Regulator of G-protein Signaling 3 (RGS3) Inhibits Gβ1γ2-induced Inositol Phosphate Production, Mitogen-activated Protein Kinase Activation, and Akt Activation*

Regulator of G-protein signaling 3 (RGS3) enhances the intrinsic rate at which Gαi and Gαq hydrolyze GTP to GDP, thereby limiting the duration in which GTP-Gαi and GTP-Gαq can activate effectors. Since GDP-Gα subunits rapidly combine with free Gβγ subunits to reform inactive heterotrimeric G-proteins, RGS3 and other RGS proteins may also reduce the amount of Gβγ subunits available for effector interactions. Although RGS6, RGS7, and RGS11 bind Gβ5 in the absence of a Gγ subunit, RGS proteins are not known to directly influence Gβγ signaling. Here we show that RGS3 binds Gβ1γ2 subunits and limits their ability to trigger the production of inositol phosphates and the activation of Akt and mitogen-activated protein kinase. Co-expression of RGS3 with Gβ1γ2 inhibits Gβ1γ2-induced inositol phosphate production and Akt activation in COS-7 cells and mitogen-activated protein kinase activation in HEK 293 cells. The inhibition of Gβ1γ2 signaling does not require an intact RGS domain but depends upon two regions in RGS3 located between acids 313 and 390 and between 391 and 458. Several other RGS proteins do not affect Gβ1γ2 signaling in these assays. Consistent with the in vivo results, RGS3 inhibits Gβγ-mediated activation of phospholipase Cβ in vitro. Thus, RGS3 may limit Gβγ signaling not only by virtue of its GTPase-activating protein activity for Gα subunits, but also by directly interfering with the activation of effectors.

Heterotrimeric G-proteins link seven transmembrane receptors to downstream signaling pathways. Receptor activation triggers the exchange of GTP for GDP by the G␣ subunit of the heterotrimeric G-protein, causing a conformational change in the G␣ subunit, which facilitates its dissociation from the receptor and G␤␥ subunits. GTP-bound G␣ and free G␤␥ subunits then bind and activate downstream effectors. However, G␣ subunits possess an intrinsic GTPase activity, which returns G␣ to its GDP bound state and thereby limits the duration of G␣ signaling. Because GDP-G␣ possesses a high affinity for G␤␥ subunits, the heterotrimeric G-protein rapidly reforms, effectively ending G␤␥mediated signaling as well (reviewed in Refs. 1 and 2).
Cells possess another important mechanism that curtails the duration in which a G␣ subunit remains GTP bound. Members of a family of proteins termed regulators of G-protein signaling (RGS) 1 dramatically accelerate the intrinsic rate that certain G␣ subunits hydrolyze GTP, a property that identifies them as GTPase-activating proteins (GAPs). The mammalian RGS proteins have a 120-amino acid region, RGS domain, or RGS box, which binds G␣ i and G␣ q subfamily members in a transition state of the GTP hydrolysis reaction, thereby lowering the free energy of the reaction (reviewed in Refs. 3 and 4). In addition, Rho guanine nucleotide exchange factors have a divergent RGS domain, which accelerates the intrinsic GTPase activity of G␣ 12 and G␣ 13 (5,6). RGS proteins with GAP activity for members of the G␣ s subfamily remain enigmatic.
