CdGAP, a Novel Proline-rich GTPase-activating Protein for Cdc42 and Rac*

Cdc42 mediates several signaling pathways leading to actin reorganization, transcriptional activation, and cell cycle control. Mutational analysis of Cdc42 has revealed that actin reorganization and transcriptional activation are induced through independent signaling pathways. The Y40C effector mutant of Cdc42 no longer interacts with many of its known target proteins, such as p65PAK and WASP, yet this mutant can still induce filopodia formation. To identify Cdc42 targets involved in actin rearrangements, we have screened a yeast two-hybrid cDNA library using the Y40C mutant of Cdc42 as a bait. We report here the identification of a novel serine- and proline-rich GTPase-activating protein, CdGAP, which is active in vitro on both Cdc42 and Rac. Microinjection of CdGAP into serum-starved fibroblasts inhibits both platelet-derived growth factor-induced lamellipodia and bradykinin-induced filopodia mediated by Rac and Cdc42, respectively. CdGAP does not show in vitro activity toward Rho, and it has no effect on lysophosphatidic acid-induced stress fiber formation when microinjected into fibroblasts. The carboxyl terminus of CdGAP reveals potential protein kinase C phosphorylation sites and five SH3 binding motifs. Thus, CdGAP is a novel GAP that is likely to participate in Cdc42- and Rac-induced signaling pathways leading to actin reorganization.

Members of the Rho family of small GTPases have been implicated in diverse cellular functions, including the reorganization of the actin cytoskeleton, cell growth control, transcriptional regulation, and membrane trafficking (1). Like all GTPases, they act as molecular switches to control downstream cellular events. The interconversion of the inactive GDP-bound form into the active GTP-bound form is regulated by members of the Dbl family of guanine nucleotide exchange factors (2), whereas inactivation of the GTP-bound form is stimulated by the RhoGAP family of GTPase-activating proteins (GAPs) 1 (3). In addition, Rho GTPases interact with guanine nucleotide dissociation inhibitors, whose precise function is unclear (1).
Three members of the Rho family, Rho, Rac, and Cdc42, have been shown to play a crucial role in regulating the organization of the actin cytoskeleton in response to extracellular stimuli (4). Activation of Rho, Rac, and Cdc42 in quiescent Swiss 3T3 fibroblasts induces the assembly of filamentous actin into stress fibers, lamellipodia, and filopodia, respectively (5)(6)(7)(8). In addition to these effects on the actin cytoskeleton, Rho GTPases are important regulators of signaling pathways leading to transcriptional activation. Several groups have demonstrated that Rac and Cdc42 (and in some cells Rho) can activate the c-jun amino-terminal/stress-activated mitogen-activated protein kinase and the p38/HOG mitogen-activated protein kinase cascade (9 -14). All three GTPases are also able to stimulate serum response factor-and NF-B-dependent transcription (15)(16)(17). Finally, Rho, Rac, and Cdc42 have each been implicated in cell cycle control; when introduced into quiescent Swiss 3T3 fibroblasts, all three GTPases induce G 1 cell cycle progression and subsequent DNA synthesis, whereas inhibition of Rho, Rac, or Cdc42 prevents serum-induced DNA synthesis (18,12). Although the signals mediating the effects of Rho GTPases on cell cycle progression are not known, their ability to stimulate G 1 progression correlates well with actin polymerization and not with c-jun amino-terminal kinase activation (19).
In an attempt to understand the biochemical mechanisms underlying the various cellular effects induced by these GTPases, there has been much interest in identifying and characterizing target (effector) proteins. To date, around 10 distinct targets for each of the three Rho GTPases have been identified (1). A mutational analysis of Rho, Rac, and Cdc42 that selectively disrupted their binding with downstream effectors has given some insight into which interactions are essential for specific downstream cellular effects induced by these GTPases (19 -22). Among the multiple Rac and Rho targets, the mutational analysis of Rac has revealed that the only interaction with POR1 or p160 ROCK correlates well with actin reorganization (20,19). On the other hand, stress fiber formation induced by Rho likely involves interaction with both p160 ROCK and a second unknown effector (22). An effector site mutation, Y40C, in Cdc42 prevents its interaction with two known targets, p65 PAK and WASP, yet this mutant can still induce filopodia formation (19). It can be concluded from these experiments that the interaction of Cdc42 with p65 PAK or WASP is not essential for actin reorganization, although the experiments do not exclude the possibility that p65 PAK or WASP are involved in filopodia formation. More recently, a WASP-related protein, N-WASP, has been identified (23). N-WASP still interacts with the Y40C mutant of Cdc42, and when expressed in cells, appears to synergize with Cdc42 in the formation of filopodia (24).
