Spontaneous Transformation of a p53 and Rb-Defective Human Fallopian Tube Epithelial Cell Line after Long Passage with Features of High-Grade Serous Carcinoma

Ovarian cancer is one of the most lethal gynecological cancers, and 80% are high-grade serous carcinomas (HGSOC). Despite advances in chemotherapy and the development of targeted therapies, the survival rate of HGSOC has only moderately improved. Therefore, a cell model that reflects the pathogenesis and clinical characteristics of this disease is urgently needed. We previously developed a human fallopian tube epithelial cell line (FE25) with p53 and Rb deficiencies. After long-term culture in vitro, cells at high-passage numbers showed spontaneous transformation (FE25L). This study aimed to compare FE25 cells cultured in vitro for low (passage 16–31) and high passages (passage 116–139) to determine whether these cells can serve as an ideal cell model of HGSOC. Compared to the cells at low passage, FE25L cells showed increased cell proliferation, clonogenicity, polyploidy, aneuploidy, cell migration, and invasion. They also showed more resistance to chemotherapy and the ability to grow tumors in xenografts. RNA-seq data also showed upregulation of hypoxia, epithelial-mesenchymal transition (EMT), and the NF-κB pathway in FE25L compared to FE25 cells. qRT-PCR confirmed the upregulation of EMT, cytokines, NF-κB, c-Myc, and the Wnt/β-catenin pathway. Cross-platform comparability found that FE25L cells could be grouped with the other most likely HGSOC lines, such as TYKNU and COV362. In conclusion, FE25L cells showed more aggressive malignant behavior than FE25 cells and hence might serve as a more suitable model for HGSOC research.


Introduction
Epithelial ovarian cancer (EOC) ranks fifth among cancer-related deaths in women. Among EOC, high-grade serous ovarian carcinoma (HGSOC) is the leading cause of death [1,2]. More than 220,000 new ovarian cancer cases happened yearly and most patients with ovarian cancer are diagnosed at a late stage of the disease with a 5-year survival rate of 30-40% only [1].
Current treatment options include debulking surgery and adjuvant chemotherapy. Recently, targeted therapy with poly ADP ribose polymerase inhibitors (PARPi) and antivascular endothelial growth factor (VEGF) has been used for the treatment of ovarian cancer [3]. Despite the introduction of these novel therapies, the survival rate of ovarian cancer has not improved [1] primarily because ovarian cancer is detected at an advanced stage.
EOC is divided into two types, based on complex pathogenic and molecular pathways [4]. Type 1 EOC includes clear cell, endometrioid, low-grade serous, and mucinous carcinomas. Type 2 ovarian cancers include HGSOC, carcinosarcoma, and undifferentiated

FE25L Cells Exhibited Faster Proliferation and More Cells in S and G2/M Phases
To determine the characteristics of FE25L cells, we compared their morphology, proliferation rate, and cell cycle distribution with FE25 cells. FE25L cells exhibited similar morphology to that of FE25 cells, showing a cobblestone-like appearance ( Figure 1A). The proliferation rate of FE25L cells was significantly faster than that of FE25 cells (p < 0.05, on day 3 and day 5, Figure 1B). Quantification of the proportion of cells in different phases of the cell cycle showed that significantly more FE25L cells were in the S and G2/M phases than that in FE25 cells (30% vs. 21%, p < 0.01) ( Figure 1C). Moreover, FE25L showed a higher DNA content or ploidy than FE25 cells.

FE25L Had Higher Migration and Invasion Activity Than FE25 Cells
To determine the migration and invasion capabilities of FE25 and FE25L cells, we performed trans-well migration and invasion assays. Migration was significantly higher in the FE25L group than in the FE25 (nearly 9-fold, p < 0.01, Figure 2A). In the invasion assay, FE25L cells were found to be more invasive than FE25 cells (nearly 6.5-fold, p < 0.01; Figure 2B). These results indicate that FE25L had significantly increased migration and invasion capabilities compared to FE25 cells.

FE25L Formed More Colonies Than FE25 Cells
Clonogenicity activity of the two cells were characterized by colony forming assay on FE25 and FE25L cells and found that there were nearly 7 fold more colonies formed by FE25L than FE25 cells (p < 0.001, Figure 3).    The error bar represents the mean ± SD. All the experiments were performed in triplicates. ** p < 0.01, *** p < 0.001.

