Matrix metalloproteinase-9 silencing by RNA interference triggers the migratory-adhesive switch in Ewing's sarcoma cells.

Enhanced expression of (pro)matrix metalloproteinase-9 (MMP-9) is associated with human tumor invasion and/or metastasis. COH cells derived from a highly invasive and metastatic Ewing's sarcoma constitutively express proMMP-9. Transfection of a double stranded RNA that targets the MMP-9 mRNA into COH cells depleted the corresponding mRNA and protein as demonstrated by reverse transcriptase-PCR, enzyme-linked immunosorbent assay, and gelatin zymography. proMMP-9 extinction resulted in the following: (i) decreased spreading on extracellular matrix (fibronectin, laminin, collagen IV)-coated surfaces, (ii) inhibition of migration toward fibronectin, and (iii) induced aggregation, which was specifically disrupted by a function-blocking E-cadherin antibody. MMP-9 knockdown concomitantly resulted in increased levels of surface E-cadherin, redistribution at the plasma membrane of beta-catenin, and its physical association with E-cadherin. Moreover, induction of E-cadherin-mediated adhesion was associated with RhoA activation and changes in paxillin cytoskeleton. Finally, an inhibitor of gelatinolytic activity of pro-MMP9 did not reduce COH cell migration confirming that the enzymatic property of COH MMP-9 was not required for migration toward fibronectin. Overall, our observations define a novel critical role for proMMP-9 in providing a cellular switch between stationary and migratory cell phases.

Invasion and metastasis of tumor cells is a multiple process that depends on uncontrolled interactions between adjacent cells and/or cells and their extracellular environment (1,2). These interactions are mediated directly by specific adhesion receptors and indirectly by extracellular proteinases that mediate degradation of the extracellular matrix (ECM). 1 Many of the relevant proteinases belong to the matrix metalloproteinases (MMPs), which are a family of related zinc-containing proteinases that have the ability to degrade ECM (3). One member of the MMP family, MMP-9 (gelatinase B, 92 kDa) is capable of degrading type I, IV, V, VII, and XI collagens and laminin (1,4). Such proteolytic ability suggests that MMP-9 ultimately regulates cell migration, tumor growth, and angiogenesis (1,2,4). MMP-9 is overexpressed in many human solid and hematological malignancies (2,(5)(6)(7). MMP-9 promoter activity is induced coincidentally with invasion during tumor progression (8,9). Further, in vitro overexpression of MMP-9 confers a metastatic phenotype (10 -12). Conversely, selective suppression of MMP-9 by antisense gene transfer impairs in vitro cell migration of glioma cells (13,14) and osteoclast-like cells (15). Different families of adhesion receptors are likely to play a role in directing cell motility and include integrins, CAMs, and cadherins (16,17). Cadherins function by connecting cells to each other by homophilic interactions, in which they bind selectively to identical cadherin types. A cytoplasmic protein termed ␤-catenin interacts directly with the cadherin cytoplasmic domain and indirectly with the cytoskeleton via ␣-catenin, which interacts with actin and ␣-actinin (18). Signal transduction pathways from cadherin to RhoA, Rac1, and Cdc42 have been identified recently (19,20).
Compiled studies indicate a clear relationship between loss of cadherin expression and increased invasiveness in tumor cells (21)(22)(23). Moreover, various carcinoma are shown to exhibit an inverse relationship between MMP-9 and E-cadherin expression (24 -31). However, the molecular links between these divergent profiles of expression for MMP-9 and E-cadherin remain unknown.
Ewing's sarcoma (ES) is a malignant childhood bone and soft tissue tumor known to be highly aggressive and invasive (32). ES is characterized by a specific recurrent balanced chromosomal translocation t(11;22) (q24;q12) (33). Previous studies from our laboratory (34 -36) have demonstrated the inhibitory action of interferons on ES cell growth and MMP-9 expression. In the present investigation, we sought to determine whether MMP-9 extinction could affect ES cell behavior. Because recent reports (37)(38)(39) have demonstrated the utility of gene silencing by si-RNA in mammalian cells, we used a ds-RNA to interfere with the expression of MMP-9 gene. Our findings support the conclusion that MMP-9 constitutes a trigger for the switch between adhesive and migratory states of ES cells through ␤-catenin/RhoA/paxillin signaling pathways in a manner independent of its enzyme activity.
Cells-Human Ewing's sarcoma COH cells, wild-type p53 cells from a metastatic tumor localized on femur (41), were maintained in RPMI 1640 (Invitrogen) containing 10% heat-inactivated FCS (Myoclone Plus; Invitrogen) and 10 g/ml gentamycin in a humidified 37°C incubator (5% CO 2 ). COH cells were tested free of mycoplasma as assessed by RT-PCR (VenorGeM; Biovalley S. A.). In preliminary experiments, we controlled that trypsin treatment did not alter the levels of COH cell surface adhesion molecules (integrins, CAMs, and E-cadherin) as assessed by flow cytometry. Thus, cells were rapidly trypsinized and washed twice in PBS before adhesion and migration assays.
