Silencing PPA1 inhibits human epithelial ovarian cancer metastasis by suppressing the Wnt/β-catenin signaling pathway

Inorganic pyrophosphatase (PPA1) activity is a key determinant of cellular inorganic pyrophosphate levels, and its expression is correlated with growth of several solid tumors. To investigate this relationship, we first examined PPA1 expression in human epithelial ovarian cancer (EOC) samples, and found that PPA1 was overexpressed in tumors from EOC patients. Higher PPA1 levels correlated with advanced grades, stages, and poor survival in EOC patients. Examination of PPA1 function in EOC revealed that silencing PPA1 inhibited EOC migration, epithelial-mesenchymal transition (EMT), and metastasis in vitro and in vivo. In addition, PPA1 may promote the dephosphorylation and translocation of β-catenin. These results demonstrate that silencing PPA1 inhibits EOC metastasis by suppressing the Wnt/β-catenin signaling pathway. Strategies for downregulating PPA1 may have therapeutic potential for the prevention and treatment of EOC.


INTRODUCTION
Epithelial ovarian cancer (EOC) accounts for more than 90% of ovarian cancers. Despite continued efforts to improve EOC treatment, it is still the fifth leading cause of cancer death among women [1]. The high mortality of EOC patients is mainly attributed to metastasizing to the abdominal viscera. Metastasis is a multistep process that includes epithelial-mesenchymal transition (EMT) of tumor cells, collective invasion, reorganization of cytoskeleton, and microenvironment remodeling [2,3]. Better understanding of the molecular mechanisms involved in EMT is important for identifying novel therapeutic targets and establishing new treatment strategies for EOC.
Inorganic pyrophosphatase (PPA1) has been identified as a soluble cytosolic pyrophosphatase, which was found to be essential for growth and development in the round-worms Ascaris and Caenorhabditis elegans [4,5]. Among mammals, PPA1 has the potential to regulate neurite growth via JNK dephosphorylation in mouse neuroblastoma cells [6], as well as the potential to induce type I collagen synthesis and stimulate calcification by osteoblasts [7]. Increases in the expression and activity of PPA1 in rat and mouse livers are correlated with aging [8,9].

Research Paper
carcinoma [14], breast cancer [15], and ovarian cancer [16]. Sang et al. found PPA1 overexpression correlated with tumor progression and poor survival of patients. They also found that PPA1 could inhibit the migration and invasion of gastric cancer cells in vitro. Their findings indicate that PPA1 might be a useful marker for gastric cancer metastasis and progression [17]. In previous studies, we performed a proteomic analysis in ovarian cancer tissues to identify proteins that were differentially expressed in ovarian cancer tissues compared to normal ovarian tissues. We identified PPA1 as a potential EOC marker [18].
We hypothesized that PPA1 is correlated with metastasis and a poor EOC prognosis. In this study, PPA1 expression levels were evaluated in EOC tissues from 55 EOC patients and functional experiments were performed to confirm the metastatic features of PPA1. Finally, the molecular role of PPA1 in regulating Wnt/β-cateninassociated EMT signaling was explored.

The expression of PPA1 in EOC patients
To explore the clinical significance of PPA1 in EOC, the expression of PPA1 was detected in frozen primary EOC samples (n=26) and normal ovarian tissues (n=8) using qRT-PCR analysis ( Figure 1A). PPA1 expression was upregulated in EOC tissues compared to PPA1 expression in normal ovarian tissues (p<0.05). The protein levels of PPA1 were also measured in frozen EOC tissues (n=3) and normal ovarian tissues (n=3) using Western blots. PPA1 was also upregulated in EOC tissues compared to normal ovarian tissues ( Figure 1B).
IHC was performed to examine PPA1 expression in EOC (n=55), borderline tumors (n=20), benign tumors (n=24) and normal ovarian tissues (n=23). We found that PPA1, which was expressed predominantly within cytoplasm, was generally weak or negative in both normal ovarian and benign tumor tissues. In contrast, strong PPA1 expression was observed in both the borderline ovarian tumors and EOC. Compared to normal ovarian tissues and benign tumor tissues, PPA1 was upregulated in EOC and borderline tumors (p<0.05). Compared to borderline epithelium tumors, PPA1 was significantly in EOC tissues (p<0.05). In short, PPA1 expression was found to be positively correlated with increased tumor malignancy (p<0.01) (Table 1, Figure 1C, and Supplementary Figure 1).

