Complexes of the G Protein Subunit Gβ5 with the Regulators of G Protein Signaling RGS7 and RGS9

A novel protein class, termed regulators of G protein signaling (RGS), negatively regulates G protein pathways through a direct interaction with Gα subunits and stimulation of GTP hydrolysis. An RGS subfamily including RGS6, -7, -9, and -11, which contain a characteristic Gγ -like domain, also has the unique ability to interact with the G protein β subunit Gβ5. Here, we examined the behavior of Gβ5, RGS7, RGS9, and Gα in tissue extracts using immunoprecipitation and conventional chromatography. Native Gβ5 and RGS7 from brain, as well as photoreceptor-specific Gβ5L and RGS9, always co-purified as tightly associated dimers, and neither RGS-free Gβ5 nor Gβ5-free RGS could be detected. Co-expression in COS-7 cells of Gβ5 dramatically increased the protein level of RGS7 and vice versa, indicating that cells maintain Gβ5:RGS stoichiometry in a manner similar to Gβγ complexes. This mechanism is non-transcriptional and is based on increased protein stability upon dimerization. Thus, analysis of native Gβ5-RGS and their coupled expression argue that in vivo, Gβ5and Gγ-like domain-containing RGSs only exist as heterodimers. Native Gβ5-RGS7 did not co-precipitate or co-purify with Gαo or Gαq; nor did Gβ5 L-RGS9 with Gαt. However, in transfected cells, RGS7 and Gβ5-RGS7 inhibited Gαq-mediated Ca2+ response to muscarinic M3 receptor activation. Thus, Gβ5-RGS dimers differ from other RGS proteins in that they do not bind to Gα with high affinity, but they can still inhibit G protein signaling.

ate intracellular effectors. Upon receptor activation, the G␣ subunit of the heterotrimer binds GTP, and the signal is terminated when the bound GTP is hydrolyzed. The intrinsic GTPase activity of G␣ subunits is relatively slow compared with the signaling seen in many physiologic responses suggesting that additional factors are needed to accelerate GTPase activity in vivo. One class of GTPase-activating proteins (GAPs) 1 for heterotrimeric G proteins are their effectors; cGMP phosphodiesterase (1), phospholipase C (2), and adenylyl cyclase V (3). Recently, a new class of GAPs, termed regulators of G protein signaling (RGS) proteins, has emerged (4 -7). The currently established role of RGS proteins is to negatively regulate G protein-linked signaling pathways by reducing the lifetime of G proteins in the active GTP-bound state. RGS proteins achieve this by functioning as GAPs for G␣ subunits (4,8,9) and/or through a GTPase-independent mechanism (10,11). RGS proteins are characterized by a homologous 120-amino acid region, referred to as the RGS domain, that is responsible for binding G␣ and stimulating its GTPase rate (12,13). Outside of this RGS domain, however, the more than 30 members of the family are structurally diverse. Other structural elements found in RGS proteins include PDZ, pleckstrin homology, proline-rich, and Dbl-homology (DH) domains, which might mediate subcellular targeting, assemble signaling complexes, or be involved in the regulation of RGS GAP activity (6,7).
A distinct subfamily of RGS proteins, including RGS6, -7, -9, and -11, contains two unique domains. One is the Disheveled/ Egl-10/pleckstrin (DEP) domain, which is found at the N-terminal part of the proteins, and has an unknown function. The other domain is the G protein ␥-like (GGL) domain, which is responsible for the specific interaction with the neurospecific G protein ␤ subunit, G␤ 5 (14 -16). Reconstitution of G␤ 5 and RGS7 in vitro has shown that G␤ 5 preferentially forms a heterodimer with RGS7 over G␥ 2 (15). The dimers of G␤ 5 and RGS6, -7, and -11 display GAP activity toward G␣ o in vitro, but not other G␣ subunits (14,17). The G␤ 5 L-RGS9 complex from photoreceptor outer segments can act as a GAP for G␣ t in concert with its effector, the ␥ subunit of cGMP phosphodiesterase (PDE␥) (18). Importantly, previous work has shown that G␤ 5 -RGS complexes exist in vivo (18 -20). In contrast, despite the fact that G␤ 5 has been shown to interact with G␥ subunits in vitro (21)(22)(23)(24), G␤ 5 ␥ complexes have never been detected in native tissues. Here, we asked if G␤ 5 is always complexed with RGS proteins, does it exist as a monomer, or is it associated with G␥ in native tissues. We found that in the cytosol and detergent extracts of membranes from both the retina and brain, G␤ 5 exists exclusively as a dimer with RGS. We also demonstrate through co-expression of G␤ 5 and RGS7 in cultured cells that a mechanism regulating the stoichiometry of these proteins exists. Furthermore, although G␤ 5 -RGS complexes do not bind to G␣ subunits with high affinity, the complex can inhibit G␣ qmediated signaling through the muscarinic M3 receptor.

EXPERIMENTAL PROCEDURES
Antibodies-Polyclonal antisera were raised against a synthetic peptide corresponding to amino acids 463-477 of bovine RGS9 by Alpha Diagnostic International. Antibodies against RGS7, G␤ 1 , and G␤ 5 were described previously (15,19). RGS7 and RGS9 antisera were affinitypurified by passing the serum over a Sulfolink column (Pierce) with covalently attached immunizing peptide. Anti-G␣ o antibody was kindly provided by Dr. Allen Spiegel (National Institutes of Health, Bethesda, MD), anti-G␣ t by Dr. Melvin Simon (Caltech, Pasadena, CA), anti-G␣ q by Dr. David Manning (University of Pennsylvania, Philadelphia, PA), anti-PDE␥ by Dr. Eva Faurobert (IPMC-CNRS, Valbonne, France), anti-arrestin by Dr. Vsevolod Gurevich (Sun Health Research Institute, Sun City, AZ), and anti-phosducin by Dr. Rehwa Lee (UCLA, Los Angeles, CA). Antibodies against the various G␥ subunits were purchased from Santa Cruz Biotechnology, Inc. Secondary antibodies were obtained from Jackson Immunologicals.