The mammalian RGS proteins can be broadly divided into two groups (7): those composed predominantly of an RGS domain such as RGS1, RGS2, RGS4, and RGS5 and those that contain an RGS domain but also other domains. The second group includes RGS6, RGS7, RGS9, RGS11, RGS12, and RGS14. The smaller RGS proteins likely function solely as G␣ GAPs, whereas some of the larger RGS proteins are undoubtedly G␣ effectors such as p115 Rho guanine nucleotide exchange factors. RGS3 exists as two isoforms, thus falling into both groups (8,9): a shorter version that encodes largely an RGS domain (RGS3CT) and a larger isoform that has a strongly acidic region and an unusual region that contains a hexapeptide repeat enriched for proline, glutamine, and acidic residues (8,9). Both versions possess GAP activity for G␣ i and G␣ q and can impair signaling through G␣ i and certain G␣ qlinked signaling pathways (10). The function of the N-terminal domain of RGS3 is unknown, although the N-terminal fragment shifts to membranes after a calcium signal (11). Transient expression of RGS3 potently inhibits the chemotaxis of a pre-B cell line, even better than does RGS1, which has excellent G␣ i GAP activity (12,13). The known importance of G␤␥ signaling in chemokine-directed migration (14,15) led us to test whether RGS3 employs another mechanism besides its G␣ GAP activity to impair G␤␥ signaling. In three in vivo models of G␤ 1 ␥ 2 signaling, inositol phosphate generation through the stimulation of phospholipase C␤, the activation of mitogen-activated protein kinase (MAPK), and the activation of Akt, we find that RGS3 potently inhibits the activation of these signaling pathways even when it lacks a functional RGS domain. Supporting the in vivo data, purified RGS3 blocks G␤␥-induced inositol phosphate production by phospholipase C␤2 in vitro.
Cell Culture-HEK 293 cells and COS-7 cells were obtained from the American Type Culture Collection (Manassas, VA). The cells were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum. The transfected cells were plated in 6-well plates. For the MAPK assays the cells were re-plated in 0.5% serum, and for the Akt assays they were re-plated in 0.5% serum for 24 h, at which point the media was replaced with Dulbecco's modified Eagle's medium without serum.
Recombinant Proteins-The His-tagged recombinant RGS3 and RGS4 were purified from Escherichia coli BL21(DE3) inducted with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside at 30°C for 2-12 h. The recombinant proteins were batch-purified under nondenaturing conditions using nickel nitrilotriacetic acid beads (Qiagen, Santa Clara, CA) and eluted with an imidazole gradient. The purified protein fractions were dialyzed against the wash buffer and stored at Ϫ70°C. In some instances RGS3 was further purified over a monoQ column (Amersham Pharmacia Biotech). The recombinant His-tagged Jun kinase was purchased from Santa Cruz (Santa Cruz, CA). GST-RGS3 fusion proteins were prepared from lysates of HEK 293 cells previously transfected with the appropriate expression vector. The fusion proteins were partially purified on glutathione-Sepharose 4B as described by the manufacturer (Amersham Pharmacia Biotech). The eluted GST-RGS3 fusion proteins were concentrated and dialyzed using Centricon 30 (Millipore, Bedford, MA) with the equilibration buffer (20 mM Hepes, 1 mM EDTA, 1 mM EGTA, pH 7.2) containing 0.5 mM aminoethylbenzenesulfonyl fluoride, 10 g/ml leupeptin, and 10 g/ml aprotinin. The resulting concentrate (1.5 ml) was applied using a 2-ml sample loop to a MonoQ HR5/5 column (Amersham Pharmacia Biotech) that had been equilibrated with the equilibration buffer in the HP1100 high performance liquid chromatography system (Hewlett Packard, Palo Alto, CA). All operations were performed at 4°C or on ice. Proteins were eluted at a flow rate of 1 ml/min by successive applications of the equilibration buffer for 5 min and an increasing linear gradient from 0 to 0.6 M NaCl for 60 min. Fractions (1 ml) were collected, and 20 l of each fraction were separated on 10% gel by SDS-PAGE. The fusion proteins were identified in SDS-PAGE gels by silver staining. All the GST-RGS3 fusion proteins were eluted around 0.4 M NaCl. Peak fractions (three fractions) were pooled, concentrated to ϳ0.1 ml by Centricon 30, divided into portions, and stored at Ϫ70°C. The recombinant G␤ 1 ␥ 2 was prepared from Sf9 cells infected with baculovirus constructs encoding G␤ 1 , G␥ 2 , and His-tagged G␣ i1 according to the published procedure (19).