Figure 2.
Migration and invasion assay in FE25 (p16) and FE25L (p116) cell lines. (A) Trans-well migration assay. Culture medium 500 μL was added to lower chamber. After cells migrated for 48 h, cells on the membranes were stained with DAPI. Blue immunofluorescence-stained migrated cells were pictured by a microscope (upper panel). Then migrated cells were quantified and compared between the two cell lines (total migrated cells on the membranes) (n = 3 in each group). (B) Trans-well invasion assay. Upper chamber was coated with Matrigel. The other steps of the experiment were the same as described above (n = 3 in each group). ** p < 0.01. Scale bar = 100 μm. The upper panel shows a representative image of the experiment. Migration and invasion assay in FE25 (p16) and FE25L (p116) cell lines. (A) Trans-well migration assay. Culture medium 500 µL was added to lower chamber. After cells migrated for 48 h, cells on the membranes were stained with DAPI. Blue immunofluorescence-stained migrated cells were pictured by a microscope (upper panel). Then migrated cells were quantified and compared between the two cell lines (total migrated cells on the membranes) (n = 3 in each group). (B) Transwell invasion assay. Upper chamber was coated with Matrigel. The other steps of the experiment were the same as described above (n = 3 in each group). ** p < 0.01. Scale bar = 100 µm. The upper panel shows a representative image of the experiment.
Clonogenicity activity of the two cells were characterized by colony forming on FE25 and FE25L cells and found that there were nearly 7 fold more colonies form FE25L than FE25 cells (p < 0.001, Figure 3).

FE25L Exhibited Greater Tumorigenesis Potential
Next, we examined the tumorigenic capability of the two cell lines in NSG mice. subcutaneous injection of 1 × 10 6 FE25 or FE25L cells into NSG mice, FE25L gene tumors in 3/3 mice on days 165 (1 mouse) and 183 (2 mice). FE25 generated tumors in one of the three mice on day 173 ( Figure 4A). The gross appearance of the xenogra wax blocks are shown in Figure 4B. Hematoxylin and eosin staining revealed more morphic cells in the FE25 tumor and more epithelioid cells in the FE25L tumor (F 4C). Nuclear atypia with significant nuclear pleomorphism with large, bizarre, and tinucleated form was noted in the xenografts.

FE25L Acquired More Severe Aneuploidy
Chromosomal instability with copy number variation (CNV) is a phenotype ch teristic of HGSOC [15]. We analyzed the karyotypes of the FE25 and FE25L cells.

FE25L Exhibited Greater Tumorigenesis Potential
Next, we examined the tumorigenic capability of the two cell lines in NSG mice. After subcutaneous injection of 1 × 10 6 FE25 or FE25L cells into NSG mice, FE25L generated tumors in 3/3 mice on days 165 (1 mouse) and 183 (2 mice). FE25 generated tumors in only one of the three mice on day 173 ( Figure 4A). The gross appearance of the xenografts in wax blocks are shown in Figure 4B. Hematoxylin and eosin staining revealed more pleomorphic cells in the FE25 tumor and more epithelioid cells in the FE25L tumor ( Figure 4C). Nuclear atypia with significant nuclear pleomorphism with large, bizarre, and multinucleated form was noted in the xenografts.

FE25L Acquired More Severe Aneuploidy
Chromosomal instability with copy number variation (CNV) is a phenotype characteristic of HGSOC [15]. We analyzed the karyotypes of the FE25 and FE25L cells. We found marked polyploidy and aneuploidy in FE25L cells compared with FE25 cells ( Figure 5). Specifically, FE25 cells were sub-diploid with 42 chromosomes at passage 31 ( Figure 5A), and FE25L cells were sub-triploid with 59-74 chromosomes at passage 139 ( Figure 5B).

Copy Number Variation in FE25L vs. FE25
DNA copy number variations were evaluated by whole-exome sequencing of FE25 (passage 20) and FE25L (passage 124) cells ( Figure 6A,B). Compared to normal cells, both FE25 and FE25L showed amplification of driver genes such as CCNE1 and NOTCH1, which were also found in the TCGA-HGSOC study [16]. Other amplified oncogenes frequently found in epithelial ovarian cancer were ERBB2, STAT3, PIK3C2B, CDK2, and CDK4 [16], as well as SWI/SNF chromatin modification complex genes, including ARID1A and SMARCC1 [16]. TCGA-confirmed tumor suppressor genes with a lower copy number in the two cells included RB1 and CDKN2A/B [16]. Other copy number reduced genes included SMAD4, HNF1B, and some gene loci that are not known to be tumor suppressor genes, such as KRAS, MYC, PIK3CA, and IL6.  Table (C) were further confirmed by qRT-PCR. *** p < 0.001. * p < 0.05.

