RM cells were derived from immortalized mouse ovarian surface epithelial cells transduced with retroviral constructs to achieve ectopic expression of mutant K-Ras (KRAS^G12D) and Myc as previously described (5).

The construction of the RM-PAX2 and RM-pWPI cell lines has been described in detail previously (5). Briefly, the murine Pax2 cDNA (pax2-b variant) was cloned from murine oviduct and inserted into the Not I site of pWPI (Addgene plasmid 12254) to generate a lentivirus expression vector (WPI-Pax2-IRES-eGFP, hereafter pWPI-Pax2). The empty pWPI vector was used as a control. The vector plasmids (pWPI or pWPI-Pax2), packaging plasmid pCMVR8.74 (Addgene plasmid 22036), and the ecotropic envelope expression plasmid, pCAG-Eco (Addgene plasmid 35617) were co-transfected into 293T cells (ATCC^® CRL-3216™) to generate lentivirus. RM cells were infected with lentivirus and then passaged at least 3 times prior to sorting for GFP expression by fluorescence-activated cell sorting (Beckman Coulter, Inc.).

The RM, RM-WPI or RM-PAX2 cells (5×10⁴) were plated in 6-well plates for 72-h growth assays. Cell numbers were counted using a Vi-Cell™ XR cell viability analyzer (Beckman Coulter, Inc.).

Animal experiments were performed in accordance with the Canadian Council on Animal Care's Guidelines for the Care and Use of Animals under a protocol approved by the University of Ottawa's Animal Care Committee. Mice were maintained in a dedicated room for immune-compromised mice (21̊C, 40-60% humidity, 12/12 h light/dark cycle). A commercial rodent diet (2018 Teklad Global 18% Protein Rodent Diet, Harlan Laboratories, Indianapolis, ID, USA) along with acidified water was available ad libitum. Housing, food and water were autoclaved, and all animal manipulations were carried out in a certified ESCO-type A2 BSC hood, following a two-person dirty/clean protocol. Mice were allowed a 1-week acclimation period prior to the initiation of any experimental manipulations.

The RM-WPI or RM-PAX2 cells (1×10⁷ in 500 µl PBS) were injected into the peritoneal cavity of eight 8-week-old SCID mice (The Jackson Laboratory), separately. The mice injected with RM-PAX2 cells intraperitoneally had an average weight of 17.7 g at the time of purchase, and the average weight of RM-WPI group was 18.8 g. No analgesics or anesthetics were administered to the mice for the i.p. injection of cells, as i.p. injections are considered routine procedures, are performed quickly and do not appear to be very painful to the mice. Eight mice were assigned to each group. Disease progression was monitored and the mice were euthanized via CO₂ asphyxiation once a humane endpoint was reached (rapid changes in weight, loss or gain of >5 g compared to the average body weight of control mice of the same age, presence of respiratory distress, a palpable mass or abdominal distention that impairs mobility). For euthanasia, the flow rate of medical-grade CO₂ was 1.5 litres/min to deliver a 30% change in the chamber volume/min. The mice were kept in the CO₂ chamber for 3-5 min until they did not respond to pain stimuli (pinching tails and paws). Tumors were then excised, imaged by fluorescent microscopy (EVOS Imaging Systems, Thermo Fisher Scientific), fixed in 10% buffered formalin for 24 h, paraffin-embedded and sectioned at a thickness of 5 µm for histological analysis. Alternatively, tumor samples were snap-frozen in liquid nitrogen for protein analysis.

The tissue microarrays (TMA) consisted of specimens from 152 patients who were diagnosed with epithelial ovarian cancer and treated from 2005 to 2013. Among these, 131 specimens were generous gifts from the Affiliated Qilu Hospital of Shandong University and the remaining were from the Guangxi Medical University Cancer Hospital. The pathological diagnosis of specimens in the TMAs was reviewed by senior pathologists of that same institution. Three tissue samples used for western blot analysis were from patients who were diagnosed with high-grade serous ovarian cancer and followed by radical hysterectomy at the Guangxi Medical University Cancer Hospital, and no malignant lesions in ampulla of oviduct was proved by pathology. Patient informed consents were obtained prior to the experiments and the protocols were approved by the Ethics Review Committees of both the Affiliated Qilu Hospital of Shangdong University and the Guangxi Medical University Cancer Hospital.