Enzyme-linked Immunoadsorbent Assays-The culture supernatants from COH cells were harvested under sterile conditions and frozen before MMP-9 and VEGF contents were determined using commercial enzyme-linked immunosorbent assay kits provided by R & D Systems. Controls included FCS-supplemented RPMI 1640 medium alone incubated under the same conditions.
RNAi Experiments-The si-RNA sequence targeting human MMP-9 chosen in this study (from mRNA sequence; GenBank TM accession number NM-004994) corresponds to the coding region 377-403 relative to the first nucleotide of the start codon (target ϭ 5Ј-AAC ATC ACC TAT TGG ATC CAA ACT AC-3Ј). Computer analysis using the software developed by Ambion Inc. confirmed this sequence to be a good target. si-RNAs were 21 nucleotides long with symmetric 2-nucleotide 3Ј overhangs composed of 2Ј-deoxythymidine to enhance nuclease resistance. The si-RNAs were synthesized chemically and high pressure liquid chromatography purified (Genset, Paris, France). Sense si-RNA sequence was 5Ј-CAUCACCUAUUGGAUCCAAdTdT-3Ј. Antisense si-RNA was 5Ј-UUGGAUCCAAUAGGUGAUGdTdT-3Ј. For annealing of si-RNAs, mixture of complementary single stranded RNAs (at equimolar concentration) was incubated in annealing buffer (20 mM Tris-HCl pH 7.5, 50 mM NaCl, and 10 mM MgCl 2 ) for 2 min at 95°C followed by a slow cooling to room temperature (at least 25°C) and then proceeded to storage temperature of 4°C. Before transfection, cells cultured at 50% confluence in 6-well plates (10 cm 2 ) were washed two times with OPTIMEM 1 (Invitrogen) without FCS and incubated in 1.5 ml of this medium without FCS for 1 h. Then, cells were transfected with MMP-9-RNA duplex formulated into Mirus TransIT-TKO transfection reagent (Mirus Corp., Interchim, France) according to the manufacturer's instructions. Unless otherwise described, transfection used 20 nM RNA duplex in 0.5 ml of transfection medium OPTIMEM 1 without FCS per 5 ϫ 10 5 cells for 6 h and then the medium volume was adjusted to 1.5 ml per well with RPMI 2% FCS. Silencer TM negative control 1 si-RNA (Ambion Inc.) was used as negative control under similar conditions (20 nM).
RT-PCR-RNA extraction from Ewing cells and subsequent cDNA synthesis were conducted as described previously (34). MMP-9 cDNA (296 bp) was amplified using the sense primer 5Ј-GGA GAC CTG AGA ACC AAT CTC-3Ј and the antisense primer 5Ј-TCC AAT AGG TGA TGT TGT CGT-3Ј according to published sequences (42). ␤ 1 -Integrin cDNA was amplified using the sense primer 5Ј-GTG AAT GGG AAC AAC GAG GTC-3Ј and the antisense 5Ј-ACA ATA AAA CGA TAA ACG GAC-3Ј according to published sequences (43). VEGF cDNA was amplified using the sense primer 5Ј-ACA TCT TCC AGG AG ACC CTG ATG AG-3Ј and the antisense 5Ј-GCA TTC ACA TTT GTT GTG CTG T-3Ј according to published sequences (44). 18 S ribosomal RNA was used as an internal control (QuantumRNA TM 18 S Internal Standard; Ambion Inc). The PCR products were visualized by electrophoresis in 1.6% agarose gel containing 0.2 g/ml ethidium bromide. The NIH Image 1.44 ␤11 software was used for the analysis.
Adhesion Assays-Twenty-four-well flat bottomed microtiter plates (Nunc) were coated overnight with fibronectin (Fn) or Laminin (Lm) or Collagen IV (Col) and blocked with BSA as described previously (40). Cells (4 ϫ 10 5 cells per well in 1 ml of RPMI without FCS) were allowed to adhere to each substrate-coated well at 37°C for 60 min. After PBS washing, adherent cells were trypsinized and quantitated using a cell Coulter Counter channelizer 256 (the diameters of living migrated cells ranging from 7 to 14 m). Results from triplicates (mean Ϯ S.D.) were expressed as relative cell adhesion (number of attached cells/total number of cells ϫ 100). Specific adhesion to substrates was determined by subtracting the nonspecific attachment of cells to BSA-coated surfaces from cell attachment to coated surfaces. Morphology of attached cells was assessed by staining with the Hemacolor kit from Merck and subsequent light microscope examination.
