Direct Binding of Activated c-Src to the b 3 -Adrenergic Receptor Is Required for MAP Kinase Activation*

Both b 2 - and b 3 -adrenergic receptors (ARs) are able to activate the extracellular signal-regulated kinase (ERK) pathway. We previously showed that c-Src is required for ERK activation by b 2 AR and that it is recruited to activated b 2 AR through binding of the Src homology 3 (SH3) domain to proline-rich regions of the adapter protein b -arrestin1. Despite the absence of sites for phosphorylation and b -arrestin binding, ERK activation by b 3 AR still requires c-Src. Agonist activation of b 2 AR, but not b 3 AR, led to redistribution of green fluorescent pro- tein-tagged b -arrestin to the plasma membrane. In b -ar-restin-deficient COS-7 cells, b -agonist-dependent co-precipitation of c-Src with the b 2 AR required exogenous b -arrestin, but activated b 3 AR co-precipitated c-Src in the absence or presence of b -arrestin. ERK activation and Src co-precipitation with b 3 AR also occurred in adi- pocytes in an agonist-dependent and pertussis toxin-sensitive manner. Protein interaction studies show that the b 3 AR interacts directly with the SH3 domain of Src through proline-rich motifs (P XX P) in the third intracellular loop and the carboxyl terminus. ERK activation and Src co-precipitation

During the past several years, transmembrane signaling traffic through G protein-coupled receptors (GPCRs) 1 has grown from the classic G protein effectors such as adenylyl cyclase and phospholipases to include novel mechanisms for activation of mitogen-activated protein (MAP) kinase cascades. These signaling systems typically involve receptor and nonreceptor tyrosine kinases as scaffolds and intermediaries (1)(2)(3)(4)(5)(6). An example of this flexibility in GPCR signaling includes the ␤ 2 -adrenergic receptor (␤ 2 AR). Although this receptor is classically known to couple to Gs and stimulate adenylyl cyclase, it can also activate the ERK1/2 MAP kinase pathway (7,8). In some cell types, the ␤ 2 AR activates ERK through its coupling to a PTX-sensitive G i protein and subsequent Ras-dependent MAP kinase activation (7,9), whereas in other systems this occurs in a PTX-independent and a G s -and cAMP-dependent process (10,11) through the activation of Rap1 (11).
In exploring the mechanisms of ␤ 2 AR-stimulated MAP kinase activation, we have found that some of the same signaling molecules required for receptor desensitization can also be intimately involved in the activation of the MAP kinase cascade. Following agonist activation, most GPCRs are phosphorylated by GPCR kinases (GRKs), with subsequent binding of ␤-arrestin to the phosphorylated receptor serving to interdict G protein coupling and signal transduction (5,12,13). However, in addition to its role in desensitization, ␤-arrestin can also participate in the events leading to MAP kinase activation. Binding of ␤-arrestin1 to the agonist-activated ␤ 2 AR rapidly recruits c-Src to the receptor (12,14). This recruitment appears to be mediated by an interaction between the amino-terminal proline-rich region of ␤-arrestin1 and the SH3 domain of c-Src (13,15).
The ␤ 3 AR is a member of the ␤AR subfamily of GPCRs that is expressed predominantly in adipocytes. Because selective ␤ 3 AR agonists have been shown to prevent or even reverse obesity and diabetes in various animal models (16 -18), increased attention has been focused upon the molecular and physiological regulation of this receptor as a therapeutic target (19). Early studies of ␤-adrenergic stimulation of adenylyl cyclase in adipocytes by Rodbell and colleagues (20) indicated the presence of a PTX-sensitive component. In examining this issue, we showed that this effect is due to the presence of the adipocyte-specific ␤ 3 AR and its ability to simultaneously couple to both G s and G i , leading to the activation of the cAMP-dependent protein kinase A and ERK1/2 pathways, respectively (9). Because GRK-mediated phosphorylation is necessary for ␤-arrestin binding (reviewed in Ref. 21), but the ␤ 3 AR lacks sites for phosphorylation (22), we concluded that the ␤ 3 AR must employ a novel mechanism of ERK activation. Here, we demonstrate that conserved proline-rich motifs in the third intracellular loop and carboxyl terminus of the ␤ 3 AR directly recruit c-Src in a ␤ 3 AR agonist-and PTX-sensitive manner. This interaction occurs specifically through the SH3 domain of c-Src. Our findings establish a new mechanism whereby some GPCRs can acquire ligand-induced tyrosine kinase activity by means of direct recruitment of Src kinases.
* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
GST fusion proteins of c-Src containing either the SH3 and SH2 domains (GST-SH3/SH2) or the SH2 domain alone (GST-SH2) were prepared as described previously (23).
Confocal Fluorescence Microscopy-Confocal microscopy was performed on a Zeiss LSM510 laser scanning microscope using a Zeiss 63X 1.4 numerical aperture water immersion lens, and fluorescent signals were collected (14). Three independent experiments were performed, and 10 or more fields/sample were analyzed in each experiment.
Protein-Protein Interaction and Co-precipitation Assays-The biotinylated fusion proteins for the wild-type or mutated Xa-2-␤ 3 AR Loop3 or Xa-2-␤ 3 AR Tail were expressed and purified (Promega). For in vitro binding, 2 g of purified biotinylated fusion protein was mixed with 4 g of GST-Src-SH3/2 or GST-Src-SH2 fusion proteins and incubated in phosphate-buffered saline containing 1% bovine serum albumin, 5% glycerol, and 0.1% Nonidet P-40. After 16 h at 4°C, the reactions were terminated by washing the immobilized complexes with 30 volumes of ice-cold phosphate-buffered saline containing 5% glycerol and 0.1% Nonidet P-40. The supernatant was removed, and 50 l of 2ϫ SDSpolyacrylamide gel electrophoresis sample buffer was added to each reaction. The biotinylated proteins were resolved by 4 -20% SDS-acrylamide gradient gel electrophoresis (Novagen), transferred to nitrocellulose membranes, and identified by staining with streptavidin-conjugated alkaline phosphatase. Immunoprecipitations of HA-tagged ␤ 2 AR and ␤ 3 AR from intact cells and immunoblotting for co-precipitated proteins were performed as described previously (25).

RESULTS AND DISCUSSION
Activation of the Ras-dependent ERK cascade by many GPCRs requires Src kinase activity (13,25). For the ␤ 2 AR, ERK activation depends on the delivery of ␤-arrestin-bound c-Src to the receptor (13). However, it is unclear whether other GPCRs utilize this same mechanism. As shown in Fig. 1, ␤ 3 ARmediated ERK activation similarly requires Src kinase activity, as demonstrated by its concentration-dependent sensitivity to the Src-specific tyrosine kinase inhibitor, PP2. Complete inhibition was achieved within the range of 1 to 5 M; a concentration previously established to selectively inhibit Src kinase (8,26). Inhibition of ␤ 3 AR-mediated ERK activation was also observed when the C-terminal Src kinase was co-expressed with ␤ 3 AR (data not shown). Because GRK-mediated phosphorylation of receptors is necessary for ␤-arrestin binding (21), but the ␤ 3 AR is not phosphorylated by GRKs, we hypothesized that agonist stimulation of ␤ 3 AR would not lead to ␤-arrestin binding. This hypothesis is confirmed as illustrated in Fig. 2, which compares the effects of agonist stimulation on the cellular distribution of a chimeric ␤-arrestin 2-GFP in HEK-293 cells expressing either the human ␤ 2 AR or the mouse ␤ 3 AR. Isoprenaline (10 M) stimulation of the ␤ 2 AR promotes the rapid translocation of ␤-arrestin 2-GFP from a diffuse cytosolic distribution to the plasma membrane where it aggregates with the receptor in membrane-associated puncta (14). In contrast, stimulation of cells expressing the mouse ␤ 3 AR with the selective ␤ 3 AR agonist CL316,243 (5 M) fails to induce ␤-arrestin 2-GFP translocation. Thus, although ␤ 3 AR-stimulated ERK activation is Src-dependent, similar to the ␤ 2 AR, the ␤ 3 AR response occurs without the formation of complexes between ␤ 3 AR and ␤-arrestin.
We previously showed that c-Src interacts with proline-con-taining motifs in the ␤-arrestin amino terminus and the SH3 domain of c-Src, although the c-Src catalytic domain also contributes significantly to this binding (13). Interestingly, although the ␤ 3 AR does not recruit ␤-arrestin, all species homologues of this receptor contain highly conserved proline residues in both the third intracellular domain and the carboxyl terminus that are completely absent from the ␤ 2 AR. Two of these proline clusters in each domain contain the sequence PXXP, which represents the minimal consensus motif for SH3 domain binding (27)(28)(29). We therefore tested the hypothesis that these proline-rich motifs within the ␤ 3 AR might directly recruit SH3 domain-containing proteins to the receptor, obviating the need for ␤-arrestin to function as an adapter protein.
