Signaling by a Non-dissociated Complex of G Protein βγ and α Subunits Stimulated by a Receptor-independent Activator of G Protein Signaling, AGS8*

Accumulating evidence suggests that heterotrimeric G protein activation may not require G protein subunit dissociation. Results presented here provide evidence for a subunit dissociation-independent mechanism for G protein activation by a receptor-independent activator of G protein signaling, AGS8. AGS8 is a member of the AGS group III family of AGS proteins thought to activate G protein signaling primarily through interactions with Gβγ subunits. Results are presented demonstrating that AGS8 binds to the effector and α subunit binding “hot spot” on Gβγ yet does not interfere with Gα subunit binding to Gβγ or phospholipase C β2 activation. AGS8 stimulates activation of phospholipase C β2 by heterotrimeric Gαβγ and forms a quaternary complex with Gαi1, Gβ1γ2, and phospholipase C β2. AGS8 rescued phospholipase C β binding and regulation by an inactive β subunit with a mutation in the hot spot (β1(W99A)γ2) that normally prevents binding and activation of phospholipase C β2. This demonstrates that, in the presence of AGS8, the hot spot is not used for Gβγ interactions with phospholipase C β2. Mutation of an alternate binding site for phospholipase C β2 in the amino-terminal coiled-coil region of Gβγ prevented AGS8-dependent phospholipase C binding and activation. These data implicate a mechanism for AGS8, and potentially other Gβγ binding proteins, for directing Gβγ signaling through alternative effector activation sites on Gβγ in the absence of subunit dissociation.

G protein-coupled receptor signaling systems play key roles in a number of biological processes (1). When bound to specific ligands, G protein-coupled receptors facilitate exchange of GDP for GTP on the G␣ subunit, leading to conformational changes in the G␣"switch" regions (2). These changes decrease the affinity of G␣ for G␤␥ and are thought to result in subunit dissociation such that G␣ and G␤␥ subunits are free to interact with downstream effectors (3). In this model, structural elements in G␣ and G␤␥ that bind downstream effectors are masked at the interface between the G␤␥ and G␣ subunits and revealed upon subunit dissociation.
Alternatively it has been argued that subunit dissociation is not a required step in the G protein activation cycle, instead, a distinct GTP-bound conformation of the ␣␤␥ complex may represent the activated form of the G protein (4). Considerable experimental evidence has accumulated in favor of this idea. Most recently, cell-based fluorescence resonance energy transfer and bioluminescence energy transfer analyses suggest that G protein heterotrimers may not dissociate upon activation but rather undergo specific conformational rearrangements (5)(6)(7). Other evidence includes: A covalently linked G␣-␤␥ fusion protein is signaling competent in yeast (8); the kinetic coupling model for G protein activation suggests that, in the presence of regulators of G protein signaling proteins, GTP hydrolysis is too rapid to allow subunit dissociation prior to effector activation (9) and the ␤2-adrenergic receptor can activate adenylyl cyclase without causing nucleotide exchange implying that subunit dissociation is not required for adenylyl cyclase activation by this receptor (10). A potential barrier to acceptance of the model is that it is not clear how conformational rearrangements of the subunits would allow for productive engagement of key signaling surfaces at the G␣-␤␥ interface with effectors. One proposal is the "clam shell" hypothesis where G␣-␤␥ contacts are maintained by contacts between the amino-terminal helix of the G␣ subunit (G␣-NT) 4 with the side of the G␤ subunit ␤-propeller, while other regions at the G␣-␤␥ interface are exposed for effector activation.
G protein-coupled receptor-independent mechanisms for G protein activation have been discovered in recent years. For example, activators of G protein signaling (AGS proteins), discovered in a yeast-based screen for receptor-independent activation of the pheromone pathway, are a group of proteins that stimulate G protein signaling (11,12). The best characterized of the AGS proteins act through interactions with G protein ␣ subunits to dissociate the G protein heterotrimer either through nucleotide exchange-dependent (Class I AGS proteins) or -independent mechanisms (Class II AGS proteins). Here we explore the mechanism for G protein activation by a recently described AGS protein, AGS8, which binds to G protein ␤␥ subunits (13). AGS8 forms a complex interacting with G␣ and G␤␥ simultaneously, occupies the G␤␥"hot spot," a critical effector binding and signal transfer region on G␤␥ (14 -17), yet does not dissociate G␣ from G␤␥ subunits. In complexes between G␣, ␤␥, and AGS8, a signal transfer region on G␤␥ that does not involve the hot spot is critical for signaling to phospholipase C (PLC) ␤. This introduces a concept for G protein ␤␥ subunit signaling where amino acids outside the G␣-␤␥ subunit interface drive G␤␥ signaling and provides a model for how non-dissociated G protein complexes can activate downstream targets and a potential modified alternative of the clam shell hypothesis.

EXPERIMENTAL PROCEDURES
Materials-Peptides (SIRK, SIRKALNILGYPDYD; SIGK, SIGKAFKILGYPDYD; and SIRK(L9A) SIRKALNIAGYPDYD) were synthesized by Alpha Diagnostics International, purified by high-performance liquid chromatography to Ͼ90% purity, and their identity was confirmed by mass spectrometry analysis. Phosphatidylinositol 4,5-bisphosphate and phosphatidylethanolamine were obtained from Avanti Polar Lipids, nickel-nitrilotriacetic acid-agarose was from Qiagen, GTP␥S was from Sigma, glutathione-Sepharose and anti-GST antibody were from GE Healthcare, TetraLink TM tetrameric avidin resin was from Promega (Madison, WI), anti-G␣ i antibody was from Oncogene.
