VE-cadherin-induced Cdc42 signaling regulates formation of membrane protrusions in endothelial cells.

The cytoplasmic domain of cadherins and the associated catenins link the cytoskeleton with signal transduction pathways. To study the signaling function of non-junctional VE-cadherin, which can form during the loss VE-cadherin homotypic adhesion, wild type VE-cadherin or VE-cadherin cytoplasmic domain (DeltaEXD) was expressed in sub-confluent endothelial cells. We observed that Cdc42 was activated in transfected cells and that these cells also developed Cdc42-dependent >70-microm-long plasma membrane protrusions. The formation of these structures required actin polymerization, and they developed specifically in endothelial cells as compared with epithelial cells. Expression of the VE-cadherin cytoplasmic domain lacking the beta-catenin binding site also induced Cdc42 activation; thus, its activation cannot be ascribed to beta-catenin binding. However, these cells were not able to form the protrusions. These results suggest that the cytoplasmic domain of non-junctional VE-cadherin can serve as a scaffold involved in Cdc42 activation at the endothelial plasma membrane. beta-Catenin and the associated alpha-catenin may serve as support sites for actin polymerization, leading to formation of long plasma membrane protrusions. Thus, non-junctional VE-cadherin actively participates in inside-out signaling at the plasma membrane, leading to the development of endothelial membrane protrusions.

The cytoplasmic domain of cadherins and the associated catenins link the cytoskeleton with signal transduction pathways. To study the signaling function of non-junctional VE-cadherin, which can form during the loss VE-cadherin homotypic adhesion, wild type VE-cadherin or VE-cadherin cytoplasmic domain (⌬EXD) was expressed in sub-confluent endothelial cells. We observed that Cdc42 was activated in transfected cells and that these cells also developed Cdc42-dependent >70m-long plasma membrane protrusions. The formation of these structures required actin polymerization, and they developed specifically in endothelial cells as compared with epithelial cells. Expression of the VE-cadherin cytoplasmic domain lacking the ␤-catenin binding site also induced Cdc42 activation; thus, its activation cannot be ascribed to ␤-catenin binding. However, these cells were not able to form the protrusions. These results suggest that the cytoplasmic domain of non-junctional VE-cadherin can serve as a scaffold involved in Cdc42 activation at the endothelial plasma membrane. ␤-Catenin and the associated ␣-catenin may serve as support sites for actin polymerization, leading to formation of long plasma membrane protrusions. Thus, nonjunctional VE-cadherin actively participates in insideout signaling at the plasma membrane, leading to the development of endothelial membrane protrusions.
Cadherin molecules are membrane receptors involved in cellcell adhesion (1,2). They form a large family of proteins with tissue specificity that associates in a homophilic manner. They are indispensable in tissue formation during development, and in addition to adhesion, they serve morphogenetic functions (3)(4)(5)(6). Classical cadherins are anchored to actin cytoskeleton by ␣-catenin, which binds to ␤-catenin (7). Catenins have multiple functions in signal transduction and transcription regulation (8) and, thus, they are directly involved in the transduction of proliferation signals (6,9,10).