The ages of the 152 patients with ovarian serous carcinoma were 33 to 79 years, with an average age of 54.9±9.9 years. The median time of overall survival (OS) and progression-free survival (PFS) was 52.0 and 28.0 months, respectively. The patient clinical information is presented in Table I.

All patients were followed-up eacg month in the first year following surgery, every 3 months in the second year, every 6 months in the third year, and once every 3 years thereafter. The endpoint was the appearance of tumor progression (including recurrence not controlled with treatment, or death). The follow-up time ranged from 5 to 129 months (average, 43±25 months). PFS was determined as the period from the day of surgery to the time of disease recurrence. Those who did not recur and survived beyond April 1, 2016 were recorded as censored data.

The expression of PAX2, RAS, acetyl-CoA acyltransferase 2 (ACAA2) and pancreatic lipase (PNLIP) was determined by immunohisto-chemistry using the Immunohistochemistry Envision HRP kit (cat. no. KIT-5004, Maixin Biotechnologies), rabbit-anti-PAX2 antibody (cat. no. ab150391, Abcam), rabbit-anti RAS antibody [Ras (D2C1), rabbit mAb, cat. no. #8955, Cell Signaling Technology], rabbit-anti ACAA2 antibody (cat. no. PAS78709, Invitrogen; Thermo Fisher Scientific) and rabbit-anti PNLIP (cat. no. PA5-80956, Invitrogen; Thermo Fisher Scientific). Optimal dilution ratio of primary antibodies was 1:100.

The percentage of positive cells in each specimen and the staining intensity were independently evaluated and scored by two senior pathologists. When scored for the percentage of positive cells, '0' stands for no positive cells, '1' for 1-25, '2' for 26-50, '3' for 51-75 and '4' for 76-100%. When scored for staining intensity, '0' stands for no coloring, '1' for pale yellow, '2' for brown and '3' for deep brown. The final scores were the multiple of these two scores. For PAX2, the total score 0 was defined as negative, ≤6 was weak expression and >6 was strong expression. For RAS, 0 was defined as negative, >0 was positive; for ACAA2, 0 was defined as negative, <5 was low expression, and ≥5 was high expression; for PNLIP, 0 was defined as negative, <5 was low expression, and ≥6 was high expression.

Tumor tissues from the murine RM model expressing either PAX2 or not were ground under liquid nitrogen and the protein was extracted using the ProteoExtract^® Complete Mammalian Proteome Extraction kit (cat. no. 539779, Merck Millipore) and quantified using the Pierce BCA Protein Assay kit (cat. no. 23227, Pierce; Thermo Fisher Scientific). Protein samples were then reduced, alkylated, digested and labeled with iTRAQ reagents according to the recommended protocol (iTraq Reagent 8 plex buffer kit, cat. no. P/N4381664; iTraq Reagent 8 plex Multiplex kit, cat. no. P/N4381663; Applied Biosystems). The samples were labeled as follows: RM-PAX2, iTraq reagents 117 and 118; RM-WPI iTraq reagents 119 and 121. Following iTRAQ labeling, samples were fractionated by HPLC and analyzed by high-resolution LC-MS/MS. Quantitative global proteome analysis was performed in the PTM Biolabs (https://www.ptmbiolabs.com/). Bioinformatics analysis was carried out to annotate quantifiable targets by protein annotation, functional classification, functional enrichment, functional enrichment-based cluster analysis, etc.

