Phosphorylation of GRK2 by Protein Kinase C Abolishes Its Inhibition by Calmodulin*

G-protein-coupled receptor kinases (GRKs) are impor-tant regulators of G-protein-coupled receptor function. Two members of this family L, GRK2 and GRK5 L, have been shown to be substrates for protein kinase C (PKC). Whereas PKC-mediated phosphorylation results in inhibition of GRK5, it increases the activity of GRK2 toward its substrates probably through increased affinity for receptor-containing membranes. We show here that this increase in activity may be caused by relieving a tonic inhibition of GRK2 by calmodulin. In vitro , GRK2 was preferentially phosphorylated by PKC isoforms a , g , and d . Two-dimensional peptide mapping of PKC a -phospho-rylated GRK2 showed a single site of phosphorylation, which was identified as serine 29 by HPLC-MS. A S29A mutant of GRK2 was not phosphorylated by PKC in vitro and showed no phorbol ester-stimulated phosphorylation and an API 100 quadrupole mass spectrom- eter as described previously (25). Two-dimensional Peptide Mapping— GRK2 was phosphorylated and digested as above. Labeled peptides were analyzed by two-dimensional phosphopeptide mapping as described (26). The experiment was also performed with a synthetic peptide comprising residues 7–34 of GRK2 with identical results.

Signaling via G-protein-coupled receptors (GPCRs) 1 is subject to a variety of regulatory processes. One of the key regulatory mechanisms is the control of receptor function by Gprotein-coupled receptor kinases (GRKs). These kinases phosphorylate the agonist-bound state of GPCRs, and this phosphorylation is often the initial step in homologous receptor desensitization, the loss of receptor responsiveness upon repeated or prolonged receptor stimulation (1,2). To date, six members of the GRK family are known (3,4). Of these, GRK2 (previously called ␤-adrenergic receptor kinase-1) as well as GRK3, GRK5, and GRK6 are widely expressed throughout the mammalian body whereas GRK1 (also known as rhodopsin kinase) and GRK4 have been found only in specific tissues.
All GRKs have a similar molecular architecture. A central catalytic domain is flanked by an N-and C-terminal domain. The function of the latter is to provide a membrane anchor, which seems to be essential for receptor phosphorylation (5). The C terminus of GRK1 is farnesylated, whereas GRK2 and GRK3 possess a pleckstrin homology domain implicated in phosphatidylinositol bisphosphate and G-protein ␤␥-subunit binding. GRK4 and GRK6 are palmitoylated, and GRK5 binds phospholipids through a poorly defined polybasic domain (3,4). The function of the N terminus is less clear. It has been shown for GRK1 that antibodies against an epitope in the N terminus prevent phosphorylation of light-activated rhodopsin but not of a peptide substrate (6). However, the epitope against which the antibodies were directed is poorly conserved in the GRK family. Recent results have indicated that the N terminus of GRK2 and GRK3 contains an RGS (regulator of G-protein signaling) domain that specifically binds to and activates G␣ q (7,8). In addition, the extreme N terminus of GRK2 and GRK5 contains a binding site for Ca 2ϩ /calmodulin. Binding of Ca 2ϩ /calmodulin to this site causes an inhibition of GRK activity, and this may represent a mechanism for the regulation of GRKs (9 -11).
In addition to GRKs, other kinases have been found to phosphorylate GPCRs. The most prominent kinases capable of phosphorylating receptors are the protein kinases A and C (PKC), which are activated by the second messengers generated after stimulation of these receptors. Recent studies have identified cross-talk between PKC and GRKs, which may provide another level of regulation of GRKs. PKC has been shown to phosphorylate GRK2 (12,13) and GRK5 (14). Whereas phosphorylation of GRK5 by PKC attenuated its activity (14,15), PKC-phosphorylated GRK2 caused significantly enhanced receptor phosphorylation and desensitization (12,13). This enhancement was at least partially caused by a PKC-mediated stimulation of GRK2 translocation from the cytosol to the plasma membrane where the substrate receptors are located (13). The molecular mechanisms of this activation have, however, not been resolved. Hence, we have investigated the exact nature of the phosphorylation and regulation of GRK2 by PKC. We report here that phosphorylation of GRK2 by PKC occurs in the N terminus and that this phosphorylation abolishes the inhibitory effect of Ca 2ϩ /calmodulin on GRK2 activity.

