Gβγ Inhibits Gα GTPase-activating Proteins by Inhibition of Gα-GTP Binding during Stimulation by Receptor*

Gβγ subunits modulate several distinct molecular events involved with G protein signaling. In addition to regulating several effector proteins, Gβγ subunits help anchor Gα subunits to the plasma membrane, promote interaction of Gα with receptors, stabilize the binding of GDP to Gα to suppress spurious activation, and provide membrane contact points for G protein-coupled receptor kinases. Gβγ subunits have also been shown to inhibit the activities of GTPase-activating proteins (GAPs), both phospholipase C (PLC)-βs and RGS proteins, when assayed in solution under single turnover conditions. We show here that Gβγ subunits inhibit G protein GAP activity during receptor-stimulated, steady-state GTPase turnover. GDP/GTP exchange catalyzed by receptor requires Gβγ in amounts approximately equimolar to Gα, but GAP inhibition was observed with superstoichiometric Gβγ. The potency of inhibition varied with the GAP and the Gα subunit, but half-maximal inhibition of the GAP activity of PLC-β1 was observed with 5–10 nm Gβγ, which is at or below the concentrations of Gβγ needed for regulation of physiologically relevant effector proteins. The kinetics of GAP inhibition of both receptor-stimulated GTPase activity and single turnover, solution-based GAP assays suggested a competitive mechanism in which Gβγ competes with GAPs for binding to the activated, GTP-bound Gα subunit. An N-terminal truncation mutant of PLC-β1 that cannot be directly regulated by Gβγ remained sensitive to inhibition of its GAP activity, suggesting that the Gβγ binding site relevant for GAP inhibition is on the Gα subunit rather than on the GAP. Using fluorescence resonance energy transfer between cyan or yellow fluorescent protein-labeled G protein subunits and Alexa532-labeled RGS4, we found that Gβγ directly competes with RGS4 for high-affinity binding to Gαi-GDP-AlF4.

of G␣. Such negative cooperative binding permits G␤␥ to regulate effectors during G protein activation. In solution, activation of G␣ by a nonhydrolyzable GTP analog can drive physical dissociation of G␤␥ (1-4) (see Refs. 5 and 6 for reviews). Complete dissociation of G␤␥ from G␣-GTP also occurs to some extent in biological membranes during receptor-initiated activation of G␣ (7,8). However, recent data suggest that G␤␥ bound to G␣-GTP can also interact productively with effector proteins without dissociating from the heterotrimer (8 -11). This implies that the G␣-GTP-G␤␥ complex somehow exposes the sites on G␣ and G␤␥ necessary for productive interaction with effector proteins while they remain bound to each other.
G␤␥-regulated effectors include K ϩ and Ca 2ϩ channels, adenylyl cyclase, phospholipase C-␤ (PLC-␤), 4 PI 3-kinase, and some protein kinases (6). Effector interaction sites on the G␤␥ molecule that exert these various functions have been mapped by a variety of chemical and mutagenic strategies. Together, they appear to cover a large fraction of the G␤␥ surface, despite significant overlap that makes several effectors bind competitively with respect to each other (6,(12)(13)(14)(15)(16). G␤␥ also binds a diverse collection of other proteins: G protein-coupled receptor kinases, phosducin, AGS1, and receptors (6).
In addition to its other functions, G␤␥ subunits have been shown to inhibit the activity of G protein GAPs in solution-phase, single turnover assays (17)(18)(19). GAPs accelerate hydrolysis of GTP bound to G␣ and thus promote G␣ deactivation. GAPs can thereby inhibit steady-state signaling, accelerate signal termination when receptor agonist is removed, alter receptor selectivity among G proteins, and/or suppress basal (receptor-independent) signal output (see Ref. 20 for review). G␤␥ inhibits the GAP activity of both RGS proteins and phospholipase C-␤s and inhibits GAP activity toward all G␣ subunits tested. The potency and extent of inhibition vary considerably among GAPs and G␣ targets.