Migration Assays-The migration of COH cells was determined using the method described previously (45). After adjusting the cell density to 1 ϫ 10 6 cells/ml in 0.1% BSA-RPMI, 100,000 cells in 100 l were added to the top chamber of a 24-transwell apparatus (6.5-mm diameter, 8-m pore size; Costar 3422, Corning Inc., Corning, NY) in the absence or presence of fibronectin (25 g) in the lower chamber. After overnight incubation at 37°C in an atmosphere containing 5% CO 2 , living cells (with diameters ranging from 7 to 14 m) that passed through the membrane were collected from the lower well and counted in a cell Coulter Counter channelizer 256. Results from triplicate wells were expressed as mean Ϯ S.D.
Protein Analysis-Following cell (1 ϫ 10 7 cells/ml) transfection for the indicated times, cells were washed twice with cold PBS. Total cellular extracts, nuclear and cytoplasmic extracts were prepared as described previously (36). Equivalent amounts of protein were separated on an 8 or 10% ProSieve® 50 gel (BioWhittaker Molecular Applications), transferred to nitrocellulose (Schleicher & Schuell), and blotted as described previously (34). For co-immunoprecipitation experiments, cells were lysed in 400 l of lysis buffer containing 0.5% Brij35 (Sigma). Cell lysates were immunoprecipitated with mAbs prebound to protein G-Sepharose at 4°C overnight. Immune complexes were washed four times and then resolved on 8% ProSieve® gels as described previously (40). For E-cadherin immunocomplexes, the ␤-catenin was first revealed followed by the total E-cadherin hybridization, and conversely for ␤-catenin complexes, E-cadherin was first revealed before total ␤-catenin hybridization.
Fractionation Studies-Following cell transfection for the indicated times, cells were washed twice with cold PBS before resuspension in HEM buffer (2 mM MgCl 2 , 10 mM Hepes, pH 7.5, 2 mM EDTA, 0.2 mM EGTA) (46) containing protease inhibitor mixture and phosphatase inhibitor mixture II. Following ice incubation for 10 min, cells were lysed in a Dounce homogenizer. After centrifugation at 4°C for 5 min at 500 ϫ g to discard nucleus and subcellular organelles, the homogenates were centrifuged at 4°C for 30 min at 45,000 ϫ g. The cytosolic pool was collected, and the pellet (membrane-enriched fraction) was rapidly washed in HEM buffer and solubilized in Laemmli sample buffer (100 l). Protein concentrations were determined using DC TM protein assay (Bio-Rad). Proteins from cytosolic and membrane-enriched pools were separated on an 8% ProSieve® 50 gel and immunoblotted as described above.
Immunofluorescence Staining and Confocal Microscopy-Cells were washed twice with cold PBS and once with PBS containing 1% BSA before low speed cytocentrifugation to polylysine-treated slides (O. Kindler GmBH & Co., Freiburg, Germany). Cells were fixed in ice-cold methanol for 7 min and then washed twice with PBS. After 60 min of blocking in PBS containing 3% BSA at room temperature, cells were incubated in the same buffer with a polyclonal E-cadherin Ab (H108; 5 g/ml), a monoclonal ␤-catenin Ab (clone 2A4H7; 1 g/ml), or an isotype mIgG1 (1 g/ml) for overnight at 4°C. The preparations were washed three times for 15 min with PBS/0.05% Tween 20 and one wash with PBS, followed by an incubation for 90 min at room temperature with a fluorescein isothiocyanate-conjugated anti-mouse Ab or a TRITC-conjugated anti-rabbit Abboth diluted 1/1000 in PBS/BSA. Preparations were then washed three times with PBS/0.05% Tween 20 and once in PBS. Slides were mounted in Vectashield (47). Labeled samples were further analyzed by confocal microscopy on an Nikon microscope equipped with the Bio-Rad Laser-Sharp MRC-1024 confocal laser scanning software, using a Nikon Fluor ϫ100 oil-immersion objective and the 488-and 568-nm excitation wavelengths of the laser (47). The fluorescence of ␤-catenin and E-cadherin was analyzed concurrently in the same cell samples.

Efficient Extinction of MMP-9 Expression in COH Cells by RNAi Strategy-
We used an RNAi method to target MMP-9 in the ES COH cell line, which constitutively expresses high levels of MMP-9. The constructs we designed encoded an RNA that targets the MMP-9 mRNA (Fig. 1A). The 21-nucleotidelong target sequence had no homology with other members of the MMP family. The ds-RNA and Silencer TM negative control si-RNA (snc) were each tested for their ability to suppress MMP-9 specifically. We first assessed whether RNAi was dose-and time-dependent. COH cells were transfected with 1-20 nM of the ds-RNA for up 3 days. By RT-PCR analysis, a MMP-9dependent ds-RNA-mediated inhibition was observed in a doseand time-dependent manner (Fig. 1B). The snc-RNA (20 nM) was incapable of inhibiting MMP-9 gene expression ( Fig. 1B) even when transfected with a 10-fold-excess of the saturating ds-RNA concentration (200 nM) (data not shown). The timecourse assay performed with 20 nM ds-RNA-transfected COH cells showed that induced MMP-9 silencing could be maintained for at least 3 days (corresponding to seven generation times) (Fig. 1C). Importantly, snc-RNA and ds-RNA transfection had no effect on the mRNA levels of two unrelated genes i.e. VEGF and integrin ␤1 (Fig. 1B).
