Overexpression of E-cadherin on Melanoma Cells Inhibits Chemokine-promoted Invasion Involving p190RhoGAP/p120ctn-dependent Inactivation of RhoA*

Melanoma cells express the chemokine receptor CXCR4 that confers high invasiveness upon binding to its ligand CXCL12. Melanoma cells at initial stages of the disease show reduction or loss of E-cadherin expression, but recovery of its expression is frequently found at advanced phases. We overexpressed E-cadherin in the highly invasive BRO lung metastatic cell melanoma cell line to investigate whether it could influence CXCL12-promoted cell invasion. Overexpression of E-cadherin led to defective invasion of melanoma cells across Matrigel and type I collagen in response to CXCL12. A decrease in individual cell migration directionality toward the chemokine and reduced adhesion accounted for the impaired invasion. A p190RhoGAP-dependent inhibition of RhoA activation was responsible for the impairment in chemokine-stimulated E-cadherin melanoma transfectant invasion. Furthermore, we show that p190RhoGAP and p120ctn associated predominantly on the plasma membrane of cells overexpressing E-cadherin, and that E-cadherin-bound p120ctn contributed to RhoA inactivation by favoring p190RhoGAP-RhoA association. These results suggest that melanoma cells at advanced stages of the disease could have reduced metastatic potency in response to chemotactic stimuli compared with cells lacking E-cadherin, and the results indicate that p190RhoGAP is a central molecule controlling melanoma cell invasion.

Although melanoma only accounts for 5% of skin cancers, when metastasis starts, it is responsible for 80% of deaths from skin cancers (7). Melanocytes express E-cadherin (8 -10), but melanoma cells at early radial growth phase show a large reduction in the expression of this cadherin, and surprisingly, expression has been reported to be partially recovered by vertical growth phase and metastatic melanoma cells (9,11,12).
Trafficking of cancer cells from primary tumor sites to intravasation into blood circulation and later to extravasation to colonize distant organs requires tightly regulated directional cues and cell migration and invasion that are mediated by chemokines, growth factors, and adhesion molecules (13). Solid tumor cells express chemokine receptors that provide guidance of these cells to organs where their chemokine ligands are expressed, constituting a homing model resembling the one used by immune cells to exert their immune surveillance functions (14). Most solid cancer cells express CXCR4, a receptor for the chemokine CXCL12 (also called SDF-1), which is expressed in lungs, bone marrow, and liver (15). Expression of CXCR4 in human melanoma has been detected in the vertical growth phase and on regional lymph nodes, which correlated with poor prognosis and increased mortality (16,17). Previous in vivo experiments have provided evidence supporting a crucial role for CXCR4 in the metastasis of melanoma cells (18).
Rho GTPases control the dynamics of the actin cytoskeleton during cell migration (19,20). The activity of Rho GTPases is tightly regulated by guanine-nucleotide exchange factors (GEFs), 4 which stimulate exchange of bound GDP by GTP, and inhibited by GTPase-activating proteins (GAPs), which promote GTP hydrolysis (21,22), whereas guanine nucleotide dissociation inhibitors (GDIs) appear to mediate blocking of spontaneous activation (23). Therefore, cell migration is finely regulated by the balance between GEF, GAP, and GDI activities on Rho GTPases. Involvement of Rho GTPases in cancer is well documented (reviewed in Ref. 24), providing control of both cell migration and growth. RhoA and RhoC are highly expressed in colon, breast, and lung carcinoma (25,26), whereas overexpression of RhoC in melanoma leads to enhancement of cell metastasis (27). CXCL12 activates both RhoA and Rac1 in melanoma cells, and both GTPases play key roles during invasion toward this chemokine (28,29).
Given the importance of the CXCL12-CXCR4 axis in melanoma cell invasion and metastasis, in this study we have addressed the question of whether changes in E-cadherin expression on melanoma cells might affect cell invasiveness. We show here that overexpression of E-cadherin leads to impaired melanoma cell invasion to CXCL12, and we provide mechanistic characterization accounting for the decrease in invasion.

