β1/β2-Adrenergic Receptor Heterodimerization Regulates β2-Adrenergic Receptor Internalization and ERK Signaling Efficacy*

β1- and β2-adrenergic receptors (β1AR and β2AR) are co-expressed in numerous tissues where they play a central role in the responses of various organs to sympathetic stimulation. Although the two receptor subtypes share some signaling pathways, each has been shown to have specific signaling and regulatory properties. Given the recent recognition that many G protein-coupled receptors can form homo- and heterodimers, the present study was undertaken to determine whether the β1AR and β2AR can form dimers in cells and, if so, to investigate the potential functional consequences of such heterodimerization. Using co-immunoprecipitation and bioluminescence resonance energy transfer, we show that β1AR and β2AR can form heterodimers in HEK 293 cells co-expressing the two receptors. Functionally, β-adrenergic stimulated adenylyl cyclase activity was found to be identical in cells expressing β1AR, β2AR, or both receptors at similar levels, indicating that heterodimerization did not affect this signaling pathway. When considering ERK1/2 MAPK activity, a significant agonist-promoted activation was detected in β2AR- but not β1AR-expressing cells. Similarly to what was observed in cells expressing the β1AR alone, no β-adrenergic stimulated ERK1/2 phosphorylation was observed in cells co-expressing the two receptors. A similar inhibition of agonist-promoted internalization of the β2AR was observed upon co-expression of the β1AR, which by itself internalized to a lesser extent. Taken together, our data suggest that heterodimerization between β1AR and β2AR inhibits the agonist-promoted internalization of the β2AR and its ability to activate the ERK1/2 MAPK signaling pathway.

␤ 1 -and ␤ 2 -adrenergic receptors (␤ 1 AR and ␤ 2 AR) are co-expressed in numerous tissues where they play a central role in the responses of various organs to sympathetic stimulation. Although the two receptor subtypes share some signaling pathways, each has been shown to have specific signaling and regulatory properties. Given the recent recognition that many G protein-coupled receptors can form homo-and heterodimers, the present study was undertaken to determine whether the ␤ 1 AR and ␤ 2 AR can form dimers in cells and, if so, to investigate the potential functional consequences of such heterodimerization. Using co-immunoprecipitation and bioluminescence resonance energy transfer, we show that ␤ 1 AR and ␤ 2 AR can form heterodimers in HEK 293 cells co-expressing the two receptors. Functionally, ␤-adrenergic stimulated adenylyl cyclase activity was found to be identical in cells expressing ␤ 1 AR, ␤ 2 AR, or both receptors at similar levels, indicating that heterodimerization did not affect this signaling pathway. When considering ERK1/2 MAPK activity, a significant agonistpromoted activation was detected in ␤ 2 AR-but not ␤ 1 AR-expressing cells. Similarly to what was observed in cells expressing the ␤ 1 AR alone, no ␤-adrenergic stimulated ERK1/2 phosphorylation was observed in cells co-expressing the two receptors. A similar inhibition of agonist-promoted internalization of the ␤ 2 AR was observed upon co-expression of the ␤ 1 AR, which by itself internalized to a lesser extent. Taken together, our data suggest that heterodimerization between ␤ 1 AR and ␤ 2 AR inhibits the agonist-promoted internalization of the ␤ 2 AR and its ability to activate the ERK1/2 MAPK signaling pathway.
The standard model describing signaling in G protein-coupled receptors (GPCRs), 1 where the receptor functions strictly as a monomer, is no longer tenable. In the past few years, a number of studies have demonstrated that oligomerization of GPCRs may play important roles in receptor trafficking and signaling (for reviews, see Refs. 1 and 2). In addition to forming homodimers, several receptors have been shown to heterodimerize with other receptor subtypes. In some cases, such as the metabotropic GABA B (3)(4)(5)(6)(7) and the gustatory receptors (8,9), heterodimerization between closely related subtypes was found to be essential for the formation of functional receptors. Although only a few examples for such obligatory heterodimerization are available to date, an increasing number of reports suggest the occurrence of heterodimerization between more or less closely related family members (10 -19). In some of these cases, heterodimerization has been proposed to lead to receptors with pharmacological and/or functional properties that are different from those of the individual receptors.
The potential regulatory influences that these oligomeric assemblies may have on the function of receptors co-existing in the same cell led us to investigate whether receptors that are ubiquitously expressed in many tissues and cell types could function as heterodimers. Two such widely distributed receptors are the ␤ 1 AR and ␤ 2 AR, which are co-expressed in a large number of tissues and cell types (20 -24). Interestingly, the two receptor subtypes were shown to form homodimers when expressed individually in heterologous expression systems (25)(26)(27). Although the two receptors display more than 50% sequence homology (28) and share transmembrane domain motifs proposed as a dimerization interface (25), no study has directly investigated their potential for heterodimerization.
Although the two receptors are known to couple to G s important differences in their functional properties have been reported. For instance, the ␤ 2 AR has been shown to be more efficiently coupled than some variants of the ␤ 1 AR to adenylyl cyclase (29 -31). Also, ␤ 2 AR activation leads to a more efficient stimulation of various MAPK signaling pathways (32,33). In further contrast to the ␤ 2 AR, which undergoes rapid internalization following agonist stimulation, the ␤ 1 AR was found to remain largely localized at the cell surface for extended periods following agonist stimulation (34 -36). Since these differences represent intrinsic properties that were determined for the individual receptors, they offer useful readouts to assess the functional consequences of heterodimerization.
In the present study, the occurrence of heterodimerization between the ␤ 1 AR and ␤ 2 AR was assessed by co-immunoprecipitation and bioluminescence resonance energy transfer (BRET) in living human cells co-expressing the two receptors. The potential functional consequences of the heterodimerization were also assessed by determining the influence of receptor co-expression on the ability of the receptors to stimulate adenylyl cyclase and MAPK pathways as well as to undergo agonist-promoted internalization. We found that co-expression of the two receptor subtypes in HEK 293 cells lead to their heterodimerization and inhibited both agonist-promoted ␤ 2 AR internalization and ERK1/2 MAPK stimulation. This suggests that heterodimerization may represent a regulatory cross-talk process arising through the creation of a receptor form that has distinct functional properties.