The RT-PCR results were confirmed by enzyme-linked immunoadsorbent assay. COH snc-RNA-transfected cells cultured up to 3 days spontaneously released high amounts of MMP-9 into the culture conditioned medium whereas ds-RNAtransfected cells showed a marked time-and dose-dependent inhibition in MMP-9 protein levels (Fig. 1D, left panel). In accordance with RT-PCR data, levels of released VEGF by COH cells were not affected by ds-RNA transfection (Fig. 1D, right panel). Zymography analysis of the conditioned media from COH cells before and after snc-RNA treatment indicated the presence of a gelatinase activity at 92 kDa (Fig. 1E, compare lanes 4 and 6) consistent with the pattern of recombinant proMMP-9 (Fig. 1E, lane 1). In contrast, MMP-9 was barely evident in conditioned medium from ds-RNA cells (Fig. 1E, compare lane 8 with lanes 4 and 6) corroborating the undetect- able MMP-9 transcript levels (Fig. 1C). Preincubation of COH supernatants with amino-phenyl mercuric acid, which activates the proform to the activated form, resulted in conversion of proMMP-9 to an active form of 82 kDa size (Fig. 1E, lanes 2, 5, and 7). Together, these findings indicate that RNAi efficiently and specifically inhibits endogenous MMP-9 gene expression in COH cells.
MMP-9 Silencing Induces E-cadherin-mediated Cell-Cell Adhesion-RNAi led neither to cell necrosis (data not shown) nor to cell apoptosis as measured by means of fluorescence flow cytometry ( Fig. 2A). Indeed, Apo 2.7 antigen, a mitochondrial membrane protein exposed on the surface of cells undergoing programmed cell death (48), was expressed on less than 10% of snc-RNA cells. The levels of Apo 2.7 on ds-RNA cells remained close to that of snc-RNA cells ( Fig. 2A). Although their proliferation rates were almost comparable (Fig. 2B), the morphologies of cultured COH ds-RNA cells appeared strikingly different from parental and snc-RNA cells (Fig. 2C). Parental COH cells cultured as adherent monolayers showed a fibroblasticlike phenotype (Fig. 2C, a) as was seen with snc-RNA cells (Fig.  2C, b). However, upon ds-RNA treatment for 48 h, typical cell aggregates were formed (Fig. 2C, c) in a dose-dependent manner (Fig. 2C, d-f). Kinetic studies indicated that aggregation was already observed within the first 24 h and was optimal at days 2-3 of ds-RNA treatment (data not shown).
Studies have linked cell-cell adhesion to CAMs and cadherins (17,22). As seen in Table I, parental COH cells expressed low to intermediate levels of ICAM-1, VCAM, E-cad-herin, and NCAM on the cell surface as assessed by flow cytometry. The levels of ICAM-1, VCAM, and NCAM remained unchanged upon snc-RNA and ds-RNA treatments for up 3 days (see Table I for day 2; data not shown for day 3). In contrast, after 2 to 3 days of transfection, aggregated ds-RNA cells consistently expressed E-cadherin more strongly (see Table I and Fig. 3A) than parental and snc-RNA cells, suggesting the importance of this adhesion molecule for cell-cell adhesion. Indeed, aggregation of ds-RNA cells was inhibited by a neutralizing antibody against E-cadherin (Fig. 3B). Together, these data indicate that MMP-9 silencing promotes cell-cell contacts through the contribution of E-cadherin. We also assessed whether MMP-9 silencing prevented E-cadherin inactivation involving the proteolytic cleavage of its extracellular domain by MMP-9. However, we demonstrated that MMP-9 inhibitor did not affect the profiles of surface expression of ICAM-1, VCAM, NCAM, and E-cadherin of MMP-9-silenced cells as compared with control and snc-RNA cells (Table II).
MMP-9 Silencing Impairs COH Cell Migration-In parallel, to analyze the impact of MMP-9 silencing on migration of COH cells, we used transmigration assays with Transwells (Costar) where the lower chamber was uncoated or coated with ECM proteins Fn, Lm, and Col. As shown in Fig. 4A, a significant percentage of parental COH cells spontaneously transmigrated, and Fn significantly increased their migration. The two other ECM proteins Col and Lm did not promote as efficient migration as Fn (data not shown). Importantly, COH cell migration was not inhibited by the MMP-9 inhibitor 2(R)-2-[(4biphenylsulfonyl)amino]-3-phenylpropionic acid (Fig. 4A). As expected, gelatinolytic activity of recombinant proMMP-9, as well as that present in the conditioned medium of COH cells, was down-regulated by this inhibitor (Fig. 4B) ascertaining its potency.