Differentially Regulated Gene Sets in FE25L Compared to FE25 Cells
The relative increase or decrease in the gene copy number in FE25L does not necessarily result in changes in expression. We performed an RNA-seq array on these two cells to reveal the gene expression changes between FE25L and FE25. Differentially regulated genes in FE25L compared to FE25 were identified and subjected to gene ontology analysis using GSEA [19]. We selected 100 upregulated and 100 downregulated genes (highest fold-change) for the analysis (these genes may be prominent genes in this phenotype). The three most significantly enriched gene sets in upregulated genes included hypoxia, epithelial-mesenchymal transition (EMT) (VEGFA, COL6A2, COL6A3, ADAM12, TFPI2, MXRA5, LAMA1, SERPINE2, and SFRP1), and NF-κB pathways ( Table 3). The three most significant enrichment of gene sets in downregulated genes included the interferon-γ signaling pathway, interferon-α signaling pathway, and EMT (TIMP3, THBS1, COL3A1, VCAN, COL4A2, IGFBP4, TAGLN, CDH6, COL1A2, APLP1, and NTM) ( Table 4). The above results suggest that FE25L contains more cancer transformation-related genes.  Table  shows genes with somatic varied copy number (SNV) upon FE25 to FE25L progression, which also showed a compatible mRNA alteration as revealed by RNA sequencing. (D) The genes listed in Table  (C) were further confirmed by qRT-PCR. *** p < 0.001. * p < 0.05. Compared to FE25, FE25L had gain sites at chr2p25, 7q21,22, 8p11, and 20q11, which are related to breast cancer and chronic myelogenous leukemia (Table 1) [17,18]. FE25L loss sites were noted on chromosome 11, related to complement, coagulation, oxidative phosphorylation, estrogen early response, and mTORC1 signaling ( Table 2).  Table 2. CNV loss sites and overlap genes of FE25L (p124) compared to FE25 (p20) cells (all at chr11).

HALLMARK_ COAGULATION <138>
Genes encoding components of blood coagulation system; also up-regulated in platelets.

Differentially Regulated Gene Sets in FE25L Compared to FE25 Cells
The relative increase or decrease in the gene copy number in FE25L does not necessarily result in changes in expression. We performed an RNA-seq array on these two cells to reveal the gene expression changes between FE25L and FE25. Differentially regulated genes in FE25L compared to FE25 were identified and subjected to gene ontology analysis using GSEA [19]. We selected 100 upregulated and 100 downregulated genes (highest fold-change) for the analysis (these genes may be prominent genes in this phenotype). The three most significantly enriched gene sets in upregulated genes included hypoxia, epithelial-mesenchymal transition (EMT) (VEGFA, COL6A2, COL6A3, ADAM12, TFPI2, MXRA5, LAMA1, SERPINE2, and SFRP1), and NF-κB pathways ( Table 3). The three most significant enrichment of gene sets in downregulated genes included the interferon-γ signaling pathway, interferon-α signaling pathway, and EMT (TIMP3, THBS1, COL3A1, VCAN, COL4A2, IGFBP4, TAGLN, CDH6, COL1A2, APLP1, and NTM) ( Table 4). The above results suggest that FE25L contains more cancer transformation-related genes.   Next, we explored the differences in the expression of related signaling pathway genes between FE25 and FE25L. TCGA ovarian cancer array analyzed 570 human HGSOC tumors using a whole-genome array. The prognosis of ovarian cancer patients with a high EMT index is worse than that of patients with a low EMT index [20]. We found lower expression of E-cadherin and increased expression of fibronectin, vimentin, and Twist in FE25L cells, which may reflect more EMT in FE25L cells than in FE25 cells ( Figure 7A). Cytokines and growth factors secreted by tumor tissues in the ovarian cancer microenvironment play a role in the immune escape, tumor progression, and cancer dissemination [21]. We found higher expression of interleukin (IL)-6, IL-8, bone morphogenetic protein 2 (BMP2), insulin-like growth factor 2 (IGF2), and VEGFA in the FE25L group than in the FE25 group ( Figure 7B). Ccnd1 overexpression is strongly associated with shortened progression-free survival in human ovarian carcinomas [22]. NF-κB is also associated with decreased survival rate in ovarian cancer [23]. Genes involved in the NF-κB pathway include Ccnd1, S100a4, Spp1, and IkBkε [24]. We found lower S100A4 and SPP1 expression in FE25L than in FE25 ( Figure 7C). HGSOC is characterized by numerous copy number alterations, among which overexpression of the myc oncogene occurs in half of the tumors [25]. We found that c-Myc expression was higher in FE25L cells than in FE25 cells ( Figure 7D). Taken together, FE25L expressed more TCGA HGSOC-related genes than FE25 did.

