G protein-coupled receptor multimers: a question still open despite the use of novel approaches

1 1 acetylcholine ABSTRACT Heteromerization of G protein-coupled receptors (GPCR) can significantly change the functional properties of involved receptors. Various biochemical and biophysical methodologies have been developed the last two decades to identify and functionally evaluate GPCR heteromers in heterologous cells, with recent approaches focusing on GPCR complex stoichiometry and stability. Yet, validation of these observations in native tissues is still lacking behind for the majority of GPCR heteromers. Remarkably, recent studies, particularly some involving advanced fluorescence microscopy techniques, are contributing to our current knowledge on aspects that were not well known till now such as GPCR complex stoichiometry and stability. In parallel, a growing effort is being applied to move the field forward into native systems. This short review will highlight recent developments to study stoichiometry and stability of GPCR complexes and methodologies to detect native GPCR dimers. D 1 R/D 2 R heteromerization involves the C-tail of D 1 R and intracellular loop 3 (IL3) of the long D 2 R isoform. Membrane-permeable fusion constructs consisting of a TAT sequence and D 1 R G 396 -L 413 (TAT-D1 C ) or D 2 R M 257 -E 271 (TAT-D2L IL3 ) peptides disrupted D 1 R/D 2 R complexes and heteromer-induced signaling in striatal neurons (Pei et al. , 2010; Hasbi et al. , 2014). D 1 R/D 2 R coimmunoprecipitation was enhanced from brain tissue derived from patients suffering major depression as compared to healthy persons (Pei et al. , 2010). TAT-D2L IL3 has antidepressant-like effects when injected in brain of rats as revealed by increased mobility in a forced swim tests and reduced escape failures in learned helplessness tests, leading the authors to suggest a prominent role for D 1 R/D 2 R heteromers in this neurological disorder (Pei et al. , 2010). The class A angiotensin receptor type 1a (AT1aR) and class B secretin receptor (SCTR) are co-expressed in osmoregulatory brain centers and form heteromeric complexes in heterologous cells. AT1aR/SCTR heteromerization was specifically inhibited by peptides derived from AT1aR-TM1 and SCTR-TM2, whereas both homo- and heteromerization were inhibited by AT1aR-TM4 and SCTR-TM4 (Lee et al. , Injection of AT1aR-TM1 in mice brain reduced hyperosmolality-induced drinking, confirming the physiological role of this class A/class B GPCR heteromer in the regulation of water homeostasis.


Introduction
G protein-coupled receptors (GPCRs) play a key role in the regulation of cells in our body by activating intracellular signaling in response to a wide variety of specific agonists (Lagerström and Schiöth, 2008;Alexander et al., 2011). Most cells express multiple GPCR subtypes and are consequently able to respond to at least a corresponding number of different agonists (Vassilatis et al., 2003;Regard et al., 2008;Insel et al., 2012;Fève et al., 2014). Distinct GPCRs can affect each other's functions to come to an integrative cellular response through direct physical interactions in heteromeric complexes. Over the last 20 years, dimerization and/or oligomerization was reported for nearly all tested GPCR subtypes using mostly engineered GPCR constructs expressed in heterologous systems ( Fig. 1) (Khelashvili et al., 2010;Cottet et al., 2012).
Class C GPCRs require dimerization to transduce transmembrane signaling in response to agonists (Kniazeff et al., 2011;El Moustaine et al., 2012). The most studied example is the GABA B receptor, for which GABA B1 and GABA B2 functionally complement each other by forming the receptor heterodimer. Binding of GABA to the N-terminal extracellular domain (NTED) of GABA B1 results in allosteric transactivation of the GABA B2 protomer and subsequent G protein coupling to the activated 7 transmembrane domain (7TMD) of GABA B2 (Galvez et al., 2001;Duthey et al., 2002;Kniazeff et al., 2002). Moreover, GABA B1 requires heteromerization with GABA B2 to traffic to the cell surface (Margeta-Mitrovic et al., 2000;Pagano et al., 2001;Brock et al., 2005). Heteromerization of the taste 1 This work is supported by the Spanish Ministry of Economy and Competitiveness (MINECO) [SAF2014-57138-C2-1-R] to Marián Castro. dx.doi.org/10.1124/mol.115.099440. receptor (T1R) 3 with T1R 1 or T1R 2 results in the sensation of umami taste or sweeteners, respectively (Zhao et al., 2003). Indeed, knockout of the individual T1R 1 -, T1R 2 -, or T1R 3encoding genes in mice largely attenuates umami, sweet, or both flavors, respectively (Damak et al., 2003;Zhao et al., 2003). Dimerization of purified metabotropic glutamate (mGlu) 2 receptors in nanodiscs is required for G protein activation in response to the endogenous agonist glutamate (El Moustaine et al., 2012). In contrast, refolding of purified class A GPCRs in detergent micelles or nanodiscs revealed that monomers are fully able to bind their cognate ligand, activate G proteins, and/or recruit G protein-coupled receptor kinases and arrestins (Bayburt et al., 2007;Hanson et al., 2007;White et al., 2007;Whorton et al., 2007Whorton et al., , 2008; Kuszak et al., 2009;Arcemisbéhère et al., 2010;Tsukamoto et al., 2010;Bayburt et al., 2011). In addition, the class B parathyroid