Lysophosphatidic Acid Initiates Epithelial to Mesenchymal Transition and Induces β-Catenin-mediated Transcription in Epithelial Ovarian Carcinoma*

Background: Aberrant activation of Wnt/β-catenin signaling is found in ovarian cancer, even when mutations are absent. Results: The bioactive lipid lysophosphatidic acid induces β1-integrin clustering, nuclear β-catenin translocation, and activation of Tcf/Lef-regulated gene expression. Conclusion: Microenvironmental lipids modulate β-catenin signaling. Significance: Understanding the mechanisms of Wnt/β-catenin activation provides a basis for therapeutic targeting of the pathway in ovarian cancer. During tumor progression, epithelial ovarian cancer (EOC) cells undergo epithelial-to-mesenchymal transition (EMT), which influences metastatic success. Mutation-dependent activation of Wnt/β-catenin signaling has been implicated in gain of mesenchymal phenotype and loss of differentiation in several solid tumors; however, similar mutations are rare in most EOC histotypes. Nevertheless, evidence for activated Wnt/β-catenin signaling in EOC has been reported, and immunohistochemical analysis of human EOC tumors demonstrates nuclear staining in all histotypes. This study addresses the hypothesis that the bioactive lipid lysophosphatidic acid (LPA), prevalent in the EOC microenvironment, functions to regulate EMT in EOC. Our results demonstrate that LPA induces loss of junctional β-catenin, stimulates clustering of β1 integrins, and enhances the conformationally active population of surface β1 integrins. Furthermore, LPA treatment initiates nuclear translocation of β-catenin and transcriptional activation of Wnt/β-catenin target genes resulting in gain of mesenchymal marker expression. Together these data suggest that LPA initiates EMT in ovarian tumors through β1-integrin-dependent activation of Wnt/β-catenin signaling, providing a novel mechanism for mutation-independent activation of this pathway in EOC progression.

(EMT), and it is proposed that this phenotypic plasticity contributes to decreased response to conventional therapeutics (17)(18)(19)(20). Furthermore, recent work has demonstrated a role for the mesenchymal phenotype in mesothelial cell invasion at metastatic sites in the peritoneal cavity (21). Understanding the mechanism(s) regulating this phenotypic plasticity may enable reversal of acquired resistance to chemotherapeutic agents.
Acquisition of mesenchymal phenotype via EMT can be initiated by extracellular matrix context, disruption of adhesion, and soluble signaling factors (22,23). Of these factors, lysophosphatidic acid (LPA) has emerged as a potentially important mediator of EMT in ovarian carcinoma due in part to its high concentration in the ovarian cancer microenvironment (up to 80 M) (24,25). LPA modulates mitogenic activity, motility, escape from anoikis, and survival effects through a class of heterotrimeric G-protein-coupled receptors, LPA 1 to LPA 5 (26,27). One mechanism by which LPA modulates motility and facilitates the gain of a mesenchymal phenotype in EOC is disruption of E-cadherin-based cell-cell adhesions (28 -31). Of particular interest is the fate of adherens junction-associated and/or cytoplasmic ␤-catenin following disruption of cell-cell adhesions. In intact epithelial tissues, ␤-catenin displays junctional localization and is associated with the cytoplasmic domain of E-cadherin. Junction disruption can induce pathways that subsequently target ␤-catenin for proteosome-mediated degradation or for nuclear translocation and transcriptional regulation (32). As previous work has demonstrated the intersection of cytoplasmic and nuclear ␤-catenin pools (33), this study investigated the consequences of LPA-induced disruption of E-cadherin-mediated cell junctions on subcellular ␤-catenin localization and Tcf/Lef/␤-catenin transcriptional activity. These results provide additional support for the role of ligand-independent ␤-catenin activity in serous epithelial ovarian carcinomas.

Experimental Procedures
Cell Culture-OVCA429 and OVCA433 cell lines were generously provided by Dr. Robert Bast, Jr. (M.D. Anderson Cancer Center, Houston, TX), and were maintained in MEM (Gibco Invitrogen), 10% fetal bovine serum (Gibco Invitrogen), penicillin/streptomycin (Gibco Invitrogen), amphotericin B (Cellgro by Mediatech), nonessential amino acids (Cellgro by Mediatech, Herndon, VA), and sodium pyruvate (Cellgro by Mediatech) at 37°C in 5% CO 2 . Multicellular aggregates were formed in 96-well plates coated with 50 l of 0.5% agarose in serum-free media by seeding 5000 cells per well in serum-free media and incubating overnight at 37°C in 5% CO 2 . Aggregate formation was confirmed by light microscopic visualization.