EXPERIMENTAL PROCEDURES
Materials-Crude phosphatidylcholine (type II-S from soybean) and calmodulin (from bovine testes, activity at least 40,000 units/mg protein) were from Sigma. The GRK2 antiserum has been described previously (13). * These studies were supported by grants from the Deutsche Forschungsgemeinschaft, the European Union, and the Fonds der chemischen Industrie. 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.
Cell Culture-Cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics in a humidified incubator with 7% CO 2 . HEK293 cells were transfected with a modified calcium phosphate method (16). COS cells were transfected using DEAE-dextran (17). Transiently transfected cells were analyzed ϳ48 h after transfection.
Mutagenesis-The construction of the cDNA coding for GST fused to human GRK2-(552-689) has already been described (13). cDNAs coding for the N terminus (amino acids 1-185) or the pleckstrin homology domain (amino acids 552-656) of human GRK2 were amplified from the full-length cDNA by PCR with Pfu polymerase (Stratagene) and cloned into the vector pGEX1T (Amersham Pharmacia Biotech) for expression as GST fusion proteins (18). Point mutations on human GRK2 were introduced by the method of Kunkel (19). All constructs were verified by automated sequencing.
Purification of Proteins-G-protein ␤␥-subunits were purified from bovine brain according to Sternweis and Robishaw (20). GST fusion proteins were expressed in Escherichia coli BL21 and purified by affinity chromatography as described previously (13). Human and bovine GRK2 were expressed in baculovirus-infected Sf9 cells and purified according to Kim et al. (21). PKC isozymes were purified from baculovirus-infected Sf9 cells as described previously (22). Protein concentrations were determined according to Bradford (23).
Determination of PKC and GRK2 Activity-PKC activity was determined using a soluble peptide substrate, [Ser 25 ]PKC␣-(19 -31) as described previously (22). GRK2 activity was usually measured by using urea-treated rod outer segments (50 pmol of rhodopsin) as the substrate in the presence of 50 nM purified G-protein ␤␥-subunits as described previously (24). Incubation was carried out at 30°C for 8 min. 32 P incorporation into rhodopsin was determined by SDS-polyacrylamide gel electrophoresis of the reaction mixture, excision of the rhodopsin band, and Cerenkov counting or quantification with a phosphorimager. To assess phosphorylation of soluble substrates by GRKs, 16 pmol of a GST fusion protein of the C terminus of the parathyroid hormone receptor were incubated with GRK2 for 30 min at 30°C. For investigation of the effects of calmodulin and/or PKC on GRK2 activity, GRK2 was preincubated with 70 units of calmodulin and/or 25 pmol of PKC for 30 min at 30°C in the dark before adding rod outer segments. PKC was activated with 10 M PMA. Mass Spectrometry-200 ng of purified GRK2 were phosphorylated with 10 ng of recombinant PKC␣ in the presence of 20 M [␥-32 P]ATP (1 Ci/mmol). The mix was separated by SDS-polyacrylamide gel electrophoresis, and the phosphorylated band was identified using autoradiography. It was then excised from the gel and digested with endopeptidase LysC, and the peptide mixture was separated by reverse-phase HPLC coupled to a radiodetector and an API 100 quadrupole mass spectrometer as described previously (25).
Two-dimensional Peptide Mapping-GRK2 was phosphorylated and digested as above. Labeled peptides were analyzed by two-dimensional phosphopeptide mapping as described (26). The experiment was also performed with a synthetic peptide comprising residues 7-34 of GRK2 with identical results.

