Specific Recognition of Rac2 and Cdc42 by DOCK2 and DOCK9 Guanine Nucleotide Exchange Factors*

Recognition of cognate Rho GTPases by guanine-nucleotide exchange factors (GEF) is fundamental to Rho GTPase signaling specificity. Two main GEF families use either the Dbl homology (DH) or the DOCK homology region 2 (DHR-2) catalytic domain. How DHR-2-containing GEFs distinguish between the GTPases Rac and Cdc42 is not known. To determine how these GEFs specifically recognize the two Rho GTPases, we studied the amino acid sequences in Rac2 and Cdc42 that are crucial for activation by DOCK2, a Rac-specific GEF, and DOCK9, a distantly related Cdc42-specific GEF. Two elements in the N-terminal regions of Rac2 and Cdc42 were found to be essential for specific interactions with DOCK2 and DOCK9. One element consists of divergent amino acid residues in the switch 1 regions of the GTPases. Significantly, these residues were also found to be important for GTPase recognition by Rac-specific DOCK180, DOCK3, and DOCK4 GEFs. These findings were unexpected because the same residues were shown previously to interact with GTPase effectors rather than GEFs. The other element comprises divergent residues in the β3 strand that are known to mediate specific recognition by DH domain containing GEFs. Remarkably, Rac2-to-Cdc42 substitutions of four of these residues were sufficient for Rac2 to be specifically activated by DOCK9. Thus, DOCK2 and DOCK9 specifically recognize Rac2 and Cdc42 through their switch 1 as well as β2–β3 regions and the mode of recognition via switch 1 appears to be conserved among diverse Rac-specific DHR-2 GEFs.

Rac and Cdc42 are extensively studied Rho family small GTPases (1)(2)(3)(4). The Rho family comprises at least 25 GTPases that regulate numerous biological processes (5) such as cell migration (6 -8), cell cycle progression (9), gene expression (10), innate immunity (11), and bacterial and viral infections (12)(13)(14). Rho GTPases have been described as molecular switches that cycle between active and inactive states (15). They are active when bound to guanosine triphosphate (GTP) and become inactive in the guanosine diphosphate (GDP)-bound state following GTP hydrolysis. Activation involves the exchange of GDP for GTP (15). This rate-limiting step provides a basis for regulation of small GTPases by guanine-nucleotide exchange factors (GEFs) 2 (1, 16 -18). GEFs control the function of small GTPases through at least two mechanisms. On one hand they catalyze the GDP/GTP exchange activity of Rho GTPases in response to appropriate activation signals. On the other hand they dictate the spatial and temporal contexts in which the activation of GTPases occurs, and thereby contribute to the specificity of Rho GTPase signaling pathways (16, 19 -24).
Two main GEF families for Rho GTPases, Dbl (diffuse B-cell lymphoma)-related and CDM (CED5, DOCK180, and Myoblast City)/DOCK180-related, have been described. The Dbl family of GEFs was the first to be identified and is the larger family with about 70 members in humans (19 -25). Dbl-related GEFs share a catalytic domain of about 200 amino acids called the Dbl homology (DH) domain. Detailed information on how GEFs interact with Rho GTPases is available for Dbl-related but not CDM GEFs. Determination of the three-dimensional structures for a number of DH domains, alone or in complex with their cognate Rho GTPases, has provided insights into how Rho GTPases are specifically recognized by GEFs as well as the mechanism of the exchange reaction. These structural studies have also revealed that divergent amino acid residues in the ␤2-␤3 regions of Rac and Cdc42 are the major determinants of specific interactions with DH-domain containing GEFs (26 -29).
The CDM family of GEFs has 11 members in humans with additional homologues across eukaryotes (30,31). The four subfamilies of CDM proteins share a catalytic domain of about 500 amino acids referred to as DOCK-homology-region 2 (DHR-2), Docker, or CDM-Zizimin-homology 2 (CZH2) domain (30 -32). Members of one subfamily include DOCK2 and DOCK180, which activate Rac to mediate cellular processes such as chemotaxis and phagocytosis (6,24,33). Interestingly, the physiological activation of Rac by these GEFs requires their association with a regulatory protein ELMO1 (Engulfment and Motility) (6,33,34). Another distantly related subfamily of CDM GEFs includes DOCK9, also called Zizimin1, which specifically activates Cdc42. DOCK9 has been shown to induce filopodia formation, however, the exact physiological roles of this subfamily have not been defined (35). To date, no information regarding elements in Rac and Cdc42 that mediate * This work was supported by United States Public Health Service Grant AI-42561 (to J. S.). 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 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Fig. S1. 1 To whom correspondence should be addressed. Fax: 516-367-8369; E-mail: skowrons@cshl.edu.