tk;2Endothelium forms a semi-permeable barrier that regulates the flux of liquid and solutes (11,12) as well as leukocyte transmigration from the microvasculature to sites of injury and infection (13)(14)(15). Adherens junctions formed between endothelial cells contain VE-cadherin, which is specifically expressed in the endothelium (16). Direct involvement of VE-cadherin to endothelial barrier function was demonstrated by VE-cadherin blocking antibodies in cell-culture and mouse models (17). In a different situation, VE-cadherin/catenins reorganize during polymorphonuclear or monocyte transmigration through the endothelial monolayer (18). VE-cadherin targeting revealed that in addition to its role in cell-cell adhesion, it is required for transduction of the vascular endothelial growth factor proliferation pathway through ␤-catenin association to phosphatidylinositol 3-kinase, which is downstream of vascular endothelial growth factor receptor 2. Ablation of VE-cadherin interferes with endothelial proliferation and causes apoptosis (6). However, it is remarkable that the VE-cadherin function was not compensated despite the presence of N-cadherin, which also associates with ␤-catenin and is expressed in endothelial cells. Thus, it is possible that the cytoplasmic domain of VE-cadherin serves functions that are not completely understood. Although VE-cadherin is classified as a "classical" cadherin in respect to its primary structure, the cytoplasmic tail is unique among the members of this subgroup (19). To study the role of VEcadherin in signaling mechanisms in endothelial cells, wild type VE-cadherin or the cytoplasmic domain of VE-cadherin mutant was expressed in sub-confluent endothelial cells. The cytoplasmic domain of cadherins has been characterized in a variety of studies as the "dominant negative mutant" for cellcell adhesion, since its expression blocked cell adhesion in cells of epithelial, neuronal, muscle, and endothelial origin (20 -24). 1 In endothelial cells a chimera between interleukin 2 receptor extracellular domain and VE-cadherin cytoplasmic domain interfered with proliferation by titrating-out ␤-catenin from its downstream effectors, and its stable expression in tissue culture cells led to down-regulation of endogenous VE-cadherin and defects in proliferation (25).
In the present study, we observed that non-junctional VEcadherin induced through its cytoplasmic domain the activation of Cdc42 and the formation of Ͼ70-m-long plasma membrane protrusions. Cdc42 activation was shown to be necessary for the formation of these protrusions. We further demonstrated that Cdc42 activation was not dependent upon ␤-catenin association to VE-cadherin. Two critical factors were required for the formation of the protrusions, (a) membrane localization of VE-cadherin and (b) association of ␤-catenin with its cytoplasmic domain. Furthermore, the formation of membrane protrusions was endothelial-specific since it did not occur in epithelial cells. These results suggest a novel role of VE-cadherin cytoplasmic domain in the formation of membrane protrusions, which may be involved in the restoration of endothelial junctional barrier function after the loss of homotypic VE-cadherin adhesion.

MATERIALS AND METHODS
Cell Culture-Human microvascular endothelial cells (HMECs) 2 were gifts from Dr. E. W. Ades (26) (NIAID, National Institutes of * 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  Health, Atlanta, GA). HMECs were cultured in MCDB 131 (Invitrogen) supplemented with 5% fetal bovine serum (Hyclone Laboratories, Logan, UT). Primary human pulmonary arterial endothelial cells were cultured in EBM 2 (endothelial basal medium) complete medium (Bio-Whittaker, Walkerville, MD). Bovine lung microvascular endothelial cells, purchased were from VEC Technologies (Renssalaer, NY), were grown in MCDB 131 medium supplemented with 5% fetal bovine serum and endothelial cell growth supplement (Sigma-Aldrich). Alveolar epithelial cells A549 were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum.
Antibodies-A monoclonal antibody against VE-cadherin was from Chemicon (Temecula, CA), and a goat polyclonal antibody was from Research Diagnostics (Flanders, NJ). Rabbit polyclonal anti-FLAG and anti-Myc-tag monoclonal were from Zymed Laboratories Inc. (San Franscisco, CA). Mouse monoclonal anti-FLAG (M2) was from Sigma-Aldrich. Monoclonal antibodies for ␤-catenin, p120, and Rac were from Transduction Laboratories (Lexington, KY). Rabbit anti-Cdc42 was from Santa Cruz (Santa Cruz, CA). Horseradish peroxidase-conjugated secondary antibodies were from Jackson Laboratories (West Grove, PA). For immunofluorescence experiments, we used either fluorescein isothiocyanate and Texas Red (Jackson Laboratories) or Alexa 488 and Alexa 568 (Molecular Probes, Eugene, OR)-conjugated antibodies. Actin was visualized with Alexa 488-conjugated phalloidin (Molecular Probes).