First, we determined whether ␤ 3 AR could directly recruit Src kinases to activate the ERK pathway in the absence or presence of over-expressed ␤-arrestin in COS-7 cells, which express little endogenous ␤-arrestin (30). As shown in Fig. 3 (lanes 1-3), agonist treatment resulted in the detectable co-precipitation of c-Src with the ␤ 2 AR only in the presence of co-expressed ␤-arrestin. As expected from earlier studies (13), the provision of ␤-arrestin dramatically enhanced ␤ 2 AR-mediated ERK phosphorylation under these conditions. In contrast, the expression of ␤-arrestin had no effect on responses mediated by the ␤ 3 AR. Robust ␤ 3 AR agonist-dependent co-precipitation of c-Src and ERK1/2 activation was observed, which did not require the presence of ␤-arrestin (lanes 4 -6). Src that co-precipitated with ␤ 2 AR and ␤ 3 AR was also in its activated (dephosphorylated) state.
Second, because the physiological site of expression of the ␤ 3 AR is the adipocyte, a key question is whether the ␤ 3 AR directly recruits Src kinase in adipocytes as observed in COS-7 cells. We performed similar co-precipitation experiments in the mouse white adipocyte cell line, C3H10T1/2 (31). As shown in Fig. 4A, by day 4 of differentiation, C3H10T1/2 cells express the adipocyte-specific genes ␤ 3 AR and the fatty acid-binding protein aP2 (32,33). Fig. 4B shows that the ␤ 3 AR-selective agonist CL316,243 is capable of triggering ERK activation in both nontransfected (NT) cells (via the endogenous ␤ 3 AR), and in the HA-m␤ 3 AR transfected cells, but ERK activation was abolished by inactivation of G i with PTX. Fig. 4C shows that, as observed in COS-7 cells, Src kinase co-precipitates with the HA-␤ 3 AR in C3H10T1/2 adipocytes, and this interaction is both agonist-and G i -dependent. These results taken together indicate that the ␤ 3 AR can mediate the ␤-arrestin-independent recruitment of c-Src. To address whether binding between proline-rich motifs in the ␤ 3 AR and the c-Src SH3 domain might be responsible for this interaction, we first tested whether peptides derived from these regions of the ␤ 3 AR, as shown in Fig. 5A, would bind to GST-Src fusion proteins in vitro. As shown in Fig. 5B, biotinylated fusion proteins derived from both the third intracellular domain (Loop3) and the carboxyl terminus (Tail) of the wild-type ␤ 3 AR bound to the GST-Src SH3/SH2 but not to the GST-Src SH2 fusion protein. There was no interaction with GST alone. Consistent with an SH3 domain-mediated interaction, mutants of the ␤ 3 AR Loop3 (L1) and Tail (T1) peptides in which Ser was substituted for Pro in the PXXP motifs were no longer able to interact with the GST-Src SH3/SH2 peptide. In addition, we found that the wild-type ␤ 3 AR Loop3 and Tail peptides could precipitate endogenous c-Src and Grb2 from whole cell lysates of HEK-293 cells (not shown). Collectively, these data suggest that proline-rich motifs in both Loop3 and the Tail of the ␤ 3 AR possess the capacity to bind SH3 domains.
To assess the functional role of the ␤ 3 AR proline-rich motifs in Src-dependent activation of the ERK cascade, mutant ␤ 3 ARs were constructed in which these motifs were disrupted by sitedirected mutagenesis. Mouse ␤ 3 ARs containing the L1, L2 (a deletion of Loop3 amino acids Ser-242 to Pro-266), or T1 mutations were expressed in COS-7 cells and assayed for the ability to co-precipitate endogenous c-Src and to induce ERK1/2 phosphorylation and cAMP production. As shown in Fig. 6, mutation or deletion of the PXXP motifs in either the third intracellular loop or the carboxyl terminus resulted in a striking dissociation of ␤ 3 AR-mediated c-Src binding and ERK activation from ␤ 3 AR-mediated stimulation of adenylyl cyclase. Fig. 6A shows that endogenous c-Src co-precipitated with wildtype ␤ 3 AR in an agonist-dependent manner, whereas co-precipitation of c-Src with each of the mutant receptors was markedly impaired. Similar effects were observed for ␤ 3 ARmediated ERK1/2 activation, which was abolished by the L1, L2, and T1 mutations (Fig. 6C). In contrast, ␤ 3 AR-mediated production of cAMP was completely unaffected by disruption of the PXXP motifs (Fig. 6E).
