Mammalian DOCK180 protein and its orthologues Myoblast City (MBC) and CED-5 in Drosophila and Caenorhabditis elegans, respectively, function as critical regulators of the small GTPase Rac during several fundamentally important biological processes, such as cell motility and phagocytosis. The mechanism by which DOCK180 and its orthologues regulate Rac has remained elusive. We report here the identification of a domain within DOCK180 named DHR-2 (Dock Homology Region-2) that specifically binds to nucleotide-free Rac and activates Rac in vitro. Our studies further demonstrate that the DHR-2 domain is both necessary and sufficient for DOCK180-mediated Rac activation in vivo. Importantly, we have identified several novel homologues of DOCK180 that possess this domain and found that many of them directly bind to and exchange GDP for GTP both in vitro and in vivo on either Rac or another Rho-family member, Cdc42. Our studies therefore identify a novel protein domain that interacts with and activates GTPases and suggest the presence of an evolutionarily conserved DOCK180-related superfamily of exchange factors.
Mammalian DOCK180 was originally identified as a 180 kDa protein that interacts with the proto-oncogene product c-Crk ( Hasegawa et al., 1996). Recent evidence have demonstrated that DOCK180 and its orthologues in Drosophila and Caenorbaditis elegans form an evolutionarily conserved protein family [denoted the CDM family ( Wu and Horvitz, 1998)] that regulates several important biological processes. The Drosophila orthologue of DOCK180, which is known as Myoblast City (MBC), has an essential role in myoblast fusion, dorsal closure and cytoskeletal organization during embryonic development ( Erickson et al., 1997; Rushton et al., 1995). CED-5, which is the C. elegans orthologue of DOCK180, has in turn been implicated in phagocytosis of apoptotic cell corpses and in distal tip cell migration ( Wu and Horvitz, 1998). Human DOCK180 is also involved in cytoskeletal reorganization. Expression of a membrane-targeted form of DOCK180 induces spreading of NIH 3T3 cells, and coexpression of DOCK180 with Crk and its binding partner p130Cas results in accumulation of DOCK180 in focal complexes ( Hasegawa et al., 1996; Kiyokawa et al., 1998b).
Genetic data in both Drosophila and C. elegans strongly suggest that DOCK180 mediates its effects by functioning as an upstream activator of the small GTPase Rac, which is a regulator of actin-based cytoskeleton ( Nolan et al., 1998; Reddien and Horvitz, 2000). In support of the genetic data, overexpression of DOCK180 in mammalian cells has been reported to lead to JNK activation, phagocytosis of apoptotic cells and enhanced cell migration, and all these events can be inhibited by coexpression of a dominant-negative form of Rac ( Albert et al., 2000; Cheresh et al., 1999; Dolfi et al., 1998; Gumienny et al., 2001; Kiyokawa et al., 1998a). Additionally, mice lacking DOCK2, which is a DOCK180 homologue exclusively expressed in hematopoietic cells ( Nishihara, 1999; Nishihara et al., 1999), are deficient in lymphocyte migration and Rac activation in response to chemokines ( Fukui et al., 2001).
The mechanism by which DOCK180 regulates Rac has remained elusive. Overexpression of DOCK180 or DOCK2 in 293 cells leads to GTP loading of Rac, and these proteins also associate with nucleotide-free Rac in cell lysates ( Kiyokawa et al., 1998a; Nishihara et al., 1999; Nolan et al., 1998). It is not known, however, whether DOCK180 itself, or one of its associated protein, interacts with Rac and catalyzes its conversion to GTP-Rac. Like other GTPases, Rac is active when bound to GTP and inactive when bound to GDP. Conversion of the GDP-bound proteins to the active state is catalyzed by guanine nucleotide exchange factors (GEFs) ( Van Aelst and D'Souza-Schorey, 1997). Of note, DOCK180, DOCK2, MBC and CED-5 lack the typical tandem Dbl-homology and Pleckstrin-homology (DH-PH) domains found in most GEFs that are involved in activation of Rho-family GTPases, such as Rho, Rac and Cdc42 ( Schmidt and Hall, 2002). Recently, ELMO1/CED-12 was identified as an upstream regulator of Rac that functions genetically at the same step as DOCK180/CED-5 in engulfment of apoptotic cells and cell migration in C. elegans ( Gumienny et al., 2001; Reddien and Horvitz, 2000; Wu et al., 2001; Zhou et al., 2001). Mammalian ELMO1 was subsequently shown to directly interact and functionally cooperate with DOCK180 in Rac-dependent phagocytosis of carboxylate-modified beads in CHO LR73 cells. ELMO1 lacks any obvious catalytic domains, and when expressed alone in mammalian cells, it fails to have a notable effect on Rac GTP loading in vivo ( Gumienny et al., 2001). It is thus plausible that formation of a multiprotein complex around DOCK180 is required for DOCK180 (or another component in the complex) to interact with and/or to activate Rac.
We report the identification of a region within DOCK180 named the DHR-2 (DOCK Homology Region-2) domain, which directly interacts with nucleotide-free Rac in vitro and induces the GTP loading of Rac both in vitro and in vivo. Furthermore, we have identified several novel homologues of DOCK180 that possess the DHR-2 domain and found that many of them bind to and exchange GDP for GTP on either Rac or Cdc42. Thus, our studies identify a conserved protein domain that directly interacts with and activates GTPases and suggest the presence of a previously unidentified, evolutionarily conserved DOCK180-related superfamily of GEFs.