FE25L Exhibited Higher Expression of EMT, Cytokine, NF-κB, and c-myc Signaling Pathways
Next, we explored the differences in the expression of related signaling pathway genes between FE25 and FE25L. TCGA ovarian cancer array analyzed 570 human HGSOC tumors using a whole-genome array. The prognosis of ovarian cancer patients with a high EMT index is worse than that of patients with a low EMT index [20]. We found lower expression of E-cadherin and increased expression of fibronectin, vimentin, and Twist in FE25L cells, which may reflect more EMT in FE25L cells than in FE25 cells ( Figure 7A). Cytokines and growth factors secreted by tumor tissues in the ovarian cancer microenvironment play a role in the immune escape, tumor progression, and cancer dissemination [21]. We found higher expression of interleukin (IL)-6, IL-8, bone morphogenetic protein 2 (BMP2), insulin-like growth factor 2 (IGF2), and VEGFA in the FE25L group than in the FE25 group ( Figure 7B). Ccnd1 overexpression is strongly associated with shortened progression-free survival in human ovarian carcinomas [22]. NF-κB is also associated with decreased survival rate in ovarian cancer [23]. Genes involved in the NF-κB pathway include Ccnd1, S100a4, Spp1, and IkBkε [24]. We found lower S100A4 and SPP1 expression in FE25L than in FE25 ( Figure 7C). HGSOC is characterized by numerous copy number alterations, among which overexpression of the myc oncogene occurs in half of the tumors [25]. We found that c-Myc expression was higher in FE25L cells than in FE25 cells ( Figure  7D). Taken together, FE25L expressed more TCGA HGSOC-related genes than FE25 did.

FE25L Cells Exhibited Increased Wnt/β-Catenin Expression
The Wnt/β-catenin signaling pathway was examined by Western blotting and was used to check protein expression in the cytoplasm and nucleus. The Wnt/β-catenin pathway is associated with cancer stem cell self-renewal, chemoresistance, and metastasis in

FE25L Cells Exhibited Increased Wnt/β-Catenin Expression
The Wnt/β-catenin signaling pathway was examined by Western blotting and was used to check protein expression in the cytoplasm and nucleus. The Wnt/β-catenin pathway is associated with cancer stem cell self-renewal, chemoresistance, and metastasis in all subtypes of epithelial ovarian cancer [26]. In our study, β-catenin protein expression was increased in both the cytoplasm and nucleus of FE25L cells ( Figure 8A-D). Taken together, FE25L expressed more β-catenin than FE25 did. all subtypes of epithelial ovarian cancer [26]. In our study, β-catenin protein expression was increased in both the cytoplasm and nucleus of FE25L cells ( Figure 8A-D). Taken together, FE25L expressed more β-catenin than FE25 did.

Correlation of Gene Expression between FE25 and FE25L Cells and Current HGSOC Lines
A previous study used numerical scores to group current ovarian cancer cell lines into three groups: likely HGSOC (Group 1), probably HGSOC (Group 2), and unlikely HGSOC (Group 3) models of HGSOC [27]. KURAMOCHI or COV362 cell lines belong to Group 1, and SKOV3 or A2780 cells belong to Group 3 [27]. We used genetic features identified in the likely HGSOC cancer cell lines (Group 1) to compare with FE25 and FE25L [27]. Using the top 500 diverse genes of Group 1 HGSOC, FE25 and FE25L cells were grouped with Group 1 cells, including JHOS2, CAOV3, COV362, and TYKNU (Figure 10).

Correlation of Gene Expression between FE25 and FE25L Cells and Current HGSOC Lines
A previous study used numerical scores to group current ovarian cancer cell lines into three groups: likely HGSOC (Group 1), probably HGSOC (Group 2), and unlikely HGSOC (Group 3) models of HGSOC [27]. KURAMOCHI or COV362 cell lines belong to Group 1, and SKOV3 or A2780 cells belong to Group 3 [27]. We used genetic features identified in the likely HGSOC cancer cell lines (Group 1) to compare with FE25 and FE25L [27]. Using the top 500 diverse genes of Group 1 HGSOC, FE25 and FE25L cells were grouped with Group 1 cells, including JHOS2, CAOV3, COV362, and TYKNU ( Figure 10).

