A GTPase-activating Protein for the G Protein Ga z

A GTPase-activating protein (GAP) specific for Gαz was identified in brain, spleen, retina, platelet, C6 glioma cells, and several other tissues and cells. Gz GAP from bovine brain is a membrane protein that is refractory to solubilization with most detergents but was solubilized with warm Triton X-100 and purified up to 50,000-fold. Activity is associated with at least two separate proteins of Mr ∼22,000 and 28,000, both of which have similar specific activities. In an assay that measures the rate of hydrolysis of GTP pre-bound to detergent-soluble Gαz, the GAP accelerates hydrolysis over 200-fold, from 0.014 to 3 min −1 at 15°C, or to ≥20 min−1 at 30°C. It does not alter rates of nucleotide association or dissociation. When co-reconstituted into phospholipid vesicles with trimeric Gz and m2 muscarinic receptor, Gz GAP accelerates agonist-stimulated steady-state GTP hydrolysis as predicted by its effect on the hydrolytic reaction. In the single turnover assay, the Km of the GAP for Gαz-GTP is 2 nM. Its activity is inhibited by Gαz-guanosine 5′-O-thiotriphosphate (Gαz-GTPγS) or by Gαz-GDP/AlF4 with Ki ∼1.5 nM for both species; Gαz-GDP does not inhibit. G protein βγ subunits inhibit Gz GAP activity, apparently by forming a GTP-Gαzβγ complex that is a poor GAP substrate. Gz GAP displays little GAP activity toward Gαi1 or Gαo, but its activity with Gαz is competitively inhibited by both Gαi1 and Gαo at nanomolar concentrations when they are bound to GTPγS but not to GDP. Neither phospholipase C-β1 (a Gq GAP) nor several adenylyl cyclase isoforms display Gz GAP activity.

G proteins mediate numerous cellular processes by traversing a cycle of GTP binding and hydrolysis. Bound GTP activates a G protein such that it can stimulate a downstream effector protein. Activation is terminated when the bound GTP is hydrolyzed to GDP, which does not activate. Each step of the cycle is controlled, such that both steady-state GTPase activity and the concentrations of the active and inactive forms are highly regulated.
Activation of heterotrimeric G proteins is promoted by sevenspan cell-surface receptors that facilitate GDP release and GTP binding. Small monomeric G proteins (Ras, Rac, Rho, Arf, etc.) are activated by cytosolic proteins that similarly facilitate GTP binding.
In many cases, hydrolysis of bound GTP, the deactivation step, is accelerated by GTPase-activating proteins, or GAPs 1 (1)(2)(3)(4). GAPs appear to fulfill at least one of four definable roles. Some GAPs for monomeric signaling G proteins, such as Ras GAP, appear to attenuate G protein signal amplitude in response to inputs from inhibitory signaling pathways (1, 2, for review). GAPs for the monomeric G proteins involved in cytoplasmic vesicle trafficking are thought to act by terminating a G protein-dependent assembly or transit step. Some effector proteins that are regulated by heterotrimeric G proteins also act as GAPs for their G protein regulators. The GAP activities of these effectors, such as phospholipase C-␤ (5, 6) and the cyclic GMP phosphodiesterase ␥ subunit (7,8), may allow effector-specific modulation of response times or may enhance the selectivity of receptor-G protein signaling (3). A fourth class of GAPs, also for the trimeric G proteins, includes members of the recently identified RGS protein family (4, 9 -13). Little is known of the physiology of RGS proteins, but they can contribute to desensitization toward a prolonged signal (Sst2p in yeast; Refs. 10,14,15) or act as long term attenuators of signal amplitude (Egl-10 protein in Caenorhabditis elegans; Ref. 9). G protein GAP activity can potentially be used to identify and purify regulators of G protein function or to point to novel inputs to G protein signaling pathways. For GAPs that are also effectors, their identification can indicate what downstream signals the G protein mediates.
We began a search for new GAPs for heterotrimeric G proteins by looking for a GAP for G z , a pertussis toxin-insensitive member of the G i family that is abundant in brain, adrenal medulla, and platelets (16 -19). There were three reasons for this choice. Although G z can mediate inhibition of adenylyl cyclase (20 -22) and respond to receptors that regulate other G i family members (21,23), the signaling pathway(s) mediated by G z remains unknown, and a G z GAP might be an effector. Second, G z hydrolyzes bound GTP very slowly (16). Its activation lifetime is about 7 min, which seems incompatible with normal signaling functions unless a GAP accelerates deactivation. Last, the slow hydrolytic rate simplifies design of an assay for a GAP. We describe here the detection of G z GAP activity in brain and other tissues, substantial purification of a G z GAP from bovine brain, and several aspects of its mechanism of action and regulation.
Cholic acid was purified as described (29), and other detergents were purchased from various suppliers. [␥-32 P]GTP was either purchased or synthesized (30) and purified as described (6).