Purification of ␤␥ and ␣ Subunits-G␤ 1 ␥ 2 or biotinylated G␤ 1 ␥ 2 (bG␤ 1 ␥ 2 ) subunits were purified from 2 liters of Sf-9 cells triply infected with His 6 -G␣ i1 , wild-type, or bG␤ 1 subunits, and G␥ 2 subunits essentially as described (18). All of the alanine-substituted G␤ 1 subunit mutants used in this study were modified with a biotin acceptor peptide at the amino terminus and were biotinylated. These mutants were expressed with G␥ 2 in 200 ml of SF9 cells and partially purified exactly as described in a previous study (16). Experiments comparing alanine-substituted bG␤␥ activities to wt bG␤␥ utilized wt bG␤␥ subunits partially purified in parallel with the mutants for direct comparison. Myristoylated G␣ i1 was purified from Escherichia coli as previously described (19) and bound GTP␥S (0.5 mol/ mol of protein) was determined by an Amido Black protein assay.
Expression and Purification of GST-tagged cAGS8-The coding sequence of cAGS8 (A1359-W1730 (cDNA#1-16)) was fused in-frame to GST in the pGEX-4T vector (GE Healthcare) and expressed and purified from E. coli BL21/DE3 cells as previously described (13) with some modification. Overnight cultures grown in Luria broth containing ampicillin (50 g/ml) were diluted to an A 600 equal to 0.5-0.6 and induced with 0.1 mM isopropyl 1-thio-␤-D-galactopyranoside for 3 h at 30°C. Fusion proteins from a 1-liter culture were purified following the previously described protocol except prior to elution the glutathione-Sepharose column was washed with 5 mM ATP and a denatured E. coli protein extract to remove contaminating GroEL. The resulting eluted protein was 80% pure and dialyzed against 50 mM HEPES, pH 7.5, 50 mM NaCl, 1 mM dithiothreitol, and protease inhibitors. Protein concentrations were determined by Amido Black protein assay, and purified proteins were snap frozen and stored at Ϫ80°C.
Protein Binding Assays-GST-cAGS8 or GST was incubated with either purified G␣ i1 , or bG␤ 1 ␥ 2 , or various bG␤ 1 mutants in 200 l of binding buffer (20 mM HEPES, pH 8.0, 1 mM EDTA, 1.2 mM MgCl 2 , 0.1% C 12 E 10 (polyoxyethylene-10-lauryl ether), 1 mM dithiothreitol, 150 mM NaCl, 10 M GDP). The binding mixture was rotated at 4°C overnight. 40 l of 50% glutathione-Sepharose was added to the binding mixture, and the mixture was further rotated for 1 h. The glutathione matrix was washed with three rounds of centrifugation/washing in binding buffer. Identical conditions were used for avidin-agarose isolations of bG␤ 1 ␥ 2 except avidin-agarose was used. Proteins bound in precipitates were identified by immunoblotting following PAGE. In some cases immunoblots were quantitated by chemiluminescence imaging with a charge-coupled device camera. This has a much larger dynamic range than film and overcomes issues with signal saturation observed with x-ray film.
Gel-filtration Chromatography-Various combinations of GST-cAGS8, G␣ i1 , G␤ 1 ␥ 2 , and PLC␤2 (100 nM each) were incubated in 500 l of gel-filtration buffer (20 mM HEPES, pH 8.0, 1 mM EDTA, 1.2 mM MgCl 2 , 0.1% C 12 E 10 , 1 mM dithiothreitol, 150 mM NaCl, 10 M GDP, and protease inhibitor mixture) at 4°C for 2 h. Then the mixture was applied to tandem Superdex 75/200 columns (GE Healthcare) pre-equilibrated with gelfiltration buffer and resolved at a flow rate of 0.4 ml/min at 4°C. 1-ml fractions were collected. An aliquot of each fraction (20 l) was analyzed by SDS-PAGE on a 12% polyacrylamide gel and visualized by silver staining. No binding of any of the proteins to GST was detected in the gel-filtration assay (data not shown).
GTP␥S Binding Assay-This assay was performed as described previously (20) except without membranes or receptors. Briefly, 120 nM G␣ i1 (6 pmol), in the presence or absence of 180 nM GST-cAGS8 and with or without 240 nM G␤ 1 ␥ 2 subunits, were incubated with 0.4 M GTP␥S ([ 35 S]GTP␥S, ϳ5000 cpm/pmol specific radioactivity in the assay) in the presence of 2 M GDP to suppress the basal rate of nucleotide exchange by G␣ i1 . Bound [ 35 S]GTP␥S detected by binding to nitrocellulose filters.
PLC Assay-PLC assays were performed as described previously (21) and in the figure legends.
Measurement of G␣-␤␥ Interactions by Flow Cytometry-Equilibrium binding of fluorescein isothiocyanate-labeled myristoylated G␣ i1 to bG␤ 1 ␥ 2 subunits was measured using flow cytometry as has been previously described (22)(23)(24). Nonspecific binding, determined by the simultaneous addition of 300 pM fluorescein isothiocyanate-labeled myristoylated G␣ i1 and 50 nM myristoylated G␣ i1 subunits to the bG␤␥ bound beads, was 10 -20% of the total signal and was subtracted from the mean channel numbers from each experiment unless otherwise indicated.
