Selective Interaction of AGS3 with G-proteins and the Influence of AGS3 on the Activation State of G-proteins*

AGS3 (activator of G-protein signaling 3) was isolated in a yeast-based functional screen for receptor-indepen-dent activators of heterotrimeric G-proteins. As an initial approach to define the role of AGS3 in mammalian signal processing, we defined the AGS3 subdomains involved in G-protein interaction, its selectivity for G-pro-teins, and its influence on the activation state of G-protein. Immunoblot analysis with AGS3 antisera indicated expression in rat brain, the neuronal-like cell lines PC12 and NG108-15, as well as the smooth muscle cell line DDT 1 -MF2. Immunofluorescence studies and confocal imaging indicated that AGS3 was predomi-nantly cytoplasmic and enriched in microdomains of the cell. AGS3 coimmunoprecipitated with G a i3 from cell and tissue lysates, indicating that a subpopulation of AGS3 and G a i exist as a complex in the cell. The coim- munoprecipitation of AGS3 and G a i was dependent upon the conformation of G a i3 (GDP >> GTP g S (guanosine 5 * -3- O -(thio)triphosphate)). The regions of AGS3 that bound G a i were localized to four amino acid repeats (G-protein regulatory motif (GPR)) in the carboxyl terminus (Pro 463 –Ser 650 ), each of which were capable of binding G a i . AGS3-GPR domains selectively is representative of five to seven images obtained by different fixation methods and using different an- tibody concentrations. Only very weak immunofluorescence signals were detected in nontransfected DDT 1 -MF2 cells as expected from the relative strengths of the signals for control and AGS3-transfected cells observed by immunoblotting in A . No immunofluorescent signal was detected in control or AGS3 transfectants in the absence of any primary antibody.

cell microdomains, receptor phosphorylation and internalization, cross-talk between signaling pathways, and proteins that regulate the basal activation state of G-proteins independently of the receptor.
We partially purified a direct G-protein activator from NG108-15 cells (1,2) and subsequently used a functional screen to identify three proteins (AGS1-3, for activator of Gprotein signaling 1-3) that activated heterotrimeric G-protein signaling in the absence of a cell surface receptor (3)(4)(5). The identification of such proteins raises many interesting and unexpected questions relative to signal processing by heterotrimeric G-proteins. As an initial approach to address these issues, we focused on the biochemical and functional characterization of AGS 1 proteins, and this report deals specifically with AGS3 (AF107723, calculated molecular weight 72,049). AGS3, isolated from a rat brain cDNA library, contains seven tetratricopeptide repeats (TPRs) and four GPR (G-protein-regulatory) motifs separated by a linker in the middle of the protein (4) 2 (Fig. 1).
AGS3 is one member of a larger protein family defined by a two-domain structure (Fig. 1). In rodents and humans, this family is defined by rat AGS3 and human LGN (U54999), which was isolated in a yeast two-hybrid screen using G␣ i2 as bait (8). A single AGS3-related protein is found in Caenorhabditis elegans (AAA81387) and Drosophila melanogaster (AF36967). Analysis of genome and expressed sequence tag data bases indicated that in addition to human LGN cDNAs, there are partial human cDNAs exhibiting higher homology to AGS3 versus LGN (e.g. AL117478 (360 amino acids), 95% sequence similarity to AGS3 and 57% sequence similarity to LGN; AI272212 (190 amino acids), 93% sequence similarity to AGS3 and 57% sequence similarity to LGN). Likewise, analysis of mouse/rat genome and expressed sequence tag data bases indicate that in addition to AGS3 cDNAs (e.g. L23316) there are mouse cDNAs (e.g. AA543923, AA166402, and AW539573) exhibiting higher sequence homology to human LGN versus AGS3. Thus, AGS3 and LGN are distinct proteins, and perhaps there are additional related proteins in the primate genome yet to be identified.
The first insight as to the functional role of LGN and AGS3 in signal processing was their identification as a G␣ i -binding protein (8) and isolation as a receptor-independent G-protein activator (4), respectively. Additional insight was provided by recent studies in D. melanogaster, where the AGS3/LGN homolog PINS (Partner of Inscuteable) is required for events involved in the asymmetric cell division of neuroblasts in the early stages of development (9,10). PINS is part of a multiprotein complex that is translocated from the cytosol to one pole of the dividing neuroblast. In this article, we report the existence of an AGS3-G␣ i complex within the cell, define the G␣-interacting domains of AGS3, and determine the selectivity of AGS3 for different G␣ subunits. AGS3, which preferentially binds to G␣ GDP , can bind multiple G␣ subunits and hence may function as a scaffolding protein to provide spatially and temporally discrete signaling events. AGS3 and G␤␥ actually competed with each other for interaction with G␣ t(GDP) , and AGS3 inhibited guanine nucleotide exchange on G␣ i1 . The properties of the AGS3-G␣ interactions add unexpected dimensions to signal processing by G-protein-regulated signaling systems.

