The Epidermal Growth Factor Receptor Modulates the Interaction of E-cadherin with the Actin Cytoskeleton*

Alterations in the expression or function of molecules that affect cellular adhesion and proliferation are thought to be critical events for tumor progression. Loss of expression of the cell adhesion molecule E-cadherin and increased expression of the epidermal growth factor receptor are two prominent molecular events that are associated with tumorigenesis. The regulation of E-cadherin-dependent cell adhesion by epidermal growth factor (EGF) was therefore examined in the human breast cancer cell line, MDA-MB-468. In this study, changes were observed in the subcellular distribution of components that mediate the cytoplasmic connection between E-cadherin and the actin-based cytoskeleton in response to activation of the EGF receptor. Serum withdrawal activated E-cadherin-dependent cell-cell aggregation in MDA-MB-468 cells, and this treatment stimulated the interaction of actin, α-actinin, and vinculin with E-cadherin complexes, despite the absence of α-catenin in these cells. By contrast, the co-precipitation of actin with E-cadherin was not detected in several α-catenin positive epithelial cell lines. Treatment with EGF inhibited cellular aggregation but did not affect either the levels of E-cadherin or catenin expression nor the association of catenins (β-catenin, plakoglobin/γ-catenin, or p120 cas ) with E-cadherin. However, EGF treatment of the MDA-MB-468 cell line dissociated actin, α-actinin, and vinculin from the E-cadherin-catenin complex, and this coincided with a robust phosphorylation of β-catenin, plakoglobin/γ-catenin, and p120 cas on tyrosine residues. Furthermore, inactivation of the EGF receptor in serum-treated MDA-MB-468 cells with either a function-blocking antibody or EGF receptor kinase inhibitors mimicked the effects of serum starvation by stimulating both cellular aggregation and assembly of E-cadherin complexes with vinculin and actin. These results demonstrate that the EGF receptor directly regulates cell-cell adhesion through modulation of the interaction of E-cadherin with the actin cytoskeleton and thus substantiates the coordinate role of both of these molecules in tumor progression and metastasis.

Tumor progression is associated with changes in the expression or activity of cell surface molecules from the growth factor receptor family (1)(2)(3) and the cadherin family of cell adhesion molecules (reviewed in Refs. 4 and 5). Abnormally high levels of the epidermal growth factor receptor (EGFR) 1 and one of its ligands, transforming growth factor ␣, have frequently been observed in human tumors and in tumor cell lines, and this is thought to play a critical role in malignant progression by increasing the transduction of mitogenic signals (6 -13). Similarly, down-regulation of the calcium-dependent cell adhesion molecule, E-cadherin, has been observed in poorly differentiated tumors and in highly invasive tumor cell lines (4, 5, 14 -17), suggesting that cell adhesion promoted by this molecule may be important for the maintenance of an epithelial phenotype and for the suppression of tumor invasion.
The roles of EGFR and E-cadherin in tumor progression have been explored using specific inhibitors and cDNA transfections. Perturbation of EGFR function with specific antibodies or kinase inhibitors was shown to inhibit proliferation of both tumor cell lines in vitro and tumor xenografts in vivo (reviewed in Ref. 13). Inactivation of E-cadherin function with neutralizing antibodies resulted in increased cellular proliferation and invasiveness (14,18). Recovery of E-cadherin function by cDNA transfection into E-cadherin-deficient tumor cell lines reversed the invasive phenotype and restored an epithelial morphology (19,20). Other transfection studies have demonstrated that changes in the expression levels of growth factor receptor tyrosine kinases and the cell adhesion molecule E-cadherin may be coordinated. For example, transfection with the Her2/neu oncogene down-regulated E-cadherin expression in a mammary cell line, and inhibition of Her2/neu autophosphorylation reversed this effect (21). Conversely, EGFR expression was shown to be down-regulated in a cervical carcinoma cell line after transfection with E-cadherin (22). It appears therefore that increased receptor tyrosine kinase activity and loss of E-cadherin function may be related cellular events that are associated with tumor progression.
