A New Family of Cdc42 Effector Proteins, CEPs, Function in Fibroblast and Epithelial Cell Shape Changes

Cdc42, a Rho GTPase, regulates the organization of the actin cytoskeleton by its interaction with several distinct families of downstream effector proteins. Here, we report the identification of four new Cdc42-binding proteins that, along with MSE55, constitute a new family of effector proteins. These molecules, designated CEPs, contain three regions of homology, including a Cdc42 binding domain and two unique domains called CI and CII. Experimentally, we have verified that CEP2 and CEP5 bind Cdc42. Expression of CEP2, CEP3, CEP4, and CEP5 in NIH-3T3 fibroblasts induced pseudopodia formation. Fibroblasts coexpressing dominant negative Cdc42 with CEP2 or expressing a Cdc42/Rac interactive binding domain mutant of CEP2 did not induce pseudopodia formation. In primary keratinocytes, CEP2- and CEP5-expressing cells showed reduced F-actin localization at the adherens junctions with an increase in thin stress fibers that extended the length of the cell body. Keratinocytes expressing CEPs also showed an altered vinculin distribution and a loss of E-cadherin from adherens junctions. Similar effects were observed in keratinocytes expressing constitutively active Cdc42, but were not seen with a Cdc42/Rac interactive binding domain mutant of CEP2. These results suggest that CEPs act downstream of Cdc42 to induce actin filament assembly leading to cell shape changes.

Cdc42 is a member of the p21 Rho family of small GTPases that act downstream of cell surface receptors to regulate the formation of different F-actin-containing structures (1). Cdc42 regulates actin polymerization to induce the assembly of filopodia, which are thin actin filaments that project from cells (2,3). Cdc42 can also regulate other biological processes such as cytokinesis (4,5) and cell polarity (6 -13). As to its role in cell polarity, Cdc42 regulates polarization of Drosophila wing epithelial cells by inducing actin polymerization at the apical cortex of these cells (6). In mammalian cells, Cdc42 regulates a number of polarized responses, including actin localization (7,8), nuclear displacement (7), protein trafficking (9), and directed cell movement (10 -13). These studies suggest that Cdc42 regulates cell shape and polarity changes by coordinating actin changes with altered cell membrane assembly.
The ability of Cdc42 to regulate the actin cytoskeleton and cell polarity in mammalian cells stems from its interaction with a large number of downstream effector proteins (14). One such effector protein, N-WASP, induces actin polymerization leading to filopodia formation (15). N-WASP mediates these actin changes by directly binding actin (15), the Arp2/3 complex (16), and by interacting with the actin-binding protein profilin (17). The p21-activated protein kinases (PAK) 1 also mediate actin reorganization required for filopodia formation (18,19), cell motility (20,21), and neurite-outgrowth (22,23). Unlike N-WASP, PAKs mediate their biological activities by phosphorylating a number of cytoskeletal targets (20,21). Despite what is known about Cdc42 and effector interactions, the molecular details of how Cdc42 regulates actin and cell polarity changes in different cell types are only beginning to emerge.