specific interactions with CDM GEFs is available. Here, we identify the main determinants in Rac2 and Cdc42 that mediate their specific activation by DOCK2 and DOCK9. Interestingly, we found that divergent residues in the switch 1 regions of Rac2 and Cdc42, in addition to divergent amino acids in ␤2-␤3 regions, play critical roles in the specific activation of these GTPases by DOCK2 and DOCK9. These observations suggest that CDM family GEFs and DH family GEFs use different elements to specifically recognize and activate cognate GTPases.
GST Pull-down Assay-5-g aliquots of GST-myc-GTPases were incubated with extracts prepared from human embryonic kidney (HEK) 293T cells expressing DOCK9, DOCK2, and ELMO1 in 500 l of LB supplemented with 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, and 1% Triton X-100. 5 l of packed glutathione-conjugated agarose beads were added and samples were rotated overnight at 4°C. Beads were washed twice with ice-cold LB, incubated for 10 min at 80°C in Lae-mmli buffer, separated by SDS-PAGE, and analyzed by immunoblotting for the myc epitope.
p21-binding Domain (PBD) Pull-down Assay-HEK 293T cells were transiently transfected with a total of 20 g of plasmid DNA (1-2 g of myc-GTPase alone or together with myc-DOCK2 and myc-ELMO1 (2 g each) or myc-DOCK9 (10 g) expression vectors and the appropriate amount of an empty pCG vector) using the calcium phosphate coprecipitation method as described previously (14). tag-Rac2, or tag-Cdc42, was used as an internal control. Twenty-four hours post-transfection, cells were washed with ice-cold phosphate-buffered saline and lysed in 1 ml of LB buffer supplemented with 10 mM MgCl 2 , 1% Triton X-100 and EDTA-free complete protease inhibitor mixture (Roche). Lysates were cleared by centrifugation at 14,000 ϫ g in an Eppendorf centrifuge (5415C) for 10 min at 4°C. 1 ml of cleared extract was rotated for 1 h at 4°C with 20 g of GST-PBD of p21-activated kinase 1 bound to glutathione-conjugated agarose beads (14). Samples were analyzed as described above. Chemiluminescent signals were quantified using FluorChem Imaging System and software (Alpha Innotech, Cannock, United Kingdom).
GTP␥S Activation Assays-Aliquots of detergent extracts containing 50 g of proteins were supplemented with 10 mM EDTA and incubated with, or without, 0.1 mM GTP␥S at 25°C for 15 min. The nucleotide exchange reaction was terminated by adding MgCl 2 to 10 mM, placing samples on ice, and subjecting them to PBD pull-down assays.
Immunoblot Analysis-Proteins were separated by SDS-PAGE, transferred onto polyvinylidene fluoride membrane and the membrane incubated overnight at 4°C in blocking buffer (5% Carnation powdered milk and 1% bovine serum albumin in TBS containing 0.1% Tween (TBS-T)) containing anti-myc 9E10 monoclonal antibody. Membranes were washed in TBS-T and incubated with horseradish peroxidase-conjugated goat anti-mouse IgG (Jackson ImmunoResearch Laboratories, Inc., Pennsylvania, PA) for 2 h at 4°C. Membranes were washed in TBS-T, treated with enhanced chemiluminescence (ECL) detection reagents (GE Healthcare), and exposed to an x-ray film, or imaged using FluorChem Imaging System.

