Wnt5A modulates integrin expression in a receptor-dependent manner in ovarian cancer cells

Wnt5A signals through various receptors that confer versatile biological functions. Here, we used Wnt5A overexpressing human ovarian SKOV-3 and OVCAR-3 stable clones for assessing integrin expression, cell proliferation, migration, invasion, and the ability of multicellular aggregates (MCAs) formation. We found here, that Wnt5A regulates differently the expression of its receptors in the stable Wnt5A overexpressing clones. The expression levels of Frizzled (FZD)-2 and -5, were increased in different clones. However ROR-1, -2 expression levels were differently regulated in clones. Wnt5A overexpressing clones showed increased cell proliferation, migration, and clonogenicity. Moreover, Wnt5A overexpressing SKOV-3 clone showed increased MCAs formation ability. Cell invasion had been increased in OVCAR-3-derived clones, while this was decreased in SKOV-3-derived clone. Importantly, αv integrin expression levels were increased in all assessed clones, accompanied by increased cell attachment to fibronectin and focal adhesion kinase activity. Moreover, the treatment of clones with Box5 as a Wnt5A/FZD5 antagonist abrogates ITGAV increase, cell proliferation, migration, and their attachment to fibronectin. Accordingly, we observed significantly higher expression levels of ITGAV and ITGB3 in human high-grade serous ovarian cancer specimens and ITGAV correlated positively with Wnt5A in metastatic serous type ovarian cancer. In summary, we hypothesize here, that Wnt5A/FZD-5 signaling modulate αv integrin expression levels that could be associated with ovarian cancer cell proliferation, migration, and fibronectin attachment.

Wnt5A alters SKOV-3 MCAs formation ability and compactness. The formation of MCAs is one of the hallmarks of a metastasizing ovarian carcinoma 29 , which supports the survival of the tumor cells in ascites and protects them from chemotherapy [30][31][32] . Here we found that Wnt5A overexpression led to a significant increase by 3.6-fold in MCAs formation ability of C9/SKOV-3 clone compared to mock (P < 0.001, Fig. 4A, left and right panels); however, those MCAs were looser than mock MCAs (Fig. 4A) which was abrogated in the Wnt5A knocked-down C9/SKOV-3 clone (Fig. 4B, lower panel). Moreover, we found an increased level of Wnt5A, c-myc, and CDH-1 expression from day 3 to day 6 in C9/SKOV-3 MCAs compared to mock MCAs ( Supplementary Fig. S4B).
Next, we assessed the cell adhesion of Wnt5A overexpressing clones to extracellular matrix components which showed increased cell adhesion to fibronectin (FN) and laminin (LN) compared to mock cells (Fig. 6A, upper and lower panels). Meanwhile, Box5 reversed adhesion to FN ( Fig. 6A lower panel), and C9/SKOV-3 clone cell adhesion to Collagen type -I (Col. I) and -IV (Col. IV) was decreased up to 80% (1 h incubation) compared to mock (Fig. 6B). Subsequently, C9/SKOV-3 clone cell adhesion to FN or LN showed increased pTyr 397 -FAK immunostaining ( Fig. 6C), indicating activation of FAK, as an important regulator of integrin signaling 33 . Moreover, FN-dependent FAK activation had been reduced in the presence of Box5 (Fig. 6C). In line with our data here, loose C9/SKOV-3 MCAs became more compact upon the addition of FN into the C9/SKOV-3 clone (Fig. 6D). Similarly, C3/OVCAR-3 clone cell adhesion to FN and LN has been significantly increased and its adhesion to FN was abrogated in the presence of Box5 (Fig. 6E). There was an increased C3/OVCAR-3 clone cell adhesion to Col I and Col IV that was not modulated by the presence of Box5 (Fig. 6E). Altogether, our results showed increased adhesion of Wnt5A overexpressing clones to FN that may require Wnt5A/FZD-5 signaling as Box5 reverts it.Wnt5A alters E-cadherin expression and morphology of Wnt5A overexpressing SKOV-3 clones.
