Hetero-oligomerization between β2- and β3-Adrenergic Receptors Generates a β-Adrenergic Signaling Unit with Distinct Functional Properties*

The ability of the closely related β2- and β3-adrenergic receptors (AR) to form hetero-oligomers was assessed by bioluminescence resonance energy transfer. Quantitative bioluminescence resonance energy transfer titration curves revealed that the β2AR has identical propensity to hetero-oligomerize with the β3AR than to form homo-oligomers. To determine the influence of heterooligomerization, a HEK293 cell line stably expressing an excess of β3AR over β2AR was generated so that all β2AR are engaged in hetero-oligomerization with β3AR, providing a tool to study the effect of hetero-oligomerization on β2AR function in the absence of any β2AR homooligomer. The hetero-oligomerization had no effect on the ligand binding properties of various β2AR ligands and did not affect the potency of isoproterenol to stimulate adenylyl cyclase. Despite the unaltered ligand binding properties of the β2/3AR hetero-oligomer, the stable association of the β2AR with the β3AR completely blocked agonist-stimulated internalization of the β2AR. Given that the β3AR is resistant to agonist-promoted endocytosis, the results indicate that the β3AR acted as a dominant negative of the β2AR endocytosis process. Consistent with this notion, the β2/3AR hetero-oligomer displayed a lower propensity to recruit β-arrestin-2 than the β2AR. The hetero-oligomerization also led to a change in G protein coupling selectivity. Indeed, in contrast to β2AR and β3AR, which regulate adenylyl cyclase and extracellular signal-regulated kinase activity through a coupling to Gs and Gi/o, no Gi/o coupling was observed for the β2/3AR hetero-oligomer. Together, these results demonstrate that hetero-oligomerization between β2AR and β3AR forms a β-adrenergic signaling unit that possesses unique functional properties.

The ability of the closely related ␤ 2 -and ␤ 3 -adrenergic receptors (AR) to form hetero-oligomers was assessed by bioluminescence resonance energy transfer. Quantitative bioluminescence resonance energy transfer titration curves revealed that the ␤ 2 AR has identical propensity to hetero-oligomerize with the ␤ 3 AR than to form homo-oligomers. To determine the influence of heterooligomerization, a HEK293 cell line stably expressing an excess of ␤ 3 AR over ␤ 2 AR was generated so that all ␤ 2 AR are engaged in hetero-oligomerization with ␤ 3 AR, providing a tool to study the effect of hetero-oligomerization on ␤ 2 AR function in the absence of any ␤ 2 AR homooligomer. The hetero-oligomerization had no effect on the ligand binding properties of various ␤ 2 AR ligands and did not affect the potency of isoproterenol to stimulate adenylyl cyclase. Despite the unaltered ligand binding properties of the ␤ 2/3 AR hetero-oligomer, the stable association of the ␤ 2 AR with the ␤ 3 AR completely blocked agonist-stimulated internalization of the ␤ 2 AR. Given that the ␤ 3 AR is resistant to agonist-promoted endocytosis, the results indicate that the ␤ 3 AR acted as a dominant negative of the ␤ 2 AR endocytosis process. Consistent with this notion, the ␤ 2/3 AR hetero-oligomer displayed a lower propensity to recruit ␤-arrestin-2 than the ␤ 2 AR. The hetero-oligomerization also led to a change in G protein coupling selectivity. Indeed, in contrast to ␤ 2 AR and ␤ 3 AR, which regulate adenylyl cyclase and extracellular signal-regulated kinase activity through a coupling to G s and G i/o , no G i/o coupling was observed for the ␤ 2/3 AR hetero-oligomer. Together, these results demonstrate that hetero-oligomerization between ␤ 2 AR and ␤ 3 AR forms a ␤-adrenergic signaling unit that possesses unique functional properties.
Cell surface receptors, which mediate their biological effects via coupling to G proteins, control major functions of eucaryotic organisms like neurotransmission, immune response, cell growth, and metabolism (1). Over the past few years, several lines of evidence have supported the notion that G protein-coupled receptors (GPCR) 1 exist and act as homo-or heterooligomeric signaling units (2)(3)(4)(5). Hetero-oligomerization among GPCR has been shown to modulate ligand binding, G proteincoupling, endocytosis, and desensitization of the receptors (6 -19).
Because of their clear and distinct properties, the ␤ 2 adrenergic receptor (␤ 2 AR) and ␤ 3 AR offer an ideal receptor pair to investigate the pharmacological and functional consequences of hetero-oligomerization. Indeed, despite their high degree of sequence homology (49%), each receptor displays characteristic ligand binding properties that can be easily distinguished (20). Also, although the ␤ 2 AR undergoes rapid and efficient agonistpromoted internalization (21)(22)(23), the ␤ 3 AR is resistant to these regulatory processes (24 -26). The fact that these two closely related receptor subtypes are naturally co-expressed in adipocytes (27)(28)(29) suggests the possibility of formation of hetero-oligomers in native tissues and makes the characterization of this hetero-oligomer of potential physiological interest.
One difficulty when investigating hetero-oligomerization of GPCR is the heterogeneity of ligand-binding sites and signaling units that can occur if two receptors are co-expressed in the same cell. To analyze the properties of a hetero-oligomer in the presence of two homo-oligomers, one must ensure that the hetero-oligomer represents the major subpopulation among the expressed receptors. Also, a good correlation between the extent of hetero-oligomerization and the functional changes attributed to the formation of hetero-oligomers should be established. Unfortunately, quantitative assessment of heterooligomers is not a trivial issue. The amount of hetero-oligomers formed will be a function of the affinity of the individual receptors for one another and of their relative expression levels. Because of the difficulty in getting good estimation for these parameters, studies investigating GPCR hetero-oligomerization remained largely qualitative.
In the present study, we established experimental conditions that ensured that the entire population of ␤ 2 AR heterologously expressed in HEK293 cells is engaged in hetero-oligomerization with the ␤ 3 AR, thus allowing the functional characterization of the ␤ 2 AR within the ␤ 2/3 AR hetero-oligomer in the absence of any ␤ 2 AR homo-oligomer.
Our study reveals that in contrast to the robust agonistpromoted endocytosis characteristic of the ␤ 2 AR, the ␤ 2/3 AR hetero-oligomer was not internalized upon agonist stimulation, suggesting that the ␤ 3 AR-negative endocytotic phenotype prevailed in the hetero-oligomer. This dominant-negative effect of the ␤ 3 AR occurred without any change in the ligand binding properties of the receptors and most likely results from a diminished ␤-arrestin-2 recruitment to the hetero-oligomer. When considering the coupling properties of the hetero-oligomer, we found that unlike ␤ 2 AR and ␤ 3 AR, which can couple to G i/o (30,31), the ␤ 2/3 AR hetero-oligomer cannot engage this signaling pathway. Taken together, our results indicate that the ␤ 2/3 AR hetero-oligomer is a ␤AR-like signaling unit distinct from ␤ 2 AR or ␤ 3 AR expressed alone.