We recently characterized MSE55 as a nonkinase Cdc42 effector that mediates actin reorganization leading to the formation of pseudopodia (24). In the present study, we identify four additional proteins CEP2, CEP3, CEP4, and CEP5 that share homology with MSE55. These proteins contain a CRIB domain that is homologous to that of MSE55 (24,25). Two other domains, designated CI and CII, are also found in MSE55/ CEP1, CEP2, CEP3, and CEP4, whereas CEP5 lacks a CI domain but instead contains a proline-rich region. Using in vitro binding assays, CEP2 and CEP5 interacted with Cdc42, but not Rac. Similar to our previous results with MSE55 (24), expression of all the CEP members induced long pseudopodia in NIH-3T3 fibroblasts. In contrast, only background levels of pseudopodia were observed in fibroblasts expressing a CRIB mutant of CEP2 or coexpressing dominant negative Cdc42 with CEP2. In primary keratinocytes, expression of Cdc42-Q61L, CEP2, and CEP5 all decreased F-actin staining in the adherens junction, but increased the number of cytoplasmic actin stress fibers. Unlike Rho-induced stress fibers, the cytoplasmic stress fibers seen in the Cdc42-and CEP-expressing cells were greater in number, were thinner, and transversed the entire cell body. In many cases, keratinocytes expressing CEPs also showed cell elongation in the direction of the actin filaments. These results suggest that CEPs act downstream of Cdc42 to induce actin filament assembly leading to cell shape changes. * This work was supported by Grant R29-CA77459-01 (to P. D. B.) from the National Cancer Institute, by a predoctoral scholarship from the ARCs Foundation (to D. S. H.), and by a Department of Defense breast cancer predoctoral fellowship (to D. M. P.). The DNA sequencing and Microscopy core facilities of the Lombardi Cancer Center were funded by Grant P30-CA51008 from the National Institutes of Health. 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM

EXPERIMENTAL PROCEDURES
Identification of CEP Clones-Clones of human CEP2 were identified from a TBLASTN search of the expressed sequence tag (EST) data base at the National Center for Biotechnology Information using the 16amino acid sequence of the CRIB domain of MSE55 as a query (24,25). An EST clone corresponding to human CEP2 was obtained from the IMAGE Consortium (clone identification number 22978). This clone was sequenced on an Applied Biosystems 377 DNA sequencer and has the GenBank Accession number AF098290. Clones containing related genes, human CEP3 (hCEP3), human CEP4 (hCEP4), and mouse CEP5 (mCEP5) were also obtained from the IMAGE Consortium (clones 488206, 360199, and 408069, respectively) and sequenced. The nucleotide and amino acid sequences of hCEP3, hCEP4, and mCEP5 have GenBank accession numbers AF104857, AF099664, and AF102773, respectively.
GST Capture Experiments-Approximately 5 g of either recombinant glutathione S-transferase (GST), Rac-Q61L, or Cdc42-G12V expressed as fusion proteins in Escherichia coli were purified on glutathione-agarose resin and loaded with GTP␥S as described previously (24). N-terminal FLAG-tagged mammalian expression vectors of CEP2, CEP2-H39A, H42A, and CEP5 were transfected by electroporation (250 microfarads/400 V) into Cos1 cells, and similar levels of expression were observed for the different constructs. After 48 h, cell lysates were prepared and used in GST capture experiments as described (24). Briefly, precleared lysates were incubated for 10 min with each of the immobilized recombinant proteins and washed, and bound proteins were separated on an 8% SDS-polyacrylamide gel electrophoresis gel. Proteins were then electrophoretically transferred to nitrocellulose. The blots were blocked and then probed with anti-FLAG M2 monoclonal antibody (Sigma Chemical Co., St. Louis, MO). FLAG-tagged CEPs were then detected by incubating with rabbit anti-mouse-horseradish peroxidase followed by incubation with enhanced chemiluminescence reagents (Pierce, Rockford, IL) and exposure to x-ray film.
