Guanine Nucleotide Exchange Regulates Membrane Translocation of RacLRho GTP-binding Proteins*

GTP-binding proteins of the Rho family are main- tained as cytosolic complexes with RhoGDI in resting cells, but are released and translocate to the membrane during the course of cell activation. Membrane association of Rac/Rho/CDC42 was specifically induced by GTP analogs and required a heat- and trypsin-labile mem- brane component. Translocation was associated with the release of Rho family proteins from RhoGDI, but such release did not occur in the absence of membranes, nor was release in the absence of guanosine B'-O-(thio-triphosphate) (GTP@) sufficient for membrane associa- tion. Membrane binding was correlated with exchange of GTPyS for GDP on Rac, and only GTPyS-bound Rac became membrane localized. We propose that translocation of Rac and other members of the Rho family is controlled by membrane-associated guanine nucleotide ex- change factors, providing a mechanism to regulate the release and activation of individual members of the Rho family during cell stimulation. GTP-binding proteins of the Rho family (Rho, Rac, CDC42Hs) have been shown to be important regulators of cell function. Rac controls the activity of the phagocytic leukocyte NADPH oxidase (1-3) and can stimulate the polymerization of actin associated with membrane ruffling and lamaellopod formation (4). Rho stimulates the

GTP-binding proteins of the Rho family are maintained as cytosolic complexes with RhoGDI in resting cells, but are released and translocate to the membrane during the course of cell activation. Membrane association of Rac/Rho/CDC42 was specifically induced by GTP analogs and required a heat-and trypsin-labile membrane component. Translocation was associated with the release of Rho family proteins from RhoGDI, but such release did not occur in the absence of membranes, nor was release in the absence of guanosine B'-O-(thiotriphosphate) (GTP@) sufficient for membrane association. Membrane binding was correlated with exchange of GTPyS for GDP on Rac, and only GTPyS-bound Rac became membrane localized. We propose that translocation of Rac and other members of the Rho family is controlled by membrane-associated guanine nucleotide exchange factors, providing a mechanism to regulate the release and activation of individual members of the Rho family during cell stimulation.
GTP-binding proteins of the Rho family (Rho, Rac, CDC42Hs) have been shown to be important regulators of cell function. Rac controls the activity of the phagocytic leukocyte NADPH oxidase (1)(2)(3) and can stimulate the polymerization of actin associated with membrane ruffling and lamaellopod formation (4). Rho stimulates the formation of actin stress fibers and focal adhesions (5,6) and can regulate motile responses of cells (7). The function of CDC42 in mammalian cells has not yet been determined, but it regulates bud site selection in Saccharomyces cerevesiae (8).
The activities of GTP-binding proteins of the Rho family are regulated by additional proteins which modulate their GTP/ GDP state. A variety of GTPase activating proteins are active on Rho family members (9)(10)(11)(12). In addition, all Rho family members bind to a cytosolic regulatory protein known as RhoGDI. This protein is an important determinant of the overall function of the Rho family since it can inhibit GDP dissociation (131, prevent GTP hydrolysis (14,151, and maintain Rho family proteins in soluble (cytosolic) forms (16). Finally, proteins which stimulate the exchange of GTP for GDP on Rho proteins are necessary for activation in the presence of cytosolic M e concentrations. Dbl has activity toward CDC42 and Rho (17), but, thus far, only the protein termed smgGDS has been grants HL48008 and GM44428, as well as by an Arthritis Foundation * This work was supported by United States Public Health Service Fellowship (to T.-H. C.). 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  shown to stimulate nucleotide exchange on Rac (18,19). Little is known about how the activities of these critical regulators of Rho family function are controlled.
In resting neutrophils, Rac is maintained as a cytosolic complex with RhoGDI (20,21). Upon cell activation, Rac is released from this complex and becomes associated with the membrane fraction which contains the assembled NADPH oxidase (22). Translocation of Rac occurs with nearly identical kinetics and stoichiometries as does the movement of the p47ph" and p67p"" oxidase components (22). A knowledge of the mechanisms which control the movement of Rac from the cytosol to membrane will be important for understanding how the active NADPH oxidase is regulated. Such information is also important for understanding the control of Rac and Rho during assembly of the actin cytoskeleton. The formation of membrane ruffles (Rac) can be stimulated by growth factors such as platelet-derived growth factor in the absence of actin stress fiber formation (Rho), while lysophosphatidic acid can stimulate the formation of stress fibers in the absence of membrane ruffling (4,5). Since both Rac and Rho are found complexed with RhoGDI in the cytosol of unstimulated cells, a form which is unable to interact effectively with target-andor regulatory proteins, mechanisms for specific release from these complexes leading to Rac and Rho activation must be controlled at the level of receptor-initiated signaling pathways.
