G Protein Activation by Serotonin Type 4 Receptor Dimers

The discovery that class C G protein-coupled receptors (GPCRs) function as obligatory dimeric entities has generated major interest in GPCR oligomerization. Oligomerization now appears to be a common feature among all GPCR classes. However, the functional significance of this process remains unclear because, in vitro, some monomeric GPCRs, such as rhodopsin and β2-adrenergic receptors, activate G proteins. By using wild type and mutant serotonin type 4 receptors (5-HT4Rs) (including a 5-HT4-RASSL) expressed in COS-7 cells as models of class A GPCRs, we show that activation of one protomer in a dimer was sufficient to stimulate G proteins. However, coupling efficiency was 2 times higher when both protomers were activated. Expression of combinations of 5-HT4, in which both protomers were able to bind to agonists but only one could couple to G proteins, suggested that upon agonist occupancy, protomers did not independently couple to G proteins but rather that only one G protein was activated. Coupling of a single heterotrimeric Gs protein to a receptor dimer was further confirmed in vitro, using the purified recombinant WT RASSL 5-HT4R obligatory heterodimer. These results, together with previous findings, demonstrate that, differently from class C GPCR dimers, class A GPCR dimers have pleiotropic activation mechanisms.

G protein-coupled receptors (GPCRs) 2 are key players in cell-cell communication. They transduce a wide range of extracellular signals, such as light, odors, hormones, or neurotransmitters into appropriated cellular responses (1,2). Signal transduction occurs via conformational changes of the ligandactivated GPCRs, leading to activation of G proteins and their downstream signaling pathways (3).
GPCRs have been considered for a long time as monomeric proteins, and the paradigm of "one ligand/one receptor/one G protein" was the driving principle (1). However, a growing number of studies revealed dimerization/oligomerization of GPCRs (4 -8) mostly in heterologous cells (homo-or heterodimers) but also in native tissues or in vivo (dimers) (9 -13). In line with these observations, a recent report described crystal structures of the chemokine receptor CXCR4 that are consistent with the formation of homodimers (14).
A relatively accepted model proposes that only one protomer in a dimer is fully activated, even when both binding sites are occupied (15)(16)(17)(18)(19). The activated protomer interacts with the G␣ subunit to accelerate GDP/GTP exchange. This is compatible with recent data showing that a rhodopsin monomer (or a ␤ 2 -adrenergic receptor monomer) is sufficient to activate its cognate G protein after purification and reconstitution in a phospholipid bilayer (20,21). However, some observations indicate that the active state of a GPCR dimer is asymmetric (22,23) and that conformational switches occur between protomers (24,25). Moreover, occupation of the second protomer of a dimer is probably not "silent" because it can either favor (26) or reduce (22) coupling efficiency. It has also been reported that occupation of both binding sites in 5-HT 2C receptor dimers is mandatory for receptor activation (27).
Collectively, these reports suggest that the second protomer within GPCR homo-or heteromeric assemblies can be involved in different regulatory mechanisms. How such a diversity may be related to the physiological roles of the corresponding receptor and how it can influence signaling efficiency are poorly addressed questions.
Here, we analyzed the functional response elicited by different combinations of wild type (WT) and mutant 5-HT 4 receptor (5-HT 4 R) dimers in a cellular context and in vitro, using purified proteins. These combinations allowed occupancy of one or both binding sites and the control of the coupling of each protomer to G proteins. We considered the receptor dimer⅐G protein entity as the minimal functional unit (28) because oligomeric entities can be viewed as multiples of dimers. Therefore, throughout we have used the term "dimer" to represent the minimal oligomeric arrangement.
We show that upon agonist occupation of both protomers, 5-HT 4 R dimers were about twice as efficient in activating G proteins as following occupation of only one binding site. This may suggest that each protomer could be independently coupled to a G protein. However, we provide experimental evi-dence suggesting that a single heterotrimeric G s protein couples to a receptor dimer. These results were confirmed by in vitro experiments showing that activation of a single G s protein was higher when both protomers within a dimer were occupied by an agonist.
Cell Cultures and Transfection-COS-7 cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% dialyzed FCS and antibiotics. They were transfected at 60 -70% confluence by electroporation as described previously (31) and processed for subsequent experiments.