Materials and Methods
BLAST searches were performed by using the National Center for Biotechnology Information (NCBI) Standard BLAST server (http://www.ncbi.nlm.nih.gov/blast). The sequences coding for the various DOCK180-related proteins identified in the BLAST searches were acquired from the NCBI Center and the Kazuza DNA Research Institute (Japan). Full-length sequence of MOCA (termed here DOCK3) was obtained from D. Schubert and Q. Chen ( Kashiwa et al., 2000). The sequences were aligned using ClustalW. With two exceptions (DOCK5 and DOCK10), all available human, Drosophila and C. elegans DNA clones, although not necessarily full-length, covered sequences that coded for the DHR-2 domain and adjacent amino acids. Equal length amino-acid sequences of these clones were subsequently used for phylogenetic analysis. The amino-acid sequence distances were determined with the PHYLIP 3.5 package. The phylogenetic tree was derived by neighbor-joining analysis applied to pairwise sequence distances by using the Kimura two-parameter method to generate unrooted trees. The final output was generated with TREEVIEW. The individual nodes, or branching points, in the generated tree were examined by bootstrap analysis with 10,000 pseudoreplicates of the data and found to be reliable. The maximum bootstrap value is 100, which denotes a highly significant branching event. This phylogenetic tree analysis formed the basis of our initial classification of the DOCK180-related proteins. DOCK5 and DOCK10 proteins were classified on the basis of their identity scores obtained with ClustalW and on the basis of the phylogenetic analysis carried out with sequences covered by these clones. The various Dictyostelium discoideum gene products were classified on the basis of identity scores. SCANSITE, PFAM and SMART software were used for protein domain identification and analysis ( Bateman et al., 2002; Schultz et al., 2000; Yaffe et al., 2001). Web-based server 3D-PSSM (http://www.sbg.bio.ic.ac.uk/~3dpssm) was used for threading analysis of the various DHR-2 domains ( Fischer et al., 1999). The ClustalW algorithm was used to determine the percentage identity in amino-acid sequence of the various protein domains to the homologous regions in DOCK180.
Plasmids and antibodies
The pCNX-Flag-DOCK180 construct and wild-type and mutant forms of Flag-DOCK2 have been previously described ( Kiyokawa et al., 1998b; Nishihara et al., 1999) and were kindly provided by M. Matsuda. The DHR-2 domain (aas 1111-1636) of DOCK180, the subdomains of DHR-2 (aas 1111-1515, 1111-1395, 1111-1335, and 1335-1515) and the DOCKER domain [aas 1111-1657, as described by Brugnera et al. ( Brugnera et al., 2002)] were amplified by using Flag-DOCK180 as a template and ligated into pGEX4T-1 (Amersham Biosciences) or pcDNA3-Myc. The DOCK180ΔDHR-2 mutant, which contains a deletion of amino acids 1111-1636, was generated by PCR and ligated into pcDNA3.1Zeo. The cDNA for CED-5 was obtained from R. Horvitz, and the cDNAs for KIAA0299, KIAA1058 and KIAA1771 were from the Kazuza DNA Research Institute. The DHR-2 domains of CED-5, DOCK2, DOCK3 (KIAA0299) and DOCK9 (KIAA1058) were amplified from their respective cDNAs and ligated into pGEX4T-1. The DHR-2 domain of DOCK7 (KIAA1771) was amplified from reverse-transcribed RNA isolated from 293-T cells (owing to sequence rearrangements in the original KIAA1771 clone obtained from the Kazuza DNA Research Institute) and subsequently subcloned into pGEX4T-1. The DHR-2 domains of DOCK180, DOCK2 and DOCK9 were also subcloned into the mammalian expression vector pcDNA3-Myc. The full-length cDNA of mouse ELMO1 was obtained from the EST database (GenBank Acc. #AI574349) and subcloned into the pcDNA3-Myc vector. pEBB-ELMO1-GFP vector coding for mouse ELMO1 with a C-terminal GFP-tag was obtained from K. Ravichandran and has been described previously ( Gumienny et al., 2001). The pET28 Vav2 DPC (DH-PH-Cysteine Rich) construct has been described previously ( Abe et al., 2000) and was a generous gift from C. J. Der. The pRK5-Myc-Rac1, -Cdc42 and -RhoA plasmids, as well as the plasmids encoding GST-Rac1 and GST-RhoA were obtained from A. Hall. The pGEX-Cdc42 construct was from J. Sondek. The pGEX PAK-BD construct has been described previously ( Abassi and Vuori, 2002). Anti-DOCK180, anti-Myc, anti-Cdc42 and anti-RhoA antibodies were obtained from Santa Cruz Biotechnologies. The pan-Rac antibody was from Upstate Biotechnology. The anti-Flag and anti-GFP antibodies were purchased from Sigma and Chemicon, respectively.
