Interactions between Intracellular Domains as Key Determinants of the Quaternary Structure and Function of Receptor Heteromers*

G protein-coupled receptor (GPCR) heteromers are macromolecular complexes with unique functional properties different from those of its individual protomers. Little is known about what determines the quaternary structure of GPCR heteromers resulting in their unique functional properties. In this study, using resonance energy transfer techniques in experiments with mutated receptors, we provide for the first time clear evidence for a key role of intracellular domains in the determination of the quaternary structure of GPCR heteromers between adenosine A2A, cannabinoid CB1, and dopamine D2 receptors. In these interactions, arginine-rich epitopes form salt bridges with phosphorylated serine or threonine residues from CK1/2 consensus sites. Each receptor (A2A, CB1, and D2) was found to include two evolutionarily conserved intracellular domains to establish selective electrostatic interactions with intracellular domains of the other two receptors, indicating that these particular electrostatic interactions constitute a general mechanism for receptor heteromerization. Mutation experiments indicated that the interactions of the intracellular domains of the CB1 receptor with A2A and D2 receptors are fundamental for the correct formation of the quaternary structure needed for the function (MAPK signaling) of the A2A-CB1-D2 receptor heteromers. Analysis of MAPK signaling in striatal slices of CB1 receptor KO mice and wild-type littermates supported the existence of A1-CB1-D2 receptor heteromer in the brain. These findings allowed us to propose the first molecular model of the quaternary structure of a receptor heteromultimer.

Receptor heteromers are the focus of intense research, as through heteromerization receptors become unique functional entities with different properties from those of either receptor when not engaged in heteromerization resulting in new therapeutic targets (1)(2)(3)(4). Their unique properties provide a "biochemical fingerprint" thus allowing their identification in native tissues (1,3). There is already a long list of discovered heteromers with two different G protein-coupled receptors (GPCRs) 2 (2,4). Furthermore, we recently obtained evidence for the existence of receptor heteromultimers, i.e. heteromers including more than two different receptors, and reported on heteromers, including the GPCRs adenosine A 2A , cannabinoid CB 1 , and dopamine D 2 receptors, in transfected cells (5). Evidence of GPCR homomultimers has also been recently demonstrated in living cells (6,7). Many important questions regarding receptor heteromers and heteromultimers remain unanswered. What is the arrangement of their GPCR units (quaternary structure)? What are the molecular determinants of their quaternary structure? Last but not least, what is their functional significance in native tissues?
It was inferred that different molecular mechanisms were involved in GPCR homo-and heteromerization. For family C GPCRs, disulfide bonds between extracellular domains as well as coiled-coil interactions between C-terminal domains seem to be necessary for the formation of functional homomeric or heteromeric receptors (8). For oligomerization of family A GPCRs, the helical transmembrane (TM) domains seem to be particularly important (7, 9 -15). In this study, by using mutated A 2A , CB 1 , and D 2 receptors, we investigated the relevance of electrostatic interactions (16) between intracellular domains in the determination of the quaternary structure of GPCR heteromers between A 2A , CB 1 , and D 2 receptors. Our initial goal was to obtain evidence for multiple intracellular interactions in the A 2A -CB 1 -D 2 receptor heteromultimer. Significantly, the same intracellular domains involved in A 2A -CB 1 -D 2 receptor heteromultimerization were also involved in A 2A -D 2 , A 2A -CB 1 , and CB 1 -D 2 receptor heteromerization. A three-dimensional model of the quaternary structure of the receptor heteromultimer was obtained by using the information from resonance energy transfer between A 2A , CB 1 , and D 2 receptors in the receptor heteromultimer. Furthermore, a biochemical property of the receptor heteromultimer was found to be dependent on its correct quaternary structure, determined by the intracellular electrostatic interactions, which allowed its identification in rodent brain tissue.
Mutant Receptors-Ser 374 in the C-terminal domain of the human A 2A receptor was mutated to Ala to obtain the A 2A A374 receptor. The sequence 199 RIFLAARRQ 207 (boldface letters indicate the amino acid involved in the interaction between the receptors and the residues that were mutated) in the cytoplasm at the end of TM5 of human A 2A receptor was mutated to 199 RIFLAAAAQ 207 to obtain the A 2A A205-A206 receptor. The sequence 462 SVSTDTSAE 470 in the C-terminal domain of human CB1 receptor was mutated to 462 SVSTDAAAE 470 to obtain the CB 1 A467-A468 receptor. The sequence 321 TSEDGKVQVT 330 in the third intracellular loop of human CB 1 receptor was mutated to 321 AAEDGKVQVT 330 to obtain CB 1 A321-A322 receptor. Mutations were performed by site-directed mutagenesis (Cellogenetics, Ijamsville, MD).
Fusion Proteins and Expression Vectors-The human cDNAs of the A 2A , CB 1 , and the mutant versions of these receptors or the human D 2 , D 2S , and D 4.4 receptors, cloned in pcDNA3.1, were amplified without their stop codons using sense and antisense primers harboring unique EcoRI and BamHI sites to clone A 2A , A 2A A374 , and A 2A A205-A206 receptors in the Rluc corresponding vector, EcoRI and KpnI to clone D 2 and D 2S receptors in the GFP 2 corresponding vector, BamHI and EcoRI to clone CB 1 , CB 1 A467-A468 , and CB 1 A321-A322 in the enhanced YFP corresponding vector, and XhoI and BamHI sites to clone D 4.4 receptor in the Rluc corresponding vector. The amplified fragments were subcloned to be in-frame into restriction sites of the multiple cloning sites of pcDNA3.1-Rluc, pGFP 2 -N3(h), pEYFP-N1 (Clontech) to give the plasmids corresponding to A 2A -Rluc, A 2A A374 -Rluc A 2A A205-A206 -Rluc, D 4 -Rluc, D 2 -GFP 2 , D 2S -GFP 2 , CB 1 -YFP, CB 1 A467-A468 -YFP, and CB 1 A321-A322 -YFP receptor fusion proteins. The cDNA of the 5HT 2B -YFP fusion protein was kindly provided by Dr. Irma Nardi (University of Pisa, Italy). Under these conditions, the fusion proteins are expressed at the membrane level, are not strongly overexpressed, and are quantitatively expressed in similar amounts (5).