EXPERIMENTAL PROCEDURES
Cells, Antibodies, and Reagents-BLM human melanoma cells were cultured as described previously (28). Anti-E-cadherin antibodies were provided by Dr. Amparo Cano (Instituto de Investigaciones Biomédicas, Madrid, Spain) and were also purchased from R & D Systems (Minneapolis, MN). Anti-␤catenin antibodies were from Cell Signaling Technology (Danvers, MA); anti-␣-catenin, anti-p120ctn, and anti-Rac1 were from Pharmingen; anti-RhoA and anti-phosphotyrosine PY20 were from Santa Cruz Biotechnology (Santa Cruz, CA); antivinculin and anti-␤-actin were from Sigma; anti-GFP was from Molecular Probes (Eugene, OR); anti-p190RhoGAP was from Upstate Biotechnology, Inc. (Lake Placid, NY); and anti-CXCR4 was from R & D Systems. We also obtained anti-p120ctn antibodies from Dr. Albert B. Reynolds (Vanderbilt University, Nashville, TN), whereas control P3X63 and anti-␤1 integrin Lia1/2.1 were provided by Dr. Francisco Sánchez-Madrid (Hospital de la Princesa, Madrid, Spain). Alexa 633-phalloidin was purchased from Molecular Probes; fibronectin was from Roche Diagnostics, and Src inhibitors PP2 and PP3 were from Calbiochem-Novabiochem.
Vectors, siRNA, Transfections, and Infections-The E-cadherin expression vector pBAT-EM2 and the empty vector pBAT-neo were gifts from Dr. Amparo Cano. Vectors coding for GFP-fused forms of wild type RhoA and Rac1 and activated V14-RhoA and V12-Rac1 were gifts from Dr. Francisco Sánchez-Madrid, and pEGFP-p190RhoGAP and pEGFP-p120ctn vectors were from Dr. Keith Burridge (University of North Carolina, Chapel Hill). A pool of siRNA for E-cadherin (ON-TARGETplus SMARTpool) and two siRNAs against p190RhoGAP, p190A(1), and p190A(2) (siGENOME duplex D-004158-03 and -04, respectively) were purchased from Dharmacon Inc. (Chicago, IL), whereas siRNA for p120ctn was from Ambion Inc. (Austin, TX). We used a control siRNA as earlier described (30). BLM cells were transfected with expression vectors or siRNA (100 nM) using Lipofectamine reagent (Invitrogen) or X-tremeGENE (Roche Diagnostics), respectively, as reported previously (30). Transfectants were tested in the different assays 48 h post-transfection. When GFP-expressing transfectants were used, cells were analyzed by flow cytometry to assess transfection efficiency, which was consistently higher than 85% (data not shown). The different E-cadherin sublines generated upon pBAT-EM2 or pBAT-neo transfection into BLM cells were obtained by limit dilution, and transfectants were selected with G418 (Invitrogen). Wild type E-cadherin and the mutant 764AAA E-cadherin form in pLZRS vectors were gifts of Dr. Albert B. Reynolds. These vectors were cotransfected with pNGVL-VSV-G and pNGVL-gag-pol vectors in 293FT cells as reported previously (31). Conditioned media containing viral particles were used to infect BLM cells that were selected with G418. Expression of E-cadherin was assessed by flow cytometry.
Invasion and Adhesion Assays-Invasions were done as earlier established (29). Briefly, cells on the upper compartments of Matrigel-coated Transwells (Falcon) were allowed to invade for 22 h to the lower compartments filled with invasion medium with or without CXCL12. Invasive cells were fixed, stained, and counted under a microscope. For adhesion assays, cells were first labeled with 2Ј,7Ј-bis(carboxyethyl)-5(6Ј)-carboxyfluorescein-acetoxymethyl ester (Molecular Probes) and loaded (3 ϫ 10 4 ) in triplicate on 96-well dishes (Costar, Cambridge, MA) previously coated with Matrigel (1 g/ml) (Falcon), collagen IV (2 g/ml), or fibronectin (8 g/ml) (Sigma). Cells were allowed to adhere for 10 min at 37°C, followed by three washes with Dulbecco's modified Eagle's medium to remove unbound cells. Adhesion was quantified using a fluorescence analyzer (POLARstar Galaxy; Offenburg, Germany).
