Characterization of the GRK2 Binding Site of Gαq*

Heterotrimeric guanine nucleotide-binding proteins (G proteins) transmit signals from membrane bound G protein-coupled receptors (GPCRs) to intracellular effector proteins. The Gq subfamily of Gα subunits couples GPCR activation to the enzymatic activity of phospholipase C-β (PLC-β). Regulators of G protein signaling (RGS) proteins bind to activated Gα subunits, including Gαq, and regulate Gα signaling by acting as GTPase activating proteins (GAPs), increasing the rate of the intrinsic GTPase activity, or by acting as effector antagonists for Gα subunits. GPCR kinases (GRKs) phosphorylate agonist-bound receptors in the first step of receptor desensitization. The amino termini of all GRKs contain an RGS homology (RH) domain, and binding of the GRK2 RH domain to Gαq attenuates PLC-β activity. The RH domain of GRK2 interacts with Gαq/11 through a novel Gα binding surface termed the “C” site. Here, molecular modeling of the Gαq·GRK2 complex and site-directed mutagenesis of Gαq were used to identify residues in Gαq that interact with GRK2. The model identifies Pro185 in Switch I of Gαq as being at the crux of the interface, and mutation of this residue to lysine disrupts Gαq binding to the GRK2-RH domain. Switch III also appears to play a role in GRK2 binding because the mutations Gαq-V240A, Gαq-D243A, both residues within Switch III, and Gαq-Q152A, a residue that structurally supports Switch III, are defective in binding GRK2. Furthermore, GRK2-mediated inhibition of Gαq-Q152A-R183C-stimulated inositol phosphate release is reduced in comparison to Gαq-R183C. Interestingly, the model also predicts that residues in the helical domain of Gαq interact with GRK2. In fact, the mutants Gαq-K77A, Gαq-L78D, Gαq-Q81A, and Gαq-R92A have reduced binding to the GRK2-RH domain. Finally, although the mutant Gαq-T187K has greatly reduced binding to RGS2 and RGS4, it has little to no effect on binding to GRK2. Thus the RH domain A and C sites for Gαq interaction rely on contacts with distinct regions and different Switch I residues in Gαq.

of the solvent-exposed surface of the ␣5 helix. This site is now referred to as the C site (22).
The residues on G␣ q required for association with the C site of the GRK2 RH domain are not known. Interestingly, a G188S mutation in G␣ q has no effect on the GRK2-G␣ q interaction (22), even though this mutation prevents the interaction of G␣ q with other RGS proteins (23,24), suggesting that the G␣ q residues critical for interaction with GRK2 are different from those used to bind RGS proteins. However, the activation-dependent association of G␣ q with GRK2 requires that at least part of the interface involves the switch regions of G␣ q (18). In support of this, GRK2 binds chimeric proteins that contain the GTPase domain of G␣ q and the helical domain of G␣ 16 , the only G␣ q family member that does not interact with the RH domain of GRK2, but not reciprocal chimeras, in an activation-dependent manner (19).
In this study we used a molecular modeling approach to identify residues on G␣ q that may interact with GRK2. Sitedirected mutagenesis followed by GST-pull-down and cellular inositol phosphate assays indicated that contact sites for GRK2 on G␣ q include the Switch I and III regions as well as residues in the helical domain of G␣ q . Some of these residues are distinct from those that are important for interactions with RGS2 and RGS4. In addition, our previous mutational studies of the GRK2 C site were based upon a homology model of the GRK2 RH domain (22). Because the structure of full-length GRK2 indicates that the ␣5 helix, the major point of G␣ q contact, is significantly longer than modeled (14), we have further refined the C site based on the G␣ q :RH model and the G␣ q mutagenesis studies described herein.

EXPERIMENTAL PROCEDURES
Materials-HEK-293 cells were from American Type Culture Collection (CRL-1573). FuGENE 6 transfection reagent was from Roche Molecular Biochemicals. Super Signal West Pico ECL reagents were from Pierce. myo-[ 3 H]Inositol was obtained from PerkinElmer Life Sciences. Cell culture media were from Mediatech Cellgro. GRK2 mouse monoclonal antibody was from Upstate Biotechnology. Horseradish peroxidase-conjugated anti-mouse secondary antibody was from Promega. Ultima Flo AF and Ultima Gold scintillation cocktails were from Packard Chemical. All other chemicals and reagents were from Sigma and Fisher Scientific.
Expression Plasmids and Mutagenesis-pcDNA3-G␣ q -R183C (G␣ q -RC) with an internal EE epitope tag was provided by Dr. C. Berlot. EE-tagged G␣ q and G␣ q -Q209L (G␣ q -QL) have been described previously (19). All G␣ q , G␣ q QL, and G␣ q RC point mutants were created in the background of EE-tagged protein using the sequential PCR method (25). The GST-RGS4 expression plasmid was provided by Dr. R. Neubig, and pcDNA3-RGS2 was provided by Dr. D. Siderovski. GST-RGS2 was created by using PCR to engineer a 5Ј BamH1 site and a 3Ј XhoI site onto RGS2 and then subcloning the RGS2 fragment into the BamH1 and XhoI sites of pGEX-5X1. GRK2 constructs have been described previously (18,22). GRK2 mutants were prepared by sequential PCR or by QuikChange Mutagenesis (Stratagene).
Cell Culture and Transfection-HEK-293 cells were maintained in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum and 100 units/ml penicillin-streptomycin at 37°C in 5% CO 2 . Cells in 6-well plates were transfected with 1 g of DNA and 3 l of FuGENE, whereas 3 g of DNA and 9 l of FuGENE were used for the transfection of cells in 6-cm dishes.
Molecular Modeling-G␣ q was homology-modeled with the SWISS-MODEL server (26) using as a template the structure of G␣ i (49% sequence identity) in complex with RGS4 (11), which represents the most complete atomic model of an activated G␣ subunit. The model was subsequently verified by ERRAT (27). As expected, the regions of the model evaluated as most unreliable (over the 95% confidence level) are the effector binding loops of G␣ q , which vary greatly among the four G␣ subfamilies. However, the three switch regions of the G protein represent the most likely binding site for GRK2 given its requirement for activated G␣ q (18). The three switch regions of the model are distant from the effector loops, include some of the most highly conserved residues among G␣ subunits, and are found in essentially the same conformation in all crystal structures that involve an activated G␣ subunit. Evaluation by ERRAT also suggested that these regions were modeled reliably.
