G Protein-coupled Receptor Kinase GRK2 Is a Phospholipid-dependent Enzyme That Can Be Conditionally Activated by G Protein (cid:98)(cid:103) Subunits*

G protein-coupled receptor kinases (GRKs) mediate agonist-dependent phosphorylation of G protein-cou-pled receptors (GPRs) and initiate homologous receptor desensitization. Previously, we reported that charged phospholipids directly interacted with the two GRK iso-forms, GRK2 and GKR3, via a pleckstrin homology (PH) domain to regulate GRK activity (DebBurman, S. K., Ptasienski, J., Boetticher, E., Lomasney, J. W., Benovic, J. L., and Hosey, M. M. (1995) J. Biol. Chem. 270: 5742– 5747). Here, evidence is provided to support the hypothesis that charged phospholipids are required for agonist-dependent phosphorylation of receptors by GRK2. In the absence of charged phospholipids, the purified human m2 muscarinic acetylcholine receptor (hm2mAChR) reconstituted in pure phosphatidylcholine vesicles or in a noninhibitory detergent was not a substrate for GRK2. However, these receptor preparations were stoichiometrically phosphorylated in an ag-onist-dependent manner upon addition of charged phospholipids. The known ability of G protein (cid:98)(cid:103) subunits to and on a YM-100 membrane using a Centricon Phospholipid vesicles (50 (cid:109) g) were added to the hm2 mAChR in deter- gent micelles, gently vortexed, and preincubated for 5 min at room temperature prior to phosphorylation. Phosphorylation of receptors by GRK2 was performed in a manner similar to reconstituted hm2 mAChRs. Phosphorylation of Casein by GRK2— A stock solution of partially dephosphorylated casein (10 mg/ml) was prepared in 50 m M pH 7.4. Phosphorylation of casein (0.25–0.5 mg/ml) was carried out with GRK2 (100 n M ) in reactions that contained the same phosphorylation buffer previously described for phosphorylation of hm2 mAChRs, with 200 (cid:109) M ATP (500 cpm/pmol) for 10 min at 37 °C. In reactions that contained phospholipid, lipids were preincubated with GRK2 for 10 min at room temperature prior to the addition of casein. The reactions were stopped with 2 ml of 10% trichloroacetic acid containing 0.4 mg of bovine albumin serum. Phosphorylated casein was collected on G/FC Whatman filters, washed 4X with 4 ml of 10% trichloroacetic acid, and phosphate incorporation was directly measured using a scintillation counter. GRK2 activity was defined as the difference in phosphate incorporation in the presence and absence of GRK2. In reactions with phospholipids, a separate blank was used due to higher background obtained with phospholipid.

Regulatory mechanisms that control biological responses involve both the amplification and desensitization of cellular signaling (1). For G protein-coupled receptors (GPRs), 1 which represent a diverse superfamily of membrane receptors, agonist-induced phosphorylation of the receptors is thought to be a critical event in the initiation of receptor desensitization (2)(3)(4). G protein-coupled receptor kinases (GRKs) are extremely selective protein kinases that phosphorylate specifically only agonist-occupied active form of GPRs (5). A variety of approaches including reconstitution studies in vitro (6 -9), overexpression of protein kinases (10), use of protein kinase inhibitors (11,12), receptor mutagenesis (13)(14)(15), antisense GRK knockout (16), dominant-negative inhibition of GRK (15,17,18), and transgenic mice that overexpress a GRK-selective inhibitory peptide (19), strongly suggest that GRKs mediate agonist-induced phosphorylation of GPRs and trigger homologous desensitization of GPRs.
Subsequently, others reported on interactions between phospholipids and GRK2 in studies of the ␤ 2 AR (57,58). Our results were confirmed by Onorato et al., (57) who observed similar effects of charged phospholipids (including an inhibition by PIP 2 ) toward GRK2-mediated phosphorylation of purified ␤ 2 ARs in dodecyl maltoside mixed micelles. On the other hand, Pitcher et al. (58) reported that phospholipids neither bound nor regulated GRK2 activity significantly. In contrast, these authors observed that activation of GRK2 specifically required the binding of both G ␤␥ and PIP 2 and implicated the PH domain in GRK2 as essential for this effect. While each of these studies pointed to an important role of phospholipids and/or G ␤␥ in regulating GRKs, several important questions remain.
This study describes the results of investigations carried out to dissect the relative contributions of phospholipids and G ␤␥ in the activation of GRK2 and addresses the specific differences observed among authors regarding the role of PIP 2 in the regulation of GRK2. To achieve this goal, we reconstituted purified hm2 mAChRs into neutral phosphatidylcholine (PC) vesicles and asked the following questions. 1) Were receptors substrates for GRKs in the absence of charged phospholipid? 2) Did charged phospholipids activate GRKs in the absence of G ␤␥ and was this activation a direct effect? 3) Did G ␤␥ effects on GRK require prior activation of GRKs by charged phospholipid, and specifically PIP 2 , to stimulate GRK2? 4) What were the most effective phospholipids in regulating GRKs? 3 and partially dephosphorylated casein were purchased from Sigma. The sources of other reagents have been previously described (56,59).

Materials-IP
Cell Culture-Sf9 insect cells were cultured as described (60), in suspension culture using Sf-900 II serum-free medium supplemented with gentamicin (50 g/ml).
Purification and Reconstitution of hm2 mAChRs-The hm2 mAChRs were purified and reconstituted in either chick heart lipids or pure PC (660 M) obtained from Avanti Polar Lipids, as described previously (34).