Materials
Unless otherwise stated, all chemicals were of reagent grade or higher and were obtained from Sigma.
HA-␤ 2 AR-The human His␤ 2 AR coding sequence was amplified using a sense primer containing a BamHI restriction site followed by the HA tag sequence and an antisense primer. The amplified HA-␤ 2 AR sequence was digested with BamHI/EcoNI and then subcloned into BamHI/EcoNI-digested pcDNA3 (RSV)-His-␤ 2 AR vector, resulting in pcDNA3 (RSV)-HisHA-␤ 2 AR plasmid.
␤ 1 AR-Rluc-The pcDNA3.1-␤ 1 AR:6:hRluc was a generous gift from BioSignal Packard Bioscience. This fusion protein contains a linker of 6 amino acids linking the carboxyl tail of the human ␤ 1 AR to humanized Rluc.
␤ 2 AR-Rluc-The humanized Rluc coding sequence (Packard Instrument Co.) was amplified using sense and antisense primers and then subcloned into the PCR Blunt II Topo vector (Invitrogen). The hRluc fragment was excised by digestion with KpnI/XbaI and subcloned into the KpnI/XbaI-digested pcDNA3.1 Zeo vector to generate the pcDNA3.1 Zeo/hRluc plasmid. The human His␤ 2 AR coding sequence was amplified without its stop codon using sense and antisense primers. The PCR product was subcloned into PCR Blunt II Topo Vector and then excised by double digestion with HindIII/KpnI and ligated into the HindIII/KpnI-digested expression vector pcDNA3.1Zeo/hRluc. The resulting construct encodes a 6-amino acid linker between the carboxyl tail of the ␤ 2 AR and the humanized Rluc sequence. In some experiments, a c-myc-tagged version (N-terminal) of the ␤ 2 AR-Rluc construct was used.

Cell Culture and Transfection
HEK 293 cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Wisent or BioMedia), 100 units/ml penicillin/streptomycin, 2 mM L-glutamine (from Wisent or Invitrogen). For transfection experiments, cells were seeded at a density of 2 ϫ 10 6 cells/100-mm dish and cultured for 24 h. Transient transfections were then performed using either the calcium phosphate precipitation protocol (39) or LipofectAMINE (Invitrogen) according to the manufacturer's recommendations. The cells were then cultured in the same medium for 48 -72 h. In some experiments, transient transfections were carried out in HEK 293 cells stably expressing the human ␤ 2 AR (a generous gift from Dr. Phil Wedergaertner, Thomas Jefferson University).

Immunoprecipitation and Western Blot Analysis
For immunoprecipitation experiments, cells co-expressing HAtagged receptors and the ␤ 2 AR-GFP were harvested 72 h post-transfection, washed three times in PBS, and incubated with blocking buffer (PBS containing 0.2% bovine serum albumin) for 1 h on ice. Subsequently, the cells were incubated with the mouse monoclonal anti-HA (12CA5) antibody (1:250 dilution) in blocking buffer on ice for an additional 1 h. After two washes in blocking buffer and two washes in PBS, cells were lysed, and proteins were solubilized in radioimmune precipitation buffer containing 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 50 mM iodoacetamide, 1 mM phenylmethylsulfonyl fluoride, 0.1 mg/ml benzamidine, 2.5 g/ml leupeptin for 30 min on ice and centrifuged at 12,000 ϫ g for 15 min at 4°C to remove cellular debris. For immunoprecipitation of the total receptor pool, lysates were incubated overnight at 4°C with additional 12CA5 antibodies (1:200 dilution) before the addition of protein G-Sepharose for 3 h. For immunoprecipitation of cell surface receptors, only protein G-Sepharose was added. Protein G-Sepharose-antibody-antigen complexes were then collected by centrifugation at 12,000 ϫ g. The immunoprecipitates were washed four times with cold radioimmune precipitation buffer and resuspended in sample buffer containing 60 mM Tris-HCl, pH 6.8, 2% SDS, 4 M urea, and 100 mM dithiothreitol and heated at 50°C. Protein concentration to be used in the immunoprecipitation was assessed using the DC Protein assay kit (Bio-Rad) with bovine serum albumin as a standard. Protein samples were resolved by 10% SDSpolyacrylamide gel electrophoresis, transferred to nitrocellulose (Protran), and subjected to immunoblotting using rabbit polyclonal anti-GFP antibody (CLONTECH, 1:100 dilution). The Renaissance chemiluminescence kit (PerkinElmer Life Sciences) was used for Western blot development.

BRET Assay
Forty-eight hours post-transfection, cells were washed twice in PBS, detached with PBS/EDTA, and resuspended in PBS plus 0.1% glucose. Cells were then distributed in 96-well microplates (white Optiplate from BioSignal Packard Bioscience) at a density of ϳ100,000 cells/well. Deep Blue C coelenterazine (BioSignal Packard Bioscience) was added at a final concentration of 5 M, and readings were collected using a modified Top-count apparatus (BRETCount) that allows the sequential integration of the signals detected in the 370 -450-and 500 -530-nm windows using filters with the appropriate band pass (Chroma). The BRET signal is determined by calculating the ratio of the light emitted by the receptor-GFP (500 -530 nm) over the light emitted by the receptor-Rluc (370 -450 nm). The values were corrected by subtracting the background signal detected when the Receptor-Rluc constructs were expressed alone. On a routine basis, the protein concentration of the samples was determined to control for the number of cells using a Bradford assay (Bio-Rad) with bovine serum albumin as a standard. To determine the maximal BRET level detectable between each partner, preliminary experiments were carried out by co-transfecting increasing amounts of the receptor-GFP plasmids with a constant quantity of receptor-Rluc construct.