Hydrolysis of G␣ z -bound [␥-32 P]GTP in routine GAP assays was measured by incubation of the substrate (usually ϳ1-2 nM) in the buffer used for its preparation but including 1 mM free Mg 2ϩ , 10 g/ml bovine albumin, and 5 mM nonradioactive GTP. Unlabeled GTP was added to inhibit nucleoside triphosphatase activity present in crude GAP fractions in the event that any [␥-32 P]GTP dissociated from G␣ z . Assays were carried out at 15°C for times that varied from 30 s to 60 min. Hydrolysis of bound [␥-32 P]GTP was measured as release of [ 32 P]orthophosphate (31). Hydrolysis followed a single exponential time course. Hydrolysis is expressed either as the amount of bound GTP hydrolyzed at early times (quasilinear time course) or as a first-order rate constant. 2 GAP-independent hydrolysis is subtracted from all data except in Figs. 1, 6, and 8B.
Other GTPase Assays-Steady-state GTPase assays were performed as described (31) at 30°C in buffer that contained 20 mM NaHepes (pH 8.0), 25 mM NaCl, 0.1 mM EDTA, 1.1 mM MgCl 2 , 20 g/ml bovine albumin, 50 nM GDP, 100 nM [␥-32 P]GTP, and either 100 M carbachol or 5 M atropine. The rate of hydrolysis of [␥-32 P]GTP bound to G␣ o and G␣ i1 was measured at 15°C essentially as described previously (26), except that the buffer for the initial [␥-32 P]GTP binding reaction contained 5 mM EDTA, and hydrolysis was initiated by adding 6.0 mM MgCl 2 plus 0.1 mM unlabeled GTP. The concentration of G␣-bound [␥-32 P]GTP at initiation of the hydrolysis reaction is given as the total amount of [ 32 P]orthophosphate released during the reaction.
Purification of G z GAP-G z GAP was purified from bovine cerebral cortical membranes prepared according to Sternweis and Pang (32). All procedures were performed at 0 -4°C except where explicitly noted. In practice, we pool active fractions from multiple runs of the Q-Sepharose column before gel filtration and pool several gel filtration peaks for G␣ z affinity chromatography. Combining preparations in this way improves both the yield and purification. However, Table II is an example of an early preparation that was completed exactly as described below.
Membranes were washed once in 50 mM Tris-Cl (pH 7.5), 1 mM DTT, 1 mM EDTA, 0.5 M NaCl, and 0.1 mM PMSF and resuspended to 2.5 mg/ml in 20 mM NaHepes (pH 7.5), 0.1 mM EDTA, 0.3 mM PMSF, and 2% Triton X-100. The suspension was sloshed for 25 min at 30°C and centrifuged for 40 min at 150,000 ϫ g. The supernatant was loaded on a DEAE-Sephacel column (700 ml) that was equilibrated with Buffer 1 (20 mM NaHepes (pH 7.5), 0.1 mM EDTA, 1% Triton X-100, 0.1 mM PMSF). The column was washed with Buffer 1 and eluted with a gradient of 0 -300 mM NaCl in Buffer 1. Both G z GAP activity and protein were eluted in parallel as a broad, asymmetric peak. Active fractions were pooled and diluted 3-fold with Buffer 2 (20 mM NaHepes (pH 7.5), 0.1 mM EDTA, 1 mM DTT, 1% cholate, 0.1 mM PMSF). DTT was added to a final concentration of 20 mM, and the solution was applied to a column of Q-Sepharose that had been equilibrated with Buffer 2. The column was washed sequentially with Buffer 2, Buffer 2 plus 0.1 M NaCl, and Buffer 2 plus 0.25 M NaCl, and then eluted with a gradient of 0.25-0.55 M NaCl in Buffer 2. Protein and GAP activity again eluted as a broad peak. Active fractions were concentrated on an Amicon PM30 membrane and chromatographed on a column of Ultrogel AcA-34 equilibrated with Buffer 2 plus 0.1 M NaCl. A typical elution profile is shown in Fig. 2. The second peak of GAP activity was pooled and concentrated by adsorption to Mono Q and elution with a gradient of 0.1-0.55 M NaCl in Buffer 2.
Standard procedures were used for SDS-gel electrophoresis (34) and staining with Coomassie Blue or silver (35). For samples in which GAP activity was to be measured after electrophoresis, samples were denatured in sample buffer (34) that contained 1% SDS and 10 mM DTT. G z GAP activity was extracted from slices of SDS-polyacrylamide gels and renatured by homogenizing the gel in 5-10 volumes of renaturation buffer (20 mM NaHepes (pH 7.5), 1 mM DTT, 0.1 mM EDTA, 1% Triton X-100) and shaking overnight at 0°C.
The concentration of free Mg 2ϩ in assay buffers that contained significant concentrations of both EDTA and GTP was calibrated as described by Huskens and Sherry (36).

RESULTS
Identification of G z -GAP Activity-We used the G␣ z -[␥-32 P]GTP complex as substrate to test for the presence of proteins that can increase the rate of hydrolysis of G␣ z -bound GTP. Purified G␣ z hydrolyzes bound GTP very slowly (16), such that G␣ z -[␥-32 P]GTP can be prepared and purified with good yield. About 25% of total G␣ z is bound to [␥-32 P]GTP after gel filtration, with the rest bound to unlabeled GDP. Under standard assay conditions at 15°C, GTP bound to G␣ z is hydrolyzed with a rate constant, k hydrol , of about 0.014 min Ϫ1 , which corresponds to a t1 ⁄2 of about 50 min (Fig. 1).