hormone receptor activates G proteins when expressed as monomers, as observed using dimer-disrupting mutations (Pioszak et al., 2010). Hence, dimerization is not required for tested class A and B GPCRs to transduce agonist-induced intracellular signaling. The question of why do these GPCRs dimerize and oligomerize obviously rises. Physically interacting GPCRs may modulate each other's activities. However, unambiguous discrimination between GPCR crosstalk as a consequence of receptor heteromerization and that resulting from their intracellular signaling events (Schmidlin et al., 2002;Vázquez-Prado et al., 2003;Natarajan et al., 2006;Kelly et al., 2008;Rives et al., 2009;Nijmeijer et al., 2010) is experimentally challenging and requires experimental perturbation of these Fig. 1. Detection of GPCR dimers. Antibodies targeting epitope-tagged (depicted) or native GPCRs specifically label receptor populations to subsequently allow detection of physical interaction by coimmunoprecipitation (co-IP) (A) or close proximity using in situ proximity ligation of DNAconjugated secondary antibodies (B) and time-resolved (TR) FRET between fluorophore-conjugated antibodies (C). Time-resolved FRET between fluorescent ligands allows detection of receptor complexes in native tissue (D). Genetic fusion of luminescent or fluorescent donor and acceptor proteins or nonfunctional fragments of these proteins to the C-terminal tail of GPCRs allows detection of close proximity by RET (E) and bimolecular fluorescence (BiFC) or luminescence (BiLC) complementation (F). Heteromer identification technology (HIT): RET between b-arrestin and one GPCR subtype that is fused to donor or acceptor proteins upon agonist activation of another untagged GPCR subtype indicates that both GPCRs are in close proximity (G). CODA-RET detects the interaction of GPCR dimers with intracellular signaling partners using the combination of bimolecular fluorescence complementation with sequential FRET (H).
GPCRs to form heteromers. Moreover, in situ validation of GPCR heteromerization and their specific functional properties in native tissues is difficult and consequently often lacking (Pin et al., 2007). Hence, the physiologic relevance of many identified GPCR heteromers remains a topic of debate, not least because recent methodologies shed distinct light on the size and stability of GPCR heteromers (Lambert and Javitch, 2014). In this review, we will first focus on recent developments to determine the size, stability, and proximal signaling of GPCR complexes and secondly on evidence for GPCR heteromerization in native tissue.

Proportion, Size, and Stability of GPCR Dimers and Oligomers
GPCR oligomerization has been the subject of significant research over the last two decades by using a number of biochemical and biophysical approaches mostly involving engineered GPCR constructs, among other approaches. Hence, receptors harboring N-terminal epitope (e.g., hemagglutinin, FLAG, or cMyc) and/or SNAP/CLIP/Halo tags can be detected using specific high-affinity antibodies and covalent labeling, respectively. Subsequent coimmunoprecipitation and time-resolved FRET reveals the physical association and close proximity of GPCRs in heterologous cells (Fig. 1, A-C) (Milligan and Bouvier, 2005;Maurel et al., 2008;Faklaris et al., 2015). Moreover, fusion of bioluminescent, fluorescent proteins, or nonfunctional fragments of these proteins to the C-terminal tail of GPCRs allows close-proximity detection of GPCR dimers and/or oligomers in living cells using bioluminescence resonance energy transfer (BRET), fluorescence resonance energy transfer (FRET), or bimolecular complementation, respectively (Fig. 1,E and F), and approaches such as the heteromer identification technology or complemented donor-acceptor resonance energy transfer (CODA-RET) have also been implemented for the investigation of complexes involving three interacting partners (Fig. 1, G and H) (Ciruela et al., 2010;Cottet et al., 2012;Kaczor et al., 2014).
Although most of these techniques have retrieved qualitative information on the formation of GPCR dimers, the proportion of GPCRs that are engaged in dimers and/or oligomers as well as the stoichiometry and stability of such complexes are generally not well characterized and require carefully controlled quantitative measurements. Saturation FRET between purified b 2 -adrenergic receptors (b 2 -ARs) site-specifically labeled with fluorophores and reconstituted in lipid bilayers suggested the predominant formation of tetramers (Fung et al., 2009). Threecolor sequential BRET-FRET and bimolecular luminescence/ fluorescence complementation in combination with resonance energy transfer (RET) revealed the formation of GPCR heteromultimers, consisting of at least three or four individual GPCRs, when expressed at physiologic levels (Lopez-Gimenez et al., 2007;Carriba et al., 2008;Guo et al., 2008;Nijmeijer et al., 2010;Armando et al., 2014). However, the interpretation of quantitative RET approaches between membrane-associated proteins has been challenged (James et al., 2006;Lambert and Javitch, 2014;Lan et al., 2015), and revised experimental designs (Szalai et al., 2014) as well as third-party RET approaches (Kuravi et al., 2010) have been proposed to improve the interpretation of results from RET experiments.