Cell Culture and Transfection-HEK293 or COS-1 cells were cultured in DMEM supplemented with 10% FBS, 100 units/ml penicillin/ streptomycin, and 2 nM L-glutamine. For transient expression of recombinant proteins, cells were seeded at a density of 2 ϫ 10 6 cells per 100-mm dish, cultured for 24 h, and then transfected by the calcium phosphate precipitation method for HEK293 cells (33) or by using the FuGENE 6 reagent according to the manufacturer's protocol for COS-1 cells. 48 h after transfection, cells were washed twice with phosphatebuffered saline (PBS), detached with 5 mM EDTA in PBS, and used immediately. HEK293 cell clones stably expressing ␤ 2 AR-Rluc or ␤ 3 AR-Rluc were obtained from transfected cells selected with 400 g/ml G418. The expression of the receptor-Rluc fusion protein was controlled by measuring luciferase activity. To obtain HEK293 cells stably co-expressing ␤ 2 AR-Rluc and ␤ 3 AR-GFP, a cell clone already expressing ␤ 2 AR-Rluc was co-transfected with the ␤ 3 AR-GFP construct along with the pED mtxr vector harboring a point-mutated dihydrofolate reductase gene that transfers resistance to methotrexate (34). After selection with 400 nM methotrexate, clones co-expressing ␤ 2 AR-Rluc and ␤ 3 AR-GFP were isolated, and co-expression of both proteins was monitored by luminescence, fluorescence, and BRET measurement (33).
BRET Measurement-To monitor receptor-receptor interactions in living cells, BRET 2 assays were performed using a TopCount NXT TM (PerkinElmer Life Sciences) as described before (33). Briefly, after catalytic degradation of the substrate DeepBlueC by the energy donor Renilla luciferase (Rluc), light is emitted with a peak at 400 nm. The energy acceptor green fluorescent protein (GFP) is excited by nonradiative energy transfer if GFP is located within a distance of less than 100 Å from the energy donor. As a result, fluorescence is re-emitted by GFP with a peak at 510 nm. The ratio of the light intensity emitted at 500 -530 nm over 370 -450 nm is defined as the BRET 2 signal. For the detection of the ␤-arrestin-2 recruitment by ␤ 2 AR, the BRET 1 technology (Fusion TM ␣-FM, PerkinElmer Life Sciences) was used as described before (32). Briefly, after the substrate coelenterazine H was added, light intensity was sequentially integrated in the 510 -590 and 440 -500 nm window. The BRET 1 signal was defined as the ratio of the light intensity measured at 510 -590 nm over 440 -500 nm. The expression level of the energy donor (Rluc) or acceptor (GFP or YFP) was controlled by measuring luminescence and fluorescence for each BRET experiment as described before (33).
Membrane Preparation-HEK293 cells were homogenized in ice-cold buffer (5 mM Tris/HCl, pH 7.4, 2 mM EDTA, 5 mg/ml leupeptin, 10 mg/ml benzamidine, and 5 mg/ml soybean trypsin inhibitor) using a Polytron (Ultra-Turrax T24, IKA) for 5-10 s at maximum speed. Lysates were centrifuged at 500 ϫ g for 10 min at 4°C. The resulting supernatant was centrifuged at 45,000 ϫ g for 20 min (4°C), and pellets were washed twice using the same buffer. Protein amount was determined (Bio-Rad Protein Assay), and membranes were stored at Ϫ80°C. Radioligand Binding Assay-For saturation binding, 2 g of total membrane preparation were incubated with increasing concentrations (0.002-10 nM) of the ␤-AR antagonist 125 I-labeled cyanopindolol ( 125 I-CYP) in 5 mM Tris/HCl, pH 7.4, and 0.1% bovine serum albumin. Specific binding of 125 I-CYP was determined as the amount of 125 I-CYP binding inhibited by 10 M alprenolol. Samples were incubated for 1 h at 37°C; the reaction was stopped by rapid filtration over Whatman GF/C glass-fiber filters, and the retained radioactivity was measured (1271 RIAgamma Counter, LKB Wallac). Competition binding assays were performed under the same conditions using 25 pM of 125 I-CYP to occupy only ␤ 2 AR (this concentration binds 1% or less of the ␤ 3 AR sites) or 250 pM of 125 I-CYP to ensure the occupancy of ␤ 2 AR and ␤ 3 AR. Specific binding of 125 I-CYP was determined in the presence of increasing concentrations of the competing ligands.
Internalization Assay-To induce receptor endocytosis, cells were stimulated for 30 min with 1 M ISO or 50 nM FEN at 37°C. After washing the cells twice with ice-cold PBS, receptor sequestration was measured by detecting total binding of 25 pM 125 I-CYP (CYP tot ) at 13°C for 3 h. Specific binding of 125 I-CYP on the plasma membrane was determined as the binding inhibited by 100 nM of the hydrophilic ligand CGP12177A (CYP CGP ), although total specific binding was defined as the binding inhibited by 10 M of the hydrophobic ligand alprenolol (CYP Alp ). The internalization rate of the receptor in percentage was calculated based on the following equation: ((CYP tot Ϫ CYP CGP )/(CYP tot Ϫ CYP Alp ) ϫ 100) basal Ϫ ((CYP tot Ϫ CYP CGP )/(CYP tot Ϫ CYP Alp ) ϫ 100) ISO/FEN . Endocytosis was also assessed after cell fractionation as described previously (25). Briefly, after stimulation of whole cells for 30 min at 37°C with 1 M ISO, cells were detached and membranes prepared as described above. Membranes were then placed at the top of a sucrose cushion (35%) and centrifuged at 150,000 ϫ g for 90 min. The light endosomal fraction containing internalized receptor (Rluc light ) was found at the 0 -35% interface, whereas the heavy plasma membrane fraction (Rluc heavy ) sedimented at the bottom of the tube. Each membrane fraction was collected, centrifuged at 200,000 ϫ g for 60 min, and resuspended in 5 mM Tris/HCl, pH 7.4, and the amount of receptor-Rluc was determined by measuring Rluc activity in each fraction. Endocytosis was calculated in percentage using the following equation: (Rluc heavy /(Rluc heavy ϩ Rluc light ) ϫ 100) basal Ϫ (Rluc heavy / (Rluc heavy ϩ Rluc light ) ϫ 100) ISO .
cAMP Production-To determine cAMP accumulation in living cells, 200,000 cells were seeded in 12-well microplates (coated with 0.1% poly-D-lysine) 24 h prior to the experiment and labeled for 4 -6 h in DMEM without FBS containing 2 Ci/ml of [ 3 H]adenine. Cells were stimulated for 30 -45 min at 37°C in DMEM containing 2.5 M IBMX and different drugs at the indicated concentrations. The reaction was terminated by removing the DMEM/IBMX/ligand solution and the addition of ice-cold 5% trichloroacetic acid. [ 3 H]ATP and [ 3 H]cAMP were separated by sequential chromatography (Dowex resin/aluminum oxide), and the accumulation of cAMP was expressed as the ratio of Adenylyl Cyclase Activity-Adenylyl cyclase activity of membranes freshly prepared (see above) from HEK293 cells was measured using the cAMP AlphaScreen TM assay from PerkinElmer Life Sciences (for details see www.perkinelmer.com/lifesciences) with slight modifications. Briefly, membranes (ϳ2 g of proteins) were incubated for 30 min at room temperature in the reaction buffer (5 mM Tris/HCl, pH 7.4, 2 mM EDTA, 53 M GTP, 112 M ATP, 2.7 mM phosphoenolpyruvate, 50 units/ml myokinase, and 10 units/ml pyruvate kinase) containing the tested activators (ISO, FEN, forskolin) and the anti-cAMP-conjugated acceptor bead. The reaction was stopped by adding a buffer containing 0.3% Triton X-100, biotinylated cAMP, and the streptavidin-coated donor bead. After 1 h of incubation at room temperature, the interaction between donor and acceptor beads was detected by measuring the emission of the acceptor bead (520 -620 nm) after excitation of the donor bead (670 -690 nm) by using the Fusion TM ␣-FM from PerkinElmer Life Sciences.