Phosphorylation of Reconstituted hm2 mAChRs-Phosphorylation of hm2 mAChRs (20 -50 nM) by GRK2 (10 nM) was performed as described (34,56). GRK2 and the GST-GRK2(C-term) fusion protein were prepared as described previously (39,61). Reactions were performed for one hour in a volume of 50 l and in the presence of agonist (unless otherwise stated). In reactions that contained GST-GRK2(C-term), the fusion protein was used at 10 M. All phosphorylation reactions, including reactions with membrane-bound GPRs and detergent solubilized m2 mAChRs (described below) were stopped with SDS-sample buffer and phosphorylated receptors were resolved by SDS-polyacrylamide gel electrophoresis using 8% polyacrylamide (62). Phosphate incorporated into receptor was measured by PhosphorImager analysis. Phospholipids dissolved in choloroform were dried under nitrogen, resuspended in 10 mM Tris-HCl, pH 7.4, 1 mM EDTA to achieve a desired stock concentration (2-5 mg/ml), sonicated for five minutes at 4°C, and used at the concentrations indicated. Reactions with G ␤␥ contained either transducin ␤␥ (␤␥ t , 250 nM) or brain ␤␥ (␤␥ b , 100 nM), unless otherwise stated.
Phosphorylation of Membrane-bound GPRs-The hm2 mAChRs in Sf9 insect cell membranes were prepared and phosphorylated by GRK in reactions containing 0.5 pmol of receptor as described previously (59). Porcine m2 mAChRs in purified membranes prepared from stably transfected mammalian Chinese hamster ovary (CHO) cells (63) were a gift from Dr. Mike Schimerlik and were phosphorylated by GRK2 in a manner identical to hm2 mAChRs in Sf9 insect cell membranes using 1 pmol of mAChR per reaction. Rhodopsin in urea-treated rod outer segments from bovine retina was prepared and phosphorylated by GRK2 using 10 pmol of rhodopsin per reaction either in the presence of absence of light (64).
Preparation and Phosphorylation of Purified hm2 mAChRs in Dodecyl Maltoside Detergent Micelles-The hm2 mAChRs expressed in Sf9 insect cells were solubilized and purified in 0.8% digitonin, 0.16% cholate (34) and subjected to detergent exchange as described by Onorato et al. (56). Briefly, the digitonin-solubilized purified hm2 mAChRs were filtered through a Sephadex G-50 column preequilibrated with 0.5 mM dodecyl maltoside, 100 mM NaCl, 10 mM HCl, pH 7.4, 5 mM EDTA and concentrated on a YM-100 membrane using a Centricon device. Phospholipid vesicles (50 g) were added to the hm2 mAChR in detergent micelles, gently vortexed, and preincubated for 5 min at room temperature prior to phosphorylation. Phosphorylation of receptors by GRK2 was performed in a manner similar to reconstituted hm2 mAChRs.
Phosphorylation of Casein by GRK2-A stock solution of partially dephosphorylated casein (10 mg/ml) was prepared in 50 mM Tris-HCl, pH 7.4. Phosphorylation of casein (0.25-0.5 mg/ml) was carried out with GRK2 (100 nM) in reactions that contained the same phosphorylation buffer previously described for phosphorylation of hm2 mAChRs, with 200 M ATP (500 cpm/pmol) for 10 min at 37°C. In reactions that contained phospholipid, lipids were preincubated with GRK2 for 10 min at room temperature prior to the addition of casein. The reactions were stopped with 2 ml of 10% trichloroacetic acid containing 0.4 mg of bovine albumin serum. Phosphorylated casein was collected on G/FC Whatman filters, washed 4X with 4 ml of 10% trichloroacetic acid, and phosphate incorporation was directly measured using a scintillation counter. GRK2 activity was defined as the difference in phosphate incorporation in the presence and absence of GRK2. In reactions with phospholipids, a separate blank was used due to higher background obtained with phospholipid.

RESULTS
GRK-mediated Phosphorylation of m2 mAChRs Reconstituted in PC Vesicles-In previous studies, we found that purified hm2 mAChRs reconstituted in lipids purified from chick heart were effective substrates for GRKs (34,56), and that phosphorylation was strikingly promoted by charged phospholipids, while noncharged PC produced little effect (56). Because the chick heart lipids were providing a mixture of undefined lipids that may affect GRK activity, we sought to develop a "null" (or neutral) background for more detailed analysis of the lipid effects on GRKs. As shown in Fig. 1A, recombinant hm2 mAChRs purified from Sf9 insect cells and reconstituted into pure PC vesicles were poorly phosphorylated by GRK2; in the presence or absence of the agonist carbachol the maximal stoichiometry was 0.1-0.2 mol P/mol R, in contrast to the stoichiometry of ϳ4 mol P/mol R obtained under similar conditions but using chick heart lipids for the reconstitution (Fig. 1A). Furthermore, increasing the PC content up to 500 M only resulted in a modest increase in agonist-dependent activity toward hm2 mAChRs (0.5 mol P/mol R) (Fig. 1A). This result demonstrated that presentation of substrate (ligand activated hm2 mAChRs) in pure PC vesicles was insufficient to yield stoichiometric phosphorylation by GRK.
Effects of Charged Phospholipids on GRK-mediated Phosphorylation of m2 mAChRs in PC Vesicles-The neutral background provided by pure PC vesicles allowed us to investigate whether the addition of charged phospholipids could activate GRK2 and allow stoichiometric, agonist-dependent phosphorylation of hm2 mAChRs. The addition of 250 M PS, phosphatidylglycerol (PG), phosphatidylethanolamine (PE), cardiolipin, or phosphatidylinositol (PI) induced agonist-dependent phosphorylation of the PC-reconstituted hm2 mAChRs by GRK2 (up to 5 mol P/mol R); substoichiometric phosphorylation was observed in the absence of agonist (Fig. 1B). The results suggest that two factors were minimally required to detect GRK2mediated phosphorylation of the hm2 mAChR in vitro: the active GPR and activating phospholipid. Phospholipid effects were concentration-dependent and saturable; high micromolar concentrations were required to maximally activate GRK2 by PI (Fig. 1C) and the other phospholipids (data not shown). Previously, we had reported that PIP 2 caused inhibition of the phosphorylation of mAChR which were reconstituted in chick heart lipids (56). When we assessed the effects of PIP 2 on the phosphorylation of mAChR in PC vesicles we observed biphasic effects, PIP 2 stimulated phosphorylation at low concentrations (5-25 M) and inhibited at higher concentrations (Fig. 1A). A detailed analysis of the PIP 2 effects will be described later.