Two Elements in the N-proximal Regions of Rac2 and Cdc42
Mediate Activation by DOCK2-ELMO1 and DOCK9-To address how Rho GTPases are recognized by their cognate CDM family GEFs, we studied the interactions of Rac2 and Cdc42 with DOCK2 and DOCK9. DOCK2 activates Rac but not Cdc42, whereas DOCK9 activates Cdc42 but not Rac (6,35,38). We focused on DOCK2 and DOCK9 because they belong to the subfamilies of CDM proteins that are most divergent from each other. The amino acid sequences of the catalytic DHR-2 domains of DOCK2 and DOCK9 share 19% amino acid identity. To assess how Rac2 and Cdc42 selectively interact with DOCK2 and DOCK9, a series of chimeric Rac2/Cdc42 proteins were generated based on the locations of amino acid differences in their sequences (Fig. 1). The critical chimeras had Rac2 and Cdc42 amino acid sequences swapped at positions 39 between the switch 1 region and ␤2 strand, and/or at positions 58 between the ␤3 strand and switch 2 region (Fig. 1).
Wild-type Rac2 and Cdc42 proteins and their chimeras were transiently expressed alone or together with DOCK9, or DOCK2 and ELMO1, in human embryonic kidney (HEK 293T) cells. As internal controls wild-type Rac2, or Cdc42, proteins tagged with a 30-amino acid long N-terminal peptide tag (tag-) were used. This modification permitted the control and test GTPases to be separated by SDS-PAGE. All proteins were tagged at their N termini with the myc epitope to facilitate their detection by immunoblotting. Detergent extracts were prepared from transfected cells and the GTP-bound forms of the GTPases were precipitated with the PBD of p21-activated kinase 1 fused to glutathione S-transferase (PBD-GST) and immobilized on glutathione-agarose beads. Total GTPases present in detergent extracts and PBD-bound GTPases were resolved by SDS-PAGE, detected by immunoblotting for the myc epitope, and quantified by direct imaging of the chemiluminescent signals.
The results of experiments to identify Cdc42 sequences that mediate specific activation by DOCK9 are shown in Fig. 2A. Expression of DOCK9 strongly stimulated the binding of Cdc42 to PBD (compare lanes 8). In contrast, no such effect was seen when DOCK9 was coexpressed with Rac2 (compare lanes 1). Thus the ability of DOCK9 to specifically activate Cdc42 was reproduced in a transient expression assay in HEK 293T cells. Interestingly, a chimeric GTPase comprising amino acids 1 to 58 of Cdc42 joined to amino acids 59 to 192 of Rac2 (C58R) was activated by DOCK9 almost as well as wild-type Cdc42 ( Fig. 2A, histogram, compare lanes 7 with 8). In contrast, the reciprocal R58C chimera was severely defective for activation by DOCK9 (compare lanes 6 with 7 and 8). These observations indicated that divergent amino acids in the N-proximal 58 residues of Cdc42 are instrumental for its activation by DOCK9. Interestingly, chimeric proteins with Rac2 and Cdc42 sequences swapped at positions 39, such as C39R and R39C58R, were severely defective for activation by DOCK9 (compare lanes 3 and 4 with 7 and 8). As a control, chimeric GTPases were loaded with GTP␥S in cell extracts and tested for their abilities to bind PBD in pull-down assays. These experiments revealed that all GTPases retained binding to PBD (see supplemental Fig. S1A). Together these observations suggested that at least one divergent amino acid residue located in the ␤1 strand and/or switch 1 region as well as at least one in the ␤2-␤3 region of Cdc42 mediate specific activation of this GTPase by DOCK9.
Similar experiments were performed to identify elements in Rac2 that mediate specific activation by DOCK2. The test and control GTPases were transiently co-expressed with DOCK2 and ELMO1 in HEK 293T cells. ELMO1 was included in these experiments because this protein was reported to enhance the ability of DOCK2 to activate Rac through binding to DOCK2 and thereby relieving the inhibition of its catalytic domain function (39). As shown in Fig. 2B, Rac2, but not Cdc42, was strongly activated by co-expression of DOCK2 and ELMO1. Significantly, the R58C chimera was activated by DOCK2-ELMO1 to about 50% of the level seen with the wild-type Rac2 (histogram, compare lanes 6 with 1 and 8), whereas the reciprocal chimera C58R was not responsive (compare lanes 7 with 6 and 8). These observations suggested that divergent amino acid residues located in the 58-amino acid long N-proximal region of Rac2 are instrumental for activation by DOCK2-ELMO1. Notably, the R39C and C39R58C chimeras were severely defective for activation by DOCK2-ELMO1 (compare lanes 2 and 5 with 1 and 6). Together, the above observations indicated that divergent amino acid residues located in the ␤1 strand and/or switch 1 and in ␤2-␤3 region of Rac2, or Cdc42, cooperate to mediate specific activation of these GTPases by their cognate CDM family GEFs.