Immunoblotting and Immunoprecipitations-Cell lysates were prepared using a solution containing 150 mM NaCl, 50 mM Tris, pH 7.5, 5 mM EDTA, and 1% Triton X-100 along with a mixture of protease inhibitors for 20 min on ice. The detergent-insoluble material was removed by microcentrifugation for 10 min at 4°C. Equal amounts of protein from each sample were fractionated by SDS-PAGE and transferred to pure nitrocellulose. Membranes were blocked with 10% milk in TTBS for 1 h and then incubated with an appropriate dilution of the primary antibody in 5% milk and 0.05% sodium azide in TTBS overnight. The blots were washed twice with TTBS before the addition of a biotinylated goat-anti rabbit immunoglobulin (DAKO, Carpinteria, CA) diluted 1:1500 in TTBS containing 5% fetal calf serum. After a 1-h incubation, the blot was washed twice with TTBS and then incubated with streptavidin conjugated to horseradish peroxidase (DAKO). The signal was detected by enhanced chemiluminescence (ECL) following the recommendations of the manufacturer (Amersham Pharmacia Biotech). The co-immunoprecipitations were performed using lysates (20 mM Tris, pH 8.0, 137 mM NaCl, 2 mM EDTA, 1% Triton X-100, 1 mM sodium orthovanadate, plus protease inhibitors) prepared from COS-7 cells or HEK 293 cells co-expressing G␤ 1 ␥ 2 and various RGS proteins. Anti-FLAG monclonal antibody or anti-G␤ polyclonal antiserum was added, and the immunoprecipitates were collected with the appropriate secondary antibody-coupled magnetic beads (Dynal Corp., Lake Success, NY). They were washed three times in lysis buffer, twice in lysis buffer with 0.5 M NaCl, fractionated by SDS-PAGE, and analyzed by immunoblotting with the appropriate antibody.
In Vivo Inositol Phosphate Production-Inositol phosphate production was measured as previously described (20). Briefly, COS-7 cells were transfected using LipofectAMINE (1:8). Twenty-four hours after transfection, the culture media was replaced with inositol-free Dulbecco's modified Eagle's medium containing 5% fetal calf serum and 1 mM sodium pyruvate for 2 h, after which 2 Ci/ml myo-[2-3 H]inositol (Amersham Pharmacia Biotech) were added and, 15 min later, 10 mM LiCl. The cells were incubated for an additional 14 h before washing with phosphate-buffered saline followed by the addition of 0.5 ml of 20 mM formic acid. Thirty min later the supernatant was collected, and a second extraction was performed. Each 1-ml extract was neutralized to pH 7.5 with 7.5 mM Hepes and 150 mM KOH. The supernatants were centrifuged for 2 min at 15,000 ϫ g, collected, and each loaded onto to a 0.5-ml Dowex AG-X8 column (Bio-Rad) that had been previously washed with 2 ml of 1 M NaOH, 2 ml of 1 M formic acid, and 5 washes of 5 ml of water. After loading the sample, the column was washed with 5 ml of water and 5 ml of 5 mM borax and 60 mM sodium formate. The columns were eluted with 3 ml of 0.9 M ammonium formate and 0.1 M formic acid. 0.2 ml of each elution was added to 10 ml of CytoScint and analyzed via scintillation counting.
MAPK and Akt Assays-Equal numbers of HEK 293 (MAPK assay) or COS-7 (Akt assay) cells were plated on 10-cm plates at an approximate density of 5 ϫ 10 5 cells/plate the day before transfection. The HEK 293 or COS-7 cells were transiently transfected with pCMVHA-ERK1 or pCMV6Akt-HA and various RGS expression vectors using a standard calcium phosphate method or Superfect (Qiagen). Transfected DNA levels were normalized with control plasmids. Forty-eight hours after the transfection, HA-immunoprecipitates were subjected to in vitro kinase assays using myelin basic protein (MAPK assay) or H2-B (Akt assay) as a substrate. The in vitro kinase assays were performed as previously described (21,22). Before the in vitro kinase assay the HA immunoprecipitates were washed three times with kinase lysis buffer (20 mM Hepes, pH 7.4, 2 mM EGTA, 50 mM ␤-glycerophosphate, 1% Triton X-100, 1 mM Na 3 V0 4 , and 10% glycerol) to which a protease inhibitor mixture tablet was added), three times with a LiCl wash buffer (500 mM LiCl, 100 mM Tris, pH 7.4, 0.1% Triton X-100, and 1 mM dithiothreitol), and three times with kinase buffer (20 mM MOPS, pH 7.2, 2 mM EGTA, 10 mM MgCl 2 , and 0.1% Triton X-100). The reactions were size-fractionated by SDS-PAGE and autoradiographed. The levels of HA-ERK1, Akt-HA, and RGS protein expression were detected by immunoblotting cell lysates. The autoradiographs were analyzed using NIH Image.