Wnt5A correlates positively with ITGA5, ITGAV, ITGA4, and ITGB6 expression in the serous type ovarian cancer. We drew the heatmap plot to classify the up-regulated and down-regulated analyzed genes including integrins, CDH-1, CDH-2, and Wnt5A in different subtypes of serous ovarian cancer and normal ovary (Fig. 7A). Next, a two factorial experimental design was developed to evaluate the difference in gene expression between groups which showed that the assessed genes were significantly different between normal and cancerous groups (P < 2.2 × 10-16). CDH-1 and CDH-2 were significantly lower in the HGSOC group compared to other groups (Fig. 7A). In HGSOC, we found a higher expression levels of ITGA2 (P < 0.001), ITGA4 (P < 0.05), ITGA5 (P < 0.05), ITGB6 (P < 0.01) compared to other groups ( Fig. S6A-D). In HGSOC, ITGAV expression levels were higher (P < 0.05, compared to BL and normal groups), and ITGB1 and ITGB3 were higher compared to normal and LGSOC groups (P < 0.01 and P < 0.05, respectively) ( Fig. S6E-G). We further investigated the existence of a relationship between Wnt5A, and integrins in the metastatic groups www.nature.com/scientificreports/ (LGSOC + HGSOC). We found that Wnt5A was positively correlated with ITGA4 (r = 0.52, P = 0.02), ITGA5 (r = 0.39, P = 0.05), ITGAV (r = 0.44, P = 0.03) and ITGB6 (r = 0.53, P = 0.01). Our results with human specimens partially support our data from in vitro experiments and suggest that Wnt5A in ovarian cancer exerts a modulatory role on integrin expression. Inhibition of αv integrin abrogates Wnt5A-induced cell proliferation and migration. Since we found that Wnt5A and ITGAV contribute to cell proliferation and migration (Fig. 7D), it was tempting to investigate the effect of specific αv inhibitor CWHM-12 on Wnt5A overexpressed cells. We found decreased cell proliferation in the presence of CWHM-12. (Fig. 8A). Moreover, cell migration was decreased by 22% and 57% in mock, and C3/OVCAR-3 clone in the presence of CWHM-12, respectively compared to untreated cells (Fig. 8B upper and lower panels). This supports the hypothesis that Wnt5A modulates cell proliferation and migration of ovarian cancer cells through up-regulation of ITGAV expression.

Discussion
Analysis of a large cohort of OVC patients and evaluation of data obtained from The Cancer Genome Atlas (TCGA) found up-regulation of Wnt5A expression in all major histotypes relative to benign controls 16,35 . Furthermore, Wnt5A protein is present in ovarian tumor ascites 16 , supporting its contribution to the ovarian cancer progression 36 . Our previous report showed the role of Wnt5A in the regulation of substrate-dependent adhesion and migration 20 , though, the exact mechanism by which Wnt5A promotes cell adhesion remains unclear. Since Wnt5A signals through different receptors related to its versatile biological effect, therefore, here we investigate first whether Wnt5A could regulate its receptors in our model. We found up-regulation of FZD-2 and FZD-5 and down-regulation of FZD-4 but ROR1 and ROR2 have been differently modulated in the Wnt5A overexpressing clones. Interestingly, changes in the expression levels of these receptors were rescued either by Wnt5A knock-down or inhibiting Wnt5A signaling by the Box5 antagonist which is expected to exert its effect via a direct FZD-5 binding 37,38 . Here, we report increased cell adhesion to FN, increased cell migration, and cell proliferation, which had been reversed in the presence of Box5. In line with our findings here, it has been reported that Wnt5A/FZD-5 signaling increased melanoma cell adhesion and migration, which was reversed in the presence of Box5 37,39 . Our findings here are further supported by the fact that Knockdown of either ROR1 or Results of RT-qPCR were normalized related to GAPDH used as an internal control. Mean ± SD of three independent experiments. a: compared to mock, and b: compared to Wnt5A overexpressing clones. *P < 0.05; **P < 0.01; ***P < 0.001 compared to mock. www.nature.com/scientificreports/ ROR2 alone did not affect cell adhesion to collagen or fibronectin, suggesting that the ROR receptors do not play a major role in regulating ovarian cancer cell adhesion 35 . However, either ROR1 or ROR2 contributes to OVC cell invasion 35 corroborating with our observation here that C9/SKOV-3 clone with down-regulated RORs receptors