In the current study, we have investigated the regulation of Rac and Rho translocation in a cell-free system derived from human neutrophils. Based upon these studies, we propose a mechanism by which specific activation of individual GTPbinding proteins of the Rho family might be controlled during cell stimulation.
EXPERIMENTAL PROCEDURES Membrane and Cytosol Preparation, Damlocation-Membrane and cytosolic fractions were prepared by the methods detailed in Curnutte et al. (23). Protein concentrations of membrane and cytosol fractions were determined using the Pierce BCA protein assay kit with bovine serum albumin as the standard. Translocation in vitro was routinely performed by mixing purified neutrophil membranes with a %fold amount (proteidprotein) of neutrophil cytosol in the presence of 100 p~ GTP@ f 10 m M EDTA and incubation at 30 "C for 15 min. Membranes were then pelleted in a Beckman 70.1 Ti rotor for 45 min at 2", the cytosol removed for Western blot analysis, and the membranes resuspended for analysis by Western blotting. The time course of Rac translocation was directly dependent upon the ratio of membrane to cytosol used, with the 1:3 ratio found to be optimal. Modifications to the basic procedure are described in the text.
Preparation of Proteins-Recombinant p47P"" was isolated from a bacterial expression system (a kind gift of Dr. Stephen Chanock, NIH) as a glutathione S-transferase fusion protein using standard procedures. Recombinant p67Ph" was isolated by modifications of the proce- Uhlinger and David Lambeth, Emory. Both proteins were greater than 70-90% pure as isolated. Racl, Rac2, and RhoA were expressed in a baculovirus-Sf9 cell expression system, purified as described previously (251, and quantitated by I:''SlGTPyS binding (26). RhoGDI was expressed in E. coli and purified as previously described (15). Complexes of Racl or RhoA with RhoGDI were prepared by incubating Rac or Rho with a 3-fold molar excess of RhoGDI for 5 min a t 30 "C; we have confirmed that nearly complete complex formation occurs under these conditions (15,20). Where indicated, Racl was preloaded with ["HlGDP prior to use, as described elsewhere (15,26).
Membrane BindinglNucleotide Exchange-For the membrane associatiodnucleotide binding experiments of Fig. 6, the indicated form of Rac or Rho was prepared and incubated in 25 mM Ms-HCI, pH 7.5, 1 mM EDTA, 1 mM dithiothreitol, 10 mM MgCI, with 150 pgof neutrophil membrane in the presence or absence of 20 p~ ["SIGTPyS (specific activity = 2 4 , 0 0 0 cpdpmol) for 15 min a t 30 "C, and the reaction was stopped with 4 ml of ice-cold buffer (25 mM Ms-HCI, pH 8.1 mM EDTA, 1 mhl dithiothreitol, 150 rnM NaCI, and 30 mM MgCI,), then pelleted for 30 min in a Beckman 70.1 Ti rotor. The supernatants were analyzed for protein bound (by filtration over Ba85 nitrocellulose filters) or total radiolabel (by direct scintillation counting). The pelleted membranes were resuspended in 300 p1 of incubation buffer and scintillation counted to assess membrane bound radiolabel. In some experiments, the resuspended washed membranes were extracted with 0.5% Nonidet P-40 plus 0.5% cholate for 30 min on ice for immunoprecipitation analysis with Rac antibody, as indicated in the text.
Cell-free NADPH Oxidase Assay-Superoxide formation was assayed exactly as previously described (25).