These data suggest that the ability of the ␤ 3 AR to mediate Src-dependent activation of the ERK cascade, but not its ability to interact with G s protein and stimulate the cAMP pathway, is dependent upon the integrity of the PXXP motifs in the third intracellular domain and the carboxyl terminus of the receptor. Each of these motifs is sufficient to bind to Src-GST SH3 domains in vitro, and site-directed mutagenesis of either motif prevents the agonist-dependent formation of complexes between the ␤ 3 AR and c-Src when the mutant receptors are expressed in intact cells. The apparent necessity for intact PXXP motifs in both the third intracellular loop and the carboxyl terminus of the ␤ 3 AR for agonist-dependent Src co-precipitation and functional ERK activation is not clear, but it implies that a novel multimeric complex containing at least activated c-Src and two domains of the ␤ 3 AR is formed on the receptor itself to trigger this signaling cascade. However, the exact nature of this complex is unknown, and the stoichiometry of the receptor-Src interaction will require further mutagenesis and structural analysis.
The distinct strategies employed by the ␤ 2 AR and ␤ 3 AR to recruit Src kinases illustrate the flexibility that characterizes the mechanisms of G protein coupling, agonist-induced receptor sequestration, and activation of tyrosine protein kinases by heptahelical receptors. In the case of the ␤ 2 AR, G protein coupling and FIG. 2. Stimulation of ␤ 2 AR, but not ␤ 3 AR, results in recruitment of ␤-arrestin 2-GFP to the plasma membrane. Cells expressing HA epitope-tagged ␤ 2 AR (upper panel) or ␤ 3 AR (lower panel) and ␤-arrestin 2-GFP chimera were stimulated for 10 min with 10 M isoproterenol or 5 M CL316,243, respectively. The subcellular distribution of ␤-arrestin 2-GFP before (left panels) and after (right panels) agonist stimulation was determined by laser confocal immunofluorescence microscopy (13). Frames shown are from 1 of 3 independent experiments.
FIG. 3. ␤-Arrestin is required for the association of c-Src with ␤ 2 AR but not with ␤ 3 AR. COS-7 cells were transfected with c-Src and HA epitope-tagged ␤ 2 AR or ␤ 3 AR with or without ␤-arrestin. Transfected cells were treated with 5 M isoprenaline (ISO) for 2 min or 5 M CL316,243 for 5 min, as indicated. Immunoprecipitation (IP) of HAtagged receptors was performed, and co-precipitated c-Src was detected by immunoblotting with specific antisera for total c-Src and activated (Tyr530 dephosphorylated) Src, as described previously (13). Aliquots of whole cell lysates were resolved in parallel and immunoblotted for ERK1/2 (9). The data shown are from 1 of 3 independent experiments.
FIG. 4. ␤ 3 AR binds to c-Src in a G i protein-and agonist-dependent manner in differentiated C3H10T1/2 adipocytes. C3H10T1/2 cells were transfected with HA-␤ 3 AR and differentiated as described under "Experimental Procedures." Immunoprecipitation assays (IP) were performed with anti-HA antibody. A, expression of aP2 and ␤ 3 AR mRNA as a function of differentiation on the indicated days. Levels of aP2 are maximal by day 4. Cyclophilin RNA (Cyclo) is a control used in Northern blotting (17,40). B, the level of phosphorylated ERK1/2 (ERK1/2-P) in cell lysates and quantification of 3 independent experiments (mean Ϯ S.D). C, the level of c-Src co-precipitated with ␤ 3 AR and quantification of 3 independent experiments (mean Ϯ S.D.).
ERK activation occur sequentially. Src recruitment requires ␤-arrestin binding, an event that simultaneously terminates receptor-G protein coupling and triggers removal of the receptor from the cell surface. In contrast, the ␤ 3 AR is expressed in adipocytes where it is stimulated by norepinephrine in response to the requirement for fuel mobilization and heat generation (34). Given its physiologic role, there is little apparent need for ␤ 3 AR desensitization, particularly in times of chronic stimulation such as exposure to cold. There is increasing evidence that the coincident activation of the cAMP-dependent protein kinase A and MAP kinase pathways could have important consequences for energy balance (35)(36)(37)(38), given the potent and robust effects of ␤ 3 AR agonists in vivo (17,39). Inclusion of SH3 domain binding motifs within the intracellular domains of the receptor incorpo-rates the adapter protein role of ␤-arrestin within the receptor, thus allowing Src recruitment and ERK activation to proceed independently of receptor sequestration.