In Vitro Reconstitution Assay for PLC Activation-Reactions were started by preincubation of GST-RGS3 fusion proteins (0.5-2 g) and G␤␥ subunits (0.5 g) in a 25-l volume for 10 min at 30°C. After incubation, reaction tubes were transferred on ice and prepared for the PLC activity. PLC activity was measured as previously described (18). Briefly, substrate was prepared as sonicated micelles of 75 M [ 3 H]phosphatidylinositol 4,5-bisphosphate (10,000 -12,000 cpm/assay) and 750 M phosphatidylethanolamine. 10 ng of PLC-␤2 was added to each tube, and CaCl 2 was added to the assay mixture to give 200 nM free Ca 2ϩ . Assays were performed in 60-l volume for 10 min at 30°C. Under these conditions, [ 3 H]IP 3 formation was linear with respect to time and enzyme concentration.

RESULTS
RGS3 Impairs the Generation of Inositol Phosphates by G␤ 1 ␥ 2 -Since RGS3 possesses GAP activity for both G␣ i and G␣ q analyzing whether RGS3 modulates G␤␥ signaling through either G q -or G i -coupled receptors is not feasible. However, the intracellular expression of G␤ 1 ␥ 1 , G␤ 1 ␥ 5 , or G␤ 5 ␥ 5 is known to activate PLC␤2, resulting in the production of intracellular inositol phosphates (23). The newly synthesized G␤␥ subunits target to the plasma membrane, where insufficient G␣ subunits exist to form inactive heterotrimers. The free G␤␥ subunits can activate PLC␤ enzymatic activity, which hydrolyzes phosphatidylinositol 4,5-bisphosphate, releasing the second messengers inositol 1,4,5-trisphosphate and diacylglycerol. Therefore, by concomitantly introducing a construct that expresses RGS3, we could test whether RGS3 modulates the production of inositol phosphates by free G␤␥ subunits. We found that expression of G␤ 1 ␥ 2 enhanced inositol phosphate production by COS-7 cells, co-expressing PLC␤2 7-9-fold over background levels. The introduction of RGS3 inhibited inositol phosphate production by nearly 70%, whereas two other RGS proteins, RGS4 and RGS10, did not impair inositol phosphate production in similar experiments (Fig. 2). The expression of RGS3 did not alter the levels of G␤ subunits or PLC␤2, although the transfected G␤ 1 subunits co-migrate with the endogenous G␤ subunits. Thus, although RGS4 has better GAP activity for G␣ i than does RGS3 and equivalent G␣ q activity (10), RGS4 does not impair G␤␥ signaling to phospholipase C␤ like RGS3 does.
RGS3 Inhibits the Activation of MAPK and Akt by G␤ 1 ␥ 2 -G␤␥ subunits directly activate other effectors beside PLC␤. Distinct sets of contacts along the edge of the ␤ propeller of G␤ subunits are thought to establish the specificity of G␤␥ for its effectors (24,25). Through yet unknown effectors G␤␥ subunits stimulate MAPK and Akt activation. Expression of G␤ 1 ␥ 2 subunits in COS-7 cells enhances the activity of the MAPK ERK2 (26). Therefore, by concomitantly expressing RGS3 we could test whether RGS3 inhibits a second G␤␥-signaling pathway. Rather than COS-7 cells, we used HEK 293 cells, thereby changing both the signaling pathway and the cell environment from the previous experiments. The introduction of G␤ 1 ␥ 2 into HEK 293 increased ERK activation approximately 7-fold as assessed by the amount of phosphorylated substrate generated in an in vitro kinase assay. The introduction of RGS3 reduced G␤ 1 ␥ 2 -mediated ERK activation to the level observed in the absence of G␤ 1 ␥ 2 . In contrast, neither RGS4 nor RGS10 appreciably altered G␤ 1 ␥ 2 signaling in these cells (Fig. 3).
