PTEN loss in the fallopian tube induces hyperplasia and ovarian tumor formation

Abstract

The signaling events involved in the onset of ovarian cancer from the fallopian tube epithelium (FTE) are crucial for early detection and treatment of the disease, but they remain poorly defined. Conditional homozygous knockout of PTEN mediated by PAX8-cre recombinase was sufficient to drive endometrioid and serous borderline ovarian carcinoma, providing the first model of FTE-derived borderline tumors. In addition, heterozygous PTEN deletion in the FTE resulted in hyperplasia, providing a model to study early events of human ovarian pathogenesis. To uncover the mechanism underlying the invasion of cancerous oviductal cells to the ovary, PTEN-deficient murine oviductal cells were developed and tagged with green fluorescent protein. Loss of PTEN increased cell migration, invasion, and upregulated WNT4, a key regulator of Müllerian duct development during embryogenesis. Further investigation revealed that WNT4 was required for increased migration and colonization of the ovary by PTEN-deficient oviductal cells in a β-catenin independent manner. Human tumor microarrays and ovarian cancer cells lines confirmed WNT4 expression in cancer and its role in migration. Together, these findings provide a novel model to study the mechanism of fallopian tube tumor initiation and invasion to the ovary mediated by loss of PTEN, which may help to define early events of human ovarian carcinogenesis.

Introduction

Epithelial ovarian cancer is the most lethal gynecological malignancy with approximately 14,000 deaths each year in the U.S. (1). Ovarian carcinomas are classified as Type I or Type II based on molecular and histopathological differences (2). Type I includes low-grade serous carcinoma (LGSC), endometrioid, some clear cell, mucinous, and Brenner carcinomas; while Type II tumors include high-grade serous carcinomas (HGSC) and undifferentiated carcinomas (2). Type II tumors harbor TP53 mutations (3) that can typically be visualized by immunohistochemistry, termed the “p53 signature,” and have a higher propensity for metastasis (2).

Evidence suggests that HGSC can originate from the fallopian tube epithelium (FTE) (4–6). In fact, early precancerous lesions have been identified in the fallopian tubes, but not in the ovaries, of women with HGSC and in women at high risk of getting HGSC (bearing BRCA½ mutations) (7–9). A meta-analysis found that salpingectomy was associated with a lower risk of developing endometrioid ovarian cancer and HGSC (10). Endometrioid ovarian cancer was recently reported to develop from both OSE and fallopian tube in murine models (11). LGSC arise from cystadenofibromas that progress to borderline tumors, and it may also be derived from fallopian tube mucosa (12). Serous borderline tumors (SBTs) have been suggested as a precursor for LGSC and occasionally of HGSC; however, no current model of SBT derived from FTE has been yet reported. Better models describing different histotypes of ovarian cancer and characterizing essential drivers of tumor formation from different epithelial cell population may improve detection, prevention, and treatment.

The importance of PTEN in high-grade serous cancer may have been underestimated. For example, over 30% of HGSC tumors present with loss of PTEN when the stromal DNA contribution is excluded (13). Immunohistochemistry studies of PTEN protein levels demonstrated total loss of PTEN in 15% (including stroma) of HGSC and partial loss in another 50% (14). To support the importance of PTEN loss in early stages of human ovarian tumorigenesis from the FTE, the loss of PTEN was found in 33% of serous tubal intraepithelial carcinoma (STIC) (15, 16). According to the TCGA, the PI3K/AKT pathway is altered in 45% of HGSC, and this could amplify downstream signaling similarly to the loss of PTEN, such as pAKT activation (17). PTEN mutations have also been found in about 20% of endometrioid ovarian carcinoma (18). In addition, the knockdown of PTEN alone in murine FTE cell lines was sufficient to generate subcutaneous tumors and peritoneal colonization in athymic nude mice (9).

Several models have helped to confirm that the FTE can give rise to different histotypes of ovarian cancer. Most murine models of FTE-derived cancer published so far require loss of PTEN, except for a model of OVGP1-driven SV40 expression in the oviduct that generated p53 stabilization, STIC and invasive adenocarcinoma (19). A transgenic mouse model with loss of BRCA½, Trp53, and PTEN driven by the PAX8-Tet-on-CRE promoter, in the fallopian tube, generated HGSC (29). A model of PTEN and Dicer deletion induced HGSC derived from FT stroma due to Amhr2-diven CRE expression (20). In addition, OVGP1-driven conditional deletion of PTEN and APC (Adenomatous polyposis coli) led to fallopian derived endometrioid cancer, suggesting that loss of PTEN in the fallopian tube may be part of both serous and endometrioid histotypes (47). A recent inducible murine model of OVGP1-CRE-ER driving BRCA1, Trp53, Rb1 and Nf1 inactivation generated HGSC with a long latency period compared to a model of BRCA1, Trp53 mutations and PTEN deletion (21). However, no reports have been published demonstrating that PTEN deletion in the FTE alone is tumorigenic, thus its role in ovarian carcinoma early pathogenesis from FTE has not yet been reported and may require a stronger CRE-driven deletion in FTE. Ovariectomy in the BRCA/PTEN/p53 murine model reduces intra-abdominal metastasis from transformed fallopian tube cells, suggesting that metastasis to the ovary is a prerequisite for further metastasis to the peritoneum (22). However, the mechanisms regulating FTE metastasis to the ovary are not known. WNT4 is essential during embryogenesis for development of the female reproductive tract (23, 24) and amplification of this locus was recently associated with increased incidence of HGSC (25). WNT4 is enriched in stem/progenitor cells of ovarian surface epithelium and oviduct fimbria (26) and it regulates cell migration of the Müllerian duct tip (27, 28). A spontaneous model of oviductal transformation that recapitulates aging by serial passaging expressed increased WNT4 levels, suggesting that it may be expressed during tumorigenesis (29).

