A conserved binding motif defines numerous candidate target proteins for both Cdc42 and Rac GTPases.

Rho, Rac, and Cdc42 are small GTPases that regulate the formation of a variety of actin structures and the assembly of associated integrin complexes, but little is known about the target proteins that mediate their effects. Here we have used a motif-based search method to identify putative effector proteins for Rac and Cdc42. A search of the GenBankTM data base for similarity with the minimum Cdc42/Rac interactive binding (CRIB) region of a potential effector protein p65PAK has identified over 25 proteins containing a similar motif from a range of different species. These candidate Cdc42/Rac-binding proteins include family members of the mixed lineage kinases (MLK), a novel tyrosine kinase from Drosophila melanogaster (DPR2), a human protein MSE55, and several novel yeast and Caenorhabditis elegans proteins. Two murine p65PAK isoforms and a candidate protein from C. elegans, F09F7.5, interact strongly with the GTP form of both Cdc42 and Rac, but not Rho in a filter binding assay. Three additional candidate proteins, DPR2, MSE55, and MLK3 showed binding to the GTP form of Cdc42 and weaker binding with Rac, and again no interaction with Rho. These results indicate that proteins containing the CRIB motif bind to Cdc42 and/or Rac in a GTP-dependent manner, and they may, therefore, participate in downstream signaling.

mammalian cells, the Rho subfamily members control the polymerization of actin and the assembly of focal complexes at the plasma membrane in response to extracellular signals (3,6). For example, microinjection of Rho into serum-starved Swiss 3T3 cells rapidly stimulates stress fiber and focal adhesion formation (7), while Rac induces membrane ruffles (8) and Cdc42 induces the formation of filopodia (9). In addition to their effects on the actin cytoskeleton, Rho GTPases also have a role in regulating kinase signaling pathways. For example, Rho, Rac, and Cdc42 stimulate a novel nuclear signaling pathway leading to transcriptional activation of the serum response element (10). Rac and Cdc42, but not Rho, have also been shown to activate the c-Jun amino-terminal kinase (JNK) signaling pathway leading to c-Jun transcriptional activation (11,12). The mechanisms by which the Rho subfamily of GTPases regulate these apparently diverse biological processes is still not clear.
A large number of mammalian nucleotide exchange factors (GEFs), related to the yeast exchange factor Cdc24 have been identified including dbl, vav, ost, ect2, lbc, and Tiam1, and they may provide tissue specificity or receptor-specific activation of Rho family members (2,3). In addition, over 10 mammalian GTPase-activating proteins (GAPs) 1 for Rho family members have been described including BCR, p50rhoGAP, chimaerin, ABR, p190-A, p122, and myr-5 (13). The ability of these GAP proteins to interact with GTPases in a GTP-dependent manner suggests that in addition to being negative regulators, they may also act as effector proteins. The multidomain nature of many of these GAPs further supports this notion; for example, in addition to the GAP domain, p190 has a GTPase domain (14), p122 interacts with phospholipase ␦ (15), and the myosin family member, myr5, binds to actin (16).
The identification of effector proteins for Rho-related GTPases is the first step toward defining their biological activity, and a number of candidate proteins have already been reported. The target protein involved in Rac-mediated activation of the NADPH oxidase complex in phagocytic cells has been identified as p67 Phox (17). Although the Rac interactive site has been mapped to the amino-terminal 199 amino acids of the protein, no significant similarities have so far been found between this sequence and other sequences in the data base. A tyrosine kinase containing a SH3 domain, p120 ACK , has been shown to bind specifically to Cdc42 in a GTP-dependent manner, although its biological role is unclear (18). Another protein, the serine/threonine kinase p65 PAK , binds to both Cdc42 and Rac (but not Rho) in a GTP-dependent manner, but again it is not known whether p65 PAK mediates any of the described biological effects of these GTPases (19,20). Interestingly, p120 ACK and p65 PAK do have some sequence similarities in a region outside the kinase domain that represents an interactive site for Cdc42/Rac GTPases (19).