Keratinocyte Transfection and Immunofluorescence-Human primary keratinocytes were isolated from human foreskin as described (28) and maintained in keratinocyte serum free media containing epidermal growth factor (5 ng/ml), bovine pituitary extract (50 g/ml), and gentamicin (10 g/ml) (Life Technologies, Inc.). Cells were grown on glass coverslips for 2 days and then transfected with one of the following constructs: pcDNA3-Myc-Cdc42-Q61L, pEXV-Myc-RhoB-G14V, pcDNA3-Myc-CEP2, pcDNA3-Myc-CEP5, or pcDNA3-Myc-CEP2-H39A, H42A using the Fu-gene6 transfection reagent (Roche Molecular Biochemicals, Indianapolis, IN) per the manufacturer's instructions. Twenty-four hours after transfection, cells were switched to Dulbecco's modified Eagle's medium high calcium media containing 10% heat-inactivated fetal calf serum to induce adherens junction formation (28). One hour later, the cells were fixed and stained. Transfected cells were detected using a rabbit polyclonal anti-Myc antibody at a 1:200 dilution (Santa Cruz Biotechnology). For analysis of F-actin and vinculin localization, cells were stained simultaneously with rabbit polyclonal anti-Myc antibody and mouse monoclonal anti-vinculin antibody (Sigma) at a 1:200 dilution. Cells were then stained with a Texas Red-conjugated donkey anti-rabbit polyclonal antibody at a 1:300 dilution to detect transfected cells, a Cy5-conjugated donkey anti-mouse antibody at a 1:300 dilution to detect vinculin, and FITC-conjugated phalloidin at a 1:300 dilution to detect F-actin. For E-cadherin staining, cells were simultaneously stained with rabbit polyclonal anti-Myc antibody and with mouse monoclonal anti-E-cadherin antibody (Transduction Laboratories, Lexington, KY) at a 1:200 dilution. Cells were then incubated with FITC-conjugated goat anti-Rabbit antibody at a 1:200 dilution to detect transfected cells and a Texas Red-conjugated goat anti-mouse antibody at a 1:400 dilution to detect E-cadherin. Confocal microscopy was performed with an Olympus Fluoview confocal microscope (Olympus America, Inc., Melville, NY) attached to an Olympus 1 ϫ 70 inverted fluorescent scope equipped with a 60ϫ oil immersion lens. Digitized images were captured using the Fluoview software (Olympus America, Inc.).

Identification of Genes for
CEPs-Previously, we identified MSE55 as a CRIB domain-containing Cdc42 nonkinase effector protein that induces long F-actin-containing extensions resembling pseudopodia (24). Using the amino acid sequence of MSE55 to query the EST data base, we identified and sequenced three different human EST clones that contain CRIB domains and two additional regions of amino acid homology to MSE55. These cDNA clones, designated CEP2, CEP3, and CEP4, encode proteins of 210, 254, and 356 amino acids, respectively (Fig. 1A). We also identified and sequenced a mouse cDNA clone designated mCEP5, encoding a protein of 150 amino acids that contains features in common with MSE55 (Fig. 1A). Analysis of the nucleotide sequences of the different CEP clones confirmed that they contain a Kozak's consensus translation initiation sequence and an in-frame stop codon 5Ј to the methionine start codon (data not shown). Based on the amino acid sequence and binding data (see below), we conclude that CEPs define a new family of Cdc42-binding proteins. Because MSE55 is a Cdc42 effector involved in actin reorganization at the cell membrane (24) and not a major serum protein (29), we propose that MSE55 be renamed CEP1 to better describe its close structural and functional relationship to other members of the CEP family. At the time of submission to GenBank, no additional MSE55-like proteins were known, although a recent contemporaneous study by Joberty et al. (30) identified the same proteins, denoted as the Borgs, as Cdc42binding proteins. The corresponding nomenclature is as follows: CEP1/MSE55/Borg5, CEP2/Borg1, CEP3/Borg2, CEP4/ Borg4, and CEP5/Borg3. Potential Signaling Domains in CEPs-Although all CEP members contain the 16 amino acid CRIB consensus sequence (25), they also show further homology extending C-terminal to the CRIB domain (Fig. 1B). This expanded CRIB region has the consensus sequence Ile-Ser-Xaa-Pro-Leu-Gly-Xaa-Phe-Arg-His-Thr-AA-His-AA-Gly-Xaa-Xaa-Gly-(Xaa) 0 -2 -Asp-AA-Phe-Gly-Asp-Xaa-Ser-Phe-Leu, where AA represents an aliphatic amino acid (Fig. 1B). The expanded CRIB region of the CEP family is different from that found in other Cdc42 effector proteins except for the Drosophila DPR2/DACK tyrosine kinase (25,31). These comparisons suggest that the C-terminal ends of the extended CRIB domain in CEPs and DPR2/DACK could be involved in regulating the biological effects of these proteins and/or are involved in Cdc42/Rac binding. Many families of Cdc42 effector proteins contain unique extended regions of FIG. 1. The CEP family of proteins. A, protein sequence alignments of the CEP proteins. The numbers at the beginning and end of each line indicate the total number of amino acid residues to that point in the sequence. Optimal alignment of the different CEP protein sequences was performed manually. Regions sharing significant homology are highlighted in gray. Completely conserved amino acids in the CI and CII domains are indicated by an asterisk. The polyproline region of CEP5 is underlined. B, sequence alignment of CRIB, CI, and CII regions with other signaling molecules. A dashed line or a plus sign identifies identical or semiconserved amino acids to CEP1, respectively. The consensus sequences are based on completely conserved amino acids between CEP family members and are shown in boldface at the top, with conserved aliphatic residues denoted by A. The CRIB domain, potential CII domains of DPR2/DACK (25,31), and a potential CI domain of two distinct C. elegans clones (F44D12.4 and S72575) are also shown. homology C-terminal to the core CRIB domain that are functionally important. In PAK kinases, this region negatively regulates its kinase activity (32). In ACK (33) and WASP (34), these regions form an alpha helix that is involved in Cdc42 binding.