Extracellular Signal-regulated Kinase (ERK) Activity-HEK293 cells were seeded on 6-well microplates at a density of 200,000 cells per well. After 24 h, cells were serum-starved for 24 h. ERK1/2 phosphorylation was induced by the addition of ␤-AR agonists or FBS for 5 min at 37°C. The reaction was stopped by washing the cells twice with ice-cold PBS and adding 200 l of Laemmli buffer (35). Samples were then analyzed by SDS-PAGE and Western blotting by using anti-phospho-ERK1/2 antibodies from mouse (1:5000) and anti-mouse horseradish peroxidase (HRP)-conjugated secondary antibodies from sheep (1:10,000). The immunoactivity was then revealed by chemiluminescence using ECL. After stripping the membranes with 25 mM glycine, pH 2.2, 1% SDS for 1 h, the amount of total ERK1/2 proteins was detected using anti-ERK1/2 antibodies from rabbit (1:10,000) and anti-rabbit HRP-conjugated secondary antibodies from donkey (1:10,000). ERK1/2 phosphorylation was expressed as the ratio of the signal provided by the phospho-ERK1/2 antibody over the signal obtained with the total ERK1/2 antibody.
Data Analysis-Data obtained in binding experiments were analyzed using Prism3.0. Isotherms for saturation or competition binding assays were plotted for one or two binding sites, and the best fit was then used to calculate K D , K i , or EC 50 values. Data obtained in ERK1/2 activity experiments were digitized on a flatbed scanner and analyzed using the Quantity One (Bio-Rad) software. Statistical significance of the differences was assessed by the two-tailed Student's t test.
To exclude further the possibility that the BRET signals obtained could result from nonspecific interaction or aggregation due to overexpression in a heterologous expression system, BRET experiments were carried out in HEK293 cells expressing a wide spectrum of receptor numbers (580 -6570 fmol/mg) in an ϳ1:1 ␤ 3 AR/␤ 2 AR ratio. If the BRET signal observed is the result of spurious interactions between the two receptors, it would be expected to increase as a function of the total receptors expressed even at constant energy acceptor/donor (␤ 3 AR-GFP/␤ 2 AR-Rluc) ratios. As shown in Fig. 1B, the BRET signals detected were constant over the entire range of receptor expression, indicating that overexpression did not drive the heterooligomerization. Most interesting, the lowest receptor density tested (580 fmol/mg) is similar to values reported previously (596 Ϯ 263 fmol/mg) for fully differentiated human brown adipocytes (38), indicating that oligomerization could occur in native tissue.
To assess the propensity of ␤ 2 AR to either homo-or heterooligomerize with ␤ 3 AR, BRET titration curves were carried out in HEK293 cells stably expressing 410 fmol/mg of ␤ 2 AR-Rluc by increasing the concentrations of ␤ 2 AR-GFP or ␤ 3 AR-GFP transiently expressed. As reported previously (33), increasing the GFP fusion proteins expressed led to a hyperbolic increase in the BRET signal that reached a maximum BRET signal once all ␤ 2 AR-Rluc molecules that can engage in oligomerization are in a complex with either ␤ 2 AR-or ␤ 3 AR-GFP. The apparent affinity of the ␤ 2 AR-GFP and ␤ 3 AR-GFP for the ␤ 2 AR-Rluc can then be determined by estimating the GFP concentration required to attain 50% of the maximal BRET signal (BRET 50 ). Titration curves for the ␤ 2 AR homo-oligomer and the ␤ 2/3 AR hetero-oligomer gave similar BRET 50 values of 2.30 Ϯ 0.54 and 2.45 Ϯ 0.67, respectively (Fig. 1C), indicating that ␤ 2 AR had similar propensity to form homo-oligomers and hetero-oligomers with ␤ 3 AR. In contrast, co-expression of GABA B R 2 -GFP in ␤ 2 AR-Rluc-expressing cells led to a marginal BRET signal FIG. 1. Detection of ␤ 2/3 AR hetero-oligomers by BRET 2 . A, the energy donor ␤ 2 AR-Rluc or ␤ 3 AR-Rluc and the energy acceptor ␤ 2 AR-GFP, ␤ 3 AR-GFP, or GABA B R 2 -GFP were transiently co-expressed in HEK293 cells and BRET 2 signals measured following the addition of the luciferase substrate DeepBlueC as described under "Experimental Procedures." Numbers above the bars indicate the mean of the expression ratios of receptor-GFP over receptor-Rluc for each pair. Results are expressed as the mean Ϯ S.E. of five different experiments. B, ␤ 2 AR-Rluc and ␤ 3 AR-GFP were transiently co-expressed in HEK293 cells and the BRET 2 signals measured. The transfections were carried out to obtain increasing numbers of total receptor numbers while keeping the ␤ 3 AR/␤ 2 AR ratio relatively constant. The receptor expression levels appear below each bar and were determined by monitoring the luminescence and fluorescence signals. These values were then transformed into picomoles of receptor/mg of proteins using a standard curve correlating the fluorescence and luminescence signals to the number of 125 I-CYP-binding sites determined in a radioligand binding assay (33). The dotted line indicates the maximal BRET signal that can be achieved. Data from six different experiments were compiled. C, increasing amounts of ␤ 2 AR-GFP, ␤ 3 AR-GFP, or GABA B R 2 -GFP were transiently expressed in a cell line stably expressing ␤ 2 AR-Rluc (410 fmol/mg) and BRET 2 signals measured 48 h after transfection. Expression levels of ␤ 3 AR-GFP varied between 100 and 2000 fmol/mg. BRET 2 signals are plotted as a function of the expression ratio of receptor-GFP over ␤ 2 AR-Rluc obtained by measuring the total luminescence and fluorescence in each condition. The expression level of the ␤ 2 AR-Rluc was not significantly affected by the transient expression of the receptor-GFP fusion proteins. The curve was drawn from data obtained in three independent experiments. The dashed line indicates the ␤ 3 AR-GFP over ␤ 2 AR-Rluc expression ratio observed in HEK293-␤ 2/3 AR cells stably co-expressing both receptor subtypes, and the bar illustrates their BRET 2 signal obtained from five different experiments. that could not be fitted to a hyperbolic function, confirming the selectivity of interaction among the ␤AR subtypes (Fig. 1C).