To test the importance of the N-terminal region of RGS3 and an intact RGS domain, we used three additional RGS3 constructs. The N-terminal portion of RGS3, expressed by the RGS3NT construct, did not block G␤ 1 ␥ 2 -mediated ERK activation but, rather, had a modest activating effect. The RGS3CT construct, which expressed the C-terminal portion of RGS3, amino acids 314 -519, suppressed G␤ 1 ␥ 2 MAPK activation as well as did wild type RGS3. The RGS3EN construct, which expressed an RGS3 with alanine (A) at positions 419 and 420 instead of glutamic acid (E) and asparagine (N), inhibited G␤ 1 ␥ 2 -induced ERK activation similar to that of wild type. A similar EN to AA mutation in RGS4 destroys its G␣ i GAP activity and its inhibitory effect on lymphocyte chemotaxis (11,27,28). Indicating that RGS3 inhibits upstream in the MAPK pathway, it did not perturb the activation of ERK by an activated form of Ras (Fig. 3). Thus, to inhibit G␤ 1 ␥ 2 activation of the MAPK pathway, RGS3 requires neither its N terminus nor an intact RGS domain.
Besides MAPK activation, RGS3 also interfered with activation of Akt. Akt is implicated in pathways leading to cell survival in response to serum and growth factors (reviewed in Ref. 29), and several G protein-coupled receptors stimulate Akt activation in part by the release of G␤␥ subunits from G i and G q (30). Expression of G␤ 1 ␥ 2 in COS-7 cells has been shown to increase the activity of a co-expressed epitope-tagged version of Akt in an in vitro kinase assay (30). We performed similar experiments in conjunction with constructs that express wild type or altered RGS proteins. G␤ 1 ␥ 2 expression increased Akt activity severalfold, and the presence of RGS3, RGS3CT, or RGS3EN reduced it nearly to background levels. RGS3NT2 (amino acids 1-458) also significantly reduced Akt activation, especially since it did not express particularly well. RGS3NT, RGS3NT1 (amino acids 1-390), RGS2, and RGS4 had little effect (Fig. 4). Despite its inhibition of G␤ 1 ␥ 2 -induced Akt activation, RGS3 did not alter Akt activation by an activated form of Ras (Fig. 4).
Mapping the Regions in RGS3 Important for Inhibiting G␤ 1 ␥ 2 -induced Inositol Phosphate Production-We also mapped the regions of RGS3 necessary for inhibiting G␤ 1 ␥ 2mediated phospholipase C␤ activation. We transfected COS-7 cells with constructs that expressed G␤ 1 ␥ 2 and PLC␤2 in the presence or absence of constructs that expressed RGS3, RGS3CT, RGS3NT, RGS3NT1, RGS3NT2, or RGS4/3 and measured inositol phosphate production. The RGS4/3 construct encoded a fusion protein that included the first 58 amino acids of RGS4 along with amino acids 390 -519 of RGS3. Similar to the MAPK experiments, RGS3CT inhibited inositol phosphate production by G␤ 1 ␥ 2 even better than did RGS3. RGS3NT2, which lacked the C-terminal 61 amino acids of RGS3, behaved like wild RGS3 did. Comparison of RGS3, RGS3NT, RGS3NT1, and RGS3NT2 indicated that amino acids 314 -390 and 391-458 both contributed to the inhibitory effect of RGS3 on G␤ 1 ␥ 2 signaling to phospholipase C␤ (Fig. 5). This differed slightly from the findings with Akt activation, where the RGS3NT1 construct did not impair activation. Transferring amino acids 390 -519 to the N terminus of RGS4 conferred upon RGS4 the ability to inhibit G␤ 1 ␥ 2 signaling equivalent to that of wild type RGS3, although not equivalent to RGS3CT. Overall these results show that RGS3 requires the residues between 314 -390 and 391-458 to fully inhibit G␤ 1 ␥ 2 signaling.