To model the GRK2-G␣ q complex, automated docking programs were tested, but not used because they generally fail to accurately model changes, such as those of side chains, upon complex formation (28,29). We instead imposed several strict constraints based on experimental data to manually dock G␣ q with GRK2. The resulting model of their complex at the very least should predict which regions of G␣ q could be responsible for both complex formation and specificity. The first constraint was to limit the GRK2-interaction surface of G␣ q to its three switch regions and to the ␣A helix and the ␣B-␣C loop in the helical domain. This constraint derives from the facts that the formation of the GRK2-G␣ q complex is dependent on the active conformation of the G protein, and therefore presumably the conformation of its switch regions (18), and that the adjacent ␣A helix and ␣B-␣C loop have also been shown to contribute to the binding of RH domains in other G␣ subunits (12,30). The second constraint was to limit the G␣ q -interaction surface of GRK2 to solvent exposed residues on the ␣5 and ␣6 helices of its RH domain, which were previously identified to be important for complex formation with G␣ q (22). The third constraint was to fix the relative orientation of G␣ q and GRK2 to the plane of a common cell membrane, as each of their orientations with respect to a cellular membrane is relatively well known from prior crystal structures and electrostatic calculations. The orientation of G␣ q with respect to the plane of the plasma membrane, defined by the GRK2-G␤␥ structure (14), was fixed by docking it against the G␤␥ subunits present in the GRK2-G␤␥ complex in a manner similar to G␣ i in the G␣ i1 G␤ 1 ␥ 2 structure (31). G␣ q was then translated and rotated along the plane of the membrane until its GRK2-interaction surface was adjacent to the G␣ qinteraction surface of GRK2. The model was manually adjusted to optimize the packing of residues at the protein interface and then minimized using simulated annealing in CNS (32) to relieve any bad contacts between side chains. Harmonic restraints were imposed on the C␣ positions during refinement to keep the backbone relatively fixed (0.28-Å root mean squared deviation between initial and final coordinates of G␣ q ). The modeled interface buries 2100 Å 2 of surface area, which is on par with or larger than those observed in crystal structures of other RH domain-G␣ complexes (e.g. 1800 Å 2 in the RGS9-G␣ t/i1 complex (12)). The final model of the complex was verified by the program PROCHECK (33), ERRAT, and VERIFY3D, which indicated that the residues involved in the interface are consistent with a reasonably packed and complementary structure.
Inositol Phosphate Production Assay-HEK-293 cells were transfected with 0.1 g of EE G␣ q RC or EE G␣ q RC mutant constructs, 0.2 g of myc-His-tagged G␤ 1 , 0.1 g of G␥ 2 , the indicated amounts of GRK2-K220R or RGS2, and pcDNA3 up to a total of 1 g of DNA. Twenty-four hours after transfection, cells were replated on 4 wells of a 24-well plate, and 3 wells were labeled for 16 h with 2 Ci/ml [ 3 H]inositol. Inositol phosphate production was determined as previously described (22). Results are the average of at least three experiments done in triplicate and represented as Percent Control. Control is the level of inositol phosphate produced in the absence of co-transfected GRK2-K220R or RGS2. Graphing and statistical analysis, as described in the figure legends, was performed using GraphPad Prism.
The fourth well of replated transfected cells was used to monitor whether the co-expression of GRK2 or RGS2 had any effect on the expression of G␣ subunits. Cells were lysed with 50 l of SDS sample buffer, vigorously homogenized, and boiled for 5 min. 20 l of the sample was then subjected to 12% SDS-PAGE and transferred to polyvinylidene difluoride. The polyvinylidene difluoride was then probed with 2 g/ml of EE monoclonal antibody followed by horseradish peroxidase-conjugated secondary antibody (1:10,000 dilution). Pierce Super-Signal West Pico reagents were used to visualize immunoblots. Blots from assays with GRK2-K220R were stripped with a 50 mM glycine buffer, pH 2.0, and reprobed with a GRK2-specific monoclonal antibody followed by horseradish peroxidase-conjugated secondary antibody (1: 10,000 dilution) to determine the effect of transfecting increasing amounts of GRK2-K220R cDNA on GRK2 expression.
GST-GRK2-(45-178), GST-RGS2, and GST-RGS4 Interaction Assays-HEK-293 cells were transfected in 6-cm dishes with 2.0 g of G␣ q or mutant G␣ q cDNA, 0.2 g of myc-His-tagged G␤ 1 , 0.1 g of G␥ 2 , and pcDNA3 up to a total of 3.0 g of DNA. 24 h after transfection cells were washed with cold phosphate-buffered saline and lysed with 0.3 ml of lysis buffer (20 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM dithiothreitol, 100 mM NaCl, 5 mM MgCl 2 , 0.7% Triton X-100, 1 mM phenylmethylsulfonyl fluoride, and 5 g/ml leupeptin and aprotinin). After 1 h of lysis at 4°C, cells were centrifuged for 3 min at full speed in a microcentrifuge. For G␣ q -RC and G␣ q -QL assays, 200 l of the supernatant was removed to a new tube and incubated with 8 g of GST-GRK2-(45-178), GST-RGS2, or GST-RGS4, all pre-bound to glutathione-Sepharose beads, for 1-2 h at 4°C. The remaining supernatant, denoted "L" for lysates, was saved for subsequent immunoblot analysis alongside pull-down samples. After incubation of the lysates with the GST-GRK2-(45-178)-, GST-RGS2-, or GST-RGS4-bound beads, the samples are pelleted at low speed in a microcentrifuge for 3 min, and the beads are washed three times with lysis buffer. Proteins were then eluted from beads in 50 l of SDS sample buffer and boiled for 5 min.
For GST pull-down assays carried out in the absence or presence of AlF 4 Ϫ , 250 l of the G␣ q -containing supernatant (described above) was removed and split equally into two tubes. To one of the tubes AlCl 3 (25 M), NaF (5 mM), and MgCl 2 (1 mM) were added. GST-GRK2-(45-178)-, GST-RGS2-, or GST-RGS4 (8 g)-bound beads were then added to each tube, and the mixtures were incubated for 4 -5 h at 4°C. The remaining supernatant, denoted "L" for lysate, was saved for subsequent immunoblot analysis alongside samples. After incubation of the lysates with the GST-GRK2-(45-178)-, GST-RGS2-, or GST-RGS4-bound beads, the samples are pelleted at low speed in a microcentrifuge for 3 min, and the beads are washed three times with lysis buffer. Proteins were then eluted from beads in 50 l of SDS sample buffer and boiled for 5 min.
In all cases, 20 l of each pull-down sample was subjected to 12% SDS-PAGE and transferred to polyvinylidene difluoride, which was probed with 2 g/ml EE monoclonal antibody followed by horseradish peroxidase-conjugated secondary antibody (1:10,000 dilution). A portion of the initial lysate that represents 4% of the protein present in the lysates was also analyzed on immunoblots alongside GST pull-down samples and indicated by "L" in the figures. Pierce SuperSignal West Pico reagents were used to visualize immunoblots. For graphical representation of pull-down assays, images of Western blots were acquired using a Kodak DC-40 digital camera, and the net intensity of each band was calculated using Kodak Digital Science 1D Image Analysis Software. The percentage of the mutant G␣ q that was pulled-down by the GST fusion protein was calculated and compared with the control, which is the percentage of G␣ q that interacted with the GST fusion protein and displayed as percent control ϮS.D. Graphing and statistical analysis, as described in the figure legends, was performed using GraphPad Prism.