Cell Surface ELISA-COS-7 cells in 96-well plates were transfected with WT and/or mutant tagged 5-HT 4 R or GPCR plasmids. Twenty-four hours after transfection, cells were fixed with 4% (w/v) paraformaldehyde at room temperature for 5 min and blocked with phosphate-buffered saline containing 1% fetal calf serum (blocking buffer). Cells were then incubated with the appropriate antibody (anti-HA at 0.6 g/ml, anti-Myc at 2.2 g/ml, anti-FLAG at 4.4 g/ml) in the same buffer for 60 min. After four washes with blocking buffer, cells were incubated with anti-rabbit (1 g/ml) or anti-mouse/HRP conjugate (0.25 g/ml) (GE Healthcare) for 60 min. After extensive washes, the chromogenic substrate was added (Supersignal ELISA femto-maximum sensitivity; Pierce). Chemiluminescence was detected and quantified using a Wallac Victor2 luminescence counter.
Co-immunoprecipitation Experiments and Immunoblotting-COS-7 cells were seeded at 10 6 /150-mm plate 48 h prior to the experiment and then transfected with RhoTag-or Myc-tagged WT and mutant 5-HT 4 Rs as indicated in the figure legends. Cross-linking was carried out in phosphate-buffered saline (PBS) completed with 25 mM of dithiobis(succinimidyl propionate) (Pierce) for 30 min, as described previously (33). After 60 min of incubation at 4°C, samples were processed as described (29), using a 1:1 mixture of Protein A/Protein G-Sepharose beads (GE Healthcare) that were precoupled with 8 g of anti-Myc antibody (Polyclonal A-14, Santa Cruz Biotechnology, Inc.). Immunoprecipitated proteins were eluted in Laemmli sample buffer, resolved by SDS-PAGE, and detected by Western blotting.
Determination of cAMP or Inositol Monophosphate (IP1) Production in Transfected Cells-COS-7 cells were transfected with the appropriate plasmids and seeded in 24-well plates (100,000 cells/well for cAMP and 500,000 cells/well for IP1 measurements). Twenty-four hours after transfection, a 10-min (cAMP) or 30-min (IP1) stimulation with the appropriate concentrations of drugs was performed as described previously (29). cAMP or IP1 production was quantified by HTRF using the cAMP Dynamic kit or the IP-One kit (Cisbio International, Bagnols-sur-Cèze, France), according to the manufacturer's instructions.
Membrane Preparation and Radioligand Binding Assay-Membranes were prepared from transiently transfected COS-7 cells plated in 15-cm dishes and grown in DMEM with 10% dialyzed FCS as described by Claeysen et al. (34). Membranes were homogenized in 50 mM HEPES (pH 7.4; 5 mg of proteins in 1 ml of solution) and stored at Ϫ80°C until use. Membrane suspensions (about 10 g), diluted in 100 l of 50 mM HEPES containing 10 mM pargyline and 0.01% ascorbic acid, were incubated with 100 l of [ 3 H]GR 113808 (specific activity, 82 Ci/mmol) and 50 l of buffer or competing drugs at 20°C for 30 min. For saturation analysis assays, various concentrations of [ 3 H]GR 113808 (0.001-0.8 nM) were used. BIMU8 (10 M) was used to determine specific binding. To quantify [ 3 H]GR 113808 bound to WT receptors in cells co-expressing WT and D100A (or DD) 5-HT 4 Rs, experiments were performed in the presence of 10 M 5-HT, which does not bind to the D100A or DD mutants. The difference between the total [ 3 H]GR 113808 binding and the remaining binding measured in the presence of 5-HT corresponded to the [ 3 H]GR 113808 binding to co-expressed D100A (or DD) receptors. Protein concentration was determined by the bicinchoninic acidic method.
Data Analysis-The dose-response curves were fitted using GraphPad Prism and the following equation for monophasic dose-response curves: y ϭ (y max Ϫ y min )/1 ϩ ((x/EC 50 ) n H ) ϩ y min , where EC 50 is the concentration of the compound needed to obtain 50% of the maximal effect, and n H is the Hill coefficient. Competition and saturation experiments were evaluated by non-linear regression analysis using Prism. All represented data corresponded to the mean Ϯ S.E. of three independent experiments performed in triplicate. Statistical significance was determined with the Student-Newman-Keuls test using Prism.