Cell culture and transfections
COS-1 and HEK 293-T cells were cultured in DMEM supplemented with 10% fetal bovine serum, penicillin and streptomycin (Gibco-BRL). The CHO cell line, subclone LR73, was obtained from P. Gros and grown in Alpha MEM Earle's salts supplemented with 10% fetal bovine serum, penicillin and streptomycin. For transfections, cells were grown to 80-90% confluency in six-well plates. Unless otherwise indicated in the figure legends, each well was routinely transfected with 2 μg of plasmids using the transfection reagents NovaFECTOR (Venn Nova, Inc.) or Lipofectamine 2000 (Gibco-BRL). Biochemical and cell biological studies were carried out 48 hours after transfection.
GTPase binding assays
COS-1 cells expressing Myc-tagged Rac1 or Cdc42 were lysed in a buffer containing 25 mM Tris, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% Triton X-100, 10 mM NaF, 0.5 mM Na3VO4 and the Complete protease inhibitor cocktail (Roche). Clarified cell lysates were incubated with the indicated GST fusion proteins pre-bound to Glutathione sepharose for 90 minutes. The beads were washed extensively with lysis buffer and the precipitated small GTPases were detected by immunoblotting using the appropriate antibodies followed by incubation with HRP-conjugated secondary antibodies and enhanced chemiluminescence analysis (Amersham Biosciences). GST-fusion proteins were expressed and purified for these experiments as described previously ( Côté et al., 1999). In some experiments, in vitro transcription and translation (TnT kit, Promega) were used according to the manufacturer's instructions to generate the various recombinant proteins and to label them with 35S-methionine. Interaction between the labeled DHR-2 domains and the small GTPases was detected by autoradiography.
In vitro GEF assays
A GDP dissociation assay was used in the in vitro GEF experiments as previously described ( Zheng et al., 1995). Purified small GTPases (10 μg) were loaded with [3H] GDP (Amersham) in a 100 μl final volume of loading buffer (final concentrations: 10 mM HEPES, pH 7.5, 100 mM NaCl, 2 mM EDTA, 0.2 mM DTT, 5 μM GDP and 5 μM [3H] GDP) and incubated for 15 minutes at room temperature. The loading reaction was terminated by adding MgCl2 to a final concentration of 5 mM. An aliquot of the loaded GTPases (20 μl) was diluted in a reaction buffer (final concentrations: 10 mM HEPES, pH 7.4, 100 mM NaCl, 10 mM MgCl2, 500 μg/ml BSA, 0.2 mM DTT, 1 mM GTP). To initiate the exchange reaction, GST, the GEF domain of Vav2 (Vav2 DPC construct) or the DHR-2 domains of DOCK180 and related proteins (0.5 to 1 μg) was added to the reaction mixture in a final reaction volume of 110 μl. Aliquots of the reaction mixture (30 μl) were removed at 0, 15 and 30 minute time points and directly added to 1 ml of STOP buffer (10 mM HEPES, pH 7.4, 100 mM NaCl, 5 mM MgCl2). Samples were filtered on nitrocellulose and washed with 5 ml of STOP buffer. The filters were subjected to scintillation counting, and the amount of bound GDP was expressed as a percentage of the 0 minute time point.
The His Vav2 DPC protein was purified for the in vitro GEF assays as previously described ( Abe et al., 2000). The DHR-2 domains of the various DOCK180-related proteins were expressed in BL21 cells and purified on Glutathione sepharose as described above. The eluted proteins were concentrated, and aliquots were digested with thrombin to remove the GST moiety. Of note, the removal of the GST moiety was required for the detection of the GEF activity of the DOCK180 DHR-2 domain and removal of the GST-domain was therefore carried out as a standard procedure in GEF assays involving any GST fusion proteins.
GST-PBD pull-down assays
In vivo GTP loading of Rac and Cdc42 was analyzed as previously described ( del Pozo et al., 2000). Briefly, 293-T or CHO LR73 cells were transfected in six-well plates with the plasmids indicated in the figure legends. 48 hours after transfection, cells were lysed in MLB buffer (25 mM HEPES, pH 7.5, 150 mM NaCl, 1% NP-40, 10 mM MgCl2, 1 mM EDTA and 10% glycerol). The clarified lysates were incubated for 30 minutes with the GST-PAK-PBD fusion protein bound to Glutathione sepharose. The beads were washed extensively with MLB buffer and the bound GTP-loaded Rac and Cdc42 were detected by immunoblotting. Equal amounts of input lysate were analyzed by immunoblotting to verify the expression levels of Rac, Cdc42 and various transfected proteins. GST-PAK-PBD was expressed and purified for these experiments as described previously ( Abassi and Vuori, 2002).
Identification of the DOCK180 superfamily of proteins
In order to gain insight into the function of DOCK180, we performed Standard-BLAST searches using the human DOCK180 protein sequence or fragments thereof as a query. Database searches indicated that DOCK180 belongs to a protein superfamily, which consists of at least 11 human members and several homologous members in Drosophila, C. elegans, D. discoideum, A. thaliana and S. cerevisiae. The sequences were retrieved from public databases, and a phylogenetic tree analysis was performed as described in detail in Materials and Methods ( Fig. 1A). The phylogenetic data demonstrated that the human, Drosophila and C. elegans proteins form four distinct subfamilies within the DOCK180-superfamily ( Table 1, Fig. 1A). Identity scores obtained with ClustalW algorithm supported this classification. For presentation purposes in this report, the 11 human DOCK180 homologues were termed DOCK1 (=DOCK180)-DOCK11 and the four subfamilies were denoted DOCK-A, -B, -C and -D. This nomenclature conforms with that used by Reif and Cyster in a recent review article for DOCK1-DOCK4 ( Reif and Cyster, 2002).