The physiological function of GAP inhibition by G␤␥ is intriguing and presently unclear. It may represent a way to shield G␣-GTP from deactivation until it encounters an effector protein, or it may support continued activation in cases where G␤␥ does not physically dissociate from G␣-GTP. The mechanism of inhibition is also not known in detail. We initially speculated that G␤␥ might inhibit GAP activity by binding G␣-GTP and thus blocking the GAP binding site, but the relatively low concentrations of G␤␥ that effectively inhibited several GAPs seemed inconsistent with the presumed low affinity of G␤␥ for G␣-GTP. G␤␥ might also interact directly with GAPs. Binding of G␤␥ to RGS proteins has not been observed, but G␤␥ does bind PLC-␤ and thus stimulate its phospholipatic activity. It was unknown whether the sites that cause phospholipase activation and GAP inhibition are identical. * This work was supported by Grants GM30355 from the National Institutes of Health and I-0982 from the R. A. Welch Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  We have now examined the mechanism whereby G␤␥ inhibits G protein GAP activity using both enzyme kinetic approaches and direct optical measurements of G␣-RGS binding and dissociation. We find that G␤␥ inhibits GAPs by competitively binding to G␣-GTP. In the case of PLC-␤, the G␤␥ binding site that mediates PLC stimulation is not required for inhibition of GAP activity. G␤␥ binds G␣-GTP with unexpectedly high affinity, with values of K d well within the range of concentrations of G␤␥ associated with regulation of its effectors. Thus G␤␥ may protect G␣ subunits from GAP-accelerated deactivation, and G␤␥ from one heterotrimer may exert this effect on a heterologous G␣.

EXPERIMENTAL PROCEDURES
Proteins-m1AChR and m2AChR, G␣ q , G␣ z , G␤1␥2, and PLC-␤s were expressed in Sf9 cells and purified as described (21). Heterotrimeric G proteins were prepared by mixing stoichiometric amounts of purified G␣ and G␤1␥2. G␤1␥2 isoforms were used in all experiments. RGS-Z1 was expressed in Escherichia coli and purified as described (22). cDNAs for fusions of G␣ i1 with enhanced GFPs (Clontech) were prepared in pQE60 (Qiagen). GFP was inserted between residues 118 and 119 of G␣ i1 with the linker peptides GTSGGGGS and SGGGGSGTAG-GHHHHHHGGG. The citrine YFP mutant, which is minimally sensitive to quenching by F Ϫ (23), was prepared using the QuikChange protocol (Stratagene). Wild-type and mutant G␣ i1 were co-expressed in E. coli with yeast protein N-myristoyltransferase and purified as described (24). The PLC-␤1⌬141 mutation, in which amino acid residues 2-141 are removed, was generated by PCR. The N-terminal hexahistidine tag was retained. Recombinant baculoviruses were produced as described previously (25). PLC-␤1⌬141 was purified essentially as described for wild-type PLC-␤1 (21). PLC-␤2 and PLC-␤3, both hexahistidine-tagged, were purified as described (21) after expression in Sf9 cells using baculovirus vectors that were a gift from Paul Sternweis (University of Texas Southwestern Medical Center). Purified phosducin was a gift from Barry Willardson (Brigham Young University).
RGS4-197, a mutant RGS4 in which all cysteine residues except Cys 197 were removed, was constructed by several rounds of QuikChange mutagenesis of the cDNA in the pGEX-2 expression vector. RGS4-197 displays essentially wild-type GAP activity both before and after its reaction with maleimides at the sole remaining cysteine residue. Mutant and wild-type RGS4 were expressed as glutathione S-transferase fusions using the pGEX-2 vector. For production of wildtype and mutant RGS4, transformed BL21DE3/pREP4 cells were grown in T7 medium (containing ampicillin, kanamycin, and isopropyl 1-thio-␤-D-galactopyranoside) for 6 h. Cell pellets were frozen in liquid N 2 . Bacteria were thawed on ice and incubated with 0.5 mg/ml lysozyme in Buffer A (20 mM Tris, pH 8.0, 100 mM NaCl, 1 mM dithiothreitol, 1 g/ml aprotinin, 10 g/ml leupeptin, and 0.1 mM phenylmethylsulfonyl fluoride) for 30 min on ice. After sonication (three 1-min bursts), the suspension was centrifuged at 4°C at 15,000 ϫ g for 40 min. The supernatant was diluted 1:2 with Buffer A and mixed with glutathione-agarose for 2 h at 4°C. The resin was washed with Buffer A and Buffer A plus 0.1% Lubrol plus 90 mM NaCl. Finally, the resin was equilibrated in 0.1 mM CaCl 2 , removed from the column, and mixed with 4 l/ml thrombin for 17-18 h at 4°C. The resin was returned to the column and eluted with another 2 volumes of Buffer A plus 0.1% Lubrol, concentrated by ultrafiltration, frozen in liquid N 2 , and stored at Ϫ80°C.
