Multiple Scaffolding Functions of β-Arrestins in the Degradation of G Protein-coupled Receptor Kinase 2*

G protein-coupled receptor kinase 2 (GRK2) plays a fundamental role in the regulation of G protein-coupled receptors (GPCRs), and changes in GRK2 expression levels can have an important impact on cell functions. GRK2 is known to be degraded by the proteasome pathway. We have shown previously that β-arrestins participate in enhanced kinase turnover upon GPCR stimulation by facilitating GRK2 phosphorylation by c-Src or by MAPK or by recruiting the Mdm2 E3 ubiquitin ligase to the receptor complex. In this report, we have investigated how such diverse β-arrestin scaffold functions are integrated to modulate GRK2 degradation. Interestingly, we found that in the absence of GPCR activation, β-arrestins do not perform an adaptor role for GRK2/Mdm2 association, but rather compete with GRK2 for direct Mdm2 binding to regulate basal kinase turnover. Upon agonist stimulation, β-arrestins-mediated phosphorylation of GRK2 at serine 670 by MAPK facilitates Mdm2-mediated GRK2 degradation, whereas c-Src-dependent phosphorylation would support the action of an undetermined β-arrestin-recruited ligase in the absence of GPCR activation. The ability of β-arrestins to play different scaffold functions would allow coordination of both Mdm2-dependent and -independent processes aimed at the specific modulation of GRK2 turnover in different signaling contexts.

G protein-coupled receptor kinase 2 (GRK2) 3 is a ubiquitous member of the GRK family, which has an important role in the modulation of a variety of membrane receptors, largely belonging to the G protein-coupled receptor (GPCR) superfamily (1,2). Agonist-triggered phosphorylation of GPCRs by GRK2 primes the recruitment of regulatory proteins termed ␤-arrestins, leading to receptor desensitization from G proteins and GPCR internalization (3,4). In addition, GRK2 and ␤-arrestins can act as agonist-regulated scaffolds and contribute to assemble macromolecular signalosomes in the receptor environment, therefore initiating G-protein-independent signal transduction pathways. In particular, ␤-arrestin1 and ␤-arrestin2 have been shown to be multifunctional adaptors able to interact with many signaling molecules such as c-Src, components of MAPK cascades, or cAMP-phosphodiesterase-4, among others (2,4,5).
The functional role of GRK2 is not limited to promoting ␤-arrestin binding to activated GPCRs. GRK2 can phosphorylate a growing number of nonreceptor substrates and associate with a variety of proteins related to signal transduction (6,7), and it has recently been shown to play unanticipated cellular roles in epithelial cell migration (8), modulation of the MEK/ERK interface (9), inhibition of TGF-␤-mediated cell growth arrest and apoptosis (10), or regulation of cell cycle progression (11,12).
The complexity of the GRK2 interactome and its central role in many signaling pathways suggest that altered GRK2 protein expression and/or activity may have profound effects on cell signaling and physiological functions. Interestingly, this kinase is up-regulated in several oncogenic signaling contexts (10,13,14) and in heart failure or hypertension (2,15), whereas it is down-regulated in inflammatory diseases such as rheumatoid arthritis (16). A tightly regulated GRK2 degradation by the proteosome pathway has been identified as a major mechanism for modulating kinase expression levels. We have shown that GRK2 is rapidly proteolyzed in a proteasome-dependent manner and that GRK2 ubiquitination and turnover are enhanced by ␤-adrenergic receptor (␤ 2 AR) activation (17), through a mechanism involving GRK2 phosphorylation by c-Src or MAPK in a ␤-arrestin-dependent manner (18,19). More recently, we have found that Mdm2, an E3 ubiquitin ligase involved in the control of cell growth and apoptosis, plays a key role in GRK2 degradation (14). Consistent with the previously reported ␤-arrestin/Mdm2 interaction (20,21), Mdm2 and GRK2 association and subsequent kinase proteolysis appeared to be facilitated by the ␤-arrestin scaffold function upon ␤ 2 AR stimulation (14).