Addition of a crude membrane fraction from bovine brain increased the rate of hydrolysis of G␣ z -bound GTP dramatically (Fig. 1A). Hydrolysis was a single component, first-order reaction over a 10-fold range of rates, and rate constants increased linearly with increasing amounts of membrane protein (Fig.  1B). These data indicate the existence of a GTPase-accelerating activity in bovine brain, i.e. a G z GAP. Based on the linearity of GTP hydrolysis with added membrane protein, we defined a unit of G z GAP activity as an increment in the hydrolytic rate constant of 1.0 min Ϫ1 . Both the basal rate of GTP hydrolysis by G␣ z and the GAP activities of several tissues were quite reproducible in this assay. The hydrolysis-accelerating activity is evidently that of a protein. In membranes, activity was destroyed by incubation with 6.7 g/ml trypsin, 20 g/ml chymotrypsin plus detergent or 0.5 mM N-ethylmaleimide. Added detergent markedly sensitized the GAP to proteolysis.
Release of [ 32 P]orthophosphate in GAP assays such as shown Fig. 1  Homogenates of several mammalian tissues and cultured cells were screened for G z GAP activity (Table I), and its distribution was found to be similar to that of G␣ z itself (16 -19; confirmed qualitatively by anti-␣ z immunoblot durning this study). Activity was highest in cerebral cortex, although there was considerable activity in membranes of spleen and retina. Peripheral fat, lung, platelets, and testis also displayed readily measurable activity. Several other tissues displayed low activity, which may represent contamination by adipose, neuronal, or vascular tissue or the action of other GAPs with low activity toward G z . We do not know how many proteins in these tissues display G z GAP activity. Among the cultured cells tested, C6 glioma cells displayed activity similar to that of brain, 20 -50 units/mg depending upon the source of the cells and the culture conditions. Several other cell lines displayed more modest activities. G z GAP activity was low in S49 murine lymphoma cells and was barely detected in Sf9 cells. Because activity was highest in brain and bovine brain is readily obtainable, we concentrated on this source and have not attempted to determine the multiplicity of G z GAPs in other tissues.
Purification of G z GAP-There is no soluble G z GAP activity detectable in brain homogenates (Ͻ2%). G z GAP behaves as an integral membrane protein and is difficult to solubilize. It is not extracted by washing at high or low ionic strength and is not solubilized by many detergents. At 0°C, neither cholate, deoxycholate, Lubrol PX, Triton X-100, CHAPS, digitonin, nor several other detergents solubilized any G z GAP activity. Small and irreproducible amounts of activity were solubilized by dodecyl maltoside and lauroyl sucrose. Fortunately, incubation of membranes with 2% Triton X-100 at 30°C released considerable G z GAP activity into a 200,000 ϫ g supernatant. This apparently soluble GAP was still badly aggregated, however, and substantial increases in specific activity were not obtained by chromatography in multiple systems. Two sequential rounds of anion exchange chromatography, which included transfer from Triton X-100 to cholate, provided little purification and substantial loss of activity (Table II) but did allow subsequent purification. Although about half of the G z GAP activity remained aggregated, the other half behaved as an apparently monodisperse species of reasonable molecular size and was thus purified about 20-fold by gel filtration in cholate (Fig. 2). This soluble material was used for further purification.
After gel filtration, G z GAP activity was appreciably purified by affinity chromatography on G␣ z -agarose. GAP activity bound to the column when the covalently coupled G␣ z was activated with either GTP␥S or with GDP/AlF 4 but not when the G␣ z was in the nonactivated, GDP-bound form (not shown). For purification, peak fractions from gel filtration were applied to Al 3ϩ /F 4 Ϫ -activated G␣ z -agarose, and after extensive washing, GAP activity was eluted by removal of Al 3ϩ , F Ϫ , and Mg 2ϩ and increasing the concentration of detergent and salt.