Trypsin Digestion of G␣ i1 -3 g of G␣ i1 was mixed with 0.15 g of L-1-tosylamido-2-phenylethyl chloromethyl ketonetreated trypsin (New England Biolabs Inc.) and incubated in 100 l of digestion buffer (20 mM Tris, pH 8.0, 0.6 mM EDTA, 5 mM MgCl 2 , 1 mM dithiothreitol, 70 mM NaCl, 10 M GDP, 30 M AlCl 3 , and 10 mM NaF) at 25°C for 1 h. The digestion reaction was terminated by adding 5 mM 1-chloro-3-tosylamido-7amino-2-heptanone. The digestion was analyzed by PAGE and visualization with Coomassie Blue staining. The protein concentration after digestion was determined with an Amido Black protein assay. To confirm that trypsin-treated G␣ i1 still bound to cAGS8, binding to GST-cAGS8 was analyzed in a GST pulldown assay as has been described in parallel with undigested G␣ i1 .

RESULTS
AGS8 Binds to a Site on G␤ 1 ␥ 2 That Overlaps the G␣ Binding Site Yet Does Not Affect Binding of G␣ to G␤␥-The carboxylterminal domain of AGS8 (amino acids A1359-W1730, cAGS8) binds to ␤␥ and activates G␤␥ signaling in yeast and in mammalian cells transfected with G protein subunits (13). To understand the molecular mechanism for how a protein can bind to G␤␥ and simultaneously propagate G␤␥ signaling, we investigated the molecular nature of the interaction between G␤␥ and cAGS8. To initially identify and characterize a binding site for cAGS8 on G␤␥ we used peptide competition and mutagenic analyses. The peptides SIGK and SIRK bind to a hot spot on G␤ 1 ␥ 2 that interacts with multiple effectors, including PLC␤2, and corresponds to the G␣ subunit switch II binding site (2,14,16,25). SIGK and SIRK compete for binding of both ␣ subunits and PLC␤2 to G␤ 1 ␥ 2 (15,23). These peptides also blocked binding of cAGS8 to G␤ 1 ␥ 2 indicating the binding sites for the peptides, PLC␤2, G␣ subunits and cAGS8 overlap on the surface of G␤ 1 ␥ 2 (Fig. 1A). A control peptide that does not bind G␤ 1 ␥ 2 , SIRK (L9A) (15), had little effect on cAGS8-G␤ 1 ␥ 2 interactions.
To identify specific amino acids within the G␤␥ hot spot site required for cAGS8 interactions, GST-cAGS8 binding to a series of alanine-substituted bG␤ 1 ␥ 2 subunit mutants within the hot spot was tested (Fig. 1, B and C; see Fig. 10). Compared with wild-type G␤ 1 ␥ 2 , G␤ 1 W99A, G␤ 1 D186A, G␤ 1 M188A, and G␤ 1 H311A (G␤ 1 H311 is outside the hot spot and is a control) mutations did not affect binding to cAGS8, whereas G␤ 1 K57A, G␤ 1 Y59A, G␤ 1 M101A, G␤ 1 L117A, G␤ 1 Y145A, and G␤ 1 N230A mutations all significantly inhibited binding of G␤ 1 ␥ 2 to cAGS8. These data complement the peptide competition data and clearly demonstrate that cAGS8 interacts with the hot spot on G␤ 1 ␥ 2 .
Many of these amino acids are required for activation of effectors, including PLC␤2, and are directly at the G␣-␤␥ binding interface (Fig. 10) (3,26). Thus binding of cAGS8 to this region would be expected to preclude G␣ subunit binding and effector activation. We used flow cytometry to analyze G␣-␤␥ interactions as previously described (22)(23)(24). As previously shown, the hot spot-binding peptide, SIGK, competes with G␣ i1 for binding to G␤ 1 ␥ 2 in a concentration-dependent man-ner ( Fig. 2A) consistent with its apparent K d for G␤␥ of ϳ1 M (22); however, cAGS8 did not inhibit G␣ binding to G␤␥, despite apparently binding to the same site as SIGK and G␣ subunit switch II. The highest concentration of cAGS8 tested, 1 M, was higher than concentrations shown to have maximal functional effects on G␤␥ signaling (see Figs. 5B and 9C for examples) and higher than the concentration required for strong binding in the GST pulldown assay and formation of stable complexes by gel filtration.
cAGS8 Binds to G␣ Subunits and G␤␥ Simultaneously-SIGK peptide binds to the switch II binding site and competes for ␣ subunit binding to G␤␥. How could cAGS8 occupy the same binding site on G␤␥ yet not compete for G␣ subunit binding to G␤␥? We hypothesized that, if cAGS8 bound both G␤␥ and G␣ simultaneously, then the complex would not be disrupted. It had previously been shown that G␣ i did not bind FIGURE 1. cAGS8 binds to a domain on the G␤ subunit that overlaps with G␣ binding. A, hot spot binding peptides block G␤ 1 ␥ 2 -cAGS8 interactions. 300 nM GST-cAGS8 was incubated with 30 nM purified bG␤ 1 ␥ 2 with or without 20 M peptides in binding buffer. Proteins bound to GST-cAGS8 were detected with immunoblotting as described under "Experimental Procedures." Samples were also immunoblotted with anti-GST antibodies to validate loading equal amounts of protein (not shown). Data shown are representative of three independent experiments. This Western blot was quantified by chemiluminescence imaging: percent inhibition values relative to control were 20% (L9A), 89% (SIGK), and 89% (SIRK). B, GST pulldown assay showing the effect of various bG␤ 1 hot spot mutants on bG␤ 1 ␥ 2 -cAGS8 interaction. 300 nM GST-cAGS8 protein was incubated with 30 nM wt bG␤ 1 ␥ 2 or with 30 nM various bG␤ 1 mutants (all partially purified and quantitated in parallel (16)) in binding buffer. The bottom panel is a Western blot of the suspension prior to pulldown assay, demonstrating that equal amounts of each G␤␥ mutant were included in each pulldown assay. Data shown are representative of three independent experiments. C, Western blots from two separate experiments as in B were quantified by chemiluminescence imaging, and the data were pooled and plotted relative to wt G␤ 1 ␥ 2 binding. Data are mean Ϯ S.E. strongly to cAGS8 (13), but perhaps a weak interaction was not detected that might be significant to formation of this complex. Indeed, G␣ i1 bound to GST-cAGS8 in the absence of G␤␥ in a manner independent of the G␣ i1 activation state, because inclusion of AlF 4 Ϫ did not affect the binding (Fig. 2B lanes 3 and   4). As a control to demonstrate that G␣ i is activated by AlF 4 Ϫ in this assay, the G␣ i subunit was tested for AlF 4 Ϫ activation-dependent protection from trypsin digestion. In the absence of AlF 4 Ϫ treatment there is no detectable binding of G␣ i pretreated with trypsin, because the inactive conformation of G␣ is not resistant to trypsin digestion and is degraded prior to binding, while a stable proteolytic fragment of G␣ i is formed in the presence of trypsin and AlF 4 Ϫ that binds to GST-cAGS8 (27). This demonstrates that the G␣ i1 subunit is activated by AlF 4 Ϫ treatment. The overall conclusion is that the AlF 4 Ϫ -dependent activation does not significantly alter binding of G␣ i to AGS8. GST-cAGS8 also did not affect the rate of binding of GTP␥S to G␣ i1 or G␣ i1 ␤ 1 ␥ 2 indicating that AGS8 is not a guanine nucleotide exchange factor (Fig. 2C). These data are consistent with the observation that the cAGS8-mediated activation of G-protein signaling in the yeast-based functional screen was independent of guanine nucleotide exchange (13).