In addition to the coordinate regulation of EGFR and Ecadherin expression levels, activation of EGFR as well as other tyrosine kinases has been shown to directly affect the adhesive function of E-cadherin via regulatory proteins known as catenins. Three catenins, ␤-catenin, plakoglobin/␥-catenin, and p120 cas , bind to the cytoplasmic domain of cadherins (23)(24)(25)(26) and, except for p120 cas , associate with the cortical actin cytoskeleton through ␣-catenin (27,28). Phosphorylation of catenins on tyrosine strongly correlates with modulation of cell adhesion. In particular, tyrosine phosphorylation of ␤-catenin was shown to be associated with loss of epithelial differentiation and increased cellular migration which was concomitant with decreased cadherin-dependent adhesion (29 -36). One study, however, has shown that tyrosine phosphorylation of ␤-catenin may not be required for modulation of cell adhesion as chimeric cadherin/␣-catenin molecules were sufficient to mediate the shift from strong to weak adhesion in the presence of v-src kinase (37). On the other hand, another study has shown that mutations in the region of ␤-catenin that is tyrosine-phosphorylated by erbB-2 results in suppression of the invasive phenotype (38). Therefore, it remains to be determined whether a direct causality exists between tyrosine phosphorylation of ␤-catenin and the loss of an adhesive phenotype. Tyrosine phosphorylation, however, did not change the levels of catenins that were associated with the cadherin cytoplasmic domain, implying that tyrosine phosphorylation may affect the interaction of catenins with adhesion components other than cadherins (26,31,34).
Several studies have implicated ␣-catenin as being particularly important in maintaining intercellular adhesion. Aberrant ␣-catenin expression has been observed frequently in human cancer cell lines with low adhesive activity (39 -43), and a mutated form of ␤-catenin lacking the ␣-catenin binding site has been detected in cancer cell lines (44). These observations suggested that the adhesive activity of cadherins is tightly regulated by ␣-catenin, possibly due to its role in linking cadherins to the cytoskeleton. In a recent study, the ␣-catenindeficient cell line MDA-MB-468 was found under conditions of serum starvation to regain E-cadherin-dependent adhesion (46). This adhesion was suggested to be mediated by vinculin which is a protein of adheren junctions sharing sequence similarity with ␣-catenin (47,48). This study implied that vinculin may connect the cadherin cytoplasmic domain to the cortical cytoskeleton in the absence of ␣-catenin by directly interacting with ␤-catenin.
In the present study, the relationship between tyrosine phosphorylation of the adhesion complex, mediated by EGFR, and the connection of E-cadherin with the actin-based cytoskeleton in the ␣-catenin-deficient MDA-MB-468 cell line were examined. The absence of ␣-catenin in this cell line appears to facilitate the isolation of detergent-soluble complexes of cadherins that contain components of the cortical cytoskeleton. Furthermore, the high levels of EGFR expression in this cell line (49) enable study of the modulation of cadherin interactions with components of the actin cytoskeleton by activation of EGFR. Here it is shown that, under serum-free conditions, actin, ␣-actinin, and vinculin co-precipitated with E-cadherin. The association of actin with E-cadherin was inhibited by cytochalasin D treatment, and this interaction was only detected in MDA-MB-468 cells but not in the MCF-10A, MDA-MB-361, or A431 cell lines, which all express ␣-catenin. E-cadherinmediated cell-cell aggregation in MDA-MB-468 cells was dramatically inhibited by EGF treatment, which correlated with prominent tyrosine phosphorylation of catenins (␤-catenin, plakoglobin/␥-catenin, and p120 cas ) and with the dissociation of actin, ␣-actinin, and vinculin from E-cadherin complexes. Furthermore, inactivation of EGFR in serum-treated MDA-MB-468 cells with a function-blocking antibody or with EGFR kinase inhibitors resulted in induction of E-cadherin-dependent adhesion and in the association of the E-cadherin complex with actin and vinculin. These findings demonstrate an interaction between E-cadherin and the cortical cytoskeleton which is sensitive to the activity of the EGF receptor. Furthermore, these results establish a molecular link between EGFR activity, cadherin-based adhesion, and the actin-cytoskeleton.

MATERIALS AND METHODS
Cell Lines-The breast cancer cell lines MDA-MB-468, MDA-MB-361, and BT549, the normal mammary epithelial cell line MCF-10A, and the epidermoid A431 carcinoma cell line were all obtained from the American Type Culture Collection (Rockville, MD). Cells were routinely cultured in Dulbecco's modified Eagle's medium/F12 (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (Life Technologies, Inc.) at 37°C in a humidified 5% CO 2 atmosphere.