Gel Electrophoresis and Immunoblotting-SDS-PAGE and Western blot analysis were performed as described previously (15,19). Visualization of protein bands was performed using ECL reagents obtained from Pierce, Inc.
Isolation of Native G␤ 5 Complexes-The G␤ 5 -containing complex was purified from the soluble fraction of bovine retina as described previously (19). For purification of brain complexes, rat brains were removed from adult female rats and homogenized in 10 ml of buffer containing 20 mM Tris-HCl, 1 mM EDTA, 50 mM NaCl, 2 mM ␤-mercaptoethanol, pH 7.5 (TEBS). Total brain homogenate was centrifuged at 50,000 ϫ g for 20 min. at 4°C. The supernatant was then directly applied to a 5-ml Q-Sepharose column and further resolved on SP-Sepharose. For analysis of the membrane proteins, the pellet was washed twice in ice-cold TEBS buffer. The pellet was then resuspended in TEBS buffer also containing 1% sodium cholate or 1% Genapol C-100. The suspensions were then mixed at 4°C for 1 h before centrifugation at 50,000 ϫ g for 30 min. The supernatant was diluted 1:2 using TEBS buffer and loaded onto a 5-ml Q-Sepharose column. Bound proteins were eluted with a linear gradient of NaCl from 50 mM to 500 mM in a total volume of 25 ml. Fractions containing G␤ 5 and RGS7, as detected by Western blot, were pooled and diluted 4-fold in TEBS buffer without NaCl and applied to a 2-ml SP-Sepharose column and eluted with a 10-ml gradient from 50 to 500 mM NaCl in TEBS. Fractions (500 l) were collected and analyzed by Western blot.
Expression of G␤ 5 and RGS7 in COS-7 Cells-COS-7 monkey kidney cells were cultured in DMEM with 10% fetal bovine serum under 5% CO 2 at 37°C. 2 ϫ 10 5 cells/well were plated into 12-well plates 1 day before transfection. 1 g of total plasmid DNA, typically containing 0.5 g of G␤ 5 , RGS7, RGS7⌬ (15), or LacZ, was mixed with 5 l of Lipo-fectAMINE (Life Technologies, Inc.) in 200 l of OPTI-MEM and added to the washed cells. For RGS7, 0.9 g of plasmid DNA needed to be used to reliably detect the protein using Western blot. Five hours later, the transfection mixture was removed and replaced with 1 ml of 10% fetal bovine serum in DMEM. Twenty-four and 48 h after transfection, cells were harvested and assayed by Western blot. All cDNAs were carried by the pcDNA3 vector (Invitrogen).
Total RNA was isolated with TRIzol reagent (Life Technologies, Inc.) from 10 6 of COS-7 cells 48 h after transfection with pcDNA3 plasmids encoding LacZ, RGS7, and G␤ 5 . The RNase protection assay was performed with 2-10 g of total RNA and 80,000 cpm of RGS7 probe using an RPA III kit (Ambion). After hybridization overnight at 45°C and RNase digestion, separation, and detection of the protected RNA probe was performed by trichloroacetic acid as described by Pham et al. (25). The protected RNA probe was precipitated with 1 ml of 0.75% sodium pyrophosphate in 5% trichloroacetic acid and 0.025% bovine serum albumin. The precipitated RNA probes were collected onto glass microfiber filters (GFC grade, Whatman) by vacuum filtration. The filters were rinsed with 5% trichloroacetic acid, dried, and radioactivity measured in 4 ml of scintillation mixture in a scintillation counter. To verify the integrity of the RNA probe after RNase digestion, an aliquot of the samples were subjected to 5% polyacrylamide electrophoresis and subsequently visualized using autoradiography (BioMax MR, Eastman Kodak Co.).
Northern Blot Analysis-Total RNA (5-15 g) prepared from transfected COS-7 cells was denatured with glyoxal/dimethyl sulfoxide, resolved on a 1% agarose gel, blotted onto BrightStar-Plus membrane (Ambion), and subsequently hybridized to RGS7-specific [␣-32 P]UTPlabeled riboprobe prepared as described above. Pre-hybridization was carried out for 30 min at 68°C in ULTRAhyb buffer (Ambion) in hybridization oven. The blot was hybridized with 5 ϫ 10 6 cpm probe/ml pre-hybridization buffer for 3 h at 68°C, washed at room temperature for 20 min (twice for 10 min), followed by another 30 min (twice for 15 min) wash at 68°C. The blot was exposed to autoradiography film (BioMax MR, Kodak).
Pulse-Chase Labeling and Immunoprecipitation-COS-7 cells were transfected with RGS7, G␤ 5 , RGS7 ϩ G␤ 5 , or LacZ cDNA constructs as described earlier. Twenty-four hours after transfection, the cells (3 ϫ 10 5 ) were incubated at 37°C in 60 ϫ 15-mm dishes in methionine-and cysteine-free DMEM (Life Technologies, Inc.) for 1 h. A pulse of 200 Ci/dish of [ 35 S]methionine and [ 35 S]cysteine (NEN Life Science Products) was given for 1 h in methionine-and cysteine-free DMEM containing 10% dialyzed fetal bovine serum. After washing the cells twice using phosphate-buffered saline (PBS), the cells were chased with serum-free DMEM at 37°C. The cells were harvested at the indicated times in 500 l of PBS containing 10 mM EDTA and 10 mM phenylmethylsulfonyl fluoride. Following freeze-thawing, the cells were centrifuged for 30 min at 14,000 ϫ g. The supernatants were then precleared with Protein A-Sepharose beads, and 200 l of the supernatant was immunoprecipitated using affinity-purified anti-RGS7 antibody (1 g of IgG/20 l of Protein A) or anti-G␤ 5 antibody for the G␤ 5 monomer. After incubating with mixing for 1 h, the beads were washed twice with PBS and then eluted using 2ϫ SDS-PAGE loading buffer. The [ 35 S]methionine/cysteine-labeled proteins were resolved by 12% SDS-PAGE, transferred to nitrocellulose, visualized by autoradiography (BioMax MR, Kodak), and quantified using Scion Image.