PPA1 expression and clinical pathological characteristics
The relationship between PPA1 expression and clinical pathological variables was analyzed using the χ 2 method or Fisher's exact test (Table 2). Patients in stage II (n=7), stage III (n=17), and stage IV (n=7) exhibited stronger PPA1 expression than patients in stage I (n=24) (p=0.011; Figure 1D and 1E). When investigating PPA1 expression and histological grades, it was found that PPA1 has the potential to inhibit EOC differentiation (G1=16, G2=13, G3=26; p=0.037; Figure 1F), suggesting that PPA1 contributes to the clinical progression of human EOC. There was no correlation observed between PPA1 expression and pathologic type (p=0.658, Supplementary Figure 2) or age (p=0.098).
Follow-up assessments with patients who underwent surgery at the Tianjin Center Hospital of Gynecology Obstetrics and the Tianjin First Center Hospital from 2005 to 2009 were conducted until the end of 2014 in this study. Kaplan-Meier curve analysis indicated that patients with strong PPA1 expression (n=35) tended to have a poor prognosis compared to patients with weak PPA1 expression (n=20; p=0.0042; Figure 1G).

PPA1 in EOC cell line migration in vitro
According to our clinical observations, PPA1 may promote EOC metastasis. Five EOC cell lines, IGROV1, A2780, SKOV3, ES2, HEY, and the normal ovarian epithelium cell line IOSE80 were chosen for the next functional investigation. In EOC cell lines, the expression of PPA1 was examined with western blots. PPA1 expression in ES2 and SKOV3 was increased slightly compared to IOSE80 (Figure 2A). To further investigate the role of PPA1 in tumor malignancy, 2 shRNAs targeting PPA1 were transfected into the PPA1 high-expression EOC cell lines ES2 and SKOV3 via a lentivirus system. The knockdown efficiency of each shRNA was validated by qRT-PCR ( Figure 2B) and western blots ( Figure  2C). After PPA1 knockdown, SKOV3 cells lost their mesenchymal-like morphology with a spindle shape, and acquired an epithelial appearance ( Figure 2D).
The effects of PPA1 in cell proliferation, migration, and EMT were investigated. We found PPA1 failed to affect cell proliferation (Supplementary Figure 3). The results of the transwell assay ( Figure 2E and 2F) and wound healing assay ( Figure 2G, 2H, 2I and 2J) suggested that downregulation of PPA1 decreased migration ability compared to control cells for both the ES2 and SKOV3 cell lines.

Expression of EMT-associated markers after silencing PPA1
Using immunofluorescent staining, we detected the levels of EMT-associated proteins α-SMA and Vimentin in the SKOV3 and ES2 cell lines. The mesenchymespecific markers α-SMA and Vimentin decreased in cells with silenced PPA1 (Figure 3A-3D). Western blot analysis showed that the expression of the epithelium-specific marker E-cadherin was increased, whereas mesenchymal markers N-cadherin and Vimentin were decreased in ES2-shPPA1 cells compared to control cells ( Figure 3E). Increased E-cadherin expression and decreased Vimentin www.impactjournals.com/oncotarget expression was also observed in SKOV3 cells. However, N-cadherin expression could not be detected in SKOV3 cell lines ( Figure 3F).
We examined the expression of PPA1, E-cadherin, and Vimentin in 55 human EOC samples. EOC tissues with high PPA1 levels appeared to have high Vimentin expression and low E-cadherin expression. Conversely, EOC tissues with low PPA1 expression exhibited high E-cadherin expression and low Vimentin expression ( Figure 4; Table 3). Pearson correlation analysis was employed to investigate the correlation between PPA1 and EMT in EOC. The results indicated that PPA1 expression was negatively associated with E-cadherin expression (R=−0.121, p=0.017), and positively associated with Vimentin expression (R=0.362, p=0.06). Therefore, PPA1 was closely correlated with decreasing epithelial traits and increasing mesenchymal characteristics.