EXPERIMENTAL PROCEDURES
Materials-[ 35 S]GTP␥S (1250 Ci/mmol) was purchased from PerkinElmer Life Sciences. Tissue culture supplies were obtained from JRH Bioscience (Lenexa, KS). Acrylamide, bisacrylamide, Bio-Rad protein assay kits, and sodium dodecyl sulfate were purchased from Bio-Rad. Ecoscint A was purchased from National Diagnostics (Manville, NJ). Guanosine diphosphate, guanosine triphosphate, and Thesit (polyoxyethylene-9-lauryl ether) were obtained from Rche Molecular Biochemicals. Polyvinylidene difluoride membranes were obtained from Pall Gelman Sciences (Ann Arbor, MI). Gammabind G-Sepharose was obtained from Amersham Pharmacia Biotech, and nitrocellulose BA85 filters were purchased from Schleicher & Schuell. Poly-L-lysine. normal goat serum, biotinylated goat anti-rabbit IgG, and Extravidin fluorescein isothiocyanate were purchased from Sigma. Immuno Fluore mounting medium was purchased from ICN Biomedicals. Purified bovine brain G-protein and antisera to the COOH-terminal 10 amino acids of G␤1, which recognizes G␤1-4, were kindly provided by Dr. John Hildebrandt (Department of Pharmacology, Medical University of South Carolina, Charleston, SC) (11,12). G␣ i1-3 and G␣ o were purified from Sf9 insect cells infected with recombinant virus as described (13) and kindly provided by Dr. Stephen Graber (West Virginia University School of Medicine, Morgantown, WV). G␣ s and G␣ q , similarly expressed in Sf9 insect cells, were kindly provided by Dr. Elliott Ross (University of Texas Southwestern Medical Center, Dallas, TX) (14). Purified G␣ t and G␣␤␥ t (15) were kindly provided by Dr. Heidi Hamm (Northwestern University Medical School, Chicago, IL). Polyclonal G␣ i3 antisera generated against the COOH-terminal 10 amino acids was kindly provided by Dr. Thomas W. Gettys (Department of Medicine, Medical University of South Carolina, Charleston, SC) (16). Purified GA antibody, which selectively recognizes G i /G␣ o , was kindly provided FIG. 1. Schematic representation of AGS3 and related proteins. Full-length rat AGS3 (AAF08683) was aligned with the human LGN protein (AAB40385), the D. melanogaster PINS protein (AAF36967), and the C. elegans protein (CE) (AAA81387) by PILEUP (University of Wisconsin GCG program) and visual adjustment. Amino acid sequence similarity and identity are indicated below the four sequences by ϩ or residue, respectively. The shaded and lined sequences represent TPR I-VII and a repeated segment of amino acids (GPR I-IV). The amino-terminal half of the AGS3 contains six TPRs, as defined by SMART analysis (7), that exist as a cluster of two (Ser 43 -Gln 116 , Gly 183 -Ile 336 ) and four motifs with a spacer of ϳ60 amino acids between the two clusters. Visual inspection of the spacer region between the two clusters of TPR motifs indicates the likely existence of an additional TPR motif defined by the I 129 GN and A 162 SEFYERNL sequence, in which the helical structure of this TPR may be extended. The carboxyl-terminal half of AGS3 contains the second functional domain consisting of four ϳ20 amino acid repeats. as discussed in the text. PINS contains three GPRs with highest homology to GPRs I, III, and IV of AGS3, LGN, and C. elegans protein.