Localization of RGS and G␤ 5 in Retinal Fractions-Bovine retinas were processed according to the procedure for the isolation of transducin as described in detail previously (26). Briefly, the retinas were resuspended, in an isotonic buffer (10 mM Tris-HCl, 100 mM KCl, 2 mM MgCl 2 , 1 mM DTT, pH 7.5) containing 45% sucrose. The suspension was passed through cheesecloth and centrifuged. Unsolubilized material was pelleted (P1) at 5,000 ϫ g for 10 min. The supernatant (S1) was diluted 1:1 in isotonic buffer. Crude rod outer segments (ROS) were collected (P2) by centrifugation at 15,000 ϫ g for 30 min. The crude outer segment (OS) pellet was further subjected to ultracentrifugation on a stepwise sucrose density gradient and the purified OS were recovered at the interface of the 1.115 and 1.135 density gradient steps.
Immunocytochemistry of Mouse Retinas-The enucleated eye from an euthanized mouse was placed in 4% paraformaldehyde in PBS for 5 min. The lens and cornea were removed, and the remaining eyecup was fixed for 1 h. After fixation, the eyecup was rinsed three times, 15 min each, in cold PBS, and infiltrated with 30% sucrose overnight. The next day, the eyecup was embedded in OCT and quickly frozen in liquid nitrogen, and 10-m cryosections were obtained. The retinal sections were blocked for 30 min in PBS containing 1% bovine serum albumin, 1% goat serum, and 0.3% Triton X-100. Affinity-purified RGS7 polyclonal antibody was diluted 1:100 in the block solution, applied to the sections, and incubated at room temperature for 1 h, after which the sections were washed three times, 5 min each, in PBS containing 1% bovine serum albumin and 1% goat serum. The fluorescein isothiocyanate-conjugated secondary antibody (Vector laboratories) was diluted 1:50 and incubated for 30 min at room temperature. After rinsing three times for 5 min each time in PBS, the sections were mounted in Vectashield and viewed under a fluorescent microscope.
Immunoprecipitation-Protein A-Sepharose beads (15 l) were washed with TEBS buffer and then incubated on ice with 1 g of affinity-purified RGS7 or RGS9 antibody for 30 min. After collecting the unbound material, the antibody-containing beads were mixed with 75 l of either brain or outer segment extract for 1 h on ice. After washing the beads twice with a large volume of TEBS buffer, a third wash using 75 l was collected for analysis on Western blot. The proteins were eluted using 75 l of 2ϫ SDS-PAGE loading buffer and subjected to Western blot analysis. For immunoprecipitation of photoreceptor outer segments, the membranes were first solubilized with 60 mM n-octyl-␤-D-glucopyranoside (Sigma) as described in He et al. (27) in either the presence or absence of AMF (25 M AlCl 3 , 10 mM MgCl 2 , 10 mM NaF).
In initial experiments, a control in which pre-immune serum was bound to the Protein A-Sepharose was tested and revealed no binding of RGS and G␤ 5 proteins. In subsequent immunoprecipitations, the negative control was extract-mixed with Protein A-Sepharose beads without antibody.
Effect of G␤ 5 -RGS7 Complex on Signaling through the Muscarinic M3 Receptor-CHO cells in six-well plates were either not transfected or transfected with the indicated combinations of cDNAs encoding the human muscarinic M3 receptor (3 g/well), RGS7 (5 g/well), and G␤ 5 (5 g/well). Twenty-four hours following transfection, cells were harvested in trypsin/EDTA and re-plated in 96 well plates. Approximately 40 h later, cells were loaded for 1 h at room temperature in a balanced salt solution with 5 M fluo-3-AM in the presence of 0.2% Pluronic F-127. Cells were washed and subsequently excited at 505 nm with emission recorded at 530 nm as an index of [Ca 2ϩ ] i using fluorometric imaging plate reader (FLIPR) according to standard protocol (28). Cells were stimulated with metacholine 20 s after the monitoring of fluorescence began. Data are presented as arbitrary units of fluorescence.

Purification of G␤ 5 -RGS Complexes from Retina and Brain-
Although complexes of G␤ 5 with RGS proteins have been reconstituted in vitro and found in native tissues, G␤ 5 -G␥ complexes have only been studied in vitro (21-24, 29, 30). In an attempt to detect native G␤ 5 -G␥, the G␤ 5 -containing fraction from the soluble fraction of bovine retina was purified to homogeneity as described previously in Cabrera et al. (19). This purified sample was analyzed using silver stain and Western blot. As shown previously in Cabrera et al. (19), G␤ 5 and RGS7 co-migrated throughout the purification. Fig. 1 shows that this G␤ 5 -RGS complex does not contain a G␥ subunit. As a control for the presence of a G␥ subunit, the SDS-PAGE lane with an equivalent amount of purified transducin heterotrimer shows a band characteristic of G␥ 1 at approximately 10 kDa. These results show that in the retinal cytosol G␤ 5 is bound to RGS and is not associated with G␥.