showed reduced cell invasion but a higher invasive index in Wnt5A overexpressing OVCAR-3 clones with up-regulated ROR2 receptor was observed. Furthermore, our findings here are substantiated by a significantly decreased level of ROR-2-inducedn MMP-13 in C9/SKOV-3 clone. Of particular note, MMP-13 is reported to be associated with the invasion ability of osteosarcoma, and thyroid cell carcinoma 27,28 . Moreover, a recent study showed that stable knockdown of ROR2 led to reverse the epithelial-mesenchymal transition in SKOV3 cells 40 . This may support our finding here that the decreased levels of ROR2 in the C9/SKOV-3 clone could be associated with its epithelial-like morphology accompanied by increased levels of E-cadherin. Wnt5A is associated with EMT in ovarian cancer 36,41 however, here, we found more motility in C9/SKOV-3 clone with increased E-cadherin levels. This may propose that Wnt5A-mediated EMT may not necessarily be associated with the down-regulation of E-cadherin. This hypothesis may be further supported by the fact that other studies report a similar phenomenon in colorectal cancer 42 in addition to, the observed mesenchymal phenotype in Wnt5A knocked-down breast cancer cells 43 . Of particular interest, we observed cytoplasmic E-cadherin immunostaining in C9/SKOV-3 clone, and in our 3D model, C9/SKOV-3 MCAs showed looser cell-cell contact and were more sensitive to Paclitaxel compared to mock (Data not shown) which may suggest that E-cadherin can be internalized and stabilized in these cells rather than forming a cell-cell junction.
To the best of our knowledge, there is no report about Wnt5A regulatory role on integrin expression and/or activation in ovarian cancer. Here we show for the first time that Wnt5A overexpression leads to the increased expression levels of αv and α5 integrins which could heterodimerize with β1, β3, β5, β6, and β8 integrins interacting with FN or LN and subsequent activation of FAK in a substrate-dependent manner. FAK is activated through autophosphorylation at Tyr 397 , which is initiated by the integrin engagement with its ligand, and the turnover of focal adhesions is required for a cell to spread and migrate 44 . One study showed that the expression of integrins ITGA1, ITGA2, and ITGAV increased over time and correlated with increased Wnt5A expression in human mesenchymal stem cell (HMSCs) differentiation to osteoblastic lineage and treatment of HMSCs with Wnt5A, increased integrin expression 25 . In this study, we found up-regulation of ITGAV in Wnt5A overexpressing clones RT-qPCR analysis of integrin subunits in C9/SKOV-3 clone with or without siRNA Wnt5A or Box5 relative to mock or scrambled (scr). (D) RT-qPCR analysis of integrin subunits in C3/OVCAR-3 clone with or without siRNA Wnt5A or Box5 relative to mock or scrambled (scr). GAPDH was used as an internal control and the data represent mean ± SD (n = 3) *P < .05; **P < .01; ***P < .001 compared to mock. www.nature.com/scientificreports/ and its reversal in the presence of Box5, suggesting that Wnt5A/FZD-5 signaling could mediate its increase. In line with this hypothesis, FZD-5 increased adhesion to FN and vitronectin in ovarian cancer cells 45 . We also found here increased expression levels of FZD-2 in Wnt5A overexpressing clones. It is worth noting that the dynamics of Wnt5A-dependent focal adhesion activity had been regulated through FZD-2, Dvl, and APC 21 . Future studies by targeting FZD-2 or FZD-5 may unravel and shed light on the understanding of Wnt5A's modulatory role in integrin expression. Integrins mediate the initial aggregation of MCAs 31,46,47 . Interestingly, this study showed that the fold change of α5, αv, and α5β1 FN binding integrin subunits were remarkably higher in MCAs compared to monolayer C9/SKOV-3 clone. It has been reported that the interaction between α5β1 and FN is involved in the formation, adhesion, and disaggregation of ovarian cancer MCAs 31,32,48,49 . Accordingly, we showed here that Wnt5A overexpressing cells were able to form compact MCAs in the presence of FN. In line with these in vitro data, we found a positive relationship between Wnt5A and α5, αv, and β6 expression in the metastatic serous ovarian cancer groups. This is consistent with the emerging data suggesting that αvβ 6 and α5β1 integrins regulate invasion and metastases of ovarian cancer 15,50 .