Generation of AGS3 Subdomains-AGS3 subdomains were generated as glutathione fusion proteins by polymerase chain reaction using the full-length cDNA of AGS3 as a template. Primers were designed to add BamHI and EcoRI sites to the 5Ј and 3Ј ends, respectively, of AGS3 subdomains to fuse the AGS3 open reading frame with the reading frame of glutathione S-transferase contained in the pGEX4T1 vector. The polymerase chain reactions were generally performed using 250 nM primers and 125 pM template DNA in a total volume of 50 l. Cycles were 1 ϫ 3 min at 94°C; 30 ϫ 1.5 min at 94°C, 1 min at 60°C, and 2 min at 72°C; and 1 ϫ 10 min at 72°C. Primers used to generate specific constructs were as follows. Preparation of Cell/Tissue Lysates-NG108-15, PC12, and DDT 1 -MF2 cells were grown as described previously (1,2). Caco-2 cells were obtained from the American Type Culture Collection and cultured in Eagle's minimum essential medium supplemented with 1% minimum essential medium nonessential amino acids. CHO cells were grown on Falcon tissue culture dishes at 37°C (5% CO 2 ) in Ham's F-12 medium supplemented with 10% fetal bovine serum plus penicillin (100 units/ ml), streptomycin (100 g/ml), and fungizone (0.25 g/ml) (19). Rat brain was homogenized in 3 ml of buffer/g of tissue of lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5 mM EDTA, 1% Nonidet P-40). Confluent 100-mm dishes of cells were washed with cell washing solution (137 mM NaCl, 2.6 mM KCl, 1.8 mM KH 2 PO 4 , 10 mM Na 2 HPO 4 ) and resuspended in 1 ml of lysis buffer/dish by homogenization. Following a 1-h incubation on ice at 4°C, the cell homogenate was centrifuged at 27,000 ϫ g for 30 min. Supernatants were collected and spun at 100,000 ϫ g for 1 h to generate a detergent-soluble fraction. The supernatant was immediately processed for immunoblotting or immunoprecipitation. In some experiments, cells and tissue were also fractionated to generate a crude membrane pellet and a 100,000 ϫ g supernatant containing cytosol. Tissues were homogenized in 5 mM Tris, 5 mM EDTA, 5 mM EGTA, pH 7.4, and centrifuged at 100,000 ϫ g for 30 min at 4°C and washed at least three times by homogenization in membrane buffer. For preparation of cell homogenates, 12 confluent 100-mm dishes were lysed in 3 ml of 5 mM Tris, 5 mM EDTA, 5 mM EGTA, pH 7.4, and centrifuged at 100,000 ϫ g for 30 min at 4°C. Cell membrane pellets were washed three times with intervening homogenization and pelleting at 100,000 ϫ g. The washed membrane pellets were resuspended in 250 l of membrane buffer (50 mM Tris, 0.6 mM EDTA, 5 mM MgCl 2 , pH 7.4) by homogenization. Protein concentrations were determined by a Bio-Rad protein assay.
Protein Interaction Assays-The interaction of AGS3 with G-proteins was assessed by both coimmunoprecipitation and protein interaction experiments using tissue/cell lysates or purified G-proteins. Protein concentrations in the lysates were determined by a Bio-Rad protein assay. For immunoprecipitation from mammalian cells, cell/tissue lysates (1-3 mg of protein in 0.5-1 ml) were pre-cleared by rotating incubation with Gammabind G-Sepharose (12.5 l of packed resin equilibrated in lysis buffer) for 30 min at 4°C. Following centrifugation, G␣ i3 antisera (1:250 dilution) was added to pre-cleared lysates and incubation continued overnight at 4°C. Protein complexes were captured by adding Gammabind G-Sepharose (12.5 l packed volume) and continuing the incubation for 30 min at 4°C. The mixture was then microcentrifuged at 4°C and the pellets washed (3ϫ 500 l of incubation buffer) and resuspended in 2ϫ Laemmli buffer. Resuspended samples were placed in a boiling water bath for 5 min and microcentrifuged for 10 min prior to loading on denaturing 10% polyacrylamide gels. Proteins were transferred to polyvinylidene difluoride membranes for immunoblotting.
For analysis of the interaction of AGS3-GPR with multiple G-protein subunits, G␣ i2 (200 nM) was incubated with G␣ i3 (50 nM) in the presence or absence of the AGS3-GPR GST fusion protein (250 nM) in 250 l of buffer A (20 mM Tris, pH 7.5, 70 mM NaCl, 1 mM dithiothreitol, 0.6 mM EDTA, 0.01% Thesit) for 1 h at 4°C. G␣ i3 antisera (1: 500) was added, and the incubation was continued for 3 h at 4°C. Protein complexes were isolated and evaluated by immunoblotting as described above.
Protein interaction assays using purified G-protein subunits were conducted as described previously (4,18). All purified G-proteins used in these studies were isolated in the GDP-bound form. Unless indicated otherwise, all G-protein interaction assays contained 10 M GDP. The AGS3-GST fusion proteins were expressed in and purified from bacteria using a glutathione affinity matrix. The AGS3-GST fusion proteins were eluted from the matrix with glutathione and desalted by centrifugation (Centricon YM-3; Millipore, Bedford, MA). For interaction assays with cell/tissue lysates, the AGS3-GST fusion protein (100 -300 nM) was incubated with purified G-protein (50 -100 nM) or cell/tissue lysate (ϳ4 mg of protein/ml) for 1 h at 24°C in a total volume of 250 l. 12.5 l of packed glutathione-Sepharose slurry was added and the mixture rotated at 4°C for 20 min, after which the affinity matrix was pelleted and washed three times with 500 l of incubation buffer. Proteins retained on the matrix were solubilized in 2ϫ Laemmli loading buffer and separated by electrophoresis on denaturing 10% polyacrylamide gels. Proteins were transferred to polyvinylidene difluoride membranes for immunoblotting. Each blot was checked by Amido Black staining to verify equal loading of fusion proteins.