We reasoned that G␤ 5 could exist in a complex with G␥ in certain non-retinal cell types and thus attempted to detect G␤ 5 -G␥ complexes in brain extracts. Because G␤ 5 is found in both soluble and membrane-associated forms (23, 31), we examined G␤ 5 in cytosolic and detergent-extracted membrane fractions of rat brain. Fig. 2A shows that in brain cytosol, similarly to the retina, G␤ 5 co-migrates with RGS7 upon consecutive anion-and cation-exchange chromatographies. We previously demonstrated that cation-exchange chromatography on SP-Sepharose can be used as an assay for the association of G␤ 5 with an RGS (15). Since the majority of proteins are negatively charged at physiologic pH, at least 90% of proteins in crude extracts do not bind to this matrix. RGS7 and other GGL-containing RGS proteins contain positively charged amino acid clusters, which perhaps interact with the cationexchanger and allow the G␤ 5 -RGS complexes to be absorbed. Thus, co-elution of G␤ 5 with RGS7 as a single peak through ion-exchange chromatographic steps indicates that they are associated. In some experiments (one of which is represented in Fig. 2B), a small portion of G␤ 5 did not bind SP-Sepharose. However, even if G␤ 5 could be detected in the unbound fraction, it represented less than 1% of the total G␤ 5 . Furthermore, with prolonged development of the Western blots, RGS7 could also be seen in the same sample. In one experiment, we re-loaded the unbound fraction on fresh SP-Sepharose and both G␤ 5 and RGS7 then absorbed to the column. This indicates that the small portion of G␤ 5 in the unbound fraction is likely due to incomplete absorption on SP-Sepharose due to some variations of experimental conditions rather than G␤ 5 existing without an RGS.
To confirm the interaction between G␤ 5 and RGS7 using an alternative approach, we co-immunoprecipitated G␤ 5 with RGS7 using an affinity-purified antibody to RGS7 (Fig. 2C). Despite the fact that both the immunoprecipitating and detect- FIG. 2. G␤ 5 -RGS7 interaction in rat brain cytosol. A, the cytosolic fraction of rat brain was separated on Q-Sepharose and eluted using a 50 -500 mM NaCl gradient. Fractions were resolved on a 12% SDSpolyacrylamide gel and then analyzed by Western blot using polyclonal antibodies to G␤ 5 and RGS7. Both G␤ 5 and RGS7 co-eluted around 200 mM NaCl. T, total; U, unbound. B, the G␤ 5 -containing pool was further resolved on the cation-exchanger SP-Sepharose. Both G␤ 5 and RGS7 co-migrate indicating that the complex remained intact. C, immunoprecipitation of both G␤ 5 and RGS7 using affinity-purified anti-RGS7 antibody. The control shows the eluate from two independent experiments and illustrates the pattern of background IgG when stained with secondary antibody only. U, unbound; W, wash; E, eluate.
ing antibodies were raised in rabbits, we were able to clearly distinguish the bands for both RGS7 and G␤ 5 over the background IgG bands in the Western blots. In all our experiments, a small portion (Ͻ10%) of G␤ 5 remained in the unbound fraction. This most likely results from G␤ 5 being associated with other RGS proteins present in the brain (i.e. RGS6, -9, and -11), that would not bind to the anti-RGS7 antibody. Supporting this interpretation, in photoreceptor outer segments where RGS9 is the only RGS species present, 100% of G␤ 5 L is immunoprecipitated with RGS9-specific antibody (Fig. 8A). Thus, two different approaches show that in the cytosolic fraction of the brain, G␤5 and RGS7 are always present as a complex.
We next examined the possibility that a G␤ 5 -G␥ complex could be present in the brain membrane. It has been shown that some ionic detergents dissociate the recombinant G␤ 5 -G␥2 complex, but in the non-ionic detergent Genapol C-100, the dimer remains intact (21). Therefore, we extracted the membranes using either sodium cholate or Genapol C-100 to minimize the effect of solubilization on the status of the native G␤ 5 complex. Both detergents solubilized G␤ 5 , RGS7, and G␣ subunits from the membranes. The protein extracts were analyzed by chromatography and immunoprecipitation using the protocols developed for the cytosol. Following 10-fold enrichment on Q-Sepharose, the G␤ 5 -RGS7-containing fractions were pooled and further resolved on SP-Sepharose (Fig. 3A). Chromatography on SP-Sepharose resulted in at least an additional 100-fold purification, and both G␤ 5 and RGS7 co-eluted during the procedure. Similarly to the results in Fig. 2 (A and B), regardless of which detergent was used, only a small portion (Ͻ1% of total) of G␤ 5 could be detected in the unbound fraction in some experiments (Fig. 3A). The chromatographic behavior of the membrane-bound G␤ 5 -RGS7 complex suggests that it is essentially identical to the cytosolic form, as well as the reconstituted dimer (15). Importantly, since neither G␤ 1 nor G␥ 2 bound to SP-Sepharose, G␤ 5 is completely resolved from "classic" G␤␥ subunits (Fig. 3B). Finally, G␤ 5 was immunoprecipitated from the detergent extracts using the anti-RGS7 antibody (Fig. 3C). Recent work by Zhang and Simonds has also shown that G␤ 5 and RGS7 can be co-immunoprecipitated (20). Based on these results, we have to conclude that in brain membranes, as well as in the cytosol, G␤ 5 is always present in a dimer with either RGS7 or one of the other GGL-containing RGS proteins.
Coupled Expression of G␤ 5 and RGS7 in Transfected Cells-Supported by previous data on the co-localization of these proteins (31)(32)(33)(34), co-purification of G␤ 5 and RGS proteins indicated that G␤ 5 and RGS7 are always dimerized in native tissues. This led us to the idea that the stoichiometry between G␤ 5 and RGS7 must be regulated. To examine this, we tested to see if expression of G␤ 5 would affect RGS7 levels and vice versa.
FIG. 3. Analysis of G␤ 5 and RGS7 in brain membranes. A, rat brain membranes were extracted with either sodium cholate or Genapol C-100 as described under "Experimental Procedures" and loaded onto a Q-Sepharose column. The G␤ 5 -containing fractions were pooled and loaded onto SP-Sepharose and eluted using a 50 -500 mM NaCl gradient. The fractions were then analyzed by Western blot using G␤ 5 and RGS7 antibodies. B, the total detergent extracts (T), SP-Sepharose unbound (U), and G␤ 5 and RGS7 containing pooled fractions (P) were stained with G␥ 2/3 and G␤ 1 antibodies. C, immunoprecipitation of G␤ 5 using anti-RGS7 antibody from detergent extracts. For the control, the sample was mixed with beads that did not contain antibody. U, unbound; W, wash; E, eluate. Data shown are representative of at least six independent experiments.