GTPyS Stimulates lFanslocation of
Rac-Upon addition of the non-hydrolyzable guanine nucleotide analog GTPyS to a mixture of human neutrophil cytosol and plasma membrane, we observed that Rac protein which had previously been present in the cytosolic fraction now became membrane-associated ( Fig. 1). This observation confirms the original report of this phenomena by Philips et al. (28). "Translocation" to the mem- 2. Time course of in uitro translocation. A, the binding of RhoGDI and Rac to neutrophil membranes was assessed after stimulation with 100 p~ GTPyS for the indicated times as described under "Experimental Procedures." Lane I , 0 min; lane 2 , 5 min; lane 3,15 min; lane 4,30 min; lane 5,60 min. Approximately twice as much sample was used in the RhoGDI blot, and the amount of RhoGDI translocated was estimated to be less than 1% of the total cytosolic pool by comparison with the intensity of the remaining cytosolic RhoGDI on the same immunoblot (not shown). The autoradiograms shown were exposed to Kodak X-Omat film for 48 h (IRhoIGDI) and 24 h (Rac2) a t -80 "C, respectively. B, time courses of Rac membrane binding after stimulation with 50 p# GTPyS were determined in the absence (A) or presence (A) of 300 p~ arachidonic acid and in the absence of GTPyS, but in the presence of 300 p~ arachidonic acid (0). In control experiments with arachidonic acid f GTPyS in the absence of membranes, there was no association of Rac with the pellet fraction, indicating that we were not pelleting lipid micelles containing Rac protein. Data shown in A and B are representative of at least three similar experiments. brane was rapid (Fig. 21, occurring within less than 5 min and peaking by 5-15 min; the rate of association was dependent on the relative ratio of membrane to cytosol. The effect of GTPyS was half-maximal at -5 PM and reached a maximum by 50 p~. GTP was variably active; in some experiments we observed substantial translocation of Rac at 1 mM GTP, but in most experiments the effect was slight, presumably due to rapid hydrolysis of the GTP. GDP, GDPPS, ATP, and ATPyS were all inactive at up to 1 mM (Fig. 1). Translocation was not stimulated by AI,F, in the presence or absence of GDP, suggesting that a heterotrimeric GTP-binding protein was not involved in this process. Prebinding of GTPyS to the membranes (1 mM for 60 min a t 30 "C, followed by three membrane washes in Mg2" containing buffers) did not stimulate Rac translocation from the added cytosol, indicating that the GTP requirement was not due to a membrane-associated GTP-binding protein. The membrane-associated Rac was not removed by washing the membranes with 500 mM NaCl or 20 mM EDTA, but it could be extracted with detergents.
The ability of GTPyS to stimulate the transfer of GTP-binding protein from cytosol to membrane was not limited to Rac; both Rho and CDC42Hs translocated in the presence of this nucleotide (Fig. 1). In contrast, we observed no GTPyS-stimulated release of Rapl from the neutrophil membrane (Fig. 1).
lFanslocation Is Dependent upon Membrane Protein Component(s)-To determine whether Rac translocation required a membrane-associated protein component, we pretreated neutrophil membranes with heat (60 "C for 15 min) or trypsin (15 pg/ml for 20 min a t 30 "C) and used these membranes in the in vitro translocation assay. Both heat and protease treatment of the membranes blocked the transfer of Rac (Fig. 3), indicating either the existence of a protein binding site for Rac in the membrane or the need for a membrane-associated protein activity for translocation to occur. The membrane protein component(s) necessary for Rac translocation does not Rac translocation. Rac translocation was assessed as described after the following treatments of the membrane fraction: lune 1, no additions; lune 2, +lo0 p~ GTPyS; lune 3, +heating 60 "C, 15 min, +lo0 PM GTPyS; lune 4 , +30 pg/ml trypsin, 20 min, 30 "C, +lo0 p~ GTPyS; lune 5, +15 pg/ml trypsin, 20 min, 30 "C, +lo0 p~ GTPyS; lune 6, +30 pg/ml trypsin with 90 pg/ml soybean trpsin inhibitor, 20 min, 30 "C, +lo0 p~ GTPyS.
Data shown are representative of two similar experiments.
appear to be any of the proteins which make up the active NADPH oxidase, as we have found that translocation of Rac induced by GTPyS still occurred when we used membrane or cytosolic materials from patients with chronic granulomatous disease that were deficient in either the membrane-associated cytochrome b,,, or the cytosolic components and p67""" (29).