Pull-down Assays with GRK2 RH Domain Mutants-Mutants of the GRK2 RH domain were assayed as GST-GRK2-(45-178) fusions using bovine brain extract as a source of WT G␣ q as described previously (22). To allow comparison of severely defective G␣ q -binding mutants, a more sensitive assay was developed by using 20 g/ml fusion protein and 500 g/ml bovine brain extract protein in the pull-down assays.

Molecular
Modeling of the G␣ q :GRK2 Interface-The binding surface for G␣ q has been localized primarily to the ␣5 helix of the GRK2 RH domain and is distinct from the protein-binding surfaces used by other RH domains (11,13,22). Accordingly, this interaction surface has been termed the C site, following the nomenclature proposed by Zhong and Neubig (17). GRK2 only interacts with activated G␣ q (18) suggesting the involvement of at least one G␣ q switch region in the interaction. However, the C site of the RH domain may bind residues on G␣ q that are distinct from those that interact with RGS proteins such as RGS2 and RGS4. We have already shown that the RGS-resistant Switch I mutant, G␣ q -G188S, retains association with GRK2 (22).
To predict which G␣ q residues could interact with the RH domain of GRK2, a homology model of G␣ q was manually docked with the RH domain from the GRK2-G␤␥ crystal structure (14) by imposing several specific constraints required by prior biochemical and structural analyses (see "Experimental Procedures"). The docking model predicts that the long ␣5 helix of the GRK2 RH domain docks into the cleft formed between the helical domain and the Ras-like domain of G␣ q and engages primarily Switches I and III of the Ras-like domain (Fig. 1A). This binding mode would be substantially different from those observed for the complexes of G␣ i and G␣ t with RGS4 (11) and RGS9 (12), respectively (Fig. 1B), although in each case the switch regions of G␣ provide the primary interaction site. In particular, the model predicts that Pro 185 of G␣ q is at the crux of the interface with its side chain packing against Asp 110 , Met 114 , and Leu 118 of GRK2 (Fig. 1, C and D). Indeed, both Asp 110 and Met 114 have previously been shown to be important for the interaction of GRK2 with G␣ q (22). Residues within the ␣A helix of G␣ q were also predicted to be in close proximity to the GRK2 RH domain and could explain why residues such as Val 137 of GRK2, which is quite distant from Asp 110 , Met 114 , and Leu 118 of GRK2, has an effect on G␣ q binding when mutated to alanine. Therefore, residues in both the Switch I and III regions and adjacent regions of the helical domain of G␣ q were targeted for mutagenesis.
Identification of G␣ q Switch Residues That Interact with the RH Domain of GRK2-To test the effects of point mutations in the Switch I and III regions of G␣ q , they were transiently overexpressed in HEK-293 cells, and GST-GRK2-(45-178) was then used to pull-down mutant G␣ q in lysates from the transfected cells. Each point mutant was made in the background of an otherwise wild-type (non-activated) G␣ q and a constitutively active G␣ q -Q209L (G␣ q -QL). Selected mutants were also generated in the constitutively active G␣ q -R183C (G␣ q -RC). Point mutants in the wild-type background were activated by the addition of AlF 4 Ϫ , which binds G␣-GDP and occupies the space normally filled by the ␥-phosphate of GTP. This causes the G␣ subunit to assume a conformation that is thought to mimic the transition state of GTP hydrolysis (34). In contrast, G␣ q -QL and G␣ q -RC are constitutively active because Glu 209 and Arg 183 are involved in the hydrolysis of the ␥-phosphate of GTP, and their mutation to leucine and cysteine, respectively, greatly decreases the rate of this reaction (34). Initially, binding of the point mutants to GST-GRK2-(45-178) was examined in the AlF 4 Ϫ and G␣ q -QL backgrounds. However, we also wanted to examine the effects of the mutations on the ability of GRK2 to inhibit the G␣ q -mediated formation of inositol phosphate in cells. Unfortunately, the co-transfection of GRK2-K220R, a kinase-deficient mutant of GRK2 (35), caused a marked decrease in the expression of several mutants, particularly K77P, Q152A, P185K, and T187K (data not shown). Previously, we had observed that relatively small amounts of co-transfected GRK2 were able to inhibit inositol phosphate production stimulated by constitutively active G␣ q -RC (data not shown). We therefore generated several of the mutants, K77A, Q81A, R92A, Q152A, P185K, and T187K, in the G␣ q -RC background and assessed their ability to bind GRK2. The effect of each mutation on binding to the RH domain of GRK2 was tested, and the results are summarized in Table I and described below.
Our model of the GRK2-G␣ q interaction predicts that Pro 185 is buried within the interface, and therefore will represent a critical specificity determinant. As expected, mutation of Pro 185 to lysine, the corresponding residue in G␣ i , has a profound negative effect on GRK2 binding (Fig. 2, A-C). Binding of P185K-QL and P185K-RC (Fig. 2, B and C) to the GRK2 RH domain is completely attenuated, whereas binding of P185K in the presence of AlF 4 Ϫ is less then 20% of wild type ( Fig. 2A). The mutation of Pro 185 , located in Switch I between the helical domain and the GTPase domain of G␣ q , does appear to decrease the expression of G␣ q by ϳ40% (data not shown). However, as will be discussed later, this mutant retains its ability to bind to RGS proteins when activated by AlF 4 Ϫ . Therefore, by mutating G␣ q Pro 185 we disrupted GRK2 binding, as predicted by the model of their complex. Two additional residues in the Switch I region of G␣ q , Val 184 and Thr 187 , were also targeted by site-directed mutagenesis. Mutation of Val 184 to aspartic acid is predicted to lessen favorable contacts with Leu 118 of GRK2. The V184D mutant has a modest effect (Ͻ80% of control) on AlF 4 Ϫ -activated G␣ q binding to GRK2, but this effect was lost in the G␣ q -QL background. Similarly, mutation of Thr 187 to lysine, the corresponding residue in G␣ 12/13 , was also predicted to destabilize the GRK2-G␣ q interface, perhaps by creating unfavorable contacts with the side chains of Lys 115 and Thr 111 of GRK2. However, GRK2 binding to G␣ q was unaffected by the T187K mutation, regardless of whether it was in the context of G␣ q ⅐GDP⅐AlF 4 Ϫ , G␣ q -RC  (11). The RH domains of GRK2 and RGS4 both interact with the switch regions of the G protein, but the surface of the RH domain used in the contact is unique. In the RGS4 complex, the ␣5 helix faces outward from the plane of the page, whereas in the G␣ q ⅐GRK2 complex it forms the principal contact surface. C and D represent views of G␣ q and GRK2, respectively, as if the complex shown in A were opened like a book. C, the GRK2-interacting surface of G␣ q . The residues shown as ball-and-stick models with green carbons are those mutated and analyzed in this study. Thick circles indicate residues that had a dramatic effect upon mutation (as per Table I), thin circles indicate an "intermediate" effect, and the absence of circles indicates no effect, at least upon GRK2 binding and inhibition of inositol 1,4,5-trisphosphate release. The residues listed in orange are those that each G␣ q residue is predicted to contact. The black sphere represents Mg 2ϩ . D, the G␣ q -interacting surface of the GRK2 RH domain. The residues shown as ball-and-stick models with green carbons are those mutated and analyzed in this (Leu 118 and Glu 130 ) and our previous study (22). Thick circles indicate residues that had a dramatic effect upon mutation, thin circles indicate an "intermediate" effect, and the absence of circles indicates no effect, at least upon G␣ q binding. The residues listed in purple are those that each GRK2 residue is predicted to contact. All panels were created using PyMOL (W. L. DeLano or G␣ q -QL ( Fig. 2 A, B and C). Substitution of residues at these positions may be permitted because they exist at the periphery of the interface and thus are partially solvent-exposed in the model. They could thereby accommodate longer side chains.