In addition to the CRIB domain, CEP family members share two other regions of homology designated the CI and CII domains that span 11 and 14 amino acids, respectively (Fig. 1B). The CI domain is located C-terminal to the CRIB domain and has the consensus sequence AA-AA-Lys-Asn-Ala-AA-Ser-Leu-Pro-Xaa-AA (Fig. 1B). The CII domain is located C-terminal to CI domain and has the consensus sequence of Asp-Leu-Gly-Pro-Ser-AA-Leu-Xaa-Xaa-AA-Leu-Xaa-AA-Met (Fig. 1B). Although CEP1, CEP2, CEP3, and CEP4 have CRIB, CI, and CII domains, CEP5 is unique in that it lacks a CI domain and instead contains a polyproline region in its place (amino acids 56 -89) (Fig. 1A). Although the functions of the CI and CII domains are unknown, several other signaling molecules contain similar domains. Two Caenorhabditis elegans cDNA clones (GenBank accession numbers S72575 and F44D12.4) encode proteins that contain potential CI domains (Fig. 1B). The Drosophila Cdc42 tyrosine kinase effector DPR2/DACK (25,31) has a potential CII domain ϳ100 amino acids downstream of the CEP-like CRIB region (Fig. 1B). The mouse PSM protein, an SH2-SH3 adapter molecule (35), also contains a CII-like domain (Fig. 1B). The fact that the CII domain occurs separately in CRIB-containing DPR2/DACK and CEP5 proteins suggests that the CI and CII domains function independently of each other.
CEPs Bind to Cdc42-Previously, we showed that CEP1/ MSE55 (24) and the Drosophila DPR2/DACK CRIB-containing kinase (25) bind strongly to Cdc42 but weakly to Rac. In light of these findings and the fact that the CRIB domain of all the CEP members and DPR2/DACK are 75% identical, we were interested in determining if other CEP family members could also interact with Cdc42. We used GST-affinity chromatography to examine binding of CEP2 and CEP5 to Rac and Cdc42. In these experiments, both CEP2 and CEP5 bound to Cdc42, but not detectably to Rac (Fig. 2). Similar experiments with CEP4 demonstrated a specific interaction with Cdc42, but not with Rac (data not shown). Yeast two-hybrid analysis also validated CEP2 and CEP5 interaction with Cdc42 (data not shown). Furthermore, a CRIB mutant of CEP2 containing alanine substitutions within two conserved histidine residues (CEP2-H39A, H42A) showed no detectable binding to Cdc42 (Fig. 2). Taken together these data suggest that CEPs interact with Cdc42 via their CRIB domain and function as downstream effector proteins.