Here, we found increased cell proliferation in the assessed Wnt5A overexpressing clones which showed increased FZD-2 and FZD-5 mRNA levels. One study has shown that FZD-2 stimulated cell proliferation and www.nature.com/scientificreports/ promoted cell migration in high-risk neuroblastoma by interfering with _β-catenin-dependent and β-cateninindependent signaling pathways 51 . Moreover, Wnt5A/FZD-2 signaling could activate Src family kinases (SFKs) and induces cervical, lung, and esophageal cancer cell proliferation 52 . It should be noted that SFKs primarily transmit signals downstream of receptor tyrosine kinases (RTKs) and integrins to regulate cell proliferation, motility, and survival 53 . Another important finding in the present study is that Wnt5A/FZD-5-mediated ITGAV up-regulation contributes to cell proliferation and migration. Furthermore, knockout of FZD-5 robustly inhibited cell growth in breast cancer cells and RNF43-mutant pancreatic ductal adenocarcinoma cells 54,55 . In line with these findings, we showed here, that antagonizing Wnt5A/FZD-5 with Box5 abrogates increased cell proliferation in Wnt5A overexpressing clones. Since GO (BP) enrichment analysis revealed positive regulation of cell proliferation by Wnt5A and ITGAV we could also suggest that Wnt5A-induced ITGAV contributes to cell proliferation in our model.
In conclusion, we identified that Wnt5A affects integrin expression, and particularly Wnt5A/FZD-5 signaling affects αv integrin expression, cell proliferation, migration, and invasion. In particular, Wnt5A-induced ITGAV may be important in increasing tumor cell adhesion, proliferation, and migration contributing to OVC progression.

Methods
All methods were carried out following relevant guidelines and regulations and, all experimental protocols were approved by the University of Tehran, college of science and Royan institute. For human specimens, informed consent was obtained from all subjects.
Cell line and culture conditions. SKOV-3 and OVCAR-3 cell lines (ovarian adenocarcinoma) were provided by Dr. Zarnani A.H from Avicenna research institute and cultured as previously described 20 . For 3D culture, cells were seeded in plates coated with 1% low melt agarose (IBI SCIENTIFIC, Tryon, NC, USA) in complete medium (RPMI containing 10% FBS, 1% L-glutamine, 1% penicillin/streptomycin). Scientific AG) known as scramble (Scr) as previously described 56 .Wnt5A expression at mRNA and protein level was detected as previously described 56 . Quantification of gene expression was performed via the standard curve method using REST-RG software version 3.
Scratch wound-healing assay and transwell invasion assay. Cells reached 90% confluency were starved overnight, then treated for 2 h with 10 µg/ml Mitomycin (Merck KGaA, Darmstadt, Germany). The wound was made by scratching the cell monolayer with a yellow tip and controlled after 3, 6, 12, 24 h with or without Box5 or in the presence or absence of CWHM-12. The invasiveness of Wnt5A overexpressing clones and mock cells was assessed (Calbiochem, USA) using 8 µm pore size transwells coated with Matrigel as previously described 56 .