Further purification of G z GAP was achieved by phenyl-Sepharose chromatography (Fig. 3). G z GAP activity was consistently eluted from phenyl-Sepharose in two peaks. This was true for multiple elution protocols that included increasing the FIG. 1. G z GAP activity in bovine brain membranes. A, G␣ z -[␥-32 P]GTP was incubated at 15°C with increasing amounts of bovine brain membranes (छ, 5 g; E, 10 g; Ç, 15 g; É, 30 g; Ⅺ, 15 g of boiled membrane). Release of [ 32 P]orthophosphate was assayed at the times shown. Data are normalized to the total amount of G␣ z -[␥-32 P]GTP at zero time (125 fmol) determined by nitrocellulose binding assay. At each time point, samples were also assayed for bound [␥-32 P]GTP (not shown). The sum of [ 32 P]orthophosphate plus bound [␥-32 P]GTP was constant over 120 min and not altered by the presence of the membranes. Hydrolysis was not altered by substitution of buffer for boiled membrane protein (not shown). In this early experiment, the assay buffer contained only 1 M free Mg 2ϩ and 0.1% Lubrol in place of 0.1% Triton X-100, resulting in the relatively low activities shown. B, each data set from A was well fit by a single component first-order reaction scheme (not shown). Values of the rate constant k app 2 obtained from least-squares fits of these data are plotted against the amount of membrane protein added. concentration of detergent and/or decreasing the concentration of salt, although it was possible to alter the peak shapes and the total and specific activities in each peak. Typically, both peaks were purified about 5000 -12,000-fold relative to the Triton extract (Table II shows data for pools). Neither peak contained homogeneous protein, as shown by SDS-gel electrophoresis (Fig. 3B). The peak fractions contained a large but reproducible group of proteins, the only identifiable one of which was the G protein ␤ subunit. G␤␥ has no GAP activity (see below), and GAP activity was not co-immunoprecipitated by anti-G␤ antibodies (not shown). Its appearance is apparently an artifact of G␣ z affinity chromatography. No single polypeptide obviously co-fractionated exactly with G z GAP activity. The phenyl-Sepharose pools contained no GTP␥S binding activity (Ͻ0.015 mol/mol GAP). We attempted to identify the polypeptide(s) that accounts for G z GAP activity by further fractionating phenyl-Sepharose peak fractions on SDS-polyacrylamide gels and then renaturing the GAP protein eluted from individual gel slices (see "Experimental Procedures"). About 50 -75% of G z GAP activity was recovered from the gel. As shown in Fig. 4, activity was broadly distributed on SDS gels between 20 and 30 kDa, with two peaks of activity reproducibly appearing at about 22 and 28 kDa. Smearing during electrophoresis is an unlikely cause of this behavior because both activity and discrete protein bands eluted from individual gel slices retain their distinct electrophoretic mobilities upon extraction and a second round of gel electrophoresis (Fig. 4B). The 22-kDa peak of activity corresponds to a relatively blurry band, but we have not assigned the 28-kDa activity to a specific stained band. In addition to the two peaks of activity, G z GAP activity is readily detected throughout the region between 22 and 28 kDa, indicating considerable heterogeneity. It is unclear whether the 22-kDa band and intermediate forms are proteolytic products of the 28-kDa form or whether they are unique G z GAP species. SDS gel analysis of GAP activity in membranes and earlier fractions during the purification have not clarified this question, although they indicate the presence of GAP activity with molecular size up to 40 kDa. Thus, the data of Fig. 4 indicate that G z GAP activity results from monomeric proteins in the size range 22-28 kDa, but the number of species remains uncertain.
In addition to indicating which molecular weight species contribute to G z GAP activity, SDS-gel electrophoresis and renaturation provide substantial purification of GAP activity. The specific activity of G z GAP extracted from SDS gels is increased more than 50,000-fold relative to the Triton extract. It is likely that these fractions are essentially pure. Such extensive purification indicates that G z GAP is of low abundance even in brain.
Mechanism of Action of G z GAP-A single GAP molecule can turn over multiple molecules of G␣ z -GTP (Fig. 5). Its behavior is most readily analyzed when it is considered as an enzyme that acts upon the substrate G␣ z -GTP and converts it to the products G␣ z -GDP plus orthophosphate. Its K m for G␣ z -GTP is about 2 nM, which represents sufficiently high affinity binding to be physiologically reasonable. The maximum GAP-stimu-  2. Ultrogel AcA-34 chromatography of G z GAP. Pooled fractions from Q-Sepharose were concentrated by ultrafiltration and chromatographed on Ultrogel AcA-34 as described under "Experimental Procedures." The first peak is at the void volume.
FIG. 3. Phenyl-Sepharose chromatography of G z GAP. Pooled affinity chromatography fractions were concentrated by adsorption to Q-Sepharose and chromatographed on phenyl-Sepharose essentially as described under "Experimental Procedures." A, elution profile. Fraction 1 begins after the high salt wash. Flow-through (FT) and wash fractions accounted for 19% of the activity and 46% of the protein originally applied. About 40% of applied protein was not eluted from the column, but all GAP activity was accounted for in the eluted fractions. This profile is from a micro-scale fractionation performed during development of the method, and the profile of the step gradient was modified subsequently. B, samples of each fraction (8.3% of total), of the load and flow-through (0.37%), and of the wash (0.83%) were analyzed by SDSgel electrophoresis followed by silver staining. Positions of molecular weight markers are shown at the left. lated hydrolysis rate can be estimated two ways. Because velocity increases linearly with the amount of GAP at low GAP concentrations (Figs. 1 and 5B), the maximum GAP-stimulated hydrolytic rate constant (k gap 2 ) can be calculated by dividing V max by the molar concentration of GAP, which is calculated according to its estimated purity and approximate molecular weight. This calculation yields a k gap of 3 min Ϫ1 at 15°C. The other estimate of k gap derives from a titration of GAP when the concentration of G␣ z -GTP is maintained at or above the K m (Fig. 5B). The maximum in such an experiment, 1.8 min Ϫ1 , can be corrected for the subsaturating concentration of G␣ z -GTP to yield a true maximum k gap of 3.1 min Ϫ1 . Thus, both determinations of k gap are about 3 min Ϫ1 , more than a 200-fold stimulation over k hydrol for G␣ z -GTP. We estimate that k gap at 30°C is over 20 min Ϫ1 , which corresponds to an average lifetime for activated G z of Ͻ2 s. Thus, because k gap is fast and because G z GAP has a high affinity for its G␣ z -GTP substrate (K m ϳ2 nM), its action is sufficient to allow G z -mediated signal transduction with physiologically appropriate rates.