Herein, we report a new transgenic mouse model and ex vivo ovarian colonization assay to study the contribution of PTEN loss from FTE to ovarian cancer pathogenesis with the intent of uncovering precursor lesions and molecular pathways regulating primary colonization of the ovary. This transgenic model expresses a CRE-recombinase driven by PAX8 to delete a floxed PTEN allele in the FTE. We demonstrated that homozygous loss of PTEN presents with tubal derived serous borderline and endometrioid ovarian carcinoma. This is the first model to report a fallopian tube-derived serous borderline cancer and, provides a model to study ovarian cancer genesis and progression from FTE. In addition, the loss of PTEN increased non-canonical WNT signaling in the FTE, which was critical for colonization of the ovary by the transformed FTE cells. These findings highlight the therapeutic potential of inhibiting the WNT non-canonical signaling pathway to decrease metastasis and invasion of the ovary by FTE-derived tumor cells, thereby blocking initial metastasis.

RESULTS

PAX8 drives CRE-recombinase expression in the oviduct but not in the ovary

To determine if loss of PTEN alone was sufficient to generate fallopian tube-derived ovarian cancer, a conditional knockout mouse model was generated by crossing mice expressing CRE-recombinase in the fallopian tube (PAX^CRE/+) with mice expressing the LoxP-PTEN-LoxP transgene (Supplementary Fig. 1SA). To validate the specificity of the model, PAX^CRE/+ mice were crossed with Rosa26 mice expressing the loxP-Stop-LoxP LacZ gene. β-galactosidase staining was performed on the tissues obtained from the mice progeny. As shown in Fig. 1A, the fallopian tube expressed the enzyme, as did the kidney, which was used as the positive control. The uterus was also positive with reduced staining compared to the oviducts. As expected, the ovary did not express PAX8 validating the specificity of the model. In addition, PCR was performed to detect recombination of the floxed PTEN allele in the oviduct of the PAX8^CRE/+PTEN^fl/fl mice as compared to wild-type and PTEN^fl/fl lacking PAX8^CRE/+ (Fig. Supplementary S1B) followed by immunohistochemistry to validate loss of PTEN protein in the FTE (Fig. 1B). Female PAX8^CRE/+PTEN^fl/fl, PAX8^CRE/+PTEN^fl/+, and PAX8^CRE/+ were aged to 7–9 months. A comparison of the reproductive tracts from PAX8^CRE/+PTEN^fl/fl, PAX8^CRE/+PTEN^fl/+, and the negative control PAX8^CRE/+ demonstrated that homozygous loss of PTEN resulted in enlarged uteri, oviducts and ovaries as well as the presence of tumors in the reproductive tract (Fig. 1C, Supplementary Fig. S1C).

Figure 1: Tissue specific knockout of PTEN in FTE results in oviductal and ovarian cancer.

A) β-galactosidase staining of tissues obtained from mice PAX8^CRE/+Rosa26 mice, showing specific staining of CRE (blue) in the oviduct and not the ovaries. Scale bar represent 200 μm. B) IHC using PTEN antibody shows immunoreactivity for PTEN in mice control (PAX8^CRE/+) as compared to PTEN knockout homozygous (PAX8^CREPTEN^fl/fl) at 7 months of age. C) Representative pictures of reproductive systems from control (PAX8^CRE/+) and PTEN knockout homozygous (PAX8^CREPTEN^fl/fl.

PAX8^CRE-driven heterozygous loss of PTEN in tubal epithelium results in hyperplasia

Histopathologic analysis demonstrated that by 3 months of age, homozygous PAX8^CRE/+PTEN^fl/fl mice presented with cancer in the oviducts and uterus whereas the ovaries were normal as were the PAX8^CRE/+ control animals. By 7 months of age, cancer cells had spread to the ovaries in the PAX8^CRE/+PTEN^fl/fl mice (Table S1). In contrast, heterozygous deletion of PTEN (PAX8^CRE/+PTEN^fl/+) in the oviduct did not result in cancer formation, however, it presented with various degrees of proliferative lesions (Fig. 2A bottom right panel), expressing WT1, PAX8, and it was immunopositive for p53 (Fig. 2B). The hyperplasia present in the oviduct showed expanded ducts with tubal epithelial proliferation, along with increased, disordered, and branched plica.

Figure 2: Heterozygous loss of PTEN in FTE results in hyperplasia.