In this report, we have localized the Cdc42 binding site of a murine p65 PAK isoform to a minimal conserved region of 16 amino acids. Using this small protein motif, we have searched the GenBank TM data bases and identified 25 potential Cdc42and/or Rac-binding proteins. In vitro binding assays confirm that several of these proteins bind to Cdc42 and/or Rac in a GTP-dependent fashion. Not all of these newly identified proteins are kinases, suggesting a role for other types of proteins in downstream signaling events mediated by the Rho family proteins.
Filter Binding Assay for p21 Proteins-Protein-protein interactions were visualized using a dot-blot assay. GST-fusion proteins were spotted onto a nitrocellulose filter which was subsequently incubated with radiolabeled GTPase. The various proteins (1 g) were spotted onto nitrocellulose (Schleicher & Schuell, BA 85) and the filter was blocked for 2 h at room temperature with 5% dried milk powder. Recombinant L63 rhoA, L61 cdc42 (G25K isoform), or L61 rac1 (0.5 g) was incubated with 10 Ci of [␥-32 P]GTP (6000 Ci/mmol) for 10 min at 30°C in 30 l of 50 mM Tris, pH 7.5, 5 mM EDTA, and 0.5 mg/ml bovine serum albumin. Nucleotide exchange was stopped on ice by adding MgCl 2 to 10 mM. The nitrocellulose filter was washed twice with buffer A (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5 mM MgCl 2 , 0.1 mM dithiothreitol) and then incubated in 10 ml of buffer A containing 5% dried milk containing the [␥-32 P]GTP-bound protein. After incubation for 5 min at 4°C, the membrane was washed three times with cold buffer A containing 5% dried milk and autoradiographed to visualize the bound GTPase.
GDP-and GTP-dependent Binding-The nucleotide requirement for binding of each of these proteins was tested using proteins immobilized to glutathione-agarose resin similar to that described previously (17). Approximately 5 g of p65 PAK -␣ and -␤, FO9F7.5, DPR2, MSE55, and GST, and 10 g of both MLK-3 and Syn-1 were used in these binding assays. The nucleotide exchange reaction was performed with either 10 Ci of [ 3 H]GTP or [ 3 H]GDP (7 Ci/mmol; Amersham) essentially as above except with 0.5 g of recombinant V12cdc42. Following the nucleotide exchange reaction, the labeled GTPases were added to the glutathione beads containing the different proteins for 5 min on ice with occasional mixing. The tubes were then microcentrifuged for 20 s, and the supernatant was discarded. Following four washes with buffer A, the washed resin was subjected to liquid scintillation counting. The binding to each of the beads was expressed as a percentage of input counts.

RESULTS AND DISCUSSION
Identification of a Minimal Cdc42 Binding Site within p65 PAK -Previously, a 40-amino acid region of rat p65 PAK was shown to contain the binding site for Cdc42 and Rac and reported to have some sequence similarity with two other Cdc42-binding proteins, p120 ACK and STE20 (19). Cloning of two murine p65 PAK isoforms, p65 PAK -␣ (the murine homologue of rat p65 PAK ) and p65 PAK -␤ revealed that they both contain high sequence conservation around this binding site. 2 The murine p65 PAK -␣ and -␤ isoforms were expressed as GST-fusion proteins and, using a filter binding assay, were shown to bind to Rac and Cdc42, but not to Rho as expected (Fig. 1). A truncated p65 PAK -␤ construct, GST-p65 PAK -␤-⌬3, continued to show strong Cdc42 binding, although binding to Rac was reduced ( Fig. 1). The ability of p65 PAK -␤-⌬3 to bind to Cdc42/Rac and its limited amino acid homology with the two other Cdc42/ Rac-binding proteins p120 ACK and STE20 suggested that a region of approximately 16 rather than 40 amino acid residues (corresponding to amino acid residues 74 -89 within rat p65 PAK -␣) might be sufficient for binding to Cdc42/Rac. Identification of Proteins Containing a CRIB Binding Motif-In order to identify potential new binding proteins for Cdc42 and Rac, we performed an iterative search of the Gen-Bank TM protein data base using the BLASTp algorithm (26). The initial query sequence with the conserved Cdc42/Rac binding site in rat p65 PAK -␣ (residues 74 -89) revealed significant matches to numerous other proteins shown in Fig. 2. Sites within these proteins along with the corresponding sequences in p120 ACK and STE20 were then used to research the data base to identify additional proteins containing this motif. Finally, a FASTA (27) search of GenBank TM with each of the potential proteins turned up partial cDNAs for additional proteins. In total, over 25 distinct proteins from a variety of organisms were identified that contain a potential Cdc42 binding motif (Fig. 2). We propose that this motif be termed CRIB for Cdc42/Rac interactive binding motif.