Lysophosphatidic Acid-LPA was purchased from Cayman Chemical (Ann Arbor, MI) and was prepared for use by dehydrating the lyophilized lipid under a tissue culture hood, on ice, overnight. LPA was resuspended in 1% BSA in PBS at a final concentration of 2 mM, allowed to dissolve on a rotator at 4°C overnight, and then aliquoted. Aliquots in use were stored at Ϫ20°C and at Ϫ80°C for long term storage. Where indicated, cells were pretreated for 15 min prior to addition of LPA with the LPA receptor inhibitor, Ki16425, or the Rho signaling inhibitor, Y27632, both of which were purchased from Cayman Chemical (Ann Arbor, MI).
Immunofluorescence-Cells were plated on 22-mm 2 glass coverslips coated by passive adsorption with type I collagen (from rat tail, BD Biosciences) and placed in 6-well tissue culture plates and were treated with LPA at the concentrations and times described in the figure legends. Following treatment, cells were gently washed in phosphate-buffered saline (PBS), fixed in paraformaldehyde (4%), and permeabilized with 0.3% Triton X-100 at room temperature, washed in PBS, and blocked in PBS, 1% bovine serum albumin (BSA), or PBS, 2.5% goat serum, followed by the addition of primary antibody diluted in PBS, 1% BSA, or PBS, 2.5% goat serum at 37°C. After two washes in PBS, coverslips were incubated with Alexa Fluor 488-or Alexa Fluor 594-conjugated goat anti-mouse or goat anti-rabbit IgG (1:500 dilution) in the dark at room temperature. Coverslips were washed twice in PBS and once in distilled water, fixed using gelvatol, and visualized using fluorescence microscopy (Nikon Microphot FXA or Leica Microsystems DM5500B). Semi-quantitative analysis of surface-associated (junctional) ␤-catenin staining was determined by counting a minimum of 12 fields per treatment and scoring as positive the number of cells with two remaining fluorescent cell-cell borders (30).
Immunohistochemistry-Immunohistochemical analysis was performed on a human tumor tissue microarray prepared at the Robert H. Lurie Comprehensive Cancer Center at Northwestern University with Institutional Review Board approval. Tumor specimens were cut 3-5 m thick (1 mm in diameter) and deparaffinized. Antigen retrieval was accomplished by heat induction at 99°C in an antigen retrieval solution (10 mM Tris, 1 mM EDTA, pH 9.0) for ϳ1 h. Immunohistochemical staining with anti-␤-catenin (BD Transduction Laboratories; 1:50) was performed according to standard procedures and as described previously (31). A total of 105 tissues of varying histotypes were  evaluated (41 serous, 26 endometrioid, 3 malignant mixed Mullerian tumors, 6 mucinous, clear cell, 19 borderline, and 1 untyped). A determination of nuclear ␤-catenin positivity or negativity was determined based on the presence of ␤-catenin staining coinciding with hematoxylin-stained nuclei (34).