To generate a cell system that would allow us to reliably assess the properties of the ␤ 2/3 AR hetero-oligomer, we selected a HEK293 cell clone (HEK293-␤ 2/3 AR cells) that stably expressed a high ratio of ␤ 3 AR-GFP/␤ 2 AR-Rluc. The high ratio observed for the HEK293-␤ 2/3 AR cells (3.6 Ϯ 0.2) corresponds to a GFP/Rluc ratio that would be expected to cause nearsaturation of the ␤ 2 AR-Rluc by ␤ 3 AR-GFP (Fig. 1C). This prediction is further supported by the high level of constitutive BRET signal observed in these cells. Indeed, the BRET signal of 0.26 Ϯ 0.02 observed was not significantly different from the maximum BRET signal (0.28 Ϯ 0.03) obtained in the titration experiments (Fig. 1C). These data indicate that most if not all of the ␤ 2 AR-Rlucs, which can engage in oligomerization, are interacting with ␤ 3 AR-GFP molecules, excluding the existence of a significant proportion of ␤ 2 AR homo-oligomers in these cells. Since previous studies have suggested that if GPCR monomers exist, they represent a very small fraction of the total receptor population (33,39), our data indicate that the HEK293-␤ 2/3 AR cells could provide a powerful tool to study ␤ 2/3 AR hetero-oligomers in the absence of appreciable ␤ 2 AR homo-oligomers or monomers. However, the saturating BRET data presented here do not allow us to formally exclude the possibility that a fraction of the ␤ 2 AR may be incapable of forming oligomers and thus that a small population of ␤ 2 AR monomer could still exist in these cells. However, as will be seen from the extent of the effect of ␤ 3 AR co-expression on the agonist-promoted endocytosis of the ␤ 2 AR, this is unlikely.
Agonist-promoted Internalization of ␤ 2/3 AR Hetero-oligomers-One of the major functional differences between ␤ 2 AR and ␤ 3 AR relates to their subcellular distribution following agonist stimulation. Indeed, as indicated in the Introduction, agonist treatment leads to a rapid endocytosis of the ␤ 2 AR while the ␤ 3 AR remains at the cell surface, even following sustained stimulation (24 -26). We thus investigated whether the ␤ 2/3 AR hetero-oligomer could undergo agonist-promoted endocytosis. In a first series of experiments, internalization was determined by assessing the translocation of receptor-Rluc fusion proteins into endosomes following cell fractionation. As expected, agonist stimulation of HEK293 cells expressing the ␤ 2 AR-Rluc alone led to 52 Ϯ 5% receptor internalization, whereas no ␤ 3 AR-Rluc internalization (4 Ϯ 3%) was observed in cells expressing this receptor subtype ( Fig. 2A). In HEK293-␤ 2/3 AR cells, the agonist stimulation failed to promote any detectable internalization (3 Ϯ 2%) of the ␤ 2 AR-Rluc, indicating that co-expression of the two receptors inhibited ␤ 2 AR endocytosis. This dominant-negative effect of ␤ 3 AR on ␤ 2 AR endocytosis was further confirmed by radioligand binding studies assessing the loss of cell surface-binding sites for the hydrophilic ligand CGP12177 (40,41). In these experiments, the CGP12177 binding to ␤ 2 AR was determined by competition against a concentration of the lipophilic ligand 125 I-CYP (25 pM) that binds almost exclusively to ␤ 2 AR (this concentration occupies only 1% of the ␤ 3 AR sites) in HEK293-␤ 2/3 AR cells. As shown in Fig. 2B, although the nonselective ␤-AR agonist ISO promoted a loss of 49 Ϯ 8% of the CGP12177-binding sites in HEK293-␤ 2 AR cells, no significant loss of cell surface ␤ 2 AR (4 Ϯ 2%) was observed in HEK293-␤ 2/3 AR cells. This inhibitory effect of ␤ 3 AR on ␤ 2 AR sequestration was not due to heterologous cross-talk resulting from ␤ 3 AR activation because the ␤ 2 AR-selective agonist FEN used at a concentration (50 nM) that does not activate ␤ 3 AR (see Fig. 8) also failed to induce sequestration of ␤ 2 AR in HEK293-␤ 2/3 AR cells (Fig. 2B). Considering that ␤ 3 AR is able to block ␤ 2 AR sequestration without being directly activated, we conclude that ␤ 3 AR inhibits ␤ 2 AR sequestration as a result of their hetero-oligomerization. The total inhibition of ␤ 2 AR sequestration observed in HEK293-␤ 2/3 AR cells is a strong indication that, as suggested by the BRET data, all ␤ 2 AR are engaged in ␤ 2/3 AR hetero-oligomers and that no functional ␤ 2 AR monomers or homo-oligomers are detectable in these cells. The lack of ␤ 2 AR endocytosis in HEK293-␤ 2/3 AR cells does not represent an idiosyncrasy of the cell clone studied because identical data were obtained with an independent cell clone expressing a similarly high ␤ 3 AR-GFP/ ␤ 2 AR-Rluc ratio (data not shown).
Pharmacological Characterization of ␤ 2/3 AR Hetero-oligomers-Since previous studies (11,12,18) reported that hetero-oligomerization can inhibit ligand binding, alterations in the binding properties of the ␤ 2 AR could be responsible for the lack of agonist-promoted endocytosis. Thus, to determine whether the ligand binding properties of the ␤ 2 AR are affected by its hetero-oligomerization with ␤ 3 AR, both saturation and competition radioligand binding studies were performed. 125 I-CYP saturation isotherms carried out with membranes derived from HEK293-␤ 2/3 AR cells were found to be biphasic, revealing high (K D 1) and low (K D 2) affinity components that agreed well with the affinities of CYP determined in cells individually expressing the ␤ 2 AR and ␤ 3 AR, respectively (Table I). Based on the fact that all ␤ 2 AR are engaged in oligomeric assembly with the ␤ 3 AR in HEK293-␤ 2/3 AR cells, the data indicate that the ␤ 2 AR component of the ␤ 2/3 AR hetero-oligomer maintains ␤ 2 AR-like affinity for CYP. The ␤ 2 AR component of the heterooligomer also maintained ␤ 2 AR-like affinities for the agonists ISO and norepinephrine. Indeed, in competition binding experiments performed with a concentration of 125 I-CYP (25 pM) that can only bind the ␤ 2 AR component, ISO and norepinephrine showed K i values that are not different from those observed in cells expressing the ␤ 2 AR alone (Table I). Most interesting, both K i(H) and K i(L) , which are characteristic of agonist affinities for the G protein-coupled and -uncoupled forms of the receptor (42), were similar, indicating that the ␤ 2 AR component of the ␤ 2/3 AR hetero-oligomer can engage G proteins. Next, to confirm that the selectivity of ligands toward ␤ 2 AR was also preserved in the hetero-oligomer, competition binding experiments with the ␤ 2 AR selective agonist, FEN, and inverse agonist, ICI118551, were carried out using a 125 I-CYP concentration (250 pM) that can detectably bind to both ␤ 2 AR and ␤ 3 AR. As shown in Table I, each compound recognized two binding sites on membranes derived from HEK293-␤ 2/3 AR cells that corresponded well with their affinities for the individually expressed ␤ 2 AR and ␤ 3 AR. Although we cannot exclude that the ␤ 2/3 AR hetero-oligomer could have altered binding properties for other ligands, our results indicate that hetero-oligomerization with the ␤ 3 AR does not radically affect the overall binding properties of the ␤ 2 AR and clearly demonstrate that the lack of agonist-promoted ␤ 2/3 AR endocytosis does not result from altered binding to ISO or FEN.