␤-Arrestins are emerging as central adaptors involved in the degradation of many cellular proteins. In addition to Mdm2, different E3 ubiquitin ligases can be scaffolded by ␤-arrestins to promote ubiquitination of other proteins present in the same multimolecular complexes (22). Moreover, ␤-arrestins can also serve as adaptors for deubiquitinating enzymes (23,24), thus being able to orchestrate the dynamic and transient ubiquitination of ␤2AR (24)  Because ␤-arrestins appear to be able to favor GRK2 degradation in response to GPCR stimulation by facilitating its phosphorylation by c-Src (18) or by MAPK (19), or by promoting GRK2/Mdm2 association (14), we decided to investigate how such diverse scaffold functions are integrated to control GRK2 turnover. We find that Src, MAPK and Mdm2 activities are independently scaffolded by ␤-arrestins to promote GRK2 degradation upon GPCR activation, whereas in basal conditions ␤-arrestins and GRK2 compete with each other for direct binding to Mdm2, thus modulating the extent of GRK2 degradation in such setting. Our results put forward an unforeseen complexity in the scaffold function of ␤-arrestins, which are able to integrate diverse signaling molecules and different ligases to modulate GRK2 turnover in different conditions. S]cysteine labeling mixture was obtained from NEN Life Science Products, and isoproterenol was from Sigma. The cDNAs encoding mutant ␤-arrestin1-V53D and ␤-arrestin2-V54D-GFP were provided by Dr. M. G. Caron (Duke University), and wild-type ␤-arrestin1-FLAG by Dr. R. J. Lefkowitz. The ␤-arrestin1-3A and ⌬7 mutants were obtained from Dr. V. V. Gurevich (Vanderbilt University, Nashville, TN). The ␤-arrestin-1-C-tail (amino acids 319 -418) construct was provided by Dr. J. L. Benovic (Thomas Jefferson Cancer Institute, Philadelphia, PA) and subsequently modified by the addition of a FLAG epitope at the COOH terminus. Expression vectors for GST-Mdm2-wt and GST-Mdm2 100 -491 constructs were generously provided by Dr. S. K. Shenoy (Duke University Medical Center, Durham, NC). The cDNAs encoding the constitutively active Y527F c-Src mutant and the active mutant construct of Raf (Raf*) lacking an NH 2 -terminal regulatory domain were provided by Dr. J. S. Gutkind (National Institutes of Health, Bethesda, MD) and Dr. J. Moscat, respectively. The source of other expression plasmids coding for wild-type Mdm2, ␤ 2 AR, ␤-arres-tin1, ␤-arrestin1 mutants S412A and S412D, wild-type GRK2, and mutants GRK2-S670A, GRK2-S670D, GRK2-Y13A,Y86A,Y92A and GRK2-Y86D,Y92D has been previously reported (8,14,17,18). All other reagents were of the highest grade commercially available.

EXPERIMENTAL PROCEDURES
Cell Culture and Transfection-HEK-293 and MEFs (either wild-type or Mdm2/p53 or ␤-arrestin1/␤-arrestin2 doublenull) were cultured in Dulbecco's modified Eagle's medium (DMEM) plus 10% (v/v) fetal bovine serum (FBS; Sigma-Aldrich) at 37°C in a humidified 5% atmosphere. Cells grown to 80% confluence in 60-mm dishes were transfected with the indicated plasmids (5 g of total DNA) by using the Lipofectamine (HEK-293) or the FuGENE 6 (MEFs) reagents. Empty vector was added as needed to keep the total amount of transfected DNA constant.
In Situ Protein Degradation Assay-Protein degradation was monitored by metabolic labeling and pulse-chase analysis as described (18). HEK-293 cells were labeled and chased in the absence of serum. When indicated, 35 S-labeled cells were challenged during the chase with 10 M isoproterenol as reported (17). At different chase times (1-3 h), cells were harvested using radioimmune precipitation assay lysis buffer (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% sodium deoxycholate, 0.5% Nonidet P-40, 0.1% SDS, with a mixture of protease inhibitors). Protein extracts were immunoprecipitated with the specific GRK2 polyclonal antibody AbFP1 or with the GRK2 monoclonal antibody clone C5/1.1 (Upstate Biotechnology) as reported (18). Immunoprecipitates were resolved by SDS-PAGE and transferred to PVDF membranes to be treated with the Enhancer Autoradiography Starter kit (EABiotech Ltd.) according to the manufacturer's protocol to amplify radioactive signal before development. Band density of 35 S-labeled GRK2 was quantified by laser densitometry analysis, and data were normalized to total GRK2 protein detected by immunoblotting with specific GRK2 antibodies and compared with values obtained at 0 h of chase.