To assess binding of cAGS8 to G protein heterotrimer, 30 nM G␣ i1 GDP and 30 nM G␤ 1 ␥ 2 subunit were incubated together with varying concentrations of cAGS8 and compared with binding to either subunit alone. G␣ i1 and G␤ 1 ␥ 2 subunit bound simultaneously to substoichiometric concentrations of GST-cAGS8 (Fig. 3, lane 9, with 10 nM cAGS8 and 30 nM for each subunit), and binding as a function of GST-cAGS8 concentration was more efficient for both subunits together than for either of the subunits alone (Fig. 3A, lanes 6 -9 compared with lanes 2-5, Fig. 3B, lanes 6 -9 compared with lanes 2-5). The data in Fig. 3 (C and D) are compiled from quantitative analysis of chemiluminescence intensities from three separate binding experiments each. We could not estimate an accurate K d for complex formation from this analysis, but GST-cAGS8 consistently gave stronger binding signals with the heterotrimer than with either subunit alone. The simplest explanation for this data is that cAGS8 binds to a complex of ␣ and ␤␥ and that the affinity for this complex is greater than for either subunit alone.
To provide further evidence for formation of the ternary complex we performed gel-filtration chromatography to evaluate the size and composition of potential complexes between G␣ i1 -GDP, G␤ 1 ␥ 2 , and GST-cAGS8 (Fig. 4). All of the individual proteins and the heterotrimer elute at approximately their predicted molecular masses, including GST-cAGS8, which elutes as an apparent monomer (monomeric molecular mass predicted to be 75 kDa). Incubation of GST-cAGS8 with either G␣ i1 -GDP or G␤ 1 ␥ 2 resulted in the appearance of cAGS8 and either G␣ i1 or G␤ 1 ␥ 2 in an earlier fraction (fraction 24) than either of the G protein subunits alone (fraction 27) at a molecular weight indicative of formation of a 1:1 complex between GST-cAGS8 and either G␣ i1 or G␤ 1 ␥ 2 . The proportion of cAGS8 associated with a G␤ 1 ␥ 2 subunit complex appears to be greater than the proportion of cAGS8 with G␣ i1 consistent with higher affinity binding of cAGS8 to G␤ 1 ␥ 2 relative to G␣ i1 . When cAGS8 was mixed with G␣ i1 and G␤ 1 ␥ 2 together, the three proteins eluted earlier (fraction 23) than the individual cAGS8-G␣ or cAGS8-G␤␥ complexes at a position consistent with the predicted molecular weight of a 1:1:1 complex, indicating formation of a stable ternary complex containing GST-cAGS8, G␣ i1 , and G␤ 1 ␥ 2 . A, 300 pM fluorescein isothiocyanate-labeled myristoylated G␣ i1 was mixed with 50 pM bG␤ 1 ␥ 2 in the presence of the indicated concentrations of SIGK (OE) or cAGS8 (f), and the amount of fluorescein isothiocyanate-labeled myristoylated ␣ i1 bound was assessed under equilibrium conditions by flow cytometry as described under "Experimental Procedures." B, AGS8 bound to G␣ i1 in an activation-state independent manner. 300 nM GST-cAGS8 was incubated with 30 nM purified G␣ i1 with or without AlF 4 Ϫ (30 M AlCl 3 , 10 mM NaF, 10 mM MgCl 2 ) in GST pulldown buffer, which includes 10 M GDP. Proteins bound to GST-cAGS8 were pulled down with glutathione-Sepharose and detected with immunoblotting as described under "Experimental Procedures." Data shown are representative of three independent experiments. This Western blot was quantified by chemiluminescence imaging: binding as percentage of input was GST (0%), G␣ i1 GDP (2.1%), G␣ i1 GDP plus AlF 4 Ϫ (2.4%), G␣ i1 GDP plus trypsin (0.4%), G␣ i1 GDP plus AlF 4 Ϫ plus trypsin (2.1%). C, AGS8 did not cause nucleotide exchange on G␣ i1 . 120 nM G␣ i1 (6 pmol) was mixed with (Ⅺ and F) or without (f and ) 240 nM bG 1 ␥ 2 , in the presence ( and Ⅺ) or absence (f and cAGS8 Binds to a Rearranged G␣␤␥ Complex-Experiments were conducted to determine how the complex between GST-cAGS8 and the G protein heterotrimer is assembled. There are two interaction sites between G␣ and G␤␥: the G␣ switch II interacts with the hot spot on G␤, and the amino-terminal helix of G␣ (␣-NT) interacts with blade 1 of the G␤ propeller (2) (see purple helix, Fig 10A). We hypothesized that cAGS8 binding competes for ␣ subunit switch II interactions with the hot spot, but interactions of G␣-NT with the G␤ blade region are maintained. In the complex, simultaneous G␣ subunit interactions with cAGS8 bound to the hot spot and with G␤ blade 1 result in a bivalent interaction that might be strong enough to maintain a stable ternary complex. To test this idea G␣ i1 subunit missing the amino-terminal 21 amino acids (G␣(tt)) of the ϳ30-amino