Reciprocal Signaling between Heterotrimeric G Proteins and the p21-stimulated Protein Kinase*

p21-activated protein kinase (PAK)-1 phosphorylated Gαz, a member of the Gαi family that is found in the brain, platelets, and adrenal medulla. Phosphorylation approached 1 mol of phosphate/mol of Gαz in vitro. In transfected cells, Gαz was phosphorylated both by wild-type PAK1 when stimulated by the GTP-binding protein Rac1 and by constitutively active PAK1 mutants. In vitro, phosphorylation occurred only at Ser16, one of two Ser residues that are the major substrate sites for protein kinase C (PKC). PAK1 did not phosphorylate other Gα subunits (i1, i2, i3, o, s, or q). PAK1-phosphorylated Gαz was resistant both to RGSZ1, a Gz-selective GTPase-activating protein (GAP), and to RGS4, a relatively nonselective GAP for the Gi and Gqfamilies of G proteins. Phosphorylation of Ser27 by PKC did not alter sensitivity to either GAP. The previously described inhibition of Gz GAPs by PKC is therefore mediated by phosphorylation of Ser16. Phosphorylation of either Ser16 by PAK1 or Ser27 by PKC decreased the affinity of Gαz for Gβγ; phosphorylation of both residues by PKC caused no further effect. PAK1 thus regulates Gαz function by attenuating the inhibitory effects of both GAPs and Gβγ. In this context, the kinase activity of PAK1 toward several protein substrates was directly inhibited by Gβγ, suggesting that PAK1 acts as a Gβγ-regulated effector protein. This inhibition of mammalian PAK1 by Gβγ contrasts with the stimulation of the PAK homolog Ste20p in Saccharomyces cerevisiae by the Gβγ homolog Ste4p/Ste18p.

Protein kinases are the eventual downstream mediators of most signals initiated by G protein-coupled receptors. Mechanisms of kinase activation are diverse, however. They include both direct stimulation of cyclic AMP-dependent protein kinase and PKC 1 by second messenger products of G protein-regulated effectors and less direct activation of tyrosine kinase and mitogen-activated protein kinase cascades. In yeast, heterotrimeric G proteins regulate members of the p21-activated protein kinase (PAK) family. In Saccharomyces cerevisiae, Ste20p is stimulated by G␤␥ subunits (Ste4p/Ste18p) in response to mating pheromones (1,2), and in Schizosaccharomyces pombe, the G␣ subunit (Gpa1p) is the signal transducer to the Ste20p homolog Shk1p (3). The PAKs are mammalian homologs of Ste20p and Shk1p that were initially recognized as kinases that are activated by Rac and Cdc42, members of the Rho family of monomeric GTP-binding proteins (4). The PAKs also respond to heterotrimeric G proteins through pathways that include regulation of both GDP/GTP exchange factors and GAPs for Rac and Cdc42 (5)(6)(7).
Conversely, protein kinases modulate upstream signaling by heterotrimeric G proteins. Receptors are subject to feedback regulation by second messenger-activated kinases and G protein-coupled receptor kinases (8); G protein-regulated effectors are modulated by phosphorylation (9 -12); and in a few cases, G proteins are themselves phosphorylated (13)(14)(15)(16). G␣ z , a sparsely expressed member of the G i family, is phosphorylated by PKC both in platelets and in cells where it has been expressed artificially (13,(17)(18)(19). PKC-catalyzed phosphorylation decreases the affinity of G␣ z for G␤␥ subunits, potentially sensitizing G z to activation because G␤␥ inhibits GDP/GTP exchange. Phosphorylation by PKC also desensitizes G z to the GAP activity of RGS proteins, which are widely thought to inhibit G protein signaling (20). PKC may potentiate G z signaling through either of these mechanisms. G z is found primarily in neurons, platelets, and adrenal chromaffin cells, and its intracellular localization suggests that it may regulate formation, transport, or release of secretory granules (21)(22)(23)(24)(25). The ability of PAK1 to cause remodeling of cytoskeletal structures points to a role in regulating processes such as cell motility and secretion, and PAK1 has been implicated directly in the Fc␥ receptor-mediated respiratory burst and cytokine secretion (26). Few natural PAK substrates are known, however; but PAK1 probably phosphorylates proteins that regulate either cytoskeletal disassembly or the cytoskeletal elements themselves.