Transient Transfection and Sample Preparation-HEK-
293T or CHO cells growing in 6-well dishes were transiently transfected with the corresponding fusion protein cDNA by the polyethyleneimine method (Sigma). Cells were incubated (4 h) with the corresponding cDNA together with polyethyleneimine (5.47 mM in nitrogen residues) and 150 mM NaCl in a serum-starved medium. After 4 h, the medium was changed to a fresh complete culture medium. Forty eight hours after transfection, cells were washed twice in quick succession in Hanks' balanced salt solution with 10 mM glucose, detached, and resuspended in the same buffer containing 1 mM EDTA. To control the cell number, sample protein concentration was determined using a Bradford assay kit (Bio-Rad) using bovine serum albumin dilutions as standards. Cell suspension (20 g of protein) was distributed into 96-well microplates; black plates with transparent bottom were used for FRET and fluorescence determinations, and white plates with white bottom were used for BRET and SRET experiments.
BRET Experiments-HEK-293T cells expressing the corresponding donor (receptor Rluc) and increasing amounts of the corresponding acceptor (receptor GFP 2 for BRET 2 or receptor YFP for BRET 1 ), as indicated in figure legends, were used. With aliquots of transfected cells (20 g of protein), three different determinations were performed in parallel. (i) To quantify fluorescence proteins expression, cells were distributed in 96-well microplates (black plates with transparent bottom), and fluorescence was read in a Fluostar Optima Fluorimeter (BMG Labtechnologies, Offenburg, Germany) equipped with a high energy xenon flash lamp, using an excitation filter at 410 nm for receptor GFP 2 reading (BRET 2 ) or 485 nm for receptor YFP reading (BRET 1 ), and emission was detected using filters at 510 nm (for GFP 2 ) or 530 nm (for YFP). Receptor fluorescence expression was determined as fluorescence of the sample minus the fluorescence of cells expressing receptor Rluc alone. (ii) For BRET 2 and BRET 1 measurements, the equivalent of 20 g of cell suspension was distributed in 96-well microplates (Corning 3600, white plates with white bottom), and 5 M DeepBlueC (BRET 2 ) or coelenterazine H (BRET 1 ) (Molecular Probes, Eugene, OR) was added. For BRET 2 experiments, readings were collected immediately (ϳ30 s) after addition of DeepBlueC using a Mithras LB 940 (Berthold Technologies, DLReady, Germany) that allows the integration of the signals detected in the short wavelength filter at 410 nm and the long wavelength filter at 510 nm. In BRET 1 after 1 min of adding coelenterazine H, the readings were collected using a Mithras LB 940 that allows the integration of the signals detected in the short wavelength filter at 485 nm and the long wavelength filter at 530 nm. (iii) To quantify receptor Rluc expression, luminescence readings were performed after 10 min of adding 5 M coelenterazine H. The net BRET is defined as ((long wavelength emission)/(short wavelength emission)) Ϫ Cf, where Cf corresponds to ((long wavelength emission)/(short wavelength emission)) for the Rluc construct expressed alone in the same experiment.
FRET Experiments-HEK-293T cells expressing the corresponding donor (receptor GFP 2 ) and increasing amounts of the corresponding acceptor (receptor YFP), as indicated in figure legends, were used. Using aliquots of transfected cells (20 g of protein), two different determinations were performed in par-allel. (i) To quantify YFP fluorescence, due to receptor YFP expression, the same procedure as described for BRET experiments was used. (ii) For FRET measurements, the equivalent of 20 g of cell suspension was distributed into 96-well microplates (black plates with a transparent bottom) and read in a Fluostar Optima fluorimeter equipped with a high energy xenon flash lamp, using an excitation filter at 410 nm and an emission filters at 510 nm (Ch x ) and 530 nm (Ch y ). Gain settings were identical for all experiments to keep the relative contribution of the fluorophores to the detection channels constant for spectral unmixing. The contribution of GFP 2 and YFP proteins alone to the two detection channels (spectral signature (17)) was measured in experiments with cells expressing only one of these proteins and normalized to the sum of the signal obtained in the two detection channels. The spectral signatures of the different receptors fused to either GFP 2 or YFP did not vary significantly from the determined spectral signatures of the fluorescent proteins alone. In determinations i and ii, linear unmixing was done taking into account the spectral signature as described by Zimmermann et al. (17) to separate the two emission spectra. For quantitation of the fluorescence emitted by each of two individual fluorophores (FluoA corresponding to the donor and FluoB corresponding to the acceptor) in FRET experiments, Equation 1 was applied, where Ch x and Ch y represent the signals in detection channels x and y, and A x , B x and A y , B x represent the normalized contributions of FluoA or FluoB to channels x and y, as they are known from the spectral signatures of the fluorescent proteins.
Sequential Resonance Energy Transfer (SRET) Experiments-The recently introduced sequential BRET-FRET (SRET) technique (5) not only allows the demonstration of heteromerization of three proteins but can also provide information about the quaternary structure of the heterotrimeric complex. By transfecting three receptors separately fused to Rluc, GFP 2 , and YFP, the detection of the SRET 2 signal demonstrates the physical interactions between the three receptors. In SRET 2 , the oxidation of the Rluc substrate DeepBlueC triggers GFP 2 excitation (BRET 2 ), which triggers a subsequent excitation of YFP (FRET) (see Fig. 1). Emission of YFP after addition of DeepBlueC is only possible if the three fusion proteins are in close proximity (Ͻ10 nm), allowing bioluminescent and fluorescent SRET to occur. For SRET experiments, HEK-293T cells were transiently co-transfected with the indicated amounts of plasmid cDNAs corresponding to receptor Rluc, receptor GFP 2 , and receptor YFP (see figure legends). In the experiments without casein kinase 1/2 inhibitors, cells were used 48 h posttransfection. When using casein kinase 1/2 inhibitors, cells were treated with casein kinase 1 inhibitor IC 261 (50 M; Cal-biochem) and casein kinase 2 inhibitor TBAC (10 M; Calbiochem) 4 h after transfection, and after 24 h, the medium was changed to a fresh complete culture medium containing the same amount of inhibitors, and cells were used 48 h post-transfection. Using aliquots of transfected cells (20 g of protein), different determinations were performed in parallel. (i) Quantification of protein-YFP expression was performed as indicated in FRET experiments. The sample fluorescence is the fluorescence calculated as described minus the fluorescence of cells expressing only protein-Rluc and protein-GFP 2 . (ii) Quantification of protein-Rluc expression was by determination of the luminescence due to protein-Rluc. Cells were distributed in 96-well microplates (Corning 3600, white plates with white bottom), and luminescence was determined 10 min after addition of 5 M coelenterazine H in a Mithras LB 940 multimode reader. (iii) BRET and FRET were combined to generate a technique called sequential BRET-FRET (SRET) that permits identification of heteromers formed by three different proteins. Cells were distributed in 96-well microplates (Corning 3600, white plates with white bottom), and 5 M DeepBlueC (Molecular Probes, Eugene, OR) was added. The SRET 2 signal was collected using a Mithras LB 940 reader with detection filters for short wavelength (410 nm) and long wavelength (530 nm). By analogy with BRET, net SRET 2 is defined as ((long wavelength emission)/(short wavelength emission)) Ϫ Cf, where Cf corresponds to long wavelength emission/short wavelength emission for cells expressing protein-Rluc and protein-GFP 2 . Linear unmixing was done for SRET 2 quantification, taking into account the spectral signature to separate the two fluorescence emission spectra (17). (iv) Using aliquots of cells transfected for SRET experiments, BRET 1 , BRET 2 , and FRET measurements were performed as indicated above. A SRET 2 saturation curve can be obtained determining SRET 2 as a function of increasing expression of the FRET acceptor (receptor YFP). From these saturation curves, an apparent SRET max was determined by fitting data to a monophasic saturation curve by nonlinear regression using the commercial Grafit curve-fitting software (Erithacus Software, Surrey, UK). These parameters have a similar meaning to these parameters when applied to BRET assays (5).