acid G␣-NT was prepared by limited trypsin digestion. Trypsin treatment of the AlF 4 Ϫ -activated G␣ subunit results in specific cleavage at Arg-21 of G␣ i1 (27,28) and formation of a stable trypsin-resistant fragment, as discussed earlier. This trypsin-digested G␣ subunit still binds to cAGS8 (Fig. 2B), but, as has been previously reported, G␣(tt) did not bind G␤␥ (Fig. 4B, top three panels). If interactions between G␣-NT with G␤␥ are required for formation of the ternary complex then it would be expected that G␣(tt) would not form a complex with cAGS8 and G␤␥. Indeed when GST-cAGS8 was mixed with trypsintreated G␣ i1 and G␤ 1 ␥ 2 , only the GST-cAGS8-G␤ 1 ␥ 2 or GST-cAGS8-G␣ i1 (tt) complexes were formed eluting in fraction 24, with no formation of the GST-cAGS8/heterotrimer complex in fraction 23. This clearly demonstrates that cAGS8 is binding to a non-dissociated complex of G␣␤␥ requiring interaction of ␣ i1 -NT with G␤␥ to maintain the complex.
G␣ i1 ␤ 1 ␥ 2 -cAGS8 Ternary Complex Is Signaling Competent-AGS8 activates G protein signaling in yeast, and studies in transfected COS7 cells showed that transfection of cAGS8 could stimulate PLC ␤2 activation in cells transfected with G␣/␤␥, suggesting that AGS8 can stimulate G␤␥ signaling from a G protein heterotrimer. Thus the data demonstrating that cAGS8 binds to the heterotrimer raises a central question: How can cAGS8 binding to G␤␥ or G␣␤␥ cause activation of G protein signaling to allow propagation of signaling by G␤␥? Based on the data presented so far a simple model where cAGS8 FIGURE 3. AGS8 binds to both G␣ i1 , G␤ 1 ␥ 2 , and the G␣-␤␥ heterotrimer. A, glutathione-Sepharose isolation of cAGS8 complexed to either G␣ alone or G␣␤␥ heterotrimer. 30 nM G␣ i1 -GDP or G␣ i1 -GDP with 30 nM G␤ 1 ␥ 2 was mixed with the indicated concentration of GST-cAGS8. Proteins bound to GST-cAGS8 were precipitated with glutathione-Sepharose and detected with immunoblotting as described under "Experimental Procedures." Data shown are representative of three independent experiments. B, GST-agarose isolation of GST-cAGS8 complexed to either G␤␥ or G␣␤␥ heterotrimer. 30 nM G␤ 1 ␥ 2 or G␤ 1 ␥ 2 with 30 nM purified G␣ i1 -GDP were mixed together with the indicated concentration of GST-cAGS8, and the binding was assayed as in A. Data are representative of three independent experiments. C, quantitative evaluation of binding curves of G␣ i1 for cAGS8 in the presence (OE) or absence (f) of G␤ 1 ␥ 2 . The immunoblot in A was evaluated with quantitative chemiluminescence imaging with a charge-coupled device camera to generate binding curves. Data were pooled from experiments performed at least three times. D, quantitative evaluation of binding curves of G␤␥ for cAGS8 in the presence (OE) or absence (f) of G␣ i1 . The immunoblot in B was quantitated as described in C. Data were pooled from experiments performed at least three times. FIGURE 4. Gel-filtration analysis of complex formation between G␣ i1 , ␤ 1 ␥ 2 , and GST-cAGS8. A, analysis of G␤ 1 ␥ 2 and ␣ i1 binding to GST-cAGS8. All proteins were mixed in 500 l at 100 nM each. B, same as in A with trypsintreated ␣ i1 (␣ i1 (tt)). Excluded blue dextran eluted in fraction 13. Gel-filtration experiments were performed as described under "Experimental Procedures." Each panel is from silver staining of SDS-PAGE separation of proteins in each fraction from the gel-filtration column. Fraction numbers at the top correspond to 1-ml fractions. On the left are the proteins that were mixed prior to gel filtration and on the right are labeling of the individual proteins. GST-cAGS8 elutes at a concentration consistent with a monomer. occupies the hot spot and competes with G␣ for binding and releasing free G␤␥ cannot be correct. Additionally, either "free" G␤␥ or the cAGS8-␣/␤␥ complex would have cAGS8 bound at a critical effector binding site and be predicted to be unable to signal to effectors such as PLC␤2, because cAGS8 binding to the G␤␥ hot spot requires amino acids also required for PLC␤2 activation. To determine if cAGS8 could block signaling by G␤␥, cAGS8 was tested for inhibition of G␤␥-dependent PLC activation. cAGS8 at concentrations as high as 1 M did not affect PLC␤2 activation by G␤ 1 ␥ 2 , whereas SIGK peptide, which also binds the hot spot, potently and effectively inhibited G␤ 1 ␥ 2 -dependent activation (Fig. 5A). Thus G␤ 1 ␥ 2 with cAGS8 bound to a key signaling interface can still regulate a downstream target in vitro.