In this context, intensity-based FRET approaches (i.e., those based on measurements of emission intensity of the fluorophores rather than of their fluorescence lifetimes), where spectral datasets are acquired, can provide quantitative information not only of the apparent FRET efficiency of a sample, but also on donor-acceptor stoichiometry of their interactions. These methods can be combined with spectral imaging microscopy for spatial resolution and have allowed investigation of the quaternary structure of GPCRs in a more quantitative manner than other previously employed nonspectral intensity-based FRET approaches based on average measurements of apparent FRET efficiency Raicu and Singh, 2013). This is possible because spectrally resolved FRET approaches allow the accurate measurement of concentrations of donors and acceptors with overlapping emission spectra. Contaminations of the FRET signal as a consequence of the donor's bleedthrough or direct acceptor excitation are corrected in spectrally resolved FRET, and possible contributions of an unpaired donor and acceptors in the sample are taken into account by applying specific algorithms. Some drawbacks of these techniques can be the instrumentation requirements, such as spectral imaging detectors for certain applications, or the requirement of donor and acceptor reference samples of a known concentration (i.e., purified fluorescent proteins) for spectral unmixing. Different spectrally resolved FRET approaches revealed the proportion of serotonin 5-hydroxytryptamine (5-HT) 1A receptors that were engaged in oligomers in transfected cells (Gorinski et al., 2012) as well as the formation of transient tetramers by stable M 3 muscarinic acetylcholine receptor dimers at the cell surface, with minimal interference from a bystander RET signal coming from nearby noninteracting partners (Patowary et al., 2013).
Fluorescence correlation spectroscopy (FCS) allows the detection of fluctuations in fluorescent intensity that result from the diffusion of fluorescent molecules in and out of an open, diffraction-limited, observation volume. Suitable for extracting two-dimensional information on membrane protein dynamics with submicrosecond temporal resolution, it constitutes a powerful approach to monitor the diffusion of GPCRs in the plasma membrane of single living cells in real time. To this end, receptors fused to a particular fluorescent protein are heterologously expressed in cells (Herrick-Davis et al., 2012Teichmann et al., 2014). As a variant, the expression of receptors fused to complementary fragments of a fluorescent protein will allow the exclusive detection of diffusing receptor complexes that become fluorescent upon bimolecular fluorescence complementation . FCS measurements achieve single molecule sensitivity and are more accurately performed in samples with very low expression levels, such as those found for some GPCRs in native cells. Hence, endogenous GPCRs can be labeled with fluorescent ligands (Cordeaux et al., 2008;Corriden et al., 2014) or specific antireceptor fragment antigen binding proteins fused to fluorescent moieties (Herrick-Davis et al., 2015) for FCS measurements (Fig. 2). In the latter case and due to the monovalent nature of the fragment antigen binding fragments opposite to antibodies, potential artifacts due to antibody-induced clustering of receptors are avoided. FCS can give information on a protein that is likely to be in the same complex based on the changes in the diffusion coefficients by the formation of larger codiffusing entities and on the brightness of the diffusing particles. However, it should be kept in mind that FCS measures codiffusing proteins that are situated within the same microdomain but do not provide proof for direct physical protein-protein interaction. In addition, due to the detection limit, FCS is not able to discriminate between diffusion times of molecules with similar masses. At least a difference by a factor of four in the molecular mass of the particles is required to resolve their diffusion times by FCS (Meseth et al., 1999). This poses a limitation of FCS to distinguish among GPCR monomers, dimers, or tetramers on the basis of diffusion times alone. Moreover, factors such as heterogeneity of membrane viscosity in microdomains or the interaction of receptors with other signaling or scaffolding proteins might affect diffusion times. Autocorrelation analysis of FCS measurements reveals the average number of molecules within the observation volume and informs on the number of diffusing entities but not on the number of fluorescent protomers within diffusing complexes. Combining FCS with a photon counting histogram analysis enables quantification of the intensity of fluorescence fluctuations, which provides information with regards to the stoichiometry of protein complexes ( Fig. 2C) (Chen et al., 1999). By doing this, the molecular brightness of the diffusing particles and consequently the number of fluorophores codiffusing in each particle can be determined. Subsequent comparison with appropriate reference proteins with a known monomeric/oligomeric status allows quantification of the oligomerization state of proteins of interest (Herrick-Davis et al., 2012. These combined techniques suggest the existence of native serotonin 5-HT 2C receptor homomeric complexes in the apical membrane of living choroid plexus epithelial cells (Herrick-Davis et al., 2015). For some class A and B GPCRs, an equilibrium between monomeric and possible dimeric species was found, with a low proportion of dimers, by using fluorescence crosscorrelation spectroscopy (Teichmann et al., 2014), a development of FCS for dual-color applications that allows discrimination of single-labeled entities carrying two different fluorophores