A relevant concern is whether in vitro stimulation of GRK2 by charged phospholipids correlates with in vivo regulation of GRK by lipids. The in vivo mol fraction percent is estimated at 1-3% for PIP 2 2 and 10% for PS (79), respectively, of the total cellular phospholipid content. The concentrations of phospholipids used throughout this work were in the physiological range; under the conditions we used, 25 M PIP 2 corresponded to 3% mol fraction, whereas the concentration of 250 M PS corresponded to 25% mol fraction. This suggests that the effects observed may well occur in vivo. Secondly, were the observed lipid effects the same if the activating lipids and receptor were both present in the same vesicle (i.e. reconstituted together)? To address this issue, the following experiments were carried out. Purified hm2 mAChRs were directly reconstituted into phospholipid vesicles that contained either 3% PIP 2 (25 M PIP 2 ) and 97% PC, or 25% PS (250 M PS) and 75% PC. Phosphorylation of receptor by GRK2 was directly compared using: (a) receptors in pure PC vesicles to which PIP 2 or PS were added after reconstitution, and (b) the receptors reconstituted in the PC/PIP 2 and PC/PS vesicles. Irrespective of whether the activating lipid was present in receptor containing vesicles or added separately, the results obtained clearly demonstrated that GRK2 was similarly activated and the results were as shown in Fig. 1 (data not shown). The possibility exists that charged phospholipids, when added, become incorporated into the existing receptor containing PC vesicles. Using the approach of directly reconstituting receptors with various combinations of lipids to extensively characterize the lipid effects on GRK2 would require multiple reconstitution schemes involving small amounts of receptor. Small scale receptor reconstitutions are impractical and result in poor yield. Therefore, given that there was little difference in activation properties of GRK, for the remainder of these studies we added phospholipids to receptors reconstituted in PC vesicles to study their effects on GRK2.
Are the effects of phospholipids due to direct effects on GRK or indirect effects on the receptor substrate? We had earlier shown that GRK2 bound to PS and PIP 2 but not to PC (56), suggesting that lipids might directly interact with GRK2 to regulate its activity. To further test this idea, we used two distinct approaches. First, we asked if phospholipids could modulate GRK2-mediated phosphorylation of casein, a soluble non-receptor substrate of GRKs (7). PS stimulated the phosphorylation of casein by GRK2 over 2-fold, while PIP 2 inhibited phosphorylation of casein by more than 60% of control ( Fig.  2A). That GRK2 exhibited activity toward casein in the absence of lipid argues for a low intrinsic activity of GRK2. Similar evidence for direct activation and inhibition of GRK2 by phospholipids also was obtained by Onorato et al. (57) who used a synthetic peptide as the nonreceptor substrate for GRKs.
Next, we used a detergent micelle assay with dodecyl maltoside (57) to test whether GRK2 required phospholipid to phosphorylate purified soluble mAChRs. Unlike most other detergents (7), dodecyl maltoside was noninhibitory toward GRK2 (57). In this method the receptors are soluble and not reconstituted into phospholipid vesicles, thus regulation of GRK2 by added phospholipids would argue against the possibility that phospholipids indirectly regulated GRK activity by modulating the conformation of hm2 mAChRs. In the absence of lipid, the hm2 mAChR in dodecyl maltoside micelles did not serve as a substrate for GRK2 either in the presence or absence of agonist 2 R. Anderson, personal communication. Receptor was phosphorylated by GRK2, as described under "Experimental Procedures," either in the presence or absence of agonist carbachol. For PC-reconstituted hm2 mAChRs, reactions were also performed in the presence of increasing concentrations of added PC vesicles (10 -500 M). Stoichiometry of phosphorylation is shown as mol P/mol R. The data represent the average of two independent experiments. B, the PC-reconstituted hm2 mAChRs (660 M PC) were phosphorylated by GRK2 in the presence of several added charged phospholipid vesicles (250 M) including PE, PG, cardiolipin (CL), PS, PI, and PIP 2 as indicated, either in the presence (solid bar) or absence (open bar) of agonist carbachol. With PIP 2 , stimulation and inhibition of GRK2 activity was observed with 25 M and 250 M PIP 2 , respectively. The data represent two to five independent experiments. C, a representative phosphorimage depicting the concentration-dependent effects of PI on GRK2-mediated phosphorylation is shown. The concentration dependence of other activating phospholipids (except PIP 2 ) on GRK2 activity was tested (1-500 M), and their potencies were found to be similar. (Fig. 2B). When PC (50 g) was added, a low basal, agonistindependent, phosphorylation of hm2 mAChRs by GRK2 was observed (Fig. 2B). In contrast, PS induced a dramatic agonistdependent phosphorylation of hm2 mAChRs which achieved a stoichiometry of 3.5 mol P/mol R (Fig. 2B). On the other hand, receptor phosphorylation in the presence of high concentrations of PIP 2 was even less than that obtained with PC (0.1 mol P/mol R). Similar effects of charged phospholipids were previously observed with the dodecyl maltoside-solubilized ␤ 2 AR and GRK2 (57). Taken together, the results with casein and dodecyl maltoside-solubilized hm2 mAChR argue that charged phospholipids directly regulate GRK2.