Discussion
FE25 cells derived from FTEC were immortalized with human papillomavirus with knockdown of TP53 and Rb genes [28]. Several studies using FE25 cells have investigated their roles in ovarian carcinogenesis [29,30]. In previous studies, treatment with ovulatory follicular fluid (FF) or growth factors within FF resulted in more malignant behavior in FE25 cells similar to HGSOC [12,28,31,32]. Given the chromosomal instability due to defective p53 and Rb, we hypothesized that FE25 cells might develop HGSOC phenotype after long-term culture. In this study, we demonstrated that higher passage FE25 cells, FE25L had an increased proliferation, clonogenicity, DNA aneuploidy, migration and invasion capabilities, and higher drug resistance. In terms of gene expression, FE25L showed upregulated EMT, cytokines, NF-κB, c-Myc, and Wnt/β-catenin pathway-associated genes compared to those in FE25 cells. FE25L cells also showed higher tumorigenicity with WT1 expression.
Distinct from human cells needing transgenes (e.g., hTERT) to conquer senescence, murine cells are easily immortalized in culture and may transform spontaneously after long-term culture [33]. McCloskey et al. observed spontaneous transformed mouse ovarian surface epithelial cells after a long passage and determined the aberrant Wnt/β-catenin and NF-κB signaling and expression of WT1, PAX8, and cytokeratin mimicking human HGSOC [24]. Another murine background cell line, ID8, is derived from the ovarian surface epithelium and transformed spontaneously after long passages [34]. Nevertheless, ID8 cannot generate HGSOC phenotypes due to not carrying the characteristic driver mutations of HGSOC (e.g., Trp53 and Rb, etc.). Quartuccio et al. studied spontaneous transformed murine oviductal epithelial (MOE) hi cells after long passages of MOE low and could generate tumors after long-term allograft. However, the tumor exhibited a sarcoma phenotype [35]. To our knowledge, FE25L is the first HGSOC cell line that spontaneously transformed from human fimbria secretory cells with the characteristic p53 and Rb/CCNE1 defect and CNV phenotypes.
Aggressive cancer cells often exhibit a high proliferation rate [36]. High-passage cancer cells also exhibit a shorter doubling time [37]. Our study was also consistent with previous studies in that FE25L showed a higher proliferation rate than that of FE25 cells.
Cancer chromosome instability (CIN) causes cell chromosome number and structural changes, resulting in tumor heterogeneity [38]. In most solid tumors, CIN with aneuploidy is associated with drug resistance and poor prognosis [39,40]. In our study, both FE25 and FE25L showed these features of CIN with DNA polyploidy and aneuploidy. Both cells showed loss of chromosome 19 commonly found in ovarian cancer [41], and showed loss of chromosome 9 (which is typically associated with copy number variation) [39], 14, 16, and 18. Compared to FE25, FE25L cells showed more severe CIN with progressive CNV. HGSOC carries a broad spectrum of diverse alterations in CNV [42]. CNV deletion of Tp53 causes further suppression or misdirection of p53 [43]. Among the FE25/FE25Laltered genes, MYC (the stem-cell transcription factor) was the most upregulated gene (42% with a 4N copy number, 37% with 3N) in TCGA. Additionally, Her2 (encoded by ERBB2) overexpression can be noted but is unrelated to CNV amplification [44]. Comparative genomic hybridization revealed that CCNE1 was significantly amplified [45]. Nevertheless, few additional single-gene drivers have been explored in TCGA [16,46]. Our study also found more CNV amplifications of ERBB2, TP53, and CCNE1 in FE25L than in FE25 cells.
CASP1 is one of the FE25L-to-FE25 copy number-altered genes which was confirmed at RNA level. Caspase 1 (casp1) is an apoptosis-related protein [47]. A previous study found that caspase 1 was abundant in the normal ovarian surface epithelium but was reduced in ovarian cancer cell lines [47]. Therefore, the downregulation of caspase 1 may be related to increased resistance to apoptosis in ovarian cancer cells. Our study showed casp1 was downregulated in FE25L cells (CNV-low), which may be related to the characteristics of advanced ovarian cancer.
Hypoxia that influences invasion and adhesion functions can promote ovarian cancer proliferation [48]. Patients with ovarian cancer with a high EMT index were associated with worse overall survival than patients with a low EMT index [20]. EMT is an important process in cancer invasion and metastasis and involves more than 30 gene expressions [49]. Cell morphology and cell-cell adhesion also affect invasion and migration [50,51]. NF-κB is also associated with decreased survival rate in ovarian cancer [23]. In our RNA-seq data, the three most significantly enriched gene sets in upregulated genes included hypoxia, EMT, and the NF-κB pathway in FE25L compared to FE25. qRT-PCR also confirmed that FE25L cells had increased expression of EMT-related genes, including vimentin, fibronectin, and Twist.
Gene expression of cytokines also indicates the malignant behavior of ovarian cancer cells. Elevated levels of cytokines in ascites are associated with patient survival [52]. We previously showed that IL-6 is increased in ovarian stromal cells and enhances the aggressive behavior of ovarian cancer [53]. Insulin-like growth factor (IGF) is also important for ovarian cancer initiation [12]. In this study, we showed increased expression of IL-6, IL-8, IGF-2, and VEGFA in FE25L cells compared to that in FE25 cells. These cytokines also correlate with the aggressive tumorigenic behavior of the cells.
In ovarian cancer, the Wnt/β-catenin pathway is one of the significant signaling pathways. A previous study showed that increased β-catenin expression reduces Dicer expression to promote ovarian cancer metastasis [54]. Conversely, inhibition of the βcatenin pathway can inhibit ovarian cancer metastasis [55]. Therefore, β-catenin is believed to be involved in ovarian cancer metastasis. In our study, after long-term in vitro passage of FE25 cells, FE25L cells exhibited increased β-catenin signaling. Therefore, our results show that FE25L exhibited an increased metastatic capability compared to FE25 cells.
Cancer metastasis is represented in the late stage and is related to a poor prognosis [56]. Metastasis is a process that involves cancer cell invasion and migration. The colony formation ability of cancer cells revealed their invasiveness. In our study, FE25L showed increased migration, invasion, and colony formation capabilities.
Following debulking surgery, adjuvant chemotherapy with carboplatin and paclitaxel is administered. However, despite the high response rate to the first chemotherapy, the disease often recurs with chemoresistance [57]. Therefore, drug resistance in cancer cell lines is also a marker of aggressive cancer behavior [58]. However, in our study, the IC 50 values of the three chemotherapeutic drugs were lower in FE25L than in FE25 cells. Low drug resistance may be due to the CNV status of FE25L.
The Cancer Genome Atlas can systematically compare genomic features of various cell lines and tumors. The previous study showed significant differences in molecular profiles between commonly used ovarian cancer cell lines and HGSOC tumor samples [27]. The results revealed that a gap between cell lines and tumors could be bridged by genomically choosing cell line models for HGSOC. Cell lines such as TYKNU and COV362 are topranked HGSOC-like cell lines other than commonly used A2780 cells [27]. Unsupervised hierarchical clustering analysis of RNA expression showed FE25L was more clustered with cell lines such as TYNKU and CAOV3 (likely HGSOC), which are characterized as better cell lines for studying human HGSOC.
The main limitation of this study was that most of the experiments were more than three technical replicates but not different cell batch replicates. Nevertheless, the FE25 cells has been extensively studied in our previous research in different cell batches with different passage numbers [28,29,31,32]. They showed consistent results. Therefore, the current results could be trusted.