A) H&E staining of tissues. Left panels show normal ovaries in control (PAX8^CRE+) and heterozygous (PAX8^CRE/+PTEN^fl/+) mice. Scale bar represents 500uM. Right panels show normal oviducts in control and hyperplastic oviduct in the heterozygous mice. Scale bar represents 50 μm. B) Immunohistochemistry (IHC) using specific antibodies targeting WT1, PAX8, and p53 in control (top) and heterozygous mice (bottom). Scale bar represents 50 μm. IHC images were scored based on staining intensity: 0=negative; 1=faint/weak; 2=intermediate/moderate; 3=strong. Top panels: p53=1; WT1=3: PAX8=3; bottom panels: p53=1; WT1=2: PAX8=2.

PAX8^CRE-driven homozygous loss of PTEN in tubal epithelia is sufficient to generate serous borderline tumors and endometrioid carcinomas

Histopathologic analysis revealed papillary and micropapillary growth patterns lined by mostly serous and focal mucinous epithelial proliferation. No definite stromal invasion was seen. This histology resembled human ovarian serous borderline or seromucinous tumors (Fig. 3A). The tumors or lesions around the oviducts in homozygous mice presented the endometrioid epithelial proliferation, complex back-to-back endometrioid glands with cribriform growth patterns. Areas of squamous metaplasia were identified within endometrioid lesions (Fig. 4A). These findings are indicative of endometrioid carcinoma (Fig. 4A) consistent with the previously reported APC^fl/flPTEN^fl/fl mouse model (11).

Figure 3: Homozygous knockout of PTEN in the oviduct generated serous borderline tumors.

A) H&E staining of tissues from homozygous PAX8^CRE/+PTEN^fl/fl mice showing serous borderline tumor as compared to human serous borderline tumors. Scale bar represents 100uM on the left panel and 50uM on middle and right panel. B) IHC of serous borderline tumors from PAX8^CRE/+PTEN^fl/fl mice using specific antibodies for CK8, WT1, acetylated tubulin (AcTUB), PAX8, Vimentin (Vim) and p53 (scale bar is 50 μm). IHC images were scored based on staining intensity: 0=negative; 1=faint/weak; 2=intermediate/moderate; 3=strong. CK8=3 membrane; WT1=3 nuclear; AcTub=3 membrane; PAX8 2; Vim: 1 focal; p53=2 (50% nuclear).

Figure 4. Homozygous knockout of PTEN in the oviduct generate endometrioid carcinoma.

A) H&E staining of tissues from homozygous PAX8^CRE/+PTEN^fl/fl mice. Scale bar represents 100 μm on the left panel and 50 μm on middle and right panel. Right panel shows squamous metaplasia, associated with endometrioid carcinoma. B) IHC from PAX8^CRE/+PTEN^fl/fl mice using specific antibodies for of CK8, WT1, acetylated tubulin (AcTUB), PAX8, Ki67 and p53. Scale bar represents 50 μm. IHC images were scored based on staining intensity: 0=negative; 1=faint/weak; 2=intermediate/moderate; 3=strong. CK8=3 membrane; WT1=2 nuclear; AcTub=0; PAX8 2 diffuse; Ki67=40% positive.

Characterization of ovarian cancer markers expression

Analysis of markers for Müllerian origin of carcinoma revealed positive staining for PAX8 and WT1 (Fig. 3B and Fig. 4B) in the tumors, suggesting that serous borderline and endometrioid carcinoma can arise from the fallopian tube. However, PAX8 expression seemed to be more diffuse in endometrioid compared to serous borderline where it presented with a distinct expression in the nucleus. p53 staining was also positive in both endometrioid and serous borderline carcinomas suggesting stabilization of p53 protein, but this was not exclusively nuclear. The stabilization of wild-type p53 in PAX8^CRE/+PTEN^fl/fl, was also confirmed by western blot in murine oviductal epithelial cells (MOE) with shRNA targeting PTEN (MOE:PTEN^shRNA ) or a scramble (SCR) shRNA (MOE:PTEN^shRNA) (Fig. S1D). Acetylated tubulin staining was more pronounced in serous borderline than in endometrioid carcinoma (Fig. 3B and 4B). Occasional stromal invasion suggested the presence of a low-grade serous ovarian carcinoma histotype.

Loss of PTEN results in increased migration and invasive potential of MOE cells

Our previous studies indicated that the loss of PTEN by shRNA increased migration of MOE cells in adherent conditions (9, 30). In our transgenic model with knockout of PTEN in the FTE, clusters were observed of invasive cells in both serous and endometrioid tumors (Fig. 5A–B H&E). These cells present, at the invasive front, with stabilization of p53 and CK8 expression (Fig. 5A–B). Consistent with previous studies (30), MOE:PTEN^shRNA cells presented increased migration in a wound healing assay (Fig. 5C) and increased invasion in a Boyden chamber assay (Fig. 5D) compared to cells MOE:SCR^shRNA. Increased ovulation number is associated with an increased risk of ovarian cancer (31, 32), leading to the hypothesis that damage during ovulation may favor invasive cells to colonize the ovary. To test if the increased migratory capability of MOE:PTEN^shRNA may reflect an increased ability to invade the ovary, an ex vivo invasion assay was developed where MOE cells stably expressing GFP were incubated with murine ovaries for 24 hrs. Ovaries were left intact or cut with a scalpel blade to mimic the physical tearing of the ovary during ovulation (hereafter referred to as ovulation mimetic). MOE:PTEN^shRNA-GFP invaded the intact and ovulation mimetic ovaries more than the MOE:SCR^shRNA-GFP, suggesting that this genetic modification is sufficient to confer oviductal cells with intrinsic invasive capabilities (Fig. 5E–F). Also, MOE:PTEN^shRNA-GFP invaded the wounded ovary to a greater extent than the intact ovary, supporting the ovulation hypothesis and the importance of loss of PTEN.