Each slice was transferred into an incubation tube containing 1 ml of ice-cold Krebs-HCO 3 Ϫ buffer. The temperature was raised to 23°C, and after 30 min, the media were replaced by 2 ml of Krebs-HCO 3 Ϫ buffer (23°C). The slices were incubated under constant oxygenation (O 2 /CO 2 , 95:5%) at 30°C for 4 -5 h in an Eppendorf thermomixer (Eppendorf-5 Prime, Inc., Boulder, CO). The media were replaced by 200 l of fresh Krebs-HCO 3 Ϫ buffer, and after 30 min, 1 M of the A 2A receptor agonist CGS-21680, 1 M of the D 2 receptor agonist quinpirole, or both prepared in Krebs-HCO 3 Ϫ buffer were added. After 10 min, the incubation solution was discarded, and slices were frozen on dry ice and stored at Ϫ80°C. When ERK phosphorylation assays were performed in cell cultures, CHO cells (48 h after transfection) were cultured in serum-free medium for 16 h before the addition of any agent. Cells were resuspended in Hanks' balanced salt solution buffer and were treated for 5 min with CGS2168 (200 nM), quinpirole (1 M), or a mixture of both ligands and rinsed with ice-cold phosphate-buffered saline. Both cells and slices were lysed by the addition of 500 l of ice-cold lysis buffer (50 mM Tris-HCl, pH 7.4, 50 mM NaF, 150 mM NaCl, 45 mM ␤-glycerophosphate, 1% Triton X-100, 20 M phenylarsine oxide, 0.4 mM NaVO 4 , and protease inhibitor mixture). The cellular debris was removed by centrifugation at 13,000 ϫ g for 5 min at 4°C, and the protein was quantified by the bicinchoninic acid method using bovine serum albumin dilutions as standard. To determine the level of ERK1/2 phosphorylation, equivalent amounts of protein (10 g) were separated by electrophoresis on a denaturing 7.5% SDS-polyacrylamide gel and transferred onto PVDF membranes. The membranes were then probed with a mouse anti-phospho-ERK1/2 antibody (Sigma, 1:2500). To rule out that the differences observed were due to the application of unequal amounts of lysates, PVDF blots were stripped and probed with a rabbit anti-ERK1/2 antibody that recognizes both phosphorylated and nonphosphorylated ERK1/2 (Sigma, 1:40,000). Bands were visualized by the addition of anti-mouse HRP-conjugated (Dako, Glostrup, Denmark) or anti-rabbit HRP-conjugated (Sigma) secondary antibodies, respectively, and SuperSignal West Pico chemiluminescent substrate (Pierce). Bands densities were quantified with LAS-3000 (Fujifilm), and the level of phosphorylated ERK1/2 isoforms was normalized for differences in loading using the total ERK protein band intensities. Quantitative analysis of detected bands was performed by Image Gauge version 4.0 software. Bifactorial ANOVA and post hoc Bonferroni tests were used for statistical comparisons.
Mass Spectometric Analysis-0.3 l of equimolar solutions of the various peptides were deposited on the sample plate followed by 0.3 l of matrix, a saturated solution of 2,4,6-trihydroxyacetophenone in 50% ethanol, and left to dry at room temperature. Spectra of each sample spot were acquired using a MALDI TOF-TOF instrument (Applied Biosystem 4700 proteomics analyzer, Framingham, MA) in positive ion mode. Each spectrum is the average of 1000 shots. All peptides were synthesized at The Johns Hopkins School of Medicine "Synthesis and Sequencing Facility." Computational Models of D 2 , CB 1 , and A 2A Receptors-The amino acid sequences of the human D 2 receptor (accession number P14416), CB 1 receptor (P21554), and A 2A receptor (P29274) receptors were obtained from UniProt. Structural simulations of the A 2A receptor are based on its crystal structure (PDB code 3EML) (20). Simulations of the D 2 and CB 1 receptors are based on computational models constructed by homology modeling techniques using the crystal structure of the ␤ 2 -adrenergic receptor (PDB code 2RH1) (21,22) as template. Because of the absence of P5.50 Ballesteros-Weinstein numbering (23) in the CB 1 receptor, we superimposed Tyr 292 (5.58) and Lys 300 (5.66) to Tyr 219 (5.58) and Lys 227 (5.66) of the ␤ 2 -adrenergic receptor. Tyr5.58 and Lys5.66 are structural and functional amino acids involved in the stabilization of the active state by interacting with Arg3.50 and Asp/Glu6.30, respectively, as revealed by the recent crystal structure of the ligand-free opsin, which contains several distinctive features of the active state (24). The highly conserved NPXXYX n ϭ 5,6 F(K/ R) motif at the junction between TM7 and Hx8 is one residue shorter in the ␤ 2 -adrenergic receptor (n ϭ 5) than in rhodopsin and D 2 or CB 1 receptors (n ϭ 6). Thus, this junction in D 2 or CB 1 receptors was modeled as in rhodopsin (PDB codes 1GZM and 2Z73) (25,26). The unambiguous assignment of the TM boundaries to a particular position is not possible. However, we have assumed that TM5 of A 2A extends to position Arg 206 (5.67) as shown in the crystal structure (20), and TM5 of D 2 extends to position Arg 220 (5.69) according to the ␤ 2 -based homology model (21,22). These definitions of TM5 cause Arg 205 (5.66)-Arg 206 (5.67) of the A 2A receptor and 215 (5.64)VLR-RRRKRVN 224 of the D 2 receptor to be located at the end of TM5 in the cytoplasm. In contrast, the Swiss Protein Database assigns these epitopes of A 2A and D 2 in IL3. The crystal structure of squid rhodopsin (PDB code 2Z73) has shown that in addition to the conserved amphipathic Hx8 that runs parallel to the membrane, the C terminus expands toward TM6 (25). However, the structural homology, among GPCRs, probably does not extend to this C-tail domain because of its high variability in length and amino acid composition among the members of the family. This C-tail is formed by 59 amino acids in the CB 1 receptor (Ser 414 -Leu 472 ), only 1 amino acid in the D 2 receptor (Cys 443 ), and 104 amino acids in the A 2A receptor (Arg 309 -Ser 412 ). Nevertheless, Ser 414 -Asn 437 of the CB 1 receptor and Arg 309 -Gly 330 of the A 2A receptor, forming part of this C-tail sequence, were modeled, in an arbitrary manner, based on the structure of squid rhodopsin.