To directly test if cAGS8 could activate signaling by a G protein heterotrimer the ability of purified GST-cAGS8 to stimulate PLC␤2 activation in vitro by G␣ i1 -GDP/G␤ 1 ␥ 2 was assessed. In the presence of stoichiometric amounts of G␣ i1 -GDP, G␤ 1 ␥ 2 -dependent PLC␤2 activation is completely inhibited (Fig. 5B). cAGS8 restored the ability of G␤ 1 ␥ 2 to activate PLC␤2 in the presence of G␣ i1 -GDP, in a concentration-de-pendent manner with an EC 50 of ϳ30 nM (Fig. 5B). Because cAGS8 does not dissociate G␣ from G␤␥, the data suggest that the heterotrimeric G protein complex can stimulate PLC␤2 when bound to cAGS8. This is surprising, because both G␣ i1 and cAGS8 bind to a region on G␤␥ thought to be required for PLC␤2 activation (14,16).
Complex Formation between cAGS8, G␣ i1 -GDP, G␤␥, and PLC␤2-To directly determine if a complex can be formed between GST-cAGS8, G␣ i1 -GDP, G␤␥, and PLC␤2, immobilized avidin was used to precipitate bG␤ 1 ␥ 2 with various potential binding partners. bG␤ 1 ␥ 2 binds PLC␤2 as expected (Fig. 6A, lane 5) and is inhibited in the presence of an equimolar concentration of G␣ i1 (Fig. 6A, lane 6). Addition of cAGS8 restored PLC␤2 binding to the level observed in the absence of G␣ i1 , and G␣ i1 remains bound in the complex (Fig. 6A, lane 8). In the absence of G␣ i1 , cAGS8 did not affect binding of PLC␤2 (lane 7). These data indicate that cAGS8, G␣, G␤␥, and PLC␤2 form a quaternary complex and AGS8-dependent binding of PLC␤2 by the heterotrimer does not involve subunit dissociation. The functional data in Fig. 5 indicate that the AGS8-bound complex is an active signaling complex.
Finally we used gel filtration to demonstrate that cAGS8 promotes binding of PLC␤2 to the heterotrimer (Fig. 7). In these experiments PLC␤2 eluted as an apparent monomer. When PLC␤2 was mixed with G␤ 1 ␥ 2 prior to gel filtration (PLC␤2 plus G␤ 1 ␥ 2 ) both PLC␤2 and G␤␥ eluted earlier than either protein alone, indicating complex formation. When PLC␤2 was mixed with the G protein heterotrimer (PLC␤2 plus G␤ 1 ␥ 2 plus G␣ i1 ) a small proportion of the PLC␤2 eluted at a higher molecular weight, but only G␤␥ was associated with this complex indicating that PLC formed a stable complex with the free G␤ 1 ␥ 2 in the mixture but not with the heterotrimer. In contrast, when GST-cAGS8 was added to the G protein heterotrimer (cAGS8 plus PLC␤2 plus G␤ 1 ␥ 2 plus G␣ i1 ), all of the proteins eluted at a much higher molecular weight than any of the individual proteins or complexes demonstrating formation of a stable complex between GST-cAGS8, G␣ i1 , G␤ 1 ␥ 2 , and PLC␤2. This demonstrates conclusively that bound cAGS8 promotes association of PLC␤2 with the cAGS8-G␣/␤␥ complex despite AGS8 binding to a key signaling surface on G␤␥.
cAGS8 Rescues the Loss of Function of G␤ 1 (W99A)␥ 2 Mutant-The currently accepted model for activation of G␤␥ signaling to downstream effectors is that separation of G␤␥ from G␣ uncovers the hot spot on G␤␥ allowing for the binding of effectors to this region (3). In the presence of cAGS8, or both cAGS8 and G␣ subunit, this surface would be expected to be occupied and the resulting complexes would be unable to signal downstream. Nevertheless, we demonstrated that both the cAGS8-G␤␥ complex and the cAGS8-G␣␤␥ complex can bind and activate PLC␤2. To understand this we hypothesized that, in these complexes, the hot spot on G␤␥ is no longer utilized for PLC␤2 activation, and a new activation site is exposed in the cAGS8-G␤␥ complexes that can activate PLC␤2. To test this idea a mutant that disables the hot spot for PLC␤2 activation, but not cAGS8 binding, was analyzed. It had been previously reported that G␤ 1 (W99A)␥ 2 is defective for PLC␤2 activation (14), but G␤ 1 (W99A)␥ 2 still bound efficiently to cAGS8 (Fig.  1C, see Fig 10A, blue amino acid). If a new binding/signal transfer surface for PLC activation was created in the presence of cAGS8, then binding of cAGS8 might rescue PLC␤2 binding and activation by G␤ 1 (W99A)␥ 2 . First binding of PLC␤2 to bG␤ 1 ␥ 2 and bG␤ 1 (W99A)␥ 2 was tested (Fig. 8A). Mutation of G␤ 1 W99 to Ala significantly inhibited binding of PLC␤2, indicating that this amino acid, previously shown to be important for activation of PLC␤2, is also important for binding to PLC␤2 (Fig. 8A, lanes 6 and 7). Addition of cAGS8 rescued binding of PLC␤2 to bG␤ 1 (W99A)␥ 2 to levels approaching binding to wt bG␤ 1 ␥ 2 (Fig. 8A, compare lanes 6, 7, and 9); whereas the binding of PLC␤2 to wt G␤ 1 ␥ 2 did not change in the absence or presence of cAGS8 (Fig. 8A, compare lanes 6 and 8).