Immunofluorescence-DDT 1 -MF2 control cells and DDT 1 -MF2 cells stably transfected with AGS3 were plated onto coverslips (18-mm round no. 1) precoated with 0.01% polylysine and allowed to grow to 60% confluence. Coverslips were then rinsed with 3 ϫ 2 ml of cell washing solution (CWS) (137 mM NaCl, 2.6 mM KCl, 1.8 mM KH 2 PO 4 , 10 mM Na 2 HPO 4 ) and fixed in 4% paraformaldehyde for 10 min, followed by two 5-min incubations in CWS containing 0.1 M glycine (3 ml/coverslip). Coverslips were then incubated in 0.01% Triton X-100 for 10 min, followed by three 5-min incubations (3 ml/coverslip) with CWS. Fixed cells were then incubated in 10% goat serum for 1 h and washed once with CWS. AGS3 antibody was diluted into CWS containing 2% goat serum and 1% fetal bovine serum and then centrifuged at 10,000 ϫ g for 10 min prior to use. Coverslips were incubated with 75 l of AGS3 antibody (0.01 mg/ml) for 1 h by placing the coverslips (cell side down) on parafilm in a humidified chamber. Following incubation with AGS3 antibody, coverslips were washed three times (3 ml/coverslip) with CWS and then incubated with goat anti-rabbit biotin conjugate (1:800) for 40 min. The fixed cells were washed three times in CWS and incubated in Extravidin fluorescein isothiocyanate (1:500) for 40 min, followed by three 10 min incubations with CWS. Washed coverslips were mounted in Immuno Fluore, sealed with nail polish, and stored at 4°C until evaluated by fluorescent microscopy. All incubations were carried out at 24°C. Mounted slips were evaluated on a Leica DMLB fluorescent microscope and by confocal microscopy using a Bio-Rad MRC-100 laser scanning confocal imaging system. The cell nucleus was identified by propidium iodide staining. Multiple series of experiments were performed to determine the optimal conditions for signal detection and to verify the specificity of observed signals. These experiments included different methods of fixation and permeabilization as well as a matrix with serial dilutions of primary antibodies and secondary conjugates. We chose to generate stable transfectants to minimize any artifacts introduced by transient transfection. Only very weak immunofluores-cence signals were detected in nontransfected DDT 1 -MF2 cells as expected from the relative strengths of the signals for control and AGS3transfected cells observed by immunoblotting. No immunofluorescent signal was detected in control or AGS3 transfectants in the absence of any primary antibody.
Additional Methods-DDT 1 -MF2 cells were stably transfected with pcDNA3.AGS3 by DNA/calcium phosphate coprecipitation (21). For antipeptide antisera, AGS3 peptides (P-32 Thr 306 -Ile 436 and P-22 Asp 528 -Gly 550 ) were synthesized and conjugated for generation of rabbit polyclonal antisera using the Peptide Synthesis and Antibody Production Facility at the Medical University of South Carolina. Each of the three antisera specifically recognized GST-AGS3 at reasonable dilutions of serum and were affinity-purified. Denaturing gel electrophoresis and immunoblotting were performed as described previously (18). For reprobing of membrane transfers, the membrane transfers were washed with buffer A containing 20 mM Tris-HCl, pH 7.6, 140 mM NaCl, 0.2% Tween each and then incubated with pre-heated stripping buffer (62.5 mM Tris-HCl, pH 6.8, 2% SDS, 100 mM ␤-mercaptoethanol) for 20 min in a 55°C water bath with gentle shaking. The membrane was then washed with buffer A and processed for immunoblotting. For Coomassie Blue staining of proteins, gels were incubated in 100 ml of staining buffer (0.25% Coomassie Blue in 45% methanol, 45% H 2 O, 10% glacial acetic acid) for 30 min at room temperature. Stained gels were then washed in 100 ml of destain solution (45% methanol, 45% H 2 O, 10% glacial acetic acid) and incubated for 30 min. Gels were then washed in fresh destain solution every 30 min until protein bands were visible.