Membrane Preparation
Membranes were prepared from cells 48 h after transfection as described previously (25,40). Briefly, cells were washed twice with ice-cold PBS. They were then disrupted by homogenization with a Polytron homogenizer in 10 ml of ice-cold buffer containing 5 mM Tris-HCl, pH 7.4, 2 mM EDTA, 5 g/ml leupeptin, 10 g/ml benzamidine, and 5 g/ml soybean trypsin inhibitor. Lysates were centrifuged at 500 ϫ g for 5 min at 4°C to remove nuclei and unbroken cells. The supernatant was then centrifuged at 45,000 ϫ g for 20 min, and the pellet was washed twice in the same buffer. Membrane preparations were used immediately for adenylyl cyclase and binding assays.

Receptor Quantification
Membranes were prepared and washed as described above. Total ␤AR number was calculated from binding experiments using a saturating concentration (300 pM) of [ 125 I]cyanopindolol (CYP; PerkinElmer Life Sciences) as the radioligand. Briefly, membrane preparations (10 g of protein) were incubated with [ 125 I]CYP for 90 min at room temperature in a total volume of 0.5 ml in the presence or absence of 10 M alprenolol or propranolol (Sigma) to define specific binding. In some experiments, the proportion of ␤ 1 AR and ␤ 2 AR expressed were determined using a concentration (0.1 M) of the selective ␤ 2 -or ␤ 1 -specific antagonists ICI 118,551 (Tocris) or betaxolol (Tocris) that fully occupies one of the subtypes but blocks less than 10% of the other.
Surface receptor expression was determined using whole cell binding assays with the hydrophilic ligand [ 3 H]CGP 12177 (CGP; PerkinElmer Life Sciences) as described (41). Cells were transiently transfected with either ␤ 2 AR-GFP10 or ␤ 2 AR-GFP10 and HA-␤ 1 AR. Briefly, ␤ 2 AR-specific internalization was determined using 5 ϫ 10 Ϫ8 M betaxolol to block CGP binding to the ␤ 1 AR, and nonspecific surface binding was determined using 10 M propranolol and subtracted from values determined in the presence of betaxolol. These experiments (except for washing steps) were conducted at 37°C. Cells were treated with 10 M isoproterenol or vehicle (10 Ϫ4 M ascorbic acid) for 2 h and then washed with cold PBS to remove ligand. For cell surface binding, cells were incubated with CGP for 30 min, and for total binding, cells were incubated with CYP for 60 min. Cells were then filtered with a Brandel cell harvester as for the membrane receptor assay described above.

Measurement of Adenylyl Cyclase Activity
Adenylyl cyclase activity was assayed in the same membrane preparation according to the method of Salomon et al. (42) using 50 g of protein in a total volume of 50 l. Enzyme activities were determined following a 15-min incubation in the presence of 1 nM to 100 M isoproterenol, 100 M forskolin, 10 mM NaF, or the vehicle at 37°C. Data were calculated as pmol of cAMP produced/min/mg of protein, normalized with respect to forskolin-stimulated adenylyl cyclase activity and analyzed by least squares regression using GraphPad PRISM.

Confocal Microscopy
Cells were cultured on coverslips overnight in Dulbecco's modified Eagle's medium and stimulated for the indicated times with 1 M isoproterenol at 37°C. They were then permeabilized and fixed with 3% paraformaldehyde (v/v) and 0.2% Triton X-100 (v/v) for 20 min at room temperature. After three washes with PBS, the cells were treated with a blocking solution containing 2% BSA and 5% preimmune normal donkey serum (Jackson Laboratories) for 1 h. Rabbit polyclonal anti-␤ 2 AR (1:100 dilution; Santa Cruz Biotechnology) or anti-␤ 1 AR (a generous gift from Dr. Brian Kobilka, Stanford University; 1:100 dilution), diluted in a solution of 2.5% normal donkey serum and 1% BSA, were subsequently incubated with fixed, permeabilized cells for 16 h at 4°C. Following three washes with PBS, CY5-labeled secondary donkey antirabbit antiserum (1:500 dilution; Jackson Laboratories) was incubated with the samples for 60 min in the dark followed again by three washes with PBS. For colocalization experiments, cells were transfected with ␤ 2 AR-GFP and ␤ 1 AR. GFP fluorescence was stimulated with a 488-nm argon laser, and these signals were overlaid with CY5 fluorescence (stimulated with a 633-nm helium-neon laser) after labeling with anti-␤ 1 AR. Confocal microscopy was performed using a Zeiss LSM-510 system with a highly corrected objective (Zeiss Plan-Apochromat ϫ63, numerical aperture 1.4 under oil). Control experiments were performed in the absence of primary antibodies and revealed a low level of background staining, indicating the specificity of the primary antibodies used.

Enzyme-linked Immunosorbent Assay to Detect Receptors Expressed at the Cell Surface
Cells were transfected with ␤ 2 AR-Rluc bearing a c-myc tag on the extracellular N terminus alone or with ␤ 1 AR-GFP10 and transferred to six-well polylysine-coated plates 12 h after transfection. 24 h posttransfection, cells were treated with 10 M isoproterenol or vehicle for 60 min at 37°C. Cells were kept on ice for all subsequent steps. After three washes with cold PBS, cells were blocked with PBS with 1% BSA (w/v) for 30 min. Blocked cells were then incubated with anti-myc (9E10) antibodies (1:500 dilution) for 60 min and subsequently washed again three times with PBS plus 1% BSA. Cells were then fixed with 3% paraformaldehyde (v/v) for 15 min and washed three times with PBS. Cells were blocked again for 15 min with PBS plus 1% BSA, incubated with anti-mouse/horseradish peroxidase conjugate (Amersham Biosciences; 1:1000) for 30 min, and finally washed with PBS plus 1% BSA three times. The substrate o-phenylenediamine dihydrochloride (Sigma) was added according to the manufacturer's instructions for 4 -7 min. The reaction was stopped with 200 l of 3 N HCl, and extinction was measured at 492 nm. Data were plotted as percentage of signal (isoproterenol) relative to time-matched control (vehicle).