Time-lapse Microscopy--Slide chemotaxis chambers (Ibidi, Martinsried, Germany) were coated with Matrigel (0.1 mg/ml) in Dulbecco's modified Eagle's medium. 6 l of cell suspensions (2 ϫ 10 6 cells/ml) were applied into the slide channel and incubated for 1 h in a humid atmosphere. Following cell attachment, chamber reservoirs were filled with Dulbecco's modified Eagle's medium, and CXCL12 (300 ng/ml) was added in the upper reservoir according to the manufacturer's instructions. Immediately after, the chambers were mounted on the stage of an inverted Live Cell Imaging microscope (Leica AF6000 LX type DMI6000B), and cell movement was recorded using a 10ϫ objective of bright field (2 min each frame for 5 h) in temperature-and CO 2 -controlled chambers. Tracks from 30 cells were determined with Manual Tracking (ImageJ plug-in, National Institutes of Health), and the effect of CXCL12 to cell directionality was analyzed with Chemotaxis and Migration Tool (ImageJ plug-in, Ibidi). Cell directionality was assessed as the ratio of Euclidean mean distance (EMD) versus accumulated mean distance (AMD).
Immunoprecipitation, Western Blotting, and GTPase Assays-For immunoprecipitation, melanoma cells were lysed as earlier described (30), and extracts were incubated with antibodies followed by specific coupling to protein A-Sepharose beads (Amersham Biosciences). Proteins were eluted in Laemmli buffer, resolved by SDS-PAGE, and subjected to Western blotting with primary antibodies, followed by incubation with horseradish peroxidase-conjugated secondary antibodies and detection with SuperSignal chemiluminescent substrate (Pierce). GTPase activity assays were performed as reported previously (29). In brief, cells were detached with EDTA-containing PBS and incubated in suspension with or without CXCL12, and upon cell lysis, aliquots from extracts were kept for total lysate controls, and the remaining volume was incubated with GST-C21 (for RhoA) or GST-PAK-CD (for Rac1) fusion proteins (32) in the presence of glutathione-agarose beads. After elution of bound proteins, they were subjected to Western blotting with anti-RhoA or anti-Rac1 antibodies.
Confocal Microscopy and Flow Cytometry-For confocal studies, cells on fibronectin-coated coverslips were fixed with 4% paraformaldehyde (Sigma) in PBS and incubated with primary antibodies, and after incubation with fluorochrome-conjugated secondary antibodies, samples were mounted with Mowiol, and images were captured using a LEICA TCS-SP2-AOBS-UV confocal microscope with 63ϫ oil immersion objective. For labeling intracellular proteins, cells were permeabilized with 0.5% Triton X-100 in PBS before staining with primary antibodies or with Alexa 633-phalloidin (Molecular Probes). Displayed images were captured at the same sections in the different samples. For flow cytometry, cells were incubated with primary antibodies, followed by incubation with Alexa 488-conjugated secondary antibodies (Molecular Probes), and analysis in a Coulter Epics XL cytofluorometer. P3X63 was used as control primary antibody.
Statistical Analyses-Data were analyzed by one-way analysis of variance followed by Tukey-Kramer multiple comparisons. In both analyses, the minimum acceptable level of significance was p Ͻ 0.05.

Overexpression of E-cadherin in Melanoma Cells Leads to
Inhibition of Chemokine-promoted Invasion-Expression of CXCR4 on BLM melanoma cells confers strong invasiveness toward CXCL12 (29). As these cells express very low levels of E-cadherin ( Fig. 1A), we investigated if its overexpression could influence chemokine-promoted melanoma cell invasion. We transfected BLM cells with E-cadherin vectors and selected DC12 and DF11 sublines expressing high levels of E-cadherin compared with DC11 or mock sublines that expressed low amounts (Fig. 1A). Control experiments indicated that CXCR4 and ␤1 integrin expression was not significantly altered in the different E-cadherin transfectants (data not shown). Overexpression of E-cadherin on BLM cells correlated with enhanced ␤-catenin levels and E-cadherin/␤-catenin co-precipitation, as well as with co-localization of these proteins at cell-cell junctions ( Fig. 1, A, top and middle, and C). Likewise, E-cadherin co-localized with ␣-catenin ( Fig. 1C) at intercellular junctions. Similar levels of p120ctn isoforms of 105 and 115 kDa were detected in mock, DC12, and DF11 sublines (Fig. 1B, left), but high amounts of this catenin were found associated with E-cadherin on DC12 cells (Fig. 1B, right), indicating that p120ctn was recruited by E-cadherin in these transfectants. A low level of co-precipitated p120ctn was also observed in E-cadherin immunoprecipitates in mock cells, possibly from the small population of E-cadherin molecules expressed on their membranes. E-cadherin co-localized with p120ctn on DC12 transfectants ( Fig. 1C), whereas it displayed a predominant intracellular distribution in mock cells (data not shown). These results show that transfected E-cadherin correctly associates with their binding partners on the membrane of BLM melanoma cells.