Drugs-Latrunculin A was purchased from Molecular Probes. VE-cadherin Constructs and Transfections-VE-cadherin-⌬EXD mutants with and without FLAG tag (DYKDDDDK) were constructed by PCR (plasmid pcDNA3-VEC-⌬EXD). The octapeptide sequence encoding for the FLAG-tag sequence replaced the VE-cadherin extracellular domain by PCR. The human wt VE-cadherin cDNA inserted in the EcoRI site of the pcDNA3 vector was used (plasmid pcDNA3-VEC) as the template. We used either primer D-FL containing the FLAG encoding sequence as sense primers (GAGTCGCAAGAATGCCGACTA-CAAGGACGACGATGACAAGACCTTCTGCGAGGATATGG) or primer D that lacks the FLAG encoding sequence (GTGAGTCGCAAGAATGC-CACCTTCTGCGAGGATATGG). Antisense primer 2002 (ACCTTCT-GCGAGGATATGG) was used in both cases. PCR products were digested with BsmI/BstEII, and these fragments were inserted in the BsmI (partial)/BstEII site of pcDNA3-VEC. The mutant version with FLAG (termed ⌬EXD) was used generally except where indicated ⌬EXD without FLAG.
To construct VE-cadherin-⌬EXD-⌬␤cat mutant, the plasmid pcDNA3-VEC-⌬EXD was digested with HindIII and Bpu10I (partial digestion and Klenow-blunt). This fragment was inserted in the HindIII/ EcoRV site of pcDNA3 (plasmid pcDNA3-VEC-⌬EXD-⌬␤cat). To make the VE-cadherin-⌬EXD-⌬p120 construct, residues 645-654 were deleted by PCR. We used as the sense primer TGGTCACCATGGACAC-CACCAGCTACGATG. To make the VE-cadherin-⌬EXD-G/A construct we substituted residues GGG (649 -651) with AAA by PCR using as sense primer TGGTCACCTACGACGAGGAGGCAGCAGCAGAGATG. For both PCR reactions, we used as the antisense primer TACAGCT-CAGCCAGCATCTTAAAC. PCR products containing the deletion or substitution were cloned in the TA cloning vector (Invitrogen) and digested with BstEII/BlpI fragments to replace the wild type BstEII/ BlpI sequence from the VE-cadherin-⌬EXD construct. All PCR products were verified by sequencing.
Transfection-Endothelial cells were transfected by electroporation or liposome-mediated DNA transfer (Qiagen, Valencia, CA). For electroporation, 10 6 cells were seeded in a 100-mm 2 dish the day before transfection. After 20 -24 h, cells were trypsinized and re-suspended in 300 l of medium. Plasmid DNA (10 g) mixed with 14 g of salmon sperm DNA was added, and cells were electroporated in 0.4-cm cuvettes (Gene Pulser II, Bio-Rad). The conditions were set at 180-mV voltage and 950-millifarad capacitance. To obtain confluent monolayers, 50 l of transfected cells were applied directly on a gelatinized coverslip and left for 1 h to adhere. Medium was added, and cells were fixed after 48 h. We typically observed 30% transfection efficiency with liposomes and 40% efficiency with electroporation.
Immunofluroscence Microscopy-Cells were grown on 1% gelatin (Sigma-Aldrich)-coated coverslips. They were fixed and stained as described previously (Kouklis et al. (29)). The stained cells were visualized either with Nikon Diaphot 200 or Zeiss Pascal confocal microscopes.
Protein Assay-Protein concentrations of cell extracts were quantified using the BCA protein assay from Pierce.
Immunoprecipitation and Western Blotting-Endothelial cells were washed, scraped with phosphate-buffered saline, and pelleted in 3 ϫ 10 3 rpm. Total cell extracts were made using immunoprecipitation buffer containing 20 mM Tris-Cl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, and 0.5% Nonidet P-40 supplemented with Complete protease inhibitor mixture (Roche Molecular Biochemicals). Cells were incubated for 10 min in 4°C in a shaker and centrifuged in 14 ϫ 10 3 rpm for 5 min to remove insoluble material. The supernatant was used for immunoprecipitation and Western-blotting experiments (thereafter called "extract"). Extracts were pre-cleared with 2 g of mouse IgG (Jackson Laboratories) and incubated with 2-3 g of the first antibody for 2 h in 4°C before the addition of 20 l of protein A/G-agarose (incubation for 1 h in 4°C). Immunoprecipitates were subjected to SDS-PAGE and Western blot analysis. Nitrocellulose membranes were blocked with TBST buffer (20 mM Tris, pH 7.4, 150 mM NaCl, 0.1% Tween 20) containing 1% gelatin (from cold water skin fish, Sigma-Aldrich), incubated for 1 h with primary antibodies, washed with TBST, and incubated for 1 h with horseradish peroxidase-conjugated secondary antibodies. For anti-Cdc42 antibodies, membranes were blocked with TBST containing 5% nonfat milk and incubated at room temperature for 3 h. The horseradish peroxidase signal was developed using Super Signal (Pierce).