Migration toward Fn was significantly reduced for day 2 ds-RNA COH cells (50% inhibition) compared with day 2 snc-RNA cells or parental cells (Fig. 4A). Visualization of migrated cells indicated cell integrity for both snc-RNA and ds-RNA cells (Fig. 4C). A kinetic course similar to that of cell-cell contact formation was observed for inhibition of migration (maximal inhibition at day 2-3; data not shown for day 3). Altogether, these experiments support a positive role for proMMP-9 in reduced COH cell migration secondary to increased cell-cell adhesion.
MMP-9 Silencing Inhibits COH Cell Spreading to Fibronectin-, Laminin-, and Collagen IV-coated Surfaces-We further examined the impact of MMP-9 silencing on adhesion of COH cells to extracellular matrix components. Parental COH cells were shown to express intermediate to high levels of (␣ 1 Ϫ␣ 6 ) ␤ 1 integrins as determined by FACS analysis (Table I), and adhere to Fn, Lm, and Col but not to BSA (Fig. 4D). No detectable level of ␤ 2 and ␤ 3 subunits was observed (data not shown). We employed function-blocking Abs to identify the cell adhesion receptors involved in cell attachment. These experiments indicated that cell adhesion was mediated via ␣ 5 ␤ 1 for fibronectin, ␣ 2 ␤ 1 , and CD44 for laminin and ␣ 1 ␤ 1 and ␣ 2 ␤ 1 for Col, respectively (Fig. 4D). Day 2 transfection with snc-RNA or ds-RNA did not modify the rates of binding of COH cells to Fn, Lm, and Col (Fig. 4E). However, a significant difference in morphology of attached cells was observed. Adherence of snc-RNA cells, like parental cells (data not shown), was accompanied by an apparent spreading of many cells on each substrate-coated well (Fig.  4F, a, c, and e) whereas adherent ds-RNA transfected COH cells retained a rounded morphology (Fig. 4F, b, d, and f). These morphological changes were, however, independent of the levels of cell surface expression of snc-RNA and ds-RNA cell adhesion molecules involved in COH cell adhesion (Table I).

Effects of MMP-9 Silencing on Paxillin Phosphorylation and
RhoA Activity-Because adhesive interactions (between adjacent cells or cells and ECM) are critically influenced by the organization of the cytoskeleton, we further explored the effects of MMP-9 silencing on various cytoskeletal related proteins. COH snc-RNA cells, like parental cells, express different proteins of the cytoskeleton including actin, talin, vinculin, and paxillin (Fig. 5A). The total steady state protein levels of actin, actinin, vinculin, and talin were not modified upon 3 days of ds-RNA treatment (Fig. 5A). With regard to ␣-actinin, a timedependent increased signal in the Western blot was unexpectedly observed under both snc-and ds-RNA conditions (Fig. 5A).
Paxillin migrates in SDS-PAGE as two close bands between 66 and 70 kDa (49) and in lower molecular mass forms (44 -46 kDa) (50,51). COH snc-RNA cells exhibited high molecular mass forms of paxillin around 66 kDa (Fig. 5A). Total steady state protein levels of 66-kDa paxillin increased with time in COH cells transfected with ds-RNA compared with the snc-RNA cells, and this was associated with the clear appearance of 44 -46-kDa paxillin forms on day 2 (Fig. 5A). The 66-kDa form of paxillin contains two critical tyrosine phosphorylation sites at Tyr 31 and Tyr 118 (52). Immunoblot analysis with an Ab against phosphotyrosine residue 118 (pTyr 118 -paxillin) showed a significant decrease (3-fold) in phosphopaxillin levels in ds-RNA COH cells compared with snc-RNA cells (Fig. 5A). Thus, MMP-9 silencing in COH cells appears associated with a decrease in 66-kDa phosphopaxillin reactivity. We further examined two tyrosine kinases, FAK and PYK2, implicated previously (52) in the phosphorylation of paxillin. Activation of the kinase activities of FAK and PYK2 leads first to autophospho-      (52,53). Our data, however, indicated that FAK/PYK2 were not phosphorylated upon MMP-9 silencing thus excluding their involvement in decreased paxillin phosphorylation (data not shown).