Spheroid formation assay. To assess the spheroid formation ability of C9/SKOV-3 clone versus mock 1.5 × 10 4 cells were cultured in 6-well plates coated with 1% low melt agarose (IBI SCIENTIFIC, Tryon, NC, USA) in serum-free RPMI supplemented with 20 ng/ml recombinant human epithelial growth factor and 20 ng/ ml recombinant human basic fibroblast growth factor (rhEGF, rhFGF, Royan Institute, Tehran, Iran) which were added every 48 h. After 6 days, the numbers of spheres greater than 70 μm in diameter were counted using an inverted microscope at 400× magnification. www.nature.com/scientificreports/ Flow cytometry assay. The cell cycle assay was performed using propidium iodide (PI) DNA staining.
Integrin array assay. Assessment of integrin proteins on the cell surface of C9/SKOV-3 clone and mock in both monolayer and MCAs was performed by using the Alpha/Beta Integrin-Mediated Cell Adhesion Array Combo Kit (Chemicon, Billerica, Massachusetts, USA) according to the manufacture protocol as previously described 57 .
Immunofluorescence and western blot analysis. pTyr 397 -FAK and E-cadherin (1:200, Santa Cruz Biotechnology, INC.) were detected in C9/SKOV-3 clone and mock cells using polyclonal rabbit anti-human pTyr 397 -FAK and polyclonal rabbit anti-human E-cadherin clone H-108 antibodies, respectively. Immunofluorescence was performed as previously described 19 . Human ovarian specimens. Serous type epithelial ovarian cancer (EOC) tumors and normal ovarian tissue specimens were obtained from surgeries performed at Imam-Khomeini University Hospital Complex. Approval was obtained from the institutional Ethics Committee on Human Investigation (Imam-Khomeini University Hospital Complex) following the World Medical Association guidelines (Helsinki Declaration of 2008) for research on human beings and informed consents were obtained from patients. All samples (n = 57) were examined by two independent and experienced gynecological pathologists for histological diagnosis and grade. The characteristics of patients are described in Supplementary Table S1. The patients (age = 24-71 and Median = 43) were divided into four groups: normal ovary (n = 10); borderline serous ovarian cancer (BLSOC, n = 12); low-grade serous ovarian cancer (LGSOC, n = 12) and high-grade serous ovarian cancer (HGSOC, n = 23). Samples were chopped into small pieces of 50 mg with a surgical bladder and immediately snap-frozen in liquid nitrogen for further RT-qPCR analysis of Wnt5A and integrins as described previously 56 . The sequences of primers are listed in Supplementary  Table S2. The expression levels of target genes are normalized related to the ACTB gene. Quantification of gene expression was performed via the standard curve method using REST-RG software version 3.
Hierarchical clustering analysis and functional and pathway enrichment analysis. A bidirectional hierarchical clustering heatmap of assessed genes in human specimens was constructed using gplots package of R language after extracting the expression values from the gene expression profile. Terms with a P-value of < 0.05 were collected and grouped into clusters based on their membership similarities. More specifically, P-values were calculated based on the cumulative hypergeometric distribution. The most significant term within a cluster was selected as the one representing the cluster. Subsequently, functional enrichment analysis was performed in 3 categories of GO terms: Biological process (BP), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment were performed to determine the involvement of genes in different biological pathways by using Enrichr (a web-based enrichment analysis tool for non-ranked gene lists that is based on Fisher's exact test). Moreover, protein-protein interaction between Wnt5A and integrins were analyzed with STRING v.11 (https ://strin g-db.org/). GO terms with a P-value of < 0.05 were considered statistically significant and annotation results for GO (BP) were displayed using VENNY v2.1.

Statistical analysis.
The normality of nominal variables was analyzed by performing the Kolmogorov-Smirnov test. Skewed and normal distributed metric variables were analyzed between two groups using Mann-Whitney U or among multiple groups using Kruskal-Wallis and one-way ANOVA tests, respectively by using R version 3.5.2. Correlations between gene expressions were analyzed by the spearman's correlation coefficient test. All experiments were performed at least three times in triplicate and the results were expressed as mean + /-SD. P < 0.05 considered significant.