Feedback inhibition of G protein-coupled receptor kinase 2 (GRK2) activity by extracellular signal-regulated kinases.

G protein-coupled receptor kinase (GRK)-mediated receptor phosphorylation and beta-arrestin binding uncouple G protein-coupled receptors (GPCRs) from their respective G proteins and initiates the process of receptor internalization. In the case of the beta(2)-adrenergic receptor and lysophosphatidic acid receptor, these processes can lead to ERK activation. Here we identify a novel mechanism whereby the activity of GRK2 is regulated by feedback inhibition. GRK2 is demonstrated to be a phosphoprotein in cells. Mass spectrometry and mutational analysis localize the site of phosphorylation on GRK2 to a carboxyl-terminal serine residue (Ser(670)). Phosphorylation at Ser(670) impairs the ability of GRK2 to phosphorylate both soluble and membrane-incorporated receptor substrates and dramatically attenuates Gbetagamma-mediated activation of this enzyme. Ser(670) is located in a peptide sequence that conforms to an ERK consensus phosphorylation sequence, and in vitro, in the presence of heparin, ERK1 phosphorylates GRK2. Inhibition of ERK activity in HEK293 cells potentiates GRK2 activity, whereas, conversely, ERK activation inhibits GRK2 activity. The discovery that ERK phosphorylates and inactivates GRK2 suggests that ERK participates in a feedback regulatory loop. By negatively regulating GRK-mediated receptor phosphorylation, beta-arrestin-mediated processes such as Src recruitment and clathrin-mediated internalization, which are required for GPCR-mediated ERK activation, are inhibited, thus dampening further ERK activation.

G protein-coupled receptor kinases (GRKs) 1 constitute a family of six mammalian serine/threonine protein kinases that phosphorylate agonist-occupied G protein-coupled receptors (reviewed in Ref. 1). GRK-phosphorylated GPCRs bind stoichiometrically to inhibitory proteins known as arrestins. The binding of ␤-arrestin1 or -2, extraretinal isoforms, prevents receptor-mediated G protein activation, recruits other molecules such as Src to the receptor, and serves to target the phosphorylated receptor for internalization via clathrin-coated pits (2). The recruitment of Src and engagement of the clathrincoated pit system by the phosphorylated ␤ 2 -adrenergic receptor (␤ 2 AR) have been shown to be essential for ␤ 2 AR-mediated ERK activation (3). The GRKs thus play a critical role in regulating GPCR-mediated signal transduction. Following agonist exposure, GRK-mediated receptor phosphorylation leads to an attenuation of G protein-mediated signaling (receptor desensitization) while at the same time initiating receptormediated ERK activation.
The activity of the GRKs is tightly regulated via a number of mechanisms. These include regulation by the ␤␥ subunits of heterotrimeric G proteins (G␤␥), lipids (phosphatidylinositol 4,5-bisphosphate and phosphatidylserine), and cytoskeletal (actin) and calcium-binding proteins (recoverin, calmodulin), as well as by protein kinase C (reviewed in Ref. 1). In most cases regulation of the GRKs is subfamily-specific. Here we identify a previously unappreciated and apparently specific mechanism by which GRK2 activity is regulated. GRK2 is demonstrated to exist as a phosphoprotein in cells. Mapping the site of phosphorylation, elucidating the functional consequences of this phosphorylation event, and identifying the GRK2 kinase in cells yields new insight into the regulatory mechanisms controlling GPCR-mediated signal transduction.
Construction of Wild Type and Mutant GRK2 cDNAs-Using polymerase chain reaction, a 5Ј Kozak sequence was added to bovine GRK2 and inserted into the EcoRI/BamHI sites of pRK5 to form the wild type construct pRK5-GRK2 wildtype. The carboxyl-terminal Ser was mutated to Ala or Asp to form pRK5-GRK2(S670A) and pRK5-GRK2(S670D).
ERK1-mediated Phosphorylation of GRK2-Purified phosphorylated or unphosphorylated GRK2 (2 M) was incubated with purified ERK1 (30 nM) in 40 mM HEPES, pH 7.5, 5 mM MgCl 2 , 2 mM dithiothreitol containing 60 M [␥-32 P]ATP (ϳ1000 cpm/pmol) in a final volume of 25 l. Incubations were performed Ϯ heparin (7 g/ml) at 30°C for 45 min. The stoichiometry of ERK-mediated GRK2 phosphorylation was determined by PhosphorImager analysis (Molecular Dynamics). When the activity of ERK-phosphorylated GRK2 was assessed, radiolabeled ATP was omitted and reactions were terminated by desalting on D-Salt Polyacrylamide 6000 Plastic Desalting columns (Pierce) to remove heparin and ATP. GRK2 remaining in solution following the desalting procedure was determined by quantitative Western blot using anti-GRK2 antibodies (10). A 32 P-labeled phosphorylation reaction was performed alongside nonradioactive reactions to assess the stoichiometry of ERK-mediated GRK2 phosphorylation.