In an attempt to identify Cdc42 targets that might be involved in actin rearrangement, we have used the Y40C mutant of Cdc42 as a bait in a yeast two-hybrid screen. We report here the identification of a novel serine-and proline-rich GTPaseactivating protein, CdGAP (Cdc42 GTPase-activating protein).
CdGAP is active on both Cdc42 and Rac but not Rho. The full-length protein has a predicted molecular mass of 89,610 Da and is rich in both charged amino acids and serine residues, and it has potential protein kinase C phosphorylation sites and SH3 binding motifs.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screen-A Ras-transformed NIH 3T3 cDNA library fused to the GAL4 activation domain in the pGAD10 vector (a kind gift of Dr. C. C. Kumar, State University of New York, Stony Brook, NY) was screened using as a bait Cdc42L61C40 fused to the GAL4 DNA binding domain of the pYTH6 vector. The bait was stably integrated into the genome of the Saccharomyces cerevisiae strain Y190, as described previously (25). Approximately 10 7 yeast colonies were screened for their ability to grow on selective medium containing 25 mM 3-amino-1,2,4-triazole. The 49 fastest growing clones were replated, and plasmids were rescued using the Wizard clean-up kit (Promega) and retransformed into the Cdc42L61C40 yeast strain. Five clones were positive in the plate lift assay for expression of the LacZ reporter gene after the second round of transformation. The 1.9-kilobase pair insert of clone 2 was sequenced and found to be a novel cDNA potentially encoding a 72-kDa protein. Sequencing was performed using Sequenase sequencing kit (Amersham Pharmacia Biotech).
Cloning of the Full-length Clone 2-A full-length sequence corresponding to clone 2 was isolated using the CLONTECH Marathon RACE polymerase chain reaction kit (mouse whole embryo cDNA library) with Advantage Klentaq polymerase. To perform a 5Ј RACE, a primer was generated corresponding to the minus strand of bp 507-529 of the original yeast clone (5Ј-GGAGGTTTGGGGCCCACACCAGG-3Ј). This primer was used with the CLONTECH adaptor primer AP2 in the standard touchdown polymerase chain reaction protocol as described in the CLONTECH manual. Analysis of the 5Ј-RACE reaction by agarose gel electrophoresis revealed a major product at 620 bp. The product was ligated into pGEX4T-3 after BglII/NotI digestion and sequenced. To perform a 3Ј-RACE, a primer corresponding to bp 1631-1657 of the original clone (5Ј-GGACACCAAGCCAGAACCTGAAGTCCC-3Ј) was used with the CLONTECH adaptor primer AP2 following the same procedure used to perform 5Ј-RACE. Analysis of the 3Ј-RACE reaction by agarose gel electrophoresis revealed a major product at 842 bp. The polymerase chain reaction product was directly ligated into pCR2.1 vector (Invitrogen), and sequence analysis provided the completed open reading frame.
Northern Blot-A mouse multiple tissue Northern blot from CLON-TECH was incubated with a radioactive DNA probe corresponding to a HindIII fragment (bp 870 -1303) of the original clone following the procedures detailed by the manufacturer. The HindIII fragment of clone 2 was labeled with [␣ 32 P]dCTP using the multiprime DNA labeling systems from Amersham Pharmacia Biotech.