Translocation of Rac was not dependent upon the activity of protein methyltransferases in the membranes, as translocation was not blocked in the presence of the methyltransferase inhibitors N-acetyl-S-farnesyl-L-cysteine or S-farnesyl-thioacetic acid (301, even though these agents effectively inhibited methylation reactions in neutrophil membranes by more than 80 and 90%, respectively, as assessed by incorporation of "H-labeled S-adenosyl methionine into membrane-associated proteins in the -20-kDa size range. These results clearly indicate that, at least in vitro, translocation of Rac is not controlled by carboxyl methylation reactions, as originally proposed by Philips et al. (28). Similarly, translocation was independent of the presence of ATP, since it took place when GTPyS was added to dialyzed cytosol which was depleted of ATP. This indicates that kinases are not involved, and we found Rac translocation in vitro to be insensitive to protein kinase and phosphatase inhibitors, as well as the phosphatidylinositol 3-kinase inhibitor wortmannin. The GTPyS-induced binding of Rac to the plasma membrane thus appears to involve an as yet unidentified protein component.
Danslocation Induced by GTPyS Is Accompanied by Dissociation of Cytosolic Rac.GDI Complexes-We have shown previously that Rac in human neutrophils is largely cytosolic and exists in a stoichiometric complex with RhoGDI (20). Translocation of Rac to the plasma membrane induced in intact neutrophils by chemoattractant stimuli and/or phorbol ester is associated with release of Rac from this complex, as very little RhoGDI became membrane-associated under these activating conditions (22). Similarly, the majority of RhoGDI did not bind to the membrane fraction when Rac translocation was induced in the broken cell system by GTPyS, suggesting complex dissociation also occurred in vitro (Fig. 1). We did, however, detect the transient association of a small fraction of RhoGDI (less than 1% of total cytosolic RhoGDI) with the membrane when stimulated with GTPyS ( Fig. 2 . 4 ) or when intact cells were activated (not shown). This observation suggests that RhoGDI does interact at the level of the membrane during the process of Rac translocation, although it does not stably associate with the membrane.
To assess the ability of GTPyS to induce dissociation of cytosolic RaeGDI complexes, we incubated cytosol with 100 J~M GTPyS for 15 min a t 30 "C in the presence or absence of 10 mM EDTA, conditions which promote nucleotide exchange onto Rac (26). As shown in Fig. 4, under GTPyS loading conditions, but We have established in a prior study that biologically active lipids such as arachidonic acid and phosphatidic acid can dissociate Rac.GDI complexes in neutrophil cytosol (20). While treatment with 300 PM arachidonic acid did cause the disruption of Rac.GDI complexes (20), this condition was not correlated with significant Rac transfer to the membrane (see Fig.  2B ). This result suggests that stable membrane binding is dependent upon both the release of Rac and its conversion to a GTP-bound form.
As shown in Fig. 2 B , arachidonic acid by itself a t concentrations up to 300 PM did not stimulate translocation of Rac. Arachidonic acid did, however, enhance the membrane association of Rac in the presence of GTPyS; this effect was concentration-dependent, with a half-maximal concentration of -150 PM. We observed that arachidonic acid shifted the concentration-response curve for GTPyS to the left as well (not shown). Phosphatidic acid similarly enhanced Rac translocation only when GTPyS was present. These effects of added lipids were not observed in the absence of membrane, and were thus not a result of pelleting of lipid micelles.
The Rac which was translocated to the membrane was fully capable of supporting NADPH oxidase activity (Fig. 5 ) . When membranes to which Rac had been translocated were assayed for 0, formation in a cell-free system to which recombinant p47P"" and p67phoX were added, they produced 0; at rates which were only marginally increased by the addition of soluble Rac. Activity was still absolutely dependent upon the addition of p47 and p67, indicating that there was no carryover of other cytosolic components. Furthermore, while the activity of soluble Rac added to the system was dependent upon the presence of exogenous GTPyS (Fig. 5 ) , we observed that after translocation of Rac to the membrane in the presence of GTPyS and subsequent removal of any residual free nucleotide by repeated washing, the system was able to produce 0, at normal rates in the absence of added GTPyS. These data indicate that the Rac which has been translocated must be in an active GTP-bound state, capable of supporting NADPH oxidase function.