The model also predicts that the GRK2 RH domain interacts with the backbone of the Switch III residue Val 240 and that there is a potential salt-bridge between Asp 243 of G␣ q and Lys 139 of GRK2. The G␣ q -V240A and D243A mutants reduce or eliminate, respectively, AlF 4 Ϫ -dependent binding to GRK2 (Table I). However, both V240A-QL and D243A-QL interact with GRK2 to the same extent as G␣ q -QL. This suggests that Switch III is more critical for binding in AlF 4 Ϫ -activated G␣ q than the QL and RC conformations of the enzyme. The structural basis for these differences is not clear, but may be due to subtle conformational changes in the three switches when bound to either GTP or a transition state complex.
Gln 152 is a highly conserved alpha-helical domain residue whose side-chain makes specific hydrogen bonds within the Ras-like domain of G␣ subunits, principally with the backbone of Switch III and with a conserved arginine residue that likewise supports Switch III. Because Gln 152 is changed to histidine in G␣ 16 , which does not bind GRK2, and because of its proximity to the modeled RH domain, it was also targeted by site-directed mutagenesis. Q152A⅐GDP⅐AlF 4 Ϫ , G␣ q -Q152A-QL, and the Q152A-RC have reduced binding to the GRK2 RH domain (Table I and Fig. 2, A and C). These results with the Q152A, V240A, and D243A mutants confirm a role for Switch III in binding GRK2.
Identification of G␣ q Helical Domain Residues That Interact with the RH Domain of GRK2-Several residues within the helical domain of G␣ q are likewise predicted to be involved in the interaction with GRK2 (Fig. 1). Leu 78 is predicted to interact with GRK2 Leu 118 , and the L78D mutant had a slight effect (Ͻ80% of control) on the AlF 4 Ϫ -dependent binding to the GRK2 RH domain (Table I). Two residues in the ␣A helix of G␣ q , Lys 77 and Gln 81 , are expected to interact with the carboxyl terminus of the GRK2 ␣5 helix. The Q81A mutant of G␣ q has reduced binding to the GRK2 RH in the presence of AlF 4 Ϫ (Table I and Fig. 2A). However, Q81A-QL and Q81A-RC (Table I and Fig. 2, B and C) retain the ability to bind to the GRK2 RH domain. Replacement of Lys 77 with a proline, the analogous G␣ 16 residue, disrupts AlF 4 Ϫ -dependent binding to GRK2 and binding to GRK2 in the G␣ q -QL background (Table I). In contrast, the K77A mutant only has an effect on binding to GRK2 in the G␣ q -QL background (Table I and Fig. 2, A-C). These data support a role for the ␣-helical domain of G␣ q in dictating the specificity and the affinity of the GRK2-G␣ q interaction.
There are additional residues in the helical domain that are in close proximity to the GRK2 RH domain in the model but that are not conserved in G␣ 16 . The V118A mutation has no effect on GRK2 binding (Table I). Mutation of Arg 92 , whose aliphatic side chain is predicted to interact with Val 137 of GRK2, to alanine does not have an effect on AlF 4 Ϫ -dependent binding or binding to G␣ q -RC (Table I and Fig. 2, A and C); however, reduced binding to GRK2 is seen with R92A-QL (Table I and Fig. 2B). In addition to the interactions between the aliphatic portion of the G␣ q Arg 92 side chain and GRK2 Val 137 , our model predicts that its guanidino group forms a salt bridge with GRK2 Glu 130 . To test this idea, E130A was introduced into GST-GRK2-(45-178). Similar to another GRK2-␣6 mutant, V137A, E130A shows a modest deficiency in its ability to bind G␣ q/11 in a GST-pull-down assay (data not shown).
G␣ q -Q152A-RC Is Less Sensitive Than G␣ q -RC to GRK2mediated Inhibition of Inositol Phosphate Production-We then tested the interaction of GRK2 with G␣ q -RC mutants in intact cells. We have previously used co-transfection of G␤␥ to stabilize the expression of G␣ subunits in the presence of RGS proteins (19). In addition, very low amounts of GRK2-K220R, 5 ng of cDNA, are able to inhibit signaling from G␣ q RC (Fig. 3,  A and B). The ability of GRK2-K220R to inhibit G␣ q -RC signaling is not affected by the co-expression of G␤␥ (data not shown). These conditions allowed us to detect differences in the sensitivity of the point mutants to GRK2 inhibition. Unfortunately, even under these conditions, the expression of the P185K-RC mutant was inversely proportional to the amount of GRK2-K220R transfected (data not shown). Therefore it was  The table summarizes the effects of the G␣ q point mutants on binding to GRK2, RGS4, and RGS2. Pull-down assays were performed with GST-GRK2-(45-178), GST-RGS4, or GST-RGS2 on cell lysates that had been transfected with G␣ q containing the different point mutations, as described under "Experimental Procedures." There is no detectable binding of G␣ q or G␣ q mutants to GST alone. Binding to AlF 4 Ϫ -activated forms of the mutants and to mutants in the G␣ q -RC background was assessed for GRK2, RGS4, and RGS2; however, RGS4 does not bind to G␣ q -QL so the effects of the point mutants in the QL form on binding could only be tested with GRK2 and RGS2.
c The statistical significance of the difference between the indicated mutants and control is indicated in Figs. 2 and 4 for GRK2 and RGS2, respectively. d For mutants indicated by a Ϫ, statistical analysis could not be performed because there generally was no observable pull-down. e ND, not determined. f Binding of G␣ q -L78D to GST-GRK2-RH is significantly different than binding of G␣ q to GST-GRK2-RH (p Ͻ 0.001) and binding of G␣ q -Q81A-RC and G␣ q -T187K-RC to GST-RGS4 is statistically different then binding of G␣ q -RC to GST-RGS4 (p Ͻ 0.001). not included in these experiments. Even so, the expressed P185K-RC is functional, because G␣ q -P185K-RC still activates phospholipase C␤ and binds G␤␥ (data not shown).
