Identification and Characterization of hPEM-2, a Guanine Nucleotide Exchange Factor Specific for Cdc42*

Guanine nucleotide exchange factors of the Dbl family regulate the actin cytoskeleton through activation of Rho-like GTPases. At present the Dbl family consists of more than thirty members; many have not been phenotypically or biochemically characterized. Guanine nucleotide exchange factors universally feature a Dbl homology domain followed by a pleckstrin homology domain. Employing data base screening we identified a recently cloned cDNA, KIAA0424, showing substantial sequence homology with Rac activators such as Tiam1, Sos, Vav, and PIX within the catalytic domain. This cDNA appears to be the human homologue of the Ascidian protein Posterior End Mark-2 (PEM-2). We refer to this exchanger as hPEM-2. hPEM-2 encodes a protein of 70 kDa and features an N-terminal src homology 3 domain, followed by tandem Dbl homology and pleckstrin homology domains. The gene is highly expressed in brain and is localized on the human X-chromosome. Employing biochemical activity assays for Rho-like GTPases we found that hPEM-2 specifically activates Cdc42 and not Rac or RhoA. Ectopic expression of hPEM-2 in NIH3T3 fibroblasts revealed a Cdc42 phenotype featuring filopodia formation, followed by cortical actin polymerization and cell rounding. hPEM-2 represents an exchange factor, which may have a role in the regulation of a number of cellular processes through Cdc42.

Similar to Ras proteins, Rho-like GTPases act as molecular switches by cycling from the active GTP-bound state to the inactive GDP-bound state. The Rho family of GTPases now comprises more than ten distinct gene products, which are conserved in all higher organisms studied. The importance of this family was first outlined when they were identified as key regulators of the actin cytoskeleton (1,2). These GTPases have since been implicated in a plethora of signaling events ranging from cell adhesion and motility, transcription activation, cell cycle progression, and cell fate determination (3). Deactivation of Rho-like GTPases is principally regulated by GTPase-activating proteins (GAPs) 1 that stimulate their intrinsic GTPase activity. Downstream signals are mediated by effector proteins, which specifically recognize the conformation of the GTPbound protein. The activation of these GTPases is regulated by proteins referred to as guanine nucleotide exchange factors (GEF proteins) (4). Furthermore, cross-talk between the members of the Rho family has been proposed. Cdc42 and Rac lie in a linear cascade in which Cdc42 activates Rac, whereas Rho in many cell types has opposing effects to Cdc42 and Rac on the actinomyosin cytoskeleton (5)(6)(7). These relationships imply that a complex network of signaling molecules is orchestrated and regulated by small GTPases.
Activators or GEF proteins for Rho-like GTPases are modular proteins that universally feature a catalytic DH-PH combination. In addition they often contain other motifs conserved in signaling molecules such as src homology 2 (SH2), src homology 3 (SH3), and post-synaptic density-disc large-ZO-1 homology (PDZ) domains, suggesting that they participate in multiprotein signaling complexes. Indeed GEF proteins have been implicated in a diverse array of cellular processes including responses to extracellular signals (8 -10), receptor capping and lymphocyte activation (11,12), determination of cellular polarity and morphology (13), and in neuronal guidance during development (14). However, the physiological function of most GEFs is unknown.
We and others have shown that membrane localization of GEF proteins is a prerequisite for their biological activity in a variety of cell types. Membrane localization is mediated by the lipid binding capacity of the PH domain (15,16). Recent studies on the DH domains from SOS, PIX, and Trio have all shown the same fundamental structure, which is entirely ␣-helical (17)(18)(19). The most conserved residues form a putative interface with the GTPase, possibly with the cooperation of the PH domain to stabilize the GTPase in the nucleotide free conformation. This facilitates the ejection of GDP from the active site and its replacement with GTP, which exceeds intracellular GDP by 10-fold.