Cell Culture
FE25 and FE25L (high passage) cells were obtained from Dr. Chu TY's lab and were cultured in MCDB105 and M199 medium (Sigma, St. Louis, MO, USA) supplemented with 10% fetal bovine serum (FBS), 100 IU/mL penicillin, and 100 µg/mL streptomycin. 5 × 10 5 cells were grown in 75 cm 2 culture flasks in an incubator maintained at 37 • C and 5% CO 2 . The culture medium was replaced every 2 days. Until reaching 75% of confluence, they were passaged at 1:3 ratio with trypsin (Sigma) treatment.

Cell Proliferation Assay
The cell proliferation kit (XTT (2,3-Bis-(2-Methoxy-4-Nitro-5-Sulfophenyl), Biological Industries Ltd., Kibbutz Beit Haemek, Israel) assay was used to measure cell proliferation according to the manufacturer's instructions. After 0, 3, 5, and 7 days of incubation, proliferation with XTT was tested. Absorbance was detected using a microplate reader at an optical density of 450 nm (Bio-Rad Model 3550, Hercules, CA, USA). The optical density values at each time point were used to construct proliferation curves of the tested cell lines.

Chromosome Evaluation
Aneuploidy is a characteristic feature of cancer cells [59]. Aneuploidy was studied in FE25 (passage 31) and FE25L (passage 139) cells at the cytogenetics core laboratory at Hualien Tzu Chi Hospital, Hualien, Taiwan. Cells were cultured in the medium until they reached an adequate amount, and 0.25% colchicine (Sigma) was added to arrest cells at metaphase. The hypotonic solution was then added to induce cell bursting. Giemsa (1:20, Sigma) staining chromosomes were reviewed by a cytogeneticist. After counting 50 metaphases, the chromosome number distribution was acquired and reported according to the 2016 International System for Human Cytogenetic Nomenclature.

Colony Formation Assay
We plated 200 cells in a 2 mL culture medium in one well of a 6-well plate. After 14 days of culture, the cells were fixed with 70% ethanol and stained with 0.8 mM crystal violet (Sigma-Aldrich), and the number of colonies was measured by ImageJ (NIH, Bethesda, MA, USA). The experiments were performed in triplicate

Migration and Invasion Assay
We placed 5 × 10 4 cells in 200 µL medium into the upper chamber of a 24-well transwell Boyden chamber (8 µm pore size; Costar, Corning Inc., Corning, NY, USA), and allowed the cells to migrate to the lower layer, where there were no cells and which only had 500 µL culture medium. After culturing for 48 h, the cells were stained with DAPI (Sigma-Aldrich) and counted using an immunofluorescence microscope. The experiments were performed in triplicate.

Drug Sensitivity Analysis
Paclitaxel and doxorubicin are common chemotherapeutic drugs used to treat ovarian cancer [60]. The effects of paclitaxel (Formoxol, Yung Shin Pharm. Ind., Co. Ltd., Taichung, Taiwan), doxorubicin (Adriblastina, Pfizer, Kent, NJ, USA), and carboplatin (SINPHAR Pharmaceutical Co., Ltd., Yilan, Taiwan) on FE25 and FE25L cells were studied. Cells were seeded at a density of 2500-3000 cells/well in a 96-well plate, cultured overnight, and XTT solution was used for calculating cell proliferation according to the manufacturer's instructions. After incubation for 2-5 h at 37 • C, the absorbance of the wells was determined using a spectrophotometer (DYNEX MRX II ELISA reader, Bustehrad, Czech) at a wavelength of 450 nm and a reference wavelength of 650 nm.
The IC 50 values of the three drugs for these two cell lines were then obtained using a non-linear regression model on GraphPad Prism (version 9.0, GraphPad Software, San Diego, CA, USA).