Purified WT⅐D100A 5-HT 4 R Dimer Preparation-5-HT 4 R production and refolding were done as described (35). The WT⅐D100A dimer complex was obtained with a two-step purification process as already done for the BLT1 dimer (36). Briefly, S-tagged (WT) and Strep-tagged (D100A) 5-HT 4 Rs were expressed as fusion proteins with KSI (the tag sequence was after the thrombin cleavage site). After removing KSI with thrombin (35), WT and D100A receptors were mixed, refolded, and purified as described (28). Refolded receptors were then immobilized on S-protein-agarose and eluted with 1 M MgCl 2 . The protein fraction recovered under these conditions was then loaded onto a Streptactin affinity column (5.0 ϫ 0.6 cm) and eluted with 2.5 mM dethiobiotin.
GTP␥S-binding Assays-BODIPY FL GTP␥S assays were carried out as described (37). Briefly, BODIPY-nucleotide binding to G␣ s subunit was determined in 10 mM Hepes, 1 mM EDTA, and 10 mM MgCl 2 , pH 8.0. Fluorescence measurements on a second to minute time scale were made in a 10-mm cell using a Cary Eclipse fluorimeter equipped with an RX.2000 rapid mixing stopped-flow unit (Applied Photophysics). In the association kinetics experiments, 1 M ligand was added to 20 nM receptor and 200 nM G␣ s ␤ 1 ␥ 2 trimer, and then binding was initiated by the addition of the BODIPY-nucleotide. The change in fluorescence was measured over time and normalized to the base-line binding of the fluorescent BODIPY analog in the absence of agonist.
Chemical Cross-linking and Size Exclusion Chromatography-The stoichiometry of the receptor⅐G protein complex was assessed as described for the BLT1 receptor. Briefly, WT dimers, purified as described above and at a concentration of 10 Ϫ6 M in the presence of the agonist (serotonin at an agonist/receptor molar ratio of 1:1), and trimeric G␣ s ␤1␥2 (receptor to G-protein molar ratio 1:10) were submitted to cross-linking for 5 h, at room temperature, after the addition of dTSP (125 mM stock solution in N,N-dimethylformamide) to a final concentration of 0.5 mM. The reaction was stopped by the addition of glycine to a final concentration of 50 mM. Cross-linked species were submitted to size exclusion chromatography, and the detergent was extracted as described (28). The resulting protein mixture was loaded onto a Superdex S200 HR column (10 ϫ 300 mm; GE Healthcare). The chromatographic retention times were standardized using a gel filtration standard kit (Bio-Rad) comprising IgG (158 kDa), bovine serum albumin (66 kDa), ovalbumin (44 kDa), and myoglobin (17 kDa). The protein composition of each peak after elution was assessed by SDS-PAGE under reducing conditions (2-mercaptoethanol).

Pharmacological Properties of the 5-HT 4 R Mutants Used in
This Study-To study the functional role of the second protomer of 5-HT 4 R dimers, we used different 5-HT 4 -R mutants. The D100A mutant (supplemental Fig. 1) does not bind to 5-HT but can still be activated by synthetic molecules, such as the full agonist BIMU8 and also ML 10375, which acts as a full antagonist at WT 5-HT 4 Rs ( Fig. 1) (34). This mutant is one of the recently designed RASSLs (receptors activated solely by synthetic ligands) (34,39). The D66N mutant poorly activates G s protein, does not display constitutive activity, and does not activate phospholipase C (PLC) (see Fig. 1 for cAMP data and Ref. 40 for cAMP and Ca 2ϩ /inositol phosphate data). The DD double mutant (D100A/D66N mutation) does not bind to 5-HT and shows impaired G s protein-coupling. (Fig. 1) (40). The T104A mutant is activated by 5-HT and BIMU8 but not by benzamides, such as zacopride ( Fig. 1) (41).