Figure 5: Loss of PTEN in the oviduct leads to increased migration and invasion.

A) Left panel, H&E staining of SBT, right panel IHC using p53 antibody. Dashed line delineates the invasive front between normal ovary=Ov and FTE cancerous cells. Scale bar=50 μm. IHC images were scored based on staining intensity: 0=negative; 1=faint/weak; 2=intermediate/moderate; 3=strong. P53=3 (80% nuclear). B) Left panel, H&E staining of endometrioid carcinoma cells invading the ovary, right panel IHC using CK8 antibody. Ov=ovary. Scale bar=50 μm. IHC images were scored based on staining intensity: 0=negative; 1=faint/weak; 2=intermediate/moderate; 3=strong. CK8=3 membrane. C) Wound healing assay in MOE cells stably transfected with two different shRNAs targeting PTEN (shRNA1 and shRNA2) and two different controls (SCR^shRNA and GFP^shRNA). Areas of scratches were measured at time zero and 8 hours later. Percent closure was measured in three different experiments, averaged and statistical analysis performed (*p<0.05). D) MOE cells stably transfected with shRNA targeting PTEN (shRNA1=PTEN^shRNA) and control (SCR^shRNA) were seeded in the insert of a Boyden chamber and invading cells counted after 24 hours (**p<0.01). E-F) Intact or ovulation mimetic ovaries from healthy mice were incubated with MOE:PTEN^shRNA-GFP or MOE:SCR^shRNA-GFP for 24 hours. GFP positive cells were counted from more than 3 independent experiments (*p<0.05; ***p<0.001).

Loss of PTEN in MOE results in alteration of the non-canonical WNT signaling

To identify the pathways regulated by loss of PTEN in FTE, RNAseq was performed on MOE:PTEN^shRNA and MOE:SCR^shRNA cells. RNAseq revealed an upregulation of components of the non-canonical WNT pathway such as WNT4, WNT10B, Sdc1, Sdc2, DKK3, DAAM2 (Fig. 6A). The WNT canonical signaling pathway was downregulated (Supplementary Fig. 4B and Fig. 6B), suggesting that a WNT non-canonical signaling pathway may mediate the invasive effect of PTEN loss. qPCR was used to confirm that CTNNB1 (gene encoding for β-catenin) as well as p-catenin targets LEF1 and TCF4 were not significantly regulated by loss of PTEN, whereas WNT4, DKK3 and Sdc1 were significantly up-regulated (Fig. 6C). In addition, WNT4 and DKK3 protein levels were increased upon loss of PTEN in cell lines (Fig. 6D–E) and in the PAX8^CRE/+PTEN^fl/fl tumors (Fig. 6F). WNT4 protein is expressed in human FTE (Fig. 6H) as well as in human ovarian cancer lines, particularly in OVCAR4 and OVCAR8 (Fig. 6G). In addition, analysis of human ovarian tumor microarrays (TMA) revealed the expression of WNT4 in serous, endometrioid and mucinous cancers (Fig. 6H) with significant reduction in poorly differentiated compared to serous (Fig. 6I) suggesting that WNT4 may play a role in human ovarian tumorigenesis. Database analysis of HGSC patient samples revealed upregulation or amplification of several of the pathways upregulated upon loss of PTEN in FTE (Supplementary Table S2).

Figure 6: Loss of PTEN in the oviduct increases WNT non-canonical signaling pathway.

A-B) Heat map of genes involved in non-canonical and canonical WNT signaling obtained from RNAseq data. C) qPCR was performed to show mRNA levels in MOE:PTEN^shRNA, compared to MOE:SCR^shRNA (*p<0.05; ***p<0.001). D-E) Western blot of lysates from MOE:PTEN^shRNA and MOE:SCR^shRNA blotted with WNT4 and DKK3 antibodies. F) IHC of tissues from control PAX8^CRE/+ and PAX8^CRE/+PTEN^fl/fl were assessed for WNT4 and DKK3 protein expression. G) WNT4 and FAK protein levels in a panel of human ovarian cancer lines by Western blot. H) WNT4 levels in normal human fallopian tube tissues (hFT) or tissues from patients with ovarian cancer (Tissue micro array=TMA) were detected using IHC. I) TMA were scored based on staining intensity: 0=negative; 1=faint; 2=intermediate/moderate; 3=strong/diffuse. SP=serous papillary; EC=endometrioid carcinoma; CC=clear cells; MC=mucinous carcinoma; PD=poorly differentiated. (*p<0.05).