Expression of E-cadherin on DC12 and DF11 BLM melanoma transfectants led to a significant inhibition of their CXCL12-promoted invasion across Matrigel and type I collagen, as compared with mock transfectant invasion ( Fig. 2A). Impaired invasiveness was directly dependent on E-cadherin, as invasion of DC12 cells was rescued when expression of this protein was silenced by RNA interference (Fig. 2B). Mock and DC12 cultures displayed similar morphologies when plated on Matrigel, both in the presence of CXCL12 (supplemental Fig.  S1) or in its absence (data not shown), suggesting that impairment in invasion of E-cadherin transfectants was not the result of differential cell aggregation. Examination of mechanisms underlying the defective invasion of E-cadherin melanoma transfectants using time-lapse microscopy performed on Matrigel in chemotaxis chamber slides revealed that mock transfectants had higher sustained migration directionality toward CXCL12 than DC12 cells, which displayed a predominant random migration. Thus, the cell directionality index based on the ratio EMD/AMD (see "Experimental Procedures") was near 3-fold higher in mock than in DC12 transfectants In addition, when we tested E-cadherin transfectants in adhesion assays to Matrigel layers, or to collagen IV or fibronectin, we observed a partial impairment in their attachment capability, which was mediated by ␤1 integrins (Fig. 2D). These data suggest that a defect in cell polarity toward CXCL12 associated with reduced adhesion shown by melanoma cells overexpressing E-cadherin represent likely mechanisms accounting for their impaired invasion in response to the chemokine.
Defective Invasion of E-cadherin Melanoma Transfectants Is Due to p190RhoGAP-mediated Impairment in CXCL12-promoted RhoA Activation-Tumor cell migration and invasion are closely linked to reorganization of the actin cytoskeleton, which is controlled by Rho GTPases (24). As CXCL12-stimulated melanoma cell invasion requires Rho GTPase activation (29), we first determined whether E-cadherin overexpression in melanoma cells affects activation of these GTPases. GTPase assays revealed that DC12 (Fig. 3A, left) and DF11 (data not shown) transfectants had defective RhoA activation in response to CXCL12 compared with mock cells. Conversely, E-cadherin transfectants displayed higher activation of Rac1 than mock counterparts (Fig. 3A, right). Moreover, DC12 and DF11 exhibited a lower degree of stress fiber and focal contact formation, as assessed by F-actin and vinculin staining, respectively (Fig. 3B). Importantly, when we expressed constitutively active forms of RhoA (RhoA CA) in DC12 and DF11 melanoma transfectants, invasion to CXCL12 was recovered to levels similar to those attained by mock cells expressing RhoA CA (Fig. 3C). These results indicate that defective melanoma cell invasion to CXCL12 upon E-cadherin expression is based on impaired RhoA activation.
Increased GAP activity on RhoA represents a potential mechanism underlying the diminished activation of this GTPase on E-cadherin melanoma transfectants. A main candidate GAP molecule involved in RhoA inactivation and stress fiber disassembly is p190RhoGAP, which preferentially inactivates RhoA (33,34). Activation of p190RhoGAP has been shown to be associated with its tyrosine phosphorylation (35)(36)(37), being Tyr-1105 the main phosphorylated residue (38). p190RhoGAP displayed higher a degree of tyrosine phosphorylation in DC12 and DF11 than in mock transfectants (mean 3.5-and 3.0-fold, respectively, n ϭ 3), which was associated with increased RhoA co-precipitation (Fig. 4A). When cells were exposed to CXCL12, we detected an increase in p190RhoGAP phosphorylation in mock cells, although it was difficult to observe it in DC12 and DF11 transfectants, possibly due to the fact that they already showed constitutively high phosphorylation. Moreover, transfection in DC12 cells of p190RhoGAP siRNA led to recovery of RhoA activation by CXCL12 to levels close to mock transfectants (Fig. 4B). Increase in p190RhoGAP tyrosine phosphorylation in DC12 and DF11 cells involved activity by Src family members, as it was inhibited by the Src inhibitor PP2 but not with the control inactive PP3 (Fig. 4C), in agreement with earlier results demonstrating the key role of Src proteins in p190RhoGAP phosphorylation (38 -41). Furthermore, p190RhoGAP-RhoA association was reduced in PP2-treated cells, indicating that Src-mediated phosphorylation of this GAP protein was required for its binding to RhoA. To investigate if increased GAP activity of p190RhoGAP on RhoA could underlie the impairment in DC12 and DF11 invasion, we knocked down p190RhoGAP and tested transfectants in invasion assays toward CXCL12. The results revealed that DC12 and DF11 invasion was recovered to levels similar to mock cells upon p190RhoGAP silencing ( Fig. 4D; supplemental Fig. S2), indicating that increased activation of this GAP protein in E-cadherin transfectants accounts for the reduction in invasion due to inhibition of RhoA activation.