Tiam1 activity appears specific for Rac in most cell types, and we became interested in identifying other putative Rac activators that might play similar roles in the regulation of cell adhesion and invasion (20 -22). A large number of DH proteins have now been identified through various sequencing projects. We wanted to see if grouping these uncharacterized proteins together according to their primary structures would allow us to predict their specificity toward each of the members of the Rho family. A number of studies have used recent developments in sequence manipulation and analysis to explore enzyme specificity and activity in relation to sequence homologies. Such analysis has already been performed for groups of enzymes such as kinases and phosphatases (23)(24)(25). We have attempted a similar analysis for the Rho family GEF proteins. We used several sequence analysis techniques with the DH domains from a number of well characterized members of the DH family to identify other putative activators of Rac. In this way we selected a GEF, KIAA0424 identified by the Kazusa sequencing project, which featured a strong homology with Tiam1 within the DH domain. A homologue of KIAA0424, termed posterior end mark-2 (PEM-2), has also been implicated in the polarization of early embryos in the Ascidian Ciona savignyi (26). For simplicity we refer to this GEF as hPEM-2. By employing novel biochemical activity assays for Rho-like Sequences were aligned with ClustalX (29). A consensus secondary structure mask was generated from the recent structures of PIX, Sos, and Trio and imposed to infer gap penalties in the production of the alignment. Sequences are all human except for those prefixed with m(Mouse) or r(Rat). Sequences of the form K0424 correspond to KIAA clones identified by the sequencing program at the Kazusa DNA Research Institute. These are mostly full-length cDNAs that have not yet been fully characterized. A consensus substrate specificity is inferred. C, modular layout of DH proteins similar to hPEM-2 in the DH domains. Approximate size is marked in kilodaltons. The accession number for KIAA0424 is AB007884.
GTPases and by determining the morphological effect on the actin cytoskeleton, we have examined the activity of this GEF toward Rho, Rac, and Cdc42.

MATERIALS AND METHODS
Plasmids and Constructions-We obtained pBluescript KIAA0424 from the Kazusa DNA Research Institute, Japan. We generated pBKHA by inserting the sequence CTA GCC ACC ATG GAA CAA AAA CTC ATC TCA GAA GAG GAT CTG GGA GGA G between the NheI and BamHI sites of pBKCMV (Stratagene). pBKHA-hPEM-2 was generated through standard polymerase chain reaction with the oligomers CTG TGG AAT TCA TGA CGT TGC TGA TCA C and AGC TCT CGA GAT CTG CCT CCC TGT AGG TA and then inserted between the EcoRI and XhoI sites of pBKHA. The construction, expression and use of pGST-PAK-CD and pGEX-C21 (pGST-C21) are described elsewhere (27,28). pUTSV-HA-Tiam1C1199 is described elsewhere (15). Dbl cDNA was the kind gift of A. Eva, Instituto G. Gaslini, Genova, Italy. The entire reading frame between the start and stop codons was amplified by polymerase chain reaction and subcloned between the SpeI and XhoI sites of pUTSV (AGA CTA GTC CTC AGG AAA TGT CCA GCG GCC and GAG CTC GAG TCA TCA ATA TAG GAG AGC C). A T7 tag was attached to the 5Ј end of the Dbl reading frame by digesting pUTSV-Dbl with SpeI and inserting the annealed oligomers CTA GTA TGG CCT CGA TGA CAG GTG GCC AAC AGA TGG GTT and CTA GAA CCC ATC TGT TGG CCA CCT GTC ATC GAG GCC ATA. Myc-tagged V14RhoA, V12Rac, and V12Cdc42 in pcDNA3 (Invitrogen) are described elsewhere (5,22).
Cell Culture, Transfection, and Phenotypic Analysis-COS-7 cells were propagated in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. 1.3 ϫ 10 6 COS-7 cells were seeded in 10-cm culture dishes and the following day transfected using the DEAEdextran method (29). 24 h later cells were lysed in an Nonidet P-40 buffer and the GTPase activation state determined with the pull-down assay. Cells were activated with 20 ng/ml PDGF for 5 min or with 5 m LPA for 15 min before lysis. NIH3T3 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% new born calf serum. Cells were seeded onto glass slides and grown overnight. The following day, cells were transfected with LipofectAMINE or Effectine (both from Life Technologies, Inc.) according to the manufacturer's instructions. After 24 h, the cells were fixed with 3.7% formaldehyde, permeabilized, and stained with rhodamine-phalloidin and either antibodies 12CA5 to HA, 9E10 to Myc, or anti-T7 (Novagen) as appropriate. Preparations were analyzed by confocal laser scanning microscopy.