We report here that G␣ z is phosphorylated specifically at Ser 16 by PAK1, thus inhibiting its interaction with both G␤␥ and RGS proteins. We have used this specificity to distinguish and delineate the functional consequences of phosphorylation at Ser 27 and Ser 16 , which we show to be the two principal substrate sites for PKC. An unexpected outcome is the finding that G␤␥, which stimulates the PAK homolog Ste20p in Saccharomyces, inhibits the activity of mammalian PAK1 toward both G␣ z and other substrates.

EXPERIMENTAL PROCEDURES
Plasmids and cDNA-Mammalian expression vectors for full-length PAK1, the constitutively active mutant PAK1-(165-544) (N-terminal truncation leaving residues 165-544) (27), and G12V Rac1 were prepared in pCMV5M (pCMV5 modified to include a Myc epitope tag (27)) as described previously (27). The G12V Rac1 mutation was prepared using a QuikChange mutagenesis kit (Stratagene), and the cDNA was inserted into pCMV5. Mammalian expression vectors for wild-type G␣ z and its S16A, S25A, S27A, and S16A,S25A mutants were constructed in pDP5 and were gifts from D. Manning (University of Pennsylvania) (17). The S16A,S25A,S27P triple mutant was prepared using the QuikChange kit with the S16A,S27A construct as template. Recombinant baculoviruses expressing the G␣ z mutants were prepared as described previously for wild-type G␣ z (28).
Protein Expression and Purification-Wild-type and mutant G␣ and G␤␥ subunits, other than G␣ i1 , were expressed in Sf9 cells and purified as described (28,29). G␣ i1 was expressed in Escherichia coli with or without yeast protein N-myristoyltransferase (30) and purified as described (31). G␣ q and G␣ s were gifts from S. Mukhopadhyay and T. Kozasa (this department). Wild-type PAK1 and PAK1-(232-544) were expressed in E. coli as glutathione S-transferase fusion proteins and purified by glutathione-agarose affinity chromatography (27). To prevent proteolysis, wild-type GST-PAK1 was purified in the presence of 20 g/ml aprotinin, 10 g/ml pepstatin A, 10 g/ml leupeptin, and 1 mM phenylmethylsulfonyl fluoride. Purified PAK1 and PAK1-(232-544) were dialyzed against 20 mM Tris-Cl (pH 8.0), 100 mM NaCl, 1 mM dithiothreitol, 1 mM EDTA, and 1 mM benzamidine and stored at Ϫ80°C. The protein kinase TAO1 (32) was a gift from K. Berman (this department), and PKC␣ was a gift from T. Kozasa (this department). Purified phosducin (33) was a gift from R. Gaudet and P. Sigler (Yale University).
Protein Kinase Assays and Protein Phosphorylation-Phosphorylation of G␣ z by PKC␣ was performed exactly as described (28,29). G␣ subunits were phosphorylated by PAK1 at 30°C for 60 min or the times indicated in 50 mM Hepes (pH 8.0), 10 mM MgCl 2 , 1 mM dithiothreitol, and 0.5 mM ATP. Phosphoamino acid analysis and tryptic phosphopeptide mapping of G␣ z were performed as described (34). Partial tryptic proteolysis after protection of phosphorylated G␣ z by GTP␥S was performed exactly as described (35). Under these conditions, trypsin cleaves G␣ z after Arg 29 (35). Protein kinase assays using MBP as substrate were performed as described (36).

FIG. 2. Determination of phosphorylation sites on G␣ z .