Tkanslocation of RaclRho Is Accompanied by Guanine Nucleotide Exchange-We examined the process of Rac and Rho translocation induced by GTPyS plus membranes to determine if membrane association was accompanied by guanine nucleotide exchange. When [3HlGDP/Rac.GDI complexes were prepared and incubated with neutrophil membranes, we observed that only between 3 and 10% of the labeled protein became membrane-associated in various experiments (Fig. 6, left), consistent with the view that GDI prevents membrane association and maintains Rac in a soluble, cytosolic complex in unstimulated neutrophils. Addition of free GDP/Rac to the membranes in the presence of [35SlGTPyS resulted in the almost quantitative transfer of Rac to the membrane, as determined by measuring the increase in membrane-bound [35SlGTPyS (Fig. 6). Such quantitative nucleotide exchange onto Rac will not occur under these Mg2' concentrations in the absence of an exchange factor (26) and confirmed in control experiments), and such an activity is provided by the membrane fraction, as little exchange took place in the absence of added membrane. We verified that [35SlGTPyS was bound t o Rac in the membrane pellet by extraction and immunoprecipitation with a Rac antibody, as described under "Experimental Procedures" (data not shown).
Similarly, when (GDP)Rac.GDI complexes were incubated with membranes under conditions where we had verified GTPyS-induced translocation to occur in the previous experiments, we observed substantial transfer of [35S]GTPyS binding to the membrane (Fig. 6). There was some reduction in the amount of Rac which bound GTPyS and became membraneassociated when compared with the free Rac; this is probably due to the presence of 3-fold molar excess of GDI in the incu- bation. Overall, these data suggest that guanine nucleotide exchange onto Rac is stimulated by a membrane component even in the presence of Rac/Rho.GDI complexes. Furthermore, this exchange is correlated with quantitative binding of (GTP)-Rac to the neutrophil membrane. Analogous results were obtained with Rho (Fig. 6, right). Interestingly, the binding of Rho to the membranes appeared to saturate at a much lower value for Rho than for Rac, suggesting that there was an intrinsic difference in the level of binding sites available to each protein. DISCUSSION Activation of the human neutrophil NADPH oxidase is associated with the translocation of the regulatory GTP-binding protein Rac2 to the cell membrane. Since it is difficult to use classical transfection approaches to manipulate Rac in myeloid cells, we used a cell-free system to investigate the biochemical mechanisms involved in the Rac activation process. We observed that the transfer of Rac (and Rho/CDC42) from soluble cytosolic complexes with RhoGDI to membrane sites requires the presence of a stable GTP analog (Fig. 1) and a membraneassociated protein, as determined by the sensitivity of translocation to heat or protease treatment of the membranes (Fig. 3). The GTP requirement is not mediated at the level of the membrane, nor is it mimicked by AI,F,, indicating that RaplA and heterotrimeric GTP-binding proteins are unlikely to be involved. Translocation of Rac is correlated with its release from cytosolic complexes with RhoGDI (Fig. 4). We have shown that the disruption of Rac(GDP).GDI complexes with arachidonic acid (Fig. 2B) in itself is insufficient to initiate translocation, even though the presence of arachidonate enhances translocation stimulated by GTPyS. Similar results were obtained when Rac.GDI complexes were disrupted with a specific RhoGDI antibody (data not shown). Addition of GTPyS alone to neutrophil cytosol is also insufficient to disrupt Rac.GDI complexes (Fig.  4). This is likely due to the fact that Rac does not readily exchange endogenous GDP for GTPyS at physiological Mg2' concentrations, nor even when Mg2' levels are lowered with EDTA when RhoGDI is present (13-15). However, the combination of neutrophil membranes and GTPyS does disrupt the complex of Rac with GDI, and translocation is initiated. These data indicate that in the presence of an as yet unidentified membrane protein, the binding of GTP(+) to Rac is stimulated, and release from RhoGDI occurs.