RGS4-197 was labeled with the thiol-reactive probe, Alexa532-C 5maleimide, which does not change its activity as a G q GAP. Prior to labeling, RGS4-197 (800 g) was dialyzed with 3000 volumes of buffer (25 mM Hepes, pH 7.5, 100 mM NaCl, 0.05% Lubrol) to remove dithiothreitol. A 10-fold molar excess of probe was added and incubated for 45 min at 0°C, and the reaction was quenched with ␤-mercaptoethanol. The reaction mixture was centrifuged at 4°C at 13,000 rpm for 10 min, and the supernatant was applied to a G-25 gel filtration column (30 ml). Peak fractions were pooled, concentrated, frozen in liquid N 2 , and stored at Ϫ80°C. Labeling efficiency was determined according to the protein concentration (26) and the absorption of Alexa532 at 528 nm (⑀ ϭ 78,000 cm Ϫ1 M Ϫ1 ).
Unilamellar receptor-G protein vesicles were reconstituted as described previously (21). Vesicles were incubated with 5 mM dithiothreitol at 0°C for 1 h to activate receptors (21). The concentration of receptors in the vesicles was assayed by [ 3 H]quinuclidinylbenzilate binding (27). Receptor-coupled G␣ was measured according to carbachol-stimulated [ 35 S]GTP␥S binding (21).
Functional Assays-Single turnover GAP assays, which measure acceleration of the hydrolysis of a preformed complex of G␣-[␥-32 P]GTP in detergent solution, were performed as described previously (28). Temperature, which was varied for some experiments, and concentrations used for specific experiments are given in the figure legends. Steady-state GAP activity was determined according to the increase in agonist-stimulated GTPase activity in phospholipid vesicles that contained trimeric G␣␤ 1 ␥ 2 and receptor (28).
To measure the rate of hydrolysis of G q -bound GTP under the conditions of steady-state turnover, m1AchR-G q vesicles and agonist were first mixed in assay buffer (21) for 25 min at 10°C to allow association of receptor and G q . An aliquot (20 l) was then mixed with 5 l of assay buffer that contained carbachol and [␥-32 P]GTP (300 nM final concentration). The mixture was incubated for 1 min to initiate steady-state hydrolysis and allow accumulation of G q -[␥-32 P]GTP without excessive production of background [ 32 P]P i . At this point, defined as t ϭ 0, further [␥-32 P]GTP binding was quenched by a 1:2 dilution with 2 mM nonra-FIGURE 1. G␤␥ inhibits steady-state GTPase activity stimulated by m1AChR and PLC-␤1. GTPase activity was measured at 30°C in phospholipid vesicles reconstituted with heterotrimeric G q and m1AChR in the presence of 1 mM carbachol, either with (A) or without (B) 2 nM PLC-␤1, a concentration slightly below its EC 50 as a G q GAP. At 5 min (arrow), additional G␤␥ was added at either 50 nM (B) or the concentrations shown at right (A). The concentration of G␤␥ contributed by the G q trimer originally was about 2 nM in the vesicles. In A, half-maximal inhibition was observed at 5 nM G␤␥ based on slopes over the 5-20-min time interval. The IC 50 was also 5-10 nM in two similar experiments, and the extent of inhibition was similarly large.
dioactive GTP, 2 mM atropine, and the indicated concentration of PLC-␤1. Hydrolysis of bound [␥-32 P]GTP was quenched at the appropriate times as described (28). Release of [ 32 P]P i was fitted to a single exponential equation to yield the rate constant k hydrol .
To measure the rate of GDP dissociation, [␣-32 P]GDP was first bound to m1AChR-G q vesicles by incubation with 2 M [␣-32 P]GDP, 1 mM carbachol, and 10 nM PLC-␤1, with or without G␤1␥2, for 15 min at 15°C. Aliquots of this mixture were diluted with antagonist and 5 mM unlabeled GTP, dissociation was allowed to proceed for various times, and bound [␣-32 P]GDP was measured as described (21) Phospholipase activity was determined according to hydrolysis of [ 3 H]PIP 2 added as mixed micelles (27). Reactions were initiated by the addition of enzyme, and the reaction was allowed to proceed at 30°C for 10 min. G␤1␥2-stimulated activity was determined in the presence of 1 M free Ca 2ϩ . G␣ q -stimulated activity was measured in the presence of G␣ q that has been activated previously by incubation with GTP␥S (29) plus 200 nM free Ca 2ϩ .