RESULTS
Expression Profile of AGS3 and Coimmunoprecipitation of AGS3 and G-protein Subunits-Immunoblots with AGS3 antipeptide antibodies indicated expression of AGS3 (M r ϳ 74,000) in rat brain, the neuroblastoma-glioma cell hybrid NG108-15 (rat/murine) the rat pheochromocytoma cell line PC12, and the DDT 1 -MF2 smooth muscle cell line derived from hamster vas deferens (Fig. 2). Immunoreactive species with an apparent M r of ϳ74,000 were not detected in rat liver, rat kidney, Caco-2 cells, CHO cells, HEK cells, or NIH-3T3 fibroblasts (Fig. 2). 3 The same immunoreactive M r ϳ74,000 species was observed with two different antibodies (P-32, P-22) generated against peptides derived from different regions of the protein (Fig. 2). Fractionation of tissues/cells expressing AGS3 indicated that AGS3 is enriched in the 100,000 ϫ g supernatant consistent with a major distribution of AGS3 in the cytosol (Fig. 3A). A similar fractionation of AGS3 was observed in DDT 1 -MF2 cells stably transfected with AGS3 (Fig. 3A). The subcellular localization of AGS3 was also addressed by immunofluorescence analysis following stable expression of AGS3 in the DDT 1 -MF2 cell line (Fig. 3B). Confocal microscopy was used to generate an image approximately through the middle plane of the cell. The immunofluorescent image indicates that AGS3 is predominantly cytosolic (Fig. 3B), as suggested by immunoblot analysis of the 100,000 ϫ g supernatant from cell lysates illustrated in Fig. 3A. Within the cell, the AGS3 signal is often punctate and occasionally enriched in microdomains of the cell.
We previously reported that the carboxyl-terminal 74 amino acids of AGS3 were active in the yeast functional screen and that this peptide fragment directly bound to G␣ i (4). We thus asked if full-length AGS3 was complexed with G␣ i3 in lysates of rat brain or DDT 1 -MF2 cells stably transfected with AGS3. As AGS3 preferentially regulated G␣ i2 and G␣ i3 in the yeast functional assay (4), we first approached this issue by immunoprecipitation of G␣ i3 . Approximately 30% of brain lysate G␣ i3 was immunoprecipated with a G␣ i3 carboxyl terminus antibody. Immunoblots of membrane transfers containing G␣ i3 immunoprecipitates indicated that AGS3 coimmunoprecipitated with G␣ i in a nucleotide-dependent manner (Fig. 4). The absence of G␤ in the GTP␥S-treated samples provided internal controls for G-protein activation and subunit dissociation by added GTP␥S/Mg 2ϩ . Immunoprecipitation experiments were also conducted with the AGS3 antisera P-32. Although AGS3 was effectively immunoprecipitated by the P-32 antisera in each cell/tissue extract, coimmunoprecipitation of G␣ i3 was variable, which may reflect lower immunoprecipitation efficiency for P-32 and/or a masking of the P-32 epitope in the AGS3-G␣ complex (data not shown). 3 Nevertheless, these data indicated that a subpopulation of G␣ i3 and AGS3 exists as a complex in the cell and that this interaction is regulated by nucleotide binding to G␣.
AGS3 Domains That Interact with G-proteins-The interaction between AGS3 and G-proteins was further explored in in vitro binding assays to define the regions of AGS3 actually involved in binding to G␣. We generated the amino-terminal half of AGS3 (AGS3-TPR, Met 1 -Ile 462 ) and the COOH-terminal  half of AGS3 (AGS3-GPR, Pro 463 -Ser 650 ) as GST fusion proteins (Fig. 5). The AGS3-TPR, AGS3-GPR, and the 74 amino acid carboxyl terminus (AGS3-CT, Met 577 -Ser 650 ) isolated in the original yeast functional screen were incubated with DDT 1 -MF2 cell lysates and proteins bound to the AGS3 subdomains identified by immunoblotting of gel transfers. The G␣ i1/2 binding domains of AGS3 were found in the COOH-terminal half of the protein (Fig. 5B, left panel). The TPR domains of AGS3 did not interact with G␣ i1/2 or G␤␥ (Fig. 5B, left panel).
Within the COOH-terminal region of AGS3 that binds to G␣, there are four repeats (ϳ20 residues each) termed GPR for G-protein regulatory motifs (4). Previous data indicated that the 74-amino acid domain at the carboxyl terminus of AGS3 was functional in the yeast functional screen and the interaction of this peptide with G␣ i was disrupted by targeted mutations in GPR IV highlighting the importance of the GPR motifs. We then asked if each GPR domain was indeed capable of binding G␣. Each GPR motif was generated as a GST fusion protein (Fig. 5) and evaluated in protein interaction assays using DDT 1 -MF2 lysates (Fig. 5B, right panel). Each GPR motif bound G␣ i1/2 , although GPR I, at least in this context, bound less G␣ than did GPR II-IV (Fig. 5B, right panel). These data suggest that interaction of AGS3 with G␣ i3 observed by coimmunoprecipitation experiments (Fig. 4) reflects interaction of G␣ i3 with the GPR domains in AGS3.