Role of p120ctn in Chemokine-stimulated RhoA Activation and Invasion of Melanoma Cells-Previous works have demonstrated that p120ctn overexpression results in inhibition of RhoA activation, potentially by a RhoGDI activity of p120ctn (42,43), and direct interaction of this catenin with RhoA was reported both in vitro and in vivo (44,45). Furthermore, it has been recently shown that p120ctn and p190RhoGAP can associate in E-cadherin complexes in NIH-3T3 fibroblasts, leading to efficient inhibition of RhoA activation (46). To investigate if p120ctn is capable of regulating p190RhoGAP-dependent impairment in RhoA activation and invasion in melanoma cells, we first analyzed whether these two proteins can associate on E-cadherin melanoma transfectants. Co-precipitation experiments with melanoma cells transfected with control or p120ctn siRNA revealed that p120ctn and p190RhoGAP were able to associate in mock, DC12, and DF11 transfectants, indicating that association occurred independently of the levels of E-cadherin expression (Fig. 5A). In addition, CXCL12 did not alter p120ctn-p190RhoGAP association on melanoma transfectants (data not shown). p190RhoGAP tyrosine phosphorylation in DC12 and mock cells was not dependent on p120ctn, as it efficiently took place in p120ctn-silenced cells (Fig. 5B). Furthermore, p120ctn-p190RhoGAP association was observed in PP2treated mock and DC12 cells (data not shown). However, p190RhoGAP-RhoA association was dependent on p120ctn, as it was largely decreased in p120ctn knockdown DC12 cells (70 -80% reduction, n ϭ 3) and to a lesser extent in mock transfectants (30 -40% decrease). Whereas p120ctn displayed cell surface localization only on cells overexpressing E-cadherin, cellular distribution of p190RhoGAP was similar in mock and DC12 cells (Fig. 5C; see also Fig. 1C for p120ctn). Thus,

FIGURE 2. Overexpression of E-cadherin on melanoma transfectants leads to impaired invasion to the chemokine CXCL12.
A, melanoma transfectants were subjected to invasion assays toward CXCL12 using Matrigel or type I collagen gels. **, invasion was significantly inhibited, p Ͻ 0.01, or *, p Ͻ 0.05, with respect to mock transfectant invasion (n ϭ 5). B, DC12 cells were transfected with control or E-cadherin (E-cadh) siRNA, and E-cadherin expression was analyzed by immunoblotting (left), or transfectants subjected to invasion assays. *, invasion was significantly stimulated, p Ͻ 0.05. C, transfectants were subjected to time-lapse microscopy experiments on Matrigel using chemotaxis chamber slides. CXCL12 (300 ng/ml) was added on the top reservoir of the chambers, and migration of individual cells (n ϭ 30) was tracked and represented in m (x and y axis) using a common starting point in the middle of the graph. Shown is a representative result of three independent experiments. D, transfectants were tested in adhesion assays to Matrigel, collagen IV (Col IV), or fibronectin (FN), in the presence of control or anti-␤1 mAb, and without (medium) or with CXCL12. Ctr, control. ***, adhesion was significantly inhibited, p Ͻ 0.001, or *, p Ͻ 0.05 (n ϭ 4).
p190RhoGAP was found intracellularly, as well as on cell surface protrusions, where it closely co-localized with p120ctn on DC12 cells (Fig. 5C, top). E-cadherin, p120ctn and p190RhoGAP co-localized only at cell membrane protrusions (Fig. 5C, bottom), but not at cell-cell junctions (data not shown). Cell fractionation assays confirmed the membrane distribution of p120ctn on DC12 cells, and its preferential association with p190RhoGAP at this localization (Fig. 5D). Moreover, plasma membrane fractions displayed higher amounts of p190RhoGAP and RhoA in DC12 than mock transfectants, which correlated with increased association of these proteins on DC12 cell membranes. CD44 and RhoGDI were used as markers for plasma membrane and cytosolic fractions, respectively. Together these results indicate that p120ctn and p190RhoGAP can already associate intracellularly under low levels of E-cadherin expression, and that overexpression of E-cadherin promotes their association on the cell membrane, which favors p190RhoGAP-RhoA interaction, leading to inactivation of this GTPase.