Divergent Amino Acid Residues in the Switch 1 Region and the ␤3 Strand of Rac2 Mediate Specific Activation by DOCK2-To identify amino acid residues in Rac2 that mediate specific activation by DOCK2-ELMO1, divergent residues in the N-terminal region of this GTPase were substituted with the corresponding residues in Cdc42 (Rac2-to-Cdc42 substitutions). The mutant GTPases were tested for their abilities to bind to and be activated by DOCK2-ELMO1 as well as by DOCK9.
The effects of Rac2-to-Cdc42 amino acid substitutions on the ability of Rac2 to bind DOCK2-ELMO1 and DOCK9 were assessed in pull-down assays. Wild-type and mutant myctagged Rac2 proteins were expressed in E. coli as GST fusions (GST-myc-GTPases). Purified GST-myc-GTPases were incubated with extracts prepared from HEK 293T cells ectopically expressing myc-tagged DOCK9, DOCK2, and ELMO1 and protein complexes precipitated with glutathione-agarose beads. The precipitates were analyzed by immunoblotting for the myc epitope. As shown in Fig. 3A, wild-type GST-myc-Rac2 precipitated DOCK2 and ELMO1 but not DOCK9 (lane 1). Conversely, wild-type GST-myc-Cdc42 precipitated DOCK9 but not DOCK2 or ELMO1 (compare lanes 1 and 10). Thus, these data demonstrate that the specific binding of Rac2 and Cdc42 to their cognate CDM family GEFs is reproduced in pull-down assay.
Interestingly, the pull-down assays identified three of the 14 divergent amino acid residues in the N-terminal region of Rac2 as major determinants of specific recognition by DOCK2-ELMO1. Rac2-to-Cdc42 substitution of alanine 27 (Ala-27) or glycine 30 (Gly-30), located in the N-proximal region of switch 1, or of tryptophan 56 (Trp-56), located in the ␤3 strand, severely compromised the ability of Rac2 to precipitate DOCK2 and ELMO1 (compare lanes 2, 3, and 7 with lane 1). We also tested the effects of substitutions of valine 51 (Val-51) and asparagine 52 (Asn-52) in the ␤3 strand, which we found to be important for the activation of Rac2 by DOCK2 (see below). However, these changes had little if any detectable effect on the ability of Rac2 to bind DOCK2-ELMO1 (compare lanes 5 and 6 with lane 1). We concluded that divergent amino acid residues Ala-27 and Gly-30 in the switch 1 region and Trp-56 in the ␤3 strand of Rac2 are required for specific binding to DOCK2-ELMO1.
Next, mutant Rac2 proteins were tested for their abilities to be activated by DOCK2-ELMO1. As shown in Fig. 3B, individual Rac2-to-Cdc42 substitutions for Ala-27, Gly-30, or Trp-56 significantly diminished Rac2 activation by DOCK2-ELMO1 (Fig. 3B, histogram, compare lanes 2, 3, and 7 with 1). Strikingly, mutating both Ala-27 and Gly-30 in the switch 1 region rendered Rac2 essentially unresponsive to DOCK2-ELMO1 (compare lane 4 with 1-3). Interestingly, Rac2-to-Cdc42 substitution of Val-51 or Asn-52 in the ␤3 region modestly decreased activation by DOCK2-ELMO1, even though neither of these changes detectably affected Rac2 binding to DOCK2-ELMO1 in pull-down assays (compare lanes 5 and 6 with 1; see Fig. 3A). The negative effects of the V51Y and N52T changes on Rac2 activation were more clearly seen when these substitutions were combined with the W56F change in Rac2-(VN51YT,W56F) (Fig. 3B histogram, compare lane 8 with 5-7). These observations indicated that divergent residues at positions 27 and 30 in the switch 1 region and those at positions 51, 52, and 56 in the ␤3 strand of Rac2 are critical for the specific activation of Rac2 by DOCK2-ELMO1.