Loss of PTEN did not increase β-catenin nuclear localization in the MOE:PTEN^shRNA cells (Supplementary Fig. S3D–E) or in tumors, where β-catenin was expressed at the membrane (Supplementary Fig. S3F). β-catenin activation was also measured using the TOP-FLASH reporter assay and did not show any significant difference in the MOE:PTEN^shRNA cells compared to MOE:SCR^shRNA (Supplementary Fig. S3C). These results indicate that canonical WNT signaling, via β-catenin, does not mediate the migratory effect of PTEN loss, but suggest that non-canonical signals might drive migration in the model.

Loss of PTEN induced migration and invasion of the ovary through non-canonical WNT signaling

The non-canonical WNT pathway has been shown to be involved in migration (27, 28), whereas the WNT canonical pathway through β-catenin has been shown to be involved in tumor initiation but not migration nor invasion (33). Therefore, WNT4 was tested for its ability to mediate migration in the MOE:PTEN^shRNA-GFP. LGK-974, an inhibitor of the WNT canonical and non-canonical pathways, inhibited loss of PTEN^shRNA-induced invasion in a Boyden chamber assay (Fig. 7A–B) and PTEN^shRNA-induced migration (Supplementary Fig. S3I), whereas β-catenin siRNA did not reduce PTEN^shRNA-induced increased migration (Supplementary Fig. S3G). In addition, LGK-974 significantly reduced invasion of MOE:PTEN^shRNA-GFP through ovulation mimetic ovaries (Fig. 7C–D). However PRI724, a specific β-catenin inhibitor, did not reduce PTEN^shRNA-induced invasion in a Boyden chamber assay (Fig. 7B) or invasion into the ovaries (Fig. 7E). Finally, siRNA targeting WNT4 reduced invasion of MOE:PTEN^shRNA-GFP cells into the ovary, whereas β-catenin siRNA did not (Fig. 7F). These results indicate that loss of PTEN-induced migration and invasion is at least partially mediated by WNT4 upregulation. Since non-canonical WNT signaling may be mediated by Rac1, we assessed whether Rac1 the inhibitor NSC27366 and show that it reduced migration (Fig. 7J), however it did not significantly reduced invasion of the ovary (Fig. 7C), suggesting that more investigation may be required to determine its role in loss of PTEN increased migration.

Figure 7: Non-canonical WNT4 signaling mediates invasion to the ovary.

A-B) MOE cells stably transfected with PTEN^shRNA, seeded in a Boyden chamber, were treated with 1μM LGK-974 (WNT inhibitor), PRI724 (canonical WNT inhibitor) or DMSO and invading cells counted after 24 hours (*p<0.05). C-D) Intact or ovulation mimetic treated ovaries from healthy mice were incubated with PTEN^shRNA-GFP treated with 1 μM LGK-974 or DMSO for 24 hours. GFP positive cells were counted from more than 3 independent experiments and averaged, (ns p>0.05; ***p<0.001). E) Ovulation mimetic ovaries were incubated with PTEN^shRNA-GFP plus DMSO or 1 μM LGK-974, 1 μM PRI724, 10uM NSC23766 (Rac1 inhibitor), or 10uM MK2206 (AKT inhibitor) for 24 hours. GFP positive cells were counted from 3 independent experiments and averaged (**p<0.01). F) Ovulation mimetic ovaries were incubated with PTEN^shRNA-GFP transfected with control siRNA (LUC^esiRNA), siRNA targeting WNT4 (WNT4^esiRNA) and siRNA targeting β-catenin (β-cat^esiRNA). GFP positive cells were counted from 3 independent experiments and averaged (*p<0.05). G) Ovulation mimetic ovaries were incubated with OVCAR8^RFP plus DMSO or SML0837 (FAK inhibitor) for 24 hours. RFP positive cells were counted from more than 3 independent experiments and averaged (*p<0.05). H) Ovulation mimetic ovaries were incubated with OVCAR4^RFP plus DMSO or inhibitor of FAK, SML0837, for 24 hours. RFP positive cells were counted from more than 3 independent experiments and averaged (*p<0.05; ***p<0.001). I) Ovulation mimetic ovaries from healthy mice were incubated with MOE:PTEN^shRNA-GFP plus DMSO or inhibitor of FAK, SML0837, for 24 hours. GFP positive cells were counted from 3 independent experiments and averaged (****p<0.0001).

Inhibiting WNT non-canonical signaling blocks human ovarian cancer lines migration and invasion

To test whether the WNT inhibitor, LGK-974, inhibited the invasive capability of human ovarian cancer cells, OVCAR4 and OVCAR8 were infected with lentiviral particles expressing RFP and incubated with healthy ovaries in the ex vivo ovarian invasion assay. The WNT inhibitor LGK-974 reduced the human ovarian cancer lines invasion into the ovary (Fig. 7G–H), suggesting that WNT non-canonical signaling may be required for these cells to colonize the ovary. The migratory effects of WNT4 in Drosophila M. have been attributed to activation of FAK, which is also downstream of loss of PTEN (28, 34, 35). FAK is overexpressed in ovarian cancer and associated with poor prognosis (34, 36). Database analysis of TCGA revealed amplification of PTK2 (gene encoding for FAK) in about 40% of samples (Supplementary Fig. S4D). Furthermore, alterations of PTK2 correlate with increased FAK protein levels (Supplementary Fig. S4E) and reduced survival (Supplementary Fig. S4F). Therefore, it was relevant to test whether inhibiting FAK would block loss of PTEN-induced migration. A FAK inhibitor, SML0837, blocked the invasive capability of PTEN-depleted MOE (Fig. 7I), and in OVCAR4 cells (Fig. 7H), whereas the AKT inhibitor, MK2206, did not significantly change invasion (Fig. 7E).