HMEC were transiently transfected with wt VE-cadherin, ⌬EXD, or ⌬EXD-⌬␤cat mutants. Cells were harvested 24 h after transfection, and extracts were tested for Cdc42 activation using the GST-PBD pull-down assay. HMEC cell extracts were incubated with GST-PBD coupled to glutathione-Sepharose beads. Beads were washed and analyzed by Western blotting for Cdc42 binding. We observed that active Cdc42 was bound to PBD in extracts from HMECs transfected with wt VE-cadherin, -⌬EXD, or -⌬EXD-⌬␤cat (Fig. 2). Cdc42 activation remained at the basal level in mock-transfected cells (Fig.  2, GST-PBD). In control experiments, GST beads did not asso-ciate with Cdc42 under the same conditions (Fig. 2, GST).
Formation of Membrane Protrusions in Endothelial Cells Expressing VE-cadherin Cytoplasmic Domain (⌬EXD) or wt VE-cadherin-At 18 -24 h post-transfection of HMECs, the expressed proteins localized at the plasma membrane and the cytoplasm in vesicular structures. We observed that an unusually high percentage of transfected endothelial cells developed very long extensions with a striking cell shape change (Fig. 3A). These extensions typically showed branching patterns. All cells from randomly selected fields (a total of 50 -100 cells/experiment) were photographed to measure protrusion lengths. Cells with extensions greater than 70 m were scored as positive. Long extensions were evident in ϳ60% of cells transfected with ⌬EXD and ϳ30% of cells transfected with wild type VE-cadherin (Fig. 3B). As controls, we transfected HMEC with the cytoplasmic domain of another endothelial adherens junction protein, desmoplakin, known to associate with intermediate filaments (29); only 9% of these cells showed protrusions. Moreover, only 4% of mock-transfected HMEC showed such protrusions (data not shown). To rule out the possibility that cell shape changes were the result of a transfection artifact, we determined the earliest point that the protrusions were seen in relation to ⌬EXD expression. Protrusions were observed as early as 3 h post-transfection at a time when expression of transfected ⌬EXD was barely detectable (Fig. 3A). Mutant molecules localized at the plasma membrane, and protrusions reached maximum length within 6 h (as in Fig. 7C). We also expressed ⌬EXD in HMEC using retrovirus infection with identical results (data not shown), indicating the development of the protrusion was not secondary to the method used.
To extend these observations to other endothelial cells, we repeated the same transfection experiments in primary endothelial cells, human pulmonary arterial endothelial cells, and BMVEC (Fig. 3A); in both cases, membrane extensions formed 24 h after transfection of wt VE-cadherin and -⌬EXD. Endothelial cells transfected with ⌬EXD mutant were treated with the actin polymerization inhibitor, latrunculin A (30), and cells were fixed 24 h post-transfection. Latrunculin A in all cases prevented the formation of membrane protrusions (Fig. 3A), indicating the requirement of actin polymerization in the response.
We also transfected alveolar epithelial cells (A531, a type II alveolar epithelial cell line) with ⌬EXD to study protrusion formation in non-endothelial cells. No cell shape change or membrane extensions were observed in these cells (Fig. 3A), demonstrating the importance of cellular background in the mechanism of the response.