Gene Ontology (GO) annotation proteome was derived from the UniProt-GOA database (http://www.ebi.ac.uk/GOA/). Identified protein IDs were first converted to UniProt ID and then mapped to GO IDs by protein ID. If some identified proteins were not annotated by the UniProt-GOA database, InterProScan software (http://www.ebi.ac.uk/interpro/) was used to determine the GO function based on a protein sequence alignment method. Proteins were then classified by the GO annotation based on 3 categories: Biological process, cellular component and molecular function.

Protein extracts were prepared by cell lysis using RIPA Buffer. Protein concentrations were quantified using the Quick Start Bradford Protein assay kit (Bio-Rad). Proteins (40 µg) were separated on 4-12% Bis-Tris gels (Invitrogen; Thermo Fisher Scientific) and transferred to PVDF membranes. Western blot analysis was performed using primary antibodies from Abcam as follows: Rabbit anti-PAX2 (ab150391), fatty acid binding protein 4 (FABP4; ab92501), LIPA (ab219113), PNLIP (ab198181), ACAA2 (ab237540) at a 1:1,000 dilution and the secondary antibody reagent, anti-rabbit DAKO EnVision-system-HRP solution (cat. no. K4002, Dako Cytomation). Immuno-reactive bands were visualized using an ECL western blotting detection reagent (GE Healthcare) and the Syngene Bio-Imaging System program (PerkinElmer).

The online meta-analysis tool 'The Kaplan Meier plotter' (http://kmplot.com/analysis/, accessed 22 December 2018) was used to assess the effects of gene expression on survival using data for 1,001 patients with stage III-IV serous ovarian cancer. The cut-off value for high or low expression was automatically determined by the Kaplan Meier plotter. The follow-up threshold was set at 60 months. The GEO datasets and TCGA datasets, as well as the bioinformatics processing used by this online tool were previously described (13,14). Biased assays and assays with a false discovery rate (FDR) of >20% were excluded.

On the basis of the number of conditions tested, statistical significance was determined by the t-test, ANOVA (Tukey's post-test), or log-rank test (Kaplan-Meier), performed using GraphPad Prism software (GraphPad Software, Inc.) or SPSS20.0 (IBM).

Previously, the expression of PAX2 in the RM model of ovarian cancer was shown to enhance cell proliferation (5). The expression of PAX2 in the RM-PAX2 cell line was shown in a previous study [please see Fig. 4A in the study by Alhujaily et al (5)]. In this study, the cell proliferation assay confirmed that PAX2 expression led to an enhanced proliferation (Fig. 1A).

To confirm the effects of PAX2 on tumorigenesis in vivo, the RM-WPI and RM-PAX2 cells were injected intraperitoneally into SCID mice. As observed under a fluorescence microscope, both RM-WPI and RM-PAX2 cells formed tumors. Compared to the RM-WPI cells, the RM-PAX2 cells seemed to form larger tumors and spread throughout the peritoneum, to the pancreas, liver, intestine, diaphragm, uterus and ovaries (Fig. 1B). As has been previously demonstrated (5), the survival time was markedly decreased in SCID mice injected with RM-PAX2 cells compared to those injected with RM-WPI cells [median survival, 11 vs. 16 days; please see Fig. 4E in the study by Alhujaily et al (5)].

Although the role of PAX2 in tumorigenesis remains undetermined, its expression was found to increase the tumor burden and shorten the survival of mice injected with RM cells. This result led us to investigate its role in ovarian cancer progression. First, a cohort of 152 epithelial ovarian cancer cases was identified. The patients were separated into the PAX2-positive (PAX2⁺, including both weak and strong expression) and PAX2-negative (PAX2^-) groups according to the expression of PAX2 examined by immunohistochemistry (Fig. 2A). In agreement with the findings of this study on the RM model, the PFS of the PAX2⁺ patients was significantly reduced compared to that of the PAX2^- patients (median survival of 24.0 vs. 75.0 months; Fig. 2B). Using the online meta-analysis tool, The Kaplan Meier plotter revealed that a high expression of PAX2 were associated with the shortened PFS of patients with stage III-IV serous ovarian cancer (Total: 16.1 vs. 19.0 months; TP53 mutant subgroup: 17.3 vs. 19.0 months; TP53 wild-type subgroup: 14.8 vs. 33.3 months, all P<0.05, Fig. 2C).