Immunoprecipitation-The ␤-catenin/Tcf co-immunoprecipitation protocol was adopted from chromatin immunoprecipitation protocols described previously (36,37). Cells were grown to 80% confluence and then serum-starved overnight. Appropriate samples were pretreated with 40 M Ki16425 (LPA receptor inhibitor, Cayman Chemical, Ann Arbor, MI) or DMSO control, followed by addition of 40 M LPA for 2 h. Nontreated cells, cells treated with Ki16425 alone, and cells treated with 40 M LiCl served as controls. Cells were collected in a hypotonic lysis buffer (20 mM HEPES, pH 7.9, 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA), incubated on ice for 20 min, and then centrifuged at 13,000 rpm for 10 min at 4°C. Nuclei isolated from the previous step were disrupted by resuspension in a "breaking" buffer (50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl, 1% SDS, 2% Triton X-100) and then passed through a 26-gauge syringe eight times. The suspension was centrifuged at 13,000 rpm for 10 min at 4°C, diluted in 1 ml of Triton buffer (50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl, 0.1% Triton X-100), and cleared of nonspecific binding by incubation with 20 l of protein A/G beads (Sigma) at 4°C overnight. After centrifugation and collection of supernatant, protein concentration was measured using a kit (Bio-Rad). Five hundred micrograms of total protein were added to 1 ml of Triton buffer, 20 l of protein A/G beads, and 5 g of anti-␤catenin antibody and then incubated on a rotator overnight at 4°C. Beads were washed five times in Triton buffer and then resuspended in 2ϫ sample buffer. Samples were boiled and then analyzed for Tcf expression by Western blot as described below. Antibody was removed from the PVDF membrane by washing in stripping buffer (50 mM Tris, pH 6.8, 1% SDS, 150 mM NaCl, 100 mM ␤-mercaptoethanol, 0.02% sodium azide) and then re-probed for ␤-catenin as an assay control. Experiments were repeated in triplicate. Western blots were quantified using MultiGauge version 2 (FUJIFILM, Tokyo, Japan).
Flow Cytometry Analysis-Cells were serum-starved overnight, followed by pretreatment of appropriate samples with the LPA receptor inhibitor, Ki16425 (Cayman Chemical, Ann Arbor, MI), for 30 min at 37°C and LPA treatment for 2 h at 37°C. Cells were trypsinized, resuspended in serum-free medium, and then incubated with anti-active ␤1 integrin (HUTS21 clone, BD Transduction Laboratories) for 1 h. Following three washings in PBS, cells were incubated and protected from light with anti-mouse IgG-Alexa Fluor 488 for 30 min at ambient temperature. Excess antibody was removed by washing, and cells were resuspended in PBS and then analyzed with the CyAn ADP analyzer (Beckman Coulter, Brea, CA). All assays were performed three times.
␤1 Integrin Cross-linking-To evaluate membrane localization of ␤1 integrins as individual heterodimers or as clusters, a cross-linking mechanism was utilized to capture integrin distribution following LPA treatment. Analysis of ␤1 integrin clustering was performed as described previously (49). Cells were serum-starved overnight, trypsinized, and resuspended in fresh serum-free medium in 1.5-ml Eppendorf tubes. Suspensions were incubated with 40 M LPA for 40 min at room temperature. Control cells were incubated with 1% BSA in PBS (LPA vehicle). To cross-link surface-expressed ␤1 integrins, anti-␤1 integrin (MAB1959) was added for a 40-min incubation on ice. Cells were washed, resuspended in serum-free medium, and then incubated with polyclonal anti-mouse IgG (Chemicon) at 37°C for 30 min. Cell suspensions were affixed to glass 22-mm 2 coverslips by cytology centrifugation using a Cytopro 7620 centrifuge (Wescor, Logan, UT) with acceleration at 2000 rpm for 10 min. Clustering was analyzed by immunofluorescent staining (primary antibody, anti-␤1 integrin (MAB2250); secondary antibody, anti-mouse IgG-Alexa Fluor 488), as described below. Fluorescent data were quantified by counting the number of cells with punctate staining patterns indicating clusters, compared with the total number of cells in a given field (Nikon Microphot FXA, ϫ40 magnification). A minimum of 10 high powered fields were analyzed per condition, and the experiment was repeated in triplicate.
Western Blotting-Samples to be analyzed by Western blotting were loaded onto SDS-polyacrylamide gels, electrophoresed, and then transferred onto polyvinylidene fluoride (PVDF) microporous membranes (Millipore). After blocking nonspecific binding to membranes in 3% BSA in TBST for 1 h at room temperature, membranes were incubated with primary antibodies for 3 h at room temperature or overnight at 4°C, and then with HRP-conjugated secondary antibodies. Immunoreactivity was determined by SuperSignal West Dura Extended Duration substrate kit (Fisher). Western blots were quantified using MultiGauge version 2 (FUJIFILM, Tokyo, Japan). PVDF membranes were stripped of antibody by washing in a stripping buffer (50 mM Tris, pH 6.8, 1% SDS, 150 mM NaCl, 100 mM ␤-mercaptoethanol, 0.02% sodium azide) and then re-probed for appropriate assay control(s).