G␣ z can hydrolyze bound GTP at very low concentrations of free Mg 2ϩ , and its intrinsic k hydrol is independent of the concentration of Mg 2ϩ up to 10 mM (Fig. 6). In contrast, G z GAP activity displays a marked Mg 2ϩ optimum at ϳ1 mM. Although the GAP is active over a wide range of Mg 2ϩ concentrations, stimulation is ϳ4-fold higher at the optimum. Neither Ca 2ϩ nor Mn 2ϩ exerted a unique regulatory effect on G z GAP, although either can replace Mg 2ϩ over approximately the same range of concentrations (not shown). G z GAP binds tightly to G␣ z , but only in its GTP-activated form. Both G␣ z -GTP␥S and G␣ z -GDP/AlF 4 thus inhibited G z GAP activity, with a K i of ϳ1.5 nM for either nucleotide (Fig. 7). This value of K i is similar to the K m for the substrate G␣ z -GTP, which suggests that G␣ z -GTP, G␣ z -GTP␥S, and G␣ z -GDP/AlF 4 all bind the GAP at a common site and with similar affinities. G␣ z -GDP did not inhibit at concentrations up to 20 nM. Note that in the experiment of Fig. 7, G␣ z -GDP/AlF 4 was formed in the reaction mixture by the addition of Al 3ϩ plus F Ϫ , which appear to inhibit GAP activity by themselves. This inhibition results from the presence of G␣ z -GDP in the preparation of G␣ z -GTP substrate, such that about 6 nM G␣ z -GDP/AlF 4 is present in the assay even when no excess G␣ z -GDP was added. Selectivity of the GAP for the active conformation of G␣ z is confirmed by the selective binding of the GAP to G␣ z -agarose when it is activated by either GTP␥S or GDP/AlF 4 .

FIG. 4. SDS-polyacrylamide gel electrophoresis of partially purified G z GAP.
A sample of the second peak from phenyl-Sepharose chromatography (fraction 6 in Fig. 3) was diluted into SDS sample buffer and electrophoresed on a 15% polyacrylamide gel. The gel lane was sliced, and protein was eluted from each slice as described under "Experimental Procedures." A, G z GAP activity was assayed in the eluate from each slice. B, an aliquot of each eluate (and of phenyl-Sepharose fraction 6, denoted c) was denatured in SDS and electrophoresed again. The gel was silver-stained. The molecular mass markers were carbonic anhydrase (29 kDa) and ␤-lactoglobulin (18.4 kDa).

FIG. 5. GAP activity at increasing concentrations of G z GAP and of G␣ z -[␥-32 P]GTP.
A, GAP activity was assayed at the concentrations of G␣ z -[␥-32 P]GTP shown on the abscissa, as determined in a [␥-32 P]GTP binding assay. The concentration of GAP was 30 pM (7.5 ng/ml, ϳ10% pure; M r ϳ25,000; purified through affinity chromatography). Inset, Hanes replot. B, GAP activity was assayed at the GAP concentrations shown on the abscissa using 2.5 nM G␣ z -[␥-32 P]GTP as substrate. Assay times in both experiments were adjusted to obtain accurate measurements of either the initial reaction rate or of the first-order rate constant k app .
The following small molecules had no effect upon G z GAP activity: inositol trisphosphate, cyclic AMP, cyclic GMP, GTP, GTP␥S, and ATP (not shown).
Selectivity of G z GAP-The selectivity of brain G z GAP among different G␣ subunits was tested initially by comparing their abilities to compete with G␣ z -GTP in the standard assay (Fig. 7, Table III). Both myristoylated, recombinant G␣ i1 and bovine brain G␣ o inhibited competitively, G␣ o -GTP␥S with a K i of ϳ5 nM (Fig. 7) and G␣ i1 -GTP␥S with a K i of about 20 nM.
Their GDP-bound forms did not inhibit (not shown). Inhibition by G␣ o or G␣ i required that they be myristoylated; nonmyristoylated G␣ subunits inhibited weakly or not at all. Based on these data, we measured the ability of the G z GAP to accelerate hydrolysis of G␣ o -GTP and G␣ i1 -GTP, using the single turnover assay of Higashijima et al. (26) to accommodate the faster basal hydrolytic rates of these G␣ subunits. Although both G␣ o and G␣ i1 hydrolyze bound GTP much faster than does G␣ z , the relative effect of G z GAP on both G␣ subunits was minimal when compared with G␣ z (Table IV): 30% and 7% compared with more than a 30-fold effect on G␣ z . The GTP␥S-bound forms of G␣ s , G␣ q , and G␣ 12 did not compete significantly in the G z GAP assay (Table III), and their activities as GAP substrates were not tested.