DISCUSSION

Many ovarian cancers likely originate in the fallopian tube epithelium and then colonize the ovary. A FTE-origin for some HGSC is well established and there is increasing evidence that some endometrioid cancers may also originate in the FTE (10, 11). However, the step-wise sequence of events that coordinate FTE cell-transformation and invasion of the ovary remains poorly defined and these steps are pivotal to early detection and new treatment options. Most of the previous animal models for FTE-derived ovarian cancer required loss of PTEN to induce tumorigenesis, but all of these models were engineered with multiple genetic modifications (11, 20, 37). More interestingly, PTEN knockdown alone in the ovarian surface epithelium has been reported to be insufficient to induce tumorigenesis unless in combination with KRAS mutation or APC loss (38, 39). Also, PTEN loss in the FTE stroma alone is not sufficient to drive tumorigenesis, and required the deletion of DICER (20). Furthermore, a model of PTEN and APC deletion, driven by the OVGP1 promoter, generated endometrioid carcinoma from FTE (11). In the current study, homozygous deletion of PTEN alone driven by PAX8 in the FTE was sufficient to generate endometrioid and serous borderline cancer. A previous model established that FTE can give rise to endometrioid cancer (11), but in that model it required PTEN and APC deletion. Our model also formed endometrial carcinoma due to expression of PAX8^CRE/+ in the uterus, and therefore PTEN knockout in that tissue. However, previous models of loss of PTEN-induced endometrial carcinoma have not reported FT or ovarian involvement (40). Furthermore, in previous models of FTE-derived HGSC, hysterectomy did not block cancer formation whereas salpingectomy did (37). Homozygous loss of PTEN generated serous borderline tumors providing the first animal model of FTE-derived serous borderline tumors. Papillary tubal hyperplasia has been reported in patients at high risk for developing ovarian cancer (41, 42) and in low-grade serous borderline tumors (41). Interestingly, mice with PTEN heterozygous deletion had normal ovaries but FTE lesions that resembled human hyperplasia, providing a unique model for monitoring molecular and hormonal changes required for tumor development. LGSC may originate from SBT, however, SBT resembles HGSC by presenting with loss of PAX2 (43), suggesting that SBT may represent a precursor for HGSC. Therefore, our model provides a novel tool to study early events of hyperplasia, endometrioid, and serous borderline tumors from the FTE.

RNAseq of MOE PTEN^shRNA cells revealed enhanced expression of players in the WNT non-canonical pathway, such as WNT4, Sdc½ and DKK3. WNT4 was required for PTEN-induced migration, whereas p-catenin activation was not. WNT4 is of fundamental importance for Müllerian tract development and is critical for FTE cell migration during oncogenesis (27). However, WNT4’s role in reproductive cancers is unclear. In addition, a spontaneous model of oviductal cell transformation also expressed enhanced levels of WNT4, supporting its role in oviductal derived ovarian cancer (29). An inhibitor of the WNT canonical and non-canonical pathways, LGK-974, as well as knockdown of WNT4 by siRNA, inhibited loss of PTEN induced migration, whereas inhibitors of the canonical pathway alone (PRI724) did not. This suggests that WNT4 mediates loss of PTEN-induced migration in a β-catenin independent fashion, as corroborated by the lack of β-catenin activation upon loss of PTEN in the RNAseq data, activity assays, and in tumors. Other studies have reported β-catenin nuclear localization in HGSC and it is possible that β-catenin is activated during metastatic progression. Our results are consistent with reports supporting the role of WNT4 in migration (28). WNT4 was expressed in human ovarian cancer lines, especially in OVCAR8 and OVCAR4, and in several cancer histotypes, as revealed by the TMA analysis using IHC. Even though these lines do not present with genomic deletion or mutation of PTEN, they express WNT4 and whether PTEN expression is regulated at the protein levels is not known. Further experiments using human fallopian tube cells as controls are required to determine PTEN protein levels in HGSC cell lines. LGK-974 also inhibited migration in the human ovarian cancer lines OVCAR4 and OVCAR8. LGK-974 has been previously studied in primary ovarian cancer cells derived from ascites and was effective at potentiating carboplatin-induced cell death (44). However, our studies indicate that blocking WNT4 with LGK-974 also has a role in inhibiting invasion or metastasis, particularly early metastasis to the ovary. Both loss of PTEN and WNT4 have been shown to regulate migration through activation of FAK (28). Inhibition of FAK was able to inhibit invasion of oviductal cells and the human ovarian cancer cell OVCAR4 but not OVCAR8, suggesting that OVCAR8 cells may have acquired alterations that make them resistant to treatment. While loss of PTEN increases AKT activation (9), an AKT inhibitor did not reduce invasion of MOE:PTEN^shRNA to the ovary, suggesting that AKT may be important for survival, but not for colonization of the ovary. In support of this finding, our previous studies confirmed that myristoylated AKT was unable to induce tumorigenesis in oviductal cells (9).