A, tryptic mapping. G␣ z (20 pmol) was phosphorylated by full-length PAK1 or PKC␣ as described in the legend to Fig. 1. After adjusting free [Mg 2ϩ ] to 1 M with EDTA, the samples were incubated for 80 min at 30°C in binding buffer with or without 0.5 mM GTP␥S. Samples were then incubated with or without L-1-tosylamido-2-phenylethyl chloromethyl ketone-treated trypsin (1%, w/w) for 40 min at 30°C. Samples (5 pmol for samples incubated without trypsin and 12.5 pmol with trypsin) were analyzed by SDS gel electrophoresis, followed by Coomassie Blue staining (CB; lower panel) and autoradiography (upper panel). B, phosphorylation of G␣ z Ser mutants. Purified wild-type (WT; 5 pmol) or mutant (10 pmol) G␣ z was incubated with either PAK1 or PKC␣ for 60 min at 30°C as described in the legend to Fig. 1. Samples were analyzed as described for A. These experiments (A and B) were performed at least twice with similar results. C, phosphorylation of wild-type and Ser mutants of G␣ z . Aliquots (10 pmol) of wild-type (E, q), S16A (Ⅺ, f), or S27A (ƒ, ) G␣ z were incubated at 30°C with 1 pmol of PAK1 (q, f, ) or 1 pmol of PKC␣ (E, Ⅺ, ‚) and [␥-32 P]ATP (3000 cpm/pmol) as described under "Experimental Procedures." Samples were resolved by SDS gel electrophoresis, and incorporation of 32 P was determined by scintillation counting of Coomassie Blue-stained protein bands. Each point represents the mean of at least three experiments where S.D. was Ͻ15% of the mean. stored at Ϫ20°C. For immunoprecipitation, lysates were sonicated at 0°C for 10 s and centrifuged at 10,000 ϫ g for 5 min. The supernatants were diluted to 800 l with radioimmune precipitation assay buffer that contained 0.2% SDS and 0.2 mM phenylmethylsulfonyl fluoride and precleared by passage through a 0.2-ml column of Sephadex G-25. The precleared lysates were incubated with 20 l of protein A-agarose and 10 g of purified anti-G␣ z antibody for 15 h at 4°C. After extensive washing with radioimmune precipitation assay buffer that contained 0.2% SDS, precipitates were solubilized with SDS sample buffer and analyzed by electrophoresis on 10% polyacrylamide gels (37). Proteins were transferred to Schleicher & Schü ll BA85 nitrocellulose membranes for autoradiography or immunoblotting with anti-G␣ z antibody. Autoradiographs and Western blot images were quantitated with a Molecular Dynamics densitometer and appropriate standards.
Other Methods-G z GAP activity was assayed at 15°C as described (28). [ 35 S]GTP␥S binding was measured at 30°C as described. GTP␥Sbound G␣ z was partially proteolyzed with trypsin as described (35). Antibody against G␣ z was raised by injecting rabbits with the Nterminal peptide GCRQSSEEKEAARRSRR conjugated to hemocyanin. Antibody was purified from serum by precipitation with (NH 4 ) 2 SO 4 and, following dialysis, immunoaffinity chromatography on a column of the immunizing peptide coupled to CNBr-activated Sepharose CL-4B.

PAK1
Phosphorylates G␣ z Selectively at Ser 16 -The protein kinase PAK1 phosphorylated G␣ z in vitro, but did not phospho-rylate several other G␣ subunits tested (Fig. 1). PAK1-catalyzed phosphorylation of G␣ z was efficient relative to that catalyzed by PKC␣, and the truncated protein PAK1-(232-544) displayed activity similar to that of the wild-type kinase. We were unable to detect phosphorylation of G␣ z by either TAO1 or protein kinase A (data not shown).
PAK1 phosphorylated G␣ z within the first 29 amino acid residues because all 32 P was removed by limited tryptic proteolysis ( Fig. 2A), which cleaves after Arg 29 (35). Phosphoamino acid analysis of phosphorylated G␣ z detected only phosphoserine (data not shown), consistent with the absence of Thr residues in this region. In contrast to the site specificity displayed by PAK1, PKC phosphorylated at least one additional site C-terminal to Gln 30 , usually accounting for ϳ10% of the total incorporation of phosphate ( Fig. 2A, lane 12). We then used serine mutants of G␣ z to determine the site of PAK1catalyzed phosphorylation near the N terminus. PAK1 catalyzed the phosphorylation of G␣ z to ϳ1 mol of phosphate/mol of G␣ z , and phosphorylation was blocked completely by mutation of Ser 16 to Ala. Phosphorylation was not diminished by mutation of either Ser 25 or Ser 27 (Fig. 2, B and C). Longer incubation with PAK1 or the use of more PAK1 did not cause further phosphorylation of G␣ z (data not shown). PAK1 thus selectively phosphorylates Ser 16 of G␣ z .