Immunoprecipitation and Western Blot Analysis-Cells were lysed in buffer A (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 10% glycerol, with a mixture of protease inhibitors) for 1 h at 4°C and then centrifuged (15,000 ϫ g, 10 min). Supernatants were incubated with a specific agarose-conjugated anti-Mdm2 monoclonal antibody (SMP14; Santa Cruz Biotechnology), with a specific polyclonal antibody directed against both ␤-arrestin1 and ␤-arrestin2 proteins (Ab186) (18) or with an anti-GRK2 monoclonal antibody (clone C5/1.1). Immune complexes were resolved in 7.5% SDS-PAGE and transferred to nitrocellulose membranes. Blots were probed with a specific affinitypurified rabbit polyclonal antibody raised against GRK2 (C-19) (Santa Cruz Biotechnology), with the ␤-arrestin antibody Ab186 or with an Mdm2 polyclonal antibody (N-20; Santa Cruz Biotechnology). After membrane stripping, the same blots were reprobed with the corresponding antibodies directed against the immunoprecipitated proteins. Lysate aliquots were analyzed by immunoblotting to assess cellular protein expression levels using the anti-GRK2, anti-Mdm2, and anti-␤-arrestin antibodies described above, the anti-FLAG monoclonal antibody (Sigma), or an affinity-purified goat polyclonal antibody raised against actin (I-19; Santa Cruz Biotechnology). Blots were developed using an enhanced chemiluminescence method (ECL; Amersham Biosciences). Band density was quantified by laser densitometric analysis. In co-immunoprecipitation experiments, the amount of coprecipitated protein was normalized to that of the immunoprecipitated protein, as assessed by the specific antibodies.
GST Pulldown Assays-For in vitro association, 50 ng of recombinant GRK2 was incubated with 50 ng of purified Mdm2-wt or Mdm2 100 -491 GST-fusion proteins in 40 l of binding buffer B (20 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Tween 20, 2 mM EDTA, 10% glycerol, and 0.05 mg/ml BSA) for 3 h at 4°C in the presence or absence of 50 -100 ng of recombinant ␤-arrestin1 protein. After the addition of glutathione-Sepharose 4B, precipitated complexes were washed and eluted with SDS sample buffer. Bound GRK2 or ␤-arres-tin1 was analyzed by immunoblotting using specific antibodies.

RESULTS
We have previously described that ␤-arrestins would favor agonist-stimulated GRK2/Mdm2 association by recruiting Mdm2 to the vicinity of GRK2 and that Mdm2 is instrumental for both basal and agonist-induced GRK2 turnover (14). To define further the contribution of ␤-arrestins to the Mdm2-dependent degradation of GRK2, we first analyzed GRK2 protein decay by pulse-chase assays in the presence of ␤-arrestin2-V54D, a point mutant that fails to interact with Mdm2, and of ␤-arrestin1-V53D, a construct harboring an equivalent mutation within the proposed Mdm2 binding site (21). Interestingly, overexpression of ␤-arrestin1-V53D completely blocked the turnover of GRK2 both in basal conditions and upon receptor activation (94 Ϯ 13% and 96 Ϯ 6% of protein remaining after 1 h of chase, respectively (see Fig. 1A), whereas overexpression of ␤-arrestin2-V54D at similar levels abrogated the isoproterenol-induced increase in GRK2 proteolysis but had no effect on the basal turnover of the kinase (Fig. 1B). The presence of extra wild-type ␤-arrestin1 ( Fig.  1A) or ␤-arrestin2 (Fig. 1B) had no effect on GRK2 degradation, indicating that endogenous ␤-arrestin levels are suffi-cient to allow basal or receptor-stimulated GRK2 degradation, as reported previously (18).
The unexpected finding that a comparable mutation in the highly homologous ␤-arrestins resulted in a different functional effect on basal GRK2 turnover led us to examine the interaction of ␤-arrestin1-V53D with Mdm2. In sharp contrast to the reported defective binding of ␤-arrestin2-V54D to Mdm2 (21) ( Fig. 2A), we found that the V53D mutant associated with Mdm2 more efficiently than the wild-type protein (Fig. 2B). Interestingly, the extent of GRK2/Mdm2 co-immunoprecipitation was markedly diminished in the presence of ␤-arrestin1-V53D, whereas wild-type ␤-arrestin1 or ␤-arres-tin2-V54D expressed at similar levels promoted either a slight reduction or even an increase in this parameter, respectively (Fig. 3, A and B).Therefore, we hypothesized that the blockade of basal GRK2 turnover in the presence of mutant ␤-ar-restin1-V53D might be due to its unusual capability to form tight complexes with Mdm2, thus sequestering the ligase activity required for GRK2 degradation. Consistent with this notion, we found that expression of extra Mdm2 rescues the impaired degradation of GRK2 caused by ␤-arrestin1-V53D (96 Ϯ 11% of protein remaining after 1 h of chase in the absence of Mdm2 versus 51 Ϯ 9% in its presence), whereas basal turnover of the kinase is accelerated in the presence of the non-Mdm2-competing ␤-arrestin2-V54D mutant (Fig. 3C).