CEPs Induce F-actin-containing Pseudopodia in NIH-3T3 Fibroblasts-FLAG-or Myc-tagged mammalian expression vectors for CEP2, CEP3, CEP4, and CEP5 were transfected into NIH-3T3 fibroblasts. After transfection, CEP protein expression was observed in both the cytoplasm and at the plasma membrane (Fig. 3, a, c, e, and g). Approximately 40% of CEPexpressing fibroblasts showed one to two pseudopodia that were longer than the cell body length (Fig. 3, b, d, f, and h). The number of CEP-expressing cells showing pseudopodia was two times greater than that of untransfected control cells. Some cells with pseudopodia also exhibited laterally placed ruffles or ruffles localized to the distal ends of the pseudopodia (Fig. 3, f  and h). The CEP-expressing cells extending pseudopodia did not show cell body rounding or collapse of the cytoplasm, suggesting these structures were distinct from retraction fibers. Furthermore, at higher magnifications, the CEP-induced pseudopodia in some of the cells contained long F-actin cables that FIG. 2. CEP2 and CEP5 bind to Cdc42. Immobilized GST, GST-Cdc42-G12V, or GST-Rac-Q61L were loaded with GTP␥S and incubated with whole cell extracts of Cos1 cells transfected with either FLAG epitope-tagged CEP5, FLAG epitope-tagged CEP2, or FLAG epitopetagged CEP2-H39A, H42A. Following washing, bound proteins were analyzed by Western blot using the M2 mouse anti-FLAG antibody with anti-mouse-horseradish peroxidase. The blots were developed using enhanced chemiluminescence (ECL).

FIG. 3. CEPs induce pseudopodia formation in NIH-3T3 fibroblasts.
FLAG-tagged CEP constructs were transfected into NIH-3T3 fibroblasts and fixed 24 h after transfection. Fluorescence micrographs show CEP2 (a and b), CEP3 (c and d), CEP4 (e and f), and CEP5 (g and h) expressing cells inducing pseudopodia. CEP expression was detected using mouse monoclonal antibodies against the epitope tag followed by staining with FITC-conjugated goat anti-mouse IgG (a, c, e, and g). F-actin was visualized with Texas Red-conjugated Phalloidin (b, d, f, and h). Bar, 10 m. extended into the cell body (data not shown). No obvious differences in pseudopodia morphology were seen between any of the CEPs, including CEP1 (24). These data demonstrate that the five CEP family members induce similar pseudopodia formation in NIH-3T3 fibroblasts.
Cdc42 Regulates CEP-induced Pseudopodia Formation-To more clearly define the relationship between Cdc42 and CEPs, we tested the effect of CEP2/Cdc42 coexpression in NIH-3T3 fibroblasts. As previously reported (2, 3), fibroblasts expressing Cdc42 showed many uniformly placed filopodia around the cell periphery (Fig. 4, a and b). In contrast, cells coexpressing Cdc42 and CEP2 induced one or two pseudopodia with multiple filopodia and/or membrane ruffles emanating from it (Fig. 4, c  and d). Additionally, the majority of coexpressing cells showed peripherally placed filopodia similar to those seen in cells only expressing Cdc42 yet showed a more elongated cell shape (compare Fig. 4, c and d, with Fig. 4, a and b). Fibroblasts coexpressing dominant negative Cdc42-T17N and CEP2, failed to produce pseudopodia (Fig. 4, e and f). However, in many of these fibroblasts, expression of the dominant negative Cdc42 mutant alone or in combination with CEP2 appeared to be toxic, inducing cell rounding in ϳ50% of the cells.
We also examined whether cells transfected with the CRIB mutant of CEP2 could induce pseudopodia. By cell counting, CEP2 was found to induce pseudopodia formation 2-fold over untransfected cells (Fig. 5). In contrast, cells expressing the CRIB mutant of CEP2 (CEP2-H39A, H42A) showed levels of pseudopodia similar to untransfected cells (Fig. 5), although the expression level and localization of the mutant was similar to that of wild type CEP2 protein. Fibroblasts coexpressing Cdc42 with CEP2 also showed a slight increase in pseudopodia formation over CEP2 expression alone (Fig. 5). Together these data are consistent with a model whereby Cdc42 is required for CEP-mediated pseudopodia formation in fibroblasts.