Computational Models of Receptor Heteromers-Cysteine cross-linking experiments have suggested that receptor oligomerization involves the surfaces of TM1, -4, and/or -5 (10,12,13). Thus, the structures of receptor heteromers were modeled in such a manner that substituted cysteines at position 1.

Quaternary Structure of the A 2A -CB 1 -D 2 Receptor Heteromer-
An obvious initial question about receptor heteromers made up of three different receptors is whether each receptor interacts with the other two or not, i.e. if they form a triangular or linear arrangement. As in a prior report (5), we first demonstrated the ability of A 2A -Rluc, D 2 -GFP 2 , and CB 1 -YFP receptors to form heteromers by determining the SRET saturation curve in transfected HEK-293T cells (Fig. 1a). In the same experimental preparation, we found significant BRET 2 and FRET signals between the A 2A -Rluc-D 2 -GFP 2 receptor pair and the D 2 -GFP 2 -CB 1 -YFP receptor pair, respectively (Fig. 1b). Furthermore, we also detected by BRET 1 assays a positive transfer of energy between A 2A -Rluc and CB 1 -YFP receptors (Fig. 1b). These data and the positive SRET signal (Fig. 1a) in cells co-expressing A 2A -Rluc, D 2 -GFP 2 , and CB 1 -YFP receptors suggest a triangular arrangement between the three receptors (Fig. 1c). In fact, taking into account the correlation between FRET efficiency and acceptor/ donor distances and that Rluc, GFP 2 , and YFP are fused to the end of the C terminus of the receptors, the distance between BRET donors and acceptors can be approximated (17). Considering the high FRET efficiency between D 2 -GFP 2 and CB 1 -YFP receptors (36 Ϯ 3%), the range of the distance between GFP 2 and YFP in the heteromer is estimated to be 5.7-6.1 nm. Thus, a linear arrangement of the three receptors could give a positive SRET signal but a very much reduced or even nonsignificant BRET 1 signal between A 2A -Rluc and CB 1 -YFP receptors, because there is a rapid dissipation of the energy transfer (to the 6th power of the distance). Therefore, assuming that the heterotrimer is the minimal unit, only a triangular arrangement of monomers ( Fig. 1c) would make both SRET (Fig. 1a) and BRET 1 (Fig. 1b) possible between A 2A -Rluc and CB 1 -YFP receptors.

Multiple Electrostatic Interactions in A 2A -CB 1 -D 2 Receptor
Heteromers-The amino acid sequence of the human CB 1 receptor contains two highly conserved epitopes with two adjacent Thr and Ser residues (supplemental Table 1), which have a high probability of CK1/2-dependent phosphorylation (Swiss Protein Database "Net Phos" program (27)). They are located in the distal portion of the C terminus (CT) of the CB 1 receptor (Thr 467 and Ser 468 ) and in the middle portion of intracellular loop (IL) 3 (Thr 321 and Ser 322 ). The initial working hypothesis was that these CB 1 receptor epitopes, with high probability of phosphorylation, would be relevant in determining the quaternary structure of the A 2A -CB 1 -D 2 receptor heteromer, by establishing electrostatic interactions with Arg-rich epitopes located in the A 2A and D 2 receptors.
Electrostatic Interaction between Phosphorylated Thr 467 -Ser 468 in the C Terminus of the CB 1 Receptor and Arg 205 (5.66)-Arg 206 (5.67) in the Cytoplasm at the End of Transmembrane Helix 5 of the A 2A Receptor-We first looked at possible alterations in heteromerization between CB 1 and A 2A and between CB 1 and D 2 receptor in cells co-expressing a mutant CB 1 receptor in which Thr 467 (CT) and Ser 468 (CT) were replaced by Ala (CB 1 A467-A468 receptor). In cells co-expressing A 2A -Rluc and CB 1 A467-A468 -YFP receptors, there was a reduction of BRET 1 values when compared with those obtained with cells expressing A 2A -Rluc and CB 1 -YFP (Fig. 2a). On the other hand, these mutations did not modify the FRET values between D 2 -GFP 2 and CB 1 A467-A468 -YFP, when compared with cells expressing D 2 -GFP 2 and CB 1 -YFP (Fig. 2b). This mutated CB 1 receptor and all the mutant receptors described below were shown to be well expressed at the membrane level (results not shown). Furthermore, the fact that the mutated CB 1 receptor selectively altered the RET signal when co-expressed with A 2A but not with the D 2 receptors demonstrates that the results cannot be explained by changes in the membrane expression of the mutant receptor or its putative partner. These results therefore show that Thr 467 (CT) and Ser 468 (CT) of the CB 1 (Fig. 2c). Furthermore, in cells expressing CB 1 A467-A468 -YFP, A 2A -Rluc, and D 2 -GFP 2 receptors, BRET 1 values between A 2A -Rluc and CB 1 A467-A468 -YFP receptors and BRET 2 values between A 2A -Rluc and D 2 -GFP 2 receptors were significantly reduced, and FRET values between D 2 -GFP 2 and CB 1 A467-A468 -YFP receptors were increased relative to cells expressing nonmutated receptors (Fig. 2d). Because the bioluminescent or fluorescent proteins are fused to the CT of the receptors, these results indicate that expression of CB 1 A467-A468 -YFP receptors induced a modification of the quaternary structure of the A 2A -CB 1 -D 2 heteromer, with separation of the CT of CB 1 and A 2A receptors and A 2A and D 2 receptors and approximation of the CT of CB 1 and D 2 receptors.