Reagents-Epidermal growth factor was purchased from Collaborative Research (Becton Dickinson, Bedford, MA). Geldanamycin was purchased from Life Technologies, Inc. The EGFR-specific tyrosine kinase inhibitor PD 130305 was a gift from Dr. David Fry (Parke Davis). Cytochalasin D and DMSO were purchased from Sigma.
Antibodies-Monoclonal anti-human E-cadherin was acquired from Zymed (San Francisco). Monoclonal antibodies to ␣-, ␤-, plakoglobin/␥catenin and p120 cas were purchased from Transduction Laboratories (Lexington, KY). Ascites derived anti-vinculin or ␣-actinin and rabbit polyclonal anti-actin antibodies were purchased from Sigma. The monoclonal anti-phosphotyrosine antibody (G410) was obtained from Upstate Biotechnology. mAb 225, recognizing the extracellular domain of the EGFR, was a gift from Dr. John Mendelsohn (MD Anderson Cancer Center), and rabbit antiserum to EGFR was provided by Dr Joseph Schlessinger (New York University).
Cell Treatments-Confluent monolayers were either incubated in medium with 10% FBS (serum-treated) or incubated for 36 h in medium with 0% FBS (serum-starved). Serum-starved monolayers were then treated with or without 200 ng/ml EGF for 10 min at 37°C prior to cell lysis. For Geldanamycin or PD130305 treatments, confluent monolayers were split at 1:10 dilution, allowed to grow for an additional 48 h to reach about 50 -70% confluence, and were treated for 2 or 18 h with various concentrations of inhibitors as indicated. Control monolayers were treated with DMSO alone.
Immunoblotting-30 g of proteins from each indicated extract, as determined by the Bradford method (Bio-Rad), were boiled in SDS sample buffer for 10 min and loaded onto a 7.5% polyacrylamide minigel (Bio-Rad). Proteins were transferred onto Immobilon membranes (Millipore, Bedford, MA), blocked in 3% BSA/phosphate-buffered saline and incubated for 2 h at 25°C with dilutions of 1:1000 of primary antibodies. After washes, membranes were probed with a 1:5000 dilution of secondary antibodies coupled to horseradish peroxidase for 1 h at 25°C and developed with enhanced chemiluminescence (ECL) reagents (Amersham Pharmacia Biotech).
Immunoprecipitation-Cell lysates (5 mg) were incubated with 10 g of primary antibodies for 2 h at 4°C and supplemented with 3 mg of Protein A-Sepharose beads for 1 h, and the beads were subsequently washed with lysis buffer. Bound proteins were eluted by boiling the beads in sample buffer for 10 min, processed by SDS-PAGE, transferred onto membranes, and probed with 1:1000 dilutions of indicated antibodies. Blots were further processed as described above.
Cell-Cell Aggregation Assays-Cell monolayers grown to 98% confluence were incubated for 36 h in medium containing 10% FBS (serumtreated) or 0% FBS serum (serum-starved). Cell monolayers were detached by incubating in HBSS containing 0.02% crystallized trypsin (Worthington) and 10 mM CaCl 2 for 5 min at 37°C. After trypsinization, single cell suspensions were made by trituration with a Pasteur pipette. Cell viability as assessed by trypan blue dye exclusion was greater than 95%. Cells were washed twice in HBSS (Life Technologies, Inc.) and incubated at 3 ϫ 10 5 cells per well in 500 l of HBSS containing 1% BSA and 100 g/ml DNase (Worthington) with or without 1 mM CaCl 2 and in the presence or absence of 100 g/ml anti E-cadherin or control antibodies as indicated.
For EGF treatment, single cell suspensions prepared as described above were incubated with 200 ng/ml EGF in the aggregation wells for the times indicated. A sample of each treatment was further controlled for EGFR levels by Western blotting.
Aggregation assays were performed at 37°C at 100 rpm for the indicated times in triplicate wells, in 24-well non-tissue culture treated plates (No. 1147; Becton Dickinson, Franklin Lakes, NJ) that had been blocked with phosphate-buffered saline, 2% BSA for 30 min at 37°C.