Constitutively Active Cdc42 Induces Thin F-actin Stress Fiber Formation in Keratinocytes-Based on the CEP-induced phenotype observed in fibroblasts, we were interested in determining whether CEPs might function in Cdc42-mediated actin reorganization in other cell types. Although Madin-Darby canine kidney cells have been used to study Cdc42-induced actin changes (36, 37), we were unable to detect CEP-induced (e and f). Twenty-four hours after transfection, cells were stained with a rabbit polyclonal anti-Myc antibody followed by staining with FITC-conjugated goat anti-rabbit to detect Cdc42 (a and c) and Cdc42-T17N (e). Texas Red-conjugated Phalloidin was used to detect F-actin (b). FLAG-tagged CEP2 (d and f) expressing cells were detected with a mouse monoclonal anti-FLAG antibody followed by staining with Texas Red-conjugated goat anti-mouse antibody. changes in actin polymerization or cell shape in these cells (data not shown). As an alternative epithelial cell type, we chose to study the function of Cdc42 and CEPs in primary keratinocytes, which form well-defined adherens junctions driven by actin filament polymerization following a switch to calcium-containing media (38). Unlike wild type Cdc42 that showed no effect on F-actin (data not shown), keratinocytes expressing Cdc42-Q61L (Fig. 6a) showed low levels of F-actin staining at the adherens belt and contained thin F-actin filaments that extended the length of the cell body (Fig. 6b). By confocal analysis, we found that these actin filaments localized at the ventral surface and within the cytoplasm of the cell (Fig.  6b). This Cdc42-induced phenotype differed from untransfected neighboring cells that showed high levels of F-actin staining at the adherens junction (Fig. 6b). Cdc42-Q61L-expressing cells showed vinculin staining at the tips of the actin filaments, indicating that these actin filaments are stress fibers (Fig. 6c). This pattern of vinculin staining was different from that of untransfected cells in that it was less organized around the cell periphery.
We also examined the effect of expressing a constitutively active Rho mutant in keratinocytes. Although a RhoA expression vector expressed poorly in these cells, a constitutively active RhoB construct (Fig. 6d) expressed at levels similar to Cdc42. These Rho-expressing cells showed fewer stress fibers that were thicker in width and shorter in length (Fig. 6e) than those observed in cells expressing Cdc42-Q61L (Fig. 6b). RhoBexpressing cells showed fewer vinculin-staining complexes that were more pronounced than the fine punctate vinculin staining observed in Cdc42-Q61L-expressing cells (compare Fig. 6f with Fig. 6c). These results suggest that expression of constitutively active Cdc42 in primary keratinocytes induced the assembly of fine actin stress fibers that are distinct from stress fibers induced by constitutively active Rho.
CEP2 and CEP5 Alter F-actin and Vinculin Organization in Keratinocytes-We next expressed CEP2 and CEP5 in primary keratinocytes. CEP5 (Fig. 7a) and CEP2 (Fig. 7d) showed membrane and cytoplasmic localization with high levels of staining in a distinct cytoplasmic compartment surrounding the nucleus. This localization differed from the cytoplasmic and membrane localization of CEPs found in fibroblasts (see Fig. 3). CEP5-and CEP2-expressing keratinocytes showed less F-actin staining at the adherens belt and had an increase in fine F-actin filaments that ran along the length of the cell (Fig. 7, b  and e). These effects resembled the F-actin changes observed in Cdc42-Q61L-expressing cells except that CEP-expressing cells tended to be more elongated than cells expressing Cdc42-Q61L. Expression of CEP2 and CEP5 prior to calcium switch also showed these F-actin changes, although they were less pronounced (data not shown). The tips of CEP-induced actin fila-ments stained positive for vinculin indicating that these filaments were stress fibers (Fig. 7, c and f). CEP-expressing cells often showed the highest vinculin levels in regions of the cell that lacked contact with neighboring cells. This differed from untransfected cells that contained uniform vinculin staining around the cell periphery even when not neighbored by other cells. The CEP2 CRIB mutant showed a similar expression pattern (Fig. 7g) as wild type CEP2, but did not show an increase in actin filaments (Fig. 7h) or altered vinculin distribution (Fig. 7i). Instead, cells expressing the CEP2-CRIB mutant showed F-actin staining (Fig. 7h) and vinculin staining (Fig. 7i) at the cell periphery that resembled that of untransfected cells. Taken together these results suggest that CEP5 and CEP2 mediate Cdc42-like actin changes by inducing the actin and vinculin reorganization that may contribute to epithelial cell shape changes.