The G␣ q mutant, other than G␣ q -P185K, that consistently inhibited the interaction with the RH domain of GRK2 in pull-down experiments was G␣ q -Q152A (Table I and Fig. 2). Q152A-RC also showed resistance to GRK2-K220R inhibition of inositol phosphate production (Fig. 3A). Although the differences in the inhibition of G␣ q -RC versus Q152A-RC signaling by GRK2-K220R are small, the effects are reproducible. For example, the transfection of 10 ng of GRK2-K220R led to 25% inhibition of the inositol phosphate stimulated by G␣ q -RC, whereas the same amount of GRK2-K220R did not inhibit Q152A-RC (Fig. 3A). The difference between G␣ q -RC and Q152A-RC is less pronounced at higher levels of GRK2-K220R expression, suggesting that this mutant lowers the affinity of but does not totally disrupt the interaction between GRK2 and G␣ q (Fig. 3A). This would agree with GST-GRK2 pull-down data that show a marked decrease in binding to Q152A in the presence of AlF 4 Ϫ and the Q209L mutation and a smaller decrease in binding to Q152A-RC (Table I and Fig. 2, A-C).
Two of the remaining G␣ q mutants tested for inhibition of inositol phosphate production by GRK2 showed only minor defects in GRK2 interaction. Although the Q81A mutant binds to the RH domain of GRK2 in both the QL and RC form (Fig. 2, B and C), in the presence of AlF 4 Ϫ the binding of Q81A is reduced (Table I and Fig. 2A). In the inositol phosphate assays with G␣ q -Q81A-RC and G␣ q -T187K-RC there are small differences in comparison to R183C in the ability of low levels of GRK2 to inhibit signaling (data not shown). However, at higher concentrations of GRK2-K220R, G␣ q -Q81A-RC and G␣ q -T187K-RC are inhibited to a level that is similar to G␣ q -RC (data not shown). The R92A-RC mutant is inhibited in a manner that is essentially identical to R183C (data not shown). Importantly, Fig. 3B shows that expression of GRK2-K220R, even at a high level, does not decrease the expression of G␣ q -RC or the G␣ q -Q152A-RC mutant, indicating that the observed differences in inositol phosphate production are due to differences in binding of GRK2 to G␣ q -RC relative to G␣ q -Q152A-RC, and not to differences in expression levels. In general, the data from the inositol phosphate assays are consistent with the pull-down assays with the RH domain of GRK2.
The A Sites of RGS2 and RGS4 Bind Different Surfaces of G␣ q Than the C Site of GRK2-We have previously shown that the binding surface for G␣ q on the GRK2 RH domain is distinct from that of RGS4, and in this set of experiments we wanted to determine the effect of the mutations made in G␣ q on binding to RGS2 and RGS4 (22). There was little or no difference in the ability of RGS2 versus RGS4 to bind each of the G␣ q mutants represented graphically as the percentage of control Ϯ S.D. B, HEK-293 cells were transfected with EE-tagged versions of G␣ q -Q209L point mutants and G␤ and G␥ constructs. Cells were lysed, and binding of the QL mutants to GST-GRK2-(45-178) was determined as described under "Experimental Procedures." Results are plotted as described in A. C, HEK-293 cells were transfected with EE-tagged versions of G␣ q -R183C point mutants and G␤ and G␥ constructs. Cells were lysed, and binding of the RC mutants to GST-GRK2-(45-178) was determined as described under "Experimental Procedures." The lanes labeled "P" represent 40% of the G␣ q or G␣ q mutant that was present in the pull-down from 200 l of lysate. The lanes in A-C labeled "L" represent 4% of total G␣ q or G␣ q mutant available in the lysate for pull-down. Results are plotted as described in A. ૺ, the amount of the marked G␣ q mutant pulled-down is significantly different (p Ͻ 0.05) by one-way ANOVA followed by a Dunnett post-test, than the amount of G␣ q pulled-down by GST-GRK2. ૺૺ, the statistical analysis could not be performed on the binding of GST-GRK2-RH to the K77A-QL, P185K-QL, or P185K-RC mutants because there was no detectable pull-down. The data are averages from three to six independent experiments.

FIG. 2. Interaction of GST-GRK2-(45-178) with G␣ q point mutants activated by AlF 4
؊ , the Q209L, or the R183C mutation. A, HEK-293 cells were transfected with EE-tagged versions of G␣ q point mutants and G␤ and G␥ constructs. Cells were lysed and binding to GST-GRK2-(45-178) in the presence (ϩ) or the absence (Ϫ) of AlF 4 Ϫ was determined as described under "Experimental Procedures." The (ϩ) and (Ϫ) lanes represent 40% of the G␣ q or G␣ q mutant pulled down from the 125 l of lysate. In these experiments we detected little to no binding of GST-GRK2-(45-178) to G␣ q or G␣ q point mutants in the absence of AlF 4 Ϫ . Underneath the representative Western blot the percentage of each G␣ q mutant pulled down by GST-GRK2-(45-178) in the presence of AlF 4 Ϫ is compared with the control, which is the percentage of G␣ q pulled down by GST-GRK2-(45-178) in the presence of AlF 4 Ϫ , and is (Table I), and Fig. 4 presents GST pull-down data with RGS2. Mutation of residues in the Switch I region of G␣ q interferes with binding to RGS2 and RGS4. The mutation that has the most profound effect on binding to GRK2, P185K, also does not bind to RGS2 and RGS4 in the context of the RC mutation (Table I and Fig. 4C). However, P185K activated by AlF 4 Ϫ did bind to RGS4 and RGS2 (Table I and Fig. 4A). Finally, the conserved Thr residue (position 187 in G␣ q and 182 in G␣ i ), which is completely buried in the G␣ i -RGS4 interaction, is essential for binding to RGS2 and RGS4 (11). Substitution of Thr 187 with a lysine drastically reduces binding to RGS2 and RGS4 in all active forms (Table I and Fig. 4, A-C) but has no affect on binding to GRK2 (Table I and Fig. 2). In addition, we have previously shown that the RGS-resistant mutant G␣ q -G188S binds to GRK2 (22). Therefore, there are substantial differences between the surface of G␣ q bound by the A site of typical RGS proteins and the C site of the GRK2 RH domain, as predicted by their modeled interactions with G␣ q (Fig. 1, A  and B).