Effect of an MMP-9 inhibitor on cell-cell adhesion molecules on COH cells in absence or presence of an MMP-9 inhibitor
Alternatively, because paxillin can be activated via a Rhoregulated pathway (51, 54), we further studied the effect of MMP-9 silencing on Rho-A protein in COH cells. RhoA activity is regulated by both translocation and GTP binding (19). No changes in total RhoA protein levels were observed in COH cellular extracts taken from snc-RNA or ds-RNA cells (Fig. 5A). However, immunoprecipitation with Rhotekin-RBD beads resulted in a marked increase in the amount of GTP-Rho complex (activated Rho) in ds-RNA cells (Fig. 5A). Together, these results indicate that MMP-9 silencing concomitantly alters two proteins, i.e. paxillin and RhoA, known to play a role in the rearrangement of the cytoskeleton.
MMP-9 Silencing Favors Complex Formation between E-cadherin and ␤-Catenin-We determined whether MMP-9 silencing affects the steady state levels of NCAM and E-cadherin. Western blotting showed that total protein levels of NCAM and E-cadherin were unchanged in COH ds-RNA cells compared with the parental and snc-RNA cells (Fig. 5B). Because the association of the E-cadherin/␤-catenin complex is necessary for cadherin-mediated cell-cell adhesion (22,55,56), we next examined the ability of E-cadherin to bind ␤-catenin. When the total cellular extracts from parental cells, snc-RNA, and ds-RNA transfected cells were subjected to immunoprecipitation with anti-E-cadherin and then analyzed for ␤-catenin and Ecadherin levels, a higher level of ␤-catenin was complexed with E-cadherin in ds-RNA cells than in parental and snc-RNA cells (Fig. 5C, left panel). This result was further confirmed by the reverse approach, that is, by assessing the level of E-cadherin associated with immunoprecipitated ␤-catenin (Fig. 5C, right  panel). These data demonstrate that upon MMP-9 silencing, E-cadherin and ␤-catenin physically associate. We further analyzed the localization of ␤-catenin before and after MMP-9 silencing. Both parental and snc-RNA-treated COH cells showed similar expression patterns of cytoplasmic and nuclear ␤-catenin whereas ds-RNA treatment resulted in a significant accumulation of ␤-catenin in the cytoplasm (Fig. 5D). We then evaluated in which subcellular compartment ␤-catenin accumulated. ␤-Catenin was detectable in the membranous pool of triplicate. Specificity of adhesion to Fn, Lm, and Col was assayed in the presence of CD44 Ab or ␣␤ 1 Abs (20 g/ml). When used at the same concentration, isotype-matched antibodies (controls) were not inhibitory. E, adhesion of day 2 snc-RNA (20 nM) and ds-RNA (20 nM) cells to Fn, Lm, Col, and BSA. Data represent the means of experiments performed in triplicate. F, light microscopy of adherent snc-RNA (a, c, and e) and ds-RNA (b, d, and f) COH cells onto Fn (a and b), Lm (c and d), and Col (e and f). May-Grü nwald staining was used. Magnification, ϫ160. Bar, 10 M.

FIG. 5. Expression of cytoskeletal-related proteins in COH cells before and after MMP-9 silencing.
A, equivalent amounts of cell extracts (40 g) obtained from COH cells treated with snc-RNA or ds-RNA (20 nM) for 1-3 days were loaded on 8% ProSieve® gels in reducing conditions. Immunoblotting and protein detection with appropriate Abs to paxillin (PXC-10), pTyr 118 -paxillin and RhoA (mIgG1, cytoskeleton), vinculin (hVIN-1), talin (8D4), ␣-actinin (EA53), and actin (AC-15) were performed as described under "Materials and Methods." The level of GTP-bound RhoA was measured in the total extracts (500 g of protein) by immunoprecipitation with beads coupled to the Rho-binding domain of the RhoA effector Rhotekin. Exposure times were Ͻ15 s for total RhoA and actin, 5 min for talin, vinculin, and ␣-actinin, and Ͼ30 min for GTP-RhoA. B, equivalent amounts of total extracts (40 g) obtained from COH parental cells or treated with snc-RNA or ds-RNA (20 nM) for 2 days were loaded on 8% ProSieve® and immunoblotted using E-cadherin (clone 34) and NCAM Abs. C, total extracts prepared were subjected to immunoprecipitation (IP) with anti-E-cadherin (clone 34) or ␤-catenin (clone 14) Abs, followed by immunoblotting with anti-E-cadherin (H-108) and ␤-catenin (clone 14) Abs. Nuclear (EN) and cytoplasmic (EC) extracts (D) or cytosolic and membranous pools obtained from COH parental cells or treated with snc-RNA or ds-RNA (20 nM) for 2 days (E) were fractioned as described under "Materials and Methods." After solubilization, equivalent amounts of proteins were loaded on 8% ProSieve® gels and immunoblotted using ␤-catenin (clone 14) and integrin ␤1 (clone DE9) Abs. parental and snc-RNA cells (Fig. 5E). A marked accumulation of membranous ␤-catenin was seen with ds-RNA transfected cells (Fig. 5E). The integrin ␤ 1 was used as control to normalize the levels of ␤-catenin (Fig. 5E). We next examined the cellular distribution of ␤-catenin and E-cadherin in snc-RNA and ds-RNA cells by indirect immunofluorescence analysis using specific Abs (Fig. 6). Diffuse intracellular ␤-catenin staining was observed in control snc-RNA cells that is consistent with absence of E-cadherin-mediated cell adhesion (Fig. 6, a). ␤-Catenin and E-cadherin were absent from cell junctions in control snc-RNA cells (Fig. 6, a and c). In contrast, MMP-9-silenced cells presented a typical pattern of ␤-catenin and E-cadherin staining at the level of cell-cell contacts (Fig. 6, b and d).