Overall, these results suggested that, in the absence of GPCR activation, ␤-arrestins would not be playing a typical scaffold role for GRK2 and Mdm2, but rather that GRK2 and ␤-arrestins would "compete" for the cytoplasmic pool of Mdm2. To test this possibility, we conducted co-immunoprecipitation assays in cells co-expressing GRK2, Mdm2, and increasing amounts of transfected ␤-arrestin1 (Fig. 4A). In line with our reasoning, the amount of Mdm2-bound GRK2 inversely correlated with ␤-arrestin1 levels, pointing that FIGURE 1. GRK2 turnover is differentially affected by ␤-arrestin1-V53D or ␤-arrestin2-V54D point mutants. HEK-293 cells were transiently transfected with GRK2 and ␤ 2 AR in the presence or absence of either wild-type or mutant constructs (V53D and V54D) for ␤-arrestin1 (A) or ␤-arrestin2 (B), respectively. GRK2 stability under basal conditions and upon receptor stimulation with isoproterenol (ISO) was determined by pulse-chase experiments as detailed under "Experimental Procedures." 35 S-Labeled proteins immunoprecipitated with the anti-GRK2 antibody AbFP1 were resolved by SDS-PAGE followed by fluorography and densitometry. Labeled GRK2 band densities were then normalized to total GRK2 present in the immunoprecipitates, as determined by immunoblot analysis. Lysates from transfected cells were subjected to immunoblot analysis with a specific ␤-arrestin antibody to monitor equal overexpression of wild-type and mutant proteins. Data are mean Ϯ S.E. (error bars) of three independent experiments performed in duplicate. *, p Ͻ 0.05; †, p Ͻ 0.05; † †, p Ͻ 0.01, comparing GRK2 decay in unstimulated versus isoproterenol-treated conditions. Representative gel autoradiographies are shown.
GRK2 and arrestins cannot be simultaneously in complex with Mdm2. ␤-Arrestin2 overexpression displayed a similar competition effect (Fig. 4B).We next asked whether association of GRK2 with Mdm2 could occur in the absence of ␤-arrestins. Pulldown assays with recombinant GRK2 protein and purified GST-Mdm2 indicated that both proteins interact directly (Fig. 4C). Interestingly, a Mdm2 construct encompassing amino acids 100 -491 was also able to interact with GRK2 (Fig. 5A), whereas a fragment consisting of the RING domain of the ligase did not (data not shown), thus preliminarily pointing to the central region of Mdm2 (amino acids 100 -430), also reported to be involved in the interaction with ␤-arrestin (20, 21) as the GRK2-interacting interface. Addition of recombinant ␤-arrestin1 resulted in a clear reduction of Mdm2-associated GRK2 (Fig. 4D), thereby suggesting that binding of arrestins to Mdm2 can either mask or compete with the region involved in GRK2 association. In keeping with the ability of GRK2 to interact directly with Mdm2, co-immunoprecipitation of endogenous GRK2 and Mdm2 in basal conditions was observed in MEFs lacking ␤-arrestin1 and 2 expression (Fig. 4E). Notably, the agonist-induced increase in GRK2/Mdm2 association detected in wild-type MEFs was absent in ␤-arrestin-deficient cells (Fig. 4E), in agreement with a scaffold role for these proteins only in such a condition (14).
Overall, our results suggested that ␤-arrestin1-V53D would exert a "dominant negative" effect with respect to basal Mdm2-dependent GRK2 degradation. In contrast, ␤-arres-tin2-V54D (unable to bind Mdm2) would not alter the interaction of Mdm2 with GRK2 or with endogenous ␤-arrestins, thus allowing GRK2 turnover in basal conditions. To address further the interplay of arrestins and Mdm2 in GRK2 stability,  we assessed the effects of different ␤-arrestin1 mutants specifically displaying altered interactions with GPCR or Mdm2 (25). ␤-Arrestin1-⌬7, a seven-residue deletion mutant in the interdomain linker that does not interact with receptors but strongly associates with Mdm2, impaired both basal and agonist-induced degradation of GRK2, similar to the V53D mutant (Fig. 5A). In contrast, the ␤-arrestin1-3A mutant, which constitutively interacts with GPCR but binds poorly to Mdm2, mimics the effect of the V54D mutant, blocking only the receptor-stimulated turnover of GRK2. These results indicated that recruitment of Mdm2 by ␤-arrestins is a key event in the receptor-induced GRK2 degradation but not in its basal turnover.