RESULTS
Characterization of the ␤ 1 AR/␤ 2 AR Heterodimer by Co-immunoprecipitation-In a first attempt to determine whether the ␤ 2 AR could form a heterodimer with the ␤ 1 AR, a co-immunoprecipitation experiment between ␤ 1 AR and ␤ 2 AR bearing different epitope tags (HA and GFP, respectively) was designed. The ability of the ␤ 2 AR to form homodimers or to heterodimerize with the ␤ 1 AR was assessed in HEK 293 cells co-expressing the ␤ 2 AR-GFP with either HA-␤ 2 AR or HA-␤ 1 AR (Fig. 1, lanes 3-6). After cell surface (S) or total cell extract (T) immunoprecipitation of the HA-tagged receptors using a mouse monoclonal anti-HA antibody, the presence of ␤ 2 AR-GFP in the immunoprecipitate was probed using a rabbit anti-GFP anti- body. Two major species corresponding to monomeric and oligomeric forms of the ␤ 2 AR were detected in both HA-␤ 2 AR/ ␤ 2 AR-GFP-and HA-␤ 1 AR/␤ 2 AR-GFP-expressing cells. In contrast, no specific GFP immunoreactivity was detected at the cell surface or in the total extract of cells co-expressing the unrelated HA-tagged GABA B -R2 receptor and the ␤ 2 AR-GFP (Fig. 1, lanes 7 and 8), confirming the selectivity of the immunoprecipitation approach. Immunoreactivity for the HA-GABA B -R2 was, however, detected in the anti-HA immunoprecipitate (data not shown), confirming that this receptor was efficiently expressed and immunoprecipitated. Taken together, these results confirm, as previously reported (25)(26)(27), that the ␤ 2 AR and ␤ 1 AR can form homodimers and indicate that stable and selective intermolecular interactions can also occur between the two receptors as well. The higher molecular weight species detected in Western analyses correspond to SDS-resistant complexes that may either represent dimers or possibly higher order oligomers such as trimers that were not fully denatured by SDS (25). The low resolution in this region of the gels and the possibility of aberrant mobility make it difficult to unambiguously distinguish between these possibilities. For the sake of simplicity, we will therefore refer to stable receptorreceptor complexes as dimers.
␤ 1 -and ␤ 2 -Adrenergic Receptors Form Homo-and Heterodimers in Living Cells-Although co-immunoprecipitation is a generally accepted method to document protein-protein interactions, the interpretation of these experiments for hydrophobic membrane proteins such as receptors is complicated by detergent solubilization that could promote artifactual aggregation. To assess whether ␤ 1 AR/␤ 2 AR heterodimers could be detected in living cells, BRET was used. This technique is a proximity assay based on the nonradiative transfer of energy between a bioluminescent donor (Rluc) and a fluorescent acceptor (GFP) that allows real time monitoring of protein-protein interaction in living cells (26,43). BRET has been used to complement biochemical approaches to studying receptor/receptor interactions for ␤ 2 AR (26), ␦-opioid (17), and thyrotropin-releasing hormone receptor (44) homodimerization. In these experiments, we have used a slight modification of the previously published assay. The new BRET 2 technology (Bio-Signal Packard Bioscience) takes advantage of the spectral properties of a distinct luciferase substrate known as Deep Blue coelenterazine (Deep Blue C), which permits a better separation between Rluc and GFP emission spectra (see "Experimental Procedures").
To assess ␤ 2 AR and ␤ 1 AR homo-and heterodimerization, fusion constructs linking the receptor carboxyl-terminal tails to either Rluc or GFP were co-transfected in HEK 293 cells, and the transfer of energy between the two partners was assessed following the addition of Deep Blue C. Upon oxidation of the luciferase substrate, the enzyme emits light with a peak at 400 nm that can excite GFP, which, in turn, re-emits fluorescence with a peak at 510 nm but only if the two partners are within the permissive distance (Ͻ100 Å). The BRET signal is determined by calculating the ratio of the light emitted by the receptor-GFP (500 -530 nm) over the light emitted by the receptor-Rluc (370 -450 nm). Fig. 2 presents the BRET level measured for the different partners considered. A strong BRET signal was detected for both ␤ 1 AR and ␤ 2 AR homodimers, and a smaller but significant BRET was also observed in cells expressing ␤ 1 AR-Rluc/␤ 2 AR-GFP or ␤ 2 AR-Rluc/␤ 1 AR-GFP, indicating that heterodimers between the receptor subtypes also form in living cells. No significant BRET was detected when either ␤ 1 AR-Rluc or ␤ 2 AR-Rluc was co-expressed with soluble GFP expressed at a similar fluorescence level as ␤ 1 AR-or ␤ 2 AR-GFP, confirming the selectivity of the detected signals. The small difference between the BRET signals observed for the homo-and heterodimers does not necessarily reflect a smaller number of heterodimers, since the level of BRET does not depend strictly on the number of dimers formed but also on the relative orientation and distance between the energy donor and acceptor within the dimers. It follows that differences in the structural organization of the dimers could account for the different BRET levels. No significant changes were detected in the measured BRET levels when the cells were stimulated with 10 M isoproterenol (data not shown).