GTPase Pull-down Assay-Recombinant Rac, Cdc42, and Rho were prepared from Escherichia coli using a bacterial expression system as described previously (30). Gpp(NH)p (Sigma), a nonhydrolyzable GTPanalogue, was exchanged for the bound GDP of the respective GTPases as described (31). 2 g of the respective GTPase in GDP-and GppNHpbound form were mixed with 10 g of GST-C21 or GST-PAK-CD in a total volume of 1 ml of a Nonidet P-40 fish buffer and rotated for 30 min (27). The samples were then washed three times, and the bound GTPases were analyzed by SDS-PAGE followed by Western blotting.
To determine cellular activation of Rho, Rac, and Cdc42, cells were harvested by scraping in Nonidet P-40 fish buffer (27). After centrifugation to remove debris, the lysates were rotated for 1 h with 5 l of GSH-Sepharose loaded with about 20 g of either GST-C21 or GST-PAK-CD. The beads were washed three times with the Nonidet P-40 buffer, and then the bound proteins were separated by SDS-PAGE  5. A, the hPEM-2 phenotype resembles that of Cdc42, but not that of Rho or Rac in fibroblasts. NIH3T3 cells were transfected with myc-tagged V14Rho, V12Rac, or V12Cdc42 and with HA-hPEM-2, HA-Tiam1, or with T7Dbl. 24 h later, the cells were fixed and permeabilized, and the transfected cells were stained with anti-myc, anti-HA, or anti-T7 (shown in green). Actin was stained with rhodamine-phalloidin (shown in red).

hPEM-2 Activates Cdc42
before Western blot analysis. Rho proteins were detected with 12C4 monoclonal and Cdc42 with P1 polyclonal antibodies (both from Santa Cruz Biotechnology, Inc.). Rac was detected with 23A8 monoclonal antibody (Upstate Biotechnology, Inc.), and HA was detected with monoclonal antibody 12CA5.
Northern Blotting-Northern blotting was performed according to standard procedures (29), using a pre-prepared poly(A)ϩ RNA blot purchased from Origene Technologies. A 5.4-kilobase insert from pBluescript KIAA0424 was excised with SalI/NotI and was used as template for the probe, which was then hybridized to the blot before washing and analysis by autoradiography, according to the supplied protocols. Glyceraldehyde-3-phosphate dehydrogenase was used as RNA loading control.

RESULTS
Identification of Tiam1-related GEF Proteins-To find GEFs that might activate Rac, we conducted data base searches with the catalytic DH domain from Tiam1. We found that most mammalian DH proteins grouped together according to their sequence homology and their reported specificity for distinct Rho-like GTPases. An uncharacterized expressed sequence tag, named KIAA0424 (hPEM-2), appeared particularly similar to Tiam1 in the DH domain (Fig. 1A). A homologue of KIAA0424 from the Ascidian C. savignyi called PEM-2 was recently identified as a patterning gene expressed during the early stages of development (26), therefore we refer to this gene as hPEM-2. Phylogenetic processing of sequences derived from DH domains of mammalian GEFs was used to classify these GEFs into subfamilies (32,33). The output depicted in Fig. 1B shows that there is a varying degree of similarity between many members of the DH family proteins that infers their specificity. In this analysis there are four principle branches, which are labeled either "mostly Rho," "mostly Rac," "Rho and Cdc42," and "no consensus" according to the published literature (34). The arm that interested us contained most of the reported exchangers for Rac such as Tiam1, Vav, PIX, and Sos.
The similarity of hPEM-2 to Tiam1 in the DH domain, the putative role in embryonic polarization and also the modular similarity to the recently cloned GEF for Rac PIX/COOL prompted us to select it for more extensive analysis. Genomic analysis using the Unigene data base at the National Center for Biotechnology Information localized hPEM-2 to a position 145 cR from the top of chromosome X in humans. The protein encoded by hPEM-2 was predicted to encode a protein of approximately 70 kDa, featuring an N-terminal src homology 3 domain, followed by a DH-PH combination, and finally a proline-rich sequence in the C terminus (Fig. 1C).