RESULTS
PKC Isoform Specificity for GRK2 and GRK5-It has been previously shown that crude preparations of PKC purified from rat or bovine brain are capable of phosphorylating GRK2 and GRK5 (12)(13)(14). To examine which of the PKC isoforms were responsible for phosphorylation of GRKs, we analyzed PKC preparations from baculovirus-infected Sf9 cells for their ability to phosphorylate GRK2 and GRK5. The activity of these fractions was calibrated with histone H1 as a substrate. Equal amounts of the various PKC isoforms were then used to phosphorylate purified GRK2 and GRK5 (Fig. 1). Whereas GRK5 was quite a good substrate for all PKC isoforms tested except for PKC⑀, GRK2 was only significantly phosphorylated by PKCs ␣, ␥, and ␦. The radioactivity incorporated into GRKs in the absence of PKC (Fig. 1, A and B, right lanes) is caused by autophosphorylation. Because GRK2 is selectively phosphorylated by only certain PKC isoforms, whereas, with the exception of PKC⑀, GRK5 serves as substrate for all PKCs, these data suggest that the two types of phosphorylation might be different. Because of their higher activity toward GRK2, all subsequent experiments were done with recombinant PKC␣ or PKC␦.
Identification of the PKC Phosphorylation Site on GRK2-We had shown previously that GRK2 could be phosphorylated by rat brain PKC preparations to a stoichiometry of about one, suggesting a single phosphorylation site on GRK2. To verify the existence of a single phosphorylation site, we performed two-dimensional phosphopeptide mapping of GRK2. GRK2 was phosphorylated in the presence of [␥-32 P]ATP with PKC␣, digested with endopeptidase LysC, and the peptides resolved by two-dimensional electrophoresis. Only a single radiolabeled spot derived from GRK2 was identified, indicating the presence of a single phosphorylation site (Fig. 2).
To map this phosphorylation site, we performed HPLC-MS analysis of the LysC-digested peptides. The peptide containing the transferred phosphate was detected online with a radiodetector, and its mass was directly analyzed by electrospray-MS (Fig. 3). The major peak of radioactivity was identified as (AT)PAARASK based on the mass of the peptide and the fragmentation spectrum. This region corresponds to residues 22-30 of human GRK2 and identifies Ser 29 as the single phosphorylation site for PKC in GRK2.
To confirm this finding, a synthetic peptide comprising residues 7-34 of human GRK2 was synthesized, phosphorylated with PKC␣, digested with endopeptidase LysC, and subjected to two-dimensional peptide mapping as described above. The resulting phosphopeptide pattern was identical to the one obtained with full-length GRK2, as shown in separate or overlay analyses (Fig. 4). These data corroborate the correct identification of the phosphorylation site.
Ser 29 Is the PKC Phosphorylation Site in Vitro and in Vivo-To verify that Ser 29 is indeed the site phosphorylated by PKC, we constructed a mutant of human GRK2 in which this residue was replaced with alanine (S29A). This mutant was expressed in Sf9 cells using the baculovirus system and purified in a way identical to the wild-type protein. Purified S29A GRK2 was able to phosphorylate rhodopsin in a manner similar to the wild-type protein although its specific activity was lower (data not shown). As expected, PKC␦ was unable to phosphorylate the S29A GRK2, in contrast to wild-type GRK2 (Fig. 5). These data confirm that Ser 29 is indeed the only phosphorylation site for PKC in GRK2.
Phosphorylation of Ser 29 in GRK2 Abolishes the Inhibitory Effect of Calmodulin-We and others have shown previously that phosphorylation of GRK2 with PKC purified from bovine or rat brain leads to increased GRK2 activity on rhodopsin (12,13). Surprisingly, when we tried to reproduce these findings with PKC␣ or PKC␦ purified from Sf9 cells, we obtained no stimulation. Instead, we observed a slight nonsignificant reduction of GRK2 activity after PKC phosphorylation (Fig. 6,  left).