We then looked for the presence of adjacent Arg residues in intracellular domains of the A 2A receptor that could potentially interact with the phosphorylated Thr 467 (CT) and Ser 468 (CT) of CB 1  Fig. 1b), in mean Ϯ S.E. of five independent experiments performed in duplicate. One-way ANOVA followed by Bonferroni test showed significant increases or decreases with respect to the control (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.005). Linear unmixing of the emission signals was applied to the data for BRET 2 and FRET values (b and d) and for YFP quantification in saturation curves (a and c). e, the spectrum of a mixture of the following three peptides SVSTDAAAE, SVSTDpTpSAE, and LRIFLAARR, shows only one noncovalent complex between SVSTDpTpSAE and LRIFLAARR at 2171.7 atomic mass units (see text).
receptor via electrostatic interactions. We found a highly conserved motif, Arg 205 (5.66)-Arg 206 (5.67) (supplemental Table 1 A205-A206 -Rluc and CB 1 -YFP receptors showed lower BRET 1 values than those expressing WT receptors (Fig. 3a). On the other hand, the BRET 1 values between A 2A A205-A206 -Rluc and D 2 -YFP receptors were similar to the values between A 2A -Rluc and D 2 -YFP receptors (Fig. 3b). Hence, the quaternary structure of the A 2A -CB 1 receptor heteromer depends on an electrostatic interaction between epitopes located in the CT of the CB 1 receptor and in the cytoplasm at the end of TM5 of the A 2A receptor. Furthermore, this electrostatic interaction is also involved in A 2A -CB 1 -D 2 receptor heteromerization (Fig. 3, c and d). In fact, low SRET values were obtained when the A 2A A205-A206 -Rluc receptor was co-transfected with D 2 -GFP 2 and CB 1 -YFP receptors (Fig. 3c). In cells co-expressing A 2A A205-A206 -Rluc, D 2 -GFP 2 , and CB 1 -YFP receptors, BRET 1 and BRET 2 between the heteromer partners were significantly reduced, and FRET values between D 2 -GFP 2 and CB 1 -YFP receptors were increased, compared with cells co-expressing nonmutated receptors (Fig. 3d). Significantly,  Fig. 1b), in mean Ϯ S.E. of five independent experiments performed in duplicate. One-way ANOVA followed by Bonferroni test showed significant increases or decreases with respect to the control (**, p Ͻ 0.01; ***, p Ͻ 0.005). Linear unmixing of the emission signals was applied to the data for BRET 2 and FRET values (d) and for YFP quantification in saturation curves (a-c). e, spectrum of a mixture of the following three peptides LRIFLAAAA, LRIFLAARR, and SVSTDpTpSAE, shows only one NCX between SVSTDpTpSAE and LRIFLAARR at 2171.7 atomic mass units (see text). this outcome is qualitatively the same as the one shown in Fig. 2d with CB 1 A467-A468 -YFP receptor, as it would be expected if both mutants disrupt the same intermolecular interaction.

Electrostatic Interaction between Phosphorylated Thr 321 -Ser 322 in Intracellular Loop 3 of the CB 1 Receptor and an Arg-rich Epitope in Intracellular Loop 3 of the D 2
Receptor-Because the Thr 467 (CT)-Ser 468 (CT)-containing epitope of the CB 1 receptor was found to interact with Arg 205 (5.66)-Arg 206 (5.67) of the A 2A receptor, it was expected that Thr 321 (IL3)-Ser 322 (IL3) of the CB 1 receptor could interact with the D 2 receptor. In fact, when co-expressing the mutant CB 1 A321-A322 -YFP and D 2 -Rluc receptors, the BRET 1 energy transfer between Rluc and YFP was reduced when compared with BRET 1 values obtained with CB 1 -YFP and D 2 -Rluc receptors (Fig. 4a). On the other hand, the BRET 1 values obtained in cells expressing CB 1 A321-A322 -YFP and A 2A -Rluc receptors were similar to those obtained with cells expressing CB 1 -YFP and A 2A -Rluc (Fig. 4b). These results therefore show that the Thr 321 (IL3)-Ser 322 (IL3) motif of the CB 1 receptor is selectively involved in the intermolecular interactions with the D 2 receptor in the CB 1 -D 2 receptor heteromer. The fact that BRET 1 is still measurable between CB 1 A321-A322 -YFP and D 2 -Rluc receptors indicates that, once more, other epitopes are also involved in CB 1 -D 2 receptor heteromerization. Also, the same Thr 321 (IL3)-Ser 322 (IL3) epitope of the CB 1 receptor interacted with the D 2 receptor in the A 2A -CB 1 -D 2 receptor heteromer. Compared with nonmutated receptors, coexpression of CB 1 A321-A322 -YFP receptor with A 2A -Rluc and D 2 -GFP 2 receptors showed a reduction in SRET values (Fig. 4c), and FRET values were significantly decreased, and BRET 2 values were increased, whereas BRET 1 values were not modified (Fig. 4d). This suggests that replacement of  Fig. 1b), in mean Ϯ S.E. of six independent experiments performed in duplicate. One-way ANOVA followed by Bonferroni test showed significant increases or decreases with respect to the control (*, p Ͻ 0.05; **, p Ͻ 0.01). Linear unmixing of the emission signals was applied to the data for BRET 2 and FRET values (d) and for YFP quantification in saturation curves (a-c). e, the spectrum of a mixture of the following three peptides AAEDGKVQVT, pTpSEDGKVQVT, and VLRRRRKRVN shows only one NCX between pTpSEDGKVQVT and VLRRRRKRVN at 2575.6 atomic mass units (see text). mBu, milli-BRET unit.
Thr 321 (IL3) and Ser 322 (IL3) by Ala in CB 1 receptor induces a modification of the quaternary structure of the A 2A -CB 1 -D 2 receptor heteromer with separation of the CT of the CB 1 and D 2 receptors and an approximation of the CT of the D 2 and A 2A receptors. Thus, CB 1 receptor uses two different CK1/ 2-dependent phosphorylatable epitopes, located in their CT (Thr 467 -Ser 468 ) and IL3 (Thr 321 -Ser 322 ) domains, to establish simultaneous electrostatic interactions with the A 2A and D 2 receptors, respectively, in the A 2A -CB 1 -D 2 receptor heteromer.