Although the Rac2-to-Cdc42-mutated Rac2 proteins did not detectably associate with DOCK9 in pull-down assays, we nev- Quadruple epitope-tagged wild-type Cdc42 (tag-Cdc42) was co-expressed as an internal control. GTP-bound forms of GTPases were precipitated from cell extracts with recombinant PBD-GST immobilized on glutathione-conjugated agarose beads. DOCK9 and total GTPases present in detergent extracts, as well as PBD-bound GTPases were detected by immunoblotting for the myc epitope and quantified by direct imaging of chemiluminescent signals. Lower panels, histogram shows relative increases in PBDbound GTPase in the presence of ectopic DOCK9 quantified as described under "Experimental Procedures." B, activation of chimeric GTPases by DOCK2-ELMO1. Experiments were performed as described in A, except that DOCK2 and ELMO1 were used instead of DOCK9, and tag-Rac2 was used as an internal control. The histogram shows relative increases in PBD-bound GTPase in the presence of ectopically expressed DOCK2 and ELMO1. FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6 JOURNAL OF BIOLOGICAL CHEMISTRY 3091 ertheless, tested their abilities to be activated by DOCK9 in the transient expression assay in HEK 293T cells. As shown in Fig.  3C, Rac2 GTPases with amino acid substitutions in the switch 1 region were not responsive to DOCK9. In contrast, Rac2-(W56F) and Rac2-(VN51YT,W56F) with changes in the ␤3 strand were modestly activated by DOCK9 as well as by DOCK2-ELMO1 (panels B and C, compare lanes 7 and 8 with 1 and 9). We concluded that the W56F substitution in the ␤3 strand of Rac2 allows modest activation by both DOCK9 and DOCK2-ELMO1 GEFs.

Specific Activation of Rac2 and Cdc42 by DOCK2 and DOCK9
Rac2-to-Cdc42 Substitutions in the Switch 1 Region and ␤3 Strand of Rac2 Direct Activation by DOCK9 Instead of DOCK2-ELMO1-We observed that the specific recognition of Rac2 by DOCK2 and ELMO1 is mediated by divergent amino acids in the switch 1 region and ␤3 strand of Rac2 and that the FIGURE 3. Divergent amino acids in switch 1 and ␤3 strand of Rac2 mediate specific activation by DOCK2-ELMO1 and DOCK9. A, Rac2-to-Cdc42 amino acid substitutions in switch 1 and ␤3 strand of Rac2 disrupt binding to DOCK2-ELMO1. GST-myc-Rac2 fusion proteins with the indicated amino acid changes were incubated with detergent extracts from HEK 293T cells ectopically expressing myc-tagged DOCK9, DOCK2, and ELMO1, precipitated with glutathione-conjugated agarose beads, resolved by SDS-PAGE, and the myctagged proteins detected by immunoblotting for the myc epitope. 0.2% aliquot of the extract used in a typical incubation reaction was analyzed as a control (input). B and C, the effects of Rac2-to-Cdc42 amino acid substitutions in switch 1 and ␤3 strand on Rac2 activation by DOCK2-ELMO1 (B) or DOCK9 (C). The indicated GTPases were expressed alone, or together with DOCK2 and ELMO1 (B) or DOCK9 (C) in HEK 293T cells. GTPase activation assays were performed as described in the legend to Fig. 2A. Rac2-(D57N), a dominant negative mutant, was used as a negative control (51). specific recognition of Cdc42 by DOCK9 is mediated by the same general region in this GTPase. Moreover, we found that the Rac2-to-Cdc42 W56F substitution in the ␤3 strand of Rac2 allowed activation of the mutant GTPase by DOCK9. These findings prompted us to examine more closely the roles of divergent amino acids in the switch 1 regions and the ␤3 strands of Rac2 and Cdc42 in recognition by DOCK2 and DOCK9.