Together, these findings indicate that MMP-9 silencing favored localization of ␤-catenin/E-cadherin complex at the plasma membrane.

DISCUSSION
Like all MMPs, MMP-9 is synthesized as a preproenzyme, secreted as an inactive zymogen form (proMMP-9), and requires proteolytic removal of a portion of its N-terminal domain to become activated (1,4). Compiled studies have implicated MMP-9 in cell invasion in physiological and pathological processes by providing the basis for the mechanism of ECM degradation (1,4). Indeed, tissue inhibitors of MMPs in general inhibit tumor cell invasion, angiogenesis, and metastasis (57). However, discrepant data showed that the active form of MMP-9 (13,14), as well as its proenzyme form (58 -64), are associated with tumor invasion in vitro. Therefore, the actual mechanism of (pro)MMP-9-mediated invasion remains unclear. In this study, we used as a model of Ewing's sarcoma, metastatic COH cells that constitutively express high levels of 92-kDa proMMP-9. Using an RNA silencing approach, we have found that MMP-9 in its proenzyme state provides a critical link between cell motility and cell-cell adhesion.
We first showed that MMP-9 silencing affects motile properties of COH cells. ProMMP-9 silencing by RNAi significantly decreases spreading of COH cells on Fn, Lm, and Col and strongly reduces their migration toward Fn. Although other authors showed recently (65) that proMMP-9 binding to gelatin or collagen type IV resulted in an enzymatic activation of proMMP-9 without loss of its NH 2 -terminal propeptide, the scenario is unlikely in our system, because cell migration is not inhibited by an MMP-9 inhibitor. Therefore, our results strongly suggest that MMP-9-silenced transfectants are less motile than their parental cells because of the loss of proMMP-9 protein devoid of enzymatic activity.
COH cells cultured as monolayers express a fibroblastic morphology characteristic of metastatic cells. Our data demonstrate that, concomitantly to the loss of migration, proMMP-9 silencing in COH cells favored a morphology different from that of the parental cells and more specifically cell aggregates. This was because of up-regulation of E-cadherin levels at the cell surface, because an Ab against E-cadherin efficiently blocked cell-cell adhesion. Malignant transformation is often characterized by disruption of cell-cell adhesion, which can be achieved by down-regulating the expression of cadherin and/or ␤-catenin (22). Moreover, clustering of cell surface cadherins is associated with the recruitment of ␤-catenin to the cadherin-associated complex (17). Although the levels of total E-cadherin and ␤-catenin remained unchanged upon ds-RNA treatment, we demonstrated an enhanced physical association between E-cadherin/␤-catenin proteins by coimmunoprecipitation. As expected, ␤-catenin was detected at the plasma membrane at cell-cell junctions of MMP-9-silenced cells. Together, our data indicate that surface stabilization of E-cadherin/␤-catenin complex in MMP-9-silenced COH cells leads to the induction of E-cadherin-mediated cell-cell adhesion, which, in turn, diminishes the motility. Our data are in agreement with the literature documenting the involvement of the cadherin/␤-catenin pathway in tumorigenesis (23,66). Disruption of adhesion systems can contribute to tumor development (22,23) whereas increased levels of cadherins reduce tumorigenic properties (67).
In addition, biological and synthetic inhibitors of MMP activities have been suggested to promote cell-cell adhesion by at least preventing the cadherin ectodomain cleavage, thus stabilizing cadherin-mediated cell-cell contacts (68,69). MMP-3 can cleave E-cadherin (67) whereas MMP-9 cleaves ICAM-1 (70). Similarly, an MMP activity seems to be responsible for the shedding of VCAM-1 (71). Here, we demonstrate that an inhibitor of MMP-9 affected neither COH cell adhesion and migration nor the profiles of surface expression of ICAM-1, VCAM, NCAM, and E-cadherin. These results therefore provide evidence of the following. (i) The integrities of CAMs and Ecadherin on COH cells are independent of proMMP-9. (ii) The impact of proMMP-9 silencing on E-cadherin-mediated cell-cell adhesion is not related to decreased proteolysis of surface E-cadherin.