Fluorescence Measurements-Equilibrium fluorescence measurements were performed on a SPEX Fluorolog 211 spectrophotometer using a 3-mm cuvette with both excitation and emission slits set at 1.25 mm. G␣ i1 -CFP was diluted in 200 l of 20 mM NaHepes, pH 8.0, 3 M GDP, 10 M AlCl 3 , 10 mM MgCl 2 , and 5 mM NaF at 25°C. After the addition of RGS4-197Fl, the mixture was incubated for 2 min, and fluorescence was recorded between 460 and 580 nm with excitation at 431 nm. FRET between G␣ i1 -CFP and RGS4-197Fl was measured as the enhancement of RGS4-197Fl emission at 552 nm or the decrease of G␣ i1 -CFP emission at 475 nm over the sum of both components at the same concentrations. The individual emission intensities of both G␣ i1 -CFP and RGS4-197Fl (both excited at 431 nm) were scaled for concentration and subtracted from total observed fluorescence to give a measure of FRET intensity.
Dissociation of G␣ i1 -CFP and RGS4-197Fl was assayed in the same buffer used for equilibrium experiments. Fluorescence stopped flow measurements were performed using a Bio-Logic SFM3 stopped flow mixer in which syringe, mixing chambers, and delay lines are all under thermostat control. G␣ i1 -CFP (20 nM) was incubated with 200 nM RGS4-197Fl in one syringe of the mixer for at least 2 min at 25°C. An aliquot of this mixture (50 l) was mixed with an equal volume of 500 nM nonfluorescent G␣ i1 in the presence or absence of 1 M G␤1␥2. Dissociation was measured by the loss of transferred acceptor fluorescence at 10-ms intervals in a 30-l cuvette. Excitation was set at 431 nm using a monochromator, and emission was measured using a 520 nm cut-off filter.

RESULTS
G␤␥ inhibits the GAP activities of RGS proteins and PLC-␤ (17)(18)(19), modulates nucleotide binding to G␣, and is required essentially for G protein regulation by receptors (6). We therefore asked how G␤␥ alters interactive regulation by receptors and GAPs during steady-state GTPase turnover, which includes all three processes. As shown in Fig.  1A, excess G␤1␥2 markedly and potently inhibited steady-state GTPase activity in a coupled system composed of a receptor, a trimeric G protein, and a GAP. In the case of G q , m1AChR, and PLC-␤1, inhibition was nearly complete, to about the level observed without a GAP (compare with Fig. 1B). Inhibition was maximal at 50 nM G␤␥ and displayed an IC 50 of about 5 nM above the amount of G␤␥ already included in the vesicles as part of the G q heterotrimer (ϳ2 nM). G␤␥ inhibition of GAP activity during steady-state GTPase turnover is thus more potent than during a single catalytic cycle (19). Onset of inhibition was rapid upon the addition of G␤␥ (at 5 min in Fig. 1) and displayed no obvious lag. Qualitatively similar fast and complete inhibition was observed when PLC-␤3 was used as the GAP (not shown) or when the test system was FIGURE 2. G␤␥ inhibits GAP-stimulated GTP hydrolysis but not GDP dissociation. m1AChR-G q vesicles, with (F) or without (E) 100 nM excess G␤␥, were incubated with 1 mM carbachol for 25 min at 10°C. [␥-32 P]GTP (0.25 volume in assay buffer) was added to a final concentration of 300 nM, and aliquots (17 fmol of m1AChR, 44 fmol of G q ) were incubated for 1 min more. At zero time, each aliquot was further diluted 1:2 with 10°C buffer that contained 2 mM unlabeled GTP, 2 mM atropine, and the indicated concentration of PLC-␤1. Hydrolysis of G q -bound GTP was measured, and hydrolysis time courses were fit to single exponential functions to yield the rate constants shown. Each data set is representative of two experiments using different batches of vesicles; composed of m2AChR, G z , and either RGS4 (see Fig. 3B) or RGS-Z1 (not shown). In contrast, receptor-stimulated GTPase activity in the absence of GAP was essentially insensitive to inhibition by G␤␥ (Fig.  1B), suggesting that the site of inhibition is associated with the GAP rather than the receptor.