Effects of Phospholipids on GRK-mediated Phosphorylation of Membrane-bound GPRs-To test the effectiveness of native membranes to provide sufficient phospholipid to support GRK activity, we analyzed GRK2 activity toward GPR in native membranes in the presence and absence of added phospholipids. Three preparations were tested: m2 mAChRs in Sf9 insect cell membranes and in mammalian CHO cell membranes, and rhodopsin in rod outer segments. Agonist-dependent phosphorylation of membrane-bound m2 mAChR was achieved in the absence of added lipid (Fig. 3). In the presence of agonist and no added lipid, the stoichiometries of phosphorylation were 1.9 and 2.2 mol P/mol R in the Sf9 cell and CHO cell membranes, respectively (averages from 2 experiments). These stoichiometries of phosphorylation of the membrane-bound mAChR were similar to that observed with reconstituted receptors in the presence of activating phospholipids. Addition of PS (250 M) did not increase phosphorylation of carbachol-activated hm2 mAChR by GRK2 (Fig. 3); the stoichiometries were 1.8 and 2.3 mol P/mol R in the Sf9 cell and CHO cell membranes, respec-tively. Similarly, light-dependent phosphorylation of rhodopsin by GRK2 was observed in membrane preparations of rod outer segments and the extent of phosphorylation was unaltered by the addition of exogenous PS (Fig. 3). The results suggest that the phospholipid content of native membranes was sufficient to support GRK2 activity toward receptor substrates. On the other hand, PIP 2 (250 M) inhibited phosphorylation of both rhodopsin and m2 mAChR by GRK2 by 50 -60% in each of these assays (Fig. 3). For the m2 mAChR, the stoichiometries of phosphorylation were reduced to 0.85 and 0.9 in the Sf9 and CHO cell membranes, respectively. That PIP 2 was able to inhibit GRK2 under these conditions suggested that PIP 2 interacted with GRK2 via a mechanism that either prevented or overcame cumulative activation of GRK by endogenous charged membrane phospholipids.
Effects of G ␤␥ on GRK2-mediated Phosphorylation of hm2 mAChRs in PC Vesicles-Previous studies demonstrated stimulation of GRK-mediated phosphorylation of GPRs by the ␤␥ subunits of G proteins (G ␤␥ ) (26 -29, 34, 56, 59). We next asked whether G ␤␥ activated GRK2 activity toward hm2 mAChRs in the absence of charged phospholipid. Two different preparations of G ␤␥ were tested: ␤␥ t purified from retinal transducin and ␤␥ b purified from bovine brain. First, we compared the effects of ␤␥ t (250 nM) on GRK2-mediated phosphorylation of hm2 mAChR in chick heart lipids and PC vesicles. While ␤␥ t markedly stimulated phosphorylation of receptors in chick heart lipids, we observed that ␤␥ t had little effect on GRK2 activity toward agonist-occupied hm2 mAChRs in PC vesicles (Fig. 4A). If we increased the PC content by further addition of PC vesicles, ␤␥ t still did not activate GRK2, and receptor phosphorylation remained substoichiometric (compare results of Fig. 1A with no G ␤␥ with results of Fig. 4A). An identical set of experiments was performed with ␤␥ b (100 nM) and similar results were obtained (data not shown). Second, we determined concentration-dependent effects of ␤␥ t and ␤␥ b on mAChR phosphorylation in PC vesicles and observed only very modest effects of either G protein, even at very high concentrations (500 nM) (Fig. 4B). These small effects of G ␤␥ are especially striking when compared to the marked stimulation of phosphorylation by charged phospholipids (Fig. 1B). These results demonstrate that GRK did not efficiently phosphorylate its sub- strate in the absence of charged phospholipid, despite the presence of both ligand-occupied GPR and G ␤␥ .
Effects of PIP 2 and G ␤␥ on GRK2-mediated Phosphorylation of hm2 mAChRs in PC Vesicles-A previous study suggested that GRK2 was activated in a synergistic manner by G ␤␥ and PIP 2 , while neither modulator alone was reported to have effects (58). In contrast, we and others have observed inhibition of GRK2 by PIP 2 (56,57) and in this study we observed stimulation by PIP 2 at low concentrations. Therefore, in order to better understand the system, we investigated in detail the ability of PIP 2 to modulate GRK2 activity. A detailed dose-dependent study of PIP 2 effects on mAChR phosphorylation in PC vesicles revealed the complex biphasic effect of PIP 2 on GRK2. At concentrations between 1 and 75 M, PIP 2 activated GRK2 in a concentration-dependent manner, and up to 1.5 mol P/mol R were incorporated into the receptor at 25 M PIP 2 (Fig. 5A). At higher concentrations, GRK2 activity was inhibited in a dose-dependent fashion until receptor phosphorylation was reduced below basal (0.1 mol P/mol R) (Fig. 5A). This is the first demonstration that PIP 2 effects on GRK2 are biphasic, being stimulatory at low micromolar concentrations and inhibitory at higher concentrations.
In order to probe interactions of G ␤␥ and PIP 2 in regulating GRK, we determined the effects of G ␤␥ at concentrations of PIP 2 from 1 to 500 M. At concentrations where PIP 2 activated GRK2 (up to 50 M), ␤␥ t (250 nM) was able to further stimulate receptor phosphorylation by 2-fold. In addition, in the presence of ␤␥ t , the peak for stimulation occurred at slightly lower concentrations of PIP 2 than in the absence of G ␤␥ (10 versus 25 M, respectively). Synergistic activation of GRK2 by PIP 2 and G ␤␥ was maximal with 10 M PIP 2 and 3 mol P/mol R was incorporated (Fig. 5A). However, ␤␥ t (tested up to 500 nM) was unable to overcome effects of inhibitory concentrations of PIP 2 (Fig.  5A). When effects of ␤␥ b were evaluated, similar results were obtained as with ␤␥ t with two notable differences. One, 10-fold less ␤␥ b (25 nM) was required with PIP 2 to synergistically activate GRK2 with similar efficacy (data not shown). Second, high concentrations of ␤␥ b (500 nM) were able to partially prevent the complete inhibition observed at high PIP 2 (the stoichiometries of phosphorylation were 0.1 mol P/mol R in the absence of G ␤␥ and 0.6 mol P/mol R in the presence of G ␤␥b ) (data not shown).