In contrast to PAK1, PKC catalyzed the addition of 2 mol of phosphate/mol of G␣ z , and phosphorylation was decreased by about half when either Ser 16 or Ser 27 was mutated (Fig. 2, B and C). Both residues are thus PKC substrate sites. The time courses of phosphorylation of S16A and S27A G␣ z suggest that Ser 27 is the kinetically preferred PKC substrate (Fig. 2C). Such preference agrees with the conclusion of Lounsbury et al. (17) that Ser 27 is the major phosphorylation site in transfected 293 cells, although cellular phosphorylation might also be influenced by selectivity of whatever protein phosphatases naturally dephosphorylate G␣ z . We have not attempted to map the minor, more C-terminal PKC phosphorylation site on G␣ z .
To determine whether PAK1 also phosphorylates G␣ z in vivo, we expressed G␣ z in HEK-293 cells and measured its differential steady-state phosphorylation upon coexpression with PAK1, with or without the PAK activator G12V Rac1 (Fig.  3). G␣ z was phosphorylated by endogenous HEK-293 cell kinases during the 3-h incubation with [ 32 P]P i , primarily at one of the Ser residues near the N terminus. Phosphorylation was slightly increased by coexpression of wild-type PAK1 and was further increased when both PAK1 and the constitutively acti-vated G12V mutant of Rac1 were present. It was difficult to quantitate the in vivo phosphorylation of G␣ z by PAK1 because an unknown fraction of Ser 16 may already be phosphorylated prior to addition of 32 P and because considerable phosphorylation by other kinases occurred at Ser 27 (Fig. 3A, lanes 6 -8). However, phosphorylation of the S25A,S27P G␣ z mutant showed that wild-type PAK1 plus G12V Rac1 can at least double the incorporation of 32 P into Ser 16 of G␣ z and that the activated mutant PAK1-(165-544) can increase labeling by 50%. For reference, phosphorylation of G␣ z by endogenous PKC was monitored by stimulating the cells with phorbol ester, which increased phosphorylation of both wild-type and S16A G␣ z by ϳ2-2.6-fold (Fig. 3) (17). Phosphorylation of G␣ z at sites C-terminal to Arg 29 was relatively minor (Fig. 3B, lanes 8 -14).
Activation of G␣ z and PAK1-catalyzed Phosphorylation-Activation of G␣ z by AlF 4 Ϫ had no effect on its phosphorylation by either PAK1 or PKC (Fig. 4). Activation by GTP␥S was also without effect (data not shown). Conversely, neither PAK1catalyzed phosphorylation of MBP or PAK1 autophosphorylation was altered by G␣ z when bound to GDP/AlF 4 Ϫ (Fig. 4), GDP, or GTP␥S (data not shown).