Expression of hPEM-2-Northern blotting revealed that hPEM-2 was expressed almost exclusively in the brain, although at longer exposures some signal was also seen in heart tissues. The size of this mRNA ran slightly smaller than the 6-kb marker, suggesting that the 5.4-kb cDNA we obtained initially was full-length ( Fig. 2A). This cDNA showed a single open reading frame containing a start codon, preceded by stop codons in all three frames, also indicating that it was fulllength. To examine the consequences of hPEM-2 expression in cells, we expressed an N-terminally HA-tagged hPEM-2 in COS-7 cells. As predicted from the cDNA sequence, we found that hPEM-2 was expressed as a 70-kDa protein in these cells (Fig. 2B).
GTPase Activation Assays-To investigate the activity and specificity of this exchange factor we devised an assay based on the fact that activated, GTP-bound Rho, Rac, and Cdc42 bind specifically to their downstream effectors. We used the CRIB domain from PAK1B (GST-PAK-CD) to probe for activated Rac and Cdc42 (27) and the REM1 domain from the Rho effector protein Rhotekin (termed GST-C21) to probe for activated Rho (28). Similar assays have previously been used to investigate the activity of small GTPases in response to stimuli, including Ras, Ral, Rap, Rac, and Rho (27,(35)(36)(37)(38). We first analyzed the specificity and capacity of the GST-linked proteins to bind to recombinant Rho family GTPases. The GTPases were used either in their inactive GDP-bound form or the in the active conformation after the GDP was exchanged for the GTP analogue Gpp(NH)p. As expected, the GST-PAK-CD fusion appeared specific for Gpp(NH)p Rac and Cdc42, but did not bind to either Gpp(NH)p or GDP Rho (Fig. 3A). GST-C21 featuring the Rho effector motif from Rhotekin was capable only of binding the Gpp(NH)p-RhoA and not the GDP form. It did not bind Rac or Cdc42. We next used growth factors to see whether we could detect stimulation of these GTPases in intact cells using these probes. In morphological studies, PDGF has been reported to activate Rac (1), whereas LPA has been described as a Rho stimulus (2). In agreement with these observations we found that PDGF stimulation of COS-7 cells led to an increase in the amount of Rac-GTP in GST-PAK-CD pull-downs (Fig.  3B). In pull-downs with GST-C21, RhoA was stimulated by LPA.
To study the activity of hPEM-2 in cells we made use of transient expression in COS-7 cells. We used the well characterized GEF proteins, Tiam1 and Dbl, as controls. As previously reported, Dbl activated both Rho and Cdc42 but not Rac (Fig. 4), Tiam1 activated Rac and Cdc42 (Fig. 4, B and C). The activation of Cdc42 by Tiam1 was somewhat surprising. However, we have observed activation of Cdc42 by Tiam1 in other cell types such as PC12 cells (7). Additionally Tiam1-induced activation of Pak and Jun N-terminal kinase (JNK) via Cdc42 has previously been reported in COS-7 cells (39). In BW 5147 T-lymphocyte cells, Madin Darby canine kidney cells and NIH3T3, cells Tiam1 specifically activates Rac, 2 suggesting that some cell-type variations may exist in the specificity of GEF proteins. In these studies, we found no evidence that hPEM-2 was able to activate either Rac or Rho but it did activate Cdc42 to the same degree as either Tiam1 or Dbl (Fig.  4, A-C). The specificity demonstrated by the two control GEF proteins confirms previously published data and validates the technique as a general method for determining exchange factor specificity in intact cells.