Additionally, there are differences in the binding of RGS2 and RGS4 to AlF 4 Ϫ -activated versus constitutively active RC and QL forms of a few of the helical domain mutants (Table I and Fig. 4). For example, the mutant Q81A-RC has decreased binding to both RGS proteins as does Q81A activated by AlF 4 Ϫ (Table I and Fig. 4, A and C); however, G␣ q -QL and the G␣ q -Q81A-QL mutant bind to RGS2 to a similar level (Fig. 4B). Also, the AlF 4 Ϫ -activated Q152A and G␣ q -Q152A-QL showed decreased binding to RGS2 and RGS4, but Q152A-RC bound RGS2 and RGS4 equally as well as wild-type G␣ q -RC did (Table  I and Fig. 4, A-C).
Q81A-RC and T187K-RC Have Reduced Sensitivity to RGS2mediated Inhibition of Inositol Phosphate Production-We next wanted to examine the ability of RGS2 and RGS4 to inhibit each of the G␣ q point mutants in the RC form. These assays were performed in a manner similar to the inositol phosphate assays with GRK2. The G␤␥ subunits were expressed in every sample, and increasing amounts of RGS2 were co-transfected with each mutant. We were not able to assess the ability of RGS4 to inhibit the G␣ q mutants, because transfection of several different RGS4 constructs decreased the expression of G␣ q or mutants of G␣ q (data not shown). The P185K-RC mutant was not included in these experiments because, like GRK2-K220R, co-expression of RGS2 decreased its expression (data not shown).
Inositol phosphate assays performed to determine the sensitivity of the G␣ q point mutants to inhibition by RGS2 agreed with the data from the GST-RGS2 pull-down experiments. The two mutants that have very little effect on the binding of RGS2 to G␣ q , R92A and Q152A, were also susceptible to RGS2-mediated inhibition of inositol phosphate production (Fig. 5A). In contrast, low amounts of transfected RGS2 DNA did not decrease the inositol phosphate production stimulated by Q81A-RC (Fig. 5A). At the highest amount of RGS2 transfected, 20 ng, Q81A-RC-stimulated inositol phosphate production was decreased by about 25%, whereas similar levels of RGS2 decreased R183C-stimulated inositol phosphate production by 43% (Fig. 5A). This agreed with the pull-down data in Fig. 5 and suggested that this residue is involved in the G␣ q -RGS2 interaction. Finally, Thr 182 in G␣ i , which corresponds to Thr 187 in G␣ q , was found at the center of the G␣ i -RGS4 interface, and therefore mutation at this position in G␣ q should disrupt any interaction with RGS4 and RGS2. Fig. 4 shows that very little T187K-RC was pulled down with GST-RGS2. Fig. 5A also shows that signaling by T187K-RC was not inhibited by co-expression of RGS2. Fig. 3 shows that this mutation had no effect on the ability of GRK2-K220R to inhibit G␣ q signaling, once again highlighting the difference between GRK2 and RGS2 binding sites on G␣ q . As with GRK2, co-expression of RGS2 did not decrease the expression levels of G␣ q -RC or any of the mutants (Fig. 5B).
GRK2 L118A Does Not Interact with G␣ q -Although our previous GRK2 mutagenesis studies were driven by an axin/ GAIP-based homology model of the GRK2 RH domain (22), the crystal structure and the docking model (Fig. 1) used in this study to predict G␣ q residues involved in the interface with GRK2 also identified additional GRK2 residues that could be involved in the interface with G␣ q . Specifically, the new model predicts that Leu 118 of GRK2 is important for the central interaction with Pro 185 of G␣ q . To test this hypothesis, the L118A mutation was introduced into the GST-GRK2-RH domain (residues 45-178) fusion, and GST pull-down assays were used to assess the ability of the mutant to bind to G␣ q/11 from bovine FIG. 3. Effect of the Q152A point mutation in G␣ q -RC on the ability of GRK2 to inhibit inositol phosphate production. A, HEK-293 cells were transfected with 0.1 g of the constitutively active G␣ q -R183C or G␣ q -Q152A-RC and 0.2 g of myc, His-tagged G␤, and 0.1 g of G␥ and increasing amounts of GRK2-K220R and empty vector up to a total of 1.0 g of DNA. 24 h after transfection the cells were labeled with 2 Ci/ml myo-[ 3 H]inositol and 16 h later inositol phosphate production was determined, as described under "Experimental Procedures." The results shown are averages from five independent experiments each done in triplicate and displayed as percent control Ϯ S.D. The control is the inositol phosphate production stimulated by G␣ q -R183C or G␣ q -Q152A-RC in the absence of any co-expressed GRK2-K220R. ૺ, a statistically significant difference (p Ͻ 0.05) by two-way ANOVA followed by a Bonferroni post-test, between the indicated G␣ q -Q152A-RC bar and the G␣ q -RC bar transfected with the same amount of GRK2-K220R. #, a statistically significant difference (p Ͻ 0.05) by one-way ANOVA followed by a Dunnett post-test, between the indicated bar and the control, either G␣ q -RC or G␣ q -Q152A-RC in the absence of co-transfected GRK2-K220R. B, Western blots of total cellular lysates from a representative inositol phosphate experiment from A probed with the EE monoclonal antibody showing that increasing GRK2-K220R expression does not effect expression of G␣ q -RC or G␣ q -Q152A-RC. The bottom panel of Fig. 4B shows the level of GRK2-K220R overexpression. The bands corresponding to 10, 25, and 100 ng of GRK2-K220R transfected can be seen after very short exposures; however, the GRK2 band corresponding to 5 ng of cDNA transfected is barely visible, even after long exposures, suggesting that comparatively low levels of GRK2 expression can significantly inhibit G␣ q signaling. brain extracts in the presence of AlF 4 Ϫ . L118A was markedly impaired in its ability to bind G␣ q/11 , affirming the importance of the solvent-exposed Leu 118 in the G␣ q interaction (Fig. 6). The G␣ q -binding deficiency of L118A is comparable to that of previously identified mutants R106A, D110A, and E116A, which also showed severe impairment in G␣ q/11 binding (22). To compare the L118A mutation to previously identified mutants, the pull-down assay was modified so as to increase its sensitivity. Some binding (except with the R106A/D110A double mutant) can be detected under these conditions allowing the mutants to be ranked based on their decreasing ability to bind G␣ q/11 : R106A Ͼ L118A Ͼ D110A Ͼ R106A/D110A (data not shown). Two additional amino acids in the extended ␣5 helix of GRK2 were also examined; however, neither T111A nor C120A had any effect on the ability of GST-GRK2-(45-178) to bind G␣ q/11 (Fig. 6). DISCUSSION Here, G␣ q -binding residues of GRK2 identified in a previous study (22) and the crystal structure of GRK2 were both used to construct a model of the G␣ q ⅐GRK2 interaction interface. This model was used to identify G␣ q residues and additional GRK2 residues that could be involved in the interface (14). Although we believe the resulting model is globally correct, there is no way short of determining the crystal structure of the complex to know if it is accurate in detail, especially given that no high resolution atomic model of G␣ q currently exists. Therefore, although the manually docked model described here is consistent with the existing biochemical data, it should not be regarded as more than a conceptual tool to help predict regions of G␣ q that are responsible for binding and specificity.