Functional Characterization of ␤ 1 AR/␤ 2 AR Interactions-We next determined the functional consequences of ␤ 1 AR/␤ 2 AR heterodimerization by measuring receptor-mediated stimulation of several effector pathways. For this purpose, cells were transfected with the ␤ 1 AR, ␤ 2 AR, or both receptors together. Transfection conditions were such that an equivalent total number of receptors were expressed in each case (␤ 1 AR, 2.0 Ϯ 0.7; ␤ 2 AR, 1.8 Ϯ 0.5; ␤ 1 AR ϩ ␤ 2 AR, 1.5 Ϯ 0.6 pmol/mg). For cells expressing the two receptor subtypes, competition radioligand binding with the selective antagonists betaxolol and ICI-118,551 revealed equivalent proportions of ␤ 1 AR and ␤ 2 AR. In a first series of experiments, the ability of the nonselective ␤-adrenergic agonist, isoproterenol, to stimulate the adenylyl cyclase was tested. In the absence of transfected receptor, isoproterenol promoted a modest concentration-dependent increase in cAMP production that most likely reflects the presence of endogenously expressed ␤ 2 AR (ϳ0.01 pmol/mg) in HEK 293 cells (Fig. 3). Transfection of the ␤ 1 AR, ␤ 2 AR, or both receptors together led to similar increases in the agonist-stimulated adenylyl cyclase activity. EC 50  For each pair considered, the quantities of DNA used for ␤ 1 AR-Rluc and ␤ 2 AR-Rluc were selected to yield equivalent luminescent signals (370 -450 nm), whereas those for ␤ 1 AR-GFP and ␤ 2 AR-GFP were selected to obtain the maximum BRET levels. The BRET level for the soluble GFP was measured at a fluorescence level equivalent to those of ␤ 1 AR-GFP and ␤ 2 AR-GFP. Cells were harvested 48 h post-transfection, counted, and transferred to 96-well plates (100,000 cells/well). The energy transfer reaction was initiated by adding 5 M Deep Blue C coelenterazine to each well, and BRET was assessed in a BRETCount microplate reader with the filter settings described under "Experimental Procedures." The results represent mean Ϯ S.E. of 3-6 independent experiments performed in triplicate. 4), and 128 Ϯ 54 nM and 25.2 Ϯ 4.7 pmol/min⅐mg for ␤ 1 AR ϩ ␤ 2 AR (n ϭ 7), respectively. In mock-transfected cells, the EC 50 and maximal stimulated values were 171 Ϯ 179 nM and 8.1 Ϯ 5.5 pmol/min⅐mg (n ϭ 3). No differences in the ability of NaF or forskolin to stimulate adenylyl cyclase were detected whether the receptors were expressed alone or together (data not shown). Since no significant difference in either the efficacy or the potency of isoproterenol was detected, we conclude that 1) the two receptor subtypes had similar abilities to couple with the adenylyl cyclase pathway and 2) that the formation of ␤ 1 AR/␤ 2 AR heterodimers had no effect on this signaling pathway (Fig. 3).
The efficacy of each receptor subtype expressed individually or together to stimulate the MAPKs p38 and ERK1/2 was then assessed. p38 was modestly activated by the ␤ 1 AR, and this was not altered by co-expression with the ␤ 2 AR (Fig. 4E). In HEK 293 cells, we could not detect a consistent stimulation of p38 by the ␤ 2 AR alone. In contrast, a dramatic difference was observed between ␤ 1 AR and ␤ 2 AR in their ability to promote phosphorylation of ERK1/2. Indeed, as seen by other groups, agonist stimulation of the ␤ 2 AR results in a consistent activation of ERK1/2, which peaked at 5 min and returned to almost basal levels by 30 min (Figs. 4A and 5A). However, no increase in the phosphorylation of ERK1/2 could be detected upon isoproterenol treatment in cells expressing the ␤ 1 AR (Fig. 4B). This difference did not result from aberrant ERK1/2 activities in the ␤ 1 AR-expressing cells, since the phorbol ester-stimulated ERK1/2 phosphorylation was normal in these cells. Interestingly, in a similar fashion to what was observed for the ␤ 1 AR-expressing cells, no ␤-adrenergic-stimulated ERK1/2 activity could be detected in cells co-expressing both ␤ 1 AR and ␤ 2 AR (Fig. 4C). For both the ␤ 1 AR alone and ␤ 1 AR and ␤ 2 AR together, no stimulation was seen after 30 min of agonist stimulation (Fig. 5B), demonstrating that the loss of ERK1/2 stimulation shown at 5 min was not simply due to changes in the time course of receptor activation of the ERK pathway (Fig. 5). In fact, we could not detect changes in ERK1/2 activation even after 2 h of agonist stimulation when the two receptors were expressed together (data not shown). Given that the two receptors were expressed at equivalent levels (ϳ52% ␤ 1 AR), these results suggest that heterodimerization inhibited the ␤ 2 ARpromoted ERK1/2 activation. Data from several experiments are summarized in Fig. 4D. To confirm that expression of the ␤ 1 AR by itself did not reduce levels of ␤ 2 AR in the transient co-transfections, we also transfected ␤ 1 AR into a stable cell line expressing the ␤ 2 AR. As determined by confocal microscopy, the transfection efficiency for the ␤ 1 AR into the ␤ 2 AR stable cell line was always between 60 and 70% in these experiments. As shown in Fig. 5, a similar loss of isoproterenol-stimulated ERK1/2 activity was also seen upon co-expression of the two receptor subtypes. Again, no stimulation was seen when the ␤ 1 AR/␤ 2 AR were co-expressed at any of the times measured. These data are summarized in Fig. 5C.   FIG. 3. Stimulation of adenylyl cyclase by ␤ 1 AR, ␤ 2 AR, and ␤ 1 AR/␤ 2 AR. HEK 293 cells expressing ␤ 1 AR, ␤ 2 AR, or the two receptors together were assayed for adenylyl cyclase activity. Expression of both receptors was confirmed by ligand binding and were as follows Ϯ S.D.: ␤ 1 AR, 2 Ϯ 0.7 pmol/mg (n ϭ 6); ␤ 2 AR, 1.8 Ϯ 0.5 pmol/mg (n ϭ 7); and ␤ 1 AR/␤ 2 AR, 1.5 Ϯ 0.6 pmol/mg (n ϭ 7). The ratio of ␤ 1 AR to ␤ 2 AR in HEK 293 cells expressing both receptors was ϳ1:1, as determined using ␤ 2 -or ␤ 1 -selective ligand ICI 118,551 or betaxolol (␤ 1 AR, 51.2 Ϯ 3.2%; ␤ 2 AR, 48.7 Ϯ 3.2). The level of endogenous ␤ 2 AR in HEK 293 cells was 0.01 Ϯ 0.01 pmol/mg (n ϭ 3). Membranes were prepared from HEK 293 cells as described under "Experimental Procedures." Control cells expressing GFP vector (Mock) alone displayed a markedly lower stimulation (presumably due to the low levels of endogenous ␤ 2 AR in these cells) of adenylyl cyclase under the same conditions used for receptortransfected cells (data not shown). Data were normalized to forskolin stimulation and represent mean Ϯ S.E. for ␤ 1 AR (n ϭ 4), ␤ 2 AR (n ϭ 4), ␤ 1 AR/␤ 2 AR (n ϭ 7), and mock (n ϭ 3).