Effects of Phosphatidylinositol 4-monophosphate (PIP), PI, and IP 3 in GRK-mediated Phosphorylation of m2 mAChRs in PC Vesicles-We asked whether the biphasic regulation of GRK2 by PIP 2 was a property shared by other inositol phosphate containing compounds. PIP also produced a biphasic regulation of GRK2 (Fig. 5B). However, when compared with PIP 2 , PIP was slightly more potent and more efficacious in its activation of GRK2-mediated phosphorylation of mAChR (maximal stoichiometry: 2.3 mol P/mol R) and inhibited GRK2 activity to a lesser extent when used at higher concentrations (stoichiometry of receptor phosphorylation was 0.8 mol P/mol R) (Fig. 5B). PI also activated GRK2 in a dose-dependent manner, however, it was less potent than PIP or PIP 2 and it did not produce inhibitory effects in the concentration range studied (Fig. 5C). IP 3 had no effect on GRK2 (Fig. 5D). The ability of PIP, PI, and IP 3 to modulate GRK2 activity in the presence of G ␤␥ was also evaluated. In the presence of G ␤␥ the inositol phospholipids increased GRK-mediated receptor phosphorylation by ϳ2-fold (Fig. 5, B and C), while IP 3 was ineffective (Fig.  5D). In addition the PIP-mediated inhibition of GRK2 was modestly relieved by G ␤␥ . That G ␤␥ and the three stimulatory inositol phospholipids synergistically activated GRK2 is significant, since others have suggested that synergism with lipid and G ␤␥ to activate GRK2 occurs significantly only with PIP 2 (58). These results have important bearing with regard to the nature of ligands of the PH domain of GRK2 (see below). The second new finding from these results was that the presence of additional phosphates on the inositol phospholipids markedly affected their ability to regulate GRK2. In terms of activation, a plot of the dose-dependent effects of PI, PIP, and PIP 2 at activating concentrations clearly demonstrates that PIP 2 and PIP activate GRKs at 7-10-fold lower concentrations than PI (Fig. 5E, Table I), but the extent of stimulation for both PIP 2 and PIP was less for PI. In addition, a second more complex effect seen only with PIP and PIP 2 was their ability to inhibit GRK2 at high concentrations. It is likely that the lower efficacy of PIP and PIP 2 to stimulate GRK activity is related to their inhibitory properties (analyzed further below).
Effect of G ␤␥ and Non-inositol-charged Phospholipids in GRK-mediated Phosphorylation of m2 mAChRs in PC Vesicles-The above results led us to assess functional interactions of G ␤␥ with other non-inositol charged phospholipids that regulated GRK2. Agonist-induced phosphorylation of hm2 mAChRs by GRK2 was evaluated in the presence of G␤␥ t and either PS, PI, phosphatidic acid, PE, PG, PIP 2 , or PC. Each charged phospholipid, but not PC, synergistically activated GRK2 in the presence of ␤␥ t (Fig. 6A). In the absence of G ␤␥ , 1-3 sites were phosphorylated on the hm2 mAChRs as a result of phospholipid activation of GRK; the addition of G ␤␥ resulted in a total of 3-6 sites becoming phosphorylated on the receptor (Fig. 6A). An analysis of the time course of receptor phosphorylation by GRK2 in the presence of either PS (250 M) or PIP 2 (10 M) revealed that the receptor became stoichiometrically phosphorylated within five minutes (Fig. 6B). The addition of ␤␥ t (250 nM) served to enhance both the rate and extent of receptor phosphorylation by GRK2 (Fig. 6B). Interestingly, a lower concentration of ␤␥ t (25 nM) only increased the rate, but not the extent of phosphorylation (data not shown).
The agonist dependence of receptor phosphorylation was compared in the presence of PI Ϯ G ␤␥ (Fig. 6C). The results demonstrated that the extent of receptor phosphorylation is increased in an agonist-dependent manner in the presence of PI alone and further with PI plus G ␤␥ , while the concentration of carbachol required for half-maximal phosphorylation of the receptors remained unchanged (Fig. 6C). This result is expected if the lipids and G proteins modulate receptor phosphorylation via effects on GRK2 rather than directly via effects on the receptors. Taken together, these results demonstrate that agonist-dependent phosphorylation of the hm2 mAChR by GRK2 requires charged phospholipids, and that G ␤␥ plays a critical role in augmenting GRK activity, however the effects of  G ␤␥ are highly dependent on the activation of the GRK by charged phospholipids.
Localization of Phospholipid Effects on GRK2 to the C-terminal, PH Domain-containing Region of GRK2-Previously we had demonstrated that charged phospholipids interacted with the PH domain containing C terminus of GRK2 to regulate enzyme function (56). These results were obtained using a GST-fusion protein of the C terminus of GRK to block the effects of phosphatidylserine to stimulate the ability of GRK2 mediated phosphorylation of the hm2 mAChR in chick heart lipids. However the complex nature of the chick heart lipids obscured the stimulatory effects of the phosphoinositides that we report here. Because of the interest in these latter compounds as potential ligands of the PH domain of GRK2, it was important to ask whether the stimulatory effects of the phosphoinositides were mediated by the C-terminal domain of GRK2. The GST-GRK2(C-term) fusion protein effectively abolished the ability of PIP 2 and PI to activate GRK (Fig. 7A). The GST fusion protein also blocked the synergistic effects of PIP 2 and G ␤␥ (Fig. 7A). In control reactions, GST alone had no effect on GRK2 activation by these ligands (data not shown). These results support the hypothesis that domains in the C terminus of GRK, perhaps including the PH domain, are responsible for the lipid mediated activation of the kinase.