FIG. 4. Expression of G␤ 5 and RGS7 in COS-7 cells.
A, COS-7 cells were transfected with constructs for either G␤ 5 , RGS7, or both. Additional experiments were also carried with the RGS7⌬ mutant. Transfection with LacZ cDNA was used as a control and to maintain the total DNA amount at 1 g/well. Total cell lysates obtained 48 h after transfection were subjected to Western blot analysis using RGS7 (top panel) and G␤ 5 (bottom panel) antibodies. Data shown are representative of three independent experiments. B, anti-RGS7 immunoprecipitation of cell lysates in which G␤ 5 and RGS7 were co-transfected and visualized by G␤ 5 antibody on a Western blot. C, quantification of Western blot data from three representative independent transfection experiments showing an approximate 6-fold increase in G␤ 5 and 10-fold increase of RGS7 levels upon co-transfection, while RGS7⌬ does not increase G␤ 5 levels. Western blots were scanned, and the images were analyzed using Scion Image. RGS7, RGS7⌬, and G␤ 5 are normalized based to the level of the protein expression in cells transfected with the individual cDNAs. Fig. 4 shows that when COS-7 cells were co-transfected with both G␤ 5 and RGS7 expression cassettes, the levels of both proteins were increased compared with transfection with the individual constructs. Importantly, the two proteins formed a complex as detected by G␤ 5 immunoprecipitation using an anti-RGS7 antibody (Fig. 4B). Co-transfection of an RGS7 mutant lacking the G␤ 5 -binding GGL domain (RGS7⌬) with G␤ 5 did not result in such an increase. This demonstrates the necessity of the protein-protein interaction in stabilizing the G␤ 5 -RGS dimer. Quantification of three independent experiments showed that levels of G␤ 5 and RGS7 increased 6-and 10-fold, respectively, when co-transfected (Fig. 4C). In many experiments, particularly when less than 0.9 g of RGS7 plasmid was used for transfection, RGS7 could not be detected without G␤ 5 , thereby making quantification of the -fold stimulation impossible. However, in the presence of G␤ 5 , RGS7 could always be detected even when lesser amounts of RGS7 cDNA were used. This result indicates that cells have a mechanism for regulating G␤ 5 -RGS7 stoichiometry and further supports the idea that G␤ 5 and RGS7 are always together in a complex.
To examine the mechanism by which the G␤ 5 -RGS7 dimer was regulated in transfected cells, we first performed an RNase protection assay using a labeled RGS7 riboprobe (Fig. 5, A and  B). Despite the fact there is more RGS7 protein in COS-7 cells transfected with G␤ 5 , the level of RGS7 mRNA does not change. Similar experiments have demonstrated that endogenous G␤ 5 mRNA also remains unchanged upon infection of rat pituitary cells with an RGS7 adenovirus, while the amount of G␤ 5 protein is increased 10-fold. 2 Additionally, Northern blot analysis shows identical levels of RGS7 RNA in cells transfected with RGS7 alone or RGS7 ϩ G␤ 5 (Fig. 5C). This suggests that protein degradation and/or synthesis is the major regulator of RGS7 levels in cells. Subsequently, protein degradation in COS-7 cells was studied using pulse-chase labeling and immunoprecipitation (Fig. 6). In cells transfected with RGS7 or G␤ 5 alone, the labeled protein was completely degraded within 18 h, with a half-life of about 1.5 h. Conversely, in the presence of G␤ 5 , the RGS7 protein was stabilized considerably (and vice versa), with a half-life of over 24 h. Along with the RNA data, these results clearly demonstrate that the increase in RGS7 protein levels in the presence of G␤ 5 is due to the slowed proteolytic degradation of dimerized RGS7 compared with the monomer.
Native G␤ 5 -RGS Complexes Do Not Bind to G␣ with High Affinity-Since at least one of the functions of G␤ 5 -RGS complexes is regulation of G protein GAP activity, we sought to determine if G␣ subunits could be detected in complex with the G␤ 5 -RGS dimer in native tissue extracts. In photoreceptor OS, transducin is the only G protein and the role of the G␤ 5 L-RGS9 complex in its regulation has been established. T. Wensel's laboratory (27,35) showed that RGS9, which is located in OS, is likely to be the only RGS in the system. However, the polyclonal antibody used in these prior studies was derived against the entire RGS domain of RGS9 and has a minor cross-reactivity with RGS7 (19,27). Therefore, using RGS7-and RGS9specific anti-peptide antibodies, we first confirmed that RGS7 is not present in OS. Purified OS were obtained from bovine retinas, and the localization of the RGS proteins was studied by Western blot throughout the fractionation procedure (Fig. 7A). Since we have previously shown that retinal G␤ 5 ("short") is 100% soluble, while G␤ 5 L is strictly OS membrane-associated FIG. 5. RGS7 mRNA levels do not change in the presence of G␤ 5 . A, the total RNA was isolated from COS-7 cells transfected with LacZ (squares), RGS7 alone (circles), and RGS7 ϩ G␤ 5 (triangles). mRNA representing RGS7 was measured using ribonuclease protection assay, as described under "Experimental Procedures," at the three indicated amounts of total RNA. Data are representative of two independent experiments. B, autoradiograph of 5% polyacrylamide gel showing the integrity of the protected riboprobe. C, Northern blot using RGS7 specific [␣-32 P]UTP riboprobe. 5 g of total RNA was loaded in each lane.

FIG. 6. Stability of RGS7 and G␤ 5 monomers compared with the G␤ 5 -RGS7 dimer in the transfected COS-7 cells.