Assays were stopped at time 0 and 30 min (unless otherwise indicated) by fixing the cells in 1% glutaraldehyde. The extent of cell-cell binding was monitored by measuring the disappearance of single cells using a Coulter counter. Standard deviations of the mean values are included.

Serum Withdrawal Increases the Levels of E-cadherin Complexes with the Actin Cytoskeleton-
In the absence of serum, MDA-MB-468 cells display E-cadherin-dependent aggregation which correlated with a 3-fold increase in E-cadherin protein levels. This increase in E-cadherin was accompanied by the association of vinculin with the E-cadherin⅐␤-catenin complex (46). This result prompted the examination of additional cytoskeletal proteins that may be recruited to the E-cadherin adhesion complex to support cell-cell adhesion in these cells as a consequence of serum withdrawal. High levels of actin (Fig.  1A, lane 2) and ␣-actinin (Fig. 1A, lane 5), in addition to vinculin (Fig. 1A, lane 8), were observed to be associated with E-cadherin immunoprecipitates. In contrast, actin, ␣-actinin, or vinculin were not detected in immunoprecipitates of E-cadherin from serum-treated cells (Fig. 1A, lanes 1, 4 and 7, respectively). To control for specificity of the antibodies, total cell lysates were reacted with antibodies to each of these proteins (Fig. 1A, lanes 3, 6, and 9). Background immunoreactivity was consistently observed when E-cadherin immunoprecipitates were probed with the monoclonal antibody to ␣-actinin (Fig. 1A, lanes 4 and 5). When 3 times the amount of lysate was used for immunoprecipitation from serum-treated cells to account for the reduced amount of E-cadherin (Fig. 1B, lanes 1  and 2), similar levels of actin (Fig. 1B, lanes 3 and 4), vinculin (Fig. 1B, lanes 5 and 6), and ␣-actinin (data not shown) were found in the immune complexes, which suggests that the association of these molecules with E-cadherin in serum-treated MDA-MB-468 cells is restricted by the low levels of E-cadherin expression.
The co-precipitation of actin with E-cadherin was completely abolished when serum-starved MDA-MB-468 cells were treated with the actin depolymerizing agent, cytochalasin D, demonstrating that the filamentous form of actin (F-actin) specifically interacts with the E-cadherin complex (Fig. 1C,  lanes 1 and 2). In contrast, cytochalasin D did not alter the levels of actin monomers in total cell lysates (Fig. 1C, lanes 3  and 4). Since extraction of cells with Triton X-100 normally removes F-actin from the soluble fraction, the association of E-cadherin with F-actin observed in Triton X-100 extracts may be due to the presence of soluble F-actin fragments that may have resulted from either differential polymerization or mechanical shearing during cell extraction (53,54). Nevertheless, these results demonstrate that E-cadherin interacts with key components of the actin-based cytoskeleton in MDA-MB-468 cells, and these interactions may support cell-cell adhesion in the absence of ␣-catenin.
Detergent-soluble Complexes of Actin with E-cadherin Are Not Observed in ␣-Catenin Positive Cell Lines-Actin had not been previously shown to co-precipitate with E-cadherin from detergent-soluble lysates, presumably because this complex is believed to be detergent-insoluble in most cell lines (45,55,56). To explore whether complexes of E-cadherin with actin are detectable in other epithelial cell lines, analysis for the coprecipitation of these molecules was carried out in MCF-10A, MDA-MB-361, and A431 cells, all of which express normal levels of ␣-catenin and display aggregation properties that are not affected by serum starvation (Ref. 46 and data not shown). In contrast to MDA-MB-468 cells, actin was not found in immunoprecipitates of E-cadherin from lysates of the epithelial cancer cell lines MCF-10A, MDA-MB-361, or A431 cells (Fig. 2,  lanes 2, 5 and 8, respectively), even though the levels of immu-noprecipitated E-cadherin were similar to those found in serum-starved MDA-MB-468 cells (Fig. 2, lanes 1, 4, 7 and 10, respectively). Actin was present in the detergent-soluble fraction of these cells (Fig. 2, lanes 3, 6 and 9), indicating that the lack of actin in E-cadherin immunoprecipitates from these cells was not due to the insolubility of actin in these lysates. These results show that the absence of detectable actin in E-cadherin  1 and 2), anti-actin (lanes 3 and 4), and anti-vinculin (lanes 5 and 6). C, lysates (5 mg) from serum-starved MDA-MB-468 monolayers that were treated with DMSO alone or 1 M cytochalasin D in DMSO for 1 h at 37°C were immunoprecipitated with anti-E-cadherin antibodies and immunoblotted with anti-actin antibodies (lanes 1 and 2). 30 g of cell lysate from each treatment were included as controls (lanes 3 and 4).
immunoprecipitates in these cell lines correlates with their expression of ␣-catenin.