The CRIB motif occurs in both kinases and non-kinase proteins. For example, one of the CRIB-containing proteins, DPR2, is a Drosophila tyrosine kinase (21), which shows similarity in the kinase domain to pp125 FAK , c-Abl, and p120 ACK . Several serine/threonine kinase including at least 3 isoforms of p65 PAK and two human and three yeast kinases also contain a potential CRIB motif. In addition, several members of the MLK family of serine/threonine kinases also contain a protein sequence resembling a CRIB motif. Two different yeast kinases encoded by X82499 and Z48149, contain identical CRIB motifs. Non-kinase CRIB-containing proteins include a human sequence MSE55 (and two related genes, accession numbers T06431 and T75138/F12871), WASP, a human gene responsible for Wiscott-Aldrich syndrome (28), three different C. elegans genes (F09F7.5, T2365.3, and B0280.2) and two potential yeast genes (P38785 and D9740.18).
CRIB Proteins Bind to Cdc42 and/or Rac-Several of the proteins containing the CRIB motif were subcloned into the pGEX-4T-3 bacterial expression vector to test the prediction that these proteins would interact with Cdc42 and/or Rac. Fusion proteins were purified by glutathione affinity chroma-2 P. D. Burbelo, D. Drechsel, and A. Hall, manuscript in preparation. FIG. 1. Delineation of a minimal Cdc42 binding region in p65 PAK . GST fusion proteins were produced and tested for p21 binding using the filter binding assay described under "Materials and Methods." The location of the 16-amino acid core CRIB sequence, also found in STE20 and p120 ACK is denoted by the black box.
Target Proteins for Cdc42 and Rac GTPases 29072 tography following induction by isopropyl-1-thio-␤-D-galactopyranoside. These proteins included the Drosophila tyrosine kinase DPR2, human MSE55, and human MLK3 and an uncharacterized C. elegans protein, F09F7.5, encoded by the cDNA cm12 g10. In addition, a synthetic 17-amino acid fragment (Syn-1) and two additional proteins with weaker consensus CRIB motifs were also tested; PLC-␤1 (4/8 match) (24) and MSP (3/8 match) (25). These fusion proteins were assessed for binding to Cdc42, Rac, and Rho using a filter binding assay and GST-rhoGAP (which binds to all three GTPases) was used as a positive control (29). The ␣ and ␤ isoforms of p65 PAK , FO9F7.5, DPR2, MSE55, and MLK3 proteins, were found to bind to the constitutively activated form of Cdc42 (Leu substituted for Gln at codon 61) (Fig. 3). All six proteins also bind equally well to wild type Cdc42 (data not shown). However, of these, only the p65 PAK proteins, and the C. elegans protein FO9F7.5, bound equally well to Rac and Cdc42, while DPR2, MSE55, and MLK3 showed weaker binding to Rac than Cdc42 (Fig. 3). A short 17-amino acid residue peptide, Syn-1, containing a proline to alanine mutation within the consensus sequence, also showed weak binding to both Rac and Cdc42. None of the CRIB proteins tested showed binding to Rho (Fig. 3). Neither PLC-␤1 nor MSP showed binding to Cdc42, Rac, or Rho (Fig. 3). A summary of the binding characteristics of potential CRIB proteins to Cdc42 and Rac is shown in Fig. 2.