Regulation of G z GAP by G Protein ␤␥ Subunits-Although G protein ␤␥ subunits had little if any effect on the rate of hydrolysis of G␣ z -bound GTP, G␤␥ inhibited G z GAP activity up to 80% (Fig. 8). Inhibition was most marked at low concentrations of G␣ z and is caused by an increase in K m of at least 5-fold (Fig. 8A). No effect of G␤␥ on V max was detected, but we were unable to achieve saturation with G␣ z -GTP at high G␤␥ concentrations, and we may have failed to observe a slight decrease in V max . G z GAP was inhibited approximately equally by G␤ 1 ␥ 2 , G␤ 2 ␥ 2 , and G␤ 2 ␥ 3 (not shown). The increase in K m caused by G␤␥ apparently reflects formation of the GTP-bound G␣ z ␤␥ heterotrimer. The IC 50 for G␤ 1 ␥ 2 (ϳ400 nM, Fig. 8B) agrees well with its affinity for GTP-bound G␣ z (37), and we have found no evidence for G␤␥ binding directly to the GAP (which would yield classical competitive inhibition). G␣ z ␤␥-GTP may be a low affinity (high K m ) GAP substrate or it may simply block GAP binding to G␣ z -GTP. These alternatives are potentially distinguishable according to the dependence of the apparent K m on the concentration of G␤␥, but we have been unable to determine K m accurately over a high enough range of G␤␥ concentrations to answer this question.
In addition to inhibiting the GAP, G␤␥ decreased the rate of dissociation of GDP, but not GTP, from G␣ z , as is true for other G␣ subunits (38). We observed no other effects of G␤␥ in this system.
G z GAP Activity During Receptor-stimulated Steady-state GTP Hydrolysis-To study the effect of G z GAP on the receptorstimulated steady-state GTPase cycle, we co-reconstituted G z GAP with m2 muscarinic cholinergic receptor and heterotrimeric G z into unilamellar phospholipid vesicles. When the muscarinic agonist carbachol was added to promote receptor-catalyzed exchange (23), G z GAP increased the steady-state GTPase rate by about 2.5-fold (Fig. 9A). This effect seems small in comparison to the 200-fold maximum effect of the GAP on FIG. 6. Effect of Mg 2؉ on G z GAP activity. G z GAP-stimulated (q) and basal (E) GTP hydrolysis by G␣ z were assayed at the concentrations of free Mg 2ϩ shown on the abscissa. The GAP-stimulated data are the apparent hydrolytic rate constants, k app , i.e. basal hydrolysis was not subtracted from that measured in the presence of GAP. The free Mg 2ϩ concentrations shown on the abscissa were obtained by varying the concentration of MgCl 2 in the presence of 1 mM EDTA and 5 mM GTP (see "Experimental Procedures") and were measured as described by Huskens and Sherry (36). The concentration of G␣ z -[␥-32 P]GTP was 1.5 nM, and the concentration of GAP was 50 pM.   (25). 100% activity was 60 milliunits. GDP-bound forms of these G␣ subunits did not inhibit (not shown). G␣ GAP activity hydrolysis of preformed G␣ z -GTP, but the steady-state concentration of G␣ z -GTP in the vesicles is in significant molar excess over that of the GAP. The effect of the GAP on the steady-state GTPase rate is consistent with its observed activity in the single turnover assay. The effect of GAP on steady-state GTPase rates is evidently exerted only at the hydrolytic step.
In the absence of agonist, where steady-state GTPase activity is limited by the GDP/GTP exchange rate, GAP had no effect, and GAP also had no effect on the rates of nucleotide binding (Fig. 9B) or release (not shown). Consistent with its effect on hydrolysis of bound GTP, G z GAP decreased the steady-state concentration of the active G z -GTP complex during stimulation by agonist. When assayed at either 5 or 15 min in the system shown in Fig. 9A, the addition of G z GAP decreased the concentration of G z -GTP by about 30%. Both of these sample times are after steady-state was reached, as indicated by the constant concentration of bound GTP in this interval and by the completion of agoniststimulated GTP␥S binding in about 2 min (Fig. 9B). Simple kinetic models predict that the relative effect of GAP on the accumulation of G z -GTP would be greater in the absence of agonist, but we were unable to measure accurately the small amount of binding of [␥-32 P]GTP binding that occurred without agonist.
The results of the experiments shown in Fig. 9 indicate that purified G z GAP can reassociate with membrane lipids and regulate G z appropriately in (or on the surface of) a phospholipid bilayer. The addition of detergent-soluble G z GAP to preformed vesicles had no effect on the steady-state GTPase rate (not shown), which is consistent with the idea that the GAP is an integral membrane protein. DISCUSSION G z hydrolyzes bound GTP (deactivates) extremely slowly. Although it is activated at a normal rate in response to receptors (23), its active state decays with an average lifetime of about 7 min at physiological temperature (16). With these kinetics, it would be hard to understand how G z can mediate signaling responses in a reasonable way, although it clearly does.