With the exception of DOCK180 and DOCK2 (see Introduction), the other human family members have been poorly characterized and most of them have not been previously identified. DOCK3 was originally characterized as a presenilin-binding protein PBP ( Kashiwa et al., 2000). Very recently, it was renamed MOCA for `modifier of cell adhesion' owing to its ability to modulate cell-substratum adhesion and amyloid-β secretion by an unknown mechanism in nerve cells ( Chen et al., 2002). While this work was in preparation, DOCK9 was identified as a novel protein termed zizimin1 ( Meller et al., 2002) (see later). In Drosophila, four homologues of DOCK180, one member in each subfamily, were identified in our database search. Phylogenetic analysis suggested that MBC is a member of the DOCK-A subfamily, supporting a previous report in which it was proposed that MBC is the DOCK180 orthologue in Drosophila ( Nolan et al., 1998). In C. elegans, three DOCK180 homologues were identified; phylogenetic analysis confirmed the notion that CED-5 is likely to be the orthologue of DOCK180 (and other human DOCK-A/B family members) in C. elegans ( Wu and Horvitz, 1998). To our knowledge, no functional data are available for the other DOCK180 homologues in either Drosophila or C. elegans. Three novel homologues of DOCK180 were found in D. discoideum, whereas A. thaliana and S. cerevisiae have one DOCK180 homologue each. Interestingly, SPIKE1 of A. thaliana was recently demonstrated to function as a modulator of actin cytoskeleton in plant cells ( Qiu et al., 2002).
As shown in Fig. 1B, several potential signaling and protein-protein interaction domains were identified in the DOCK180 superfamily of proteins (as predicted by the PFAM, SMART and SCANSITE programs). Thus, the DOCK-A and DOCK-B family members are characterized by the presence of an N-terminal SH3 domain. Recently, members of the ELMO family of proteins were demonstrated to be direct binding partners for DOCK180 ( Gumienny et al., 2001). Although the molecular mechanism of this interaction is not fully understood, the SH3-domain of DOCK180 has been suggested to play a role in it. Similar to DOCK180, several members of the DOCK-A and DOCK-B subfamilies contain proline-rich regions that are potential binding sites for SH3-domain-containing proteins in the C-terminus. DOCK180 interacts with Crk and another adapter protein, NCKβ ( Tu et al., 2001), via these sites. An N-terminal PH-domain was detected in the members of the DOCK-D subfamily. PH domains are known to bind to differentially phosphorylated phosphoinositides, and they can also participate in protein-protein interactions ( Lemmon et al., 2002). The functional significance of the PH-domain in the DOCK-D proteins is currently unknown.
Pairwise alignment analysis identified two regions of high sequence homology that are conserved throughout the DOCK180 superfamily. These regions were named DHR-1 (DOCK Homology Region-1) and DHR-2. Of note, the homology between the yeast ylr422Wp protein and the other family members is restricted to the DHR-2 region. Also, the similarity between DOCK-A/B and DOCK-C/D subfamilies is largely restricted to the DHR-1 and DHR-2 regions. Sequence alignments of the DHR-1 and DHR-2 domains from representative human members of each subfamily are shown in Figs. 1C and 1D, respectively. PFAM analysis suggested that the DHR-1 domain of DOCK180 is likely to be a C2 domain (E value: 0.47), which is a versatile signaling domain that interacts with lipids in a Ca2+-dependent or -independent manner ( Merithew and Lambright, 2002). The putative C2 domain of DOCK180 is most similar to the C2 domain of VSP34, a phosphatidyl inositol 3′-kinase orthologue in C. albicans, which is predicted to function in a Ca2+-independent manner (Fig. S1A, available at jcs.biologists.org/supplemental) ( Nalefski and Falke, 1996). A C2 domain was also identified in other DOCK-A, as well as in DOCK-B proteins, but not in the members of the DOCK-C and -D subfamilies. Although the existence of the C2 domain would need to be experimentally verified, it could play a role in, for example, localizing DOCK180 to the plasma membrane.
No obvious domains were identified in the DHR-2 region by sequence analysis. Interestingly, however, threading analysis by 3D-PSSM suggested the presence of a DH-domain within the DHR-2 domain of DOCK9/zizimin1 that folds similar to the DH-domain of β-PIX (E-value: 0.635, 50% certainty) (Fig. S1B, available at jcs.biologists.org/supplemental). A tandem DH-PH domain resembling the fold found in known GEFs for Rac and Cdc42, such as SOS, Intersectin and β-PIX, was identified in the DOCK180 DHR-2 domain by threading analysis, albeit not with a significant E-value (data not shown). Of note, the DHR-2 domain of DOCK180 (aas 1111-1636) overlaps with a region in DOCK180 that Matsuda and coworkers found to be necessary for DOCK180-mediated induction of Rac signaling (aas 1472-1714) ( Kobayashi et al., 2001). Although structural analysis will be needed to confirm the potential presence of a GEF-like structure in the DHR-2 domain, these preliminary data prompted us to examine the potential role of the DHR-2 domain in DOCK180-mediated Rac activation.