First we tested whether combinations of Rac2-to-Cdc42 substitutions that disrupt Rac2 activation by DOCK2-ELMO1 would allow the recognition and activation of the mutant GTPase by DOCK9. In the initial experiment, a mutant Rac2 protein bearing Rac2-to-Cdc42 substitutions of all five divergent amino acid residues found to be important for the activation of Rac2 by DOCK2-ELMO1 was constructed and then tested for its abilities to associate and be activated by DOCK9 in HEK 293T cells. As shown in Fig. 4, the mutant GTPase was activated by DOCK9 to ϳ30% of the level seen with the wildtype Cdc42, even though it was unable to form a stable complex with DOCK9 in pull-down assays (panels A and B, compare lanes 2 with 9). Next, similar analyses were carried out with Rac2 proteins bearing various combinations of the five Rac2to-Cdc42 substitutions in switch 1 and the ␤3 strand. Unexpectedly, we found that omitting the G30A substitution resulted in a modest increase in both binding to and activation by DOCK9 (panels A and B, compare lane 3 with 2 and 9). Thus the combination of four Rac2-to-Cdc42 substitutions comprising the A27K substitution in the switch 1 region and the VN51YT and W56F changes in the ␤3 region appeared to be optimal for activation of the GTPase by DOCK9 (panel B, compare lanes 4 -8  with 3). We also observed that the majority of mutant GTPases showed little to no interaction with full-length DOCK9 as detected by GST pull-down assays although their activation was detected by PBD-binding assay (compare panels A and B). Of note, the mutants formed readily detectable interactions with the DOCK9 DHR-2 domain in GST pull-down assays (data not shown). Thus, the mutant GTPases probably still interacted with the full-length DOCK9, albeit weakly.
The effects of the combined Rac2-to-Cdc42 changes on activation by DOCK2-ELMO1 are shown in Fig. 4C. The Rac2-(A27K,VN51YT,W56F) and other mutant Rac2 proteins, except those lacking the W56F or A27K substitutions, were not detectably activated by DOCK2-ELMO1 (compare lanes  2-6 with 1 and 7-9). We concluded that the combination of Rac2-to-Cdc42 substitutions of residues Ala-27 in the switch 1 region and Val-51, Asn-52, and Trp-56 in the ␤3 strand of Rac2 directs optimal and specific activation in vivo by DOCK9 instead of DOCK2-ELMO1.
Rac2-to-Cdc42 Substitutions in Rac2 Switch 1 and ␤3 Strand Direct Activation by DOCK9 in Vitro-To confirm that DOCK9 specifically recognizes and catalyzes guanine nucleotide exchange on Rac2-(A27K,VN51YT,W56F), an in vitro GDP dissociation assay with recombinant proteins was used. Wild type and selected mutant Rac2 and Cdc42 GTPases, and DOCK2 and DOCK9 DHR-2 domains, were purified from E. coli. The recombinant GTPases were loaded with [ 3 H]GDP, incubated in the presence or absence of recombinant DOCK9, or DOCK2, DHR-2 domains and the GTPase-bound GDP were monitored over time. As shown in Fig. 5, the DOCK2 and DOCK9 DHR-2 domains specifically enhanced GDP dissociation from their cognate GTPases (compare panel 6 with 1 and 2 with 7). As expected, the dissociation of GDP from Rac2-(A27K,VN51YT,W56F) was accelerated in the presence of the DOCK9 DHR-2 domain, whereas the DOCK2 DHR-2 domain did not have such an effect (compare panel 3 with 8). Notably, Rac2-(A27K,VN51YT) that lacked the W56F change was not detectably activated by the DOCK9 DHR-2 domain (panel 4), thus underscoring the crucial role of phenylalanine at position 56. Indeed, the W56F substitution alone allowed a modest activation of the mutant Rac2-(W56F) GTPase by DOCK9 DHR-2 (panel 5). However, Rac2-(W56F) was also activated by the DOCK2 DHR-2 domain (panel 10), consistent with observations from activation assays in HEK 293T cells (see Figs. 3, B and C). In summary, we conclude that four Rac2-to-Cdc42 substitutions of divergent amino acid residues at position 27 in the switch 1 region and 51, 52, and 56 in the ␤3 sheet are sufficient to direct specific activation of the mutant Rac2-(A27K,VN51YT,W56F) GTPase by DOCK9 but not DOCK2.