Tcf Reporter Assay-TOPFlash (TCF reporter Plasmid) and FOPFlash (TCF mutant reporter plasmid) were generously provided by Dr. Hans Clevers (Hubrecht Laboratory and Utrecht University, Utrecht, The Netherlands). The Renilla luciferase vector, pRL-CMV, was purchased from Promega (Madison, WI). OVCA433 cells were plated at 40 -50% confluence in 6-well plates and transiently co-transfected with a Renilla luciferase reporter construct (pRL-CMV) and either the firefly luciferase TOPFlash TCF reporter plasmid or the FOPFlash TCF mutant reporter plasmid using FuGENE 6 transfection reagent according to the manufacturer's instructions (Roche Diagnostics, Mannheim, Germany). Approximately 18 h after transfection, cells were cultured in low calcium (0.1 mM CaCl 2 ), serumcontaining MEM (S-MEM (Invitrogen) for 1 h before the addition of LPA (40 M) for 2, 4, or 30 h. Cells were then lysed in passive lysis buffer (Promega). Both Renilla and firefly luciferase readings were taken on a Veritas Microplate Luminometer (Turner Biosystems, Sunnyvale, CA) using the reagents and protocol provided in the Dual-Luciferase reporter assay system (Promega). Firefly luciferase readings were first normalized to the reading for the corresponding Renilla luciferase reading to account for transfection efficiency. The adjusted TOPFlash reading was then normalized to the corresponding adjusted FOPFlash reading to account for background reading of the TOPFlash construct.
Quantitative Real Time PCR-Total RNA was extracted from treated and control cells using TRIzol reagent (Life Technologies, Inc.) according to manufacturer's instructions. cDNA was synthesized from 1 g of RNA using the RT 2 First Strand cDNA synthesis kit (SA Biosciences, Frederick, MD). Amplification was performed using iCycler (Bio-Rad) for 40 cycles, with each cycle consisting of 15 s denaturation at 95.0°C followed by 1 min of annealing at 60.0°C. SYBR Green Master Mix and primer sets for RAC1, VIM, LRP6, and WNT5A were purchased from SA Biosciences. GAPDH was used as an internal control in each reaction.
Statistics-p values were determined using the t test function (two sample, unequal variance, one-tailed distribution) using Excel (Microsoft Corp., Redmond, WA).

LPA Initiates EMT and Loss of Junctionally Localized
␤-Catenin in Ovarian Carcinoma-Previous work by our laboratory and others has demonstrated a potential link between LPA and EMT via LPA-induced activation of RhoA, an initial target in the EMT transduction pathway (38), and via regula-tion of cell-cell adhesion in EOC (28 -31, 39, 40). Ovarian cancers metastasize as both single cells and multicellular aggregates, and it has been proposed that E-cadherin stabilizes aggregate formation and contributes to chemotherapy resistance (12,16,21). To assess the potential mechanistic link between LPA and EMT, multicellular aggregates generated from EOC cell lines were treated with LPA and evaluated for changes in mesenchymal (vimentin) and epithelial (E-cadherin) marker expression. E-cadherin cell surface expression was down-regulated in multicellular aggregates (Fig. 1, A-D, green), whereas vimentin expression increased in response to LPA treatment (Fig. 1, A-D, red). Further confirmation of gain of mesenchymal phenotype was demonstrated by evaluation of F-actin stress fiber arrangement and vinculin expression. Reorganization of F-actin stress fibers (Fig. 1, E-H, red) and punctate vinculin staining (Fig. 1, E-H, green) were observed, indicating gain of mesenchymal phenotype in response to LPA treatment.