Silencing of PPA1 inhibited EMT by preventing β-catenin to translocate to the nucleus
As previously reported, the Wnt/β-catenin signaling pathway promotes EMT in cancer development [19,20]. To explore how PPA1 promotes EMT in EOC cells, we  performed western blotting and immunofluorescent staining to examine β-catenin expression in total cell, nucleus, and cytoplasm lysates. Western blots showed that total β-catenin expression was decreased in PPA1silenced cells compared to control cells ( Figure 5A). As predicted, nuclear β-catenin expression was downregulated in PPA1-silenced cells. Cytoplasmic β-catenin levels were not altered in SKOV3 cells, and decreased slightly in ES2 cells with PPA1 knockdown ( Figure 5B and 5C). Immunofluorescence staining results showed less nuclear translocation of β-catenin in PPA1 knockdown cells compared to control cells ( Figure 5D-5G). TopFlash/ FOPFlash reporter genes were used to measure the transactivation of TCF after silencing PPA1. Decreased luciferase activity was detected from the TopFlash reporter for the PPA1-silenced group compared to the scrambled control group ( Figure 5H). To clarify PPA1's effect on the Wnt/β-catenin pathway, we established SKOV3-PPA1 cells which overexpressed PPA1 in the SKOV3 cell line ( Figure 5I). The SKOV3-PPA1 cells were treated with a series of Wnt/β-catenin-specific inhibitors. IWP-2 is a Wnt-specific inhibitor that prevents palmitoylation of Wnt protein, therefore blocking Wnt secretion and activity. IWR-1-endo inhibits Wnt-induced accumulation of β-catenin through stabilization of the destruction complex member AXIN2. KY021111 selectively inhibits the Wnt/β-catenin pathway by targeting downstream GSK3β. The GSK3β inhibitor blocked the PPA1-induced β-catenin translocation. We infer that PPA1 may promote β-catenin dephosphorylation ( Figure 5J).

Silencing PPA1-inhibited metastasis of EOC in the syngeneic mouse model
To validate the correlation between PPA1 expression and tumor cell metastasis, a xenografted tumor model was performed. NOD/SCID mice were intraperitoneally (i.p.) injected with SKOV3 shCtrl (n=4) or SKOV3 shPPA1-2 cells (n=4). Forty-five days after injection, primary and metastatic tumors were measured based on the luminescence of luciferase. Photon counts decreased in the metastatic site of shPPA1-2, but did not change in primary tumors ( Figure 6A and 6B). Mice injected with shPPA1-2 cells showed less ascetic fluid accumulation in the abdominopelvic cavity compared to control animals (data not shown). Three of the mice injected with SKOV3 shCtrl cells developed metastases (multiple small metastatic nodules) under the diaphragm and liver surface, whereas none of the mice injected with the shPPA1-2 cells exhibited metastatic tumors ( Figure 6C). The results indicate that silencing PPA1 can inhibit metastasis of SKOV3 cells in vivo.
We investigated the expression of PPA1, E-cadherin, and Vimentin in xenografted mouse EOC tumor slices. As predicted, the shCtrl group with high PPA1 expression had high levels of Vimentin and low levels of E-cadherin. Conversely, xenografted tumor tissues from the shPPA1-2 group expressing low levels of PPA1 exhibited a strong E-cadherin and weak Vimentin expression profile ( Figure 6D).

DISCUSSION
As a pyrophosphatase driving biosynthetic reactions, PPA1 was found to be upregulated in multiple cancers. From our previous study [18], we identified PPA1 as a biomarker candidate through proteomic analyses, which was confirmed by using tissue microarrays. The present study was designed to evaluate PPA1 expression in different types of human ovarian tumor tissues and normal ovarian tissues to investigate how PPA1 functions.
Based on several results, high PPA1 expression can promote the occurrence and development of ovarian cancer, and PPA1 expression is positively correlated to a poor prognosis for ovarian cancer. First, compared to normal ovarian tissues and benign ovarian tumors, we found that PPA1 was overexpressed  Fourth, tumor metastasis was inhibited after PPA1 was silenced in vivo. Metastasis in ovarian cancer is established by the EMT-driven delamination of ovarian cancer cells from the primary tumor in situ, and subsequent penetration into the surrounding peritoneal cavity [21][22][23]. Based on our results, we showed that PPA1 may promote EMT. First, the EOC cell line SKOV3 gained epithelial characteristics and lost mesenchymal-like properties after PPA1 was silenced. Second, western blot and immunofluorescence revealed that silencing of PPA1 down-regulated the expression of markers associated with mesenchymal cells (e.g., N-cadherin, α-SMA, or Vimentin), while it up-regulated the expression of epithelial cell markers (e.g., E-cadherin). Third, PPA1 expression was negatively associated with E-cadherin expression, but positively correlated with Vimentin expression.
In our study, nuclear β-catenin expression decreased after PPA1 was silenced compared to cells in the control group. We demonstrated that the regulatory effect of PPA1 in EMT in EOC cells might promote the activation of the Wnt/β-catenin signaling pathway and the β-catenin nuclear translocation.
In summary, our study demonstrated that PPA1 was overexpressed in EOC with advanced grades and stages, and high PPA1 expression correlated with poor survival of EOC patients. PPA1 expression increased metastasis by stimulating EMT both in vitro and in vivo. Furthermore, the Wnt/β-catenin signaling pathway promoted this process via translocation of β-catenin. Thus, our data   support PPA1 as a potential marker for evaluating EOC progression and a useful target for EOC therapy.

