a marked reduction of potency of DOI to stimulate binding of [35S]GTPγS to pertussis toxin-sensitive G proteins in membranes prepared from cortical cells maintained in primary culture (González-Maeso et al., 2008). These studies suggest that although mGlu2 agonists appear effective in the treatment of schizophrenia, the effectiveness may stem, at least in part, from regulation of the 5-HT2A–mGlu2 hetero-dimer. This is certainly not the only case in which GPCR hetero-dimers have been implicated as therapeutic targets for the treatment of disease. For example, a substantial literature on interactions between dopamine D2 and adenosine A2A receptors in systems ranging from transfected cells to the brain (Fuxe et al., 2007; Ferréet al., 2008) has highlighted the potential for therapeutic strategies that target this complex. However, despite these interesting observations much more needs to be established in terms of tissue distribution, function and differential pharmacology of distinct sets of GPCR hetero-dimers before they will be considered widely as tractable therapeutic targets.

Acknowledgements

The author thanks the Biotechnology and Biosciences Research Council (grant numbers BB/E006302/1 and BB/D001951/1), Medical Research Council (grant number G9811527) and Wellcome Trust (grant number 076774/Z/05/Z) for support of work on GPCR dimerization/oligomerization.

Conflict of interest

The author states no conflict of interest.

Ancillary