Membrane Protrusions in Endothelial Cells Expressing VEcadherin Cytoplasmic Domain Require Cdc42-In HMEC cotransfected with dominant-negative mutant N17Cdc42 and ⌬EXD mutant we observed that co-expression of N17Cdc42 inhibited the formation of protrusions (Fig. 4). We also cotransfected HMECs with dominant-negative mutants of Rac1 (N17Rac) or RhoA (N14RhoA) and the ⌬EXD mutant. N17Rac co-expression resulted in shorter and thinner protrusions in contrast to the inhibition seen with Cdc42 dominant negative mutant (Fig. 4). N14RhoA co-expression had no effect on the ⌬EXD-induced protrusion formation (Fig. 4).
To quantify the degree of inhibition caused by co-expression of dominant negative mutants of Rho, Rac, and Cdc42 along with ⌬EXD we measured protrusion lengths in a large number (90 -120) of double-transfected cells (Fig. 5). Expression of N17Cdc42 blocked protrusion formation in 100% of doubletransfected cells, whereas ϳ56% of N19Rho-expressing cells developed protrusions, a percentage very similar to ⌬EXDalone transfected cells. N17Rac blocked protrusion formation to a certain degree but not entirely, i.e. ϳ20% of N17Rac-expressing cells formed protrusions of Ͼ70 m. In addition, a significant percentage of double-transfected cells (ϳ15%) formed slightly shorter protrusions (60 -70 m) in these cells.
Differential Roles of ␤-Catenin and p120 in Mediating Endothelial Cell Shape Change-Expression of wt VE-cadherin and VE-cadherin cytoplasmic domain had profound effects on endothelial cell shape as shown above. Because the cadherin cytoplasmic domain associates with ␤and p120-catenins, we investigated the possible roles of these interactions in cell shape change. We generated VE-cadherin cytoplasmic domain mutants ⌬EXD-⌬␤cat lacking the binding site for ␤-catenin and ⌬EXD-G/A and ⌬EXD-⌬p120 lacking the binding site for p120. wt VE-cadherin, ⌬EXD, or ⌬EXD-⌬␤cat were transfected in HMEC, and their association with ␤-catenin was determined by immunoprecipitation and Western blotting. We showed that ␤-catenin associated with VE-cadherin and -⌬EXD and did not associate with the ⌬EXD-⌬␤cat mutant (Fig. 6A). Also, extracts
Because Cdc42 was activated in ⌬EXD-⌬␤cat-transfected endothelial cells (Fig. 2), we compared cell shape changes induced by expression of ⌬EXD-⌬␤cat and the activated mutant of Cdc42. Transfection of HMEC with the constitutively active Cdc42 mutant (V12Cdc42) induced a large number of filopodia (Fig. 6D). Expression of ⌬EXD-⌬␤cat and V12Cdc42 induced similar filopodia structures secondary to the activation of Cdc42. No striking cell shape changes were observed when ⌬EXD-⌬␤cat was co-transfected with N17Cdc42 (Fig. 6E), indicating that Cdc42 is involved in filopodia formation induced by ⌬EXD-⌬␤cat. N17Cdc42 was also shown to co-localize with ⌬EXD-⌬␤cat at distinct plasma membrane sites as shown by confocal microscopy (Fig. 6E).
Mutant ⌬EXD-G/A did not associate with p120 (Fig. 7A). To study the localization of ⌬EXD-G/A with endogenous p120, cells were fixed at 6 h post-transfection when mutant expression level was low and examined by confocal microscopy. The mutant localized at membrane patches distinct from p120 and cytoplasmic vesicular structures similar to ⌬EXD (Fig. 7, B and  C). Importantly, at this time point, ⌬EXD-G/A-transfected cells also developed long protrusions similar to those after expression of ⌬EXD; thus, formation of these protrusions is independent of p120 binding to the VE-cadherin cytoplasmic domain. Quantification of the effect of ⌬EXD-G/A expression in protrusion formation showed that ϳ55% of ⌬EXD-G/A-transfected cells developed protrusions (data not shown).
We generated an additional deletion mutant (⌬EXD-⌬p120) of VE-cadherin cytoplasmic domain lacking the entire domain for p120 binding (residues 645-654) (28). We observed that ⌬EXD-⌬p120 localized exclusively in the cytoplasm and that membrane protrusions did not develop in these cells (Fig. 7D). This finding suggests the importance of membrane localization in the mechanism of protrusion formation.