GAP Assay-Recombinant Rho, Rac, and Cdc42 (5 g) were preloaded with [␥- 32  The mixture was incubated at room temperature, and 5 l samples were removed at 0, 5, 10, and 15 min, diluted in 1 ml of ice-cold buffer A (50 mM Tris-HCl (pH 7.5), 50 mM NaCl, 5 mM MgCl 2 ), and filtered through nitrocellulose filters (prewetted with buffer A). Filters were washed with 10 ml of cold buffer A, dried, and counted.
Cell Culture and Microinjection-Swiss 3T3 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum and antibiotics and maintained at an atmosphere of 10% CO 2 . For the preparation of subconfluent, serum-starved Swiss 3T3 cells, cells were plated at a density of 1 ϫ 10 5 into 60-mm dishes. After 7 to 10 days, cells were serum-starved for 16 h in Dulbecco's modified Eagle's medium containing 2 g/liter NaHCO 3 , trypsinized for 1 min, and resuspended in serum-free medium containing 0.5 mg/ml soybean trypsin inhibitor (Sigma). The cells were pelleted and resuspended in serum-free medium before replating at a density of 6 ϫ 10 4 onto glass coverslips treated overnight with fibronectin (50 g/ml). Cells were allowed to attach and spread for 1-2 h before microinjection. Cells FIG. 1. Interaction of clone 2 with Cdc42L61, RacL61, and RhoL63. Yeast strains containing integrated GAL4 DNA binding domain plasmids with Cdc42L61C40 or Cdc42L61A37 effector mutants (A) or Cdc42L61, RacL61, or RhoL63 (B) were transformed with pACTII (Ϫ), pACTII-p50rhoGAP, or pGAD10-clone 2 and allowed to grow for 2 days at 30°C. Colonies were repeated in the presence of 50 mM 3-amino-1,2,4-triazole and allowed to grow for 2 days at 30°C.

FIG. 2. Sequence of mouse CdGAP.
A, protein sequence of full-length CdGAP. The RhoGAP domain is underlined, the glutamic acid/lysine-rich region is doubleunderlined, and the proline-rich regions are in bold. The yeast clone 2 corresponds to amino acids 3-662. B, homology between CdGAP and RhoGAP domains. A BLAST homology search program (NCBI) aligned a region of CdGAP with regions of proteins that display a RhoGAP domain. Identical residues are in bold. Conserved residues are indicated by a ϩ. C, sequence alignment of the proline-rich regions of CdGAP with SH3 binding motif consensus sequence. The critical arginine (in class II consensus) and prolines are in bold; p, proline-preferred; X, nonconserved residues. treated with bradykinin were plated as described previously (8). Briefly, cells were plated at a density of 1 ϫ 10 5 onto glass coverslips, and the following day, cells were serum-starved for 2 days before microinjection. Eukaryotic expression vector pRK5myc encoding the region of CdGAP encompassing the GAP domain (amino acids 3-662) was microinjected at 0.1 mg/ml into the nucleus of ϳ50 cells over a period of 10 min. Cells were returned to the incubator for a further 2 h before treatment with extracellular factors. During microinjection, cells were maintained at 37°C with an atmosphere of 10% CO 2 .
Growth Factor Treatment and Immunofluorescence-2 h after microinjection of pRK5myc-CdGAP (amino acids 3-662) into subconfluent serum-starved Swiss 3T3 cells, cells were stimulated with 3 ng/ml platelet-derived growth factor for 12 min, 200 ng/ml lysophosphatidic acid for 30 min, or 100 ng/ml bradykinin for 5 min at 37°C and fixed for 10 min in 4% (w/v) paraformaldehyde. All steps were carried out at room temperature, and coverslips were rinsed in phosphate-buffered saline between each of the steps. Cells were permeabilized with 0.2% Triton X-100 for 5 min, and free aldehyde groups were reduced with 0.5 mg/ml NaBH 4 for 10 min. Cells were incubated in the presence of the primary monoclonal antibody anti-Myc diluted in phosphate-buffered saline for 60 min. Coverslips were transferred to a second fluorescein isothiocyanate-conjugated goat anti-mouse antibody (Sigma) and tetramethylrhodamine isothiocyanate-conjugated phalloidin (Sigma) for 30 min. Coverslips were mounted by inverting them onto 8 l of mowiol 4 -88 (Calbiochem) mountant containing p-phenylenediamine (Sigma) as an antibleach agent. After 2 h at room temperature, cells were examined on a Zeiss axiophot microscope using Zeiss 63 ϫ 1.4 oil immersion objective. Fluorescence images were recorded on Kodak T-MAX 400 ASA film.