Clinical samples and cell lines
Formalin-fixed, paraffin-embedded human specimens, including EOC (n=55), borderline tumors (n=20), benign tumors (n=24), and normal ovarian tissues (n=23), were collected from patients who underwent surgery at the Tianjin Central Hospital of Gynecology Obstetrics and the Tianjin First Center Hospital between 2005 and 2009 (n=122). The classification of clinical staging and histologic grading of EOC were determined according to the FIGO 2006 system. Patient ages ranged from 25 to 67. Within the follow-up period of 60 months, 24 patients died of the disease. Among these 122 patients, 34 fresh tissue samples (EOC=26, normal ovarian tissues=8) were snap frozen and stored at -80°C to obtain RNA and protein samples. Normal ovary specimens were sectioned from patients with uterine fibroids during total abdominal hysterectomy and bilateral salpingooophorectomy and were diagnosed by post-operative pathological examination. Approval from the Institutional Research Board at these two hospitals was obtained prior to the study.
Human EOC cell lines (IGROV1, A2780, SKOV3, ES2, HEY) and a human ovarian surface epithelial cell line (IOSE80) were stored at the Nankai University School of Medicine. The cell lines were purchased from ATCC and cultured according to the instructions provided. Human embryonic kidney 293T (HEK293T) cells were cultured in Dulbecco's Modified Eagle Medium (Corning, USA) enriched with 10% FBS, 1% P/S, 2% L-glutamine, 1% Sodium pyruvate and 1% non-essential amino acid. All cells were grown at 37°C and 5% CO2 in a humidified incubator.

Transwell and wound healing assays
Cells from each sample group were seeded into the top chamber of a 24-well plate with 4 μm polyethylene terephthalate membrane inserts (Millipore). The chamber was filled with RPMI-1640 (Corning, USA) containing 10% FBS, and the inserts were filled with 1% FBS RPMI-1640. After 12h incubation, the filter membrane was fixed with 100% methanol and stained with crystal violent for another 12h, then washed with 50% acetic acid. The OD value was measured by microplate reader (Promega). Cell migration was determined by the mean ± SD.
For the wound healing assay, ES2 and SKOV3 cells were plated for 24 hours in a 6-well plate. A scratch was made in the confluent cell layer using a sterile tip. Medium was changed and replaced with RPMI 1640 with 2% FBS. Cellular migration was assessed after 12 h and 24 h.

Luciferase reporter gene assay
A TOPFlash reporter plasmid containing two sets of three copies of the transfection grade T-cell factor (TCF) binding site upstream of the Luciferase open reading frame was used in the assay. A FOPflash plasmid containing mutated TCF binding sites was included as a negative control. TOPFlash and FOPFlash reporter plasmids (Millipore, CA) were individually co-transfected with shRNA (shPPA1-2 and shCtrl) and a plasmid with a Renilla reporter gene into 293T cells. After 48 h, cells were lysed with lysis buffer, and the activity of each reporter genes activity was detected with a Dual-Luciferase® Reporter Assay (Promega, Madison, WI).

Tumor xenograft and bioluminescence imaging
Female NOD/SCID mice, 6-8 weeks old, were kept under Specific Pathogen Free conditions according to Beijing Medical Experimental Animal Care guidelines. The animal experiments were approved by the Institutional Animal Care and Use Committee of Tianjin Nankai University School of Medicine. Ovarian cancer cells (SKOV3) stably transfected with firefly luciferase and shRNA (shPPA1-2 and shCtrl) were generated, and 1.25×10 6 cells were injected intraperitoneally into the NOD/SCID mice. Bioluminescence imaging to determine the fate of transplanted cells in living mice was performed as previously described [26]. Briefly, mice underwent BLI to determine firefly luciferase expression using the in vivo imaging system (IVIS 200) (Xenogen Corporation, Hopkinto, MA). Following anesthesia with 2% isoflurane, mice were injected with D-Luciferin IP (150 mg/kg; Biosynth International, Naperville, IL) and 1 s to 5 min scans were performed to assess firefly luciferase expression.

Statistical analysis
All of the data were presented as mean ± standard deviation. Statistical analyses were conducted using SPSS 17.0 software. A chi-square (χ 2 test) test, Pearson's correlation test and Student's t-test were used. A p < 0.05 was used as the criterion for statistical significance.

Author contributions
LW and PQ were the principal investigators and take primary responsibility for the paper. LW, PQ, RX, HN, and WZ designed the experiments. HN, WZ, YX, ZY, WS, and ZY performed the experiments. YC, YL, and SY analyzed the data. LW, HN, and RX wrote the paper.