Dimerization of 5-HT 4 R at the Cell Surface-Previous studies reported that 5-HT 4 Rs form constitutive dimers in living cells (42). We used TR-FRET tools (32) to confirm that WT 5-HT 4 R monomers could form homodimers and also heterodimers with mutant 5-HT 4 Rs at the cell surface. To this aim, WT and mutant 5-HT 4 Rs were tagged N-terminally with HA or FLAG epitopes and transiently co-transfected in COS-7 cells. As a positive control of constitutive dimerization, we used HA or FLAG-tagged GB 1 and GB 2 GABA B receptor subunits. Similar amounts of 5-HT 4 Rs and GABA B Rs were correctly expressed at the cell surface ( Fig. 2A). However, the TR-FRET signal detected when HA-and FLAG-5-HT 4 R were co-expressed ( Fig.  2B) represented only 30% of the signal obtained for GABA B R heterodimers. This difference might be explained by the fact that the GABA B Rs expressed at the cell surface were obligatory heterodimers, whereas HA-5-HT 4 R monomers could associate with either HA-5-HT 4 R or FLAG-5-HT 4 R. Hence, HA-5-HT 4 R⅐FLAG-5-HT 4 R dimers, which are the only couples producing FRET, represented only half of the real amount of dimers at the cell surface. One can thus assume that the real signal for 5-HT 4 R dimers was around 60% of the GABA B R FRET signal. Similarly, all of the 5-HT 4 R mutants were correctly expressed at the cell surface ( Fig. 2E) and produced FRET signals comparable with those of WT 5-HT 4 R dimers (Fig. 2F). Formation of 5-HT 4 R homo-or heterodimers was further confirmed by co-immunoprecipitation (supplemental Fig. 2).
We then maintained a constant density of HA-5-HT 4 Rs (donors) and increased the density of FLAG-5-HT 4 R or FLAG-GB 2 (acceptors). Using WT or DD 5-HT 4 Rs, we obtained saturating FRET curves, whereas the signal between 5-HT 4 R and GB 2 remained linear and unsaturable (Fig. 2C). These results indicate that 5-HT 4 R homodimerization was specific, whereas the 5-HT 4 R⅐GB 2 signal reflected a collisional and nonspecific contact as suggested also by their FRET emission that corresponded to only 8% of the signal for GABA B R heterodimers ( Fig. 2A).
To further investigate the specificity of 5-HT 4 R homodimerization, we performed competition experiments. Constant amounts of HA-and FLAG-5-HT 4 R were co-expressed with increasing amounts of competing GPCRs belonging to different classes: class A (5-HT 7 R and thyroid-stimulating hormone receptor), class B (PAC 1 R), and class C (mGluR 7 and GABA B2 R). GABA B2 R (GB 2 ) was unable to compete with 5-HT 4 R homodimerization (Fig. 2D). Thyroid-stimulating hormone receptor, mGluR 7 , PAC 1 R, and 5-HT 7 R competed with the formation of 5-HT 4 R homodimers to different extents but only at high and probably not physiological concentrations. By contrast, Myc-5-HT 4 R potently reduced the FRET signal and competed with HA-and FLAG-5-HT 4 R for dimer formation.
Agonist Occupation of Two 5-HT 4 R Protomers Activates the G s Signaling Pathway More Efficiently than Occupation of a Single Protomer-We first verified that the densities of the expressed receptors (assessed by ELISA and radioligand binding assays; see Fig. 1C for WT 5-HT 4 R) were proportional to the quantity of transfected cDNA (up to 250 ng/10 7 cells). The magnitude of second messenger accumulation (cAMP or IP1) was also proportional to the receptors' densities, a prerequisite to compare the responses obtained with different dimers (Fig. 1C).