Phosphorylation of G␣ z at Either Ser 16 or Ser 27 Inhibits Binding to G␤␥-To determine whether PAK1-catalyzed phosphorylation of G␣ z decreases its affinity for G␤␥, as is true for PKC (18), we monitored the effect of phosphorylation on the concentration dependence with which G␤␥ inhibits GDP/ GTP␥S exchange. As shown in Fig. 5A, phosphorylation of G␣ z by PAK1 markedly attenuated the ability of G␤␥ to inhibit nucleotide exchange on G␣ z , but had no effect on the intrinsic nucleotide exchange rate. This attenuation reflected a 10 -20fold decrease in the affinity of phospho-G␣ z for G␤␥ (Fig. 5, B and D). PAK1 and PKC inhibited G␤␥ binding equally. Furthermore, phosphorylation of Ser 16 by PAK1 alone (Fig. 5, B and D) or of Ser 27 by PKC in the S16A mutant (Fig. 5C) decreased G␣ z -G␤␥ affinity equally. Phosphorylation of both residues by PKC had no greater effect than phosphorylation of only one or the other. Although mutation of either residue to Ala itself decreased affinity for G␤␥ somewhat (IC 50 shifted FIG. 5. Effects of phosphorylation of G␣ z at Ser 16 and/or Ser 27 on its interaction with G␤␥. G␣ z was phosphorylated by either PAK1 or PKC␣ by incubation at 30°C for 90 min as described in the legends to Figs. 1 and 2. After adjusting free [Mg 2ϩ ] to 1 M with EDTA, the samples (2.2 pmol/assay point) were preincubated for 10 min at 0°C in binding buffer with or without G␤␥. Binding of [ 35 S]GTP␥S (20 M, 1500 cpm/pmol) was then determined as described (28). A, GTP␥S binding to either PAK1-phosphorylated (OE, ‚) or control (E, q) wild-type G␣ z was measured in the presence (OE, q) or absence (‚, E) of 100 pmol of G␤␥. Each curve was fit to a first-order reaction scheme, and all four curves share the same value of maximal GTP␥S binding. B-D, the rate constants for GTP␥S binding to wild-type (WT) G␣ z , S16A G␣ z , and S27A G␣ z , respectively, were determined at increasing concentrations of G␤␥ (shown as the molar ratio G␤␥/G␣ z ). G␣ z was either phosphorylated (closed symbols) or not (open symbols; no kinase added) as shown. Data are means of duplicate determination in two or three experiments (n ϭ 4 or 6).

FIG. 6. Interactive effects of phosphorylation or mutation at Ser 16 and Ser 27 and of G␤␥ on GAP-stimulated hydrolysis of G␣ z -bound GTP.
Samples of wild-type (WT) or mutant G␣ z were phosphorylated by either PAK1 or PKC␣ as described in the legend to Fig. 2C, presumably to nearly stoichiometric incorporation of phosphate at Ser 16 and/or Ser 27 . Control (non-phosphorylated) samples were treated identically, except that protein kinase was omitted. Samples were then bound to [␣-32 P]GTP, and hydrolysis of G␣ z -bound GTP (2 nM) stimulated by 50 pM RGSZ1 was measured as described (28) Table I. from ϳ1 M to ϳ2.5 M in this and other experiments), the additional decrease caused by phosphorylation was much greater, ϳ12-fold in Fig. 5B and ϳ20-fold in Fig. 5 (C and D) and similar experiments.
Phosphorylation of G␣ z at Ser 16 Blocks the GAP Activity of RGS Proteins-Phosphorylation of G␣ z by PKC blocks the GAP activity of RGS proteins (29,38). Selective phosphorylation of Ser 16 by PAK1 also substantially inhibited the GAP activities of both RGSZ1 and RGS4 (Fig. 6 and Table I). RGS4 was inhibited by Ͼ96%. RGSZ1 was inhibited by ϳ60 -80% under the conditions shown. G z GAP purified from bovine brain, which contains RGSZ1 and at least one other member of the RGSZ subfamily (29), was inhibited to the same extent as was recombinant RGSZ1 (data not shown). Inhibition of GAP activity by G␣ z phosphorylation is caused by an increase in K m , which probably indicates a decrease in affinity (29). Fractional inhibition will therefore vary with the concentration of phospho-G␣ z relative to its K m , and the greater fractional inhibition of RGS4 thus reflects its lower affinity for GTP-bound G␣ z .
Because PAK1 phosphorylates Ser 16 exclusively and PAKcatalyzed phosphorylation inhibits responses to GAPs as much as does phosphorylation of both Ser 16 and Ser 27 by PKC, phosphorylation of Ser 16 can account for the inhibitory effect of PKC on GAP activity. Phosphorylation of Ser 27 in S16A G␣ z had no effect on its sensitivity to either GAP. Thus, PAK1 will be just as efficacious an inhibitor of the RGSZ family of G z GAPs as is PKC.
