Epithelial ovarian carcinoma cell lines (SKOV3, CAOV-3, OV-1063, and OVCAR-3) and human ovarian epithelium cell line IOSE80 were purchased from Chinese Academy of Sciences (Beijing, China). The cells were cultured in RPMI-1640 medium supplement with 10% fetal bovine serum and maintained in a condition of 37°C and 5% CO₂.

Approved by the Ethics Committee of Hangzhou First People’s Hospital (approval no. IRB#2021-20210406-01), with informed consent by patients, we collected 22 tissue samples from 11 patients. These specimens were human tissue specimens that were pathologically diagnosed as epithelial ovarian cancer during staging ovarian cancer surgery, including cancerous tissues and para-cancerous tissues.

The total RNA was extracted from EOC cells and tissues using TRIzol (TaKaRa Inc. Tokyo, Japan). Reverse transcription kit (TaKaRa Inc., Tokyo, Japan) was used to transform RNA to cDNA. The mRNA levels of NTPCR in tumor cells and tissues were detected using SYBR Green kit (Thermo Inc., Waltham, MA, United States), and GAPDH was set as an endogenous control. The relative expression levels of NTPCR in different samples were determined by the method of comparing CT values (ΔΔCT) and normalization with internal reference. The primers were NTPCR-hF: ACCCGTCTTGAGGAATGTGA and NTPCR-hR: CTCTTGAACTGGGCACTCCT, and GAPDH-hF: TGACAACTTTGGTATCGTGGAAGG and GAPDH-hR: AGGCAGGGATGATGTTCTGGAGAG.

The overexpression vector of NTPCR used a commercial vector (Sino Biological, Beijing, China, cat: HG17161-NY); shRNA was synthesized by Shanghai Gima Biopharmaceutical Technology (sequence as follows): shRNA1, 5′-GGCCTTTATCGAGAGTTGGGTTAGA-3′; shRNA2, 5′-CCTCTGGTGTGCCTGTTGATGGATT-3′; and shRNA3, 5′-CAGTATGTGGTCGACCTGACTTCTT-3′. shRNA was inserted into pGPU6/Neo to obtain the corresponding plasmid. SKOV3 and OVCAR-3 cells were taken in the logarithmic growth phase and transfected with Lipofectamine^TM 2000. Six hours later, the mixture was aspirated and changed to a normal medium to continue culturing. After 72 h, the cells were harvested and used to detect molecular expression by RT-PCR.

Cell viability was evaluated using a CCK-8 kit (Beyotime Biotechnology, China) following the protocol of the manufacturers. SKOV3 and OVCAR-3 cells were seeded into 96-well plates with 3,000 cells per well. After transfecting the overexpression vector or interference shRNA vector of NTPCR for 24, 48, 72, and 96 h, a total of 10 μl CCK-8 solution was added to wells, and the optical density (OD) values were read at 450 nm using a microplate reader system.

Cell proliferation was detected by BeyoClick^TM eyoClick cell proliferation detection kit (Biyuntian, C0078S). The following are the general steps. Add EdU to the cell culture medium at a final concentration of 10 μM, incubate for 10 h, and fix with 4% paraformaldehyde for 15 min after phosphate-buffered saline (PBS) immersion. Remove the fixative solution by PBS immersion, and then incubate with 0.5% Triton X-100 at room temperature for 15 min to permeate the membrane. After removing the permeabilization solution, immerse in PBS, add 0.5 ml of Click reaction solution, and incubate for 30 min at room temperature in the dark. Remove the reaction solution, after immersion in PBS, add Hoechst 33342 reaction solution dropwise and incubate at room temperature for 15 min in the dark. Mount the slide with mounting solution containing anti-fluorescence quencher, and then observe and collect the images under a fluorescence microscope (Olympus, BX53).

After cell transfection, SKOV3 and OVCAR-3 cells were harvested for flow cytometry analysis. Differential groups of cells were washed with PBS three times and then stained with propidium iodide (PI)/RNase solution following the protocol of the manufacturer. Cell cycles were detected by using FACScan flow cytometer (BD Biosciences) and analyzed using FlowJo software.

We evaluated the effect of NTPCR on the cell invasion and migration abilities by using the Transwell system. The membranes of inserts were precoated with Matrigel for invasion assay overnight. After that, SKOV3 and OVCAR-3 cells were harvested and suspended using serum-free RPMI-1640 medium. Thus, 200 μl of cell suspension including 2.0 × 10⁴ cells was added into the upper chamber, while 500 μl of culture medium with 10% fetal bovine serum (FBS) was added into the lower chamber. Following incubation for 24 h, the cells were fixed with 4% paraformaldehyde and stained with 0.1% crystal violet for 10 min. The results were analyzed by using a microscope (Olympus IX73; Olympus Corporation, Tokyo, Japan).

The protein samples prepared in ice-cold RIPA buffer were separated on SDS-PAGE gels and then transferred onto polyvinylidene fluoride (PVDF) membranes. The membranes were then blocked in 5% skimmed milk at room temperature for 1 h and then incubated with the primary antibody solution (NTPCR: Biodragon BD-PT3202 anti-rabbit monoclonal, 1:1,000; GAPDH: Proteintech 60004-1-lg mouse monoclonal, 1:1,000) at 4°C overnight. The following day, the secondary antibodies (Peroxidase AffiniPure Goat Anti-Rabbit IgG (H + L), 1:10,000; Peroxidase AffiniPure Goat Anti-Mouse IgG (H + L), 1:5,000) were added and then incubated at room temperature for 2 h. For chemiluminescence development, Millipore ECL system was used; to perform grayscale analysis on the data, TanonImage was used, and then statistical analysis was performed on the grayscale analysis results. The data mapping software is GraphPad Prism5.