Inhibition of steady-state GTPase by G␤␥ predominantly reflects inhibition of the GAP-accelerated GTP hydrolysis reaction itself rather than of other steps in the GTPase catalytic cycle. G␤␥ significantly inhibited hydrolysis of G␣ q -GTP in the vesicles (Fig. 2). Excess G␤␥ did not alter the receptor-promoted dissociation of GDP from G protein (Fig. 2, inset), the other relatively slow step in the GTPase cycle (30). Thus, the modest effect of G␤␥ in the single turnover assay (Fig. 2) is magnified when GAP activity is measured at steady-state (Fig. 1).
For both RGS proteins and PLC-␤, G␤␥ added to the receptor-G protein vesicles increased the amount of GAP needed to accelerate GTPase hydrolysis but had little if any effect on the maximal hydrolytic rate that could be attained. In a single turnover assay of GTP hydrolysis in vesicles, G␤␥ increased the EC 50 for PLC-␤1 by more than 4-fold (Fig.  2). G␤␥ also shifted the concentration dependence of GAP activity during steady-state hydrolysis stimulated by agonist-bound receptor. Similar effects were observed with RGS4 or PLC-␤1 (Fig. 3A). Vesicles that contained m2AChR and G z showed similar shifts in RGS4 potency in the presence of G␤␥ (Fig. 3B). In each case, the maximal GTPase activity at high GAP concentration was about the same with or without G␤␥, but the EC 50 for GAP was increased 4 -6-fold. In contrast to its effects on GAP potency, excess G␤␥ had no effect on the potency of agonistbound receptor to accelerate steady-state GDP/GTP exchange (Fig.  3C), again indicating that G␤␥ acts primarily at the GAP-promoted hydrolytic step of the GTPase cycle rather than at GDP/GTP exchange. G␤␥ Inhibits GAP Activity by a K m -based Mechanism-According to x-ray crystallographic structures, the two surfaces of G␣ that interact with G␤␥ and RGS proteins overlap substantially. These surfaces include both the switch regions and the N-terminal helix (22,(31)(32)(33). Inhibition might thus reflect competition between G␤␥ and GAP for binding to G␣-GTP, although affinity of G␤␥ for G␣-GTP is low. Alternatively, inhibition could result from an allosteric decrease of GAP-G␣ affinity caused by the binding of G␤␥ to the GAP-G␣-GTP complex.
An increase in K m without a change in V max is the hallmark of competitive inhibition, but it is also characteristic of a negative cooperative interaction between inhibitor and substrate. We therefore examined the effect of G␤␥ on the K m of RGS-Z1 and RGS4, using G␣ z -GTP as substrate because the K m for these reactions is substantially lower than for other GAP-G␣ pairs. For both RGS-Z1 and RGS4, the major effect of G␤␥ was to increase the K m , with a much smaller or negligible effect on V max (Fig. 4). This pattern was observed when K m and V max were determined by varying the concentration of either G␣ z -GTP or the GAP. The increase in K m by added G␤␥ was mirrored by an increase in the concentration of G␤␥ needed to inhibit the reaction at higher concentrations of GAP (Fig. 5). The half-maximal inhibitory concentration of G␤␥ (IC 50 ) increased nearly linearly over a 30-fold range of RGS4 concentration. The linear relationship suggests competitive inhibition, in this case caused by G␤␥ binding to the G␣ z -GTP substrate. If G␤␥ bound to RGS with negative cooperativity with respect to G␣-GTP, or if it bound G␣ while in a GAP-G␣-G␤␥ ternary complex with similar effect, the curve would be expected to saturate at concentrations consistent with the affinity of G␤␥ for the target. However, such a deviation from linearity is only predicted near or above the K d for the inhibitor, and the lowest observed IC 50 for G␤␥ was about 10 nM. Thus, although these data clearly show that G␤␥ exerts its inhibition via an effect on K m , they do not distinguish the precise molecular mechanism.
Phosducin binds G␤␥ competitively with respect to G␣ (34 -36), and phosducin blocks inhibition of GAP activity by G␤␥ (Figs. 3B and 6). Phosducin decreased the inhibitory potency of G␤␥ and at high concentrations completely blocked its effect. The potency of phosducin is similar to its affinity for G␤␥ (34,35).