Interestingly, both V53D and V54D mutants display other altered functions in addition to Mdm2 association that also could hamper the agonist-induced degradation of GRK2. These ␤-arrestin mutants bind weakly to agonist-stimulated receptors (26,27) and have been reported to block receptormediated MAPK activation (28). Therefore, these mutants would be expected to be inefficient in recruiting to the receptor complex the c-Src and ERK activities previously reported to be required for GRK2 phosphorylation and the agonistinduced increase in GRK2 turnover (18,19). In line with this notion, we found that a mutant ␤-arrestin-1construct (␤-arrestin1-C-tail), the expression of which has been shown to impair agonist-promoted Src/MAPK activation (29) and which lacks the Mdm2-binding region, abrogates the agonistinduced degradation of GRK2 but does not affect its basal turnover (Fig. 5B), mimicking the effects of the ␤-arrestin2-V54D mutant. Remarkably, ␤-arrestin1-S412D, a Src-binding-defective mutant that impairs the agonist-induced decay of GRK2 protein (18), can interact with Mdm2 as efficiently as wild-type ␤-arrestin or a S412A mutant (Fig. 5C), neither of which alters GRK2 proteolysis (18). These results suggest that the adaptor functions of ␤-arrestin relating to Mdm2 and c-Src recruitment can be dissociated and are both required for an efficient agonist-triggered GRK2 degradation.
We next addressed whether ␤-arrestin-mediated phosphorylation of GRK2 by c-Src and/or MAPK (18,19) could favor the recognition of GRK2 by Mdm2 or help to bring the kinase in close proximity to the ligase. No significant changes were observed in the interaction of Mdm2 with different GRK2 mutants unable to be phosphorylated at defined residues (GRK2-Y13F,Y86F,Y92F, GRK2-S670A) or mimicking such modifications (GRK2-Y86D,Y92D, GRK2-S670D), compared with the wild-type kinase (Fig. 5D). However, we found that overexpression of Mdm2 was unable to promote the degradation of GRK2-S670A, a protein resistant to agonist-induced proteolysis (19), whereas it clearly stimulated (Fig. 5E) the turnover of GRK2-Y13F,Y86F,Y92F, a mutant that displays an altered basal and agonist-induced turnover (18). Therefore, MAPK-mediated phosphorylation of GRK2 at Ser 670 seems to facilitate Mdm2 function, whereas GRK2 tyrosine phosphorylation is not required for the action of the ligase. Interestingly, overexpression of a mutant construct of Raf (Raf*) that promotes a robust MAPK activation and Ser 670 phosphoryla- . C, GRK2 can interact directly with Mdm2 in pulldown assays. Recombinant GRK2 was incubated with GST alone or the indicated GST-Mdm2 fusion proteins as described under "Experimental Procedures." Proteins bound to glutathione-Sepharose beads were detected with specific anti-GRK2 and anti-GST antibodies. Remaining unbound proteins were monitored in parallel. D, ␤-arrestin1 competes Mdm2/GRK2 binding in vitro. Pulldown assays were performed as above in the presence or absence of purified ␤-arrestin1. Free and proteins bound to the glutathione-Sepharose beads were inmunodetected with specific antibodies. In C and D, experiments were performed at least three times with similar results. E, knockdown of both ␤-arrestin1 and 2 proteins does not prevent GRK2/Mdm2 association. Lysates from wild-type or ␤-arrestin1/2-deficient MEFs, stimulated or not with the adrenergic agonist isoproterenol, were subjected to immunoprecipitation with Mdm2 antibodies. The amount of GRK2 present in the immune complexes was detected and quantified as in previous figures. Data are mean Ϯ S.E. of 3 independent experiments. **, p Ͻ 0.01 for the indicated comparison. Representative co-immunoprecipitation and total cell lysate expression gels are shown.

Role of ␤-Arrestins in GRK2 Degradation
tion of GRK2 (19) notably restores the impaired degradation of GRK2 in the presence of ␤-arrestin1-V53D (48 Ϯ 4% of protein remaining after 3 h of chase compared with 99 Ϯ 11% in the absence of Raf* (Fig. 6A). Therefore, phosphorylation of GRK2 by MAPK can rescue the "inefficient targeting" of Mdm2 to GRK2 caused by the presence of the competing ␤-arrestin1-V53D mutant.