We have previously demonstrated that LPA disrupts E-cadherin junctional integrity in ovarian cancer monolayer cultures (30). To validate LPA-induced adherens junction disruption in multicellular aggregate cultures and to evaluate whether adhe- rens junction disruption alters the subcellular localization of ␤-catenin, multicellular aggregates of OVCA 433 cells were treated with LPA (40 M LPA, 2 h) and evaluated for E-cadherin and ␤-catenin expression by immunofluorescence microscopy. Junctionally localized ␤-catenin is significantly decreased following LPA treatment compared with control ( Fig. 2, C, D, and K) and corresponds temporally to loss of surface-associated E-cadherin (Fig. 2, A and B). Furthermore, ␤-catenin perinuclear and nuclear accumulation can be observed in response to LPA (Fig. 2D). To determine whether LPA-mediated loss of ␤-catenin junctional localization is an LPA receptor-dependent event, cells were either treated with LPA (40 M) alone, pretreated with the pharmacologic LPA receptor inhibitor Ki16425 (40 M, 30 min), or pretreated with inhibitor and then treated with LPA. LPA treatment led to loss of ␤-catenin junctional localization with nuclear/perinuclear accumulation of ␤-catenin (Fig. 2, G and L) compared with untreated control (Fig. 2E), vehicle control (Fig. 2F), and Ki16425 alone (Fig. 2H). This loss of junctional localization was rescued by pretreatment with Ki16425 prior to LPA treatment (Fig. 2, I and L). In positive controls, loss of junctional ␤-catenin was also observed following treatment with lithium chloride (LiCl, GSK3-␤ inhibitor; Lysophosphatidic Acid Induces ␤1 Integrin Activation and Clustering-We have demonstrated that ligand-induced ␤1 integrin clustering activates pathways leading to enhanced nuclear translocation of ␤-catenin and activation of ␤-catenin target genes in ovarian cancer (24). Furthermore, the LPA/LPA receptor interaction has been previously shown to trans-activate a number of surface-expressed signaling receptors (27). To assess the potential for cross-talk between LPA signaling and integrin activation, OVCA433 cells were treated with either 30 or 70 M LPA for 1 h at 37°C and then evaluated for ␤1 integrin activation by flow cytometry analysis using a conformationspecific antibody that detects activated ␤1 integrin (HUTS21). Both 30 and 70 M LPA induced ␤1 integrin activation, and this activation was partially blocked by pretreatment with the LPA receptor inhibitor Ki16425 (Fig. 3, A and B). Disseminating ovarian tumor cells are exposed to LPA in ascites as anchorageindependent cells and multicellular aggregates. To evaluate LPA-induced ␤1 integrin clustering in anchorage-independent cells, OVCA433 cell suspensions were treated with 20 or 40 M LPA for 1 h at room temperature. Surface-expressed ␤1 integ- . Cells were fixed as described above and processed for immunofluorescent staining (anti-␤-catenin, green, 1:200). The experiment was repeated in triplicate. K and L, quantitation of junctional ␤-catenin staining was performed by counting a minimum of 12 fields per treatment and scoring as positive the number of cells with two remaining fluorescent cell-cell borders (30). K, quantitation of data represented by C and D above. L, quantitation of data represented by E-J above. *, p Ͻ 0.05 rin was cross-linked using a nonactivating anti-␤1 integrin antibody, and cells were processed for immunostaining for ␤1 integrin (41). Representative high powered fields are shown in Fig. 3, C and D. ␤1 integrin clustering was potentiated by LPA treatment in a dose-dependent manner, with a 1.5-fold increase of the number of ␤1 integrin clusters (green) following 20 M LPA treatment and a 2-fold increase of clusters following 40 M LPA (Fig. 3, C and D).
To further examine the effect of LPA on ␤1 integrins and junctional integrity, cells were treated with 30 M LPA for various time points prior to processing for dual label immunofluorescence microscopy to examine LPA-induced changes in localization of ␤1 integrin and E-cadherin. Following LPA treatment, a time-dependent aggregation of ␤1 integrin was observed in both OVCA429 and OVCA433 cells (Fig. 4, A and  B). Higher magnification examination of sites of ␤1 integrin clustering shows recruitment of E-cadherin into clustered integrin complexes (Fig. 4, C and D).