FIG. 4. Stimulation of ERK1/2 phosphorylation by ␤ 1 AR, ␤ 2 AR, and ␤ 1 AR/␤ 2 AR. ERK1/2 activation is shown in the upper part of each panel (as measured by anti-phospho-ERK1/2 antibodies) in response to stimulation by 10 M isoproterenol for 5 min of serum-starved HEK 293 cells transiently transfected with ␤ 2 AR (A), ␤ 1 AR (B), or the two receptors together (C). As a loading control, the total ERK1/2 pool in the lysate was measured using anti-ERK1/2 antibodies, and the anti-phospho-ERK data are normalized with respect to the total ERK pool. Expression of both receptors was confirmed by ligand binding, Western blotting, and immunocytochemistry. 200 g of cell lysate was loaded into each lane. SDS-PAGE was followed by immunoblot, developed with enhanced chemiluminescence. Representative experiments are shown. D, p42/44 ERK data from separate experiments were digitized on a flatbed scanner and analyzed using Quantity One (Bio-Rad) software and are presented here as mean Ϯ S.E. for ␤ 1 AR (n ϭ 4), ␤ 2 AR (n ϭ 3), and ␤ 1 AR/␤ 2 AR (n ϭ 4). *, significant differences (p Ͻ 0.05 using Student's t test) between control and isoproterenol-stimulated samples. E, representative example (n ϭ 3) of p38 MAPK activation (as measured using anti-phospho-p38 antibodies) by ␤ 1 AR (28 Ϯ 10% increase over basal), ␤ 2 AR (20 Ϯ 0.07% decrease from basal), or the two receptors together (35 Ϯ 20% increase over basal).
Given the distinct internalization profiles previously reported for the two receptors, the consequences of heterodimerization on the internalization of each receptor were then assessed. Confocal immunocytochemistry analysis revealed that, under control conditions, both receptors are found primarily at the plasma membrane (Fig. 6, a and b). Stimulation of the ␤ 2 AR expressing cells with isoproterenol led to a rapid internalization indicated by the disappearance of receptor from the plasma membrane that could be detected as early as 5 min poststimulation (Fig. 6A, top panel). For ␤ 1 AR-expressing cells, the same treatment had no detectable effect on the expression of the receptor at the cell surface (Fig. 6B, top panel). No significant internalization of the ␤ 1 AR was detected even following a 2-h treatment with the agonist, confirming earlier studies that indicated that the ␤ 1 AR is resistant to agonistpromoted internalization (34 -36). In cells co-expressing the two receptor subtypes, isoproterenol stimulation failed to promote significant internalization of either the ␤ 1 AR or the ␤ 2 AR (Fig. 6, A and B, bottom panels), indicating that expression of the internalization-resistant ␤ 1 AR inhibited the internalization of the ␤ 2 AR. We also performed colocalization experiments in cells transiently transfected with both ␤ 1 AR and ␤ 2 AR-GFP. Under control conditions, both receptors are expressed at the cell surface, where they colocalize (Fig. 7A). As described above for the untagged ␤ 2 AR, isoproterenol-promoted internalization was readily observed in cells only or predominantly expressing ␤ 2 AR-GFP (Fig. 7, B and C). In contrast, no such internalization of ␤ 2 AR-GFP was observed in those cells co-expressing significant amounts of the ␤ 1 AR (Fig. 7, B and C). The modest ␤ 2 AR-GFP internalization, seen in some of cells co-expressing both receptors, most likely reflects internalization of ␤ 2 AR homodimers.
To quantify the proportion of ␤ 2 AR internalized in the presence and absence of the ␤ 1 AR, we took two separate approaches using transient transfections of the ␤ 2 AR in the absence or presence of co-transfected ␤ 1 AR. First, using the hydrophilic ligand CGP, we determined that 54.5 Ϯ 4.5% of ␤ 2 AR (as measured in the presence of 5 ϫ 10 Ϫ8 M betaxolol to isolate the ␤ 2 AR signal) remained at the surface after 2 h of agonist stimulation (Fig. 8A). In the presence of the ␤ 1 AR, the amount of ␤ 2 AR that remained at the cell surface after 2 h of agonist stimulation was 108 Ϯ 25% confirming our observations made using confocal microscopy. No changes were seen in the total numbers of receptors under these conditions as determined with CYP (data not shown). To further verify these results, we also measured cell surface expression of extracellularly tagged (N terminus) ␤ 2 AR using an enzyme-linked immunosorbent assay. Again in the presence of the ␤ 1 AR, the amount of ␤ 2 AR FIG. 5. Stimulation of ERK1/2 phosphorylation by stably transfected ␤ 2 AR in absence and presence of transiently co-transfected ␤ 1 AR. ERK1/2 activation is shown in the upper part of each panel (as measured by anti-phospho-ERK1/2 antibodies) in response to stimulation by 10 M isoproterenol for various times of serum-starved HEK 293 cells stably transfected with ␤ 2 AR in the absence (A) and presence (B) of transiently co-transfected ␤ 1 AR. Expression of both receptors was confirmed by ligand binding, Western blotting, and immunocytochemistry. Estimates using confocal microscopy of transfection efficiency for transiently expressed ␤ 1 AR were between 60 and 70% using our procedures. 200 g of cell lysate was loaded into each lane. SDS-PAGE was followed by immunoblot, developed with enhanced chemiluminescence. Representative experiments are shown and were repeated at least three times. C, p42/44 ERK data from separate experiments were digitized on a flatbed scanner and analyzed using Quantity One (Bio-Rad) software and are presented here as mean Ϯ S.E. for ␤ 2 AR (n ϭ 5) and ␤ 1 AR/␤ 2 AR (n ϭ 5). *, significant differences (p Ͻ 0.07 using Student's t test) between control and isoproterenolstimulated samples. internalized is markedly reduced. When the ␤ 2 AR is expressed alone, 52 Ϯ 10% of c-myc-tagged receptors remain at the surface after 60 min of 10 M isoproterenol in comparison with 84 Ϯ 6.6% when the ␤ 1 AR is present (Fig. 8B). Taken together, these results demonstrate that ␤ 2 AR trafficking is altered in the presence of the ␤ 1 AR. DISCUSSION We have demonstrated that, in addition to forming homodimers, ␤ 1 AR and ␤ 2 AR can also form heterodimers when co-expressed in HEK 293 cells. These interactions were shown both by co-immunoprecipitation and BRET assays in living cells. Cell surface co-immunoprecipitation experiments revealed that the heterodimer was expressed at the plasma membrane. Further, an analysis of the processing of receptors, as assessed by their sensitivity to endoglycosidase H or peptide N-glycosidase F (data not shown) indicated that fully mature receptors were forming dimers.