On the other hand, expression of a constitutively active c-Src mutant (c-Src Y527F) known to promote GRK2 phosphorylation efficiently in tyrosine residues and to enhance degradation of wild-type GRK2 markedly in normal cellular settings (18,19), was able to induce GRK2 turnover in MEFs lacking Mdm2 expression, whereas activation of the MAPK pathway did not (Fig. 6B). Overall, these results suggested that in response to GPCR activation, the ␤-arrestin scaffold would facilitate Mdm2-mediated GRK2 turnover by favoring both the recruitment of the ligase to the vicinity of the receptor and by promoting MAPK phosphorylation of GRK2 at Ser 670 . In this regulatory context, the ␤-arrestin-triggered phosphorylation of GRK2 by c-Src serves to facilitate the subsequent action of MAPK (19), resulting in enhanced GRK2 proteolysis. The fact that GRK2 tyrosine phosphorylation appears not to be required "per se" for this Mdm2-mediated, agonist-induced GRK2 degradation pathway but can stimulate basal kinase turnover in the absence of Mdm2 strongly argues for the existence of additional GRK2 ligases. Remarkably, the slow degradation rate of GRK2 in MEFs lacking Mdm2 can be rescued by extra ␤-arrestin1 expression (Fig. 6C), similar to the effect observed upon increased c-Src activation as discussed above (see Fig. 6B). Interestingly, overexpression of ␤-arrestin1 in the absence of Mdm2 is not sufficient to restore the agonist-dependent increase in GRK2 degradation (Fig.  6C), despite receptor stimulation is able to promote GRK2-pS670 phosphorylation in these cells (data not shown). These data put forward Mdm2 as the main ligase mobilized upon GPCR activation and suggest that receptor-triggered, ␤-arrestin-mediated phosphorylation of GRK2 at tyrosine and serine residues by c-Src and MAPK cooperate to enhance Mdm2mediated GRK2 turnover. In addition, an alternative, ␤-arrestin and c-Src-triggered GRK2 degradation pathway, can also take place in basal conditions in the absence of Mdm2.

DISCUSSION
The agonist-induced turnover of GRK2 has been reported to require the efficient recruitment of the Mdm2 E3 ligase and the serine and tyrosine phosphorylation of GRK2, a process that appears to render this protein more suitable for ubiquitination and degradation. GPCR stimulation would position FIGURE 5. Different activities affecting GRK2 degradation are engaged independently by ␤-arrestins. A, ␤-arrestin mutants with different Mdm2-scaffolding abilities alter GRK2 turnover in distinct ways. HEK293 cells were co-transfected with GRK2, ␤ 2 AR, and either wild-type or ␤-arrestin1-⌬7 and -3A mutants. GRK2 stability was determined in the presence or absence of the agonist isoproterenol by pulse-chase experiments as in previous figures. B, agonist-induced GRK2 turnover is hampered by a ␤-arrestin mutant unable to recruit Mdm2, c-Src, and MAPK activities. HEK-293 cells were transiently co-transfected with GRK2, ␤ 2 AR, and wild-type or mutant ␤-arrestin1 (amino acids 319 -418) and GRK2 protein decay in basal and stimulated conditions determined as above. Data are mean Ϯ S.E. (error bars) of three independent experiments performed in duplicate. *, p Ͻ 0.05 and **, p Ͻ 0.01 comparing agonist-challenged cells expressing wild-type ␤-arrestin versus those expressing the mutant. †, p Ͻ 0.05 and † †, p Ͻ 0,01 comparing vehicle-and agonisttreated cells. C, c-Src-scaffolding function of ␤-arrestin1 is not required for the interaction of Mdm2 with ␤-arrestin. HEK-293 cells were transiently transfected with Mdm2 and either wild-type ␤-arrestin1 or mutant S412D and S412A constructs that display impaired and unaffected c-Src binding, respectively. Co-immunoprecipitation (IP) of Mdm2 and ␤-arrestin was determined as in previous figures. IB, immunoblotting. D and E, phosphorylation of GRK2 at Ser 670 is necessary for the Mdm2-mediated degradation of the kinase but not for the interaction with Mdm2. Association of GRK2 with Mdm2 (D) and GRK2 stability (E) were analyzed by means of co-immunoprecipitation and pulse-chase assays, respectively, in HEK-293 cells transiently transfected with wild-type GRK2 or the indicated GRK2 phosphorylation site mutants in the presence or absence of Mdm2. Data (mean Ϯ S.E. (error bars)) from three or four independent experiments are shown. *, p Ͻ 0.05; **, p Ͻ 0.01, for the indicated comparisons. Control conditions refer to cells expressing wild-type proteins.
Mdm2, c-Src, and MAPK in the proximity of the receptor-or G␤␥-associated GRK2, either as a result of their engagement in ␤-arrestin-orchestrated multiprotein complexes or by allowing the simultaneous, timely local accumulation of these arrestin-binding partners to the vicinity of the plasma membrane (14,18,19). In this report we show that ␤-arrestins play different scaffold functions to orchestrate both Mdm2-dependent and -independent processes involved in the control of GRK2 stability in basal conditions and upon GPCR stimulation.