A, COS-7 cells were metabolically labeled with a pulse of [ 35 S]methionine/cysteine and RGS7 was immunoprecipitated at the indicated chase times as described under "Experimental Procedures." The identity of the indicated bands as RGS7 and G␤ 5 were confirmed by Western blot. The data shown are representative of two independent experiments. B, autoradiographs from two independent pulse-chase experiments were quantified using Scion Image and graphed as mean Ϯ S.E. G␤ 5 alone, triangles; RGS7 alone, circles; G␤ 5 ϩ RGS7, squares. (19,30), these molecules were used as internal markers for the fractions. Fig. 7A shows that RGS7 is found in the soluble fraction (S2) along with G␤ 5 . Immunostaining with anti-RGS7 antibody confirmed that it is not found in OS, but rather is localized mostly to rod bipolar cells (Fig. 7B). RGS9 is located along with G␤ 5 L in the membrane-associated form in the OS (P2), in accord with previous data utilizing immunostaining (35). Next, we solubilized the OS with n-octyl-␤-D-glucopyranoside under conditions preserving the GAP activity of the native RGS9 complex (27) and immunoprecipitated the extract with the anti-RGS9 antibody. Probing the obtained fractions with antibodies to G␤ 5 and G␣ t revealed that 100% of G␤ 5 L was absorbed on the beads while G␣ t was not co-immunoprecipitated (Fig. 8A). Incubation of the samples with 100 M AlCl 3 , 10 mM MgCl 2 , and 10 mM NaF (AMF), which is known to activate G proteins and promote the RGS-G␣ interaction (9, 36 -40), did not lead to binding between G␣ t and the G␤ 5 -RGS complex. Since it has been shown that PDE␥ is necessary for the GAP activity of RGS9 (18,27), we also added an excess of recombinant PDE␥ to the assay; however, the results were identical with and without PDE␥. Rhodopsin, phosducin, arrestin, and endogenous PDE␥ also remained in the unbound fraction while the G␤ 5 L-RGS9 complex was quantitatively immunoprecipitated. 3 A similar series of immunoprecipitation experiments were carried out with brain membrane extracts. The fractions obtained from immunoprecipitation were stained with antibodies to G␣ q/11 and G␣ o , as G␣ q -mediated signaling have been shown to be affected by RGS7 (41) and G␣ o has been shown to be a target of G␤ 5 -RGS7 GAP activity in vitro (17). Fig. 8 (B and C) shows that neither G␣ subunit was present in the eluate of the immunoprecipitations. These experiments were also performed in the presence of AMF, and again no binding of G␣ to RGS7 was detected. 3 Previous data using GAP assays have demonstrated that G␤ 5 -RGS dimers interact with G␣ (14,17,18). The apparent conflict between GAP assays and our findings here and other data using pull-down assays (15,17) can be explained by the higher sensitivity of the GAP assay. This indicates that, compared with that of other RGS proteins, the interaction between the G␤ 5 -RGS complexes and G␣ is weak or transient (i.e. rapid off-rate).
Expression of G␤ 5 -RGS7 Complex in Transfected Cells Inhibits G␣ q -mediated Signaling-To investigate the potential functional role of G␤ 5 -RGS7, we studied the effect of these proteins on G␣ q -mediated signaling in transiently transfected mammalian cells. CHO cells were also transfected with the muscarinic M3 receptor known to be coupled to G␣ q (42)(43)(44), and the agonist-induced change in [Ca 2ϩ ] i was measured in the presence and absence of G␤ 5 and RGS7 using a FLIPR. Fig. 9 (A  and B) shows that both RGS7 alone and RGS7 co-expressed with G␤ 5 can inhibit G␣ q -mediated Ca 2ϩ release in response to the muscarinic agonist metacholine. We observed that, in the presence of G␤ 5 , this inhibition is stronger. However, as shown in Fig. 9C, RGS7 levels are increased dramatically in the presence of G␤ 5 . Therefore, the stimulation of RGS7 activity by G␤ 5 could be due to either increased RGS7 levels or the increased potency of the G␤ 5 -RGS dimer. Despite the difficulty with the quantitative aspects of these results, it is clear that the G␤ 5 -RGS7 complex, as well as monomeric RGS7, can inhibit G␣ qmediated signaling through the M3 receptor.

DISCUSSION
In Native Tissues G␤ 5 and RGS7 Are Only Present as a Complex-Previous studies have shown that G␤ 5 -RGS complexes are found in tissue extracts (18,19), but it has not been 3 D. S. Witherow and V. S. Slepak, unpublished observations. FIG. 7. Localization of RGS7 and RGS9 in the retina. A, bovine retinas were fractionated by ultracentrifugation as described under "Experimental Procedures" and the fractions (see text for details) were probed with antibodies to RGS7, RGS9, and G␤ 5 . B, sections of mouse retinas were prepared as described under "Experimental Procedures" and stained with affinity-purified RGS7 antibody. Bipolar cells with somata in the upper aspect of the inner nuclear layer are immunoreactive for RGS7. The axonal arbors of these bipolar cells extend to the outer portions of the inner plexiform layer and ends with large varicosities (60). Some amacrine cells also are immunopositive. RPE, retinal pigmented epithelium; ONL, outer nuclear layer; INL, inner nuclear layer; OPL, outer plexiform layer; GCL, ganglion cell layer.

FIG. 8. G␣ subunits do not co-immunoprecipitate with G␤ 5 -RGS.