Cdc42-to-Rac2 Amino Acid Substitutions in Cdc42 Switch 1 and ␤2-␤3 Regions Direct Activation by DOCK2 Instead of DOCK9-Next we characterized the roles of the divergent amino acids in switch 1 and ␤2-␤3 regions of Cdc42 in recognition by DOCK9. Cdc42-to-Rac2 substitutions were introduced at positions 27 and 30 in the switch 1 region and 51, 52, and 56 in the ␤3 strand of Cdc42 and the GTPase was analyzed as described above. As shown in Fig. 6, the combined changes disrupted the ability of Cdc42 to bind DOCK9 in pull-down assays but did not totally abolish the ability to be activated by this GEF (panels A and B, for example, compare 2 with 1 and  10). Thus, the mutant GTPase probably retained a residual  FEBRUARY 8, 2008 • VOLUME 283 • NUMBER 6 interaction with the GEF that was sufficient to catalyze nucleotide exchange, even though it was not readily detectable in GST pull-down assays. Analysis of mutant GTPases with additional Cdc42-to-Rac2 substitutions in the ␤2 strand (AVT41SAN) and the loop connecting the ␤2 and ␤3 strands (IGGE46VDSK) revealed that at least some of these changes are required to further suppress activation by DOCK9 (panel B,  compare lanes 3 with 2 and 4 -9).

Specific Activation of Rac2 and Cdc42 by DOCK2 and DOCK9
Next, we assessed the effect of Cdc42-to-Rac2 substitutions on activation by DOCK2-ELMO1 in HEK 293T cells. As shown in Fig. 6, amino acid changes at positions 27, 30, 51, 52, and 56 allowed a weak activation of the mutant Cdc42 by DOCK2-ELMO1, whereas the GTPase was unable to form a stable complex with this GEF in pull-down assay (panels A and C, compare lanes 2 with 1 and 10). Significantly, additional Cdc42-to-Rac2 substitutions in the ␤2 strand and the loop connecting the ␤2 and ␤3 strands allowed specific binding and activation of the mutant GTPase by DOCK2-ELMO1, but not DOCK9 (panels A-C, compare lanes 3 with 2 and 4 -9).
To confirm the above findings, wild-type Rac2 and Cdc42, and the Cdc42 mutant that was the least responsive to DOCK9 (see Fig. 6, lane 3 in panel B, referred to as Cdc42-(C3 R)) were analyzed in an in vitro GDP dissociation assay. As shown in Fig.  7, the DOCK2 DHR-2 domain modestly enhanced GDP dissociation from Cdc42-(C 3 R) (panel 3), whereas this effect was not seen with the DOCK9 DHR-2 domain (panel 6). Thus, evidence from both in vitro and in vivo assays indicates that Cdc42-to-Rac2 substitutions in the N-proximal region of switch 1 and the ␤2-␤3 region of Cdc42 allow specific activation by DOCK2 instead of DOCK9.
Divergent Amino Acid Residues in the Switch 1 Region of Rac Mediate Specific Activation by Diverse DHR-2 GEFs-Our finding that both DOCK2 and DOCK9 recognize their cognate GTPases via divergent amino acid residues in their switch 1 regions suggested that alanine 27 and glycine 30 play a larger role in specific recognition of Rac by DHR-2 GEFs. To assess this possibility we asked how substitutions at these and other divergent N-proximal positions affect specific recognition of Rac by DHR-2 domains of DOCK180, DOCK2, DOCK3, and DOCK4 (38,40,41).
Wild-type Rac2 and Cdc42 proteins, Rac-(A3T,A27K,G30S,I33V) variant with Cdc42-to-Rac2 amino acid substitutions at positions 3, 27, 30, and 33, and Rac-(A3T,I33V) were tested for their  abilities to be activated by the DHR-2 domain of DOCK180, DOCK2, DOCK3, or DOCK4 in HEK 293T cells. As shown in Fig. 8, all four DHR-2 domains activated Rac2 without having a detectable effect on Cdc42. Notably, substitutions at all four divergent N-proximal positions rendered the GTPase almost completely unresponsive to DOCK180, DOCK2, and DOCK3, and only somewhat responsive to DOCK4. Furthermore, restoration of Rac2-specific alanine 27 and glycine 30 residues in switch 1 restored to a large degree activation by DOCK180, DOCK2, and DOCK3 and enhanced the activation by DOCK4 to a level seen with wild type Rac2 (compare lanes 3 with 2). These observations underscore the importance of the divergent amino acids in the switch 1 region for Rac recognition by DHR-2 GEFs.