Next it was determined if cAGS8 could restore the ability of G␤ 1 (W99A)␥ 2 to activate PLC␤2 (Fig. 8B). Wild-type G␤ 1 ␥ 2 stimulated PLC␤2 12-fold, whereas mutation of ␤W99 to Ala greatly impaired this activation. cAGS8 restored the capacity of G␤ 1 (W99A)␥ 2 to stimulate PLC␤2 to near wt G␤ 1 ␥ 2 levels.  cAGS8 itself did not stimulate PLC␤2, as described earlier and cAGS8 did not affect wt G␤ 1 ␥ 2 -dependent activation of PLC␤2. Together these data indicate that cAGS8 unmasks and/or creates a new binding and signal transfer site for PLC␤2 that does not involve the hot spot on G␤␥.
cAGS8 May Unmask a New Binding Surface at the Amino Terminus of ␤␥-The data support a model where AGS binds to the heterotrimer with simultaneous interactions with G␤␥ at the hot spot and the G␣ subunit and that the hot spot is no longer used for signal transfer to PLC␤2 activation, because it is occupied with cAGS8. To accommodate this model we propose that another site for effector interactions is unmasked on G␤␥ that signals to downstream effectors. We recently presented evidence that PLC␤2 can bind to two distinct regions on G␤, at the amino-terminal coiled-coil domain, and at the hot spot (29). In this earlier work it was proposed that association of PLC␤2 with the G␤␥ coiled-coil region inhibited PLC activation.
This indicates that the amino-terminal coiled-coil binding region on G␤ binds PLC␤2 when the hot spot is occupied by cAGS8. We next tested whether activation of PLC␤2 by cAGS8-G␤␥ complex was affected by the amino-terminal mutation. We predicted that, if this alternate binding site for PLC␤2 was disabled, the G␤␥ would no longer be able to activate PLC␤2 with cAGS8 bound to the hot spot. cAGS8 did not significantly inhibit wild-type G␤ 1 ␥ 2 -stimulated PLC␤2 activity (see also Fig. 5A). However, cAGS8 inhibited G␤ 1 (23-27)␥ 2stimulated PLC␤2 activity with an IC 50 of ϳ15 nM, indicating that the amino-terminal interaction region was in fact a second site required for stimulation of PLC␤2 activity by the cAGS8-G␤␥ complex. This also suggests that the apparently inhibiting activity of the amino terminus toward PLC␤2 participates in activation of PLC␤2 upon cAGS8 binding.

DISCUSSION
Here we describe a mechanism for activation of G protein signaling that does not involve nucleotide exchange or subunit dissociation. Multiple assays of direct binding demonstrate that cAGS8 binds to the G protein heterotrimer and promotes binding of PLC␤2 to G␤␥ subunits to form a quaternary signaling complex. Gel-filtration analysis demonstrates that the quaternary complex is stable, indicating the proteins in the complex interact with relatively high affinity. In addition, AGS8 modulated G protein-dependent PLC regulation with EC 50 or IC 50 values of ϳ30 nM (Figs. 5B and 9C). A key experiment, supporting the idea that subunit dissociation is not needed for signaling in the quaternary complex, demonstrated that cAGS8 could rescue the function of a mutant G␤␥ (G␤ 1 W99A) with the hot spot at the G␣/␤␥ interface disabled for PLC␤2 binding and activation. This experiment rules out the possibility that undetected free G␤␥ might be generated upon AGS8 binding that somehow transiently reveals the hot spot for PLC␤2 binding In this experiment basal PLC activity was 5.4 mmol/mg/min and with wt bG␤ 1 ␥ 2 -stimulated activity was 25 mol/mg/min, and bG␤ 1 (23-27)␥ 2 -stimulated activity was 42 mol/mg/min (as previously reported, bG␤ 1 (23-27)␥ 2 stimulates PLC activity more effectively than wt bG␤ 1 ␥ 2 (29)). Data are representative of experiments performed at least three times. and activation. Clearly cAGS8 cannot be acting simply by dissociating G␣ from G␤␥ to expose the hot spot on G␤␥.
It was initially surprising that cAGS8 binding to the hot spot on G␤␥ did not result in subunit dissociation, whereas the peptide SIGK that binds to an overlapping site promotes subunit dissociation (16,22). These observations can be reconciled by the observation that AGS8, but not SIGK, also exhibits some affinity for G␣ subunits. In this scenario, when AGS8 binds the hot spot, G␣ loses contact with the hot spot but gains new interactions with AGS8 sufficient to maintain the ternary complex in cooperation with interactions between the G␣-NT and G␤␥. Alternatively, a model could be suggested based on a recent structural model of GRK2 simultaneously bound to both G␤␥ and G␣ q . Here G␣ subunits are dissociated from the G␤␥ subunits but remain bound to two different sites of GRK2 (30). In support of the first model, the amino-terminal helix of G␣ i1 is required for maintenance of the quaternary complex but not AGS8 binding. Thus we propose a modified version of the clam shell hypothesis, where the activator (AGS8) rather than the effector binds to a rearranged complex of G␣ and G␤␥ that is maintained by contacts between the amino-terminal helix of G␣ with G␤␥ (see Fig. 10B).
Because G␤W99A disables the hot spot for PLC␤2 binding and activation, cAGS8 itself did not bind to or activate PLC␤2, and cAGS8 occupied the region on the G␤ subunit required for PLC␤2 binding. The only way that cAGS8 could rescue the binding and activation of PLC␤2 by G␤W99A would be by creating a new binding site for PLC␤2 in the cAGS8-G␤␥ complex. Site-directed mutagenesis has shown that there are contact surfaces outside of the hot spot that are important for effector recognition (29,31).