Next step was finding out which intracellular epitope of the D 2 receptor is involved in CB 1 -D 2 receptor heteromerization. D 2 receptor contains two highly conserved Arg-rich epitopes (supplemental Table 1), 215 (5.64)VLRRRRKRVN 224 , located at the end of TM5 in the cytoplasm (according to the homology modeling using the ␤ 2 -adrenergic receptor as a template; see under "Experimental Procedures"), and 266 NRRRVEAARR 275 , in the middle of IL3. Because the VLR-RRRKRVN epitope is most probably involved in A 2A -D 2 receptor heteromerization (28,29), we explored the possibility that IL3 of the D 2 receptor could interact with IL3 of the CB 1 receptor (phosphorylated Thr 321 -Ser 322 ), The D 2 short isoform (D 2S ), an alternative splicing that lacks 29 amino acid residues of IL3 (30), including 266 NRRRVEAARR 275 , was used. SRET values were clearly reduced when D 2S -GFP 2 receptor was co-expressed with A 2A -Rluc and CB 1 -YFP receptors (Fig. 5c). Significantly, the D 2S receptor led to the same qualitative modifications of the quaternary structure of the A 2A -CB 1 -D 2 receptor heteromer as those induced by CB 1 A321-A322 -YFP receptor (Fig. 5d). Thus, in cells expressing A 2A -Rluc, CB 1 -YFP, and D 2S -GFP 2 receptors, FRET values between D 2S -GFP 2 and CB 1 -YFP receptors were significantly decreased, whereas BRET 2 values between A 2A -Rluc and D 2S -GFP 2  Fig. 1b), in mean Ϯ S.E. of five independent experiments performed in duplicate. One-way ANOVA followed by Bonferroni test showed significant increases or decreases with respect to the control (***, p Ͻ 0.005). Linear unmixing of the emission signals was applied to the data for BRET 2 and FRET values (a, b, and d) and for YFP quantification in saturation curves (a and c). e, spectrum of a mixture of the following three peptides AAEDGKVQVT, pTpSEDGKVQVT, and NRRRVEAARR, shows only one NCX between pTpSEDGKVQVT and NRRRVEAARR at 2506.8 atomic mass units (see text). receptors were increased, and BRET 1 values between A 2A -Rluc and CB 1 -YFP receptors were not modified, when compared with cells co-expressing A 2A -Rluc, D 2 -GFP 2 , and CB 1 -YFP receptors (Fig. 5d). These results indicate that in the A 2A -CB 1 -D 2 receptor heteromer, CB 1 receptors interact with the Argrich domain located in IL3 of the D 2 receptor.
Notably, expression of D 2S -GFP 2 or D 2 -GFP 2 receptors with either CB 1 -YFP or A 2A -Rluc or receptors gives similar FRET (Fig. 5a) or BRET 2 (Fig. 5b) (19,20). We found a dramatic reduction of BRET 1 values in cells co-expressing a mutant A 2A -Rluc receptor, in which Ser 374 (CT) was replaced by Ala (A 2A A374 -Rluc receptor), and D 2 -YFP receptor (Fig. 6a). On the other hand, co-expression of A 2A A374 -Rluc and CB 1 -YFP receptors gave similar BRET 1 values than WT receptors (Fig. 6b). These results confirm that Ser 374 (CT) of the A 2A receptor is  Fig. 1b) in mean Ϯ S.E. of five independent experiments performed in duplicate. One-way ANOVA followed by Bonferroni test showed significant increases or decreases with respect to the control (*, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.005). Linear unmixing of the emission signals was applied to the data for BRET 2 and FRET values (e) and for YFP quantification in saturation curves (a and b). e, spectrum of a mixture of the following three peptides SAQEAQGNT, SAQEpSQGNT, and VLRRRRKRVN shows only one NCX between SAQEpSQGNT and VLRRRRKRVN at 2353.6 atomic mass units (see text). mBu, milli-BRET unit.
involved in the molecular interaction with the D 2 receptor. Not surprisingly, Ser 374 (CT) of the A 2A receptor was also found to be involved in providing the quaternary structure of the A 2A -CB 1 -D 2 receptor heteromer. Low SRET values were obtained when A 2A

A374
-Rluc was co-expressed with D 2 -GFP 2 and CB 1 -YFP receptors (Fig. 6c), compared with cells co-expressing the nonmutated receptors. From the analysis of BRET 1 , BRET 2 , and FRET occurring between partners in cells expressing A 2A A374 -Rluc, CB 1 -YFP, and D 2 -GFP 2 receptors, it was observed that BRET 2 values were significantly reduced, but FRET and BRET 1 values were not significantly modified (Fig.  6d). These results indicate that the CT-mutated A 2A receptor induces a modification of the quaternary structure of the A 2A -CB 1 -D 2 receptor heteromer, with separation of the CT of the A 2A and D 2 receptors. Therefore, the A 2A receptor uses a double-Arg motif (Arg 205 (5.66)-Arg 206 (5.67)) located in the cytoplasm at the end of transmembrane helix 5 and a CK1/2-dependent phosphorylatable epitope located in CT (Ser 374 ) to establish selective electrostatic interactions with the CB 1 and D 2 receptors, respectively. Hence, mass spectrometric analysis of a mixture of peptides corresponding to the cytoplasmic epitope at the end of TM5 of the D 2 ( 215 (5.64)VLRRRRKRVN 224 ) and the CT epitopes of the A 2A receptor ( 370 SAQEpSQGNT 378 ) and the mutant A 2A receptor ( 370 SAQEAQGNT 378 ) resulted in noncovalent complexes between the D 2 and the A 2A receptor epitopes, but not in the case of the mutant A 2A receptor (Fig. 6e).

Role of Casein Kinase 1/2-mediated Phosphorylation in the Quaternary Structure of A 2A -CB 1 -D 2 Receptor
Heteromer-To demonstrate the actual involvement of casein kinase-induced phosphorylation in the electrostatic interactions between A 2A , CB 1 , and D 2 receptors in the A 2A -CB 1 -D 2 receptor heteromer, we studied the effects of co-administration of casein kinase 1 inhibitor IC 261 and casein kinase 2 inhibitor TBAC on SRET saturation experiments in HEK-293T cells co-transfected with A 2A -Rluc, D 2 -GFP 2 , and CB 1 -YFP receptors. As expected, the casein kinase inhibitors significantly decreased SRET values (Fig. 7), supporting a role of casein kinases on maintaining a phosphorylated state of the intracellular domains of A 2A and CB 1 receptors involved in A 2A -CB 1 -D 2 receptor heteromerization.