GRK2-mediated ␤ 2 AR Phosphorylation in Whole Cells-32 P-labeled FLAG-tagged ␤ 2 AR was immunoprecipitated from transiently transfected HEK293 cells (12). Cells were treated with staurosporin (1 M) during the 32 P i labeling period and with the ␤ 2 AR agonist (isoproterenol, 10 M) for 5 min prior to harvest. 32 P-labeled ␤ 2 AR was immunoprecipitated from cells transiently transfected with GRK2 and a dominant-negative or constitutively active MEK construct. The phosphorylation status of ERK was determined by Western blot analysis using an anti-phosphoERK antibody (Promega) (13).

RESULTS AND DISCUSSION
GRK2 Is a Phosphoprotein in Cells-GRK2 was purified from baculovirus-infected Sf9 cells using previously published procedures (4). The recombinant protein appeared pure by Coomassie Blue staining, but isoelectric focusing revealed two distinct major bands and a number of minor associated bands. GRK2 purified enzyme was resolved into two peaks of GRK2 protein and activity on Source-S (Fig. 1). Notably, as compared with the first peak (P1), the second peak (P2) exhibited a higher specific activity.
Mass spectrometric analysis revealed that P1 contained a major and a minor species of GRK2 with molecular mass values of 79,777 and 79,646 Da, respectively, whereas P2 contained a single species of 79,690 Da. Because the predicted molecular mass of bovine GRK2 is 79,656 Da, and amino-terminal sequencing of the protein failed, P2 appears to correspond to amino-terminally acylated, full-length GRK2. The major P1 species is larger than the mass of P2 by the mass of one phosphate. The minor P1 species appears to lack the aminoterminal methionine with respect to the major species. P1 and P2 GRK2 were both subjected to tryptic digestion, and the peptides from each were analyzed on a MALDI-TOF mass spectrometer. By comparing these spectra, a single candidate peptide (Met 664 -Lys 677 ) was found whose mass appeared augmented by 80 Da in P1 versus P2. This peptide (MKNKPRSPV-VELSK) contains two phosphorylatable residues, Ser 670 and Ser 676 . Thus P1 and P2 likely differ by virtue of a single phosphorylation event at either Ser 670 or Ser 676 . Notably, Ser 670 resides in a sequence that conforms to an ERK consensus phosphorylation sequence, i.e. PX(S/T)P. Indeed, the peptide sequence immediately surrounding Ser 670 in GRK2 is identical to the ERK phosphorylation site in c-ELK (PRSP) (14).
Phosphorylated GRK2 Is Less Active Than Its Dephosphorylated Counterpart-Fractions corresponding to P1 or P2 were assessed for their ability to phosphorylate purified reconstituted ␤ 2 AR or the soluble GRK2 substrate tubulin. As shown in Table I, P1 exhibited a higher affinity for substrate than P2 as reflected in the lower K m of phosphorylated GRK2 for ␤ 2 AR and tubulin. The maximal rate of substrate phosphorylation (V max ) was, however, significantly lower for P1 than P2. The most profound difference was observed when ␤ 2 AR, in the presence of G␤␥ subunits, was utilized as substrate. Under these conditions P1 had an approximately 8-fold lower V max than P2 (Table I).
We next examined the dose dependent activation of the two GRK2 species by G␤␥. As shown in Fig. 2A, P1 exhibits dramatically impaired G␤␥-mediated activation of GRK2. These results reveal a profound functional consequence of GRK2 phosphorylation, an impairment of the rate of GRK2-mediated phosphorylation, and further suggest that this modification may serve to attenuate receptor-mediated allosteric activation of GRK2.