To assess whether both protomers of 5-HT 4 R dimers needed to be activated to induce the G s signaling pathway, we transfected COS-7 cells with WT 5-HT 4 R and/or the double mutant (DD) that does not bind to 5-HT and does not induce cAMP production (Figs. 1A and 3A). Exposure to 5-HT of cells transfected with the DD mutant alone did not induce cAMP production (Fig. 3B). Conversely, exposure to 5-HT of cells that expressed WT 5-HT 4 R alone or in combination with DD 5-HT 4 R provided overlapping dose-response curves (Fig. 3, A  and B). Similar densities of WT or DD 5-HT 4 Rs were measured in cells expressing WT or DD 5-HT 4 R alone (4.3 Ϯ 0.7 and 4.9 Ϯ 1.0 pmol/mg protein, respectively) and in cells that coexpressed the two protomers (4.5 Ϯ 0.6 and 5.6 Ϯ 0.8 pmol/mg protein, respectively). Therefore, the total receptor density was 2 times higher in co-transfected cells than in cells transfected with either WT or DD 5-HT 4 R. Hence, if we arbitrarily set the dimer density in co-transfected cells at 4d, the density in cells transfected with either WT or DD 5-HT 4 R will be 2d (Fig. 3C). Specifically, in co-transfected cells, three dimer species are theoretically formed (WT⅐WT, WT⅐DD, and DD⅐DD) and their respective densities will be d, 2d, and d (Fig. 3, A and C). If the occupation of one protomer by 5-HT is sufficient to fully activate dimers, WT⅐WT and WT⅐DD dimers would be similarly activated by 5-HT, and the total activation would be more elevated in co-transfected cells than in cells expressing WT 5-HT 4 R alone (H1 hypothesis; Fig. 3C). This differed from what we observed experimentally. We therefore reasoned that WT⅐DD dimers, in which only one protomer is activated by 5-HT, are able to generate cAMP but less efficiently (about 2 times less) than WT⅐WT homodimers (H2 hypothesis; Fig. 3C). In this hypothesis, the 5-HT dose-response curves in cells  expressing WT 5-HT 4 R alone or in combination with DD 5-HT 4 R should be comparable, which corresponds to experimental results (Fig. 3D). Further supporting this hypothesis, the D66N⅐D100A heterodimer occupied by ML 10375, which activates the D100A protomer but not the D66N protomer, was 50% less active than the D100A⅐D100A dimer in which both protomers were occupied by ML 10375 (supplemental Fig. 3).  Then, to compare the maximal cAMP level obtained by activating one or two protomers in the same dimer population, COS-7 cells, in which WT and D100A 5-HT 4 Rs were co-expressed at equivalent densities (4.1 Ϯ 0.6 and 4.8 Ϯ 0.9 pmol/mg protein, respectively), were incubated with 5-HT, ML 10375, or BIMU8 (5-HT activates only the WT protomer, ML 10375 only the D100A protomer, and BIMU8 both protomers; Fig. 4A). The maximal cAMP level measured upon BIMU8 stimulation (normalized to 100%) was about twice the level obtained following incubation with 5-HT or ML 10375 (52.9 Ϯ 1.1 and 49.4 Ϯ 1.3% of the BIMU8 maximal response, respectively; Fig. 4B). This is consistent with our hypothesis that turning on one protomer within a dimer produces about half of the activity obtained by activating both protomers.
We next modified the equilibrium between 5-HT 4 R dimers by progressively increasing the number of non-responding protomers (DD) while the amount of responding protomers (WT) was kept constant (Fig. 5, A and B). This resulted in a decrease in WT⅐WT dimers and a concomitant increment in WT⅐DD dimers, as indicated by TR-FRET (Fig. 5C) and coimmunoprecipitation experiments (Fig. 5D). However, 5-HTinduced cAMP production remained constant, whatever the amount of DD mutant expressed (Fig. 5E), further supporting our hypothesis that WT⅐DD heterodimers are less active than WT⅐WT homodimers (see supplemental Fig. 4 for theoretical results). As expected, cAMP production upon exposure to BIMU8, which activates both WT and DD protomers, increased concomitantly with DD expression, indicating that the cAMP detection system was not saturated (Fig. 5F). Similarly, an increase in cAMP production was observed following expression of increasing amounts of WT 5-HT 4 R (supplemental Fig. 5). To ensure that these observations were independent of the mutants and the drugs used, we investigated the functional response generated by different dimer couples. Activation of only the D100A protomer by ML 10375 (34) in D66N⅐D100A dimers or induction of only the WT protomer by zacopride (41) in WT⅐T104A dimers did not change the production of cAMP in COS-7 cells in which constant amounts of D100A or WT 5-HT 4 Rs were co-transfected with increasing concentrations of D66N or T104A 5-HT 4 Rs (Fig. 6).