Our results demonstrate that of the G␣ q residues tested, Pro 185 is the most critical for the G␣ q -GRK2 interaction (Fig.  2), although mutation of other residues within Switch I, Switch III, and the helical domain were also found to influence complex formation. Moreover, we show that mutation of G␣ q residues differentially affect interaction with GRK2 and the canonical RGS proteins, RGS2 and RGS4. Specifically, the T187K mutation significantly reduces binding to and inhibition of signaling by RGS2 but does not affect the GRK2-G␣ q interaction (compare Figs. 2 and 3 with Figs. 4 and 5). These results are consistent with the G␣ q ⅐GRK2 interface being distinct, but overlapping, with that of G␣-RGS proteins (Fig. 1, A and B).
Several lines of evidence are consistent with the proposal that the RH domain of GRK2 interacts with the switch regions of G␣ q . First, the activation-dependent nature of the interaction between the GRK2 RH domain and G␣ q strongly suggests that the switch regions of G␣ q are involved. Second, in a study using G␣ q -G␣ 16 chimeras, we have shown that GRK2 binds to a chimeric G␣ protein containing the switch regions of G␣ q but not a chimeric protein containing the switch regions of G␣ 16 , a member of the G␣ q family that does not interact with GRK2 "Experimental Procedures." Results are plotted as described in A. C, HEK-293 cells were transfected with EE-tagged versions of G␣ q -R183C point mutants and G␤ and G␥ constructs. Cells were lysed, and binding of the RC mutants to GST-RGS2 was determined as described under "Experimental Procedures." The lanes labeled "P" in B and C represent 40% of the G␣ q or G␣ q mutant that was present in the pull-down assay from 200 l of lysate. The lanes in A-C labeled "L" represent 4% of total G␣ q or G␣ q mutant available in the lysate for pull-down assay. Results are plotted as described in A. ૺ, the amount of the marked G␣ q mutant pulled-down is significantly different (p Ͻ 0.05) by one-way ANOVA followed by a Dunnett post-test, than the amount of G␣ q pulled-down by GST-RGS2. ૺૺ, statistical analysis could not be performed on the binding of GST-RGS2 to the K77A-QL, P185K-QL, P185K-RC, T187K, or T187K-QL mutants, because there was no detectable pulldown assay result. The data are averages from three to six independent experiments.

FIG. 4. Interaction of GST-RGS2 with G␣ q point mutants activated by AlF 4
؊ , the Q209L or the R183C mutation. A, HEK-293 cells were transfected with EE-tagged versions of G␣ q point mutants and G␤ and G␥ constructs. Cells were lysed, and binding to GST-RGS2 in the presence (ϩ) or the absence (Ϫ) of AlF 4 Ϫ was determined as described under "Experimental Procedures." The (ϩ) and (Ϫ) lanes represent 40% of the G␣ q or G␣ q mutant pulled down from the 125 l of lysate. In these experiments we detected little to no binding of GST-RGS2 to G␣ q or G␣ q point mutants in the absence of AlF 4 Ϫ . Underneath the representative Western blot the percentage of each G␣ q mutant pulled down by GST-RGS2 in the presence of AlF 4 Ϫ is compared with the control, which is the percentage of G␣ q pulled down by GST-RGS2 in the presence of AlF 4 Ϫ , and is represented graphically as the percentage of control Ϯ S.D. B, HEK-293 cells were transfected with EE-tagged versions of G␣ q -Q209L point mutants and G␤ and G␥ constructs. Cells were lysed, and binding of the QL mutants to GST-RGS2 was determined as described under (19). Third, modeling G␣ q onto the RH domain of GRK2 predicts that Pro 185 of G␣ q makes significant contacts with several residues in GRK2, such as Asp 110 and Met 114 , previously identified as being important for the interaction (22). The fact that Pro 185 resides at the crux of the interface would also explain why GRK2 selectively interacts with G␣ q rather than G␣ i , where the corresponding residue is a lysine. Consistent with this idea, the G␣ q -P185K mutant does not bind to GRK2 when activated by AlF 4 Ϫ , but it does bind to RGS4 and RGS2 (Table I  and Figs. 2 and 4). P185K-RC not only fails to bind GRK2 but also does not bind RGS2 or RGS4 (Figs. 2 and 4). Apparently, in the context of G␣ q with the RC mutation, P185K cannot be tolerated. A previous report suggested that Pro 185 and Ile 190 of G␣ q contribute to the higher affinity of RGS2 for G␣ q , versus G␣ i , by affecting the position of Thr 187 relative to the RGS binding pocket (36). If this is true, then it is possible that the combined effects of R183C and P185K change the conformation of Switch I so that it is incompatible with GRK2, RGS2, and RGS4 binding.
The G␣ q residues that mediate critical interactions with the A site of RGS proteins are distinct from those that interact with the C site of GRK2. In the model of the G␣ q ⅐GRK2 complex, Thr 187 is close enough to the interface that mutation to lysine could potentially disrupt the interaction. However, mutation of this residue does not affect the G␣ q ⅐GRK2 interaction and therefore does not represent a critical contact site. In contrast, the T187K mutation has a profound effect on the interaction of both RGS2 and RGS4 with G␣ q (Figs. 4). In addition, RGS2 does not inhibit signaling from the constitutively active, G␣ q -T187K-RC, form of this mutant (Fig. 5). As mentioned previously, the position of Thr 187 in G␣ q relative to the binding pocket of RGS2 may determine in part the selectivity of the RGS2-G␣ q interaction (36).