We found that, in the absence of GPCR activation, the association of Mdm2 with GRK2 and with ␤-arrestins seems to be mutually exclusive, indicating that ␤-arrestins do not display an adaptor role in such conditions, but rather compete with GRK2 for Mdm2 binding. Several lines of evidence support this notion. First, we observed a diminished GRK2 Mdm2 co-immunoprecipitation in the presence of increasing amounts of co-expressed ␤-arrestin1 or 2, thus pointing to a competition between GRK2 and ␤-arrestins for binding to Mdm2. This effect contrasts with the "bridge" scaffolding function of ␤-arrestins in linking the ligase to different targets for ubiquitination (30 -32). Consistent with this interpretation, different ␤-arrestin1 mutants that interact strongly with Mdm2 have a marked inhibitory effect on GRK2/Mdm2 association (V53D mutant) and a dominant negative effect on GRK2 degradation in basal conditions (V53D and ⌬7 mutants), which can be rescued upon expression of extra Mdm2. In this regard, it is striking that similar mutations in ␤-arres-tin1 and 2 (V53D and V54D, respectively) have such different effects on ␤-arrestin/Mdm2 binding (see supplemental text for an extended discussion on this matter).
Second, we unveiled that Mdm2 can associate with GRK2 in a ␤-arrestin-independent manner because these proteins can be co-immunoprecipitated from MEFs lacking expression of both ␤-arrestin isoforms. Moreover, recombinant GRK2 can interact directly with purified GST-Mdm2 in pulldown experiments, whereas addition of purified ␤-arrestin1 protein in these assays inhibits GRK2/Mdm2 association. Different regions of Mdm2 (from residues 383 to 410 (21) and from positions 161 to 321 (20)) have been found to be important for ␤-arrestin binding. Our preliminary data suggest that the occurrence of overlapping binding sites in Mdm2 possibly underlies the competition between GRK2 and ␤-arrestins for Mdm2 association.
We propose that in the absence of GPCR stimulation, the pool of Mdm2 bound to GRK2 would support the basal turnover of the kinase as inferred by the incomplete blockade of GRK2 degradation in cells devoid of ␤-arrestins (what might be a surrogate of receptor inactivation in terms of kinase degradation) (14). In this context, cytosolic ␤-arrestins would interfere with GRK2/Mdm2 binding by competing for the ligase, which might help to maintain enough GRK2 protein within the cell to respond adequately to a receptor challenge. The fact that receptor-unbound ␤-arrestins redistribute Mdm2 to microtubules might also contribute to keep the ligase away from GRK2 (33). On the other hand, in response to acute or sustained receptor activation, co-recruitment of ␤-arrestin-Mdm2 complexes will allow the local delivery of extra Mdm2 to GRK2 to reduce kinase levels, thereby limiting the extent of desensitization to preserve signaling receptor competence toward later challenges (19). Consistent with this key role for ␤-arrestins in the control of the disassembly or assembly of GRK2-Mdm2 complexes in different signaling contexts, Mdm2 preferentially binds to arrestins in their basal inactive state (34), suggesting that arrestins might be "preloaded" with the ligase in a "standby" state prior to their receptor-induced activation. The apparent paradox that ␤-arrestins serve as adaptors for Mdm2-dependent degradation of GRK2 in GPCR-stimulated conditions, but as competitors in the absence of stimulus can be reconciled assuming that only the agonist-triggered conformational activation of receptorbound arrestins could locally release Mdm2-competent and/or additional partners to interact with and ubiquitinate different targets as suggested previously (34).
In addition to the recruitment of Mdm2, ␤-arrestins are able to act as signaling adaptors for c-Src and MAPK/ERK upon GPCR stimulation, and these activities have been reported to facilitate receptor-dependent degradation of GRK2 (18,19). We find that these ␤-arrestin activities cooperate to enhance GRK2 turnover in different ways. Overall, our data suggest that ␤-arrestin-mediated c-Src tyrosine phosphorylation can drive GRK2 degradation independently of Mdm2 in basal conditions, whereas upon receptor activation it "primes" MAPK-mediated phosphorylation of GRK2 at Ser 670 (19) to facilitate Mdm2-mediated GRK2 proteolysis. In this regard, it is worth noting that agonist-induced GRK2 turnover is inhibited in the presence of different ␤-arrestin scaffolding mutants. ␤-Arrestin1-S412D, ␤-arrestin1-V53D, or ␤-arrestin 2-V54D, and the ␤-arrestin1-C-tail fragment (residues 319 -418), are unable to recruit c-Src to the receptor complexes for different reasons (26 -29, 34), thus impairing GRK2 tyrosine and serine phosphorylation (19). Some of these mutants interact with Mdm2 as efficiently as the wild-type proteins (S412D), whereas others do not interact at all (V54D or ␤-arrestin-C-tail), suggesting that the scaffolding roles of ␤-arrestins on Mdm2 and on c-Src or MAPK can be dissociated. Conversely, ␤-arrestin1-3A, which is able to interact with both receptors and Src/MAPK but not with Mdm2 (25) also blocks agonist-induced decay of GRK2 despite its efficient phosphorylation on Ser 670 (data not shown), thus pointing to Mdm2 as the final effector in this process.