Identification of Ser 670 as the Site Phosphorylated on GRK2-To demonstrate that Ser 670 represents a site of phosphorylation, two mutant forms of GRK2 were constructed. In these mutants one of the candidate sites of phosphorylation, S 670 , was mutated to either an alanine (GRK2(S670A)) or an aspartic acid residue (GRK2(S670D)). If S 670 indeed represents the site of GRK2 phosphorylation, it would therefore be predicted that mutating this site to alanine would produce a form of the kinase that behaves like unphosphorylated GRK2. Conversely, mutating this site to an aspartic acid residue would be predicted to produce an enzyme that mimics phosphorylated GRK2. If Ser 676 , and not Ser 670 , is a site of phosphorylation, then mutation of Ser 670 would be predicted not to affect the phosphorylation status of GRK2. The two mutant forms of GRK2 were expressed in and purified from Sf9 cells. The chromatographic profile of these enzymes on Source-S and their responsiveness to G␤␥ subunits when rhodopsin was used as a substrate was subsequently examined. GRK2(S670A) and GRK2(S670D) chromatographed as single peaks on Source-S eluting at 100 mM NaCl for GRK2(S670A) and 70 mM NaCl for GRK2(S670D). These elution positions are identical to P2 and P1, respectively, observations consistent with the identification of Ser 670 as the site of GRK2 phosphorylation. Furthermore, the G␤␥ sensitivity of GRK2(S670A) and GRK2(S670D) is es-  for ␤ 2 AR and tubulin Purified reconstituted ␤ 2 AR (0.3-0.9 M) and tubulin (0.3-9.0 M) were phosphorylated as previously described (11). Phosphorylation reactions were performed for 10 min at 30°C in either the presence or absence of G␤␥ (100 nM Regulation of GRK2 Activity by ERKs 34532 sentially identical to that observed using the unphosphorylated (P2) and phosphorylated (P1) forms of wild type GRK2 (compare Fig. 2, panels A and B). These results identify Ser 670 , a potential site of ERK phosphorylation, as the GRK2 residue phosphorylated in Sf9 cells.
ERK Phosphorylates GRK2 in Vitro-To determine whether ERK may indeed be the enzyme responsible for phosphorylating GRK2 in cells, the ability of purified ERK1 to phosphorylate unphosphorylated or phosphorylated GRK2 was examined in vitro. ERK1 phosphorylated unphosphorylated but not phosphorylated GRK2. However, this phosphorylation occurred only when heparin, an inhibitor of GRK2, was present in the phosphorylation reaction (Fig. 3A, compare lanes 5 and 6). These results demonstrate that ERK1 can phosphorylate GRK2 in vitro and that this phosphorylation occurs on Ser 670 , because GRK2 phosphorylated at this site is not a substrate for ERK1 (Fig. 3A, compare lane 6, upper and lower panels).
There are several potential explanations for the requirement of heparin in these in vitro phosphorylation reactions. GRK2 may phosphorylate and inactivate ERK1. ERK1-mediated GRK2 phosphorylation would thus be predicted to occur only when GRK2 activity is inhibited, i.e. in the presence of heparin. Alternatively, ERK1-mediated GRK2 phosphorylation may be conformation-dependent. Heparin may mimic a natural ligand of GRK2 that upon binding exposes Ser 670 . GRK2 has no effect on ERK1 activity as assessed by the ability of ERK1 to phosphorylate myelin basic protein in the presence or absence of GRK2 (data not shown). Additionally, ERK1 fails to phosphorylate GRK2 in the presence of heparin when GRK2 is denatured. These results suggest that the conformation of GRK2 is critical for ERK1-mediated phosphorylation (data not shown).
The nature of the putative ligand required for ERK-mediated GRK2 phosphorylation in cells remains to be elucidated. Known regulators of GRK2 activity including, G␤␥ and/or PIP 2 , phosphatidylserine, and Ca 2ϩ /calmodulin failed to support ERK1-mediated GRK2 phosphorylation (data not shown).
The ERK1-mediated phosphorylation of GRK2 observed in vitro would be predicted to alter the functional properties of this enzyme if ERK1 phosphorylates Ser 670 and phosphorylation at this site is responsible for attenuating the G␤␥ sensitivity of this enzyme. Unphosphorylated GRK2 incubated in the presence of heparin in either the presence or absence of ERK1 was subjected to gel filtration to remove heparin, and its ability to phosphorylate rhodopsin in the presence of varying concentrations of G␤␥ was examined. As shown in Fig. 3B ERK1-mediated phosphorylation of GRK2 dramatically impairs GRK2 activity and sensitivity to G␤␥ activation. These results again support the assignment of Ser 670 as the phosphorylated residue in GRK2 and suggest that under the appropri-ate conditions ERK1 is capable of mediating this phosphorylation event.
The requirement of heparin in the in vitro ERK1 phosphorylation assays raises the question of whether this kinase is responsible for phosphorylating GRK2 in cells. To examine a potential role for ERKs in phosphorylating and regulating GRK2 activity in cellular systems, we examined the effect of modulating ERK activity on GRK2-mediated ␤ 2 AR phosphorylation.