Further differences between G␣ q binding to GRK2 and RGS2 or RGS4 can be seen by mutating the residues in the helical domain of G␣ q . One such mutation, Q152A, located in a loop between the ␣D and ␣E helices, disrupted binding to the RH domain of GRK2 and inhibition by full-length GRK2 in cellular inositol phosphate assays (Figs. 2 and 3). The G␣ q -Q152A-RC mutant interacted with both RGS2 and RGS4 in pull-down assays, and its stimulation of phospholipase C␤ was inhibited by RGS2 (Figs. 4 and 5), indicating that the Q152A mutation selectively disrupts the interaction of G␣ q with GRK2. We also identified a mutation in the helical domain of G␣ q , Q81A, that has unique effects on RH domain binding specificity. The binding of GST-RGS4, GST-RGS2, and GST-GRK2-(45-178) to AlF 4 Ϫ -activated Q81A was reduced (Table I and Figs. 2 and 4). In contrast, G␣ q -Q81A-RC displayed decreased interaction with RGS2 and RGS4 but exhibited no defect in its interaction with GRK2 (Figs. 2 and 4). Relative to G␣ q -RC, RGS2 inhibition of G␣ q -Q81A-RC-stimulated inositol phosphate production is decreased (Fig. 5), whereas GRK2 inhibits both G␣ q -RC-and G␣ q -Q81A-RC-stimulated inositol phosphate production to similar levels (data not shown). The finding that the Q81A mutation disrupts the interaction of G␣ q with RGS2 and RGS4 is novel. Additionally, this glutamine is conserved only among the G␣ q family and therefore may be a residue in the helical domain of G␣ that could contribute to the targeting of RGS proteins to specific G␣ subunits.
Mutation of other residues in the helical domain of G␣ q decreased the interaction with GRK2. Both the K77A and R92A mutations in G␣ q decreased binding to GRK2 in the context of G␣ q -QL (Fig. 2B). These results likewise suggest that the RH FIG. 5. RGS2 inhibition of inositol phosphate production stimulated by G␣ q -RC and G␣ q -RC mutants. A, HEK-293 cells were transfected with 0.1 g of the constitutively active G␣ q -R183C, G␣ q -Q81A/RC, G␣ q -R92A/RC, G␣ q -Q152A/RC, or G␣ q -RC/T187K and 0.2 g of myc, His-tagged G␤, and 0.1 g of G␥ and increasing amounts of RGS2 and empty vector up to a total of 1.0 g of DNA. 24 h after transfection the cells were labeled with 2 Ci/ml myo-[ 3 H]inositol, and 16 h later inositol phosphate production was determined, as described under "Experimental Procedures." The results shown are averages from three independent experiments each done in triplicate and displayed as percent control Ϯ S.D. The control is the inositol phosphate production stimulated by each mutant in the absence of any co-expressed RGS2. The statistical significance of the difference between the indicated bar and G␣ q -R183C, in the absence of any additional mutations, transfected with equal amounts of RGS2 is denoted by ૺ (p Ͻ 0.05) by one-way ANOVA followed by a Dunnett post-test. #, statistically significant difference (p Ͻ 0.05) by one-way ANOVA followed by a Dunnett posttest, between the indicated bar and the control, either G␣ q -RC or a G␣ q -RC mutant in the absence of co-transfected RGS2. B, Western blots of total cellular lysates from a representative inositol phosphate experiment from A probed with the EE monoclonal antibody showing that increasing RGS2 expression does not effect expression of G␣ q -RC or any of the mutants. We were not able to detect the level of RGS2 overexpression in these experiments; however, the decrease in inositol phosphate production suggests that RGS2 expression was increased.
FIG. 6. Further mapping of the G␣ q/11 binding site on the GRK2 RH domain. Upper panel, glutathione-agarose beads bearing GST fusion proteins, either WT GST-GRK2-(45-178) or GST-GRK2-(45-178) substituted as indicated, were incubated with bovine brain extract (as a source of G␣ q/11 ) in the presence (ϩ) or absence (Ϫ) of aluminum fluoride (AlF 4 Ϫ ). Bound G␣ q /11 was visualized by immunoblotting. Lower panel, fusion proteins used in the GST pull-down assay above were separated by SDS-PAGE and visualized by Coomassie Blue staining. domain of GRK2 interacts with the ␣-helical domain of G␣ q . There is evidence from other studies that the helical domain imparts some of the G␣ selectivity upon interactions with RH domains. Skiba et al. (30) used G␣ t /G␣ i chimeras to show that the specificity of RGS9 for G␣ t resides in the helical region. A second study used RGS2/RGS4 chimeras and point mutants to identify residues in RGS2 that confer G␣ q selectivity (36). The RGS2 residues identified in this study would interact with residues in the ␣A helix in the helical domain of G␣ q and would be repelled by analogous residues in G␣ i (36). In combination with our results, such studies demonstrate that the helical domain of G␣ subunits, in particular ␣A, plays a critical role in the specificity of RH domain-G␣ subunit interactions.
Results from several experiments in this study revealed differences in the ability of G␣ q to interact with GRK2, depending upon whether the G␣ subunit is activated by AlF 4 Ϫ , the R183C mutation or the Q209L mutation. For example, the Q81A mutation disrupts interaction with the RH domain of GRK2 when G␣ q is activated by AlF 4 Ϫ but has no effect in the presence of the activating RC and QL mutations (Fig. 2). In contrast, G␣ q -K77A-QL and G␣ q -R92A-QL display greatly decreased interaction with the RH domain of GRK2, but G␣ q -K77A-RC, G␣ q -R92A-RC, and AlF 4 Ϫ -activated G␣ q -K77A and G␣ q -R92A efficiently interact with the RH domain of GRK2 (Fig. 2). In the co-crystal structure of RGS4 and G␣ i , Asn 128 of RGS4 projects into the active site of G␣ i and contacts the catalytic residue Gln 204 (11). Our model predicts that Gln 209 in G␣ q could also make direct contact with the RH domain of GRK2, by forming hydrogen bonds with the side chain of Asp 110 in GRK2. However, to do so it would have to adopt a different, more extended conformation than that observed for the analogous G␣ i Gln 204 residue in the G␣ i ⅐RGS4 complex (11). Although this extended conformation would be predicted to inhibit GTPase activity, such a conformation of G␣ q may be appropriate for binding to GRK2, which exhibits little or no GAP activity (18). The constitutively active RC and QL forms of several G␣ subunits have been used extensively and somewhat interchangeably to investigate G␣ subunit signaling; however, our results suggest that there are functional differences between these constitutively active mutants that may warrant further investigation.
In conclusion, we have used molecular modeling and mutation studies to identify residues that are important for the interaction between the RH domain of GRK2 and G␣ q . These data confirm the unique characteristics of the interaction between G␣ q and the C site of the GRK2 RH domain and identify new residues in the helical domain of G␣ that selectively disrupt the interaction between G␣ q and RGS2 or RGS4 but not GRK2. The crystal structure of GRK2 and G␤␥ in complex allows for a model in which GRK2 simultaneously interacts with agonist bound receptor, G␤␥, and G␣ q (14). It would also be interesting to investigate the possibility that G␣ q plays a role in directing the G␤␥-recruited GRK2 to specific, activated G␣ q -coupled receptors.