EGF Treatment Attenuates Cell-Cell Adhesion and Dissociates Cytoskeletal Components from the E-cadherin Complex-EGF treatment has been shown to promote cell scattering and reduce cell-cell adhesion by an unknown mechanism that correlated with increased tyrosine phosphorylation of ␤-catenin (34,35,36). The effect of EGFR activation on cell-cell aggregation was therefore compared with its effects on the association of cytoskeletal proteins with E-cadherin in MDA-MB-468 cells. EGF caused an abrupt decline in E-cadherin-dependent aggregation after 10 min of EGF treatment of serum-starved cells (Fig. 3A). The effect of EGF on cell-cell aggregation was abolished after 30 min of treatment, consistent with decreased EGFR levels due to down-regulation of the protein (Ref. 9, data not shown). The total levels of E-cadherin or catenins (␤-catenin, plakoglobin/␥-catenin, and p120 cas ) were not reduced by treatment with EGF (Fig. 3B, lanes 1-8) and therefore do not account for the reduced cellular aggregation caused by EGFR stimulation.
To verify whether the EGF-induced dissociation of cytoskeletal components from E-cadherin was accompanied by tyrosine phosphorylation of the adhesion complex, the phosphotyrosine levels of ␤-catenin, plakoglobin/␥-catenin, p120 cas , and vinculin were examined by immunoprecipitation (Fig. 5). High levels of phosphorylated EGFR were observed upon treatment with EGF (Fig. 5, lane 2), consistent with EGF-induced receptor autophosphorylation (9). Immunoprecipitation with antibodies to ␤-catenin revealed three proteins that were tyrosine-phosphorylated upon EGF treatment, including ␤-catenin and two other proteins that corresponded in molecular weight to plakoglobin/␥-catenin and p120 cas (Fig. 5, lane 4). In addition, immunoprecipitation with antibodies to plakoglobin/␥-catenin (Fig. 5, lane 6) and p120 cas (Fig. 5, lane 8) revealed that both of these molecules were tyrosine-phosphorylated in response to EGF. It remains possible that the tyrosine-phosphorylated 120-kDa band observed in catenin immunoprecipitates (Fig. 5,   lanes 4, 6, and 8) may be E-cadherin as it co-migrates with p120 cas in SDS-PAGE. However, since tyrosine phosphorylation of E-cadherin has not been reported and was not detected in EGF-treated MDA-MB-468 cells (data not shown), the identity of the 120-kDa protein band is most likely p120 cas . Additionally, immunoprecipitation with vinculin antibodies (Fig. 5, lane 10) revealed faint tyrosine-phosphorylated bands corresponding in molecular weight to EGFR, vinculin or p120 cas , ␤-catenin, and plakoglobin/␥-catenin (Fig. 5, lane 10). Since vinculin and p120 cas co-migrate in SDS-PAGE, it is not possible to discriminate between these two proteins. In contrast to the robust phosphorylation observed in response to EGF treatment, very little tyrosine phosphorylation of the above proteins was detected in untreated cells (Fig. 5, lanes 1, 3, 5, 7, and 9). These results suggest that activation of the EGF receptor disrupts cell-cell adhesion by uncoupling actin, ␣-actinin, and vinculin from the E-cadherin adhesion complex, an event that may be caused by phosphorylation of ␤-catenin or plakoglobin/ ␥-catenin on tyrosine residues. p120 cas , although tyrosinephosphorylated, likely does not regulate the linkage of E-cadherin with the cortical cytoskeleton as it does not associate with ␣-catenin (26), but it remains possible that it does interact with vinculin. Furthermore, the co-precipitation of ␤-catenin, plakoglobin/␥-catenin, and p120 cas may result from the simul-  1, 4, 7, and  10) or polyclonal anti-actin antibodies (lanes 2, 5, and 8). 30 g of cell lysate from each cell line (L) was reacted with anti-actin antibodies (lanes 3, 6, and 9). E-cad, E-cadherin.