RESULTS
Yeast Two-hybrid Screen-To identify target proteins of Cdc42 potentially involved in the control of actin assembly, a yeast two-hybrid mouse cDNA library was screened using the Y40C effector mutant of constitutively activated Cdc42L61 as bait. This Cdc42 effector mutant is no longer able to interact with at least two known Cdc42 targets, p65 PAK and WASP, yet can still induce filopodia when introduced into fibroblasts. Approximately 10 7 clones were screened as described under "Experimental Procedures" and the 49 fastest growing clones picked from the selection plates. Five clones tested positive upon rescreening and were negative in the absence of the Cdc42L61C40 bait. Fig. 1A shows the interaction of one of the five clones (clone 2) with Cdc42L61C40 mutant. Clone 2 interacts with another effector mutant (F37A) of Cdc42L61, which is also still able to induce the formation of filopodia. Clone 2 binds to Cdc42L61 (i.e. without effector mutations) and to RacL61 but not RhoL63 (Fig. 1B).
Isolation of a Novel Cdc42-interacting Protein with a RhoGAP Domain-Analysis of the five positive clones (1.9kilobase pair insert) revealed an identical DNA sequence with an incomplete open reading frame, since no stop codon was present either at the 3Ј end or the 5Ј end of the insert. Using two primers derived either from a region close to the 5Ј or the 3Ј ends of clone 2, additional upstream and downstream sequences were isolated by performing 5Ј-RACE and 3Ј-RACE protocols. A 5Ј-RACE product was obtained (620 bp), and sequence analysis of this confirmed that the amino terminus was missing from the original clone 2; the first methionine of the now-extended open reading frame is preceded by an in-frame stop codon. A major 3Ј-RACE product was also obtained (842 bp), and sequence analysis of this provided a complete open reading frame. The full open reading frame encodes a protein of 820 amino acids, and the predicted molecular mass is 89,610 Da ( Fig. 2A).
Sequence analysis of clone 2 corresponded to the Stratagene mouse heart partial clone 919612 in the GenBank Expressed Sequence Tag. The amino terminus of clone 2 revealed a striking homology with several GAPs for Rho-like proteins. Fig. 2B shows that the RhoGAP homology domain extends over 146 amino acids, and there is 32% identity between CdGAP and human n-chimerin. Searching through a GenBank data base revealed no additional recognizable domains. However, the region comprising amino acids 229 to 385 of CdGAP reveals 33% identity to a serine-rich, 54.2-kDa protein from Schizosaccharomyces pombe. Indeed, CdGAP has multiple stretches of 2 serine residues extending from the end of the RhoGAP domain to the carboxyl terminus of the protein. In addition, the region extending from amino acids 515 to 615 shows 30% identity with the glutamic acid/lysine-rich carboxyl terminus of neurofilament triplet H protein. In neurofilament proteins, this region is thought to form a charged scaffolding structure suitable for interaction with other neuronal components. The carboxyl-terminal region of CdGAP also contained five prolinerich sequences, potential SH3 binding motifs (28) (Fig. 2C).
Using a commercial mouse tissue Northern blot, a single mRNA at ϳ7.5 kilobases can be seen in all tissues, although heart and lung (lanes 1 and 4, respectively) appear to have much higher levels compared with other tissues (Fig. 3). Spleen and liver (lanes 3 and 5, respectively) reveal little detectable signal; however, at longer exposure low levels of mRNA were observed (data not shown). We conclude that CdGAP is expressed in most organs and tissues but at different levels of expression.