The data presented here describe the identification of a GAP for G z in brain membranes, its purification, and mechanistic behavior. Similar activity was identified in membranes of several other tissues and cultured cells. By accelerating GTP hydrolysis, a G z GAP reconciles aberrant deactivation kinetics with normal signaling functions. However, its precise role in signaling physiology remains unclear. G z GAP may be a G zregulated effector protein, in analogy with phospholipase C-␤ and cyclic GMP phosphodiesterase. These effectors are both regulated by G proteins and have GAP activity specific for their G protein regulators, G q and G t (5, 7, 8). The low K m of the G z GAP, about 2 nM, is in the same range of affinities as that displayed by G␣ q , G␣ s , or G␣ t for their effectors (5,6,39,40), and its selectivity for the activated form of G␣ z is also consistent with this role. Alternatively, the GAP may be a negative regulatory component of the G z pathway, involved either in desensitization or in mediating negative input from another signaling pathway. The model for such regulation could be either the GAPs for p21 RAS and related small, monomeric G proteins (2) or the RGS proteins, a large family of related proteins that inhibit signaling (4,9,10) and whose prototypes are GAPs (12). Whether the cerebral G z GAP is an effector or a modulator of inhibition will probably be elucidated when its cDNA can be used to manipulate its expression in cells.
Purification-G z GAP was initially purified about 12,000fold according to the specific activity of phenyl-Sepharose fractions, and the specific activity of the purest fractions from SDS gels is about 4-fold higher (Fig. 4). G z GAP is thus a rare protein in brain, its richest source, and is perhaps expressed in  Hydrolytic rate constants were determined at 15°C as described under "Experimental Procedures," either in the presence or absence of 0.5 nM purified bovine brain G z GAP. Concentrations of substrate were 2.7 nM G␣ z -GTP, 5.5 nM G␣ o -GTP, and 35 nM G␣ i1 -GTP (about the K m for G␣ z and the K i for G␣ o and G␣ i1 ). Data are the average of duplicate determinations in two separate assays. The results for G␣ o were confirmed qualitatively at 20°C. G␣ o was purified from bovine brain, and G␣ i1 was the recombinant myristoylated form purified from E. coli (25). only a few cell types. By rough comparison with published data (16,41) and with immunoblots performed during this study (not shown), G z GAP is about 5-10-fold less abundant than is G␣ z . This is not surprising, however, whether the GAP is an effector or purely a negative regulator. G proteins are generally in molar excess over their effector proteins. Alternatively, because G z GAP acts catalytically, it could readily function as an efficient inhibitor of G z signaling. Despite extensive purification, preparations of G z GAP remain heterogeneous. GAP activity in peak fractions from phenyl-Sepharose is distributed bimodally between 22 and 28 kDa. We do not know if the 22-kDa GAP is distinct from the 28-kDa GAP or if it is a proteolytic product, although we have been unable to proteolyze the larger form to the smaller form. There is also obvious GAP activity and protein between the two major peaks. Because the ratio of GAP activity to silver-stained protein is low between the peaks, we suspect that major silverstained bands in this region are contaminants and that the activity represents distribution of active proteolytic fragments of the 28-kDa GAP. We approached the question of proteolysis during purification by analyzing unfractionated brain membranes by SDS-gel electrophoresis. The principal peak of activity was at about 28 kDa, with tailing to about 20 kDa, but we could also detect small peaks of activity higher in the gel. These larger forms are not observed in purified preparation; peptides of M r Ͼ 30,000 in the phenyl-Sepharose fractions have no GAP activity.
Mechanism, Selectivity, and Regulation of G z GAP-Purified cerebral G z GAP is highly specific in its action on G␣ z . It displayed only slight activity with either G␣ i1 -GTP or G␣ o -GTP as substrates under conditions where hydrolysis of G␣ z -GTP was accelerated over 30-fold (Table IV). According to competitive inhibition, however, the affinity of the GAP for G␣ z is only about 3-fold greater than that for G␣ o and about 10-fold higher than for G␣ i1 (Table III). Evidently, G z GAP can bind these other G proteins with high affinity but cannot efficiently promote their deactivation. This unusual pattern of selectivity suggests that other members of the G i family may inhibit G z GAP in cells, where they are much more abundant than is G z . Given the selectivity of the purified GAP for G␣ z , it was initially surprising that there is significant activity in cells and tissues that express little if any G␣ z (Table I). It is likely that this activity is that of GAPs for other G i family members but which act on G z with low efficiency (47,48).
Cerebral G z GAP behaves generally as an integral membrane protein, although it was unusually refractory to solubilization by nondenaturing detergents. This behavior is reminiscent of caveolar proteins (42)(43)(44). However, caveolae are reported to be solubilized by octyl glucoside (G z GAP was not) and, in one experiment, G z GAP activity did not co-fractionate with caveolin in lysates of MA104 cells. We suspect that G z GAP is a markedly hydrophobic protein because of its resistance to solubilization and its tendency to aggregate. This conclusion is supported by its functional co-reconstitution with m2 muscarinic receptors and trimeric G z into phospholipid vesicles ( Fig. 9), in contrast to its inactivity when added to preformed receptor-G z vesicles. We have been unable to perform the standard tests for monomeric solubility of purified G z GAP because removing detergent by dilution or chromatography before assay led to loss of GAP activity. This was true even though the assay was performed in the presence of Triton X-100. Some of this behavior is similar to difficulties encountered in solubilizing adenylyl cyclase, a G protein-regulated effector that is a much larger, multi-span membrane protein.