The DHR-2 domain of DOCK180 binds to and activates Rac in vitro and is necessary and sufficient for DOCK180-mediated Rac activation in vivo
The specificity of GEFs toward Rho GTPases is in part determined by their ability to directly interact with these GTPases ( Gao et al., 2001; Snyder et al., 2002). Following binding, GEFs catalyze nucleotide exchange by destabilizing the strong interaction between the GTPase and GDP and stabilizing the nucleotide-free state. This higher affinity transition state intermediate is then dissociated by binding of GTP. Thus, GEFs can be distinguished from other GTPase-interacting proteins by their ability to bind to the nucleotide-free state of GTPase ( Snyder et al., 2002). We therefore examined whether the DHR-2 domain of DOCK180 could interact with nucleotide-free forms of small GTPases of the Rho-family, including RhoA, Rac1 and Cdc42. To this end, a panel of GST fusion proteins of the DOCK180 DHR-2 domain was generated ( Fig. 2A,B). The boundaries of the highly hypothetical DH and PH domains (with no significant E-value in 3D-PSSM) are denoted in Fig. 2A. As shown in Fig. 2B, the DHR-2 domain of DOCK180 readily interacted with nucleotide-free Rac1 but not with Cdc42 or RhoA. None of the smaller fusion proteins corresponding to the predicted subdomains of DHR-2 were able to precipitate Rac1 or any of the other GTPases tested. These findings do not rule out the possibility that a rudimentary tandem DH-PH domain exists in the DHR-2 domain; regions covering both the DH and PH domains and some of the adjacent sequences of Trio, Dbs or Vav2 are required for these exchange factors to demonstrate maximal GTPase binding or GEF activity ( Booden et al., 2002; Liu et al., 1998; Rossman et al., 2002).
We next investigated whether the DHR-2 domain of DOCK180 possessed GEF activity in vitro. Bacterially expressed and purified Rac1, Cdc42 and RhoA proteins were loaded with [3H]GDP and the ability of the DHR-2 domain of DOCK180 to catalyze the exchange of the labeled GDP for cold GTP was examined. As a positive control, we used the GEF domain of Vav2 since it is known to act as a GEF for these three GTPases ( Abe et al., 2000). As shown in Fig. 2C, we found that the DHR-2 domain of DOCK180 indeed contains GEF activity for Rac1 but not for Cdc42 or RhoA. As expected, the GEF domain of Vav2 readily exchanged the nucleotides for Rac1, Cdc42 and RhoA. No GEF activity was observed within the subdomains of DHR-2 (data not shown). Thus, the capability of the DHR-2 region to interact with nucleotide-free Rac1 correlated with its capability to catalyze nucleotide exchange for Rac1.
Our subsequent studies suggested that the DHR-2 domain of DOCK180 is both necessary and sufficient for DOCK180-mediated Rac activation in vivo in 293-T cells. For these experiments, we generated a mutant form of DOCK180 that lacks the DHR-2 domain (DOCK180ΔDHR-2). In addition, we generated an expression vector for Myc-tagged DHR-2 domain alone and a Myc-tagged subfragment of the DHR-2 domain corresponding to the hypothetical DH domain. As shown in Fig. 2D, and as previously reported ( Kiyokawa et al., 1998a; Kobayashi et al., 2001), overexpression of the full-length DOCK180 construct in 293-T cells resulted in a significant increase in the GTP loading of Rac, as measured by the PBD pull-down assay. The DOCK180ΔDHR-2-construct in turn failed to activate Rac, suggesting that the DHR-2 domain of DOCK180 is absolutely required for DOCK180-mediated Rac activation. Exogenous expression of the isolated DHR-2 domain resulted in an increase in the GTP loading of Rac to about the same extent as was observed when full-length DOCK180 was expressed. Expression of the hypothetical DH domain within the DHR-2 domain alone failed, as expected in the light of the in vitro data, to stimulate the GTP loading of Rac.
As noted above, ELMO1 was recently identified as an upstream regulator of Rac that functions synergistically with DOCK180 in engulfment of apoptotic cells and cell migration ( Gumienny et al., 2001; Wu et al., 2001; Zhou et al., 2001). While the present work was being finalized, Brugnera et al. reported that although DOCK180 directly binds to Rac in vivo, formation of the DOCK180-ELMO1 complex is needed for GTP loading of Rac in vivo ( Brugnera et al., 2002). This conclusion was based in part on the finding that mutant forms of DOCK180 that are incapable of interacting with ELMO1 failed to activate Rac. Thus, Brugnera et al. identified a domain within DOCK180 (denoted Docker) that recognizes nucleotide-free Rac and that can mediate GTP loading of Rac in vitro ( Brugnera et al., 2002). This domain is slightly larger on the C-terminal side than the DHR-2 domain that we have identified here. These authors reported, as data not shown, that expression of the Docker domain alone in 293-T cells failed to stimulate Rac-GTP loading in vivo. Brugnera et al. also reported that coexpression of ELMO1 with full-length DOCK180 greatly enhanced the capability of DOCK180 to activate Rac in 293-T cells and that in LR73 cells DOCK180-induced Rac GTP loading was completely dependent on coexpression of ELMO1.