To analyze if E-cadherin/p120ctn binding is required for the regulation of p190RhoGAP and RhoA, we expressed in BLM melanoma cells a mutant E-cadherin form (E-cadh-764AAA) that uncouples binding to p120ctn (47). Co-precipitation experiments indicated that E-cadh-764AAA expression in BLM cells led to a reduction in the extent of E-cadherin association with p120ctn to levels close to mock transfectants, as compared with E-cadherin wild type (E-Cadh-WT) counterparts (Fig. 6A). Control flow cytometry analyses revealed that E-cadh-WT and E-cadh-764AAA transfectants expressed similar levels of cell surface E-cadherin (data not shown). Associa- tion of p120ctn and p190RhoGAP took place independently of E-cadherin expression levels, in agreement with results obtained with DC12 cells (see Fig. 5A). Tyrosine phosphorylation of p190RhoGAP was not affected in E-cadh-764AAA transfectants (Fig. 6B), confirming that E-cadherin/p120ctn association is not essential for phosphorylation of the GAP protein. However, reduced levels of RhoA bound to p190RhoGAP were found in E-cadh-764AAA transfectants compared with E-cadh-WT cells, which was associated with a CXCL12-stimulated higher degree of RhoA activation in the mutant E-cadherin transfectants in relation with E-cadh-WT cells (Fig. 6, B  and C).
Finally, to determine the potential involvement of p120ctn in the impaired invasion toward CXCL12 shown by E-cadherin melanoma transfectants, we tested mock and DC12 cells transfected with control or p120ctn siRNA in Matrigel invasion assays. Silencing p120ctn in mock transfectants neither affected their invasion to CXCL12 nor their RhoA activation by the chemokine (Fig. 7, A and B). Notably, p120ctn-silenced DC12 cells showed a rescue of RhoA activation and a partial recovery of invasion. Associated with these results, we observed a decrease in E-cadherin cell surface expression (Fig. 7C). These data suggest that p120ctn regulate invasion in E-cadherin melanoma transfectants possibly by controlling p190RhoGAP-dependent RhoA activation.

DISCUSSION
Loss or reduction of E-cadherin expression is frequently associated with acquisition of an invasive phenotype in cancer.
Expression of E-cadherin in normal melanocytes is largely reduced at initial steps of melanoma progression (8 -10). However, different levels of E-cadherin are found at advanced stages of the disease (9,11,12), raising the possibility that recovery of E-cadherin expression could influence the invasive potential of melanoma cells.
Both in vitro and in vivo data indicate that the chemokine receptor CXCR4 plays crucial roles in the invasion and metastasis of melanoma cells (18,29). In this study we have used the A, melanoma transfectants were incubated for 20 min with or without CXCL12, and upon lysis, extracts were subjected to immunoprecipitation (IP) with anti-p190RhoGAP mAb, followed by Western blotting (WB) using the indicated antibodies. B, mock and DC12 cells were transfected with control or p190RhoGAP (p190) siRNA, and transfectant lysates were subjected to Western blotting with anti-p190RhoGAP antibodies (top) or to GTPase assays to detect RhoA activation in response to CXCL12 (bottom gel). Bottom panel represents densitometer analyses in arbitrary units showing RhoA activation levels. C, transfectants were incubated with PP2 or PP3 (2 h, 3 M) and subjected to immunoprecipitation with control or anti-p190RhoGAP mAb, and p190RhoGAP tyrosine phosphorylation and bound RhoA were analyzed by Western blotting. D, control or p190RhoGAP siRNA melanoma transfectants were tested in Matrigel invasion assays *, invasion was significantly stimulated, p Ͻ 0.05 (n ϭ 3).