Lysophosphatidic Acid Potentiates Nuclear Accumulation of ␤-Catenin and Transcriptional Activation-Loss of junctional E-cadherin can target ␤-catenin to the nucleus or for degradation (33). Although the exact mechanism of ␤-catenin nuclear transport is unknown, cytoplasmic and perinuclear accumulation of ␤-catenin is coupled to increased nuclear activation of ␤-catenin target genes. To evaluate the effect of LPA on nuclear ␤-catenin accumulation, OVCA429 and OVCA433 cells were treated with LPA; nuclear proteins were isolated by subcellular fractionation as described under "Experimental Procedures," and fractions were evaluated by Western blot. In control experiments, cells were treated with LiCl (positive control treatment) or left untreated. Results show that LPA treatment potentiates an ϳ50% increase in nuclear ␤-catenin localization in both OVCA429 and OVCA433 cells (Fig. 5, A and B). Mechanisms for activation of Wnt signaling in human EOC tumors exhibit histotype dependence, with endometrioid EOC exhibiting mutations in ␤-catenin resulting in constitutively active Wnt signaling, although other histotypes do not harbor a significant level of activating mutations in this pathway (42). Immunohistochemical evaluation of a human ovarian tumor tissue microarray demonstrates nuclear localization of ␤-catenin in each of the four EOC subtypes (Fig. 5, C-F, and Table 1). Levels of nuclear ␤-catenin in endometrioid tumor samples were consistent with literature values (84.6%). In small cohorts of mucinous and clear cell samples, 66.6 and 62.5% were positive for nuclear ␤-catenin expression, respectively, although 50.0% of serous tumors exhibited positive nuclear staining. Eight metastatic lesions (seven serous and one clear cell) were also analyzed for nuclear ␤-catenin with five of the seven serous tumors exhibiting positive staining (71.4%), in addition to positive immunoreactivity in the clear cell metastasis. These data sup-port the hypothesis that factors other than mutational activation may regulate Wnt/␤-catenin signaling in human EOC.
The ␤-catenin protein sequence does not contain a nuclear localization signal domain, and the mechanism of nuclear transport is unclear (43)(44)(45). Nevertheless, ␤-catenin activates transcription by displacing the transcriptional repressor Groucho in Tcf/Lef-binding site-containing gene promoters and subsequently binding the co-activators Tcf/Lef (and potentially other co-factors such as CBP) to initiate transcription (46 -48). To assess the potential for LPA regulation of protein/protein interactions between ␤-catenin and Tcf, nuclear proteins were isolated as described and subjected to immunoprecipitation using anti-␤-catenin antibodies. Immunoprecipitates were then probed for ␤-catenin or Tcf. Results show that Tcf/␤-catenin interaction was enhanced in cells treated with LPA relative to untreated or DMSO treated controls (Fig. 6A, inset). Pretreatment with Ki16425 reduced this protein/protein interaction to baseline levels, suggesting that LPA-induced Tcf/␤-catenin interaction is LPA receptordependent To measure activation of the ␤-catenin⅐Tcf⅐Lef transcriptional complex, cells were transfected with a Renilla luciferase reporter construct and either a TOP (Tcf) reporter construct or a FOP (control) reporter construct and then were treated with LPA for 2, 8, or 24 h. ␤-Catenin/Tcf/Lef transcriptional activity was increased after a 2-h LPA treatment and was sustained at 8 and 30 h in Fig. 6, B and C. Consequentially, LPA-induced ␤-catenin/Tcf/Lef transcriptional activity led to up-regulation of five known ␤-catenin/Tcf/Lef target genes: VIM (vimentin), WNT5A, LRP6, PTGS2 (Cox-2), and SNAI1 (Snail) (Fig. 7A). To evaluate a potential functional link between LPA-induced ␤1 integrin clustering and transcriptional activation of Wnt/␤-catenin targets, cells were treated with LPA, and activation of FAK was evaluated by Western blotting for phospho-Tyr-397 (Fig.7B,inset).LPAtreatmentenhancedFAKphosphorylation (Fig. 7B, lane b, inset), and this effect was abrogated by the addition of antibodies that prevent ␤1 integrin clustering as well as by the LPA receptor inhibitor Ki16427 (Fig. 7B, lanes  c and d, inset). As recent studies have demonstrated that LPA can modulate integrin function through Rho/ROCK activation, we also examined the effect of the ROCK inhibitor Y27632 (Fig.  7B, lane e, inset), also resulting in inhibition of FAK phosphorylation. To address the resulting effects on transcription, we examined expression of vimentin, as this ␤-catenin target gene was most significantly up-regulated following LPA treatment (Fig. 7A). LPA-induced vimentin expression was blocked by inhibition of ␤1 integrin clustering (Fig. 7C), supporting the hypothesis that LPA-integrin cross-talk and subsequent integrin clustering regulate Wnt/␤-catenin signaling and gene expression. Similar results were obtained following inhibition of LPA receptor or Rho/ROCK activity (Fig. 7C).