This discrepancy between our data suggests that the brain PKC preparations might have contained additional proteins that influence the activity of GRK2. One such protein might be calmodulin. Calmodulin has been shown to attenuate GRK2 activity, and the calmodulin binding site encompasses Ser 29 of GRK2 (11). To investigate whether calmodulin could in fact be the postulated "missing protein," we examined the influence of calmodulin on GRK2 activity.
Calmodulin inhibited the activity of bovine GRK2 by more than 75% as measured by the GRK2-mediated phosphorylation of light-activated rhodopsin (Fig. 6, right). This inhibition was completely reversed when active PKC was present in the assay (Fig. 6, right). These data show that PKC activates GRK2 by relieving the Ca 2ϩ /calmodulin-mediated inhibition.
Mutation of Ser 29 Abolishes PKC-mediated Activation of GRK2 in Intact Cells-Finally we wanted to verify that Ser 29 represents the functionally relevant PKC phosphorylation site in intact cells. To achieve this, HEK293 cells were transiently transfected with wild-type or S29A mutant GRK2. When the cells were stimulated with PMA, cells transfected with wildtype GRK2 showed a roughly 2-fold increase in cytoplasmic GRK activity (Fig. 7, left). In the case of the S29A mutant, this stimulation of GRK activity by PMA was almost completely abolished (Fig. 7, right).
The residual effects of PMA seen under these conditions are probably caused by endogenous wild-type GRK2. These results indicate that PKC-mediated phosphorylation of GRK2 at Ser 29 does indeed cause its activation in intact cells by relieving a constitutive inhibition through calmodulin or a related protein. DISCUSSION The function and activity of G-protein-coupled receptors is controlled by several classes of proteins via a multitude of mechanisms. A general feature of these mechanisms is that they are interrelated and that they appear to be controlled at several different levels. A particularly intriguing example of this complex control is the inhibition of receptor activity by GRKs, which is itself subject to control by PKC. Whereas GRK5 activity toward receptors was decreased following phosphorylation by PKC (14), GRK2 activity toward small soluble receptor substrates was also decreased by PKC, but its activity versus membrane-bound receptors was increased (12,13). These data suggested that PKC might activate GRK2 by increasing its association with membrane-bound receptors (13).
PKC phosphorylation studies utilizing a protein comprised of the C terminus of GRK2 fused to GST had initially suggested the presence of a C-terminal phosphorylation site in GRK2 (13). In the present study, using intact GRK2 and purified recombinant PKC isozymes, we have mapped the PKC phosphorylation site in GRK2 unequivocally to serine 29 in the N terminus, and we can exclude phosphorylation of the GRK2 C terminus by PKC. We have further shown that PKC phosphorylation of GRK2 does not directly stimulate the activity of the isolated GRK2 in vitro.
On the other hand, our previous experiments clearly indicated that phosphorylation of GRK2 with PKC preparations obtained from bovine brain caused an ϳ2.5-fold increase in the incorporation of phosphate into rhodopsin (13). Similarly, Chuang et al. (12) also observed activation of GRK2 after phosphorylation with a PKC preparation purified from bovine brain.