from the duallabeled diffusing particles resulting from their interaction. In this approach, the emission of the two fluorophores are separately registered by two different detectors, and when the two differently labeled partners comigrate, the fluctuations in the intensity of the two fluorophores will correlate as they diffuse into and out of the observation volume together. The crosscorrelation function provides more sensitive information than the diffusion coefficient alone for the detection of interactions. The information gathered in the work of Teichmann et al. (2014) confirmed single molecule total internal reflection fluorescence microscopy (TIRFM) studies at b 1 -AR or M 1 muscarinic acetylcholine receptors (M 1 Rs) (Hern et al., 2010;Calebiro et al., 2013). However, a homogeneous population of dimers without evidence of the coexistence of the monomeric species in equilibrium was found in the case of serotonin 5-HT 2C and also b 1 -AR or M 1 R in other FCS/photon-counting histogram-based studies (Herrick-Davis et al., 2012. These discrepancies might be related to different receptor expression levels in the cellular models employed, where very low expression levels of the receptors compatible with single molecule resolution might favor the occurrence of monomeric species in dynamic equilibrium with the dimer population. Recently, super-resolution dual-color photoactivation localization microscopy using photoactivatable dyephotoactivated localization microscopy allowed imaging of the spatial arrangement of individual GPCR molecules in dimers and oligomers at the plasma membrane with a resolution of ∼8 nm. To this end, hemagglutinin-and FLAGtagged luteinizing hormone receptor (LHR) mutants that were either ligand binding (LHR B2 ) or signaling (LHR S2 ) deficient (Rivero-Müller et al., 2010) were expressed in human embryonic kidney 293 cells and specifically labeled with antibodies conjugated with CAGE 552 and CAGE 500 photoswitchable dyes for dual-color visualization (Fig. 3A) (Jonas et al., 2015). The number and identity (either LHR B2 or LHR S2 ) of GPCRs within a 50-nm radius of each single GPCR are determined, and the irreversible activation and bleaching of the dyes allowed quantification of dimer and oligomer complexes (Fig. 3, B and C). Expression of wild-type LHR at levels in the physiologic range observed in the ovary and testis revealed a diverse organization in monomers, dimers (14.6%), and oligomers (26.8%, varying from trimers to oligomers consisting of more than nine protomers) (Fig. 3D). In this study, agonist stimulation did not elicit any change in either the percentage of associated molecules or relative proportions of dimers and oligomers (Jonas et al., 2015). Yet, in spite of the detailed information on the size and spatial arrangement of GPCRs in oligomeric complexes by photoactivatable dye-photoactivated localization microscopy, this technique involves the acquisition of a time series of images that needs cell fixation, and therefore it is not well suited for the investigation of complex stability and real-time dynamics.
The stability and dynamics of GPCR dimers and oligomers has been monitored by real-time single-molecule imaging and tracking using fluorescent ligands (Hern et al., 2010;Kasai et al., 2011) or SNAP-tagged GPCR constructs (Calebiro et al., 2013). In the former approach, one has to consider the number of ligands bound per dimer. If agonist binding of an agonist to one protomer prevents ligand binding to the associated protomer, then the dynamics of the dimer rather than monomers are observed (Albizu et al., 2010). Singlemolecule TIRFM studies reported the existence of an equilibrium between monomeric and oligomeric species for different GPCRs, with fast dynamics of less than 1 second (Hern et al., 2010;Kasai et al., 2011;Calebiro et al., 2013). In these approaches and differently to PALM, the x-y spatial resolution is limited by the optical diffraction limit (∼220 nm) so that the possibility of interpreting two monomers within the resolution limit as an apparent dimer must be considered. In particular, two-color TIRFM revealed the transient nature of M 1 R and N-formylpeptide receptor homodimers at the cell surface (Hern et al., 2010;Kasai et al., 2011). Yet, the stability of the interactions might differ between GPCR subtypes. For example, b 1 -AR displays more transient interactions than b 2 -AR, as revealed by both TIRFM and dual-color fluorescence recovery after photobleaching studies (Dorsch et al., 2009;Calebiro et al., 2013). Affinity-based corecruitment of differentially tagged mGlu receptors into forced microdomains on the cell surface confirmed the stable nature of class C GPCR dimerization, which is mediated by disulfide bridges between the NTED of the protomers (Gavalas et al., 2013). On the other hand, differentially tagged b 2 -AR or m-opioid receptor (MOR) fully segregated into distinct artificial microdomains, suggesting that homodimerization between these class A GPCRs is not sufficiently stable (Gavalas et al., 2013). Similar segregation was observed as a reduced BRET signal between wild-type and binding-deficient b 2 -AR upon agonist-induced internalization of only active (wild-type) b 2 -AR, whereas the inactive mutant remained at the cell surface (Lan et al., 2011). In contrast, coexpression of the wild-type b 2 -AR with receptors activated solely by a synthetic ligand (i.e. RASSL) b 2 -AR mutant revealed that agonist stimulation of either protomer induced internalization of the dimer (Sartania et al., 2007). Altogether, a discrepancy in the transient nature of at least some GPCR dimers is observed between single molecule labeling strategies and studies demonstrating cointernalization and cotrafficking of receptors (Milligan, 2010).