cAMP Accumulation Induced by ␤ 2/3 AR Hetero-oligomers-Hetero-oligomerization between AT 2 and AT 1 angiotensin receptors has been proposed to inhibit AT 1 receptor activation (14). Thus to determine whether the lack of internalization could result from a similar inhibition of the ␤ 2 AR activity by ␤ 3 AR, the functional characteristic of the ␤ 2/3 AR hetero-oligomer was assessed. When expressed individually, ␤ 2 AR and ␤ 3 AR are well characterized positive regulators of adenylyl cyclase (AC) through their coupling to G s (43). To test whether the ␤ 2/3 AR hetero-oligomeric complex is still able to stimulate AC activity, we examined ISO-induced cAMP production in HEK293-␤ 2/3 AR cells. As shown in Fig. 3A, the concentrationdependent increase in ISO-stimulated AC activity was very similar for ␤ 2/3 AR-, ␤ 3 AR-, and ␤ 2 AR-expressing cells (EC 50 , 9.8 Ϯ 2.6, 28.7 Ϯ 5.8, and 13.9 Ϯ 3.9 nM, respectively). Because the AC activity observed in ␤ 2/3 AR cells could be entirely due to ␤ 3 AR, the functional integrity of ␤ 2 AR within the hetero-oligomer was assessed by using the ␤ 2 AR-selective antagonist ICI118551. The EC 50 for ISO in HEK293-␤ 2 AR cells increased about 65-fold in the presence of ICI118551 (894 Ϯ 64 nM), whereas the response was not affected in HEK293-␤ 3 AR cells (Fig. 3A). For HEK293-␤ 2/3 AR cells, the presence of the ␤ 2 ARselective antagonist led to a biphasic dose-response curve. The first portion of the curve was insensitive to ICI118551, with an EC 50 of 6.5 Ϯ 2.2 nM, and thus represented the ␤ 3 AR response.
The second portion was significantly right-shifted by ICI118551, with an EC 50 of 1254 Ϯ 276 nM, and thus reflected a ␤ 2 AR-mediated cAMP production. Considering the undetectable level of ␤ 2 AR homo-oligomers in HEK293-␤ 2/3 AR cells, the ICI118551-sensitive response most likely resulted from the ␤ 2 AR component of the ␤ 2/3 AR hetero-oligomer. Taken with the detection of a high affinity ISO-binding site (Table I), these results indicate that ␤ 2/3 AR can functionally interact with G s and thus the lack of endocytosis does not reflect an inactive receptor. As in the case of the ␤ 2 AR expressed alone (44), the spontaneous (agonist-independent) activity of the ␤ 2/3 AR hetero-oligomer could be efficiently inhibited by the inverse agonist ICI118551 (Fig. 3B), confirming the unaltered functional properties of the ␤ 2 AR within the hetero-oligomer.

TABLE I
Ligand binding properties of ␤ 2/3 AR hetero-oligomers Saturation binding or competition binding studies using 125 I-CYP were performed with membranes derived from HEK293 cells stably expressing ␤ 2/3 AR, ␤ 2 AR, or ␤ 3 AR. 125 I-CYP concentrations for saturation binding experiments varied from 0.002 to 10 nM. 25 pM (isoproterenol/norepinephrine) or 250 pM (fenoterol/ICI118551) of 125 I-CYP were used for competition binding experiments. Binding curves were analyzed by using the program Prism 3.0 and the curves best-fitted to either one or two binding sites. K D and K i values are expressed in nM, and the B max are given in pmol/mg. When two sites were identified in saturation binding experiments, they were defined as binding site 1 and 2. In competition binding experiments, high and low affinity binding states for agonists were defined as K i(H) and K i(L) , respectively. Results represent the mean Ϯ S.D. of 2-3 independent experiments performed in triplicates. ND, not determined. ␤-Arrestin Recruitment by ␤ 2/3 AR Hetero-oligomers-Because agonist-induced ␤-arrestin recruitment is a crucial step in the ␤ 2 AR internalization process (45), the lack of agonistpromoted internalization of ␤ 2/3 AR raises the question of whether ␤ 2/3 AR hetero-oligomers are able to recruit ␤-arrestin. To address this question, we monitored the recruitment of ␤-arrestin-2-YFP by ␤ 2 AR-Rluc in the absence or presence of co-expressed ␤ 3 AR using the BRET 1 technology (32,46,47) after transient expression in HEK293 cells. As shown in Fig.  4A, stimulation with 1 M ISO promoted a significant BRET signal between ␤ 2 AR-Rluc and ␤-arrestin-2-YFP, reflecting the translocation of ␤-arrestin-2 to the receptor. In contrast, no BRET signal could be observed between ␤ 3 AR-Rluc and ␤-arrestin-2-YFP under the same conditions (data not shown), an observation consistent with the lack of agonist-promoted internalization of this receptor. Co-expression of the ␤ 3 AR with ␤ 2 AR-Rluc and ␤-arrestin-2-YFP significantly reduced the ISOpromoted BRET signal (Fig. 4A). Given that the ␤-arrestin-2-YFP/␤ 2 AR-Rluc expression ratio was unaffected by the ␤ 3 AR co-expression, the decreased BRET signal most likely reflects a decreased ability of the hetero-oligomer to recruit ␤-arrestin-2. However, because the extent of energy transfer does not only reflect the number of BRET pairs generated but also the distance and orientation between the energy donor and acceptor within the pairs, one cannot exclude the possibility that the reduced BRET signal could reflect a different conformational arrangement of the ␤ 2 AR-Rluc/␤-arrestin-2-YFP complex within the hetero-oligomer.
To determine whether the decreased BRET signal results from a reduced ability of the hetero-oligomer to recruit ␤-arrestin-2 or from an altered conformation of the complex, the potency of ISO to recruit ␤-arrestin-2-YFP to ␤ 2 AR-Rluc was assessed in the presence or absence of co-expressed ␤ 3 AR. Indeed, given that the hetero-oligomerization with ␤ 3 AR did not affect the affinity of the ␤ 2 AR for ISO (Table I), the potency of ISO to promote the BRET signal will reflect the relative affinity of ␤-arrestin-2 for the receptor. In the absence of ␤ 3 AR, the potency of ISO to promote ␤-arrestin-2 recruitment was 17.6 Ϯ 4.1 nM (Fig. 4B), a value consistent with the potency of ISO to stimulate AC (13.9 Ϯ 3.9 nM; Fig. 3A) and with its high affinity binding site (15.6 Ϯ 3.6 nM; Table I). In cells co-expressing ␤ 3 AR, the ISO-promoted BRET signal between ␤ 2 AR-Rluc and ␤-arrestin-2-YFP was found to be biphasic, with a high potency close to the one revealed in the absence of ␤ 3 AR (35.6 Ϯ 5.6 nM) and a second potency that was significantly rightshifted (EC 50 , 26,680 Ϯ 3192 nM). The high potency component is not statistically different from the one observed in cells expressing the ␤ 2 AR alone and thus most likely reflects the potency of the ␤ 2 AR homo-oligomer to recruit ␤-arrestin-2. The presence of a ␤ 2 AR homo-oligomer component indicates that the amount of ␤ 3 AR expressed in this transient expression system used was not sufficient to saturate all the ␤ 2 AR. When expressed alone, the ␤ 3 AR cannot recruit ␤-arrestin-2 (48) (data not shown). Thus, the low potency component of the ␤-arrestin-2 recruitment detected in the biphasic curve probably corresponds to the recruitment of ␤-arrestin-2 to the ␤ 2/3 AR hetero-oligomer, consistent with the hypothesis that the hetero-oligomer has a decreased ability to recruit ␤-arrestin-2.