highly invasive CXCR4 ϩ BLM melanoma cell line, which expresses minimal amounts of E-cadherin, to investigate whether chemokine-stimulated melanoma cell invasion is altered upon overexpression of E-cadherin. Here we show that E-cadherin melanoma transfectants have defective invasion toward CXCL12 chemotactic cues, as a result of impairment in cell migration directionality and decreased adhesion. Reduction in migration and adhesion was independent of potentially increased cell-cell contacts as a result of E-cadherin expression. Thus, time-lapse microscopy experiments indicated that defective migration directionality in E-cadherin transfectants was an intrinsic property of individual cells. Furthermore, there was not a general defect in cell response to CXCL12, as these trans-fectants were capable of up-regulating their Rac1 activation in response to the chemokine. Therefore, decreased cell polarity toward chemotactic stimuli led to random migration and reduced invasion of E-cadherin melanoma transfectants.
Mechanistic characterization of the impaired invasion revealed that a defective capacity of E-cadherin melanoma transfectants to activate RhoA in response to CXCL12, because of increased activity of p190RhoGAP on this GTPase, as well as a reduction in stress fiber formation, accounted for the inhibition in invasion. Thus, expression of a constitutively active form of RhoA or silencing p190RhoGAP expression led to rescue of E-cadherin melanoma transfectant invasion. These data suggest that chemokinestimulated RhoA activation contributes to the cell polarity response toward CXCL12 that results in directed migration and invasion, and the data indicate that p190RhoGAP plays a master role in controlling melanoma cell invasion.
Tyrosine phosphorylation of p190RhoGAP has been linked with an increase in its activation (35)(36)(37)41). Melanoma cells overexpressing E-cadherin exhibited higher levels of p190RhoGAP tyrosine phosphorylation than mock cells, which was dependent on the activity of Src proteins, in accordance with their reported involvement in the phosphorylation of the GAP protein (38 -41). The basis for the increased tyrosine phosphorylation most likely arises from E-cadherin engagement in cell-cell contacts that triggers Src activity at adherens junctions, as proposed previously (48). Src-dependent phosphorylation of p190RhoGAP was required for its association with RhoA, as it was decreased in PP2-treated melanoma cells. Using a constitutively active form of RhoA (63LRhoA), Noren et al. (48) reported that 63LRhoA co-precipitated with tyrosine-phosphorylated p190RhoGAP in E-cadherin Chinese hamster ovary transfectants. Furthermore, phosphorylation of p190RhoGAP at Tyr-1105 is required for actin stress fiber disassembly (49), a process associated with RhoA inactivation. Together with our results, these data stress the importance of p190RhoGAP tyrosine phosphorylation for its binding to and inactivation of RhoA.
p120ctn is a E-cadherin binding partner that is also involved in inhibition of RhoA activation by direct interaction with this FIGURE 5. p120ctn controls p190RhoGAP association with RhoA in melanoma cells. A and B, melanoma cells were transfected with control or p120ctn siRNA, and transfectants were subjected to immunoprecipitation with control or anti-p190RhoGAP mAb followed by immunoblotting with antibodies to p120ctn, p190RhoGAP, phosphotyrosine, or RhoA. C, top, mock and DC12 cells were transfected with p190RhoGAP-GFP and subsequently stained with anti-p120ctn antibodies, followed by analysis by confocal microscopy. Shown are two different representative cell morphologies as follows: cells with ruffle-like protrusions (left) or with lamellipodia (right). Bottom left, co-localization of p120ctn-GFP, E-cadherin (Alexa 647), and p190RhoGAP (Alexa 568) on DC12 cell membrane protrusions. Right, co-localization of p190RhoGAP-GFP and E-cadherin.
White arrows indicate areas of protein co-localization. D, cytoplasmic and membrane fractions from melanoma transfectants were subjected to immunoblotting (left) or to immunoprecipitation (IP) with control or anti-p190RhoGAP mAb followed by Western blotting with antibodies to p120ctn, p190RhoGAP, or RhoA (right).
GTPase (50). Overexpression of E-cadherin in melanoma cells led to a predominant p120ctn-E-cadherin association on the cell membrane, whereas p120ctn displayed a cytoplasmic localization in mock cells. Interestingly, we found that p120ctn was associated with p190RhoGAP both in mock and E-cadherinexpressing melanoma transfectants, but cell membrane co-localization and association of these proteins was mostly observed in transfectants overexpressing E-cadherin. Furthermore, we found higher amounts of p190RhoGAP, as well as RhoA, in plasma membrane fractions from E-cadherin transfectants than from mock cells, although confocal microscopy analyses revealed similar distribution of p190RhoGAP at cell lamellipodia in both transfectants. Notably, E-cadherin, p120ctn, and p190RhoGAP co-localized at cell membrane protrusions, together suggesting that E-cadherin promotes p120ctn/p190RhoGAP association on the cell membrane.