There is inadequate information to compare this aspect of G z GAP with RGS proteins, although RGS4 and GAIP are both water-soluble (11,12) and Sst2p binding to membranes is sensitive to ionic strength (15).
The enzymologic mechanism of G z GAP action is apparently straightforward. 2 G␣ z -GTP is essentially stable over the usual assay interval. The GAP binds the GTP-bound form of G␣ z with nanomolar affinity (Figs. 5 and 7), and the GAP-G␣ z -GTP complex then hydrolyzes GTP fairly quickly (t1 ⁄2 ϳ15 s at 15°C; t1 ⁄2 Ͻ2 s at 30°C). The GAP binds G␣ z -GDP weakly if at all, such that the complex rapidly dissociates after hydrolysis. This mechanism allows G z GAP to act catalytically; i.e. one GAP molecule can cycle among multiple molecules of G␣ z -GTP. A corollary to this behavior is that the rate-limiting step in the GAP-mediated GTPase reaction is hydrolysis of the GAP-G␣ z -GTP complex. This conclusion is supported by the finding that the maximum reaction rate at saturating and super-stoichiometric concentrations of GAP (Fig. 5B) is the same as the FIG. 9. G z GAP amplifies agonist-stimulated steady-state GTPase activity in phospholipid vesicles that contain trimeric G z and m2 muscarinic cholinergic receptor. M2 muscarinic receptor and trimeric G z were co-reconstituted into phospholipid vesicles, with (E, q) or without (Ç, å) G z GAP, essentially as described previously (23) and under "Experimental Procedures". A, steady-state GTPase activity was measured in the presence of either 100 M carbachol (q, å]) or 5 M atropine (E, Ç). In parallel with the measurement of GTP hydrolysis, the steady-state concentration of G z -[␥-32 P]GTP in the assays with carbachol was assayed both at 5 and 15 min. The concentration of G z -[␥-32 P]GTP at both times was 460 pM in vesicles with GAP and 640 pM in vesicles without GAP. The concentration of GAP in the assay, when present, was about 30 pM. B, GTP␥S binding was measured in the same vesicles. maximum specific activity of the GAP at saturating G␣ z -GTP (Fig. 5A). In apparent contrast, the effect of RGS4 on G o and G i seems to be limited by substrate binding (12).
The activity of G z GAP during receptor-stimulated steadystate GTP hydrolysis is evident when it is co-reconstituted in phospholipid vesicles with G z and m2 muscarinic receptor (Fig.  9). The relative effect of the GAP activity was limited by its concentration and/or by the ratio of its concentration to that of G z . G z was in molar excess over GAP in the vesicles, the probable physiologic condition, and stimulation of steady-state hydrolysis increased if either more GAP or less G z was used. We have not yet pursued this relationship quantitatively because the availability of GAP and its concentration in stock solutions were both limiting. The need for G␤␥ in the vesicles to permit receptor-G␣ z coupling probably diminished the effect of the GAP (Fig. 8), and we also suspect that there was less than one molecule of GAP per vesicle. It is important to note that GAP does not alter the rates of dissociation of either GDP or GTP from G␣ z . It will not, therefore, influence activation rates and will only accelerate the deactivation limb of the GTPase cycle.
G z GAP binds the activated form of G␣ z with about the same affinity when it is bound to GTP (according to K m ), to GTP␥S, or to GDP/AlF 4 (according to K i ). In contrast, RGS4, a GAP for the G i family and G q , is inhibited much more potently by an GDP/AlF 4 -bound G␣ than by the same G␣ bound to GTP␥S (47,48). Because GDP/AlF 4 binds to G␣ i1 and G␣ t as a transition state analog (45,46), these authors suggested that RGS4 acts as a GAP by favoring the transition state structure of a G␣ over its GTP-bound form. This mechanism would presumably differentiate the GAP activity of RGS proteins from that of effectors, which are activated both by GTP␥Sand GDP/AlF 4 -liganded G proteins, and would thus suggest that the cerebral G z GAP is an effector. This distinction may not be generally valid, however, or may perhaps not extend to G z . In preliminary experiments, we found that RGS4 is potently inhibited by the GTP␥S-bound form of G␣ z , a behavior similar to that of the cerebral G z GAP. The active site and enzymatic properties of G z differ markedly from those of other G i family members (16 -18), and it is possible that GDP/AlF 4 is not a transition state analog at the active site of G␣ z . If true, however, this argument would favor an effector function for the G z GAP.
Regardless of any yet unknown cellular roles of the G z GAP, its presence and regulation will influence G z -modulated signaling. The first mode of regulation so far observed is inhibition of the GAP by G␤␥ subunits. Inhibition of GAP activity by G␤␥ over a reasonable range of concentrations allows modulation of G z signaling by other G protein pathways, where activation will release G␤␥ in large excess over G␣ z . Other controls of GAP activity are also likely, and their understanding should help us understand the cellular pathways uniquely regulated by G z .