The identification of ligands for PH domains has been the objective of many recent studies (65). Studies with pleckstrin and other PH domain proteins suggested that these proteins could bind PIP 2 (45). Several recent studies have demonstrated IP 3 bound to the crystal structures of the PH domains of ␤-spectrin (49) and phospholipase C-␦1 (48). IP 3 binding to the PH domain of PLC-␦1 has been shown to compete with PIP 2 binding (50). We asked whether IP 3 could antagonize activation of GRK2 by the phosphoinositides (Fig. 7B). IP 3 (600 M) reduced by 50 -65% the stimulation of receptor phosphorylation by PIP 2 and PI, as well as the combination of PIP 2 and G ␤␥ (Fig. 7B). In control reactions, IP 3 alone had little effect on GRK2 in the presence or absence of G ␤␥ (Fig. 7B). That none of the phosphoinositide effects were completely prevented by IP 3 suggests that IP 3 either has a lower affinity for the lipid interaction domain(s) in GRK2 and/or, as a soluble ligand, is unable to effectively compete for binding to GRK2 with PIP 2 (10 M) or PI (250 M). Taken together, the results with the GST-GRK(Cterm) fusion protein and IP 3 suggest that the stimulatory effects of the phosphoinositides may arise from interactions with the PH domain in the C terminus of GRK2.
Inhibitory Effects of PIP 2 and PIP Are Mediated at a Site Distinct from the Stimulatory Site(s)-Neither the PH domaincontaining GST-GRK2(C-term) fusion protein nor IP 3 were able to prevent PIP 2 from inhibiting GRK2 at high PIP 2 concentrations (data not shown), suggesting that inhibition mediated by PIP 2 was mediated by another mechanism. In view of the inhibitory effects we observed with the multiply phosphorylated inositol lipids, we tested the possibility that the inhibitory effects on GRK2 might be mediated via competition with ATP at its binding site. If this is the case, elevating the concentration of ATP should alleviate the PIP-or PIP 2 -mediated inhibition of GRK activity. First, as a control, we tested the effects of varying ATP from 1-200 M on the extent of receptor phosphorylation by GRK using PI, the phosphoinositide that exhibited "pure" stimulatory properties (Fig. 5C). In the presence of PI Ϯ G ␤␥ , receptor phosphorylation by GRK increased as a function of ATP in the range of 1-50 M, after which a plateau was reached which was maintained up to concentrations of 400 M (Fig. 8 and data not shown). In marked contrast, changes in the concentration of ATP from 50 M to 400 M had marked effects on the results obtained with PIP 2 (Fig.  9A). In reactions containing 10 M PIP 2 , the extent of receptor phosphorylation increased from 1 mol P/mol R at 50 M ATP up FIG. 7. The effects of GST-GRK2(C-term) fusion protein and IP 3 on activation of GRK2 by phospholipid and G ␤␥ . A, effects of GST-GRK2(C-term) fusion protein on phosphorylation of hm2 mAChRs by GRK2 in reactions that were carried out for 10 min with either PIP 2 (10 M) in the presence or absence of ␤␥ t (250 nM), or with PI alone (250 M). In reactions that contained the fusion protein, phospholipid and G ␤␥ were preincubated with the fusion protein (10 M) for 10 min at room temperature before adding GRK. B, effects of IP 3 were assessed as in A, except that reactions contained IP 3 . GRK2 was preincubated with IP 3 for 10 min at room temperature before adding phospholipid and G ␤␥ . All reactions are means (ϮS.E.) of three independent experiments. to 3 mol P/mol R at 400 M ATP (Fig. 9A). Similarly, at 100 and 500 M PIP 2 , GRK activity was significantly increased as ATP was increased, however the inhibitory phase of PIP 2 was still apparent. Inhibition caused by PIP 2 was also not overcome in experiments performed with 1 mM ATP (data not shown). The results suggest that even low concentrations of PIP 2 (such as 10 M) exerted mixed stimulatory and inhibitory effects, and that the inhibition can be relieved, at least partially, by increasing the concentration of ATP. The complete concentration dependence curve for PIP 2 was examined at a high and low concentration of ATP (Fig. 9B). Increasing ATP increased both the extent of stimulation by PIP 2 at low concentrations and significantly reduced the extent of inhibition at high PIP 2 (Fig. 9B). However, inhibition at high concentrations of PIP 2 was not completely removed. Elevations in ATP affected both the stimulatory and inhibitory phases of the PIP 2 curve, and the EC 50 and IC50 appeared to be unchanged.