Silencing p120ctn expression did not affect CXCL12-stimulated mock cell invasion, indicating that a potential inhibition of RhoA activation by cytoplasmic p120ctn did not significantly influence this process. However, when p120ctn was knocked down in melanoma cells overexpressing E-cadherin, a recovery of chemokine-promoted RhoA activation linked to a large decrease in p190RhoGAP-RhoA association and a partial rescue of invasion was observed. As shown above, silencing p120ctn or p190RhoGAP expression leads to substantial recovery of RhoA activation in melanoma cells overexpressing E-cadherin, but only p190RhoGAP knockdown cells fully rescued the invasion. Our results suggest that RhoA inactivation in these cells is primarily due to p190RhoGAP, but p120ctn is required for efficient association between the GAP protein and RhoA.
On the other hand, E-cadherinmediated, Src-dependent tyrosine phosphorylation of p190RhoGAP in E-cadherin melanoma transfectants did not require p120ctn or p120ctn binding to E-cadherin, as it efficiently took place in cells expressing a p120ctn-uncoupled E-cadherin form (E-cadh-764AAA). Src-mediated tyrosine phosphorylation of p190RhoGAP requires the presence of p120RasGAP (51), a protein that directly interacts with the middle domain of p190RhoGAP (38,52); therefore, it is likely that efficient phosphorylation of p190RhoGAP in FIGURE 6. Uncoupling E-cadherin-p120ctn interaction impairs p190RhoGAP-RhoA association. BLM cells were transfected with empty vector (mock), wild type E-cadherin (E-Cadh-WT), or E-cadherin-764AAA (E-cadh-764AAA) and subjected to immunoprecipitation (IP) with anti-E-cadherin or anti-p190RhoGAP antibodies (A and B) followed by Western blotting to the indicated proteins or to RhoA GTPase assays (C). Numbers below the gel represent densitometer analyses in arbitrary units showing RhoA activation levels.  cells is contributed by p120RasGAP. Together these results suggest a model for chemokine-stimulated invasion of melanoma cells depending on the levels of E-cadherin expression. When this protein is absent or expressed at very low levels, GEF activation in response to chemokines overcomes GAP activity leading to Rho GTPase activation that favors invasion (30). It is noteworthy that this happens even in the presence of a preformed cytoplasmic p120ctn-p190RhoGAP-RhoA complex. When E-cadherin is expressed at high levels, tyrosine phosphorylation of p190RhoGAP is increased, enhancing its p120ctn-dependent association with RhoA on the cell membrane. This leads to RhoA inactivation, overtaking chemokinestimulated GEF activity and finally resulting in defective invasion toward the chemokine, which is based on impaired cell migration directionality. Our results indicate that E-cadherin-p120ctn interaction is important to stimulate or to stabilize p190RhoGAP-RhoA association, as it was significantly reduced in E-cadh-764AAA transfectants, resulting in a partial rescue of RhoA activation. It has been shown earlier that when p120ctn is bound to E-cadherin on the cell membrane, it cannot directly mediate inhibition of RhoA activation (42), and that p190RhoGAP can take this role in a Rac-and p120ctn-dependent manner in NIH-3T3 fibroblasts (46). Our data confirm the direct involvement of p190RhoGAP in p120ctn-regulated RhoA inactivation using a cancer cell model, demonstrating that this process can profoundly affect cell invasion toward chemokines.
Our results raise the possibility that subpopulations of E-cadherin-positive melanoma cells at advanced stages of the disease could have reduced metastatic potency in response to chemotactic stimuli compared with cells lacking E-cadherin, suggesting that regulation of its expression constitutes an important therapeutic pathway to limit melanoma dissemination. Indeed, in a skin reconstruction model, ectopic E-cadherin expression led to blockade of melanoma cell invasion into dermis (53). Activation of p190RhoGAP would represent a key step leading to inhibition of RhoA activation and decreased invasion. In this regard, a tumor suppressor role has been associated with p190RhoGAP (54,55). Further studies are needed to ascertain its involvement in melanoma cell metastasis.