In Vitro Activity of CdGAP-To determine whether CdGAP encodes a functional GAP activity toward Rho GTPases, amino acids 3-662 of CdGAP were subcloned into the pGEX-4T3 E. coli expression vector as described under "Experimental Procedures." The GST fusion protein was eluted from glutathioneagarose beads and migrated at the expected molecular mass of ϳ99 kDa after SDS-polyacrylamide gel electrophoresis gel electrophoresis (data not shown). Fig. 4 shows that the GAP domain of CdGAP is as efficient as p50RhoGAP in stimulating the intrinsic GTPase activity of Cdc42. CdGAP was also active on Rac, but unlike p50RhoGAP, CdGAP showed no activity toward Rho. To confirm that the loss of filter-associated counts resulted from the stimulation of GTP hydrolysis by CdGAP and not from trivial effects such as proteolysis of the GTPases, metal chelation, or nucleotide release, we have incubated the GAP domain of CdGAP with Cdc42 bound to [8-3 H]GTP. Under the same conditions used for the GAP assay, no loss of filterassociated counts was observed (data not shown). Therefore, CdGAP reveals in vitro GAP activity toward Cdc42 and Rac but not Rho, and it seems unlikely that it has some guanine nucleotide exchange factors activity for Cdc42 and Rac.
CdGAP Down-regulates Cdc42 and Rac in Vivo-Ridley et al. (29) showed that p50RhoGAP can stimulate the GTPase activity of Cdc42, Rac, and Rho in vitro, but after microinjection into Swiss 3T3 fibroblasts, p50RhoGAP inhibited Rho-mediated stress fiber formation but not Rac-induced membrane ruffling. It is likely that p50RhoGAP is specific in vivo for Cdc42 and Rho. To assess the specificity of the GAP domain of CdGAP in vivo, a eukaryotic expression vector pRK5 encoding Myc-tagged CdGAP (amino acids 3-662) was microinjected into serumstarved, subconfluent Swiss 3T3 cells, and its effects on actin changes induced by the addition of extracellular agonists was examined. Fig. 5 shows that Rac-dependent, platelet-derived growth factor-induced membrane ruffling and Cdc42-dependent, bradykinin-induced filopodia were completely abolished by microinjection of CdGAP before stimulation. Microinjection of CdGAP had, however, no effect on Rho-dependent, lysophosphatidic acid-induced stress fiber formation. These results demonstrate that CdGAP is able to down-regulate both Cdc42 and Rac in vivo. As shown in Fig. 5E, microinjection of CdGAP into serum-starved cells disrupts the residual cortical actin and, after a longer period of time, cells round up and completely detach from the plate. It appears that basal Cdc42 and/or Rac activity is required to maintain spread and attached cells. DISCUSSION We have identified a cDNA encoding a novel protein, CdGAP, that interacts in vitro with Cdc42 and Rac but not Rho. The full-length cDNA encodes a protein of 820 amino acids with an expected molecular weight of 89,610 Da. According to Northern blot analysis, CdGAP is ubiquitously expressed, although it is relatively higher in heart and lung tissues.
The amino-terminal sequence of CdGAP contains three con- FIG. 5. Inhibition of platelet-derived growth factor-induced membrane ruffling and bradykinin-induced filopodia but not lysophosphatidic acid-induced stress fibers by CdGAP. Serumstarved subconfluent Swiss 3T3 cells were fixed after stimulation with no addition (A and E), 3 ng/ml platelet-derived growth factor (B and F) for 12 min, 200 ng/ml lysophosphatidic acid for 30 min (C and G), 100 ng/ml bradykinin for 5 min (D and H). In panels E to H, cells were microinjected with pRK5myc-encoding CdGAP (3-662) (0.1 mg/ml) 2 h before stimulation. Actin filaments were visualized with fluorescently tagged phalloidin, and injected cells were localized by co-staining with an anti-Myc antibody and by indirect immunofluorescence (not shown). Approximately 50 cells were microinjected per coverslip. sensus sequence boxes found in several GAP proteins active on Rho family GTPases (30). Expression of recombinant CdGAP reveals in vitro GAP activity toward Cdc42 and Rac but not Rho, and this is reflected in the yeast two-hybrid interaction assay. To date, more than 13 proteins containing a RhoGAP domain have been identified in mammalian cells, although one of them, the p85 regulatory subunit of phosphatidylinositol 3Ј-kinase, does not exhibit any GAP activity (1). Several RhoGAP-containing proteins have also been discovered in other eukaryotic organisms such as S. cerevisiae, Drosophila, and Caenorhabditis elegans (31)(32)(33)(34). These proteins all share a related GAP domain that spans 140 amino acids, and their substrate specificity varies, although in vitro at least most display GAP activity toward more than one Rho GTPase. Overexpression of CdGAP in fibroblasts can down-regulate Cdc42 and Rac but not Rho, in agreement with its in vitro activity.