FIG. 3. EGF inhibits E-cadherin-dependent adhesion induced by serum starvation with no effect on the levels of E-cadherin or catenins.
A, serum-starved MDA-468 cell monolayers (ϪFBS) were made into single cell suspensions by low trypsin, 10 mM CaCl 2 treatment and aggregated in HBSS, 1% BSA Ϯ 1 mM CaCl 2 both in the presence or absence of 100 g/ml anti-E-cadherin antibodies for 10 and 30 min (data not shown) and in the presence or absence of 200 ng/ml EGF at both time points. The accumulation of aggregates at times 0, 10, and 30 min was determined using a Coulter counter and was of the same magnitude at 10 and 30 min in untreated samples. Serum-treated monolayers (ϩFBS) were aggregated for 10 min with no additions. The results are average Ϯ S.E. of six separate experiments. B, serum-starved MDA-468 monolayers were treated with or without 200 ng/ml EGF for 10 min at 37°C and were lysed in 1% Triton X-100 buffer including tyrosine phosphatase inhibitors. Cell lysates (5 mg), were immunoprecipitated (IP) with antibodies to E-cadherin (E-cad) (lanes 1 and 2), ␤-catenin (␤-cat) (lanes 3 and 4), plakoglobin/␥-catenin (␥-cat) (lanes 5 and 6), and p120 cas (lanes 7 and 8). Proteins were resolved by SDS-PAGE and probed with the same antibodies used for immunoprecipitation. taneous co-precipitation of all three catenins with EGFR (Ref. 32 and data not shown).

Inactivation of EGFR Function Stimulates Cell-Cell Adhesion and Association of E-cadherin with the Actin Cytoskeleton in Serum-treated MDA-MB-468
Cells-To verify whether EGFR activity was directly associated with the non-adhesive state of serum-treated MDA-MB-468 cells, cells were incubated with EGFR function-blocking reagents, and both cell-cell aggregation and the co-precipitation of actin and vinculin with E-cadherin were examined. Treatment of serum-treated MDA-MB-468 cells for 18 h with either the EGFR function-blocking antibody (mAb 225) (52) or Geldanamycin (50) down-regulated EGFR levels (Fig. 6A, lanes 1-4) and caused a 16-and 14-fold increase in E-cadherin-dependent cell adhesion, respectively (Fig. 6C). Similarly, treatment of MDA-MB-468 cells with 1 M PD 130305, a specific EGFR tyrosine kinase inhibitor (51), which inhibited EGFR autophosphorylation at 0.1 M (Fig. 6B,  lanes 1-3), caused a 9-fold increase in E-cadherin-dependent cell-cell aggregation (Fig. 6C). In comparison, the increase in cell aggregation induced by serum starvation was of greater magnitude (32-fold) (see Fig. 3A) and may be caused by the more complete inhibition of EGFR autophosphorylation (Fig. 5,  lane 1).
The effects of EGFR inhibitors on cell-cell aggregation were directly associated with changes in the co-precipitation of actin and vinculin with E-cadherin complexes. Short (2 h) or long term (18 h) treatment of MDA-MB-468 cells with 1 M PD 130305 or 1 M Geldanamycin (18 h) stimulated the co-precipitation of actin (Fig. 7, lanes 1-7) and vinculin (Fig. 7, lanes 8,  9 and 10) with E-cadherin. These results demonstrate that inactivation of EGFR induces the interaction of the E-cadherin adhesion complex with the actin-based cytoskeleton. DISCUSSION The present study examines the mechanism of regulation of E-cadherin-dependent cell adhesion by tyrosine phosphorylation. The effect of EGFR activation on cell adhesion was compared with the association of E-cadherin complexes with actin and the actin bundling proteins, vinculin and ␣-actinin. The MDA-MB-468 cell line was used as a model because it has been shown to express high levels of the EGF receptor (49) and contains complexes of cadherins with the cytoskeleton which are detergent-extractable, thus permitting the biochemical analysis of these molecular interactions (46). A recent study FIG. 4. EGF inhibits the interaction of E-cadherin with the actin cytoskeleton but does not alter the association of catenins with E-cadherin. A, serum-starved MDA-468 monolayers were incubated with or without 200 ng/ml EGF for 10 min at 37°C and lysed in 1% Triton X-100 buffer including tyrosine phosphatase inhibitors. Cell lysates (5 mg) were immunoprecipitated (IP) with anti-E-cadherin antibodies. Proteins were resolved by SDS-PAGE and probed with the following antibodies: polyclonal anti-actin (lanes 1 and 2), monoclonal anti-␣-actinin (lanes 3 and 4), and monoclonal anti-vinculin (lanes 5 and 6). Control or EGF-treated cell lysates from serum-starved MDA-468 monolayers were immunoprecipitated with anti-E-cadherin antibodies, electrophoresed, and probed with antibodies to ␤-catenin (lanes 1 and  2), plakoglobin/␥-catenin (lanes 3 and 4), and p120 cas (lanes 5 and 6).