Our studies above indicated that the DHR-2 domain of DOCK180 is both necessary and sufficient for Rac activation in 293-T cells, and we therefore decided to examine the requirement for ELMO1 in more detail. First, we tested the Docker domain identified by Brugnera et al. in our experimental model, and, consistent with the results obtained with the DHR-2 domain, we found that the Docker domain readily activates GTP loading of Rac in 293-T cells ( Fig. 2D). We then examined the capability of full-length DOCK180 to activate Rac in LR73 cells. As shown in Fig. 3A, expression of DOCK180 alone was sufficient to induce Rac GTP loading in LR73 cells, and coexpression of ELMO1 had no additional effect on the GTP loading. To rule out the possibility that the effect of DOCK180 on Rac activity was saturated in this experiment, we carried out an extensive titration of the DOCK180 plasmid (0.05μ g to 1 μg of plasmid/well) and expressed it in LR73 cells with or without coexpression of ELMO1. As shown in Fig. 3B, expression of DOCK180 resulted in a dose-dependent activation of Rac GTP loading in LR73 cells. Of note, a clear induction in Rac GTP loading was observed when the expression level of the transfected DOCK180 protein was about 1.5—two-fold compared to the endogenous level of DOCK180, suggesting that physiological levels of DOCK180 are sufficient to induce Rac GTP loading. Coexpression of ELMO1 failed to enhance DOCK180-induced GTP loading of Rac at all DOCK180 concentrations tested, including those that resulted in submaximal GTP loading of Rac. Similar results were obtained in 293-T and NIH 3T3 cells and also upon extensive titration of the amount of cotransfected ELMO1 plasmid (data not shown). Of note, we used an ELMO1 construct with an N-terminal Myc-tag in Fig. 3A, whereas Brugnera et al. utilized an ELMO1 construct harboring a C-terminal GFP-tag in their experiments. To exclude the possibility that the N-terminal Myc-tag would somehow affect the function of ELMO1, the ELMO1 construct used by Brugnera et al. was utilized in Fig. 3B. Taken together, our results demonstrate that exogenous expression of DOCK180 at physiological levels is sufficient to induce GTP loading of Rac at least in 293-T and LR73 cells, and coexpression of ELMO1 was not found to be necessary for this activity (see Discussion).
DHR-2 domains of various DOCK180-related proteins bind to nucleotide-free GTPases and have GEF activity in vitro and in vivo
We next explored the possibility that the GTPase-binding and GEF activity function is a general feature of the DHR-2 domains within the DOCK180 superfamily, and we found that this indeed is the case. As shown in Fig. 4A, an interaction between the DHR-2 domain of CED-5 (C. elegans DOCK180) with nucleotide-free human Rac1, but not with Cdc42 or RhoA, was detected, suggesting that the ability of the DHR-2 domain to interact with Rac is evolutionarily conserved within the DOCK-A subfamily. We then examined the capability of the DHR-2 domains of various human DOCK180-related proteins to interact with Rac1, Cdc42 and RhoA under nucleotide-free conditions. As shown in Fig. 4B, GST-DOCK2-DHR-2 (a member of the DOCK-A subfamily) precipitated endogenous Rac from cell lysates to the same extent as the DHR-2 domain of DOCK180. The DOCK2-DHR-2 was found to be highly specific for Rac so that it failed to interact with Cdc42 and RhoA (data not shown). Nucleotide-free Cdc42, but not Rac1 or RhoA, in turn robustly precipitated the in vitro translated DHR-2 domain of DOCK9/zizimin1 (member of the DOCK-D family) ( Fig. 4C). We were unable to detect a significant interaction by the DHR-2 domain of DOCK3/MOCA (member of the DOCK-B subfamily) with Rac1, Cdc42 or RhoA ( Fig. 4B; data not shown). Similarly, DOCK7 (a member of the DOCK-C family) demonstrated a barely detectable interaction with Cdc42 and RhoA, but no interaction with Rac1 ( Fig. 4C). Thus, these data suggest that some (but not all) of the DOCK180 family members tested are able to interact with Rac1 or Cdc42 via their DHR-2 domains.
We next used both in vitro and in vivo approaches to determine whether the various DHR-2 domains contained GEF activity. First, DHR-2 domains of DOCK2, DOCK3/MOCA, DOCK7 and DOCK9/zizimin1 were produced as GST fusion proteins and tested for their ability to exchange GDP for GTP on Rac or Cdc42 in vitro. As shown in Fig. 5A, the DHR-2 domain of DOCK2 exhibited clearly detectable GEF activity for Rac1. DHR-2 domains of DOCK3/MOCA, DOCK7 and DOCK9/zizimin1 in turn failed to demonstrate any GEF activity toward the GDP-loaded Rac1. These findings correlated well with the capabilities of these DHR-2 domains to interact with nucleotide-free Rac1 in vitro (see Fig. 4). As shown in Fig. 5B, the DHR-2 domains of DOCK2, DOCK3/MOCA and DOCK7 had no effect on the nucleotide-loading status of Cdc42. By contrast, the DHR-2 domain of DOCK9/zizimin1 was found to contain specific GEF activity for Cdc42. Thus, the capability of the DHR-2 of DOCK9/zizimin1 to catalyze nucleotide exchange on Cdc42 correlated well with its ability to interact with Cdc42 under nucleotide-free conditions ( Fig. 4).