A, photoreceptor outer segments were solubilized in the presence and absence of AMF using 60 mM n-octyl-␤-D-glucopyranoside. The extract was then immunoprecipitated using anti-RGS9 antibody, and the fractions were examined by immunoblots using G␤ 5 L and G␣ t antibodies. B and C, Western blots illustrating the immunoprecipitation of brain membrane extracts stained with either G␣ o (B) or G␣ q (C) antibodies. For the control, a 1:1 mixture of Genapol and cholate extracts was used. In A-C, the control represents the extracts mixed with beads not containing any antibody as described under "Experimental Procedures." U, unbound; W, wash; E, eluate. determined whether G␤ 5 can also be bound to a G␥ subunit. We show here that in the soluble retinal extract, the entire pool of G␤ 5 is present as a G␤ 5 -RGS dimer lacking a G␥ subunit (Fig.  1). In the brain, G␤ 5 and RGS7 can be found not only in cytosolic but also in membrane-associated forms. Cytosolic G␤ 5 and RGS7 formed a complex, as demonstrated by both chromatography and immunoprecipitation (Fig. 2). Since neither G␤ 5 nor RGS7 have putative sites responsible for membrane attachment, the composition of such membrane complexes might be different from those in the cytosol. G␤ 5 could be bound to the membrane by G␥, which is prenylated and known to be responsible for anchoring G␤␥ complexes to the membrane (45)(46)(47). RGS7 could be bound to a protein other than G␤ 5 , for example an acylated G␣ subunit or the ion channel polycystin, which has been shown to interact with RGS7 in the yeast two-hybrid system and co-transfected cells (48). To investigate the mem-brane-bound G␤ 5 and RGS7, we solubilized membranes with mild detergents and examined the behavior of G␤ 5 and RGS7 by chromatography and immunoprecipitation similarly to the cytosolic forms.
In our assays we made an emphasis to quantitatively evaluate the yield and distribution of the complex in all fractions, including in the unbound material. Cation-exchange chromatography completely separates G␤ 5 from G␤ 1 and G␥ subunits and, importantly, shows that essentially 100% of the G␤ 5 pool is bound to RGS proteins (Fig. 3). This conclusion is corroborated by our experiments utilizing immunoprecipitation with RGS7 and RGS9 antibodies, as well as the very recent work of Zhang and Simonds (20), who in a reciprocal experiment immunoprecipitated RGS6 and RGS7, but not G␥, from brain extract using a G␤ 5 antibody.
As in all experiments with membrane proteins, it is possible that detergent could alter the composition of native complexes, i.e. G␥ could dissociate from G␤ 5 allowing RGS to bind. Indeed, Jones and Garrison (21) have recently demonstrated that the ionic detergents CHAPS and cholate can dissociate the recombinant G␤ 5 -G␥ 2 dimer. In contrast, the non-ionic detergent Genapol C-100 did not disrupt the interaction (21). In light of this, we studied membrane extracts using different detergents including cholate, which has been used to purify G proteins (49), Genapol C-100, shown to preserve G␤ 5 -G␥ (21), and n-octyl-␤-D-glucopyranoside, which preserves the GAP activity of the native RGS9 complex (27). The results presented in Fig. 3 show that, regardless of the detergent used, G␤ 5 and RGS7 from brain membranes exist as a complex that is similar in stability to that of the "classic" G␤␥ dimers. The question why some of the heterodimer is cytosolic and some is membranebound remains open and requires additional investigation. The studies of retinal and brain G␤ 5 -RGS presented here are in agreement with our previous data with in vitro translated proteins that G␤ 5 preferentially binds to RGS7 even in the presence of excess G␥ 2 (15). Although it is possible in principle that in a specific small population of cells, G␤ 5 and RGS proteins may exist apart from each other, current analysis of native extracts indicates that in situ G␤ 5 and RGS are present as a tightly associated complex regardless of their subcellular localization.
Control of G␤ 5 :RGS Balance in Cells-The apparent absence of monomeric forms of G␤ 5 and RGS7 and G␤ 5 -G␥ complexes in situ, as well as the co-localization of G␤ 5 and RGS7 in brain (31)(32)(33)(34), suggested that their association is tightly controlled. Indeed, as we show here, in cells transiently transfected with RGS7 and G␤ 5 cDNAs, the level of expression of one protein is significantly increased in the presence of the other (Fig. 4). The increased protein level is based on the enhanced stabilization of the complex and requires direct protein-protein interaction between G␤ 5 and RGS7 (Figs. 4 -6). In another recent study, we have shown that infection of neuroendocrine cells in primary culture with an adenovirus expressing RGS7 increases endogenous G␤ 5 levels through a post-transcriptional mechanism requiring direct interaction between RGS7 and G␤ 5 . 2 Our finding of post-transcriptional regulation of G␤ 5 and RGS7 is supported by research by Benzing et al. (50), who demonstrated degradation of RGS7 through the ubiquitin-proteosome pathway. Co-regulation of G␤ 5 and RGS protein levels has also independently been shown by others and thus appears to be a universal phenomenon. The knockout of the RGS9 gene in mouse leads to disappearance of G␤ 5 L from photoreceptors (51). Snow et al. (16) observed greater levels of RGS6 and G␤ 5 in cell lysates when they were co-expressed. In contrast, Kovoor et al. (52) recently reported that the amount of RGS7 expressed in Xenopus oocytes does not change upon co-expression of G␤ 5 . This might be due to innate technical difficulties with protein detection in oocytes. The researchers metabolically labeled the cells with [ 35 S]methionine and immunoprecipitated the extract with an RGS7 antibody. G␤ 5 did not co-immunoprecipitate with RGS7 because the complex was denatured due to the presence of 4% SDS during the protein extraction. Furthermore, even though the SDS was diluted to 0.4% for immunoprecipitation, it is possible that some of the IgG was denatured, making quantitative analysis of RGS7 inaccurate. Even if RGS7 levels do not increase when co-expressed with G␤ 5 in the oocytes, it is clear that G␤ 5 -RGS balance is maintained in mammalian cells. Our results show this mechanism to be analogous to regulation of G␤␥ dimers where the G␤:G␥ stoichiometry is also controlled through proteolytic degradation of unassociated subunits (53)(54)(55)(56).