Activation of Cdc42 by VE-cadherin Cytoplasmic Domain
Signal Formation of Endothelial Membrane Protrusions-We expressed wt VE-cadherin or VE-cadherin cytoplasmic domain mutant in subconfluent endothelial cells. We showed that expression of either of these constructs induced Cdc42 activation and that the expressed cadherins localized at the plasma mem-  7. Plasma membrane localization of cadherin cytoplasmic domain but not p120 association is required for membrane protrusion formation. A, HMEC were transfected with ⌬EXD-⌬␤cat, ⌬EXD, ⌬EXD-G/A, or ⌬EXD-⌬p120 (lanes ⌬␤cat, ⌬EXD, G/A, and ⌬p120). Cells were lysed, and the mutants were immunoprecipitated (ip) with M2 monoclonal antibody. Immunoprecipitates were subjected to Western blot analysis using anti-p120 antibody. Note that ⌬EXD-G/A and ⌬EXD-⌬p120 did not associate with p120. B, HMEC were transfected with ⌬EXD, fixed 24 h post-transfection, and double-stained with anti-FLAG and anti-p120 antibodies. C, HMECs transfected with ⌬EXD-G/A, fixed 6 h post-transfection, and double-stained with anti-FLAG and anti-p120 antibodies. ⌬EXD-G/A induced protrusions similar to ⌬EXD. Arrowheads indicate ⌬EXD-G/A, and the arrow shows p120 localization at the membrane. D, HMECs transfected with ⌬EXD-⌬p120 fixed after 24 h and double-stained with anti-FLAG and anti-p120 antibodies. ⌬EXD-⌬p120 localized exclusively in the cytoplasm with no apparent cell shape changes in the transfected cells. All of the above experiments were repeated at least three times.
brane. These findings suggest that the non-junctional VE-cadherin and especially its cytoplasmic domain was involved in Cdc42 activation. Rho GTPases have been implicated in formation of actin-driven membrane rearrangements (31). Cdc42 in particular is considered as an important GTPase, signaling the formation of filopodia (32,33). We showed that expression of wt VE-cadherin and VE-cadherin cytoplasmic domain induced the formation of membrane protrusions. Co-expression experiments using dominant negative mutants of Rho, Rac, or Cdc42 along with ⌬EXD mutant demonstrated that formation of these protrusions was predominantly the result of Cdc42 activation. It is not clear how the cytoplasmic domain of VE-cadherin activates Cdc42. The guanine exchange factor Tiam1 is localized at the cell-cell junctions of epithelial cells and promotes invasiveness of T lymphoma cells across the epithelial barrier (34). It is possible that a similar guanine exchange factor may be localized at the VE-cadherin cytoplasmic domain that enables activation of Cdc42 in close proximity to the plasma membrane.
To rule out the possibility that formation of membrane protrusions induced by the VE-cadherin cytoplasmic domain is not restricted to a transformed endothelial cell line, we also studied two primary endothelial cells with similar results. Development of long membrane protrusions was clearly evident within 3 h after transfection of the ⌬EXD mutant; that is, at a time when mutant expression was barely detected. Thus, it is unlikely that the protrusions were the result of massive overexpression of the transfected mutant. In a control experiment, the expression of desmoplakin, a vimentin binding protein (35), did not affect endothelial cell morphology. In another control experiment expression of VE-cadherin cytoplasmic domain in type II alveolar epithelial cells (A531 cells) also failed to induce membrane protrusions, implying that the effect is endothelial cell-specific. The latter finding is consistent with observations in cultured epithelial cells that expression of E-or N-cadherin ⌬EXD mutant failed to induce a cell shape change (23,36). It is well established that cadherin and associated catenins mediate outside-in signaling (6,9,10). In the present study, we show that non-junctional VE-cadherin induces signaling initiated inside endothelial cells, which in turn induces a cell shape change involving formation of membrane protrusion that may restore junctional integrity after the loss of homotypic VEcadherin adhesion.