Modulating ERK Activity in Cells Regulates GRK2 Function-HEK293 cells transiently transfected with FLAG-tagged ␤ 2 AR, and GRK2 were additionally transfected with either a dominant-negative or a constitutively active form of ERK kinase-1 (MEK1). MEK1 is a dual specificity threonine/tyrosine kinase that phosphorylates and activates ERK1 and -2. Transfection of dominant-negative (MEK(K/A)) or constitutively active (MEK(DD)) forms of MEK thus regulates ERK activity in cells. Transfected HEK293 cells were labeled with 32 P i and the ability of GRK2 to phosphorylate ␤ 2 AR assessed by immuno-

Regulation of GRK2 Activity by ERKs 34533
precipitation of 32 P-labeled receptor following agonist treatment. Transfection of dominant-negative MEK (MEK(K/A)) and inhibition of ERK activity led to an approximately 2.5-fold increase in GRK2-mediated ␤ 2 AR phosphorylation (Fig. 4). In contrast, transfection of a constitutively active MEK (MEK(DD)) construct and ERK activation mildly attenuated GRK2-mediated ␤ 2 AR phosphorylation (Fig. 4). These results are consistent with the identification of ERK as the kinase responsible for phosphorylating GRK2 in cells. Inhibition of the GRK2 kinase would be predicted to increase the amount of unphosphorylated GRK2 present in this cellular system and thus to increase GRK2 activity. In contrast, activation of the GRK2 kinase would be predicted to increase the amount of phosphorylated GRK2 impairing GRK2-mediated phosphorylation events. The high basal level of GRK2 phosphorylation observed in unstimulated cells (Fig. 1) may provide an explanation for the relatively small effect of MEK(DD) transfection on GRK2 activity (Fig. 4).
The discovery that GRK2 exists as a phosphoprotein in cells provides a striking analogy between GRK2 and ␤-arrestin1 (13,15). ␤-Arrestin1 binds to GRK-phosphorylated GPCRs and serves to uncouple the phosphorylated receptor from G proteins and to target it to clathrin-coated pits for internalization. ␤-Ar-restin1, like GRK2, exists as a phosphoprotein in cells (13). Indeed, ERK has recently been shown to phosphorylate ␤-ar-restin1 in cells (15). Dephosphorylation of ␤-arrestin1 occurs at the plasma membrane following agonist occupancy of GPCRs and is required for the high affinity interaction of ␤arrestin1 and clathrin (13). Thus, two proteins required for mediating the internalization of agonist-occupied GPCRs, GRK2 and ␤-ar-restin1, are ERK substrates. Engagement of the clathrincoated pit endocytosis machinery has been shown to be required for ERK activation by the ␤ 2 AR (3). By phosphorylating and regulating the activity of two key players in this internalization process, ERK can thus be envisioned as participating in a negative feedback regulatory loop.
It is of interest to note that ERK phosphorylation at Ser 412 is specific to ␤-arrestin1, because this residue is not conserved in other members of the arrestin family. Similarly Ser 670 of GRK2 is not conserved in GRK3. It will be of interest to determine whether phosphorylation by ERK is a regulatory mechanism specific to GRK2 and ␤-arrestin1 or if other members of these protein families are regulated by ERK phosphorylation. The observation that ␤-arrestin1 is dephosphorylated at the plasma membrane following agonist occupancy of GPCRs suggests, by analogy, that a similar mechanism for dephosphorylating GRK2 may exist. These two proteins may be dephosphorylated by the same phosphatase. Although several questions remain concerning the exact mechanism of ERK-mediated GRK2 phosphorylation, the elucidation of a potential role for ERK in regulating this enzyme provides novel insights into the mechanisms controlling GPCR-mediated signal transduction. FIG. 4. Modulating ERK activity in cells regulates GRK2 activity. A, GRK2-mediated ␤ 2 AR phosphorylation was monitored in HEK293 cells by immunoprecipitation of FLAG-tagged ␤ 2 AR as described under "Experimental Procedures." Cells were transfected with empty vector (EV), GRK2, GRK2 ϩ MEK (K/A), or GRK2 ϩ MEK(D/D) and stimulated with ϩ/Ϫ isoproterenol (Iso) as indicated. An autoradiograph representative of two separate experiments is shown in the upper panel. The lower panel shows a Western blot in which 25 g of cell lysate from each transfection condition is probed with an anti-phos-phoERK antibody. B, a quantitative representation of receptor phosphorylated following GRK2, GRK2 ϩ MEK(K/A) and GRK2 ϩ MEK(D/D) overexpression. Agonist-dependent receptor phosphorylation observed in the cells transfected with empty vector has been subtracted from each condition, and the receptor phosphorylation obtained in the presence of GRK2 transfection only arbitrarily set to 100%. Results represent the average of two experiments.