Protomers of 5-HT 4 R Dimers Do Not Independently Couple to G Proteins-Our results suggest that agonist-induced activation of both protomers of 5-HT 4 R dimers induces a 2-fold greater activation of cAMP production than occupation of one protomer. We next explored whether simultaneous occupation of the two protomers conferred to 5-HT 4 R dimers a structure able to activate more efficiently a single G protein than occupation of a single protomer or whether each occupied protomer independently activated its own G protein.
To discriminate between these two possibilities, we needed a dimer combination in which a fully active protomer (D100A) was associated with a protomer that could bind to a ligand, and thus adopt an "active" conformation, but was "completely inactive" for coupling to G protein (Fig. 7A). Because a protomer that does not activate the G s -cAMP pathway was not available, we took advantage of the capacity of 5-HT 4 Rs to activate PLC in COS-7 cells. D66N 5-HT 4 R binds to all agonists, including BIMU8 (40), but it did not induce inositol monophosphate (IP1) production upon activation by BIMU8 (Fig. 7B). When BIMU8 was used as an agonist, the D66N⅐D100A dimer exhibited a level of activity comparable with that of a dimer (D100A⅐D100A) in which both protomers can activate the PLC effector pathway (Fig. 7B). This ruled out the hypothesis that each protomer activates its own G protein.

Conformational Switch between Protomers and Faster Activation of G s Proteins upon Occupation of both Protomers-To
further confirm these observations in vitro, we produced a recombinant WT⅐D100A 5-HT 4 R dimer stabilized in detergent solution, as previously described for the BLT1 receptor (35, 36) (see Fig. 8A for the purification steps). Ligand binding experi- ments to evaluate the formation of WT⅐D100A dimers indicated that 5-HT displaced only half of the tritiated antagonist GR 113808 bound to the purified dimeric complex (Fig. 8B). The stoichiometry of the 5-HT 4 R dimer⅐G s protein assembly was investigated by chemical cross-linking followed by size exclusion chromatography (28). A major species was observed with a chromatographic mobility compatible with that of a complex resulting from the association of a single heterotrimeric G s protein with a receptor dimer (Fig. 9A). Indeed, the calculated mass for such a complex (181,926 Da) corresponded to the experimental value estimated from the chromatographic mobility of the molecular weight standards (Fig. 9A).
To monitor receptor activation, we mutated all of the dyeaccessible cysteines of WT or D100A protomers into serine residues except for cysteine 262, which is located at the cytoplasmic end of TM6 in both protomers (see supplemental Fig. 1 for Cys 262 location). This residue was labeled with the environment-sensitive dye IANBD to selectively monitor ligand-induced changes in the conformation of the modified protomer (38). This modification did not alter their binding properties and functional status (data not shown). As demonstrated for the BLT1 receptor (36), two conformational changes were observed upon agonist activation of one protomer. The conformational change of the unliganded protomer of the dimer was smaller than the change of the agonist-liganded protomer. However, a clear conformational switch between protomers within the dimer was detected upon agonist occupation of only one protomer (Fig. 8C).
We then analyzed GTP␥S binding to G␣ s induced by activation of one or two protomers in purified WT⅐D100A dimers. The initial rate of WT⅐D100A-catalyzed GTP␥S binding to G proteins was faster in the presence of BIMU8, which activates both protomers, than in the presence of 5-HT or ML 10375, which activate only one protomer, (WT and D100A, respectively; Fig. 9B). The kinetic constants for receptor-catalyzed GTP␥S binding to G␣ s were 0.18 Ϯ 0.01 min Ϫ1 (with 5-HT), 0.17 Ϯ 0.02 min Ϫ1 (with ML 10375), and 0.25 Ϯ 0.02 min Ϫ1 (with BIMU8), confirming that the G s protein was activated more efficiently when both protomers were turned on.