To further elucidate the mechanisms through which PAX2 affects ovarian cancer progression, iTRAQ proteomic technology was exploited to identify differentially expressed proteins between RM-PAX2 and RM-WPI tumors (Fig. 3A). The expression of PAX2 in these tissue samples was confirmed by immunohistochemistry (Fig. 3B). In total 2,964 proteins were identified from the mouse tumors, of which 2,016 proteins were quantified. When setting quantification ratio thresholds of >1.2 as upregulated and <0.83 as downregulated, in a comparison of protein expression in RM-PAX2 vs. RM tumors, 252 proteins were upregulated and 122 proteins were downregulated (Fig. 3C and D).

KEGG signaling pathway enrichment analysis revealed that the top upregulated signaling pathways in RM-PAX2 xenograft tumors were those that regulate cell metabolism, particularly fatty acid metabolism (Fig. 3E and F). The most downregulated signaling pathways were those that regulate cell cycle, extracellular matrix receptor interaction and cell junctions (data not shown).

The marked upregulation of cell metabolic pathways suggests that PAX2 may promote the reprogramming of tumor cell metabolism to support cell survival. Fatty acid catabolism involves fatty acid activation, translocation to the mitochondria and β-oxidation. In the cytoplasm, fatty acid is activated by Acyl-CoA synthetase (Acsl), which catalyzes fatty acid into fatty acyl-CoA (15). Subsequently, fatty acyl-CoA is transferred into the mitochondria by carnitine acyltransferase I (Cpt1) and carnitine acyltransferase II (Cpt2). Cpt1 is the rate-limiting enzyme of fatty acid metabolism. After fatty acyl-CoA enters the mitochondrial matrix, it decomposes into acetyl-CoA catalyzed by a series of fatty acid β-oxidases, including Echs1, Acads, Acadm, Acadl, Acadvl, Acaa1 and Acaa2 (Fig. 3F). The acetyl-CoA finally enters the tricarboxylic acid cycle and is completely degraded to water, CO₂ and ATP through oxidative phosphorylation by the mitochondrial respiratory chain (Fig. 3F). In the PAX2-overexpressing allograft tumor tissue, almost all enzymes involved in fatty acid metabolism were significantly increased, as shown in Fig. 3F and Table II (all P<0.05).

The above-mentioned results suggested that PAX2 promoted fatty acid metabolism in RM tumors; however, it is unclear whether fatty acid substrates arose through de novo synthesis or by uptake from surrounding adipocytes. Lipases and fatty acid-binding proteins (FABPs) are required for transferring lipids from adipocytes to cancer cells (16). In RM-PAX2 tumors, Pnlip, inactive pancreatic lipase-related protein 1 (Pnliprp1) and pancreatic lipase-related protein 2 (Pnliprp2) were elevated 2.35, 5.45 and 2.47-fold. Murine homologues of FABPs, Fabp1 and Fabp4 were also increased by 2.31- and 2.91-fold, respectively (Table II). Moreover, the expression of fatty acid synthase (Fasn), a key enzyme for fatty acid synthesis, was not altered (Table II).

To confirm the fatty acid metabolic reprogramming discovered by iTRAQ analysis, the expression of FABP4 and a lipase (LIPA) was analyzed in tumor tissue from both the murine RM-PAX2 model and patients with ovarian cancer. The FABP4 and LIPA expression levels were upregulated in the RM-PAX2 tumor tissue compared to the RM-WPI tissue (Fig. 4A). In accordance with the murine model, the expression of FABP4 and LIPA was higher in the ovarian tumors exhibiting a strong PAX2 expression (PAX2⁺⁺⁺) when compared to the tumors with a weak PAX2 expression (PAX2⁺) (Fig. 4B). While FABP4 was expressed at higher levels in tumors expressing high levels of PAX2, FABP4 was also upregulated in the PAX2-low expressing tumors when compared to the normal contralateral fallopian and ovary tissues (Fig. 4A and B). This indicates that even low levels of PAX2 may be sufficient to induce the expression of FABP4, or perhaps other factors in the tumor, in addition to PAX2, leads to the upregulation of FABP4.