Function of the G␤ 5 -RGS7 Complex-The hallmark function of RGS proteins is to act as GAPs for G␣ subunits, and therefore studies have been performed to analyze the GAP activity of recombinant G␤ 5 -RGS complexes. In contrast to the isolated RGS domains of large RGS proteins or relatively small monomeric RGSs, such as RGS4, the functionally active, full-length GGL-containing RGS proteins cannot be easily obtained in the large quantities required for GAP assays. Despite this difficulty, initial experiments presented by Snow et al. (14) showed that the complex of G␤ 5 with RGS11 lacking the DEP domain can be produced in a baculovirus system, and clearly demonstrated that this truncated mutant had GAP activity toward G␣ o . In later studies of recombinant complexes of G␤ 5 with full-length RGS6 and RGS7, Posner et al. (17) confirmed the GAP activity of G␤ 5 -RGS dimers. This GAP activity is remarkably specific to G␣ o , but not G␣ i or G␣ q (14,17). Interestingly, in cell-based assays, RGS7 can inhibit G␣ i -mediated (52,56) as well as G␣ q -mediated signaling (41). Here, in accord with these previous studies, we show that G␤ 5 -RGS7 can inhibit G␣ qmediated Ca 2ϩ mobilization caused by stimulation of the M3 muscarinic acetylcholine receptor in CHO cells (Fig. 9). The controversy between the GAP analysis and functional cellular assays can be reconciled by two considerations. First, it is possible that G␤ 5 -RGS7 acts via a non-GAP mechanism, similarly to that described for RGS4 and GAIP (10,11). Second, the interaction of G␤ 5 -RGS7 with the G protein and the GAP activity may require the presence of other components of the pathway, such as effector or receptor. It is known that the photoreceptor RGS protein, RGS9, acts only in the presence of the G protein effector PDE␥ (18,27). Furthermore, partial purification of native G␤ 5 L-RGS9 complex leads to the loss of GAP activity even in the presence of PDE␥, suggesting the necessity of an additional component (18). Importantly, it appears that RGS7 can only inhibit G␣ q -mediated signal transduction from certain receptors, as a different G␣ q -coupled receptor (gonadotropin-releasing hormone receptor) cannot be inhibited by RGS7. 2 The receptor specificity of RGS action has been previously demonstrated for RGS1, -4, and -16 (58,59).
In an attempt to identify molecules that interact with G␤ 5 -RGS in situ, we carried out a series of experiments utilizing chromatography and immunoprecipitation. Because G␤ 5 -RGS7 is a GAP for G␣ o and RGS7 inhibits G␣ q -mediated signaling, these G␣ subunits are obvious candidates for such a direct protein-protein interaction. We found that neither G␣ o nor G␣ q co-immunoprecipitates or co-migrates with the G␤ 5 -RGS7 complex during purification (Fig. 8). No binding was detected in the presence or absence of aluminum fluoride (AMF), a promoter of G protein-RGS binding. It is unlikely that we tested the samples for the "wrong" G␣ subunits, as these assays also did not reveal the interaction between G␤ 5 L-RGS9 and transducin (Fig. 8C) for which the interaction has been established (18,27,51). We have also could not pull-down the native or in vitro translated G␤ 5 L-RGS9 complex using immobilized G␣ t . 4 The results with native G␤ 5 -RGS complexes are in accord with previous data using recombinant G␤ 5 -RGS dimers (15,17) showing that G␤ 5 -RGS7 interacts with G␣ with low affinity. This affinity, however, must be sufficient to confer the functional effects upon G␣ i and G␣ q in cells and on G␣ o in GAP assays.
Although it is clear that the G␤ 5 -RGS complexes can act as GAPs for G␣ subunits, the role of G␤ 5 in this interaction is unknown. In pull-down binding assays, full-length RGS7, as well as its RGS domain alone, binds to G␣ (15,57), while G␤ 5 -RGS7 complexes do not (15,17). This implicates the role of G␤ 5 as an inhibitor of G␣-RGS binding. Similarly, the RGS domain of RGS9 exerts GAP activity toward G␣ t alone, while the native G␤ 5 L-RGS9 complex does not. The GAP activity of the native complex requires the presence of the effector enzyme PDE␥. Also consistent with the idea that G␤ 5 constrains RGS in a less active form is the fact that the GAP activity of the RGS7 domain (41), and also full length RGS4, is much stronger (Ͼ10-fold) than that of the G␤ 5 -RGS7 complexes (17). Another possible role of G␤ 5 is to increase the specificity of the complex toward G␣ subunits. For example, the RGS7 domain alone binds to G␣ o , G␣ i3 , and G␣ z (57), while the G␤ 5 -RGS7 complex has only been shown to be a GAP for G␣ o (17). Thus, data obtained in vitro indirectly suggests that G␤ 5 may either attenuate RGS function or confine G␣ specificity. In contrast, Kovoor et al. (52) have recently shown that G␤ 5 augments RGS7-mediated inhibition of G␣ i signaling in oocytes. Our current results also allude that G␤ 5 increases RGS7 activity. However, since G␤ 5 enhances the amount of RGS7, the correct interpretation of the data must require further experimentation where the expression levels of G␤ 5 , RGS, and other molecules are controlled.
The data presented here show that G␤ 5 and RGS7 always exist as a complex in vivo, and as with G␤␥ subunits, cells coordinate their stoichiometry. G␤ 5 -RGS7 dimers do not bind to G␣ subunits with a high affinity, but can inhibit G␣ q -mediated signaling through the M3 receptor in a cellular assay. Previous experiments, even very recent studies (34,50,61,62), investigated the roles of recombinant G␤ 5 (i.e. G␤ 5 ␥ 2 complexes) and GGL-containing RGS proteins outside the context of each other. However, the current data dictate that studying G␤ 5 and these RGS proteins requires viewing them not as separate entities, but as heterodimeric proteins that potentially have several functions.