The correlation between the reduced ␤ 2 AR endocytosis and the reduced ability of the ␤ 2/3 AR hetero-oligomer to recruit ␤-arrestin was further investigated in an additional cell system. For this purpose, COS-1 cells expressing ␤ 2 AR-Rluc and ␤-arrestin-2-YFP in the presence or absence of excess ␤ 3 AR were used to measure the agonist-promoted ␤ 2 AR endocytosis and ␤-arrestin recruitment, in parallel, in the same cells. In these experiments, COS-1 cells expressed 1.4 Ϯ 0.4 pmol/mg of ␤ 2 AR-Rluc and 6.2 Ϯ 1.6 pmol/mg ␤ 3 AR, resulting in a ␤ 3 AR/ ␤ 2 AR ratio of 4.4 Ϯ 0.8. As shown in Fig. 5A and in agreement with the results obtained in the HEK293-␤ 2/3 AR cells, the robust ISO-promoted internalization of ␤ 2 AR observed in cells expressing the ␤ 2 AR was inhibited by 95 Ϯ 7.8% in cell coexpressing an excess of ␤ 3 AR. This confirms that the dominantnegative effect of the ␤ 3 AR on ␤ 2 AR endocytosis is not limited to HEK293 cells and is most likely a general phenomenon. The endocytosis inhibition was paralleled by a 90 Ϯ 4.8% decrease in agonist-promoted ␤-arrestin recruitment in cells co-expressing the ␤ 3 AR (Fig. 5B). Taken together, these data support the hypothesis that the dominant-negative effect of the ␤ 3 AR on ␤ 2 AR endocytosis could result from a reduced ability of the hetero-oligomer to recruit ␤-arrestin efficiently.
ISO-induced Desensitization of Agonist-promoted AC Activity in HEK293-␤ 2/3 AR Cells-Because of the central role played by ␤-arrestin in the agonist-promoted desensitization of the ␤ 2 AR, we next examined the ability of the ␤ 2/3 AR hetero-oligomer to undergo ISO-promoted desensitization. For this purpose, ISOpromoted AC activity was assessed in cell membranes derived from HEK293-␤ 2 AR or HEK293-␤ 2/3 AR cells pretreated or not with ISO for 1 h. As shown in Fig. 6, ISO stimulated AC with almost identical potency in membranes derived from naive HEK293-␤ 2 AR and HEK-␤ 2/3 AR cells (EC 50 : ␤ 2 AR, 32.5 Ϯ 4.2 nM; ␤ 2/3 AR, 29.2 Ϯ 3.5 nM). The agonist-promoted desensitization of the ␤ 2 AR can be readily seen by the significant decrease in ISO potency observed in membranes derived from HEK293-␤ 2 AR cells pretreated with ISO (EC 50 , 7560 Ϯ 137 nM). In the case of the HEK293-␤ 2/3 AR cells, ISO pretreatment led to a complex dose-response that was best fitted to a two component curve. The first component displayed an EC 50 (39.4 Ϯ 4.8 nM) almost identical to that obtained in membranes derived from untreated cells, whereas the EC 50 of the second component was significantly right-shifted to a value similar to that obtained in the desensitized ␤ 2 AR-expressing membranes (EC 50 , 12,470 Ϯ 1730 nM). Because ␤ 3 AR does not undergo agonist-promoted desensitization (24,26), and no ␤ 2 AR homo-oligomer exists in the HEK293-␤ 2/3 AR cells, we conclude that the first component of the curve represents the non-desensitized ␤ 3 AR, whereas the second component corresponds to the desensitized ␤ 2/3 AR hetero-oligomer. These data therefore suggest that despite its inefficient recruitment of ␤-arrestin, the ␤ 2/3 AR hetero-oligomer can undergo agonist-promoted desensitization, indicat-ing the contribution of ␤-arrestin-independent mechanism(s). ␤ 2/3 AR-mediated Activation of ERK1/2-Several studies (49 -52) have suggested that the activation of the ERK1/2 signaling pathway by ␤ 2 AR involves ␤-arrestin recruitment and ␤ 2 AR endocytosis. Particularly relevant to the present study is the observation that hetero-oligomerization between ␤ 2 AR and either ␤ 1 AR or -opioid receptors blocked both ␤ 2 AR endocytosis and ␤ 2 AR-mediated ERK1/2 activation (15,16). To determine whether the inhibition of ␤ 2 AR endocytosis resulting from its hetero-oligomerization with ␤ 3 AR also blocked its ability to activate ERK1/2, the agonist-stimulated ERK1/2 activity was assessed. As shown in Fig. 7, ISO promoted ERK1/2 phosphorylation in HEK293-␤ 2/3 AR cells as well as in cells expressing ␤ 2 AR or ␤ 3 AR alone. To determine whether the ␤ 2/3 AR heterooligomer contributed to the response or if it could be entirely attributed to the ␤ 3 AR, cells were stimulated with the ␤ 2 ARselective agonist FEN. Fifty nM FEN activated ERK1/2 in HEK293-␤ 2/3 AR cells but not in cells expressing ␤ 3 AR alone, indicating that ␤ 2/3 AR hetero-oligomers are able to activate ERK1/2. ␤ 2 AR and ␤ 3 AR-stimulated ERK1/2 activity was previously shown to be PTX-sensitive, suggesting a role for G i/o in this pathway (30,31). To determine whether this is also a property of the ␤ 2/3 AR hetero-oligomer, the effect of PTX was investigated. In cells expressing the ␤ 2 AR or ␤ 3 AR alone, PTX significantly decreased the agonist-stimulated ERK1/2 activation (Fig. 7), confirming previous reports. In contrast, ERK1/2 activity stimulated by the selective ␤ 2 AR agonist FEN was found to be resistant to PTX treatment in HEK293-␤ 2/3 AR cells (Fig.  7), indicating that the ␤ 2/3 AR activates ERK1/2 in a G i/o -independent manner. This PTX resistance cannot be attributed to a cell-specific difference such as those recently found among various HEK293 cell clones (53) because the ISO-stimulated ERK1/2 activity (resulting from the ␤ 3 AR homo-oligomers) was found to be PTX-sensitive in the same HEK293-␤ 2/3 AR cell clone (Fig. 7). Together, these results indicate that the ␤ 2/3 AR hetero-oligomer activates the ERK1/2 through a signaling cas- Receptor-stimulated ERK1/2 activation. HEK293-␤ 2 AR, HEK293-␤ 2/3 AR, and HEK293-␤ 3 AR cells pretreated or not with pertussis toxin (PTX) were stimulated for 5 min at 37°C with 1 M ISO, 50 nM FEN, or FBS. ERK1/2 activity was then assessed following cell lysis using phospho-specific ERK1/2 antibodies. Densitometric analysis of the phospho-ERK1/2 bands were carried out and normalized to the intensity of the bands obtained for the total ERK1/2 population following reprobing with anti-ERK1/2 antibodies. The graphical representation of the ERK1/2 activity is expressed as fold over basal. Asterisk indicates a significant (p Ͻ 0.05) difference between PTX-treated and -untreated cells.
cade that is independent of the activation of G i/o .