RGS3 Co-Immunoprecipitates with G␤␥ Subunits-We determined whether RGS3 associated with G␤ 1 ␥ 2 in vivo by cotransfecting HEK 293T cells with constructs directing the expression of full-length RGS3 and G␤ 1 ␥ 2 and analyzing immunoprecipitates for the respective proteins. The FLAG antibody readily immunoprecipitated the FLAG-tagged RGS3 and in addition co-immunoprecipitated G␤␥ subunits (Fig. 6A). Conversely, a polyclonal antibody directed against G␤ subunits immunoprecipitated both endogenous G␤␥ subunits and the expressed G␤ 1 ␥ 2 subunits and co-immunoprecipitated the FLAG-tagged RGS3 (Fig. 6A). Control ERK and Myc antibodies failed to immunoprecipitate either FLAG-RGS3 or G␤␥ subunits. Next, we examined whether RGS3 directly bound G ␤␥ subunits by combining recombinant His-tagged RGS3 with purified G␤␥ subunits and then examined anti-His immunoprecipitates for RGS3 and G␤␥ subunits. Anti-His immunoprecipi- tates contained both RGS3 and G␤␥ subunits (Fig. 6B). In contrast, His-tagged Jun kinase failed to co-immunoprecipitate G␤␥ subunits, and neither RGS3 nor G␤␥ subunits immunoprecipitated with an HA antibody. These results indicate that RGS3 can directly bind G␤␥ subunits.
RGS3 expressed in cell lines localizes predominantly in the cytosol and can be shifted to membranes by G-protein signaling (11). A previous study indicates that endogenous RGS3 localizes predominantly in a membrane fraction rather than the cytosolic fraction of ␣T3-1 cells (31). The RGS3 antiserum used in these experiments, while raised against recombinant RGS3, reacted only with RGS3 and not with RGS3CT. Here we used an anti-peptide antiserum prepared against a peptide from the N-terminal portion of RGS3. This antiserum recognized recombinant and transfected full-length RGS3 and detected a band that co-migrated with recombinant RGS3 in COS cells (10). Fractionation of COS-7 cells into a membrane-enriched and -depleted fractions revealed that the membrane-enriched fraction contained the majority of the RGS3, although some RGS3 also localized in the cytosolic fraction (data not shown). When we examined G␤ immunoprecipitates using the broadly reactive G␤ antiserum, we found that they contained small amounts of RGS3 (Fig. 6C). Conversely, RGS3 immunoprecipitates also contained small amounts of G␤ subunits. Thus, in COS-7 cells, endogenous RGS3 localizes predominantly at intracellular or plasma membranes, and a small portion is constitutively associated with G␤␥ subunits.
Both the N-Terminal Portion of RGS3 and RGS3CT Co-Immunoprecipitate with ␤␥ Subunits-Based on the signaling experiments we expected to co-immunoprecipitate RGS3CT with G␤␥ subunits but not RGS3NT. In contrast, we found that RGS3NT, RGS3NT1, RGS3NT2, and RGS3CT all co-immunoprecipitated with G␤␥ subunits when we co-expressed them along with G␤ 1 ␥ 2 . RGS3NT very readily co-immunoprecipitated with a polyclonal antibody against G␤ subunits but failed to immunoprecipitate when we used a polyclonal antibody against the HA epitope (Fig. 7A). When we co-expressed G␤ 1 ␥ 2 with RGS2, RGS3, or RGS4, G␤ subunits did not co-immunoprecipitate with either RGS2 or RGS4, but they did with RGS3 (Fig. 7B) 4. RGS3 reduces G␤ 1 ␥ 2 -induced Akt activation but not RasV12induced Akt activation. COS-7 cells were transfected with constructs directing expression of G␤ 1 , G␥ 2 , and HA-Akt along with various RGS constructs or not. COS-7 cells were also transfected with constructs directing the expression of RasV12 and HA-Akt, and RGS3 or a control. HA-immunoprecipitates were subjected to an in vitro kinase assay using H2-B as a substrate. Scanning the autoradiographs assessed the amounts of 32 P incorporated into substrate. The fold induction compared with the culture without G␤ 1 ␥ 2 is indicated. The levels of HA-Akt, G␤, and RGS proteins in the cell lysates are shown. The results are representative of one of three experiments performed. PLC␤2 activity by nearly 50% and RGS3CT markedly inhibited it, whereas GST-RGS3NT slightly enhanced the generation of IP 3 (Fig. 8). Thus, the results with GST-RGS3 fusion proteins closely paralleled the in vivo results we had previously obtained. Of note, the purified GST-RGS3 was less stable than either the GST-RGS3NT or GST-RGS3CT. DISCUSSION RGS3 inhibits G␤ 1 ␥ 2 -mediated Akt activation and inositol phosphate production in COS-7 cells and MAPK activation in HEK 293 cells. The inhibition of these pathways by RGS3 does not depend upon an intact RGS domain or on its N-terminal 390 amino acids but requires a region in RGS3 that overlaps the RGS domain. These data are best explained by RGS3 blocking G␤␥-mediated activation of its effectors. Less likely, RGS3 could directly inhibit the effectors per se or impair downstream elements in the signaling pathways. Arguing for RGS3 acting at the level of G␤␥ signaling to its effectors, RGS3 does not reduce phorbol ester-induced ERK activation (8) or Ras-mediated ERK or Akt activation (this study).