Several GPCRs have arranged as dimers or oligomers in recent high-resolution crystal structures, whereas others have crystallized as monomers. These crystal structures suggest that GPCRs can assemble in multiple ways, which might explain the possible formation of complexes larger than dimers. Antagonist-bound MOR crystallizes as oligomers, with a large Fig. 3. Dual-color photoactivatable dyes and localization microscopy (PD-PALM) to investigate spatial and structural organization of GPCR complexes. Epitope-tagged binding-deficient (LHR B2 ) and signalingdeficient (LHR S2 ) receptors were labeled with specific antibodies conjugated with CAGE 552 or 500 dyes, and cells were subsequently fixed (A). PD-PALM images were acquired using a total internal reflection fluorescence-equipped microscope by iterative cycles of photo-conversion of the CAGE dyes upon UV illumination and simultaneous dual-channel single-molecule imaging and irreversible photo-bleaching (B). Localization coordinates (x-y) were assigned to individual fluorescent particles detected in each imaged channel. Analysis of the number of associated protomers was conducted based on a recursive search for neighborhood particles within a radius of 50 nm from each individual protomer (C). In combination with molecular modeling studies, this approach allowed us to resolve the composition and spatial arrangement of each associated group of molecules (dimers/ oligomers) visualized by PD-PALM (D). The structural models, depicted from the extracellular side, are taken from Jonas et al. (2015) under Creative Commons Attribution Unported License to Author Choice articles. contact interface involving transmembrane (TM) 5/TM6 and a smaller interface formed by TM1/TM2/helix 8 (Manglik et al., 2012), whereas a ligand-free b 1 -AR is arranged as oligomers via TM4/TM5 and TM1/TM2/helix8 interfaces (Huang et al., 2013). The antagonist-bound k-opioid receptor is arranged as a dimer via a TM1, TM2, and helix 8 interface (Wu et al., 2012). CXCR4 crystallizes as dimers via a TM5/TM6 interface when occupied by antagonistic chemokine vMIPII or small-molecule (IT1t) and cyclic peptide (CVX15) antagonists (Wu et al., 2010;Qin et al., 2015), whereas antagonist-bound histamine H 1 receptor has a TM4 dimerization interface (Shimamura et al., 2011). However, it should be kept in mind that receptor modifications, bound ligand, and crystal packing conditions might affect (forced) dimerization interfaces and require systematic experimental validation using site-directed mutagenesis and/or interfering peptides. Indeed, TM1 forms the dimer interface in the mGlu 1 7TMD crystal structure, whereas a TM4/TM5 interface is actually observed in full-length mGlu receptors in cells (Xue et al., 2015).
Hence, observed discrepancies in both the quaternary structure and stability of GPCR complexes between different studies emphasize the need for further refinement and systematic comparison of methods to monitor GPCR interactions in time.

Proximal Evidence for GPCR Heteromer Specific Signaling
GPCR heteromerization can significantly affect signaling and/or trafficking characteristics of individual GPCR subtypes (Jordan and Devi, 1999). However, unambiguous separation of heteromer-specific signaling from downstream crosstalk is difficult and requires experimental disruption of dimer formation (Prezeau et al., 2010;Vischer et al., 2011). Supportive evidence for heteromer-specific signaling comes from RET-based detection of signaling events very proximal to GPCRs in transfected cells. For example, agonist stimulation of a given (unmodified) GPCR subtype changes BRET between another coexpressed GPCR subtype and specific G proteins and/or b-arrestins (Fig. 1G) (See et al., 2011;Mustafa et al., 2012;Watts et al., 2013;Jonas et al., 2015). In addition, fusion of GPCRs to bimolecular luminescence or fluorescence complementation protein fragments and subsequent coexpression with compatible G protein or b-arrestin RET fusion constructs (i.e., CODA-RET) allowed simultaneous detection of GPCR heteromerization and engagement of intracellular signaling partners upon agonist stimulation (Fig. 1H) Armando et al., 2014;Guitart et al., 2014;Bellot et al., 2015;Frederick et al., 2015). Heteromerization of G s -coupled D 1 R and the G i/o -coupled dopamine D 2 receptor (D 2 R) induces intracellular Ca 21 mobilization upon agonist activation, which could be impaired by the G q/11 inhibitor YM254890 Rashid et al., 2007). Moreover, agonist-induced recruitment of G q -green fluorescent protein to D 1 R-Renilla luciferase requires coexpression of D 2 R and could be inhibited by a membrane-permeable peptide that disrupts D 1 R-D 2 R heteromerization (Hasbi et al., 2014). In contrast, however, a recent study reported the lack of both G q recruitment to D 1 R-D 2 R heteromers in a CODA-RET assay and G q activation, as measured by BRET, between Ga q -Renilla luciferase 8 and Gg 2 -Venus (Frederick et al., 2015). Although these engineered biosensor-expressing cells can reveal the potential of GPCRs to modify each other's signaling upon heteromerization, it should be kept in mind that protein expression levels should be kept to a minimum to avoid nonspecific interactions as a consequence of membrane (microdomain) overcrowding.