Modeling the CB 1 -D 2 Receptor Heteromer-Initially, to discern which of these TM interfaces most favorably permits the proposed electrostatic interaction between phosphorylated Thr 321 (IL3)-Ser 322 (IL3) of CB 1 and 266 NRRRVEAARR 275 (IL3) of D 2 in the CB 1 -D 2 receptor heteromer, all possible dimeric interfaces were constructed (supplemental Fig. 1). It is impor-tant to acknowledge the difficulty of modeling IL3 of either CB 1 or D 2 receptors unambiguously (see under "Experimental Procedures"); thus, the exact location of these epitopes in IL3 cannot be determined. Nevertheless, it seems clear to us that the TM1-TM1, TM4-TM4 invago , and TM4-TM4 ago interfaces position IL3 of CB 1 and D 2 receptors in opposite sides of the TM bundles (supplemental Fig. 1, a-c), which makes the proposed electrostatic interaction difficult. In contrast, the TM5-TM5 interface places IL3 of the CB 1 receptor contiguous to IL3 of the D 2 receptor (supplemental Fig. 1d), facilitating their electrostatic interaction. It thus seems reasonable to propose that the Arg-rich epitope of the D 2 receptor located in the cytoplasm at the end of TM5 is involved in CB 1 -D 2 receptor heteromerization.
Modeling the A 2A -CB 1 Receptor Heteromer-The A 2A -CB 1 receptor heteromer was also modeled through the entire set of TM interfaces (supplemental Fig. 2) to reproduce the electrostatic interaction between phosphorylated Thr 467 -Ser 468 in the CT of the CB 1 receptor and Arg 205 (5.66)-Arg 206 (5.67) in the cytoplasm at the end of TM5 of the A 2A receptor. CT of the CB 1 receptor is made of 59 amino acids (Ser 414 -Leu 472 ), in addition to the conserved Hx8 that runs parallel to the membrane (Ser 401 -Pro 413 ). It is thus difficult to determine with precision the position of Thr 467 (CT)-Ser 468 (CT). However, although GPCRs CT vary greatly in length and sequence, we have assumed that the CT of CB 1 unfolds toward TM6 as found in the crystal structure of squid rhodopsin (22). Taking these facts into account, TM4-TM4 invago , TM4-TM4 ago , and TM5-TM5 interfaces between CB 1 and A 2A receptors would allow the electrostatic interaction between Thr 467 (CT)-Ser 468 (CT) and Arg 205 (5.66)-Arg 206 (5.67) in the A 2A receptor (supplemental Fig. 2, b-d), whereas the TM1-TM1 interface would not (supplemental Fig. 2a). Modeling the A 2A -CB 1 -D 2 Receptor Heteromer-The quaternary structure of the A 2A -CB 1 -D 2 heteromer was finally obtained by combining the CB 1 -D 2 (TM5-TM5 interface) and CB 1 -A 2A (TM4-TM4 invago ) models described above (Fig. 8a). This combination of TM-TM interactions was selected among the others because it best reproduces the distance between GFP and YFP in the proposed A 2A -CB 1 -D 2 receptor heteromer within the 5.7-6.1-nm range experimentally determined from FRET efficiencies (see above).  (Fig. 8, a and c).
As shown above, expression of the CB 1 A467-A468 -YFP or A 2A

A374
-Rluc mutant receptors leads to a separation of the CT of CB 1 from A 2A and the CT of A 2A from D 2 , respectively. This clearly suggests that phosphorylated Thr 467 (CT)-Ser 468 (CT) in CB 1 or Ser 374 (CT) in A 2A serves to maintain the large and flexible CT of the receptors in the proper conformation by interacting with the Arg-rich epitope of the corresponding promoter. It thus seems reasonable to suggest that the absence of Thr 467 (CT)-Ser 468 (CT) in CB 1 or Ser 374 (CT) in A 2A modifies the CT of the mutant receptors, whereas the packing of the TMs in the A 2A -CB 1 -D 2 heteromer remains similar.
Structure-Function Relationship in the A 2A -CB 1 -D 2 Receptor Heteromer-We explored the possibility that changes in the quaternary structure of A 2A -CB 1 -D 2 receptor heteromer after disruption of the electrostatic interactions could correlate with changes in the receptor heteromer function. We first looked for differences in signaling (activation of the MAPK pathway) in cells co-expressing A 2A and D 2 receptors in the absence and presence of CB 1 receptors (Fig. 9). In cells co-expressing A 2A and D 2 receptors, coactivation of both receptors with their respective selective agonists CGS 21680 (200 nM) and quinpirole (1 M) produced a similar degree of ERK1/2 phosphorylation than activation of either A 2A or D 2 receptors. As shown in Fig. 9a, the additional co-expression of CB 1 receptor produced a qualitatively different pattern with a significantly higher effect of coactivation of A 2A and D 2 receptors compared with cells expressing only A 2A and D 2 receptors. We then demonstrated that this pattern of MAPK activation is a biochemical characteristic of the A 2A -CB 1 -D 2 receptor heteromer, because it depends on the integrity of its quaternary structure. In fact, we found that it particularly depends on the integrity of the intracellular electrostatic interactions that the CB 1 receptor forms with the D 2 receptor in A 2A -CB 1 -D 2 receptor heteromer. Thus, in cells expressing CB 1 A321-A321 or D 2S receptors (which lose the ability to establish electrostatic interactions with the D 2 or the CB 1 receptors, respectively, in the A 2A -CB 1 -D 2 receptor heteromer), the pattern of MAPK activation was significantly altered and qualitatively similar to that observed in cells only co-expressing A 2A and D 2 receptors (Fig. 9b).
The pattern of MAPK activation could then be used as a biochemical fingerprint of the A 2A -CB 1 -D 2 receptor heteromer to detect its presence in the brain (3). In fact, comparing the pattern of ERK1/2 phosphorylation upon activation of A 2A and D 2 receptors in striatal slices from wild-type mice and CB 1 receptor knock-out mice, we found the same qualitative differences as those observed in co-transfected cells with and without CB 1 receptors (compare Fig. 8, a and c). Thus, in striatal slices from CB 1 receptor knock-out mice, there was a significantly lower ERK1/2 phosphorylation upon co-activation of A 2A and D 2 receptors compared with that obtained from striatal slices from wild-type animals. A bifactorial ANOVA demonstrated a significant genotype effect (p Ͻ 0.05) and significant treatment/ genotype interaction (p Ͻ 0.05), and post hoc Bonferroni tests only showed a significant difference between both groups when the striatal slices were co-treated with CGS 21680 (1 M) and quinpirole (1 M) (Fig. 9c).