Addition of G␤␥ to phosphorylated G␣ z had little further inhibitory effect on GAP activity (Fig. 5). The slight inhibition that was observed occurred primarily below 200 nM G␤␥, well below the K d of phospho-G␣ z for G␤␥. Residual inhibition by G␤␥ probably reflects effects on the small fraction of G␣ z that was not phosphorylated during incubation with kinase.
In addition to blocking phosphorylation, mutation of either Ser 16 or Ser 27 to Ala also inhibited the intrinsic sensitivity of G␣ z to GAP activity ( Fig. 6 and Table I). Sensitivity of nonphosphorylated S16A G␣ z to either RGSZ1 or RGS4 was equivalent to that of phospho-Ser 16 G␣ z and was not further altered by phosphorylation of Ser 27 . S27A G␣ z was more sensitive to GAP activity than was S16A G␣ z , but was still a far worse GAP substrate than wild-type G␣ z . It was striking that PKC-catalyzed phosphorylation at Ser 27 in the S16A mutant reproducibly increased its sensitivity to GAPs. Although this effect was relatively small, it was a consistent finding in multiple experiments. The small inhibitory effect of G␤␥ on the sensitivity of mutated or phosphorylated G␣ z to GAPs was similarly reproducible.
The intrinsic rates at which S16A and S27A G␣ z hydrolyzed bound GTP were diminished in comparison to the wild type, by ϳ15 and 25% respectively ( Table I). Proteolysis of the N-terminal 29 amino acid residues of G␣ z provided a control to show that the effects of mutation or phosphorylation of Ser 16 and Ser 27 are local. Limited tryptic proteolysis increased the k hydrol for each of these proteins to ϳ0.022 min Ϫ1 (data not shown), the value characteristic of the proteolyzed wild-type, non-phosphorylated protein (35).
G␤␥ Inhibits the Protein Kinase Activity of PAK1-During the course of the experiments described above, we noticed that G␤␥ reproducibly inhibited the protein kinase activity of PAK1-(232-544) (Fig. 7A). Inhibition was detectable by 80 nM and was half-maximal at ϳ200 nM, well within the range of concentrations over which other regulatory actions of G␤␥ have been described (39). Similar results were obtained with full-length, wild-type PAK1 both before and after removal of the fused GST domain (data not shown). The buffer used to store G␤␥ had no effect on PAK1 activity (Fig. 7A). G␤␥ also inhibited the ability of PAK1 to phosphorylate either MBP (Fig. 7) or MEK1 (data not shown). G␤␥ thus appears to inhibit PAK directly rather than simply binding the G␣ z substrate and blocking access to the kinase. The inhibitory activity of G␤␥ was relatively specific for PAK. It had no effect on the protein kinase activities of either protein kinase A or TAO1 in two separate experiments and inhibited PKC insignificantly (10 -15% at the highest concentrations tested).
G␤␥ slightly but reproducibly stimulated PAK1 autophosphorylation in the presence or absence of added substrate (example in Fig. 7A), indicating that G␤␥ binds directly to PAK1 to alter its function. In yeast, G␤␥ also binds (and stimulates) the PAK homolog Ste20p directly (1, 2), and its stimulatory activity is blocked by binding to G␣. Surprisingly, G␤␥ was a potent inhibitor of PAK1 protein kinase activity both when free or when complexed as a heterotrimer with GDP-bound G␣ z or G␣ i (Fig. 7, A and B) or G␣ q (data not shown). Addition of the G␤␥-binding protein phosducin (up to 6.7 M) also failed to reverse inhibition by G␤␥ (Fig. 7B).

DISCUSSION
Mammalian PAKs were discovered as effectors of the small G proteins Rac and Cdc42, although the yeast homolog Ste20p was first identified as a protein kinase activated by G␤␥. Our current findings now indicate that PAKs function both upstream and downstream of heterotrimeric G proteins in animal cell signaling pathways. First, PAK1-catalyzed phosphorylation potentiates G z activation by inhibiting the GAP activity of RGS proteins, including the RGSZ subfamily of G z -selective GAPs. Second, phosphorylation of G␣ z decreases its affinity for G␤␥ subunits and thus attenuates the inhibitory effects of G␤␥ on G z activation. The net result is a two-pronged potentiation of G z signaling by PAK. The reduced affinity of phospho-G␣ z for G␤␥ will promote G␤␥ release and might thereby potentiate G␤␥ signaling, but G␣ z is expressed at such low levels that it may release too little G␤␥ to have significant impact on intracellular signaling.