Gene Expression Analysis by qRT-PCR Extraction of Total RNA from Cells
RNeasy Mini kit (QIAGEN, Hilden, Germany) was used to isolate RNA from 5 × 10 5 FE25 or FE25L cells plated in a 10-cm culture dish. Briefly, after 24 h of culture, the medium was removed and the cells were washed twice with 1 × PBS, detached using 0.05% trypsin, and pelleted. The cells were then lysed in 700 µL RLT lysis buffer by aspirating them several times with a 1 mL micropipette until the cells were completely lysed and the solution became transparent. Reagents were then added to extract RNA according to the manufacturer's instructions.

Preparation of cDNA
One µg of total RNA was used for cDNA synthesis using Reverse-iT TM 1st strand synthesis kit (ABgene, Portsmouth, NH, USA) using the manufacturer's instructions. Briefly, 1µg total RNA and 1 µL anchored oligo dT primers were mixed with DEPC-treated (diethylpyrocarbonate) water to a total volume of 12 µL, incubated at 70 • C for 5 min, placed on ice. Then 8 µL reaction mixture (4 µL 5× First-strand synthesis buffer, 2 µL dNTP mix 5 mM each, 1 µL 100 mM DTT, Reverse-iT TM RTase blend 1 µL) was added followed by incubation at 47 • C for 50 min, after which the reaction was terminated at 75 • C for 10 min, followed by storage at −20 • C until use.

Quantitative Polymerase Chain Reaction
We used the ABI Step One Plus system (Applied Biosystems, Waltham, MA, USA) and FastStart Universal SYBR Green Master (ROX, Basel, Switzerland) gene expression analysis reagents for quantification of gene expression. mRNA expression levels were normalized to the mean Ct of the GAPDH, which was used as an endogenous control. The primer sequences are listed in Table 5.

RNA Sequencing and Gene Set Enrichment Analysis (GSEA)
Gene expression profiling was conducted using high-throughput RNA sequencing (Illumina NovaSeq 6000 platform) with 28,264 human genome probes. Upregulated and downregulated genes in FE25L compared to FE25 were identified and subjected to gene ontology analysis using GSEA [19]. Online GSEA (https://www.gsea-msigdb.org/gsea/ index.jsp, assessed on 19 May 2022) was used to identify gene function sets with a significant false detection rate (FDR) q-value of <0.05.

Cross-Platform Comparability
Cluster analysis of FE25, FE25L, and ovarian cancer cell lines, including A2780, CP70, A1847, KURAMOCHI, etc., was performed. Human OneArray version HOA 6.1 (Phalanx Biotech Group, Hsinchu, Taiwan) platform containing 28,264 human genome probes was used. Additionally, public domain data of ovarian cancer cell lines (Cancer Cell Line Encyclopedia; CCLE RNAseq gene expression data for 1019 cell lines) and the Affymetrix Human Genome U133 Plus 2.0 Array (Platform GPL 15308) were used to cluster FE25 and FE25L with the analyzed ovarian cancer cell lines. The previous study grouped current ovarian cancer cell lines into group 1 (likely HGSOC), group 2 (probably HGSOC), and group 3 (unlikely HGSOC) [27]. We used the group 1 genetic feature to identify the cell lines that resemble HGSOC cells. The top 500 diverse genes of group 1 were selected to differentiate the difference between the cell lines.

Library Preparation for the Whole Exon Sequencing
The DNA library was prepared using 1 µg of genomic DNA as input material. An Agilent SureSelect Human All Exon kit (Agilent Technologies, Santa Clara, CA, USA) was used to generate sequencing libraries. A hydrodynamic shearing system (Covaris, Woburn, MA, USA) was used to generate 160-280 bp DNA fragments. The exonuclease/polymerase activities were used to convert the remaining overhangs into blunt ends. The 3 ends of the DNA fragment were adenylated and adapter oligonucleotides were ligated. Polymerase chain reaction (PCR) was used to enrich the DNA fragments and ligate adapter molecules at both ends. After the PCR, the biotin-labeled probe in the liquid phase was hybridized into the library. The exons of the genes were captured using streptomycin-coated magnetic beads. PCR was used to enrich the captured libraries for hybridization. An AMPure XP system (Beckman Coulter, Beverly, MA, USA) was used to purify the final products. An Agilent high-sensitivity DNA assay on the Agilent Bioanalyzer 2100 system was used to quantify the products.

Copy Number Detection and Visualization
To detect copy number variation (CNV) in FE25 and FE25L, a software toolkit CNVkit (v0.9.8) was used to detect and visualize CNVs from whole-exome sequencing data [61]. Briefly, to determine the genome-wide copy number, the read depths of the on-target sequencing reads were compared and provided comparable copy number estimates in the target regions. The normalized copy ratio was visualized using a scatter plot.