Mammalian expression vectors encoding full-length DOCK2, three deletion mutants of DOCK2 and the isolated DHR-2 domains of DOCK2 and DOCK9/zizimin1 (see Fig. 5C) were next transfected in 293-T cells, and the in vivo GTP-loading status of Rac and Cdc42 was analyzed by the PBD pull-down assay as above. As shown in Fig. 5D, full-length DOCK2, DOCK2 (939-1854) and the isolated DHR-2 domain of DOCK2 were found to be potent activators of Rac. Two C-terminal deletion constructs of DOCK2, which either disrupt or eliminate the DHR-2 domain, in turn failed to activate Rac. Thus, these results demonstrate that the DHR-2 domain of DOCK2, similar to that of DOCK180, is both necessary and sufficient for DOCK2-mediated Rac activation in vivo. As expected, we failed to detect enhanced GTP-loading of Rac when overexpressing the DHR-2 domain of DOCK9/zizimin1. Somewhat surprisingly, the DHR-2 domain of DOCK9/zizimin1 also failed to activate Cdc42, despite the fact that it was highly expressed in the transfected cells and it robustly activated Cdc42 in vitro. Recently, Meller et al. reported on identification of zizimin1 (DOCK9), and they also found that expression of a region of zizimin1 that corresponds to DHR-2 was not sufficient to induce Cdc42 activation in vivo ( Meller et al., 2002). Expression of full-length zizimin1 resulted in activation of Cdc42 in vivo ( Meller et al., 2002), suggesting that additional sequences within DOCK9/zizimin1 might be needed to stabilize the DOCK9/zizimin1-Cdc42 interaction in vivo or to target the protein to an appropriate subcellular localization.
We report here on the identification of a domain termed DHR-2 within DOCK180 that binds to and activates the small GTPase Rac in vitro and that is both necessary and sufficient for DOCK180-mediated Rac activation in vivo. Our additional findings indicate that the DHR-2 domain and its functional capabilities are evolutionarily conserved within a superfamily of DOCK180-related proteins. Our studies therefore identify a novel protein domain that interacts with and activates GTPases and suggest the presence of a previously unidentified DOCK180-related superfamily of GEFs.
While our manuscript was being prepared, Brugnera et al. reported that, similar to what we find here, a domain within DOCK180 specifically recognizes nucleotide-free Rac and mediates GTP loading of Rac in vitro ( Brugnera et al., 2002). In contrast to our findings, these authors reported that in cells binding of DOCK180 to Rac alone is insufficient for GTP loading, and a DOCK180-ELMO1 interaction is required. We in turn found that mutant forms of both DOCK180 ( Fig. 2) and DOCK2 ( Fig. 5) that probably fail to interact with ELMO proteins readily promote GTP loading of Rac in vivo. At present, the reason for the discrepancy between our findings and those by Brugnera et al. remains unclear, since the same constructs and cell lines were used in both studies. We note that we utilized a lipid-based transfection method in our studies, which in our hands gives approximately 50-70% transfection efficiency in 293-T cells. Brugnera et al. in turn utilized calcium phosphate precipitation; we have observed only 10% transfection efficiency of 293-T cells with this method (data not shown). The differences in the transfection methods may explain the different results obtained when the various DOCK180 constructs were expressed alone, but it remains unclear why coexpression of the ELMO1 construct would yield different results in the two studies. Owing to the lack of specific antibodies, we have not been able to determine the expression levels of endogenous ELMO proteins; it is possible that if ELMO1 does potentiate DOCK180-mediated Rac activation, saturating levels of ELMO1 may be present in the 293-T and LR73 cells we have used in our experiments. Thus, although our results with the isolated DHR-2 domains suggest that binding to ELMO1 is not an absolute requirement for DOCK180- and DOCK2-mediated Rac activation in vivo, we can not rule out the possibility that under different conditions, ELMO1 enhances the capability of DOCK180 to activate Rac.
It has been reported previously that coexpression of DOCK180 with ELMO1 and Crk is required for Rac-dependent cellular events, such as actin reorganization ( Gumienny et al., 2001). Also of note, genetic evidence strongly supports an indispensable role for all three proteins, DOCK180, Crk and ELMO1, in proper Rac signaling in vivo ( Gumienny et al., 2001; Wu et al., 2001; Zhou et al., 2001). We favor a hypothesis in which DOCK180, via its DHR-2 domain, is mainly responsible for the binding and GTP loading of Rac in vivo, whereas other protein-protein interactions by DOCK180 would either potentiate this activation and/or mediate appropriate subcellular localization of the DOCK180 complex, which in turn would be a requirement for proper Rac signaling. In other words, DOCK180 alone may be sufficient for activation of Rac as monitored by GTP loading, but it may not be sufficient for activation of Rac signaling. Previously, del Pozo and coworkers demonstrated that both GTP loading of Rac and membrane localization of activated Rac are needed for Rac to activate downstream signaling events ( del Pozo et al., 2000). In support of this, we have recently found that the adapter protein Crk mediates membrane translocation of GTP-loaded Rac, and this event is required for active Rac to functionally couple to its effectors, such as PAK, and to activate downstream signaling pathways leading to actin reorganization ( Abassi and Vuori, 2002). Thus, the role of Crk, and perhaps also ELMO1 via its PH domain, may be to mediate membrane localization of the DOCK180-Rac complex. Clearly, more work is needed to be able to fully understand the functional regulation of DOCK180 signaling.