The fact that the delayed degradation of GRK2 in Mdm2deficient cells in basal conditions can be similarly accelerated by overexpression of ␤-arrestin1 or of a constitutively active c-Src mutant strongly supports the occurrence of an alternative route for GRK2 turnover in the absence of GPCR activation, likely based on additional E3 ubiquitin ligases recruited by a specific ␤-arrestin cellular pool, that would preferentially target tyrosine-phosphorylated GRK2. Conversely, when such a pathway is not operative, as upon expression of the tyrosinephosphorylation and proteolysis-defective mutant GRK2-Y13F,Y86F,Y92F, Mdm2 overexpression can markedly enhance GRK2 degradation.
On the other hand, the defective GRK2 turnover caused by ␤-arrestin1-V53D expression can be rescued in the presence of a constitutively active Raf protein, which promotes GRK2 phosphorylation at Ser 670 (19). Notably, such enforced phosphorylation of GRK2 at Ser 670 fails to promote GRK2 degradation in the absence of Mdm2, and overexpressed Mdm2 is unable to rescue the retarded turnover of a GRK2 mutant with impaired phosphorylation at Ser 670 (GRK2-S670A), indicating that Ser 670 phosphorylation is key to facilitate the action of Mdm2 on GRK2. This modification could induce a conformational change that would favor GRK2 ubiquitination by Mdm2 and/or help to bring the kinase in close proximity to the ligase, as has been recently suggested for the ERK-triggered degradation of FOXO3a by Mdm2 (35).
Given the multifactor positive role of ␤-arrestins as GPCRdependent co-factors in GRK2 degradation, it could be hypothesized that changes in ␤-arrestin cellular levels would inversely correlate with GRK2 expression. Interestingly, in several pathophysiological conditions such as cerebral hypoxia/ischemia (36) or in the hypothyroid liver (37), GRK2 downregulation (mainly taking place at the protein level) is accompanied by an up-regulation of ␤-arrestin expression. An inverse correlation between GRK2 and arrestin protein levels has also been described in other hypothyroid organs (37) and in macrophages sustainedly stimulated through Toll-like receptors (38). However, a simultaneous increase in GRK2 and ␤-arrestin levels is observed after myocardial infarction (39), in Parkinson disease (40) or during acquisition of analgesic tolerance in the brain (41), indicating that such functional interplay is more complex. Interestingly, up-regulation of GRK2 protein caused by dopamine depletion in animal models of Parkinson disease is reverted upon restoration of receptor signaling with L-dopa treatment (42). It should be then considered that, in principle, enhanced ␤-arrestin levels would favor GPCR-promoted GRK2 degradation, whereas in the absence of GPCR activation its role as a cytosolic Mdm2quenching factor would be potentiated, leading to opposite effects.
Signaling pathways triggered by non-GPCR receptors might also have an effect on GRK2 turnover by altering the scaffolding ability of ␤-arrestins. Activation of tyrosine kinase receptors such as the insulin receptor causes phosphorylation of ␤-arrestin1 at position Ser 412 (43) which would hamper c-Src recruitment, thus potentially leading to inhibition of the ␤-arrestin/GRK2 tyrosine phosphorylation-dependent degradation pathway (18). It is tempting to suggest that these mechanisms may cooperate in promoting GRK2 cellular accumulation upon increased stimulation of certain growth factor receptors (14).
Finally, our data also put forward the possibility that Mdm2 may also regulate GRK2 turnover in a ␤-arrestin-independent manner. Interestingly, we have recently described an active GRK2 protein decay during the G 2 /M cell cycle transition by a mechanism involving the pS670-dependent recruitment of the prolyl-isomerase Pin1 and subsequent degradation (11). Such GRK2 down-regulation, which is critical for timely cell cycle progression, also takes place in cells lacking both ␤-arrestins. Noteworthy, the Mdm2 ligase is specifically up-regulated during G 2 /M transition (44) in parallel to enhanced GRK2 phosphorylation at Ser 670 (11). Because we found here that Ser 670 phosphorylation can facilitate Mdm2-dependent GRK2 degradation, it is tempting to suggest that Mdm2 is the ligase involved in GRK2 down-regulation during the cell cycle.
In summary, our results show that ␤-arrestins can act as positive or negative modulators of GRK2 stability in different signaling contexts, by differentially scaffolding Mdm2 and possibly other ligase(s) as well. The mechanisms by which ␤-arrestins can direct Mdm2 or other E3 ligases toward a particular target among a broad repertoire of potential ubiquitination substrates in response to specific signals remains to be investigated further.