Western Blot
Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) was performed in 1× running buffer (0.025 M Tris, 0.192 M glycine, 0.1% SDS) as the electrolytic buffer. The extracted protein (50 µg), and 6× protein-loading dye were boiled at 95 • C for 10 min, and 18 µL/well was loaded into the gel. The gel was first run at a voltage of 80 V and after the sample passed through the stacking gel, the voltage was increased to 120 V. After electrophoresis was completed, a transfer buffer was used to transfer the protein separated by SDS-PAGE colloid to the transfer membrane. The transfer membrane was washed with 1× PBST (PBS Tween-20 buffer, 0.13 M NaCl, 0.01 M NaH2PO4, 0.05% (v/v) Tween 20) for 10 min, followed by incubation in the blocking buffer for 1 h. The membranes were then incubated with the primary antibody diluted in the blocking buffer for 1 h. The samples were washed three times with PBST for 10 min. Membranes were then incubated with HRP-linked secondary antibody diluted in blocking buffer for 1 h, and washed three times with PBST for 10 min. The protein bands were visualized by incubating in HRP substrate in the dark for 1 min. The samples were clamped in a plastic bag and detected using a luminometer (FUJIFILM LAS-3000). The antibodies used were β-catenin (1:1000, Sigma-Aldrich). β-actin (1:10,000, Sigma-Aldrich) and histone-H3 (1:10,000, Sigma-Aldrich) were used as internal controls.
For protein quantification, the data were normalized to total protein loading (actin or histone-H3), and relative quantification was compared with that of the original cells for fold-change analysis. ImageJ software (National Institutes of Health, Bethesda, MA, USA) was used to perform quantification according to the methods described in the previous study [62].
The ReadyPrep protein extraction kit (cytoplasmic/nuclear, Bio-Rad) was used to isolate proteins from the nucleus and cytoplasm following the manufacturer's instructions. Western blotting was performed to analyze the resulting cytoplasmic/nuclear proteins.

Xenograft Tumorigenesis
NSG immunodeficient mice [NOD/Shi-scid/IL-2Rγ null ] were purchased from Jackson Laboratory and raised at the Tzu Chi University Laboratory Animal Center (Hualien, Taiwan). The Laboratory Animal Care and Use Committee of Hualien Tzu Chi Hospital approved the experimental protocol.
Mice were divided into two groups: FE25 and FE25L (n = 3 each). A total of 1 × 10 6 tested cells were transplanted into the subcutaneous region of each mouse's back. We observed by palpation and measured the tumor growth in each group of mice. Tumor volume was calculated as (width) 2 × length/2 (mm 3 ). Three to six months later, when the tumor volume of the mouse was greater than 500 mm 3 , the mice were sacrificed, and the tumor was removed and stored in neutral formalin.
The percentage of tumorigenesis and xenograft proliferation rate of the test cells was calculated and compared between the two groups.

Immunohistochemistry (IHC)
Hematoxylin and eosin staining and immunostaining were used to evaluate the histology of the tumors formed by the tested cells. First, xylene was used to remove the formalin, followed by alcohol stratification and rehydration. The antigen unmasking solution (Dako, Agilent, Santa Clara, CA, USA) and 3% hydrogen peroxide in distilled water were used to block the intrinsic peroxidase activity. Immunostaining was performed with antibodies against PAX8 (1:500, Dako) and WT-1 (1:500, Dako), CK7 (1:500, Dako) according to the manufacturer's instructions (Vector). The specimens were then placed in a refrigerator at 4 • C overnight and treated with anti-rabbit horseradish peroxidase-labeled polymer (Dako) at room temperature for 30 min. All immunostained slides were stained with hematoxylin. WT-1 positive cells were calculated by the positive staining cell numbers among 50 cells counted in each of the three fields.

Statistical Analysis
This study (proliferation, gene expression, colony number, migration and invasion cell numbers, and tumor proliferation rate) used the non-parametric test (Mann-Whitney U test) for two groups to analyze whether the experimental results were statistically significant. Fisher's exact test was used to calculate the percentage of tumorigenesis. Statistical significance was set at 0.05. The results are presented as mean ± standard deviation (SD).

Conclusions
We demonstrated that FE25L cells showed more aggressive behavior in cancer than FE25. Genome instability, proliferation rate, migration, invasion, colony formation, and tumor formation capabilities were higher in the FE25L group than in the low passage FE25 cells. More HGSOC gene signatures and CNV were observed in the FE25L group. Therefore, FE25L might be a more suitable model for ovarian cancer research. Future studies should aim at comprehensively analyzing the differences in gene expression between FE25L and FE25 cells. Data Availability Statement: The datasets generated during and analyzed during the current study are available from the corresponding author on reasonable request.