Asymmetric Positioning of the G s Protein upon Receptor Activation-To explore how the ␣ subunit of G s protein (G␣ s ) interacts with each protomer in a dimer, we performed FRET FIGURE 7. A single G protein is similarly activated by a dimer composed of a "G-protein coupling-deficient" protomer and a fully functional protomer or by a dimer composed of two fully functional protomers. A, theoretical receptor⅐G protein complexes according to the two hypotheses: the D66N⅐D100A heterodimer is either coupled to one heterotrimeric G protein (hypothesis 1) or to two heterotrimeric G proteins (hypothesis 2). BIMU8 can bind to both protomers and induces the stabilization of their "active" state. The D100A protomer is coupled to G q and can induce IP1 production (B), whereas the D66N mutant cannot activate the Gq-PLC pathway. In hypothesis 1, an allosteric positive conformational switch between protomers results in full activation of one single G protein. BIMU8 stabilizes the "active" state of both protomers (even if the functional coupling of the D66N is dead) and induces an activity level of 1, in terms of IP1 production. In hypothesis 2, each protomer is coupled independently to one G protein. experiments in which the WT or D100A protomers of purified WT⅐D100A dimers were labeled with the fluorescence donor (Alexa Fluor 468) at cysteine 262 (see above), and G␣ s was labeled at its N terminus with the acceptor (Alexa Fluor 568; see "Experimental Procedures"). This modification did not affect G␣ s functionality, as shown by the similar receptor-catalyzed GDP/GTP exchange rates at unmodified and modified G␣ s proteins (0.26 and 0.23 min Ϫ1 , respectively).
When the WT subunit was labeled with the fluorescence donor, challenging with 5-HT (which activates only the WT protomer) produced a much lower FRET signal than following challenge with ML 10375 (which induces the D100A protomer) (Fig. 10A). This indicates that the labeled N terminus of G␣ s was near the labeled WT protomer when ML 10375 was bound to the dimer and that it was in the inverse orientation (i.e. the N-terminal part of G␣ s was close to the D100A protomer) following challenge with 5-HT (Fig. 10B). In other words, the C terminus of G␣ s was close to the activated protomer, consistent with previous reports (43,44). Moreover, the FRET signal induced by BIMU8, which binds to both protomers, was about half of the signal observed following ML 10375 stimulation and about twice the 5-HT-induced signal (Fig. 10A). These results are consistent with the assumption that the C-terminal part of G␣ s can move near the WT or D100A protomer, when BIMU8 activates both protomers (Fig. 10B).
Positioning the FRET donor on the D100A (instead of WT) protomer confirmed these results. Indeed, upon 5-HT binding to the WT protomer, the labeled N terminus of G␣ s moved close to the D100A protomer, which resulted in a clear FRET signal (Fig. 10, C and D). Conversely, upon challenge with ML 10375, which activates only D100A, the N terminus of G␣ s moved away from D100A, and, thus, the FRET signal was lowered ( Fig. 10, C and D). BIMU8 induced an intermediate FRET signal, like when WT was used as FRET donor (Fig. 10A). Collectively, these experiments confirm the asymmetric positioning of G␣ s proteins upon activation of 5-HT 4 R dimers, with the C terminus being located near the ligand-occupied protomer.

DISCUSSION
Recent data indicate that some class A GPCRs, such as rhodopsin, as well as the ␤ 2 -adrenergic, -opioid, and neurotensin receptors can activate G proteins as monomers, at least in small lipid vesicles (20,21,45,46). Furthermore, monomeric rhodopsin in solution has been shown to activate transducin at the diffusion limit (47). This clearly indicates that class A GPCRs do not need to dimerize to transduce signals. However, this conclusion should not mask many other data indicating that dimer (and presumably oligomer) formation occurs in class A GPCRs (7).
Dimerization has been mostly studied in terms of assembly and intracellular trafficking. Much less is known about the consequences of dimerization on receptor signaling efficiency. Here, we have addressed this question by using 5-HT 4 R as a class A GPCR model and by taking advantage of a large variety of mutants, which bind or do not bind to agonists, couple or do not couple to G protein, and can be studied in cellular contexts as well as in vitro (34,35,40,41).