The preceding data also suggested that AGS3 is capable of binding multiple G␣ i subunits. To address this issue, we asked if a GST-AGS3 fusion protein containing GPRs I-IV indeed bound more than one G␣ i at the same time. A GST-AGS3 fusion protein containing GPRs I-IV was incubated with a mixture of G␣ i3 and G␣ i2 . Samples were then immunoprecipitated with antisera directed against the carboxyl terminus of G␣ i3 . In the presence of AGS3, G␣ i2 was also found in the G␣ i3 immunoprecipitate (Fig. 6). G␣ i2 was not found in the G␣ i3 immunoprecipitate in the absence of AGS3 (Fig. 6). These data clearly indicate that AGS3 is capable of binding more than one G␣ i subunit consistent with a putative role of AGS3 as a scaffolding protein within a larger signal transduction complex.
Selectivity of AGS3 for G-proteins-The preceding data clearly established the interaction of AGS3 with G␣ within the cell and defined the regions of AGS3 involved in G-protein binding. We then asked if the interaction of AGS3 with G␣ was selective for different G-protein families. We approached this question using crude tissue/cell lysates and purified G␣ subunits. The AGS3-GPR GST fusion protein was incubated with rat brain lysate and bound proteins identified by immunoblotting with G␣ specific antisera. AGS3-GPR effectively bound G␣ i1-3 , but not G␣ s , G␣ o , G␣ q , or G␤␥ (Fig. 7A). Based upon the comparison of the signal intensity in the input versus sample lane, it is estimated that AGS3-GPR binds ϳ20-% of the total G␣ i protein in the lysate sample. Similar results were obtained in DDT 1 -MF2 cell lysates. Each of the protein interaction experiments in the tissue/cell lysates were done in the presence of GDP which would stabilize heterotrimeric G␣␤␥; however, immunoblotting with G-protein ␤ subunit antisera indicated that AGS3 was complexed with G␣ i in the absence of G␤ (Fig. 7A). Thus, either AGS3 effectively promoted subunit dissociation or there is a population of G␣ i that exists free of G␤␥. The selectivity of AGS3 for different G-proteins was also observed using purified G␣ subunits. AGS3 bound to G␣ i1-3 and purified G␣ t , but it did not interact with G␣ s and weakly bound G␣ q and G␣ o (Fig. 7B). A similar profile of AGS3 selectivity for G␣ subunits was also observed in a yeast functional assay (4). Comparison of the relative intensities of the bound G␣ versus input G␣ for G␣ o /G␣ q and G␣ i (Fig. 7B) indicated a higher apparent affinity of AGS3 for G i versus G o /G q , which may account for the inability of AGS3 to interact with G␣␤␥ q and G␣␤␥ o in brain lysates (Fig. 7A).
AGS3 and G-protein Activation-As both AGS3 and G␤␥ interact with the GDP-bound conformation of G␣, the two proteins may actually compete with each other for interaction with G␣ and thus AGS3 would essentially promote subunit dissociation in the absence of nucleotide exchange. This issue was addressed by determining the influence of G␤␥ on the interaction of AGS3 with G␣ t . We first compared the ability of AGS3 to interact with purified G␣ t versus heterotrimeric G t (Fig. 8A). At equimolar concentrations of purified G␣ t and heterotrimeric G t , AGS3 bound equivalent amounts of G␣ t . As observed with the AGS3-G␣ i complex isolated from tissue/cell lysates, G␤␥ was not present in the AGS3-G␣ t complex isolated from purified heterotrimeric G t , indicating that AGS3 effectively dissociated G t from G␤␥. We thus asked if G␤␥ would interfere with formation of the AGS3-G␣ t complex. In these experiments, G␣ t was first incubated with equimolar or excess G␤␥ to generate heterotrimeric G t prior to exposure of the complex to AGS3. The interaction of AGS3 with G␣ was not altered by G␤␥ at concentrations equivalent to G␣, as observed in the experiments using heterotrimeric G t (Fig. 8B), but it was completely blocked by 10-fold higher concentrations of G␤␥ (Fig. 8B), indicating that AGS3 and G␤␥ are effectively competing with each other for binding to G␣ GDP . DDT-AGS3, DDT 1 -MF2 cells stably transfected with rat AGS3. B, immunofluorescent analysis of AGS3 distribution in DDT 1 -MF2 cells stably transfected with AGS3. Cells were fixed and processed as described under "Experimental Procedures." Confocal microscopy was used to evaluate images through different planes of the cells, and the micrograph shown is the image taken from approximately the middle plane of the cells. The large rounded area in the middle of the cell devoid of signal corresponds to the cell nucleus as defined by propidium iodide staining. This image was generated with a Bio-Rad MRC-100 laser scanning confocal imaging system (magnification, ϫ63; 30% laser power; gain, 1250; IRS, 0.7). This figure is representative of five to seven images obtained by different fixation methods and using different antibody concentrations. Only very weak immunofluorescence signals were detected in nontransfected DDT 1 -MF2 cells as expected from the relative strengths of the signals for control and AGS3-transfected cells observed by immunoblotting in A. No immunofluorescent signal was detected in control or AGS3 transfectants in the absence of any primary antibody.