Inhibition of the GAP Activity of PLC-␤ Does Not Require the Known G␤␥ Binding Site-PLC-␤ family members are stimulated both by activated G␣ q and by G␤␥ (37). The principal G q -binding domain of PLC-␤ is a coiled coil region that is C-terminal to the catalytic and C2 domains. This domain displays G q GAP activity when expressed separately, and its deletion eliminates both the GAP activity and the G q responsiveness of phospholipase activity (38 -41). An N-terminal region that includes the PH domain is required for response to G␤␥, although other regions have also been implicated as important (42)(43)(44)(45). To determine whether G␤␥ binding to PLC-␤ at the PH domain is required for its inhibition of GAP activity, we compared the ability of G␤␥ to inhibit the GAP activity of intact PLC-␤1 and an N-terminally truncated mutant (⌬141). As shown in Fig. 7A, the GAP activity of the truncation mutant was inhib-  FEBRUARY 24, 2006 • VOLUME 281 • NUMBER 8 ited by G␤␥ to the same relative extent and with similar potency as was wild-type PLC-␤1, suggesting that the PH domain is not required for inhibition of the G q GAP activity of PLC-␤1. As expected, this truncation completely blocked its responsive to stimulation by G␤␥ (Fig. 7B), consistent with previous data that showed that the PH domain is required for interaction of PLC-␤ with G␤␥ (43). The G q GAP activity of the truncation mutant was essentially unaltered (Fig. 7A, potency data not shown), as was its ability to respond to stimulation by G␣ q -GTP␥S (Fig. 7C), although the maximal specific activities of this mutant are somewhat lower in all assays, with or without stimulation. This effect may reflect decreased stability. Regardless, the retention of GAP inhibi-tion by the ⌬141 mutant indicates that if G␤␥ inhibits GAP activity by binding directly to PLC-␤, it must do so via a previously unknown site.

Inhibition of G Protein GAPs by G␤␥ Subunits
G␤␥ Inhibition of G␣-RGS Binding-To determine whether and how G␤␥ blocks the binding of GAPs to G␣-GTP, we measured its effect on RGS4 binding to G␣ i1 . Binding was detected as FRET between a G␣ i1 -CFP fusion protein and RGS4 that was covalently labeled with Alexa532-maleimide at Cys 197 , which is in the conserved RGS domain but in which alkylation does not interfere with GAP activity (data not shown). These proteins bind tightly in the presence of GDP-Al ϩ3 -F 4 Ϫ (31), and this complex displayed considerable FRET from the CFP donor to the Alexa acceptor. According to this direct assay, G␤␥ inhib-  (22) because these assays were performed at 5°C to decrease background hydrolysis. K m is temperature-dependent (data not shown).  ited RGS4 binding to G␣ i1 with IC 50 Ӎ 290 nM (Fig. 8A). Three determinations yielded values between 230 and 500 nM, the range in which G␤␥ half-maximally inhibits the G␣ i GAP activity of RGS4.
To differentiate competitive binding and negatively cooperative binding, we asked whether G␤␥ increases the rate of dissociation of the G␣ i1 -G␤␥ complex. Negatively cooperative interactions, although defined at equilibrium, generally result from an increase in the dissociation rates of the two negatively interacting ligands (G␤␥ and RGS in this case) from a low-affinity ternary complex (G␣-G␤␥-GAP). In contrast, a competitive mechanism has no such kinetic effect, because competing ligands never bind simultaneously. Dissociation of the complex of G␣ i1 -GDP-AlF 4 and RGS4 was measured in the presence or absence of 1 M G␤␥ under the same conditions used for the experiment in Fig.  8A. G␤␥ had no effect on the rate of dissociation of G␣ i -RGS4 (Fig. 8B) even though it markedly inhibited G␣ i -RGS4 binding (Fig. 8A). Thus, if G␤␥ accelerates the dissociation of G␣ i -RGS4, its affinity for the G␣ i -RGS4 complex must be extremely low, such that concentrations much higher than 1 M are required (see "Discussion").