DISCUSSION
GEFs contribute to the specificity of Rho GTPase signaling by recognizing cognate GTPases and coupling GTPase activation to downstream signaling pathways (16,18,19,21,(42)(43)(44). To gain insight into how CDM GEFs recognize their cognate GTPases, we studied interactions of Rac2 and Cdc42 with DOCK2 and DOCK9. Using a series of chimeric and pointmutated GTPases in binding and activation assays, two elements in the N-terminal regions of these proteins were found to be important for specific activation by their cognate CDM GEFs. One element comprises divergent amino acid residues at positions 27 and 30 in the conformationally flexible switch 1 regions and the other element consists of three divergent residues at positions 51, 52, and 56 in the ␤3 strands. The side chains of these critical amino acid residues, except for those at positions 51, are surface accessible and thus may contact the cognate CDM GEFs directly (28,29,45).
Previous studies have revealed that DH domain GEFs discriminate between Rac and Cdc42 based on their ␤2 and ␤3 strands (45,46). Similar to DH domain GEFs, our findings indicate that both DOCK9 and DOCK2 recognize cognate GTPases through their ␤3 strands. In contrast, both CDM GEFs also recognize divergent residues in switch 1 regions, which are usually not utilized by DH family GEFs. Thus, CDM/DHR-2 and DH domain GEFs use both common and distinct structural elements in Rac and Cdc42 to discriminate between the GTPases.
Even though both the DOCK9/DOCK2 and the previously studied DH domain GEFs recognize GTPases through their ␤3 strands, they probably do so differently. For example, previous studies demonstrated that the DH domain GEFs distinguish between Rac and Cdc42 mainly via an aromatic residue at position 56 (47,48). This is best illustrated by the observation that a Rac-to-Cdc42 substitution of the tryptophan residue at position 56 of Rac1 with phenylalanine (W56F) allows Rac1-(W56F) to be specifically activated by the otherwise Cdc42specific DH domain GEF ITSN1, but not the Rac-specific DH domain GEF Tiam1 (48). Conversely, the reciprocal substitution in Cdc42 allows Cdc42-(F56W) to be specifically activated by Tiam1 (47). In contrast, our observations indicate that whereas the hydrophobic residues at positions 56 are also important for the recognition of the GTPases by DOCK2 and DOCK9, they are not sufficient to allow these GEFs to fully discriminate between Rac and Cdc42. This is evident from the observations that phenylalanine substitution for Trp-56 allows Rac2-(W56F) to be activated by DOCK9, an otherwise Cdc42specific GEF, however, this change does not prevent activation of the mutant Rac2-(W56F) by the Rac-specific DOCK2. Thus, whereas the aromatic amino acid residue at position 56 in Rac or Cdc42 is sufficient for specific recognition by the Dbl-related GEFs Tiam1 and ITSN1 (45,47,48), additional elements in these GTPases are required for their specific recognition by DOCK180-related GEFs DOCK2 and DOCK9.
Our observations indicate that the switch 1 region is important for the recognition of Rac by diverse Rac-specific DHR-2 GEFs and Cdc42 by DOCK9. This observation has interesting implications. Since the divergent residues at positions 27 and 30 in the switch 1 regions have not been implicated in specific recognition of these GTPases by any of the studied DH domain GEFs it seems that DH and DHR-2 GEFs evolved to probe distinct regions to discriminate between small GTPases such as Rac and Cdc42. Also, the crystal structures of DH domain GEFs such as Tiam1, ITSN1, and Dbs in complex with their cognate GTPases show that positions 27 and 30 of the Rho GTPases remain accessible in the complexes (28,29,45). Moreover, previous structural and biochemical studies indicated that divergent amino acids at positions 27 and 30 in the switch 1 region of Rac are important for specific recognition and recruitment of its downstream effector p67 phox , a component of the NADPH oxidase (49,50). p67 phox is not a downstream effector of wildtype Cdc42, however, Cdc42-to-Rac substitutions at positions 27 and 30 (K27A and S30G) in Cdc42 allows p67 phox to serve as an effector of the resulting mutant Cdc42 (49,50). This evidence and our findings together indicate that the divergent amino acid residues at positions 27 and 30 in the switch 1 regions of Rac and Cdc42 can mediate functional interactions with both downstream effectors (e.g. p67 phox ) as well as upstream regulators (e.g. DOCK2 and DOCK9), and also suggests that CDM GEFs may influence the recruitment of specific downstream effectors by restricting their access to the cognate GTPase.