The interaction of AGS3 with G␣ may actually stabilize the GDP-bound or nucleotide-free conformation of G␣ and "free up" G␤␥ for downstream signaling. Indeed, this conjecture would account for the biological activity of AGS3 in the yeast functional assay (4), where G␤␥ is responsible for subsequent activation of the pheromone response pathway. To address this issue, we asked if AGS3 influenced the guanine nucleotide binding properties of G␣ i . AGS3-GPR blocked the binding of GTP␥S to G␣ i1 (IC 50 ϳ 0.1 M) (Fig. 9A). We had previously identified key amino acid residues in GPR-IV that disrupted binding of AGS3-CT to G␣ i (4), and we then examined the effect of this series of AGS3-CT mutants on GTP␥S to G␣ i1 . The AGS3-CT peptides containing GPR mutations that resulted in a loss of binding to G␣ i in protein interaction assays (F609R, R624F) (4) were also ineffective at inhibiting GTP␥S binding to G␣ i1 (Fig. 9B). These data, and the results obtained in protein interaction experiments where AGS3 preferentially binds G␣ GDP versus G␣ GTP␥S , suggest that AGS3 actually stabilizes the G␣ GDP or nucleotide-free conformation and functions as an inhibitor of guanine nucleotide exchange on G␣. These biochemical data are consistent with the functional properties of AGS3 in S. cerevisiae in that the action of AGS3 did not require the generation of G␣ GTP and it was not antagonized by overexpression of the GTPase activating protein RGS4 (4). 4 DISCUSSION A large number of diverse signaling mechanisms within the cell utilize guanine nucleotide-binding proteins as a molecular switch to process biological signals. Due to the central place of these events in signal propagation, several mechanisms have evolved to turn this switch on and off. Such mechanisms include the regulation of guanine nucleotide exchange (e.g. guanine nucleotide exchange factors, guanine nucleotide dissociation inhibitors), hydrolysis (GTPase-activating proteins) and the subcellular targeting of G-proteins themselves. In general, signal processing by G-protein-regulated systems, as is the case for single membrane span receptors, likely operates within the context of a dynamic signal transduction complex. Such a multiprotein complex may provide coordinated and integrated functionality for heterotrimeric G-protein signaling systems, which process a myriad of external stimuli via G-protein-coupled receptors. Within such a complex there are likely accessory proteins distinct from receptor, G-protein and effector that influence various aspects of signal propagation. Such proteins may: 1) determine the specific pathway that the signal travels, 2) provide a cell-specific mechanism for signal amplification, 3) influence the population of activated G-protein/effector within the cell independent of receptor activation, 4) be "effectors" subject to receptor regulation providing attractive targets for cross-talk between diverse signaling systems, 5) provide alternative modes of input to G-protein-regulated signaling path-ways independent of a classical G-protein-coupled receptor, and/or 6) serve as scaffolding proteins to organize a signal transduction complex.
AGS3 is one of three mammalian cDNAs isolated in an expression cloning system in S. cerevisiae as receptor-independent activators of heterotrimeric G-protein signaling (3)(4)(5). Epistasis experiments in the yeast system indicated that the three cDNAs activated the pheromone response pathway at the level of G-protein, and the proteins were therefore termed activators of G-protein signaling (AGS1-3). Both cellular and/or in vitro studies indicated that these proteins exhibited selectivity for G-proteins and used different mechanisms to activate G-protein signaling. AGS1 is a novel Ras-related protein that directly increases GTP␥S binding to G␣ (3,5). AGS2 is identical to mouse Tctex1, a protein that exists as a light The two domain structure (TPR and GPR motifs) of AGS3 is highly conserved. TPR motifs serve a range of functions for diverse proteins (9,10,22,23). The TPR domains of Rapsyn, which contains an organization of TPR motifs most closely related to the AGS3 TPR domains, are involved in clustering of nicotinic receptors at the neuromuscular junction (23). Studies with the D. melanogaster AGS3/LGN homolog PINS suggest a role for the TPR domains in trafficking of AGS3 within the cell (9,10). In D. melanogaster PINS binds to Inscuteable, and this interaction is required for placement of key proteins (Inscuteable and Bazooka) involved in polarization and proper orientation of mitotic spindles of neuroblasts during asymmetric cell division of neuroblasts (9,10). G␣ i /G␣ o is apparently complexed with the PINS/Inscuteable complex, where it presumably plays a signaling function. In mammalian tissues, AGS3 is expressed at highest levels in brain where it is primarily found in a 100,000 ϫ g supernatant following homogenization. AGS3 may oscillate between cytosol and membrane compartments as observed for PINS (AGS3/LGN ortholog) in D. melanogaster.