We evaluated the affinity of G␤␥ binding to G␣ i -GDP-AlF 4 to test the feasibility of the competitive inhibition mechanism because G␤␥ binds to active G␣ subunits with reduced affinity. We found that G␣ i1 binds G␤1␥2 with K d ϳ 150 nM under our standard assay conditions (Fig. 8C). The K d can also be calculated from the IC 50 obtained in the experiment of Fig. 8A, assuming a competitive mechanism and given the affinity with which G␣ i binds RGS4. 5 From these data, the K d for G␣ i -G␤␥ binding is 114 nM, in good agreement with the value obtained in Fig. 8C. This affinity is also consistent with the potency with which G␤␥ inhibited GAP activity given the likely assumption that K m ϭ K s for the GAP-catalyzed hydrolysis of G␣ i -GTP (see Ref. 28 for calculations).

DISCUSSION
These data show that G␤␥ subunits inhibit G protein GAP activity in a membrane environment during stimulation by receptor, thus providing a model for the regulation of GAPs in agonist-stimulated cells. Inhibition displayed rapid onset (Fig. 1) and occurred in the range of 20 to 200 nM G␤␥ (Figs. 2 and 3), similar to the concentrations required for activation of known G␤␥-regulated effectors such as K ϩ or Ca 2ϩ channels, PI 3-kinase, PLC-␤, or adenylyl cyclase (6). Inhibition of GAPs provides a novel mechanism for cross-talk among G protein pathways, most simply by allowing G␤␥ from one G protein to regulate a GAP that is modulating the activity or kinetics of another. For example, inhibition of G q GAP activity by G␤␥ provides a plausible mechanism for the G i -mediated potentiation of Ca 2ϩ signaling by G q , an effect that is hard The responses to G␤␥ of wild type PLC-␤1 (Ⅺ) and the ⌬141 mutant (f), which lacks the N-terminal PH domain, were compared in several assays. A, inhibition of steady-state G q GAP activity measured with M1AChR-G q vesicles in the presence of carbachol. G␤␥ was added at the indicated concentrations. B, phospholipase C activity was assayed with the soluble enzyme and micellar substrate at increasing concentrations of G␤␥. C, phospholipase activity on the micellar substrate was measured at increasing concentrations of GTP␥S-activated G␣ q . These data are representative of two or three replications of each experiment. In all cases, both the phospholipase and GAP-specific activities of the mutant are somewhat lower than that of wild-type PLC-␤1, probably because of diminished stability. to rationalize by other known mechanisms of G i action. Most G i effects reflect the action of G␤␥ rather than of G␣ i , at least in part because G i (and G o ) is expressed at high levels compared with other G proteins and releases its G␤␥ subunits more readily (5,8,47). Signal potentiation by G i /G o may frequently reflect inhibition of GAPs acting on the G q family, G z or G 12/13 . In addition to potentiating other signals, GAP inhibition by G␤␥ may allow one G protein to prolong the duration of a signal conveyed by another. GAPs, particularly the RGS proteins, thus form a large new group of candidate G␤␥-regulated effectors.
Whether free G␤␥ actually is found in the plasma membrane upon G protein stimulation and what its concentration might be are unknown and currently not measurable. Recent studies by Bünemann et al. (9) suggest that complete dissociation of G␤␥ from G␣ i is not favored in native plasma membranes during stimulation by agonist, although G␤␥ signaling was observed. Dissociation of G␣ o from G␤␥ was observed under similar conditions (8). G␣ o binds G␤␥ less tightly than does G␣ i (47). G␣ i -GTP and G␤␥ apparently also remain associated during receptor-promoted stimulation of K ϩ channels by G␤␥ (10,11), and PLC-␤ can be stimulated by both G␣ q and G␤␥ (37). It is thus likely that conformational isomerization in a G␣␤␥ heterotrimer is sufficient to allow signaling by G␤␥ without complete dissociation. Because activation decreases affinity of G␣ for G␤␥, it is also likely that the rate of G␤␥ dissociation also increases, allowing G␤␥ a kinetic window for G␣ exchange. The possibility that G␤␥ interacts with multiple partners simultaneously, or changes partners in a multiprotein complex, is thus plausible, but its reality awaits direct physical measurement. Nevertheless, G␤␥ inhibits GAP activity and stimulates known cellular effectors over the same concentration range, which argues that GAP inhibition can occur physiologically regardless of whether completely free G␤␥ is the active agent. More substantial evidence will depend on the development of intracellular assays of GAP interactions with G␣-GTP substrates.