Loss of PTEN in the FTE, in both the transgenic mouse model and in the oviductal cell line model, demonstrated a significant reduction of PAX2 (30, 43) and Fig. S1E. PAX2 loss is thought to be one of the earliest events in the development of HGSCs from the FTE, and it is also lost in ovarian serous borderline tumors (43). Since the cell model harbors shRNA that reduces PTEN, while the mouse model is a full deletion, these data suggest that some reduction in PTEN, not necessarily complete loss, is sufficient to reduce PAX2 levels. In addition, stabilization of p53 protein visualized by IHC was observed in the PAX8^CRE/+PTEN^fl/fl model and in the MOE:PTEN^shRNA cell line (9). However, no downregulation of p53 mRNA or mutation was detected by RNAseq suggesting that the stabilization maybe through inhibition of wild-type p53 binding to MDM2 or other mechanisms, which require further investigation. A previous murine model demonstrated that mutation of R270 was not sufficient to generate stabilization of p53 in the FTE, suggesting that other alterations may be required (45). PAX2 and PAX8 regulate WNT4 during development (46, 47); however, their mechanism of regulation of WNT4 during tumorigenesis has not been studied. In the allograft model of loss of PTEN-induced tumorigenesis, the cells produced multiple tumors in the peritoneal space, whereas in the transgenic model by 7 months of age, there was no evidence of metastasis to the peritoneum. This suggests that colonization from the FT to the ovary likely occurs first, as also suggested recently by Coffman and Buckanovich (22) and that further alterations beyond loss of PTEN are required for metastatic progression. One possible alteration could be mutation of p53 that has been found in 96% of HGSC or as was shown in other models PTEN in combination with loss of Rb, NF1, or BRCA (21, 37).

Overall, our findings report for the first time that loss of PTEN alone from FTE is sufficient to induce tubal carcinoma and subsequent ovarian involvement. Furthermore, homozygous loss of PTEN generated serous borderline tumors from FTE and endometrioid carcinoma, whereas, heterozygous loss of PTEN generated hyperplasia, resembling human precursor lesions. Upregulation of the non-canonical WNT4 ligand and its signaling mediated invasion of the ovary, providing a potential novel role for inhibitors of non-canonical WNT pathway in preventing human ovarian cancer early progression.

MATERIAL AND METODS

Cell culture

MOE and HGSC cell lines (OVCAR3, OVCAR4, OVCAR8, OVCAR432, KURAMOCHI, OVSAHO) were cultured as previously described (48, 49). All lines were authenticated by STR analysis at the University of Illinois at Chicago in May 2015. MOE cell lines (originally provided by Dr. Barbara Vanderhyden, U. of Ottawa) were transfected using TransIT LT1™ with plasmids containing the gene of interest as described previously (9). Cells expressing the shRNA #2 targeting PTEN and the new control shRNA targeting GFP was generated using shRNA from Sigma (respectively TRCN0000322487 and SHC005). siRNAs targeting WNT4 (Sigma, EMU047101), β-catenin (Sigma, EMU047621) and control (Sigma, EHURLUC) were transfected using TransIT LT1™. WNT inhibitors LGK-974 (14072) and PRI724 (205865) were obtained respectively from Cayman and MedKoo. The FAK inhibitor (SML0837) was obtained from Fisher. Rac1 inhibitor (NSC23766) and Akt inhibitor (MK2206) were obtained from Cayman.

Animals

All animals were treated in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals and the established Institutional Animal Use and Care protocol at the University of Illinois at Chicago. The C57b/6 LoxP-PTEN-LoxP was obtained from MMHCC (Mouse Models of Human Cancer Consortium) were bred with mice expressing CRE-recombinase under control of Pax8 promoter from Research Institute of Molecular of Pathology, Vienna, Dr. Bohr-Gasse (50). For the pathological analysis a minimum of 3–4 mice per genotype were used.

ß-galactosidase staining

The β-Galactosidase Reporter Gene Staining Kit was adapted from (Sigma GALS). Organs were washed 2 times with 1 mL 1X PBS, fixed and incubated for 30 min at RT on orbital rocker. After wash with PBS, organs were incubated with 1 mL of X-gal staining solution I at 37°C in humidified incubator overnight for 18 hours. Organs were then processed as for IHC.

PCR reaction to assess recombination and genotyping

PCR analysis for intact PTEN was performed using primer 1 (5’ AAAAGTTCCCCTGCTGATGATTTGT-3’) and primer 2 (5’-TGTTTTTGACCAATTAAAGTAGGCTGTG-3’). To detect the recombined allele, primers 1 and 3 (5’-CCCCCAAGT- CAATTGTTAGGTCTGT-3’) were used. PCR conditions were 35 cycles (30 sec at 95°C, 1 min at 55°C, and 1 min at 72°C) using HotStarTaq Master Mix (Qiagen, Valencia, California, United States), primer (0.25 lM), and DNA (50 ng). For genotyping, the following primers were used: CRE 5’-AGCTGGCCCAAATGTTGCTGG-3’, Pax8 exon 3 5’-CCCTCCTAGTTGATTCAGCCC-3’, Pax8 intron 2 TCTCCACTCCAACATGTCTGC. PCR conditions for CRE: 95°C 5min, 95C 30 sec, 56 30sec, 72 °C 45sec, 72 5min. Floxed PTEN primers: P1 5’-AAAAGTTCCCCTGCTGATGATTTGT-3’, P2 TGTTTTTGACCAATTAAAGTAGGCTGTG.