As shown here, we were unable to reproduce these results using PKC isozymes purified from baculovirus-infected Sf9 cells. These discrepancies raised the possibility that the GRK2 activation observed after phosphorylation with PKC might be the result of additional protein(s) present in the brain PKC preparations, but not in the PKCs purified from Sf9 cells. One candidate for such a protein was calmodulin, which has previously been shown to inhibit GRK2 activity (10,11). Indeed, Cells were then lysed, cytosol was prepared, and GRK2 activity was measured using rhodopsin as the substrate as described under "Experimental Procedures." Shown are the mean Ϯ S.E. from three experiments. The inset shows a representative experiment. The amounts of transfected GRK2 were roughly equivalent as determined by Western blotting (data not shown). purified recombinant GRK2 was inhibited by calmodulin, and this inhibition could be fully reversed when GRK2 was phosphorylated by PKC. We and others have previously reported that PKC activation of GRK2 also occurs in intact cells (12,13). Here, we show that removal of the PKC phosphorylation site on GRK2 also abolishes GRK2 activation through PKC. This indicates that a pool of GRK2 in the cell may be tonically inhibited by forming a complex with calmodulin or a related protein from which it is released through PKC phosphorylation.
Whereas our experiments offer a mechanistic explanation for the phenomenon of GRK2 activation by PKC, they also raise several questions. First, how does calmodulin inhibit GRK2, and how is this inhibition altered following phosphorylation of GRK2 by PKC? It is conceivable that calmodulin competes with the agonist-activated receptor for GRK2. This suggestion is supported by findings obtained with GRK1 (rhodopsin kinase) and rhodopsin. An antiserum directed against amino acids 17-34 of GRK1 was able to inhibit phosphorylation of lightactivated rhodopsin but not of a peptide substrate (6). Interestingly, this region of GRK1 is also able to bind rhodopsin, as is the corresponding region of GRK5 (11). Furthermore, Nterminal truncation of GRKs 2 and 5 impairs their ability to phosphorylate rhodopsin (27). These data indicate that the N terminus of GRKs might be involved in binding to the receptor substrates. The sequence homology of the calmodulin-binding sites in GRK2 and GRK5 is, however, quite low.
The proximal C terminus of GRK2 (residues 457-546 of 689) has been implicated in receptor binding as well (28). This suggests that both C-and N-terminal regions of GRK are involved in receptor association. The C terminus has been shown to interact with light-activated rhodopsin by itself, suggesting that it may be sensitive to the conformational changes occurring in G-protein-coupled receptors upon activation (28). In contrast, truncation of the N terminus does not abolish the ability of GRKs to be activated by agonist-occupied receptors (27). It therefore appears that the N terminus is not involved in sensing the activated state of the receptor but may contribute some weak interaction that is required for efficient substrate recognition. Evidently, additional experiments are necessary to prove this hypothesis.
Another question is how phosphorylation of GRK2 by PKC affects its regulation by Ca 2ϩ /calmodulin. Such inhibition might be because of a reduction in direct binding affinity (measured to be around 40 nM, see Ref. 11), but it might also be more complex and involve other regions of the GRKs. (A second calmodulin binding domain has been mapped to residues 466 -689 of GRK2, Ref. 11.) Such a more complex interaction is suggested by the fact that phosphorylation of GRK2 by PKC (as well as of GRK5) has a direct inhibitory effect on the activity of the catalytic core region of GRKs if, instead of a membranebound receptor, a soluble substrate was phosphorylated. This was either a peptide (13,14) or the soluble C terminus of the parathyroid hormone receptor (data not shown).
It is remarkable that GRK2 seems to possess a special versatility for regulating the activity of G q -coupled receptors. It is inhibited by Ca 2ϩ /calmodulin and thereby regulated by G q -mediated alterations in intracellular free Ca 2ϩ -levels. This inhibition is released through PKC activity, which can also be controlled via G q . In addition, GRK2 contains an RGS domain, which leads to fairly selective inhibition of G q -mediated signal transduction pathways (7,8). The reasons for this specialization of GRK2 are at the moment completely unclear, in particular because GRK2 was initially thought to be particularly active toward the G s -coupled ␤-adrenergic receptors.
Taken together, our data delineate a previously unknown mechanism of regulation of G-protein-coupled receptors, which is exerted by an interplay of PKC and Ca 2ϩ /calmodulin in regulating the activity of GRK2. These observations are a further step in the complex machinery regulating the function of these receptors.