FIG. 5. EGF stimulates tyrosine phosphorylation of ␤-catenin, plakoglobin/␥-catenin, and p120 cas . Serum-starved MDA-468 monolayers were treated with or without 200 ng/ml EGF for 10 min at 37°C and lysed in 1% Triton X-100 lysis buffer including tyrosine phosphatase inhibitors. 0.5 mg of cell lysate was used for detection of phosphorylated EGFR, and 5 mg of total cell lysate was used for detection of all other proteins. Cell lysates were immunoprecipitated (IP) with antibodies to EGFR (lanes 1 and 2), ␤-catenin (␤-cat) (lanes 3 and 4), plakoglobin/␥-catenin (␥-cat) (lanes 5 and 6), p120 cas (lanes 7 and 8), and vinculin (Vinc) (lanes 9 and 10). After SDS-PAGE, proteins were probed with anti-phosphotyrosine antibodies (Ptyr). showed that high levels of vinculin were associated with the E-cadherin/␤-catenin complex in these ␣-catenin-deficient cells, and vinculin was shown to bind ␤-catenin, suggesting that it could mediate the interaction of E-cadherin with the actin-based cytoskeleton (46). In the present study, the loss of cell adhesion caused by EGF treatment was associated with the decoupling of actin, ␣-actinin, and vinculin from E-cadherin complexes. The withdrawal of these cytoskeletal components from the E-cadherin adhesion complex was accompanied by a dramatic phosphorylation of ␤-catenin, plakoglobin/␥-catenin, and p120 cas on tyrosine residues. These results suggest a model whereby tyrosine phosphorylation of ␤-catenin or plakoglobin/ ␥-catenin causes vinculin to dissociate from the E-cadherincatenin complex, thus causing the loss of cell-cell adhesion in these cells.
Previous studies have indirectly linked tyrosine phosphorylation of ␤-catenin to reduced cell-cell adhesion and increased cell migration (29 -31, 34, 36), but no mechanism was elaborated in these studies. In the present study, while tyrosine phosphorylation did not affect the association of catenins with E-cadherin, it was associated with the loss of actin, ␣-actinin, and vinculin from E-cadherin immunoprecipitates. In accord with some studies (37), vinculin tyrosine phosphorylation was undetected in these cells, although it has been found at low levels in src transformed cells (57,58). The results from the present study therefore support a model whereby tyrosine phosphorylation modulates the interaction of the E-cadherincatenin complex with the actin cytoskeleton. In this model, vinculin dissociation from phosphorylated ␤-catenin or plakoglobin/␥-catenin also removes ␣-actinin and actin from the adhesion complex; these two molecules have both been shown to be associated with vinculin (59,60). However, EGF appears to affect more dramatically the association of E-cadherin with actin and ␣-actinin than with vinculin, which suggests that other molecules may be involved in the linkage of the E-cadherin complex with the actin cytoskeleton.