Discussion
LPA regulates a multitude of ovarian tumor cell responses, including proliferation, migration, and invasion (19,24,28). LPA is expressed as high as 80 M in the ascites fluid and serum of patients with ovarian cancer (24, 25, 49 -51), underlying the importance of understanding its pathophysiological role in ovarian cancer. Previous studies in cancer cells have shown that LPA induces ␤-catenin nuclear translocation. For example, in colon cancer cells, LPA activates ␤-catenin via modulation of glycogen synthase kinase 3␤, resulting in enhanced transcription of a ␤-catenin/TCF reporter gene (52,53). Similar results were obtained in A431 epidermoid carcinoma cells (33,54), resulting in loss of junctional E-cadherin and gain of vimentin expression. Data in this study define a role for LPA in activation of ␤-catenin-regulated transcription and induction of an EMT program in ovarian cancer cells and multicellular aggregates and provide novel mechanistic insight on the role of LPA-induced clustering of ␤1 integrins in modulation of this signaling pathway.
We have recently shown that LPA modulates loss of epithelial cohesion as a functional result of disrupting cell-cell junctions, through protease-dependent cleavage and altered cadherin trafficking in ovarian carcinoma (30). The fate of FIGURE 5. LPA induces LPA receptor-dependent nuclear translocation of ␤-catenin. A and B, cells were treated as labeled (Con, untreated control; LPA, 40 M; LiCl, positive control, 40 M) and then subjected to subcellular fractionation as described under "Experimental Procedures." Fractions were electrophoresed on 9% SDS-polyacrylamide gels and immunoblotted for ␤-catenin expression (anti-␤-catenin, 1:1000). From analysis of triplicate blots, ␤-catenin staining in the nuclear fraction of both OVCA429 (A) and OVCA433 (B) cell lines was increased ϳ50% in LPA-treated cells compared with control (graphs). Control blots were probed with HDAC1 (middle panels) or ␤-actin (lower panels) to detect the presence of nuclear or cytoplasmic compartments, respectively. WCL designates unfractionated whole cell lysate. Molecular mass of ␤-catenin, 92 kDa. C-F, representative samples of each of the four major EOC histotypes were stained by immunohistochemistry (IHC) with anti-␤-catenin (BD Transduction Laboratories, 1:50) and a biotinylated secondary antibody (1:200; Vectastain ABC, Vector Laboratories). Slides were finally subjected to 3,3Ј-diaminobenzidine peroxidase (Vector Laboratories) exposure and hematoxylin staining. ␤-Catenin staining is present at cell-cell junctions, in the cytoplasm, and in the nucleus. Nuclear ␤-catenin is found in serous (C), endometrioid (D), clear cell (E), and mucinous (F) tumors.  Where indicated, inhibitor was preincubated for 15 min prior to addition of LPA. Following treatment, cells were lysed, and ␤-catenin was immunoprecipitated (IP) as described under "Experimental Procedures." Lysates were electrophoresed on 9% SDS-polyacrylamide gels and immunoblotted (IB) with an anti-Tcf antibody (inset, 1:1000) or anti-␤-catenin (inset, 1:1000). Band intensity was quantified using FUJIFILM MultiGauge version 3.0 and is represented as relative intensity (percent of control  ␤-catenin in response to LPA-mediated adherens junction dissolution, however, had not been previously reported. Data presented here are consistent with our previous work and demonstrate loss of cell surface-associated ␤-catenin following LPA treatment. Interestingly, immunofluorescence microscopy results demonstrate that perinuclear accumulation of ␤-catenin is observed following LPA treatment, corresponding with data that first identified an intersection between the cellcell adhesion-related and transcription-related cytoplasmic pools of ␤-catenin (33). The theory that cytoplasmic ␤-catenin pools intersect nuclear ␤-catenin pools is further supported by data demonstrating increased nuclear localization, Tcf/Lef reporter activity, and ␤-catenin target gene transcription in response to LPA treatment. Although the precise mechanism by which LPA treatment modulates ␤-catenin translocation is unknown, the current data support the hypothesis that LPA-integrin cross-talk results in ␤1 integrin activation. Many studies have suggested convergent signaling between growth factors and integrin-mediated adhesion processes (55)(56)(57). Several G-protein-coupled agonists, including LPA, have been shown to induce phosphorylation of the integrin effector p125FAK (58). Additionally, LPA-induced migration in fibroblasts is dependent on ␤1 integrin expression (59), and LPA/LPAR 2 interaction activates TGF-␤ signaling in an integrin-mediated manner (60). Interestingly, LPA treatment also leads to increased ovarian cancer cell adhesion to collagen, as well as enhanced ␤1 integrin protein expression (29). We have previously reported that ␤1 integrin aggregation alters ␤-catenin dynamics via disruption of adherens junctions and inhibition of glycogen synthase kinase 3␤ activity, leading to enhanced transcriptionally active ␤-catenin (23). This is consistent with data in the current report showing co-localization of E-cadherin at sites of clustered ␤1 integrins. Furthermore, three-dimensional collagen culture induces ␤1 integrin aggregation and down-regulates the expression of dickkopf-1, an inhibitor of canonical Wnt signaling (61,62). These data support a mechanism whereby LPA-induced ␤1 integrin aggregation activates ␤-catenin signaling.