AGS3 selectively binds to G␣ i in the presence of GDP, and protein interaction assays with AGS3-GPR indicate that the AGS3-GPR-G␣ complex is free of G␤ suggesting the following possibilities. First, AGS3 binds to G-protein heterotrimer (G␣␤␥) and actively promotes subunit dissociation, while maintaining G␣ in the GDP-bound state. Second, AGS3 "catches" a transient nucleotide-free conformation of G␣ i and this interaction is stabilized by binding of GDP to G␣ with AGS3 replacing G␤␥ as a G␣ binding partner. Third, during "basal" cycling of the G␣ through its various states of activation/inactivation, there is a period when G␣ is free of G␤␥, allowing AGS3 to bind G␣ and exclude rebinding of G␤␥. Each possibility could account for the activity of AGS3 in the yeast functional screen, where the pheromone response pathway is activated by G␤␥ (4), and each is consistent with the biochemical data indicating that G␤␥ and AGS3 compete with each other for interaction with G␣. Thus, the activity of AGS3 as a receptor-independent activator of G-protein signaling may actually involve dissociation of G␣ and G␤␥ in the absence of nucleotide exchange "releasing" G␤␥ from G␣ GDP to activate downstream effectors. In such a scenario, G␣ bound to AGS3 may be functionally inert and signal termination would require dissociation of AGS3 and G␣ GDP with rebinding of G␤␥ and G␣ GDP . Each of these scenarios are also of note relative to the role of AGS3 (PINS)-G␣ i / G␣ o complexes in neuroblast processing in D. melanogaster and the defect in orientation of the mitotic spindle observed in the absence of PINS (9,10). In the latter situation, interaction of AGS3 (PINS) with G␣ i(GDP) /G␣ o(GDP) could "release" G␤␥ for effector regulation, which may involve the apparent localization of G␤␥ to microtubules and/or mitotic spindles (24,25). The detailed analysis of the interaction of AGS3 with G␣ presented in this report should greatly facilitate efforts to further FIG. 8. Influence of G␤␥ on the interaction of AGS3 with G␣ t . A, AGS3-GPR (300 nM) was incubated with purified G␣ t or G␣␤␥ t (100 nM) and processed for protein interaction studies as described under "Experimental Procedures." The data presented are representative of five individual experiments with G␣ t and two experiments with G␣␤␥ t using different batches of fusion proteins. B, AGS3-GPR (100 nM) was added to tubes containing G␣ t (50 nM) that had been preincubated with G␤␥ (50 or 500 nM) and samples processed for protein interaction assays. Similar data were obtained in two experiments. The blot in A was first probed with the G␣ antisera and then stripped for reprobing with the G␤ antisera. The input lane in A and B, respectively, contain one-tenth and one-fifth of the lysate volume used in each interaction assay. define the role of AGS3 in signal processing in mammalian systems.
Another possible explanation for the detection of AGS3-G␣ complexes that do not contain G␤␥ is that there is a population of G␣ i in the cell that exists free of G␤␥. In such a case, AGS3 may regulate the activation state of G␣ in much the same way as does G␤␥. AGS3-G␣ complexes may be regulated by unexpected modes of signal input, which promote nucleotide exchange on the AGS3-G␣ GDP complex and release "activated" or "functional" AGS3 and G␣ i(GTP) . An interaction of AGS3 with G␣ in the absence of G␤␥ may also function to hold G␣ in the right place so that the signal input is more effectively processed and as such the two-domain structure of AGS3 may serve as a type of scaffold within a larger signal transduction complex. By virtue of its potential to bind up to three and possibly four G␣ subunits, AGS3 could "seed" oligomeric structures of G␣ i (26,27), which might mesh with arrays of other signaling molecules. AGS3 may actually be complexed with a mixture of G␣ i1-3 .
As noted earlier, the GPR motif is also found in other proteins that interact with or regulate G-protein ␣ subunits, and one would hypothesize that the GPR motif has evolved to serve as an anchor for proteins to bind to G␣ subunits. 5 The GPR domains are found in proteins that have apparently different effects on the activation state of G-protein. The GTPase-activating proteins RGS12 and RGS14 both contain a GPR motif, as does the Purkinje cell-specific protein Pcp2. The latter protein was reported to activate brain G-protein by accelerating guanine nucleotide exchange on heterotrimeric G-proteins (28). The functional and biochemical studies with the AGS3-GPR motifs indicate that this motif actually behaves as an inhibitor of guanine nucleotide exchange on G␣. As a relatively small discrete structure that binds to G-protein ␣ subunits, the GPR motif may serve as a template for rational design of peptides/ small molecules that directly influence the activation state of G-protein.