Similar results were obtained in experiments comparing PIP at high and low ATP. At low concentrations of PIP (10 M) receptor phosphorylation was increased ϳ2-fold in the presence of 400 M ATP compared to 50 M ATP (stoichiometries were 3.5 mol P/mol R and 1.5 mol P/mol R, respectively). Similarly at high (inhibitory) concentrations of PIP, receptor phosphorylation was 0.6 mol P/mol R at 50 M ATPand 1.3 mol P/mol R at 400 M ATP (data not shown). These results with PIP and PIP 2 were in marked contrast to those seen when the "pure" stimulatory lipid PI was used (Fig. 8) where the extent of receptor phosphorylation was constant from 50 -400 M ATP (Fig. 8, and data not shown). Since neither the GST-C(term) fusion protein nor IP 3 could prevent the PIP 2 -mediated inhibition, and since elevations in ATP partly overcame the inhibition, taken together, these results suggest that the inhibitory effects observed with PIP and PIP 2 may be due to actions at a site distinct from that mediating stimulation of GRK activity by these compounds. The preliminary assessment provided here suggests that the 4Ј-and 5Ј-phosphate groups on PIP and PIP 2 participate in this inhibition. Further experiments will be necessary to more completely assess the mechanism(s) involved. The additional finding from this series of experiments is that because higher ATP partly relieved the PIP/PIP 2 mediated inhibition of receptor phosphorylation by GRK2, the efficacies of these ligands as stimulators of GRK2 were increased at high ATP. Thus, it is likely that at low ATP the true activation properties of PIP 2 and PIP were masked by expression of inhibitory properties (compare extent of stimulation at 50 M ATP in Fig. 5E to that achieved at higher ATP in Fig. 9). Under conditions of high ATP, the efficacies of PIP 2 and PIP as stim-ulators of GRK activity were as favorable as PI and the other charged phospholipids. DISCUSSION Several previous reports have concentrated on the regulation of GRK2 and/or GRK3 by charged phospholipids (47, 56 -58) and the combinations of lipids and ␤␥ subunits of G-proteins (56,58). However, significant differences in these reports prompted us to further analyze the mechanisms underlying the complex regulation of GRK2 exerted by these compounds. In our initial report, we demonstrated that several charged phospholipids bound to GRK2 and that certain phospholipids activated the phosphorylation of hm2 mAChR by GRK2, while PIP 2 produced inhibition of activity (56). The effects of stimulatory phospholipids and G ␤␥ were found to be additive and the site(s) responsible for stimulation were shown to be within the PH domain containing C terminus (56). Subsequently, studies with the ␤ 2 AR confirmed the stimulatory properties of the charged phospholipids and the inhibitory properties of PIP 2 , and the use of a mixed micelle system allowed the investigators to conclude that the activity of GRK2 required activation by charged phospholipids (57). In contrast, Pitcher et al. (58) found no evidence for activating effects of phospholipids alone but demonstrated synergistic activation of GRK2 with combinations of PIP 2 and G ␤␥ . Here we report on detailed studies of the interactions of charged phospholipids and G ␤␥ using a method that afforded a "neutral" background which allowed us to clearly assess these complex interactions. The results obtained strongly support the hypothesis that GRK2 is a lipid-dependent kinase (57). In the absence of a charged phospholipid GRK2 exhibited only a very low activity toward receptor substrates (substoichiometric phosphorylation). In addition, GRK2 alone was relatively poor in phosphorylating non-receptor substrates such as casein ( Fig. 2A) and peptides (7,57). However, activity was dramatically increased upon the addition of a charged phospholipid, particularly toward the GPRs. Although there are likely to be other factors that contribute to GRK2 activation, particularly the receptors themselves (31,35), the data reported here, and by Onorato et al. (57), establish GRK2 as a phospholipid-dependent kinase. While not explicitly tested in the present studies, our previous study suggested that GRK3 is also a phospholipid-dependent kinase, as both GRK2 and GRK3 (previously referred to as ␤ARK1 and ␤ARK2) were regulated by lipids in an identical manner (57). Notably, our present finding that charged phospholipids specifically activated GRK2 correlated well with GRK2: phospholipid binding studies in our previous report (56). In both of these studies, GRK2 neither bound to nor was activated by pure PC vesicles (Fig. 1, C and D, in DebBurman et al. (56), and Fig. 1A, this study). Furthermore, these effects of PS and PIP 2 to bind (Fig.  1, C and D, in DebBurman et al. (56)) and activate GRK2 (Fig.  1B, this study) were dose-dependent. Thus, the combination of results from the previous and present studies provide a strong correlation between GRK:lipid interaction and enzyme activation. The major findings reported here are summarized and discussed below.
First, the phosphoinositides PI, PIP, and PIP 2 markedly stimulated GRK2 activity at low concentrations. Both PIP and PIP 2 stimulated activity with EC 50 values in the 10 -20 M range, and were 10-fold more potent than PI and other charged phospholipids that stimulated GRK2 activity. Thus of the various lipids that activated GRK2, PIP and PIP 2 appear the likely regulators of GRK2 activation in vivo. However, this cannot be ascertained with certainty from the present results. Furthermore, the concentrations of several charged phospholipids may be transiently changed during signal transduction, especially near the plasma membrane (67, 68). The stimulatory effects of the phosphoinositides appeared to be mediated through the C terminus of GRK2, and may involve the participation of the PH domain in this region. This is suggested by: 1) the GST-GRK2(C-term) fusion protein containing the PH domain prevented the stimulation by PIP 2 and PI; 2) IP 3 , which has been postulated to be the critical component of PIP 2 binding to several other PH domains (48 -50), partially blocked the stimulatory effects of PIP 2 ; and 3) the EC 50 for the functional effects of PIP 2 on GRK2 activity was similar to the K d value for PIP 2 binding to the isolated PH domain of GRK2 (45). Nevertheless, this assignment of functional effects of the phosphoinositides to the PH domain of GRK2 awaits extensive structural data to support or refute this contention. Furthermore, several observations might suggest that some of the lipid effects reported here might involve interactions at additional domains and/or that the PH domain of GRK is quite different than other PH domains. For example, how would PI and the non-inositol phospholipids interact with a PH domain if the GRK2 PH domain is similar to those whose structures have been recently solved? PIP 2 has been reported to bind to the PH domains of several different proteins (45,46), and several crystal structures have been described with IP 3 bound in a pocket of PH domains, where the 4Ј and 5Ј phosphates of the inositol ring are critical determinants of the binding (48 -50). Since PI and the other non-inositol-charged phospholipids lack these determinants, other factors would be necessary to accommodate their binding.
A second finding from the present results that has not been previously appreciated is that the ␤␥ subunits of G-proteins are unable to stimulate GRK activity in the absence of a charged phospholipid that is capable of stimulating GRK activity. Thus, the G ␤␥ subunits are conditional activators of GRK2. In a previous report, Pitcher and colleagues (58) proposed that neither the G ␤␥ subunits nor PIP 2 alone could support GRK2 activity toward ␤ 2 AR, but that a synergistic regulation of activity was observed in the presence of both compounds. These results are different from those of Onorato et al. (57) who observed stimulation of phosphorylation of the ␤ 2 AR by GRK2 by several phospholipids, and those reported here with the hm2 mAChR where clear effects of phospholipids on GRK activity were observed and were required for G ␤␥ to modulate activity. Although we are unable to reveal the reason for the differences in these reports, a possibility is that the requirements for phosphorylation of the ␤ 2 AR in reconstituted phospholipid vesicles is different from those required for phosphorylation in the mixed detergent micelles used by Onorato et al. (57) and for the mAChR used here. Nevertheless, it is clear that G ␤␥ effects cannot be observed in the absence of a charged activating phospholipid, which itself is capable of activating GRK2.