We have demonstrated that co-expression of the two receptors leads to altered functional properties in the putative heterodimer, suggesting that the ␤ 1 AR and ␤ 2 AR interact as a heterodimer to yield a novel ␤AR subtype with unique functional properties. This is manifested by the complete loss of ERK1/2 MAPK stimulation by isoproterenol when the two receptors were co-expressed. Given that the ␤ 1 AR, when expressed alone, was also incapable of stimulating ERK1/2, our results suggest that the heterodimers display a coupling pattern characteristic of the ␤ 1 AR. Previous studies have also demonstrated that the ␤ 1 AR is less effective than ␤ 2 AR in stimulating ERK1/2 MAPK (32,33). To our knowledge, however, this is the first report indicating that cells co-expressing the two receptors may have blunted ERK1/2 responses to ␤-adrenergic stimulation.
Both the ␤ 1 AR and ␤ 2 AR are expressed in the cardiovascular system, and more particularly, both receptors are co-expressed in cardiomyocytes. One potential consequence of this cross-talk regulation of the ERK1/2 pathway through receptor heterodimerization may manifest itself in the modulation of signaling by ␤-adrenergic stimulation. For example, recent reports have demonstrated that phospholipase A 2 translocation to the plasma membrane, stimulated by the ␤ 2 AR, is an important mediator of ␤ 2 AR-mediated positive inotropy in embryonic chick cardiomyocytes (45,46). In these cells, ␤ 2 AR stimulation leads to arachidonic acid release via phospholipase A 2 , resulting in an increased release of calcium from sarcoplasmic reticulum stores. Further, these authors have also demonstrated that this ␤ 2 AR response is ERK1/2-dependent. These effects of ␤ 2 AR stimulation on intracellular free calcium and positive inotropy in embryonic chick heart myocytes might be masked by activation of ␤ 1 AR, and this possibility was considered by FIG. 7. Colocalization of transiently co-transfected ␤ 1 AR and ␤ 2 AR-GFP in HEK 293 cells in response to agonist stimulation. Confocal images were taken from fixed and permeabilized cells expressing ␤ 2 AR-GFP and ␤ 1 AR before (A) and after (B and C) stimulation for various times with 10 M isoproterenol. Images of GFP fluorescence and staining with CY5-conjugated secondary antibody following labeling with anti-␤ 1 AR. Shown in order from top to bottom are anti-␤ 1 AR signals, GFP signals, and the overlay. A representative image is shown for unstimulated cells (A), which demonstrates cell surface colocalization of both receptors, and cells 5 min (B) and 120 min (C) after stimulation with 10 M isoproterenol. Note that, in cells expressing both receptors, receptors remain colocalized at the cell surface (red arrows), whereas cells expressing only or predominantly ␤ 2 AR show rapid receptor internalization (yellow arrows).
FIG. 8. Quantification of ␤ 2 AR internalization in the absence and presence of co-transfected ␤ 1 AR. A, loss of cell surface (CGP 12177) binding to ␤ 2 AR after 120 min of stimulation at 37°C with 10 M isoproterenol was determined in the presence of 5 ϫ 10 Ϫ8 M betaxolol to isolate the ␤ 2 AR signal. Shown are mean Ϯ S.E. for percentage of cell surface ␤ 2 AR remaining after agonist stimulation in the presence and absence of co-expressed ␤ 1 AR (n ϭ 3 for both conditions). Data were compared with cell surface binding in unstimulated cells, which was normalized to a value of 100%. Samples were controlled for changes in total receptor binding as determined with CYP. B, enzyme-linked immunosorbent assay to measure cell surface ␤ 2 AR. Shown are mean Ϯ S.E. for percentage of N-terminally tagged myc-␤ 2 AR-Rluc receptors remaining at the cell surface after 60 min of agonist stimulation at 37°C in the presence and absence of co-expressed ␤ 1 AR-GFP10 (n ϭ 3 for both conditions). Data are shown relative to cells stimulated with vehicle. Total ␤AR expression levels in all of these experiments were between 1 and 3 pmol/mg of membrane protein. At these levels, the ␤ 2 AR internalizes efficiently when expressed alone, suggesting that our results with co-expression of the ␤ 1 AR are not simply due to saturation of the internalization mechanism. *, significant differences (p Ͻ 0.05 using Student's t test) between ␤ 2 AR expressed alone and ␤ 1 AR/␤ 2 AR co-expressed together.
the authors of the studies cited above (46). It is likely that this type of regulation may also occur in mammalian cardiomyocytes. It has been demonstrated that ␤AR stimulation of ERK1/2 may play a role in the development of cardiac hypertrophy (47). Up-regulation of a ␤ 2 AR-specific signal may also be important when one considers the proapoptotic effects of ␤ 1 AR stimulation and the antiapoptotic effects of ␤ 2 AR stimulation on the myocardium (32,48). This may be particularly important during the progression to heart failure when the ␤ 1 AR is selectively down-regulated, potentially unmasking and increasing the impact of signaling pathways specific to the ␤ 2 AR.