We also report here on the identification of a novel, evolutionarily conserved superfamily of DOCK180-related proteins, members of which are likely to function as GEFs for various GTPases. Our results indicate that members of the DOCK-A subfamily ( Table 1) function as GEFs for Rac. Interestingly, some subfamilies are likely to be GEFs for other GTPases, such as members of the DOCK-D subfamily for Cdc42. Meller et al. recently reported on purification and identification of a novel protein termed zizimin1, which functions as a GEF for Cdc42 ( Meller et al., 2002). Importantly, zizimin1 is identical to the gene product termed DOCK9 in this report, which is a member of the DOCK-D subfamily. Similar to us, Meller et al. concluded that DOCK9/zizimin1 belongs to a family of DOCK180-related proteins. By using a biochemical approach, these authors identified a minimal region within zizimin1 that interacts with Cdc42 and named this domain CZH2; interestingly, this domain is identical to the DHR-2 domain that we identify in this report as being a conserved domain in all DOCK180-related proteins. Meller et al. were able to demonstrate that the CHZ2-domain of zizimin1, when immunoprecipitated from mammalian cells, contained GEF activity toward Cdc42 in vitro. These authors were unsuccessful, however, in detecting GEF activity with bacterially produced and purified CHZ2-domain of zizimin1 and therefore were unable to exclude the possibility that another protein with GEF activity was present in the zizimin1 immunoprecipitates. Our results with purified components conclusively demonstrate that the DHR-2 domain of DOCK9/zizimin1 contains intrinsic and specific GEF activity toward Cdc42 ( Fig. 5). Similar to Meller et al., we found that expression of the DHR-2 domain of DOCK9/zizimin1 was not sufficient to induce the GTP-loading of Cdc42 in vivo. Meller et al. were able to demonstrate a CHZ2/DHR-2-mediated activation of co-expressed SAAX-Cdc42, which is a Cdc42 mutant that is entirely cytoplasmic but does not bind to the inhibitory protein Rho-GDI. Thus, it is possible that membrane localization of DOCK9/zizimin1 and/or removal of Rho-GDI from Cdc42 by an unknown mechanism is required for DOCK9/zizimin1 to activate Cdc42. Our preliminary data support the latter possibility, as we found that a myristylated, membrane-targeted form of DHR-2 domain of DOCK9/zizimin1 fails to induce GTP loading of Cdc42 (data not shown).
At present, the functional activity of the members of the DOCK-B and DOCK-C families remains to be determined. We had anticipated that members of the DOCK-B subfamily, such as DOCK3/MOCA, would function as GEFs for Rac owing to their high sequence homology to DOCK-A subfamily. Likewise, we expected that DOCK7, a member of the DOCK-C subfamily, could potentially have the same specificity as DOCK9/zizimin1 owing to the similarities between their DHR-2 domains. Our studies indicated, however, that the DHR-2 domains of DOCK3/MOCA and DOCK7 are likely to lack binding and catalytic activities towards Rho, Rac1 and Cdc42. Nevertheless, we can not rule out the possibility that regions outside of the DHR-2 domain would be required for these proteins to interact with Rho-family GTPases or any other GTPases. Chen and coworkers have reported that exogenous expression of DOCK3/MOCA has a significant effect on cell-substratum adhesion in nerve cells ( Chen et al., 2002), and it is therefore tempting to speculate that this protein would affect the function of small GTPases known to be involved in the regulation of actin cytoskeleton or cell adhesion receptors, such as integrins. These possibilities are currently being examined.
At present, the molecular mechanisms of GEF activity within the DHR-2 domain are not known. As noted above, we have observed resemblance to a DH-PH domain structure within some of the DHR-2 domains by threading analysis. Also of note, secondary structure predictions demonstrated that the DHR-2 domain is likely to be highly helical, similar to known structures of other GEFs ( Cherfils et al., 1998) (data not shown). The precedent for a `non DH-PH' Rho-family GEF comes from Salmonella typhimurium, in which SopE, which lacks discernible DH-PH-domains, functions as a GEF for Rac ( Hardt et al., 1998; Rudolph et al., 1999). Atomic-level structural studies are likely to yield important information on the catalytic mechanisms and the molecular basis for the GTPase specificities observed for the various members of the DOCK180 superfamily.
We thank K. Becherer, D. Kedra, E. Larsen, and E. Lau for excellent technical support and members of the Vuori and Pasquale labs for helpful discussions. We are grateful to C. J. Der and M. Booden for critical advice and gift of reagents. We also thank P. Gros, A. Hall, R. Horvitz, M. Matsuda, J. Sondek and K. Ravichandran, as well as the Kazuza DNA Research Institute for the various cDNAs and reagents used in this study. The study was supported by grants from NIH (to K.V.). J.-F. C. is a recipient of a post-doctoral fellowship from the Canadian Institute of Health and Research (CIHR).
- Accepted October 11, 2002.
- © The Company of Biologists Limited 2002