Competition between G␤␥ and GAPs-Both steady-state kinetics and protein-protein binding data indicate that G␤␥ inhibits GAP activity competitively, competing with GAP for binding to G␣-GTP and protecting it from GAP-accelerated deactivation. Initial indication for a competitive mechanism came from the predominant increase in K m caused by G␤␥ with little if any effect on V max . This mechanism is further supported by the ability of G␤␥ to increase the EC 50 of RGS proteins in GAP assays (Fig. 4) and the ability of GAPs to increase the IC 50 of G␤␥ (Fig. 5). Formally, such behavior would also be consistent with negatively cooperative binding interactions; G␤␥ might bind either G␣-GTP or GAP (or both) and decrease its affinity for the other. The dependence of K m on inhibitor is commonly used to differentiate these mechanisms, but we were unable to measure K m at high enough concentrations to observe its saturation at high concentrations of G␤␥. However, the absence of an effect of G␤␥ on the rate of dissociation of the RGS4-G␣ i complex (Fig. 8B) lends support to a competitive mechanism. Finally, G␤␥ inhibits the GAP activity of the ⌬141 truncation mutation of PLC-␤1, which lacks the known G␤␥ binding site. If inhibition derived from negative cooperativity, G␤␥ would have to bind to a yet unknown site on the PLC-␤ molecule.
Competitive inhibition depends on the surprisingly high-affinity binding of G␤␥ to the GTP-activated form of G␣. The data in Fig. 8 are the first direct measurement of the affinity of the active form of a G␣ subunit for G␤␥ and indicate that affinity is much higher than was previously supposed. The affinities of G␤␥ for G␣ i -GTP␥S and G␣ i -GDP-AlF 4 are very similar and therefore probably reflect the affinity for the GTP-bound form as well. High-affinity binding of G␣-GTP and G␤␥ is also supported by the observations that G␣ and G␤␥ may remain bound during the receptor-stimulated GTPase cycle (48,49). The lifetime of the G␣-GTP-G␤␥ complex ( ϭ 1/k d ) is thus significantly longer than the lifetime of the GTP-activated state of the G␣ subunit ( ϭ 1/k hydrol ).
Structural Basis of Competition between G␤␥ and GAPs-At one level, competition between GAPs and G␤␥ for binding to G␣-GTP is predictable because both RGS proteins and G␤␥ bind to the same surface of the G␣ subunit, which includes the switch regions (31,32,50). Binding here is thought to mediate both the GAP activity of RGS proteins and the GDP dissociation inhibitor activity of G␤␥. The second surface in G␣ to make contact with both G␤␥ and RGS proteins is the N-terminal helix (22,32,51). However, the structure of this part of the G␣-RGS interface is unknown. The only available crystallographic structure of a G␣-RGS protein complex shows the G␣ N-terminal helix making crystal contact with an RGS protein in the adjacent unit cell (31), most likely as a crystallization artifact. Little structural information is available about the binding of G␣ q -GTP to PLC-␤. The C-terminal region of PLC-␤ is required both for stimulation of phospholipase activity by G␣ q and for G q GAP action, but the precise interface remains speculative (38 -40). Again, more than one binding site may be involved.
A complication in interpreting competition between G␤␥ and GAPs for G␣-GTP is that crystallographic data on these interfaces were obtained with the GDP/AlF 4 -bound form of G␣ for the G␣-RGS complex and with the GDP-bound form for the G␣␤␥ trimer. Although the structure of the G␣ globular domain is fundamentally the same in both cases, the orientations of the N-terminal helix are markedly different. Indeed, the N-terminal region of G␣ subunits are markedly flexible, although mostly helical when determined by x-ray crystallography. During the GTPase catalytic cycle, the conformation of G␣ presumably oscillates between the active and inactive conformation several times per second. It is unclear in this case that G␣ ever assumes either canonical conformation. It is plausible to propose an oscillating, or pivoting, binding mode in which G␣ may remain bound to both GAP and G␤␥ during the GTPase cycle. Binding might be maintained via the G␣ N-terminal helix, with the switch regions on the globular domain alternately interacting with GAP and G␤␥. This mechanism would remove protein association/dissociation reactions from the rapid GTPase cycle and thus accelerate cycle transit. It might also permit quasi-continuous activity of G␤␥ as long as receptor-catalyzed GTP binding was sufficiently rapid. How it would impact on the rate of GAP-accelerated GTP hydrolysis is uncertain. Mutation of G␣ that differentially alters its affinity for either GAP or G␤␥ may help address these questions, as will dynamic studies of the conformations and orientations of G␣ during the GTPase cycle.