Cdc42-Q61L-and CEP-expressing Keratinocytes Show Reduced E-cadherin Staining at Adherens Junctions-Because both Cdc42 and CEP expression resulted in the loss of F-actin at adherens junctions, we next examined the effect of Cdc42 and CEPs on E-cadherin distribution at adherens junctions. One hour following calcium switch, cells expressing Cdc42-Q61L (Fig. 8, a and b), CEP2 (Fig. 8, c and d), or CEP5 (Fig. 8, e and f) showed reduced levels of E-cadherin at adherens junctions ( Fig. 8, b, d, and f). Cdc42-expressing cells also showed an increase in E-cadherin localized within vesicles. In contrast, untransfected cells showed strong E-cadherin staining at cellcell junctions (Fig. 8, b, d, and f). These data suggest that overexpression of Cdc42 or CEPs interfere with the assembly of normal adherens junctions. DISCUSSION Here we report the structure and function of a new family of Cdc42 effector proteins that we have called CEPs. This family contains five members: CEP1/MSE55, CEP2, CEP3, CEP4, and CEP5. CEP1 through CEP4 all contain a previously identified CRIB domain (25) and two new domains, designated CI and CII. CEP5 is structurally different because it lacks a CI domain. Transfection experiments indicate that CEPs induce actin reorganization and cell shape changes that lead to pseudopodia formation in fibroblasts. Interestingly, pseudopodia normally occur in NIH-3T3 fibroblasts and are increased in cells overexpressing v-Fos (39), CD-44 (40), or an F-actin-bundling protein (41). Although the biological relevance of the CEP-induced pseudopodia with relation to Cdc42 signaling is still unclear, there are several possible explanations. The CEPinduced pseudopodia in fibroblasts may represent an exaggerated activity of these molecules, producing a phenotype that is not generally observed with endogenous levels of CEPs. Previous studies have shown that inducible recruitment of Cdc42 to FIG. 6. Constitutively active Cdc42 induces thin actin stress fibers in keratinocytes. Keratinocytes were transfected with Myc-tagged Cdc42-Q61L (a-c) or RhoB-G14V (d-f). Twenty-four hours after transfection, the cells were changed to high calcium complete media for 1 h. Myc-tagged Cdc42-Q61L (a) and RhoB-G14V (d) were detected with rabbit anti-Myc antibody followed by staining with Texas Red-conjugated donkey anti-rabbit antibody. FITC-conjugated Phalloidin was used to detect F-actin (b and e). Vinculin was detected by staining with mouse anti-vinculin followed by Cy5-conjugated donkey anti-mouse antibody (c and f). Bar, 10 m.
the membrane induces a single filopodium (42) rather than the array of filopodia that is conventionally seen with expression of wild type Cdc42. Based on these findings, CEPs might function in fibroblasts at specific membrane sites resulting in a net effect of directed membrane protrusion. CEPs may also represent targets of other Cdc42-related GTPases that normally function in fibroblasts to generate pseudopodia formation rather than filopodia formation. Consistent with this model, Cdc42 homologues such as TC10 (43) and TCL (44) interact with CRIB-containing proteins in vitro and induce actin structures that are distinct from Cdc42.