Further support for this mechanism is provided by the observation that the following five pro-metastatic ␤-catenin target genes are up-regulated following LPA treatment: VIM (vimentin), WNT5A, LRP6, PTGS2 (Cox-2), and SNAI1 (Snail1). Expression of these genes was also up-regulated following multivalent ␤1 integrin engagement (23) and inhibition of LPA-induced ␤1 integrin aggregation blocked vimentin transcription. Cox-2 contributes to tumorigenesis by inhibiting apoptosis, increasing growth factor expression to promote angiogenesis, and by enhancing matrix metalloproteinase expression to stimulate invasion (63). Cox-2 protein is expressed in ovarian carcinoma and functions as a downstream effector of LPA-mediated ovarian tumor cell migration and invasion (64). Genes commonly associated with EMT, including vimentin, snail, LRP6, and Wnt5a, were also induced by LPA treatment. Ovarian cancers typically display both epithelial and mesenchymal characteristics, and the mesenchymal marker vimentin is widely expressed in human tumor specimens (16). Recent molecular profiling studies support these findings, showing up-regulation of EMT and EMT-associated invasion programs in primary ovarian tumors and their matched metastases (65,66). Snail is a key inducer of EMT and functions as a negative regulator of E-cadherin transcription. Several studies have demonstrated that nuclear localization of Snail correlates with tumor progression, with enhanced Snail immunoreactivity in metastatic lesions (67,68). Snail and other mesenchymal genes were found to be up-regulated in a panel of ovarian cancer cells competent for mesothelial clearance, a surrogate assay that models initial events in intraperitoneal metastasis (21). Furthermore, patients with both primary and metastatic tumors positive for Snail expression showed a significant decrease in overall survival (69). LRP6 functions as a Wnt coreceptor that recruits Axin and Dishevelled to the plasma membrane, thereby disrupting the degradation of ␤-catenin and facilitating ␤-catenin nuclear translocation (70). LRP6 is expressed by ovarian carcinoma cell lines and tissues (71). Expression of Wnt5a, a ligand for the Frizzled receptor, is high in ovarian cancer tumors and ascites fluid and is linked to poor overall survival (33,69,70). Emerging data suggest that activation of canonical versus noncanonical Wnt signaling by Wnt5a is dependent on the receptor/co-receptor context, and further studies in ovarian cancer cells are warranted (72)(73)(74)(75).
In summary, the current data demonstrate that lysophosphatidic acid, expressed in high concentration in ascites fluid of women with ovarian cancer, potentiates ␤-catenin-dependent Wnt signaling via a mechanism that involves activation and aggregation of ␤1 integrin. Currently, a number of therapeutic agents that target Wnt signaling, both loss of function and gain of function alterations, are being investigated in the treatment of cancer, bone disease, cardiac and vascular disease, and arthritis (76,77). Therapeutics under development are designed to target each level of Wnt signaling (extracellular, cytoplasmic, nuclear, and pathway cross-talk) and include small molecule and biotherapeutic agents (78 -80).
Although not yet extensively evaluated in ovarian cancer, recent preclinical studies support a role for further evaluation of Wnt/␤catenin signaling as a potential therapeutic target (71,81). Understanding the mechanisms by which microenvironmental factors such as lysophosphatidic acid may contribute to ovarian cancer progression and metastasis via modulation of this pathway may thereby contribute to identification of alternative treatment options for women with ovarian cancer.