The guanine nucleotide dependence of Cdc42 binding to these proteins was investigated using a glutathione-agarose bead assay. Significant binding of p65 PAK -␣ and -␤, FO9F7.5, MSE55, DPR2, MLK3, and Syn-1 was detected to the GTPbound form of Cdc42 (Fig. 4), but not to the GDP-bound form of Cdc42. The results are consistent with previous reports showing the GTP-dependent binding of p120 ACK (18) and p65 PAK (19,20) and are consistent with these proteins being candidate effectors of Cdc42 and Rac biological activities.
In conclusion, we have delineated a short motif (the CRIB motif) in over 10 distinct proteins that confers binding to the Cdc42 and/or Rac GTPases. At least another 15 proteins have been identified by a data base search that contain a potential CRIB site. The length of the CRIB motif is approximately 16 amino acids containing a region of variable length between the two halves of the binding motif. The CRIB motif contains eight core amino acids with the sequence I-S-X-P-(X) 2-4 -F-X-H-X-X-H-V-G. It is interesting to note that proteins with one or two differences within the core sequence can still show binding to Cdc42/Rac.
Of the proteins shown to contain a functional CRIB motif, only the kinases have a defined biochemical activity. In addition to the SH3-containing kinase p120 ACK , a new Drosophila tyrosine kinase, DPR2, was found to contain the CRIB motif. Although the CRIB domains are found at a relatively similar distance after the kinase domains, DPR2 is unlikely to be the Drosophila homologue of p120 ACK , since it lacks a SH3 domain and the overall amino acid similarity is low. In addition to the p65 PAK family of kinases, a new serine/threonine kinase family, the MLK kinases, were found to contain a potential CRIB motif. To date, there appears to be at least four members of the MLK family including MLK1 (30), MLK2/MST (30,31), MLK3/ SPRK/PTK1 (32,23,33), and DLK (34). All four of these kinases share a characteristic hybrid kinase domain between serine/threonine and tyrosine kinases, although SPRK/MLK3 (23) and DLK (34) have now been shown to possess serine/ threonine catalytic activity. In addition, all four MLK members contain putative ␣-helical leucine zipper motifs COOH-terminal to the kinase domain, although little is known of the function of this region. One might expect that the activity of at least MLK2 and MLK3 kinases will be stimulated following Cdc42/ Rac binding as seen with p65 PAK (19,20), but this has yet to be tested.
The biochemical function of the non-kinases containing the CRIB motif is not clear at all. Some of these proteins also contain proline-rich regions (e.g. F09F7.5 and MSE55) raising the possibility that they might act as adapters and interact with SH3-containing protein(s) or other polyproline-binding proteins such as profilin. and As, Ascaris. Amino acid sequence comparison is shown between the different CRIB proteins. The number of residues matching the strict consensus sequence of eight amino acid residues is also shown. Sequence analysis suggests that C09B8.7 is a C. elegans homologue of p65 PAK . Binding of Cdc42 and Rac determined experimentally is shown as positive (ϩ) or negative (Ϫ). p21 binding to p65 PAK -␣/R, p65 PAK -␥/H, p120 ACK , STE20, and Shk1 were from Refs. 19  The identification of so many potential effectors for Cdc42 and Rac is quite surprising. However, numerous activities have already been ascribed to Rac and Cdc42, and each of these activities may require a distinct effector(s). Rac regulates the assembly of focal complexes and the polymerization of actin in lamellipodia, it has an essential role in Ras-induced cellular transformation and can act as an oncogene in its own right (35), and it regulates the activity of cPLA 2 (36). Rac and Cdc42 can also activate the JNK kinase cascade (11,12) and in Drosophila both Rac and Cdc42 have been shown to be involved in the extension of neuronal growth cones (37). The identification of proteins containing the CRIB motif should aid in dissecting the molecular mechanisms by which Cdc42/Rac GTPases regulate these processes.