Recently Joberty et al. (30) identified and characterized a set of Cdc42-binding proteins, termed Borgs, which are identical to the CEPs. Although they observed similar morphological changes following overexpression in fibroblasts, they suggested that Borgs/CEPs function to inhibit Rho activity, leading to the  H42A (g-i). Twenty-four hours after transfection, the cells were changed to high calcium complete media for 1 h. Myc-tagged CEP constructs were detected with rabbit anti-Myc antibody followed by staining with Texas Red-conjugated donkey anti-rabbit antibody (a, d, and g). FITC-conjugated Phalloidin was used to detect F-actin (b, e, and h). Vinculin was detected using a mouse anti-vinculin antibody followed by staining with Cy5-conjugated donkey anti-mouse antibody (c, f, and i). Bar, 10 m. production of retraction fibers. Our results are difficult to reconcile with this retraction fiber model. First, CEP-induced pseudopodia and retraction fibers are morphologically distinct and are formed in different ways. Retraction fibers appear as dendritic-like processes (45) and are formed following collapse of the cytoskeletal architecture. The pseudopodia formed in our studies were produced by an active process leading to the formation of single or bipolar membrane protrusions, as visualized by time-lapse videography (24). The CEP-induced pseudopodia are also not dendritic-like, but are limited to one or two extensions per cell without much branching. CEP-expressing cells did not show cell body rounding or collapse, which is normally seen in cells exhibiting retraction fibers. Second, in Madin-Darby canine kidney epithelial cells, CEPs had no noticeable effect on stress fiber formation. Third, we were unable to reproduce the finding (30) that coexpression of constitutively active Rho reverses CEP-induced morphological changes in NIH-3T3 fibroblasts. 2 The reason for this is unknown, but could be due to differences in cell culture conditions. Fourth, in keratinocytes, CEPs induced thin actin stress fibers that were similar to what was observed following transfection with an activated mutant of Cdc42. Thus, we propose an alternative model in which CEPs, acting as classical Cdc42 effector molecules, increase actin filament assembly to cause cell shape changes.
Although less is known about the function of Rho GTPases in epithelial cells as compared with fibroblasts, recent studies indicate that they function in the assembly of adherens junctions (36,37,46,47). For example, overexpression of constitutively active Rho or Rac in Madin-Darby canine kidney cells increases F-actin at the adherens junctions while dominant negative constructs block the formation of adherens junctions (46,47). Likewise, Madin-Darby canine kidney cells overexpressing constitutively active Cdc42 show increased F-actin at adherens junctions along with other associated molecules such as ␤-catenin and E-cadherin (36,37). Our studies using primary keratinocytes implicate CEPs in Cdc42-mediated assembly of thin actin stress fibers that extend across the cell body length. The phenotype observed here with both Cdc42 and CEPs resembles that of early stage adherens junction formation, where actin filaments radiate across the cell body to drive adherens junction formation and would have been missed using immortalized epithelial cells (38). Although Cdc42-Q61L and CEPs induced the formation of fine actin stress fibers, Rho induced short stress fibers with thicker actin bundles. Vinculin staining in the Cdc42-and CEP-expressing cells was also less prominent than in Rho-expressing cells. These results are in agreement with studies in fibroblasts, demonstrating that Cdc42 and Rho induce distinct actin and focal adhesion reorganization (3). CEPs and Cdc42 also interfered with E-cadherin localization to adherens junctions thus providing further support that they may function in actin reorganization during the early stages of adherens junction formation.
In fibroblasts and keratinocytes, CEP-induced actin changes were dependent on Cdc42. In NIH-3T3 fibroblasts, coexpression of dominant negative Cdc42 with CEP2 or expression of a CEP2 CRIB mutant blocked pseudopodia formation. Similarly, CEP2-induced cytoskeletal and morphological changes were also abrogated in keratinocytes expressing the CEP2-CRIB mutant. Based on these observations, we propose a model in which Cdc42 binding to the CRIB domain of CEPs induces conformational changes leading to exposure of the previously inaccessible CI, CII, and/or other domains. Further clues to the normal function of these molecules may come from the obser-vation that the Drosophila CRIB-containing ACK-like tyrosine kinase, DPR2/DACK, also contains a CII-like domain downstream of a similar CRIB domain and functions to phosphorylate a SH3-containing adapter molecule (31). These results suggest the possibility that the CII domain recruits additional actin-binding proteins or other signaling molecules and/or acts as an autoinhibitory domain as seen in the WASP proteins (48). In the case of CEP5, we propose that the unique proline-like adapter sequences replace the function of the CI domain to induce similar biological activities. Like the proline-rich sequences found in N-WASP (17), these sequences may bind profilin and/or SH3-containing molecules.