Effect of PTX on FEN-mediated cAMP Accumulation in Cells Expressing ␤ 2 AR, ␤ 3 AR, or ␤ 2/3 AR-To determine whether the lack of G i/o dependence for the ␤ 2/3 AR-stimulated ERK1/2 activation could also be observed in the AC signaling, we tested the effect of PTX on agonist-induced cAMP accumulation. As shown in Fig. 8, PTX treatment increased the net agoniststimulated cAMP production generated by a nonselective concentration of FEN (1 M) in ␤ 2 AR-, ␤ 3 AR-, and ␤ 2/3 AR-expressing cells, indicating that it alleviated an inhibitory influence of G i/o on the receptor-mediated responses. However, cAMP production induced by a concentration of FEN (50 nM), which activates only the ␤ 2 AR, was PTX-sensitive in HEK293-␤ 2 AR but not in HEK293-␤ 2/3 AR cells, whereas ␤ 3 AR-expressing cells did not respond at all (Fig. 8). Thus, in contrast to ␤ 2 AR or ␤ 3 AR, ␤ 2/3 AR-stimulated AC activity is not limited by a concomitant activation of G i/o proteins, confirming that the heterooligomer has a distinct G protein coupling pattern.

DISCUSSION
Resonance energy transfer approaches similar to the BRET assay used in the present study are becoming tools of choice to monitor protein-protein interactions in living cells (32,54). Because BRET between Rluc and GFP can occur only if the two molecules are within ϳ100 Å from one another, the occurrence of BRET between Rluc-and GFP-bearing proteins indicates a molecular proximity that is consistent with oligomerization. Although dimerization is the simplest form of oligomerization, the detection of BRET does not provide any direct information about the stoichiometric arrangement of the protomers and the proportion of molecules involved in the oligomeric assembly. Thus, the term oligomer rather than dimer, which is often used to offer a simple model, is used throughout this article. Despite these limitations, proper quantitative analysis can provide useful information about the oligomerization process. For instance, BRET 50 values obtained from titration assays can provide information about the relative affinity of the partners involved (33). Also, by assuming that all the considered molecules are free to oligomerize, the maximal BRET values can be used to estimate the proportion of Rluc-bearing molecules that are engaged in oligomerization with the GFP-bearing molecules. Here, we took advantage of some of these features of BRET to quantitatively assess the formation of ␤ 2/3 AR hetero-oligomers.
The similar BRET 50 observed in the present study for the ␤ 2 AR homo-oligomer and the ␤ 2/3 AR hetero-oligomer indicate that the ␤ 2 AR has similar propensity to homo-and to hetero-oligomerize with the ␤ 3 AR. This is not unique to this receptor pair because similar high affinity hetero-oligomerization has been reported for several closely related GPCR subtypes (33,(55)(56)(57). The similar propensity of the ␤ 2 AR to form homo and ␤ 2/3 AR hetero-oligomers suggests that rather small changes in the relative expression levels of the two subtypes could significantly affect the homo/hetero-oligomer ratio. We took advantage of this characteristic to generate a cell line where the overexpression of the ␤ 3 AR would favor the formation of ␤ 2/3 AR hetero-oligomers so that the properties of the hetero-oligomer could be studied in the absence of significant amounts of ␤ 2 AR homo-oligomers or monomers. One line of evidence suggesting that the HEK293-␤ 2/3 AR cells fulfill this criterion was provided by the quantitative analysis of the BRET data. Indeed, the BRET signals observed between ␤ 2 AR-Rluc and ␤ 3 AR-GFP in HEK293-␤ 2/3 AR cells correspond to the theoretical value expected for the hetero-oligomerization of all ␤ 2 AR-Rluc with ␤ 3 AR-GFP. Although one cannot exclude the possibility that a proportion of ␤ 2 AR could be incapable of forming homo-or hetero-oligomers (leading to the persistence of monomer despite a saturation of the BRET signal), the observation that the excess ␤ 3 AR expressed in HEK293-␤ 2/3 AR cells completely inhibited the agonist-promoted endocytosis and G i/o coupling of the ␤ 2 AR strongly indicates that all functional ␤ 2 AR were engaged in hetero-oligomerization. Thus, HEK293-␤ 2/3 AR cells provided a tool to characterize selectively the property of ␤ 2/3 AR hetero-oligomers without the confounding effects of the ␤ 2 AR homo-oligomers, thus solving a problem often encountered in studies assessing the functional outcome of GPCR hetero-oligomerization.
The analysis of HEK293-␤ 2/3 AR cells revealed that the ␤ 2/3 AR hetero-oligomeric complex maintains some of the properties of the individual protomers but acquired distinct ones. The identical properties includes unchanged ligand binding profiles and a preserved positive coupling to both AC and ERK signaling pathways. The hetero-oligomer differs from the ␤ 3 AR by its ability to undergo agonist-promoted desensitization and from the ␤ 2 AR by its lack of efficient ␤-arrestin recruitment and absence of endocytosis in response to activation. Finally, the ␤ 2/3 AR hetero-oligomer differs from both ␤ 2 AR and ␤ 3 AR by its inability to functionally interact with G i/o .
Although several studies (7,11,12,18,58) reported that hetero-oligomerization changes the ligand binding properties of one or both protomers, no such effect was observed for the ␤ 2/3 AR hetero-oligomer. Indeed, neither the binding affinity, potency, nor efficacy of any of the ␤ 2 AR ligands tested was found to be affected by ␤ 2/3 AR hetero-oligomerization. This lack of changes in the pharmacological properties suggests that each protomer maintains its cognate ligand binding pocket. Given the role of the transmembrane domains in defining binding affinity and selectivity among ␤-AR subtypes, the preservation of a ␤ 2 AR-like binding profile argues against a heterooligomerization model involving swapping of transmembrane segments between ␤ 2 AR and ␤ 3 AR and would be more compatible with a contact type hetero-oligomer, where each protomer maintains its original transmembrane domain organization (59,60).
Despite the apparent lack of effect on ligand binding, heterooligomerization had a significant impact on receptor endocytosis. The association of the endocytosis-prone ␤ 2 AR with the endocytosis-resistant ␤ 3 AR led to a complex that could not be internalized upon agonist stimulation, indicating that the ␤ 3 AR acted as a dominant-negative of the ␤ 2 AR endocytosis. Changes in the internalization profile of a given receptor upon co-expression of a distinct receptor have often been reported and attributed to hetero-oligomerization (10, 12, 13, 17-19, 61). In the case of the ␤ 2 AR, co-expression with ␤ 1 AR (16), -opioid receptors (15), or V 2 -vasopressin receptors (62) has been shown to inhibit its internalization. In the latter case, the inhibition required the concomitant activation of the ␤ 2 AR and V 2 -vasopressin receptors and was attributed to the sequestration of ␤-arrestin by the V 2 -vasopressin receptors resulting in a decreased concentration of ␤-arrestins available for the ␤ 2 AR (62). Such a mechanism cannot be invoked to explain the inhibitory effect of the ␤ 3 AR on the ␤ 2 AR internalization because: 1) the ␤ 3 AR blocked ␤ 2 AR internalization promoted by a concentration of FEN that does not activate the ␤ 3 AR, and 2) even when activated, ␤ 3 AR is not able to recruit ␤-arrestin (48). The observation that endocytosis was blocked in the absence of ␤ 3 AR activation also rules out a possible involvement of downstream signaling cross-talk in the dominant-negative effect of the ␤ 3 AR. Thus, as in the case of the ␤ 2 AR/-opioid receptor co-expression, the inhibition of the ␤ 2 AR internalization by the ␤ 3 AR most likely results from their hetero-oligomerization.