Several lines of evidence are consistent with the hypothesis that PAK phosphorylates G␣ z under physiological conditions in cells. In vitro, PAK1-phosphorylated purified G␣ z to a stoichiometry of 1 mol of phosphate/mol of G␣ z . The rate of phosphorylation was also reasonably fast in comparison with PKC, which phosphorylates G␣ z in platelets stimulated by either thrombin or phorbol ester (13,19). Phosphorylation of G␣ z in HEK-293 cells was increased by expression of constitutively active PAK1 or of wild-type PAK1 and its activator Rac. PAK1driven incorporation of 32 P into G␣ z was of the same order as that catalyzed by PKC in response to 12-O-tetradecanoylphorbol-13-acetate despite the fact that PKC phosphorylates G␣ z on two sites rather than one. Finally, phosphorylation of G␣ z by PAK is associated with altered function of the protein as discussed more fully below. We conclude that stimulation of PAKs, via the activation of either Rac or Cdc42, provides a novel means of potentiating the cellular function of G z . PAK1 displays marked selectivity for G␣ z relative to other G␣ subunits and for Ser 16 relative to other potential phosphorylation sites in G z . Selectivity for Ser 16 allowed us to delineate the individual contributions of phosphorylation of Ser 16 and Ser 27 to regulation of G␣ z in a manner not possible using mutagenesis. Phosphorylation of Ser 16 was both sufficient and necessary to decrease sensitivity to the GAP activity of RGS proteins ( Fig. 6 and Table I). It was also sufficient to decrease affinity for G␤␥. PKC-catalyzed phosphorylation of Ser 27 in the S16A mutant also decreased affinity for G␤␥. On the other hand, phosphorylation of both residues in wild-type G␣ z had no more effect than phosphorylation of Ser 16 alone. These data confirm the idea that the extreme N-terminal helix of G␣ subunits is crucial for interaction with RGS proteins (29,35), despite the fact that this interaction was not observed in the crystal structure of the G␣ i1 -RGS4 complex (40).
The ability of G␤␥ subunits to inhibit the protein kinase activity of PAK1 is provocative for two reasons. First, it suggests that the PAKs may be a new family of heterotrimeric G protein-regulated effectors. Inhibition of PAK1 by G␤␥ was nearly complete, was effective with multiple protein substrates, and occurred over a physiological range of G␤␥ concentrations. Second, PAK inhibition by G␤␥ extends the pattern of G protein regulation of the PAK family that was established in yeast. However, whereas Shk1p is activated, perhaps indirectly, by G␣ in S. pombe and Ste20p is directly activated by G␤␥ in S. cerevisiae, PAK1 is directly inhibited by mammalian G␤␥. The ability of G␤␥ to regulate PAK1-(232-544) also indicates that the G␤␥-binding site on PAK is unrelated to the binding site for Rac and Cdc42, which lies near the PAK N terminus (41). A tantalizing possibility is that mammalian G proteins may regulate PAKs through multiple mechanisms.
Inhibition of PAK by G␤␥ in vitro fulfilled most criteria for physiological validity, so it was initially puzzling that inhibition was not blocked either by GDP-bound G␣ i or by phosducin. However, yeast G␤␥ binds Ste20p through the N-terminal helix of G␤ (2, 42), which is not occluded in the phosducin-G␤␥ complex (33). Whereas the N-terminal helix of G␤ makes extensive contact with G␣ (43,44), the face of the helix that binds PAK may remain accessible in some conformations of the G␣-G␤␥ heterotrimer. Contact sites for other G␤␥-regulated effectors cluster on the face of the G␤ torus rather than near its N terminus (45) and are thus fully blocked by G␣. It should be possible to use such structural information to evaluate the biological relevance of PAK inhibition by G␤␥ in cells where PAK activity can be monitored in response to receptor-regulated G␤␥ release.