Pharmacological Evidence for GPCR Dimers and Oligomers in Native Tissues
Ligand binding to one GPCR within a dimer or oligomer can rapidly change the conformation of an associated GPCR, as shown by the inhibition of norepinephrine-induced intramolecular FRET in an engineered a 2A -AR upon stimulation of the MOR with morphine (Vilardaga et al., 2008). This transconformational change is slightly faster than the rate for G protein activation, indicating direct allosterism between both receptor subtypes within the heteromer (Xue et al., 2015). Negative-binding cooperativity was observed in the 1970s for native b-AR and the thyroid-stimulating hormone receptor (TSHR) in membrane preparations of frog erythrocytes and human thyroid samples, respectively, already suggesting the existence of GPCR dimers (Limbird et al., 1975;Limbird and Lefkowitz, 1976;Powell-Jones et al., 1979).
Hence, binding cooperativity between ligands acting at different GPCRs suggests the existence of GPCR heteromers in native tissues. However, concerns have been raised on the interpretation of binding cooperativity since agonist binding to GPCRs can be G protein dependent (Chabre et al., 2009;Birdsall, 2010). Consequently, competition for a shared pool of G proteins between GPCRs could result in the observation of apparent negative agonist binding cooperativity. This G protein depletion is, in particular, possible in equilibrium competition binding on membrane preparations as G protein coupling to activated GPCRs might be almost irreversible in the absence of free GTP (Chabre et al., 2009) but can be prevented by overexpression of G proteins (Nijmeijer et al., 2010).
Importantly, differential spatiotemporal expression of GPCR subtypes in tissues or even disease states may result in distinct heteromer-specific pharmacology. This may be particularly challenging for drug discovery programs that generally measure drug activity at a single target, and such initial screens may consequently not be adequate predictors for the in vivo effectiveness of drugs.

Detection of GPCR Dimers and Oligomers in Native Tissues
Although recombinant technologies using engineered GPCRs provide supportive evidence that many GPCRs might exist as dimers and/or oligomers or at least exist in close proximity, these approaches are not easily applicable to identify GPCR complexes in native tissues. However, in a recent study, double knock-in mice expressing the MOR and d-opioid receptor (DOR) fused in frame to mCherry and enhanced green fluorescent protein (eGFP), respectively, showed colocalization of these receptors in the midbrain and hindbrain (Erbs et al., 2015). Coimmunoprecipitation using anti-mCherry and anti-eGFP antibodies revealed that MOR-mCherry and DOR-eGFP belong to the same complex in the hippocampus from these mice, which confirms earlier detection of these MOR-DOR complexes in this brain region using a heteromer-specific antibody (Gupta et al., 2010). Interestingly, this antibody revealed increasing MOR-DOR complex levels in the cortex of animals chronically treated with morphine, whereas no coexpression of MOR-mCherry and DOR-eGFP was observed in the cortex of double knock-in mice in the absence or presence of the DOR agonist (Erbs et al., 2015). This discrepancy might be related to the difference in the receptor expression level and/or detection sensitivity.