DISCUSSION
This study shows, for the first time, that GPCR heteromers display emerging properties that depend on their folding into a certain quaternary structure, determined not only by interactions between TM domains but also involving interactions between hydrophilic intracellular domains. Significantly, we have found that each receptor, A 2A , CB 1 , and D 2 , contains two key intracellular domains to interact in a selective manner with intracellular domains of the other two receptors by means of electrostatic interactions in the formation of the quaternary structure of the A 2A -D 2 , A 2A -CB 1 , CB 1 -D 2 , and A 1 -CB 1 -D 2 receptor heteromers. Thus, the D 2 receptor contains two Arg-rich epitopes, 215 VLRRRR-KRVN 224 and 266 NRRRVEAARR 275 , that interact with potential CK1/2-dependent phosphorylatable Ser/Thr residues in CT (Ser 374 ) of the A 2A receptor and in IL3 (Thr 321 -Ser 322 ) of the CB 1 receptor, respectively; CB 1 receptor contains adjacent phosphorylatable Ser and Thr residues in IL3 (Thr 321 and Ser 322 ) and the CT (Thr 467 and Ser 468 ) that interact with Arg residues in IL3 ( 266 NRRRVEAARR 275 ) of the D 2 receptor and Arg 205 -Arg 206 of the A 2A receptor, respectively; and the A 2A receptor contains Arg residues at the end of TM5 in the cytoplasm at Arg 205 -Arg 206 and a phosphorylatable Ser residue in the CT (Ser 374 ), which interact with phosphorylatable Ser/Thr residues in the CB 1 receptor CT (Thr 467 and Ser 468 ) and an Arg-rich epitope of the D 2 receptor located in the cytoplasm at the end of TM5 ( 215 VLRRRRKRVN 224 ), respectively. The fact that each of these three receptors forms electrostatic interactions involving evolutionarily conserved adjacent Arg residues and CK1/2-dependent phosphorylatable Ser and Thr residues with the other two receptors suggests that these particular electrostatic interactions constitute a general mechanism for receptor heteromerization. In studies using synthetic peptides, it has been shown that these electrostatic interactions are particularly stable. Thus, the Arg-phosphate interaction is so stable that when using collision-induced dissociation, the noncovalent interactions between the Arg guanidinium groups and the phosphate group remain intact even though the covalent bond between the serine and phosphate breaks (16,(31)(32)(33).
Using bioluminescence resonance energy transfer techniques with mutant receptors, we propose for the first time the quaternary structure for three interacting GPCRs. Characterization of protomer organization within the A 1 -CB 1 -D 2 receptor heteromer requires, in addition to our findings, integration of information from a variety of different approaches. Most compelling are studies that apply disulfide cross-linking to map TM interfaces between protomers (10,12,13). Modeling of the CB 1 -D 2 , A 2A -CB 1 , and A 2A -D 2 receptor heterodimers was performed through the entire set of proposed TM interfaces (i.e. TM1, TM4, or TM5). Our results are compatible with models proposed for other family A GPCRs, where oligomerization involves primarily TM4 and TM5 interfaces (Fig. 8). Thus, our study supports a triangular rather than a linear arrangement of receptors in the A 2A -CB 1 -D 2 heteromer. This arrangement allows the possibility of simultaneous homodimerization of A205-A206 receptors, 1 g of cDNA of the D 2 , 0.8 g of cDNA of the D 2S receptor, and 1 g of cDNA of the CB 1 , CB 1 A467-A468 , or the CB 1A321-A322 receptors). Cells were treated for 5 min with 200 nM CGS 21680 (CGS), 1 M quinpirole (Quinp), or both (CGSϩQuinp) and ERK1/2 phosphorylation was determined as indicated under "Experimental Procedures." The immunoreactive bands from four experiments performed in duplicate were quantified, and the values represent the mean Ϯ S.E. of % of phosphorylation relative to the basal levels found in untreated cells. c, assays were performed in striatal slices from wild-type (WT) or CB 1 knock-out mice (CB 1 -KO). The slices were treated for 10 min with 1 M CGS 21680 (CGS), 1 M quinpirole (quinpirole) or both, and ERK1/2 phosphorylation was determined as indicated under "Experimental Procedures." The immunoreactive bands from four to eight slices obtained from five to nine animals were quantified, and values represent the mean Ϯ S.E. of the % of phosphorylation relative to basal levels found in untreated slices. Significant differences respect to the wild-type mice were calculated by bifactorial ANOVA followed by post hoc Bonferroni's tests (**, p Ͻ 0.01; ***, p Ͻ 0.001). each receptor unit using the TM1 interface, which is a well established phenomenon in the GPCR field. Thus, this study opens up a new conceptual challenge in the field of receptor heteromerization, which is the idea that GPCRs can form not only heteromultimers of three different receptor units but also higher order heteromultimers or "receptor nets." The interactions of the intracellular domains of the CB 1 receptor with A 2A and D 2 receptors were found to be fundamental for the correct formation of the quaternary structure needed for the function of the A 2A -CB 1 -D 2 receptor heteromers. Thus, mutant receptors lacking the interacting amino acids significantly disrupted RET and a specific qualitative pattern of ERK1/2 phosphorylation induced by co-activation of A 2A and D 2 receptors. The fact that such a disruption of the quaternary structure of the A 2A -CB 1 -D 2 receptor heteromer (as demonstrated by SRET experiments) was associated with a significant qualitative change in signaling indicates that electrostatic interactions between intracellular domains are also key determinants for the specific biochemical properties of the A 2A -CB 1 -D 2 receptor heteromer. These biochemical characteristics and the specific qualitative pattern of MAPK activation could be used as a biochemical fingerprint of the A 2A -CB 1 -D 2 receptor heteromer presence in the brain. CB 1 receptor KO mice experiments provided strong support for the existence of A 2A -CB 1 -D 2 receptor heteromer in the striatum. It has been hypothesized that A 2A -CB 1 -D 2 receptor heteromers are mostly located in one subtype of striatal neuron, the GABAergic enkephalinergic neuron, where the three receptors are highly co-expressed and exert a significant control of basal ganglia function (34). Most probably, the results obtained with MAPK signaling are just a minor but the first described example of many of the potential properties of the A 2A -CB 1 -D 2 receptor heteromers.