How might RGS3 block G␤␥-mediated effector signaling? The failure of RGS4 and the success of RGS3EN and the truncated RGS3 proteins in inhibiting G␤ 1 ␥ 2 signaling indicates that the G␣ i and G␣ q GAP activity of RGS3 cannot explain its success in inhibiting G␤ 1 ␥ 2 -triggered ERK activation. Also the N-terminal portion of RGS3 does not assist in the inhibition of G␤ 1 ␥ 2 signaling since RGS3CT also more potently inhibits. Since RGS3 binds G␤ 1 ␥ 2 subunits both in vitro and in vivo and since overexpressed RGS3 localizes in part in the cytosol, the transiently expressed RGS3 could sequester G␤ 1 ␥ 2 away from its usual microenvironment, thereby inhibiting access of G␤ 1 ␥ 2 to its effectors. However, although both RGS3NT and RGS3CT bind G␤␥ subunits, only RGS3CT potently inhibits G␤ 1 ␥ 2 signaling. Furthermore, RGS3 and RGS3CT block the activation of PLC␤2 in vitro, whereas RGS3NT does not. This last result indicates that RGS3 can directly inhibit PLC␤ activation of G␤␥ subunits. Thus, we favor the possibility that, when localized in the vicinity of heterotrimeric G-proteins undergoing GDP-GTP exchange, RGS3 can interact with GTPbound G␣ subunits, promoting G␣ GTP hydrolysis, and with G␤␥ subunits, reducing their availability for effector activation.
Among the known RGS proteins. is RGS3 unique in its ability to inhibit G␤␥ signaling? Neither RGS4 nor RGS10 inhibits G␤ 1 ␥ 2 -mediated MAPK activation or the generation of inositol phosphates, indicating that not all RGS proteins share this property. Furthermore, neither RGS2 nor G␣ interacting protein inhibits the induction of inositol phosphates after G␤ 1 ␥ 2 expression in COS-7 cells. 3 Thus among five RGS proteins, only RGS3 reduces G␤ 1 ␥ 2 -mediated signal transduction, suggesting that the inhibition of G␤␥ signaling may be a unique property of RGS3 or, if not, only possessed by a minority of the RGS proteins. Alignment of RGS3CT with RGS2, RGS4, RGS10, and G␣ interacting protein reveals considerable differences between the five proteins in the regions N-terminal to the RGS domains. Within the RGS domains these proteins share ϳ25% of their amino acids, and RGS3 possesses 22 unique residues not found in the other four proteins. To delineate critical residues in RGS3 needed for its inhibition of G␤ 1 ␥ 2mediated signaling, we created 5 constructs with triple-residue point mutations in the region that encodes amino acids 350 -415 of FLAG-RGS3. Four or the five proteins expressed in COS-7 cells; however, all of the expressed mutant RGS3 proteins lost their ability to inhibit G␤ 1 ␥ 2 signaling. This suggested that we had introduced conformational changes is this region rather than identifying specific residues needed for the inhibition. 3 Further mapping with RGS3 proteins containing a single point mutation is in progress.