Hitherto, only a limited number of the GPCR dimers and oligomers that were identified in heterologous cells have been validated in native tissues due to the absence of highly specific antibodies for most GPCR subtypes or heteromers (Michel et al., 2009). Indeed, coimmunoprecipitation of GPCR complexes from solubilized native tissues has been used for nearly two decades (Fig. 1A) (Kaupmann et al., 1998;González-Maeso et al., 2008;Pei et al., 2010) but requires critical analysis to ensure that physically interacting GPCRs are detected rather than aggregation artifacts due to the hydrophobic nature of GPCRs (Milligan and Bouvier, 2005). More recently, in situ GPCR complexes have also been detected using immunohistochemical antibodies. Labeling of native GPCR subtypes with specific primary antibodies followed by matching secondary antibodies that are conjugated to unique oligonucleotide sequences allows enzymatic ligation of these DNA strands if secondary antibodies are in close proximity (,16 nm), which corresponds to a theoretical distance of ,40 nm between the GPCR subtypes (i.e., epitopes). The formed circle DNA strand is subsequently amplified and hybridized with fluorescent complementary oligonucleotide probes for high-sensitivity fluorescence microscopy analysis (Fig. 1B) (Weibrecht et al., 2010). This socalled proximity ligation assay (PLA) confirmed the close proximity of various GPCR pairs in the central nervous system. In situ PLA between D 2 R and adenosine A 2A receptor (A 2A R) in the striatum of mice, rats, and monkeys (Trifilieff et al., 2011;Bonaventura et al., 2014;Fernández-Dueñas et al., 2015) confirmed D 2 R-A 2A R heteromer coimmunoprecipitation from the rat striatum (Cabello et al., 2009) as well as detection by coimmunoprecipitation, pull-down, FRET, BRET, and sequential BRET-FRET in heterologous cells (Canals et al., 2003;Kamiya et al., 2003;Cabello et al., 2009). Interestingly, in the striatum of an experimental Parkinsonism rat model, this PLA signal was significantly decreased as a consequence of reduced codistribution and proximity between D 2 R and A 2A R (Fernández-Dueñas et al., 2015). In addition, PLA was detected between native cannabinoid CB 1 and CB 2 receptors in the rat brain pineal gland and nucleus accumbens (Callén et al., 2012) and between D 2 R and oxytocin receptors in the rat dorsal striatum and the neuropil of nucleus accumbens (Romero-Fernandez et al., 2013). Interestingly, PLA between the dopamine D 4 receptor and a 1B -or b 1 -AR was observed in a rat pineal gland dissected 1 hour after sunrise, but not when pineal glands were isolated at sunset, which reflects the circadian variation in dopamine D 4 receptor expression levels (González et al., 2012). Dopamine inhibits adrenergic receptor signaling within these heteromers and consequently limits serotonin and melatonin synthesis and release in the pineal gland. Dopamine D 1 receptor (D 1 R) and D 2 R form heteromers in heterologous cells (So et al., 2005;Frederick et al., 2015). D 1 R/D 2 R heteromers have been proposed to play an important role in various neuropsychiatric disorders. However, PLA was absent in the shell of the nucleus accumbens, despite coexpression of native D 1 R and D 2 R in these cells, but readily observed upon D 1 R/D 2 R overexpression using viral gene transfer (Frederick et al., 2015). On the other hand, native D 1 R/D 2 R heteromers were detected in situ in rat striatal neurons using antibody-based confocal FRET analysis (Fig. 1C) (Hasbi et al., 2009(Hasbi et al., , 2014Perreault et al., 2010;Verma et al., 2010). The discrepancy between these antibody-based techniques requires further systematic comparison on similar regions. Importantly, specificity of D 1 R-and D 2 R-primary antibodies in these studies was confirmed in cells heterologously expressing dopamine receptor subtypes and in at ASPET Journals on June 13, 2021 molpharm.aspetjournals.org D 1 R or D 2 R knockout mice in situ Perreault et al., 2010).
In addition to antibodies, fluorescent ligands have been used to detect native GPCR heteromers (Fig. 1D). For example, ghrelin receptor heteromerization with D 2 R was shown in a mice hypothalamus by confocal FRET and time-resolved FRET between fluorescently labeled agonist ghrelin and D 2 R antibody-secondary antibody complexes, with the latter conjugated to Cy3 fluophore or cryptate, respectively (Kern et al., 2012). TrFRET between the D 2 R and A 2A R antagonists that were conjugated to Lumi4-Terbium and a red acceptor (dy647), respectively, confirmed the PLA signal in the rat striatum (Fernández-Dueñas et al., 2015). Likewise, native oxytocin receptor dimers were detected in the mammary gland using fluorescent antagonists, but to a much lesser extent with fluorescent agonists due to negative-binding cooperativity between agonists (Albizu et al., 2010).

Functional Evidence for GPCR Dimers in Native Tissue
Coexpression of two nonfunctional GPCRs to form a functional receptor provides convincing evidence for GPCR heteromerization, as exemplified by the functional complementation of native class C GPCRs GABA B1 and GABA B2 , which is strictly required for cell-surface expression of a functional GABA B receptor in vitro and in vivo (Prosser et al., 2001;Gassmann et al., 2004). The a 1D -AR is retained in the endoplasmic reticulum when individually expressed in heterologous cells. Systematic coexpression with 28 other class A GPCRs revealed that heteromerization with a 1B -or b 2 -AR is required for the cell surface of a 1D -AR (Hague et al., 2004;Uberti et al., 2005). However, a 1D -AR stimulates the contraction of carotid arteries in a 1B -AR knockout mice (Deighan et al., 2005), suggesting that effective trafficking of a 1D -AR to the cell surface is mediated by, for example, native b 2 -AR (Pernomian et al., 2013).