The expression of PNLIP and ACAA2 was further examined in tumor tissues of 6 patients with serous ovarian cancer with varying levels of PAX2. The expression of PNLIP and ACAA seemed to increase along with that of PAX2 (Fig. 4C). However, when assessing the potential correlation of PAX2, PNLIP and ACAA2 expression in the TMA cohort in this study, Spearman's correlation coefficients among PAX2, PNLIP and ACAA2 were not found to be significant (data not shown). The expression levels of PAX2, PNLIP and ACAA2 in the TMA cohort were determined by immunohistochemistry, as indicated in the Material and methods section.

Based on the results of IHC, the patients were separated into the high or low expression subgroups. The PFS of ACAA2 in patients with a high expression was reduced compared to those with a low ACAA2 expression patients (median survival of 29.0 vs. 75.0 months; Fig. 5A). The PFS of patients with a high PNLIP expression also exhibited a trend of a reduced PFS compared to those with a low PNLIP expression (median survival of 38.0 vs. 48.0 months, Fig. 5B), although the differences were not statistically significant.

The amplification of RAS signaling was common in epithelial ovarian cancer, but not the KRAS mutation which was used in the RM model in this study. Thus, the expression of RAS was further assessed in the TMA cohort in this study. As a result, 132 cases were RAS-positive, while only 20 was RAS-negative (Table I). In the RAS-positive subgroup, the PAX2⁺ patients exhibited a significantly shorter PFS compared with the PAX2^- patients (median survival of 20.0 vs. 38.0 months, Fig. 5C). A high expression of ACAA2 or PNLIP in patients was also associated with a reduced PFS, although this was not statistically significant (29.0 vs. 56.0 months, 36.0 vs. 53.0 months, respectively, Fig. 5D and E). The RAS-negative subgroup was not analyzed due to the small sample size (n=20).

The dependency of cancer cells on glycolysis, also known as the Warburg effect, was once recognized as the most outstanding feature of tumor cell metabolism (17) and was considered an essential property of most tumor cells, including ovarian cancer (18). It has been shown more recently that enhanced mitochondrial oxidative phosphorylation (OXPHOS) also plays an important role in ovarian cancer (19). In this study, proteomics analysis revealed that a key enzyme for glycolysis, pyruvate kinase (Pkm) was downregulated by PAX2. Other enzymes involved in glycolysis and Glut1, the enzyme responsible for glucose uptake, were not significantly altered (Table III). By contrast, proteins participating in the OXPHOS pathway and components of mitochondrial electron transport chain were all significantly upregulated (Table III) in PAX2-expressing tumor tissues.

Glutamate is also an important energy source for OXPHOS (20). Glutamate is transferred into cells primarily through the amino acid transporter (Slc1a5), but its expression was not affected by PAX2 (Table IV). Glutamate can also be generated through the degradation of amino acids. Given that the valine, leucine and isoleucine degeneration pathway was found to be significantly upregulated (Fig. 3E) and enzymes responsible for glutamate synthesis [alanine aminotransferase (Gpt), glutamate dehydrogenase (Glud1) and glutaryl-CoA dehydrogenase (Gcdh)] were increased (Table IV), it is possible that PAX2 also altered cellular glutamate metabolism.

As the high expression of PAX2 and ACAA2 shortened the PFS of patients with ovarian cancer (Figs. 2B and 5C), whether the upregulation of fatty acid metabolism is also associated with a shorter PFS was then determined. All genes listed in Table II were analyzed. The results of the Kaplan-Meier plotter analysis clearly revealed that a high expression of FABP4 and ACAA2 were associated with the shortened PFS of patients with serous ovarian cancer (Table V); this was observed in all patients regardless of the TP53 status.