The structural basis of RGS3 interaction with G␤ 1 ␥ 2 is not revealed by comparing the primary amino acid sequence of RGS3 to those of proteins known to directly interact with G␤␥. A motif, QXXER, present in the G␤␥ effectors adenylyl cyclase II and Ca 2ϩ and Na ϩ channels may specify their interaction with ␤␥ subunits (32,33). However, no such motif occurs in RGS3. Phosducin, a phosducin like protein (PhLP), and beta disruption mimic factor-1 (BDM-1) all share a highly conserved 11-amino acid region needed for G␤␥ binding (34 -36). However, again no such stretch of amino acids exists in RGS3. Interesting, like full-length RGS3, phosducin has two separate domains that bind G␤␥. The N-terminal domain of phosducin binds loops on the "top" of the G␤ t surface, overlapping the G ␣ binding surface, whereas the N-terminal domain binds the outer strands of G␤ t seventh and first blades, which may disrupt the normal orientation of G␤␥ t relative to the membrane and receptor (37). Suggesting a regulatory role for phosphorylation, phosphophosducin no longer competes with G␣ t for binding to G␤␥ t (38). Similar structural studies of the RGS3/G␤␥ complex should provide insights into the biologic role of the interaction. Finally, RGS3 does not possess a PH domain such as PLC-␤2 or ␤-adrenergic receptor kinase through which it could interact with G␤␥ (39,40).
Why did the E. coli and mammalian-expressed RGS3 differ in their ability to block G␤␥-mediated activation PLC␤ in vitro? Despite binding purified G␤␥ subunits in vitro, E. coli-expressed recombinant RGS3 minimally affected the ability of G␤␥ subunits to activate recombinant PLC␤, 4 whereas the GST-RGS3CT-purified from mammalian cells markedly inhibited the ability of G␤␥ subunits to activate recombinant PLC␤. One attractive explanation is that an in vivo post-translational modification of RGS3 controls whether it inhibits G␤␥ effector activation. Phosducin undergoes such a regulation. Phosducin and phosphophosducin possess similar affinities for G␤␥ t , yet phosducin competes with G␣ t for binding to G␤␥ t , whereas phosphophosducin does not (38). Since RGS3 undergoes extensive phosphorylation in vivo, 5 perhaps the mammalian-expressed RGS3 is phosphorylated, allowing it to inhibit G␤␥induced PLC␤ activity. Another possibility is that the bacterially expressed RGS3 is mis-folded. Although still able to bind G␤␥ in vitro, it is unable to inhibit G␤␥ signaling, because a portion of the molecule is disordered. Further studies with truncated RGS3-GST fusion proteins should allow a more precise mapping of the regions in RGS3 necessary for inhibiting G␤␥-induced PLC␤ activity.
Does RGS3 interact with heterotrimeric G-proteins? Although we show that G␤ immunoprecipitates from COS-7 cell contain RGS3, neither G␣ q nor G␣ i immunoprecipitates from the same cells do. 3 The G␣ q and G␣ i immunoprecipitates contain as much if not more G␤ than does the G␤ immunoprecipitates. These data argue that RGS3 does not constitutively associate with G i or G q and is consistent with previous studies where RGS3 coimmunoprecipitates with G␣ 11 or G␣ i only after exposure of the cell lysates to AlF 4 Ϫ (11), 5 which dissociates the heterotrimeric G-proteins and activates the G␣ subunits. Together these observations suggest that endogenous RGS3 may associate with a pool of G␤␥, which is not bound G␣. Several studies indicate the presence of an intracellular pool of G␤␥ not associated with G␣ (41,42). A pool of free G␤␥ subunits has been postulated to function in receptor-mediated endocytosis (42).
In summary, elevating the levels of RGS3 potently inhibits 3  the activation of several distinct signaling pathways stimulated by G␤␥ subunits. A portion of RGS3 located between amino acids 314 -458, which partially overlaps the RGS3 RGS domain, is required to observe the inhibition. RGS3 co-immunoprecipitates with transfected G␤ 1 ␥ 2 subunits, binds G␤␥ subunits in vitro, and inhibits the activation of PLC␤ by G␤ 1 ␥ 2 subunits. Overall, these results indicate that RGS3 can modulate G␤␥ signaling not only by virtue of its G␣ GAP activity but also by directly interfering with G␤␥ signaling. Such a mechanism may contribute to the efficacy of RGS3 as an inhibitor of chemokine-directed cell migration.