PCR conditions for PTEN: 95°C 5min, 95°C 30 sec, 55 60sec, 72°C 60sec, 72°C 10min.

RNA isolation, cDNA synthesis and RT-PCR

RNA extraction was performed as described in (51). iScript™ cDNA synthesis kit (BioRad, Hercules, CA) and SYBR green (Roche, Madison, WI) were used according to manufacturer’s instructions. All qPCR runs were performed on the ABI ViiA7 (Life Technologies, San Diego, CA). Primers used are in Supplementary Table S4.

Western blot analyses

Cells were lysed in RIPA buffer (50mM Tris pH 7.6, 150mMNaCl, 1% Triton X-100, 0.1% SDS) with protease (Sigma) and phosphatase inhibitors (tablets from Roche). Protein lysate (30μg) was loaded onto a SDS-PAGE and transferred to nitrocellulose membrane. Blots were blocked with 5% milk or BSA in TBS-T and probed at 4°C overnight with primary antibodies (Supplementary Table S3). Anti-rabbit HRP-linked secondary antibody was used for 30 min in blocking buffer. Membranes were incubated and developed as described previously (51).

Immunohistochemistry (IHC)

Reproductive tracts and tumors were embedded in paraffin and prepared for immunohistochemistry or hematoxylin and eosin stain as described in past publications (9, 48). Tissues microarrays (TMAs) were obtained from the cooperative human tissue consortium (CHTN, Charlottesville, VA) and stained as previously described. Tissues were incubated with primary antibodies overnight (Supplementary Table S3). Images were acquired on a Nikon Eclipse E600 microscope using a DS-Ri1 digital camera and NIS Elements software. IHC images were scored based on staining intensity: 0=negative; 1=faint/weak; 2=intermediate/moderate and the percent (%) of positive or nuclear over total is reported for Ki67 and nuclear p53, respectively.

Luciferase assay

Cells were plated in 24 well plate (30,000 per well). The next day cells were transfected with plasmid encoding Top-Flash luciferase driven by TCF-LEF binding sites and plasmid encoding β-galactosidase (100ng each) with 0.5ul of Mirus Transit LT1 per well in 50ul Opti-MEM. After 20 min at RT 52ul of transfection mixture was added dropwise to each well containing cells in 500 ul of complete medium. After 48 hours cells were lysed 120ul of lysis buffer for luciferase activity and β-galactosidase absorbance reading as previously described (52).

Wound healing assay

Cells were seeded at 5 × 10⁴ cells per well and assay was performed after 48 hours as described previously (53, 54). Briefly, scratch was preformed, medium removed and substituted with medium containing vehicle or inhibitors. Pictures were taken right after the scratch and 8 hours later using an AmScope MU900 with Toupview software (AmScope, Irvine, CA).

Boyden chamber invasion assay

Matrigel was thawed out on ice and diluted to 300 ug/ml with serum-free medium. 120 ul of diluted matrigel was added to the 0.8 um insert of the Boyden chamber and incubated at 37C for 1 hr for matrigel to solidify. Excess of matrigel was removed from insert. 500ul of complete medium (containing 10% FBS) was added to 24 well plates at the bottom of the insert. Cells were trypsinized and collected in media with FBS. Cells were washed 2x with serum free MEM and resuspended in serum-free medium. 120 ul of MTEC4 PTEN^shRNA or SCR^shRNA at 1×10⁴ cells/well was added and incubated for 24 hours. Cells from the top of insert were removed with cotton swap. Inserts fixed with 4% PFA for 5 min, permeabilized with 70% methanol for 5 min, stained with 0.2% crystal violet in 10% ethanol for 10 min. Inserts were then rinsed 2x with PBS and dried overnight. Images of each insert were taken using AmScope MU900 with Toupview software (AmScope, Irvine, CA) and invading cells (at the bottom of the insert) were counted in ImageJ.

Ex vivo colonization assay

Ovaries were removed from day 16 or 17 day old CD1 mice and were left intact or cut with a scalpel blade to mimic ovulation (termed ovulation mimetic). Each ovary was incubated with 30,000 fluorescently labeled cells and appropriate inhibitors overnight at 37°C in an orbital shaker at speed 40. The next day ovaries were washed several times to remove non-attached cells, all observable cells were counted, and representative pictures were taken with an AmScope MU900 with Toupview software (AmScope, Irvine, CA).

Statistical analyses

All data are represented as mean ± standard error. Statistical analysis was carried out using Prism software (GraphPad, La Jolla, CA). All conditions were tested in three replicates in at least triplicate experiments. Statistical significance was determined by Student’s t -test, or one-way ANOVA with Dunnett’s post-hoc test. *p < 0.05 was considered significant.