Several lines of evidence derived from transgenic and knockout mice also support the functional importance of ␤ 1 AR/␤ 2 AR heterodimerization. For example, high levels of ␤ 2 AR overexpression led to an impairment of cardiac ␤ 1 AR function, which is relieved using ␤ 2 AR inverse agonists such as ICI 118,551 (49). Also, responses to ␤ 2 -specific agonist stimulation were lost in mice with homozygous deletion of the ␤ 1 AR, although cardiac ␤ 2 AR number and distribution remained similar to wild type animals (37). 2 Taken together, these data are consistent with the notion that each receptor in some way depends on the other for function.
The fact that ␤ 1 AR and ␤ 2 AR had similar abilities to modulate adenylyl cyclase precludes determination of whether or not heterodimerization affects this pathway. Previous studies had suggested that the ␤ 1 AR had a reduced efficacy to stimulate adenylyl cyclase when compared with the ␤ 2 AR (29 -31). It was later demonstrated that this decreased efficacy was characteristic of a particular ␤ 1 AR variant (glycine at position 389), whereas the most common polymorphism (Arg 389 ) has similar efficacy as the ␤ 2 AR (38). Since we were using the Arg 389 ␤ 1 AR variant, it is not surprising that ␤ 1 AR and ␤ 2 AR had similar abilities to stimulate adenylyl cyclase. In the case of p38 MAPK, we again demonstrate that the ␤ 1 AR phenotype dominates over the ␤ 2 AR when the two receptors are co-expressed in HEK 293 cells, although ␤-adrenergic stimulation of p38 MAPK is much more modest than for ERK1/2 MAPK. We could not detect stimulation of p38 MAPK via the ␤ 2 AR alone. This may not be the case for all cell types, since stimulation of both ␤ 1 AR and ␤ 2 AR were shown to be coupled to the activation of p38 MAPK in adult murine cardiomyocytes (50,51).
We have demonstrated that the loss of p42/44 ERK MAPK stimulation in the heterodimer is accompanied by a reduced ability for the receptor to be internalized. A number of previous studies have demonstrated that internalization of the ␤ 2 AR is required for these distal signaling events to occur (16,(52)(53)(54). Further experiments will be needed to determine whether the presence of the ␤ 1 AR in the putative heterodimer is directly responsible for the alteration in receptor trafficking and ERK activation. Another recent study has shown that a ␤ 2 AR/opioid receptor heterodimer was also defective for both p42/44 ERK MAPK activation and agonist-mediated receptor internalization (16). These authors also showed that a ␤ 2 AR/␦-opioid receptor heterodimer that could be internalized could still stimulate the ERK pathway. Taken together, these results suggest that novel heterodimeric receptors can have altered trafficking itineraries, which have functional consequences.
Several studies have demonstrated that ␤ 2 AR-stimulated ERK1/2 activity requires the recruitment of ␤-arrestin, which acts as a scaffolding protein to assemble the ERK1/2 signaling cascades (see Ref. 55 for a review). The fact that the ␤ 1 AR has been shown to recruit ␤-arrestin much less efficiently than the ␤ 2 AR (35, 36) may provide a mechanistic explanation for the inhibitory effect of receptor heterodimerization on ERK1/2 activation. The heterodimer may therefore also have a reduced ability to promote the assembly of the ERK1/2 signaling complex.
Although co-expression of the ␤ 1 AR completely inhibited ␤ 2 AR-stimulated ERK1/2 MAPK activity in HEK 293 cells, ␤ 2 AR stimulation does result in MAPK activation in cardiac tissue and in isolated cardiomyocytes. Given that these cells express both ␤ 1 AR and ␤ 2 AR, there must be a mechanism of sequestering the ␤ 2 AR away from the ␤ 1 AR. One can suppose that there are specialized targeting events that exclude the ␤ 1 AR from subcellular domains where the ␤ 2 AR is found. Interestingly, agonist stimulation causes translocation of the ␤ 2 AR out of caveolae in neonatal cardiomyocytes, whereas the relative distribution of the ␤ 1 AR between caveolae and the bulk plasma membrane remains constant (56,57). This may provide a mechanism for physical segregation of the ␤ 2 AR, allowing it to escape dominant negative regulation by the ␤ 1 AR, which in cardiomyocytes may outnumber the ␤ 2 AR by a factor of 3.
Putative functional importance for GPCR heterodimerization is not limited to the ␤ 1 AR/␤ 2 AR. For example, obligate heterodimerization is required for signaling via GABA B receptors (see Ref. 58 for a review) and for mammalian taste receptors (8,9). Another example of cross-talk between signaling pathways involving heterodimeization is provided by the observation that AT 1 and AT 2 angiotensin II receptor association negatively regulates signaling via AT 1 receptors (59). Alterations in ligand binding properties have also been attributed to the formation of heterodimers. For instance, co-expression ofand ␦-opioid receptors (14) or of D2 and D3 dopamine receptors (60) leads to novel ligand binding properties not seen when either single receptor was expressed alone.
In summary, our results demonstrate that ␤AR subtypes can form heterodimers with altered functional properties. Given the wide distribution of ␤ 1 AR and ␤ 2 AR, it will be critical to assess the importance of heterodimerization in the trafficking, subcellular distribution, and signaling of each receptor subtype in native tissues. In view of the increasing number of reports of GPCR heterodimerization, cross-talk between receptor subtypes that involves oligomerization may represent a general regulatory mechanism.