GRK2 and GRK3 are thought to translocate from the cytosol to the plasma membrane upon agonist stimulation of cells (5). However, recent studies have determined that these GRKs are also significantly associated with microsomal membranes and the plasma membrane (69,70). While the functional significance of distinct GRK pools and GRK translocation is not clear and will require further investigation, the important question is how these enzymes become associated with the membrane and precisely targeted to the active GPR. Since G ␤␥ is known to bind GRK2 and GRK3 via the C terminus, a role of G ␤␥ in targeting GRK2 and GRK3 to the plasma membrane has become a popular and appealing hypothesis (5). However, despite considerable efforts, the involvement of G ␤␥ in membrane targeting of GRK has not been unequivocally proven. In fact, most studies implicating G ␤␥ in membrane targeting of GRK have been performed using the GST-GRK2(C-term) fusion protein as a G ␤␥ scavenger and inhibitor (19,47,71). How specific was this fusion protein as a G ␤␥ -binding protein? We and others have subsequently determined that this same fusion protein is capable of interacting with charged phospholipids in vitro (45,47,56). Furthermore, we previously demonstrated that GRK2 specifically associated with charged phospholipids, suggesting that lipids may also serve as membrane anchors for GRK2 (56). Interestingly, a dual role for phospholipids to aid in membrane localization and activation of proteins has been observed for several other signaling and adaptor proteins, including grb2, SOS, Ras, Ras-GAP, and Shc (72,73), and protein kinases, such as protein kinase C (74), and even another GRK, GRK5 (75). Thus an alternate hypothesis is that phospholipids and G ␤␥ subunits act in concert to both anchor GPRs to the plasma membrane and optimally regulate enzyme function. An additional factor that must be considered is the active GPR. Since active GPRs, in the presence of charged phospholipids, can directly associate with and activate GRKs (31,36,37), it is possible that the active GPR directly recruits GRKs and, in lieu of G proteins, also participates as an agonist-dependent membrane anchor.
The third new observation reported here is that GRK2 activity is inhibited by multiply phosphorylated inositol phospholipids through a mechanism that appears to be distinct from that mediating the stimulatory effects of these compounds. The 4Ј and 5Ј phosphate groups of the phosphatidylinositols are likely to contribute to the inhibitory effects of these compounds as the inhibitory effects were not observed with PI. While increases in ATP were able to increase the relative stimulation and decrease the inhibition seen with compounds that produce both stimulatory and inhibitory effects, further studies will be needed to reveal the site(s) mediating inhibition by PIP and PIP 2 . PIP 2 has been reported to inhibit casein kinase I (76) and may interfere with protein substrate binding to this kinase (66).
The biphasic (stimulatory and inhibitory) effects of PIP 2 observed here contrast markedly to the purely inhibitory effects of PIP 2 that we observed earlier (56). This is explained by the use of a mixture of undefined lipids from chick heart for reconstitution of the hm2 mAChR in the previous study, compared to the use of pure PC which resulted in a "neutral" background for reconstitution of the receptors in this study. It is important to note that the mix of chick heart lipids contained sufficient amounts of as yet undefined lipids which resulted in a "basal" level of activity that was similar to that observed in this study in the presence of a stimulatory lipid. Thus when chick heart lipids were used for reconstitution, the extent of phosphorylation observed in the presence of agonist was 4 -5 mol P/mol R, compared toϳ 0.2-0.5 mol P/mol R when pure PC was used herein for reconstitution. Only when a stimulatory lipid was added to the receptors in pure PC did the stoichiometries of phosphorylation approach those seen using the mix of chick heart lipids. It seems likely that the chick heart lipids contained phosphoinositides or other stimulatory phospholipids which could support receptor stimulation, and that when PIP 2 was added to this mixture, it increased the concentration into the range that produced inhibition. Since a relative degree of inhibition can be observed even with low concentrations of PIP 2 at low ATP (Fig. 9), the inhibitory effects of PIP 2 observed earlier are readily explained. Interestingly, the heart lipid mix must also contain other compounds that are favorable for receptor phosphorylation, as the maximum stoichiometries of phosphorylation that are achieved using the heart lipid mix plus added lipids and G ␤␥ (56) were more than double the highest stoichiometries reported in this study.
Regulation of GRKs by lipids has emerged as a conserved theme within the GRK family. Like GRK2 and GRK3, GRK1 also translocates to the membrane in an agonist-dependent manner via a mechanism that appears to involve isoprenylation of the C terminus (5). GRK4 and GRK6 are found associated to the plasma membrane via a mechanism that could involve palmitoylation (77,78). GRK5 interacts with membrane phospholipids via hydrophobic interactions, and as a result, becomes activated via an increase in autophosphorylation of the kinase (75). It is apparent that all GRKs employ diverse mechanisms to interact with lipids to achieve the two goals central for proper function of protein kinases: localization to substrate and activation of enzyme.
In summary, we support the hypothesis that GRK2 and GRK3 are lipid-dependent protein kinases that require charged phospholipids for kinase activation. Since GRK2 exhibits micromolar affinity for phospholipids, and since these compounds are likely to exist in the membrane in this concentration range, the available evidence would suggest that at least some fraction of the cellular GRK2 may be present in a phospholipid bound form. The activation of GPRs by agonist results in activation of G proteins which results in dissociation of the heterotrimeric G proteins in G ␣ and G ␤␥ subunits. Dissociated G ␤␥ may serve two functions: one, to help localize the membranebound or cytosolic GRKs near their target GPRs, and, second, to contribute to a synergistic regulation of GRK activity by phospholipids to facilitate regulation of the GPRs. The GPRs may also contribute to GRK activation, however, how this occurs in concert with the phospholipid and G-protein effects remains to be elucidated.