The high detergent extractability of cadherin complexes with the actin cytoskeleton appears to correlate with the absence of ␣-catenin expression in MDA-MB-468 cells. Cadherin complexes with actin were not detectable in ␣-catenin positive cell lines, and except for studies showing binding of E-cadherincatenin complexes with the actin-binding protein DNase I (61) or the co-sedimentation of filamentous actin (F-actin) with ␣-catenin in vitro (28), no previous evidence exists for the interaction of E-cadherin with F-actin in epithelial cells in vivo. The adhesion proteins, E-cadherin, ␤-catenin, and plakoglobin/ ␥-catenin in MDA-MB-468 cells, were predominantly found in the Triton-soluble pool, and a comparable amount of these proteins was observed in both soluble and insoluble-Triton fractions of ␣-catenin positive cell lines (data not shown). These results are consistent with the notion that coupling of the E-cadherin-catenin complex to the actin cytoskeleton via ␣-catenin renders these complexes insoluble in Triton X-100, which results in their removal from the soluble pool (45,55,56). In contrast, both actin and ␣-actinin have been found in Tritonsoluble immunoprecipitates of N-cadherin, in chicken retinal cells, and in fibroblasts (62,63). The greater detergent solubility of complexes of actin with N-cadherin may be due to a more labile association of N-cadherin with the cytoskeleton. These results are consistent with the proposed dynamic nature of N-cadherin-mediated adhesion (64,65).
The conditions used in this study appear to permit the extraction of the filamentous form of actin in the Triton-soluble phase since the association of actin with E-cadherin in the Triton-soluble fraction was inhibited by the actin depolymerizing agent, cytochalasin D. Triton X-100 extraction is known to preserve the integrity of the actin cytoskeleton (53,54), and thus the interaction of E-cadherin with the actin cytoskeleton has been observed in Triton X-100-insoluble fractions (45). However, F-actin is generally assumed to be associated with the Triton-insoluble pool (53,54), and thus its association with E-cadherin complexes may not be detected in soluble Triton X-100 extracts. In the present study, shorter fragments of F-actin that may be either arrested in an earlier stage of polymerization or that result from mechanical shearing during cellular extraction (53, 54) may allow the detection of E-cadherin/actin interactions in the soluble phase. It remains also an intriguing possibility that vinculin may couple the E-cadherin complex to a pool of F-actin that is distinguishable in its Triton solubility from that which is coupled through ␣-catenin.
A model emerges from the collective data in which ␣-catenin and vinculin may differentially support cell adhesion. The increased solubility of cadherin-vinculin-actin complexes implies that vinculin may be promoting a more dynamic interaction of cadherins with the cytoskeleton than does ␣-catenin. The reversible effects of serum starvation and EGF on both cell-cell aggregation and the association of E-cadherin with the cytoskeleton in MDA-MB-468 cells suggest that the adhesion promoted by vinculin may also be more sensitive to changes in growth factors levels. In support of this hypothesis, in the ␣-catenin positive A431 or BT549 cell lines, which also express both EGFR and vinculin (32,46), EGF had no effect on both cell-cell adhesion or on the association of ␣-catenin with cadherin complexes, even though it induced ␤-catenin tyrosine phosphorylation. 2 These results raise the possibility that vinculin may bind ␤-catenin with a lower affinity than does ␣-catenin or to a site on ␤-catenin distinct from the ␣-catenin binding site that is sensitive to tyrosine phosphorylation.
Other possible mechanisms that modulate the connection of E-cadherin with the cytoskeleton may be downstream of EGFR tyrosine phosphorylation. One such mechanism may be the activation of the Rho family of small GTP-binding proteins that is dependent on tyrosine kinase activity (66) and results in rapid reorganization of the actin cytoskeleton (reviewed in Refs. 67 and 68) and decreased cadherin activity (69). In addition, vinculin function may be regulated by depletion of phosphatidylinositol pools in response to EGF (for review, see Ref. 70). Phosphatidylinositol has been shown to regulate vinculin activity by inducing conformational changes that unfold the tail from the head domain of vinculin (59,60), thus exposing both the ␣-actinin and actin binding sites (71,72). Thus, reduced levels of phosphatidylinositol as a result of breakdown mediated by EGF stimulation may convert vinculin from an open/active to a close/inactive conformation that is unable to bind actin, ␣-actinin, and possibly ␤-catenin. Another likely possibility is that EGF-induced tyrosine phosphorylation of paxillin, which may increase its affinity for vinculin (73), could result in recruitment of vinculin from adherens junctions to focal contacts, thus resulting in weaker cell-cell adhesion.
The present study has shown that EGFR directly modulates E-cadherin-dependent adhesion through the connection of Ecadherin to the cytoskeleton and that vinculin may play a significant role in this modulation. Elucidation of the precise roles of the various intracellular proteins associated with cadherins will provide a further understanding of cadherin-based cellular adhesion.