A binding site for PLC␤2 in the amino-terminal coiled-coil region of the ␤␥ subunits has previously been identified by chemical cross-linking and site-directed mutagenesis (29,32).
In those studies it was demonstrated that disabling this binding surface by site-directed mutagenesis resulted in increased G␤␥dependent activation of PLC␤2 leading to the conclusion that this binding site for PLC was inhibitory while the binding site in the hot spot was stimulatory. Here it is demonstrated that the G␤␥ coiled-coil region is required for binding and activation of PLC␤2 when AGS8 is bound. This suggests that this binding site is converted from an inhibitory to a stimulatory site in the presence of cAGS8. It is also possible that PLC␤2 binding and activation in the complex requires both binding to the G␤␥ amino terminus and cAGS8, although no binding or activation of PLC␤2 by cAGS8 alone was detected. These data suggest a signaling mechanism for non-dissociated G protein complexes that uses binding sites outside the G␣/␤␥ interface for effector activation.
Interestingly, the signal transfer surfaces of the G␤␥ subunit in yeast, Saccharomyces cerevisiae, have not been shown to involve amino acids at the G␣/␤␥ interface but rather have been mapped to a region in the amino-terminal coiled-coil at a position that is nearly identical to the region we described for PLC binding (33)(34)(35). In contrast to the hot spot region where G␥ subunits cannot directly contribute to binding, the more variable G␥ subunits could conceivable directly contribute to this contact interface. Thus some signaling specificity could be supplied by binding effectors to this region.
Mechanisms for activation of G protein signaling by group II AGS proteins are distinct from receptor-mediated mechanisms for G protein activation but involve an apparently straightforward subunit dissociation-based mechanism where the GPR motif of the AGS protein binds G␣-switch II and either promotes subunit dissociation or competes for G␤␥ subunit binding, leading to accumulation of free G␤␥ (22,36,37). Here we describe a model for G␤␥ activation that does not involve revealing the hot spot on G␤␥ for PLC binding but rather involves alternative utilization of binding sites outside the G␣-␤␥ interface upon binding of the activator AGS8. This mechanism has significant implications for mechanisms of sig- FIGURE 10. Model for cAGS8-promoted G␤␥ signaling from G protein heterotrimers. A, three-dimensional representation of the structure of G␤␥ showing relevant binding surfaces and mutants. In dark gray is the ␤ subunit with bright green and red spaces showing amino acids important for cAGS8 binding. In red are amino acids important for both cAGS8 and PLC␤2 interactions. Trp-99, whose mutation to alanine disables PLC activation but not cAGS8 binding, is in blue. The orange ribbon helix representation is the ␣ subunit switch II, and the purple ribbon helix is the ␣ subunit amino-terminal helix. The gamma subunit is represented by the red helical ribbon. In yellow are amino acids mutated in the G␤ 1 (23-27)␥ 2 mutant. B, in the classic model for G protein activation, subunit dissociation after activation by G proteincoupled receptors leads to exposure of the hot spot required for binding and activation of PLC␤2. During activation by cAGS8, cAGS8 binds to the hot spot and ␣ remains associated through a bivalent interaction with the blade regions on ␤␥ and bound cAGS8. This creates a hot spot-independent signaling surface that can bind and activate effectors.
naling by non-dissociated G protein complexes that probably extends beyond AGS8 regulation.
It is not clear from our studies how AGS8 is regulated. One possible mechanism is transcriptional up-regulation of AGS8 as was observed in cardiac myocytes subjected to ischemic stress (13). Another possibility is that AGS8 is regulated posttranslationally by an upstream stimulus. It is important to emphasize that this study examined only a carboxyl-terminal portion of the full-length AGS8 protein. We elucidated how this G protein-binding domain of AGS8 can regulate heterotrimeric G protein function, but this carboxyl terminus may in turn be regulated by intramolecular interactions in full-length AGS8. Thus, it is important that we understand how this domain operates in the context of the full-length AGS8 protein.
Full-length AGS8 has been difficult to express and purify. It has been expressed in COS7 cells and does not inhibit G␤␥-dependent PLC␤2 activation (13), consistent with the data presented here. We are currently exploring various strategies to address the properties of full-length AGS8 as we move forward with these studies.
The subject of G protein subunit dissociation has been the subject of significant interest driven in part by the advent of cellular resonance energy transfer methods that allow alterations in protein interactions to be monitored in intact cells. Based primarily on the observation that, in some cases receptor activation leads to increased resonance energy transfer between G␣ and G␤ or G␥ subunits, it has been proposed that G protein subunits undergo conformational rearrangement rather than dissociation (5,7). Recent fluorescence recovery after photobleaching studies, on the other hand, favors a subunit dissociation-based model for at least some G protein isoforms (38). Analysis of G protein subunit interactions with GIRK channels suggests that G␣ and G␤␥ subunits are prebound to the channel and that conformational rearrangements lead to channel activation (39). These experiments suggest that conformational alterations of assembled G protein heterotrimer-effector complexes occur in cells that are presumably sufficient to activate effector functions, but the nature of these conformations are difficult to discern from this type of analysis. Recent determination of the structure of GRK2 bound to G protein ␣ q and G␤␥ subunits simultaneously provides atomic level detail for how a G protein trimer can interact with a target to form a quaternary complex, but in this complex the G protein subunits are dissociated from each other with each subunit separately interacting with different domains of GRK2 (30). The model presented here, although based on in vitro data, provides molecular details for one potential model for how non-dissociated G protein trimers may regulate effector targets that would be difficult to approach in intact cells.