Most proteins in the RhoGAP family contain additional domains and motifs suggesting that they are multifunctional. In addition to the GAP domain, p190, for example, contains an amino-terminal GTP binding domain (35), whereas Bcr contains a Dbl family guanine nucleotide exchange factors domain. 9 of the 13 known mammalian RhoGAP proteins contain proline-rich sequences and are, therefore, likely SH3 binding proteins. p50rhoGAP has been shown to interact with the SH3 domains of c-Src and p85, whereas 3-BP1 interacts with the SH3 domain of the Abl-tyrosine kinase (36,37). The carboxyl terminus of CdGAP reveals five proline-rich sequences containing the potential SH3 binding motif pXPpXP (p, proline-preferred; X, nonconserved residues) (28). Interestingly, the proline-rich sequence at amino acids 550 to 557 corresponds to the class II consensus sequence XPpXPXR found in various proteins such as dynamin, p47 phox or Sos1 (38). Peptide and mutational analysis have revealed that P2, P5, and R7 all contribute to SH3 binding affinity. Whether the SH3-containing protein(s) binds to CdGAP and participates in Rho GTPasesignaling pathways remains to be assessed. The mechanisms by which GAP proteins are recruited to and interact with Rho GTPase-signaling pathways is unknown. One possibility, however, is that the interaction of RhoGAPs with SH3-containing proteins is a mechanism to recruit GAP activity to specific protein complexes. Bem2p, a protein exhibiting GAP activity toward Rho1p in S. cerevisiae, has been shown to interact with Bem1p, a scaffold-like SH3-containing protein needed for bud emergence and mating projection formation, two processes that require asymmetric reorganizations of the cortical cytoskeleton in S. cerevisiae (39).
In addition to accelerating the intrinsic GTPase activity of GTP-binding proteins, it has been proposed that GAPs may also mediate downstream signals. This has been clearly established for phospholipase C ␤, which in addition to being the target of the ␣ subunit of the heterotrimeric G protein, Gq, is also a GAP for ␣q (42). Others have reported that p120RasGAP, in addition to regulating the GTPase activity of Ras, acts as a signaling molecule in its own right (40). There has been one report to date suggesting a signaling function for a member of the RhoGAP family. n-chimerin has GAP activity toward Rac and to a lesser extent Cdc42, and microinjection of its GAP domain into fibroblasts prevents both Rac-and Cdc42-induced cytoskeletal changes (41). Surprisingly, however, microinjection of full-length n-chimerin or a chimerin mutant lacking GAP activity resulted in the induction of lamellipodia and filopodia (41). The formation of these cytoskeletal structures was inhibited by coinjection of dominant negative N17Rac or N17Cdc42, suggesting that n-chimerin acts synergistically with endogenous Rac and Cdc42 activity to promote actin reorganization. So far, CdGAP has been shown to act as a nega-tive regulator of both Rac and Cdc42. Attempts to identify Cdc42 downstream effectors specifically involved in actin rearrangements using the Y40C mutant of Cdc42 as a bait in the yeast two-hybrid screen has not been of particular significance in the case of CdGAP. It will be, however, of great interest to determine whether full-length CdGAP shows any morphological function in addition to down-regulation of Rac and Cdc42 in fibroblasts. A CdGAP mutant with no GAP activity and impaired binding to Rac and Cdc42 or a mutant lacking GAP activity alone might also provide further clues to its cellular role.