Phenotypic Analysis-We then transfected HA-hPEM-2 into NIH3T3 cells, because COS-7 cells are not well suited to morphological analysis. As controls we also used the GEFs Tiam1 and Dbl. 24-h post-transfection we fixed and stained the cells for expression of HA-hPEM-2 and polymerized actin. Although there was no specific localization of hPEM-2 in these cells, we did notice the formation of filopodia, associated with a profoundly rounded phenotype (Fig. 5). The rounded phenotype, which featured a thick ring of cortical actin, seemed more reminiscent of the phenotype exhibited by V12Cdc42 than that of V12Rac, which induced cell spreading (Fig. 5). Transfection of V14Rho also induced some rounding but profoundly effected the architecture of focal adhesions and stress fiber organization, which we failed to observe in cells expressing hPEM-2 (Fig. 5). Although Dbl caused rounding similar to that induced by hPEM-2, there was a concomitant increase in stress fibers that was not seen in cells expressing hPEM-2. In cells that expressed Tiam1, as previously reported, we found an induction of cell spreading and plasma membrane ruffling (22). Thus, the observed phenotype was consistent with the results of the pull-down assay. DISCUSSION In this report we have attempted to group together the specificities of the DH group of GEF proteins using computa-tional methods. A large number of reports have implied that the specificity of DH proteins toward their downstream GTPases is inferred by the DH domain alone. For Lfc, it appears that the PH can bind to tubulin, but is not absolutely required for activity of the DH domain. The PH domain can be replaced with an isoprenylation signal to tag it to the plasma membrane that restores its transforming activity (16,40). We too have found that a chimeric protein featuring the DH domain from Dbl inserted into Tiam1 retains the capacity to activate Rho and Cdc42. 2 This indicates that for Dbl, the DH domain alone is sufficient for both specificity and activity provided it is localized at the plasma membrane. With this information, we attempted to see if we could identify a group of related exchange factors based upon sequence analysis using the DH domain from Tiam1 as a probe. BLAST searches with the DH domain from Tiam1 identified other proteins that were reported GEFs for Rac such as PIX/COOL, Vav, and Sos (41)(42)(43). The cladogram shown in Fig. 1B summarizes our findings and correlates reasonably well with the published data regarding GTPase specificity (reviewed in Ref. 34).
Although sequence comparisons led us to believe that hPEM-2 would be an exchange factor for Rac, it activates Cdc42 alone. When we expressed this protein in fibroblasts we saw cell rounding rather than the characteristic flattening and plasma membrane ruffling associated with Rac. We examined the specificity of this enzyme using pull-down assays and found that hPEM-2 was specifically a GEF protein for Cdc42 rather than Rac1 or RhoA. This is in agreement with the data from the phenotypic analysis. Moreover, the results from the activity assay for Tiam1 and Dbl also correlated with the published data from previous morphological studies. In these experiments Dbl activated both Rho and Cdc42 (44), whereas Tiam1 activated Rac1 (22,27). Tiam1 expression also activated Cdc42 in COS-7 cells, as it also does in PC12 cells (7).
PEM-2 was originally identified as a transcript in the Ascidian C. savignyi, as a maternally encoded gene involved in blastocyst patterning. The polarized staining of PEM-2 transcripts in Ascidian embryos is very reminiscent of V12Cdc42 localization in microinjected blastocysts (26,45). Furthermore, Cdc42 is a principle regulator of cellular orientation to diverse stimuli such as wounding, hormonal, and chemotactic gradients in yeast and in mammalian cells (46 -48). Cdc42, but not Rho or Rac, is implicated in lymphocyte polarization toward antigen presenting cells during activation (49). Experiments in yeast demonstrated a requirement for Cdc42 to determine the bud site in dividing cells and polarity in mating cells. Recently this has been shown to be principally regulated by the GEF Cdc24 (13). Taken together, these studies point to a role for Cdc42 in the initiation and maintenance of cellular polarity. This might infer a role for hPEM-2 in the localized organization and polarization of the actin cytoskeleton through Cdc42 activity. There is now a wealth of evidence that points to a key role for the members of the Rho family GTPases in the spatial organization of cerebral neuronal processes, where hPEM-2 is most highly expressed in adults. Other work has indicated a role for the Rac and Cdc42 effector PAK in neurite outgrowth in PC12 cells (17,50), whereas another report indicates a role for Cdc42 in the cytoskeletal reorganization, which accompanies differentiation and dendrite formation in cultured neurons (51). These reports imply a diverse role for Cdc42, particularly in the regulation of cellular orientation and polarization. Given the possible role of PEM-2 in the regulation of polarity in development, it might be interesting to examine the role of hPEM-2 in neuronal polarization.
In conclusion, we have identified a new GEF protein and have shown by morphological analysis and with a biochemical GTP loading assay that it specifically activates Cdc42. Homologues of this gene have been linked to the development of polarity in other cell systems. We are presently investigating whether this DH protein may link polarization and directional organization of the cytoskeleton to Cdc42.