Different Roles of ␤-Catenin and p120 in Induction of Membrane Protrusions in Endothelial Cells-␤-and p120-catenins associate directly with cadherins at well conserved cytoplasmic domains (37,38). Because their association with VE-cadherin may be important in the formation of membrane protrusions, we studied their role by expressing the two VE-cadherin cytoplasmic domain mutants lacking either ␤-catenin or p120 binding sites. An extended domain of 100 residues in cadherin cytoplasmic domain was identified in mediating association with ␤-catenin (39), but a short conserved motif, SLSS, is required for this association (27,39). Expression of a construct (⌬EXD-⌬␤cat) in which this motif was deleted abolished the binding of this mutant to ␤-catenin without affecting the p120 association; thus, it is likely that ⌬EXD-⌬␤cat mutant was inserted in the membrane in its proper conformation. We showed that the expression of ⌬EXD-⌬␤cat induced the activation of Cdc42 in endothelial cells, but interestingly, this mutant resulted in the formation of thin, needle-like structures resembling filopodia, similar to those induced by expression of the constitutively active Cdc42 mutant. Therefore, VE-cadherin lacking the ␤-catenin binding site is unable to induce formation of the characteristic membrane protrusions seen with the expression of cytoplasmic domain of VE-cadherin. It is known that ␤-catenin mediates the association of actin cytoskeleton with cadherin through the actin binding proteins ␣-catenin (38), ␣-actinin, and vinculin (40); thus, ␤-catenin plays a crucial role as a linker regulating interaction of cadherin with actin. The differences in cell shape change induced by ⌬EXD versus ⌬EXD-⌬␤cat can be explained by the absence of ␤-catenin/␣catenin association in the latter case. Thus actin may fail to bind to VE-cadherin-catenin complex in an appropriate manner to form the characteristic long membrane protrusions.
Because it is possible that p120 may also contribute to the formation of membrane protrusions, we expressed a VE-cadherin mutant that was unable to associate with p120. Substitution of GGG to AAA (residues 649 -651) abolished p120 binding to VE-cadherin (mutant ⌬EXD-G/A), consistent with studies using E-cadherin (28). Interestingly, the expression of ⌬EXD-G/A mutant resulted in a similar phenotype as ⌬EXD, indicating that the p120-VE-cadherin association is not essential for protrusion formation. In other studies we compared the results of the ⌬EXD-G/A mutant with another mutant, ⌬EXD-⌬p120, lacking 10 residues at the p120 binding domain. This deletion mutant did not localize at the membrane, and it also failed to induce the cell shape change. Thus, these results suggest that sorting of VE-cadherin at the membrane is required for the formation of membrane protrusions.
In previous studies, expression of p120 was shown to induce membrane protrusions in fibroblasts and epithelial cells (a "dendritic" or "branching" phenotype) (41,42). The expression of cytosolic p120 inhibited RhoA activity in both studies. Noren et al. (42) show that p120 expression resulted in Rac1 and Cdc42 activation and that p120 co-precipitated with the guanine exchange factor vav2, suggesting a direct role of cytosolic p120 in Rac1 and Cdc42 activation. Anastasiadis et al. (41) propose that the branching phenotype could be the result of RhoA inhibition and the resultant loss of stress fibers and re-organization of ERM family proteins in the cortical cytoskeleton. Both studies suggest that p120-activated signaling was initiated in the cytoplasm. Our results in endothelial cells show that VE-cadherin is involved in Cdc42 activation at the plasma membrane, but they do not rule out a parallel regulation of GTPases in the cytoplasm induced by p120.
In the present study, we provide evidence for the an important role of VE-cadherin in the formation of long protrusions in endothelial cells. Our data show that membrane protrusions form as a result of actin polymerization through Cdc42 activation and the anchorage of the newly synthesized actin filaments to the VE-cadherin cytoplasmic domain. Actin-binding proteins associated with VE-cadherin-␤-catenin complex are necessary for the extension of actin cytoskeleton after Cdc42 activation. The dual role of VE-cadherin cytoplasmic domain in Cdc42 activation and formation of actin cytoskeleton extensions may be important in the reassembly of adherens junctions and restoration of endothelial barrier function upon the loss of adherens junctional integrity.