In contrast to the majority of class A GPCRs, the folliclestimulating hormone receptor (FSHR), TSHR, and LHR are characterized by a large leucine-rich repeat-containing NTED, which is exclusively involved in the selective and high-affinity binding of their corresponding glycoprotein hormones (Osuga et al., 1997;Vischer et al., 2003a,b;Fan and Hendrickson, 2005). All three glycoprotein hormone receptors form homomers and heteromers in transfected cells (Urizar et al., 2005;Feng et al., 2013). However, only heteromerization between FSHR-LHR might be physiologically relevant as both receptors are shortly coexpressed in granulosa cells during follicle maturation (Thiruppathi et al., 2001). Taking advantage of the modular architecture of these glycoprotein hormone receptors, nonfunctional LHR mutants were created to impair either hormone binding to the NTED (i.e., LHR B2 ) or G protein activation by the 7TMD (i.e., LHR S2 ) (Osuga et al., 1997;Ji et al., 2002;Lee et al., 2002). Coexpression of LHR B2 with LHR S2 in transfected cells rescued hormone-induced cAMP production, suggesting that both nonfunctional constructs are at least organized as dimers. Similar functional complementation was observed in transfected cells coexpressing FSHR B2 and FSHR S2 (Ji et al., 2004) and TSHR B2 and TSHR S2 (Urizar et al., 2005). Moreover, coexpression of LHR B2 with LHR S2 in LHR knockout mice using a bacterial artificial chromosome to preserve correct spatiotemporal expression rescued both gonadal development and full spermatogenesis (Rivero-Müller et al., 2010). In contrast, expression of the individual loss-offunction mutants was ineffective. However, the same LHR B2 mutant (i.e., C 22 A) was more recently reported to induce some cAMP signaling in transiently transfected human embryonic kidney 293 cells , which contrasts with earlier in vitro and in vivo observations Lee et al., 2002;Rivero-Müller et al., 2010).

Perturbation of GPCR Heteromers in Native Tissue
D 1 R/D 2 R heteromerization involves the C-tail of D 1 R and intracellular loop 3 (IL3) of the long D 2 R isoform. Membranepermeable fusion constructs consisting of a TAT sequence and D 1 R G 396 -L 413 (TAT-D1 C ) or D 2 R M 257 -E 271 (TAT-D2L IL3 ) peptides disrupted D 1 R/D 2 R complexes and heteromerinduced signaling in striatal neurons (Pei et al., 2010;Hasbi et al., 2014). D 1 R/D 2 R coimmunoprecipitation was enhanced from brain tissue derived from patients suffering major depression as compared with healthy persons (Pei et al., 2010). Importantly, TAT-D2L IL3 has antidepressant-like effects when injected in the brain of rats, as revealed by increased mobility in forced swim tests and reduced escape failures in learned helplessness tests, leading the authors to suggest a prominent role for D 1 R/D 2 R heteromers in this neurologic disorder (Pei et al., 2010). The class A angiotensin receptor type 1a (AT1aR) and class B secretin receptor (SCTR) are coexpressed in osmoregulatory brain centers and form heteromeric complexes in heterologous cells. AT1aR/SCTR heteromerization was specifically inhibited by peptides derived from AT1aR-TM1 and SCTR-TM2, whereas both homomerization and heteromerization were inhibited by AT1aR-TM4 and SCTR-TM4 (Lee et al., 2014). Injection of AT1aR-TM1 in mice brains reduced hyperosmolality-induced drinking, confirming the physiologic role of this class A/class B GPCR heteromer in the regulation of water homeostasis.

Conclusions
Dimerization is essential for class C GPCR functioning, whereas class A and B GPCRs can activate G proteins and recruit b-arrestins as monomers in response to agonists. Nevertheless, most tested GPCRs form dimers and oligomers in heterologous cells, resulting in an apparent plethora of functional consequences. However, the stability as well as the stoichiometry of GPCR complexes appear to vary considerably, with only class C GPCRs forming stable complexes. Hitherto, only a small percentage of GPCR dimers and oligomers have been validated in native tissues, despite the guidelines proposed by the International Union of Basic and Clinical Pharmacology in 2007 (Pin et al., 2007). Ex vivo and in vivo detection of native GPCR dimers largely rely on the availability of specific antibodies and/or fluorescent ligands. Recent progress in the generation of llama-derived nanobodies targeting GPCRs might facilitate future detection of native GPCR dimers and oligomers in situ. The higher affinities of engineered bivalent and biparatopic nanobodies in comparison with their monovalent counterparts might suggest binding to receptor dimers, although experimental proof for these observations is still required (Jähnichen et al., 2010;Maussang et al., 2013). Binding cooperativity might be a feasible pharmacological approach to detect GPCR homomers and heteromers in ex vivo samples; however, influence from signaling molecules should be considered if using agonists, whereas both the absence and presence of binding cooperativity has been observed for antagonists. In vivo functional complementation has so far been performed for one class A GPCR subtype and requires well characterized mutants that are present at physiologic levels with correct spatio-temporal expression patterns. Finally, confirmed disruption of heteromerization using interfering peptides followed by changes in the phenotypical response provided evidence for the presence as well as (patho)-physiologic function of some GPCR heteromers. Translation of in vitro observations for more GPCR heteromers to native tissues is required in the near future to confirm that GPCR dimers and oligomers exist beyond engineered model systems.