Using FRET and saturation binding experiments, we show that 5-HT 4 Rs behave as dimers or oligomers. We then provide evidence consistent with a model in which complete activation of G signaling by 5-HT 4 R dimers requires the activation of both protomers. Moreover, our data argue against independent interaction of each protomer of a dimer with a G protein to induce its activation: 1) a dimer, in which both protomers (D66N⅐D100A) bind to BIMU8 but one (D66N) is unable to couple to G protein, was as active as the (D100A⅐D100A) dimer composed of two protomers that were both capable of binding to BIMU8 and of interacting with the G protein; 2) purified 5-HT 4 R dimers associated with an unique heterotrimeric G protein in detergent and formed a pentameric complex, as already proposed for some class A and class C GPCRs (28,48,49). Further supporting this pentameric organization, we also demonstrated that the heterotrimeric G protein bound asymmetrically to 5-HT 4 R homodimers or heterodimers, as shown previously for the BLT1 and mGlu1 receptors (18,36).
Allosteric interactions between protomers of dimeric GPCRs and their consequences for G protein activation seem to be pleiotropic. Negative cooperativity, which has been reported for chemokine receptor heterodimers (50) and for glycoprotein hormone receptors (12), is consistent with an activation mechanism in which a single ligand molecule binds to a receptor dimer (at least at low concentration). Javitch and co-workers (22) recently reported maximal activation of the dopamine D2 receptor fused to a Gqi5 chimera upon agonist binding to a single protomer. Arcemisbéhère et al. (51) found that purified BLT2 receptor monomers in solution activated G i2 proteins (GTP␥S binding) more efficiently than a solution of dimers at the same receptor concentration. In metabotropic glutamate receptors (mGluRs) (18) as well as in BLT1 receptors (36), only one heptahelical domain is turned on upon activation of these homodimeric receptors when both binding sites are occupied. Conversely, occupation of both protomers is required for full G protein activation by mGluRs (26). Similarly, binding of two ligand molecules is mandatory for 5-HT 2C receptor function (27). Furthermore, each subunit of the yeast ␣-factor receptor is activated independently by agonists, and the subunits cooperate in activating G proteins (52). Other experiments suggest that interaction of M3 muscarinic receptor dimers with ␤-Arrestin-1 and the subsequent activation of ERK1/2 require binding of an agonist to each receptor protomer (53). In agreement with these last reports, our study demonstrates that the binding of two agonists to 5-HT 4 R dimers is required to provide full G protein activation.
How two 5-HT 4 R protomers cooperate to almost double the level of G protein activation in comparison with when a single protomer is activated is unknown. Induction of two subunits leads to a higher GTP␥S incorporation rate than following activation of a single subunit. Further work is without any doubt necessary to determine how 5-HT 4 R dimers enhance the GDP/GTP exchange rate. The partial conformational change induced by the activated protomer in the unresponsive protomer observed in biophysical experiments might reflect cross-reactions between both protomers. Thus, when one protomer is occupied by the agonist, the other one adopts a conformation that is different from that of a naive protomer. The hypothesis that binding of a ligand to one protomer might induce a structural change in the other protomer is in agreement with structural data obtained for the chemokine receptor CXCR4 (14). Accordingly, two liganded protomers might form an asymmetric structure when they are occupied by the same agonist (see model in supplemental Fig. 6). Because GPCRs contact both G␣ and G␤␥ subunits (43), we propose that one 5-HT 4 R protomer might contact G␣ and the other one might contact G␤␥, as already suggested by Johnston and Siderovski (44). This is consistent with the observations made on family A and C GPCRs (19,48) and with the present findings, which show asymmetric binding of the G protein to a dimer.
The mono-, di-, or oligomeric structure of rhodopsin molecules is still a matter of debate. As rhodopsin forms dimers, one can easily understand that the possibility to obtain a response upon occupancy of a single protomer is physiologically relevant (54). Despite a high similarity in their secondary structure, GPCRs respond to a wide range of agonists (from photons to proteins) present at very different concentrations (from one photon in twilight vision to millimolar concentrations of neurotransmitters in synapses) via pleiotropic molecular mechanisms. In this context, the possibility to generate a graduated response, depending on the occupation of one or two binding sites in a GPCR dimer, is an adaptive advantage. This might be especially relevant in the brain, where GPCRs are often localized at the periphery of synapses (e.g. group I mGluRs (55)), far from presynaptic releasing sites (e.g. 5-HT receptors (56)).