Two main hypotheses could be proposed to explain how hetero-oligomerization with an endocytosis-resistant receptor can block ␤ 2 AR internalization: 1) the hetero-oligomerization leads to a complex that cannot efficiently interact with ␤-arrestin, or 2) the interaction of the complex with ␤-arrestin occurs but does not result in the targeting of the hetero-oligomer to the clathrin-coated pits and endosomes. The results obtained in the BRET-based ␤-arrestin recruitment studies clearly favor the first hypothesis. Indeed, the very low propensity of the ␤ 2/3 AR to recruit ␤-arrestin indicates that the ␤ 3 AR efficiently blocks the interaction between ␤-arrestin and ␤ 2 AR via hetero-oligomerization, hence inhibiting ␤-arrestin-mediated endocytosis of the ␤ 2 AR. The difference in the potency of ISO to promote ␤-arrestin recruitment in face of an unaltered receptor binding and AC stimulation potency suggests that the hetero-oligomerization selectively affected the ISO-promoted interaction with ␤-arrestin but not with G s . Because G protein-coupled receptor kinase-mediated phosphorylation of the ␤ 2 AR has been shown to increase its affinity for ␤-arrestin, it could be hypothesized that hetero-oligomerization prevents efficient G protein-coupled receptor kinase-promoted phosphorylation.
Despite the reduced recruitment of ␤-arrestin by the ␤ 2/3 AR hetero-oligomer, this oligomeric complex was found to undergo agonist-promoted desensitization. This is somewhat surprising when considering the important role played by ␤-arrestin in this regulatory process. However, several ␤-arrestin-independent processes such as cAMP-dependent protein kinase-mediated receptor phosphorylation (63,64) and endocytosis-independent receptor degradation (65) have been suggested previously. Additional studies will be needed to determine the precise mechanism underlying the hetero-oligomer desensitization.
It has been suggested previously that receptor endocytosis and/or ␤-arrestin recruitment contributes to the activation of ERK1/2 by ␤ 2 AR (49 -52). Consistent with this notion, the inhibition of ␤ 2 AR endocytosis upon hetero-oligomerization with either ␤ 1 AR or -opioid receptors was accompanied by an abolition of the ␤ 2 AR-stimulated ERK1/2 activation (15,16). In contrast with these studies, the ␤ 2/3 AR, which failed to undergo endocytosis, still evoked ERK1/2 activation in response to the ␤ 2 AR-selective agonist FEN, indicating that endocytosis is not required for the activation of ERK1/2 by the ␤ 2/3 AR complex. The low apparent affinity of the ␤ 2/3 AR for ␤-arrestin also suggests that the ERK1/2 activation promoted by the heterooligomer occurs in a ␤-arrestin independent manner. These characteristics are reminiscent of the ␤ 3 AR-stimulated ERK1/2 that also takes place in the absence of ␤-arrestin recruitment and receptor endocytosis (31,66). These results raise the intriguing possibility that agonist binding to the ␤ 2 AR could transactivate the ␤ 3 AR within the ␤ 2/3 AR hetero-oligomer. Such transactivation within receptor oligomers has been proposed recently for GPCR of the family A (67, 68) and family C (69). However, as reported previously (31,66,70), the ␤ 3 ARmediated ERK1/2 activation was found to be PTX-sensitive, whereas the ␤ 2/3 AR-promoted activation was not. Although this does not exclude the possibility of transactivation within the ␤ 2/3 AR, it clearly indicates that the ␤ 2/3 AR hetero-oligomer induces ERK1/2 activity by a mechanism that differs from the ␤ 3 AR.
The lack of PTX sensitivity of the ␤ 2/3 AR-stimulated ERK1/2 also clearly distinguishes the response of the hetero-oligomer from that observed for the ␤ 2 AR expressed alone. Indeed, in the present study as well as in numerous previous studies (71)(72)(73)(74)(75)(76)(77)(78), ␤ 2 AR was found to couple to both G s and G i/o . This dual coupling of the ␤ 2 AR, observed in both ERK1/2 and AC pathways, was entirely lost for the ␤ 2/3 AR hetero-oligomer. Similar loss of G i/o coupling upon hetero-oligomerization has been suggested previously for the CCR2/CCR5 (79) and ␦/-opioid receptor (10) pairs. In these two studies, coupling of the hetero-oligomer to a G protein that is not classically engaged by each of the protomers, G q/11 in the case of the CCR2/CCR5 and not identified for the ␦/-opioid receptors, was invoked to explain the maintained signaling. In the present study, no coupling to an additional G protein was observed. Rather, all the signaling properties of the ␤ 2/3 AR could be explained by its maintained coupling to G s .
The data presented here demonstrate that upon heterologous co-expression, human ␤ 2 AR and ␤ 3 AR can associate to form a hetero-oligomeric complex that maintains some of the properties of each of the protomers but acquires distinct ones. The distinct signaling characteristics could have important functional consequences on the catecholamine signaling efficacy in cells endogenously co-expressing these two receptors. Most interesting, mammalian adipocytes co-express ␤ 2 AR and ␤ 3 AR with a high ␤ 3 AR/␤ 2 AR expression ratio that resembles the one observed in HEK293-␤ 2/3 AR cells established in the present study (27)(28)(29)80). Moreover, the hetero-oligomerization between human ␤ 2 AR and ␤ 3 AR could be observed at expression levels that are similar to those reported for human adipocytes (38). If, as in the case of HEK293-␤ 2/3 AR cells, hetero-oligomerization with the ␤ 3 AR blocks the G i/o coupling of the ␤ 2 AR in adipocytes, it would undoubtedly lead to a more efficient activation of AC and ensuing lipolysis (27,28,38). The hetero-oligomerization would thus contribute to increase the AC responsiveness of this tissue to the endogenous catecholamines. The inhibition of both ␤-arrestin recruitment and ␤ 2 AR endocytosis, resulting from the hetero-oligomerization with the ␤ 3 AR, would also be predicted to increase the adipocyte responsiveness to catecholamine stimulation. Although the experiments were not performed in the adipocytes, the present work carried out in heterologous expression systems allowed us to make predictions about the potential functional consequences of hetero-oligomerization in cells endogenously co-expressing ␤ 2 AR and ␤ 3 AR. The testing of these predictions in adipocytes derived from knock-out mice lacking either one of the two receptors should now help to determine the physiological importance of hetero-oligomerization for these receptor subtypes.