Several pieces of data indicate that GTPyS binds to cytosolic Rac to induce stable membrane association. The fact that translocated Rac was able to support NADPH oxidase activity even in the absence of exogenous guanine nucleotide is a strong indication that the membrane-bound protein must be in a GTPyS-bound active form (Fig. 5). In experiments done with recombinant Rac and GDI in the absence of cytosol, transfer of [35S]GTPyS to Rac was correlated with the binding of this radiolabeled protein to the membrane (Fig. 6). Finally, this conclusion is supported by the observation that experiments in which L3H1GDP-labeled recombinant Rac.GDI complexes were stimulated to translocate by the addition of [35S]GTPyS showed only 35S-label in membrane Rac immunoprecipitates.
These data suggest a model which can account for translocation of Rac and related Rho family proteins both in the broken cell system and in intact cells. The model is shown schematically in Fig. 7. Step 1 would involve the activation of a membrane-associated exchange factor (GEF) as a result of cell activation by receptor stimuli or phorbol esters. We have shown in this study, as well as in an earlier report (25), that neutrophil membranes contain a factor which promotes exchange of GTP for GDP on Rac. This exchange factor appears t o be constitutively active in the membranes used for these experiments. This might be due to either the partial activation of the exchanger during preparation of the membranes andor the loss of a n inhibitory influence normally present. Alternatively, we may be picking up the effect of a small amount of "basal" exchange factor activity, since quite high concentrations of membrane are required for optimal exchange and translocation to be observed. We hypothesize that the putative exchange factor can act upon Rac(GDP) in the complex with RhoGDI to catalyze nucleotide exchange (Step 2 ) , since nucleotide exchange appears to be the only obvious signal for inducing translocation from cytosolic andor recombinant GDI complexes in these experiments. We have noted herein that there is a small amount of RhoGDI that binds to the plasma membrane during cell activation (Fig; 2 A ) . The normally cytosolic Rac.GDI complex may transiently associate with a binding site formed by the activated exchange protein. We do not rule out however that exchange is occurring subsequent to the release of a fraction of Rac upon interaction of the GDI complex with active lipids in the membrane bilayer (20). Indeed, such an effect could account for the kinetically defined "GDI-displacement factor" postulated to release Rab proteins from RabGDI prior to nucleotide exchange (33).
After Rac is converted to the GTP form, two events take place. The first is the release of RhoGDI (Step 3 ) . Rac in the GTP-bound form has a lower affinity for RhoGDI than does Rac(GDP) (15, 31), and this may be sufficient when the two proteins are present in equimolar amounts to cause dissociation of the proteins. In addition, the generation of various lipid mediators during cell activation may contribute to dissociation of the complex. We have shown that arachidonic acid, phosphatidic acid, and various phosphatidylinositols have the ability to disrupt Rac.GDI complexes, and that this effect is more evident when Rac is in the GTP-bound form (20). The released Rac(GTP1 will then bind to membrane target proteins involved in activation of the NADPH oxidase andor the actin cytoskeleton (Step 4 ) . Rac must be in the GTP-bound state to stimulate 0; formation or actin assembly (1, 5 , 25). Upon termination of the response, Rac would be converted back to the inactive GDPbound state through the action of various GAPS (Step 5). The Rac(GDP) would then have a high affinity for free RhoGDI and would re-associate to leave the membrane and return to the cytosol where it can again begin another cycle of activation.
I t has been shown that cells can differentially regulate functions controlled by Rho uersus Rac. Our data establish that such regulation is likely to occur through the activation of GTP-binding protein-specific guanine nucleotide exchange proteins. These exchange factors appear to be present in the plasma membranes of human neutrophils, and in membranes from other cell types as The localization of exchange proteins at the membrane may mediate the quantitative transfer of Rac/Rho/CDC42 to the membrane, where these regulatory proteins can then control the relevant protein targets which lead to NADPH oxidase activation, actin assembly, etc. Regulation of-activation in conjunction with membrane targeting appears to be a general mechanism to control small GTPbinding proteins whose cellular localization is determined by interaction with GDIs, as a similar process has recently been described for the Rab proteins involved in vesicular transport (32)(33)(34). Membrane transfer of Rabs from soluble Rab.RabGD1 complexes was correlated with exchange of GTP for GDP, and RabGDI appears to contribute to the process of specific membrane targeting of Rab proteins (34). We are currently attempting to identify membrane-associated guanine nucleotide exchange factors active on the Rho family, with the ultimate goal of understanding how they are regulated by the activation of cell surface receptors, including those for growth factors, chemoattractants, and integrins.