In MMP-9-silenced COH cells, alterations of spreading, migration, and E-cadherin-mediated cell-cell adhesion were indicative of changes in the functional association of cell adhesion molecules (integrins and E-cadherin) with the cytoskeletal network. The proteins of the Rho subfamily Cdc42, Rac, and Rho are signaling molecules primarily involved in remodeling the actin cytoskeleton (17,20). In particular, Rho proteins are required for cadherin-mediated cell-cell adhesion and consequently influence many aspects of cell shape and movement (17,19,20). More specifically, ␤-catenin, depending on its relative amounts in the cytoplasm/nucleus or associated with cadherins at the plasma membrane, can influence cells toward migration or cell-cell adhesion by regulating the function of Rho proteins (20). Previous studies (20,72) reported that RhoA activation increases the accumulation of cadherin/␤-catenin complexes at cell-cell contact in keratinocytes and epithelial cells. We similarly found a correlation among RhoA activation, preferential accumulation of cadherin/␤-catenin complexes at cell-cell junctions, and E-cadherin-mediated adhesion of MMP-9-silenced COH cells.
Among cytoskeletal proteins that play a critical role in cell spreading and migration, paxillin localizes primarily to sites of cell adhesion to the ECM called focal adhesions (51,52). Paxillin is a 66-kDa cytoplasmic protein that binds to several cytoskeletal, intracellular signaling, and adaptor molecules such as vinculin, pp 125 FAK, ILK, the cytoplasmic domain of integrin ␣4, and modulators/effectors of Rho proteins (51,73). In MMP-9-silenced cells, reduction of cell spreading to ECM is accompanied by an increase in paxillin steady state protein levels but a decrease in paxillin tyrosine phosphorylation. Our data are consistent with prior studies (52), which demonstrate in various cell types that paxillin phosphorylation is an early and requisite step in cell spreading.
Cell shape changes result in modulation of various phenotypic changes and alteration of MMP-9 expression (55,74). Inversely, our data indicate that proMMP-9 extinction induces cell shape changes (particularly those resulting from paxillin/ RhoA/␤-catenin pathways) that alter the adhesive and migratory properties of COH cells. Therefore, an unanswered question regarding the critical role of proMMP-9 for E-cadherinmediated cell-cell adhesion and migration is how can proMMP-9 influence motile properties of COH cells? A number of recent papers have demonstrated that the hyaluronan receptor CD44 can serve as a docking molecule to retain secreted MMP-9 (under both pro and active forms) at the cell surface (75,76). CD44 is involved in cell adhesion and trafficking, as well as in tumor growth, invasion, and metastasis (76 -78). The cytoplasmic domain of CD44 interacts with several cytoskeletal proteins including several guanine nucleotide exchange factors for Rho family GTPases. The interaction of CD44 with these guanine nucleotide exchange factors leads to activation of Rac1 and, under certain conditions, results in increased Rho activation (76 -78). Parental COH cells constitutively express CD44, and MMP-9 silencing does not affect CD44 expression (Table I). Thus, one can hypothesize that extracellular proMMP-9 influences cytoskeleton rearrangements (involving RhoA/paxillin/␤catenin signalings) through its association with surface CD44.
Alternatively, a yet unidentified signaling pathway in response to MMP-9 may lead to nuclear localization of ␤-catenin, which, via members of the LEF/TCF family of DNA-binding proteins, transactivates certain genes culminating in cell-cell adhesion. More specifically, association of nuclear ␤-catenin with TCF-4 promotes the expression of several molecules that have important roles in the development and progression of tumors (18,79,80). Indeed, the transcriptional activity of MMP-7, MMP-26, and MT1-MMP is regulated by TCF-4 (79,81,82), and other MMPs including MMP-9 also possess a consensus TCF-4 regulatory element (79). Our results show that COH parental cells that express low levels of surface E-cadherin have high levels of nuclear ␤-catenin while inversely increase of surface E-cadherin in MMP-9-silenced cells is associated with decreased nuclear localization of ␤-catenin. Whether MMP-9 silencing results in the withdraw of nuclear ␤-catenin via the coordinating signals transduced through ␤-catenin/TCF-4 pathways has to be considered.
In conclusion, our observations highlight a novel aspect of the role of MMP-9, independently of its enzymatic activity, for influencing tumor cell fate. Loss of motility and concomitant induction of E-cadherin-mediated cell-cell adhesion of MMP-9silenced COH cells are associated with stabilization of E-cadherin/␤-catenin complex and reorganization of the cytoskeleton through paxillin/RhoA pathways. Our data suggest a model (Fig. 7) in which proMMP-9 can be critical for progression of malignancy by triggering the switch from a stationary state to a migratory state.