We performed RNA-sequencing on an Illumina system for shNTPCR and control shNC group with three duplications per group. Firstly, Trimmomatic version 3.6 (Bolger et al., 2014) tool was used to clean up the reads by removing adapters and low-quality reads (quality value less than 20), and the reads with N ratio exceeding 10% or length less than 50 bp were also removed. Hisat2 software (v2.05) (Siren et al., 2014) was used for the alignment of clean reads to human genomic database (GRCH38, Gencode) (Harrow et al., 2012). Moreover, human genome annotation in the resulting files was mapped to corresponding read count by using featureCounts tools (Liao et al., 2014). Further filtering was performed for different results according to the fragments per kilobase of exon per million fragments mapped (FPKM) method, and the genes with FPKM value less than 0.1 in any samples were removed.

Quasi-likelihood F-tests in edgeR software (Lun et al., 2016) were used to analyze DEGs between shNTPCR and shNC samples by setting | logFC| > 1.585 and false discovery rate (FDR) < 0.05 as cutoff criteria. Two-dimensional clustering analysis and functional enrichment analysis were conducted for these DEGs.

We further screened the gene–disease associations from the DisGeNET database (Piñero et al., 2016), a comprehensive platform containing information on human disease genes and related variants. As for the DEGs screened from the database, the expression status in EOC was investigated by using Gene Set Enrichment Analysis (GSEA) algorithm, a computational algorithm to assess significant differences of defined gene sets among biological phenotypes (Subramanian et al., 2007).

We analyzed the interactions of protein–protein according to the STRING online tool and constructed the protein–protein interaction (PPI) network using Cytoscape software. It has been shown that proteins with similar functions tend to cluster together in the PPI network, and thus, we mined the functional modules using MCODE tool (Bader and Hogue, 2003; Davis, 2015).

To further explored the regulation mechanism of EOC, we screened the miRNA–gene and TF–gene pairs using Enrichr (Chen et al., 2013) and ENCODE tools (Rosenbloom et al., 2011) and constructed an miRNA–TF–gene network.

The EOC cell extract samples derived from the shNTPCR and shNC groups were analyzed by the ultra performance liquid chromatography (UPLC) method on a quadrupole time-of-flight (Q-TOF) liquid chromatography–mass spectrometry (LC-MS) system. Duplicates from 10 controls together with five QC samples were obtained to evaluate the assay reproducibility.

Metabolomics data under positive ion mode and negative ion mode were obtained and first normalized before multivariate analysis. We performed unsupervised principal component analysis (PCA) and supervised partial least-square discriminant analysis (PLS-DA) to explore the metabolic differences in NTPCR samples.

As for the differential performance of metabolites between two groups, metabolite analysis was conducted using MBROLE 2.0, which is a web server designed to permit systemic analysis for metabolomic data (Chong et al., 2018). Corresponding KEGG pathways were screened with p < 0.05 as cutoff criteria.

We used the KEGG pathway-based method to analyze differential metabolites and genes in a biological process. The thresholds were set as num-overlapping-metabolites/genes > 0 and p joint/metabolites < 0.05.

Prism9.0 statistical software was used for data analysis. All data were expressed as mean ± standard deviation (±S). Pairwise comparisons between different groups were performed by one-way analysis of variance, and p < 0.05 was considered statistically significant.

The expressed status of NTPCR was detected in EOC tissues via RT-qPCR analysis and Western blot analysis. The results showed that the expression of NTPCR in cancer tissues was significantly increased compared to para-cancerous tissues (Figure 1, p < 0.05).

To further investigate the potential role of NTPCR on the biological function of EOC, we first constructed the overexpression and interference vector of NTPCR, and qRT-PCR verified its effect (Figure 2A). Simultaneously, analysis of NTPCR expression levels in different EOC cell lines (SKOV3, CAOV-, OV-1063, and OVCAR-3), and normal ovarian epithelial cell line IOSE80 as a control, found that the expression levels of NTPCR in SKOV3 and OVCAR-3 cells were higher (Figure 2B, p < 0.05); hence, we chose these two strains of cells for follow-up experiments.

In the CCK-8 assays, cell viability exhibited significant differences between NTPCR over expression (OE) and shNTPCR groups (Figure 2C, p < 0.01), indicating that overexpression of NTPCR inhibited EOC cell viability and interference with NTPCR expression increased cell viability. EdU results showed that NTPCR downregulation of NTPCR expression promotes the DNA replication process (Figure 2D), which shows that knockdown of NTPCR promoted cell proliferation in EOC cells. The effect of NTPCR on the cell cycle was also analyzed by flow cytometry analysis, and we found that overexpression of NTPCR increased the proportions of G0/G1phase cells and decreased the proportions of S- and G2/M-phase cells, while downregulating NTPCR expression was just the opposite (Figure 2E). Finally, through the Transwell chamber experiment, we found that NTPCR inhibited the metastatic and invasive abilities of EOC cells (Figures 2F,G). On the whole, we have enough reasons to believe that NTPCR can inhibit the biological process of EOC and act as a tumor suppressor.

On the basis of clarifying that NTPCR inhibits the biological process of EOC, we intended to initially explore the mechanism of NTPCR regulating the progress of EOC. First, we analyzed the mRNA expression profile of the shNTPCR and shNC groups through transcriptomics and screen out DEGs, and under the cutoff criteria of | log₂FC| > 1.585 and FDR < 0.05, we screened a total of 1,408 DEGs between the shNTPCR group and shNC group, including 783 upregulated and 625 downregulated genes. Bidirectional clustering analysis and Volcano Plot for DEGs are shown in Figure 3A.

We performed the Gene Oncology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis for DEGs in shNTPCR groups. The results showed that these genes were mainly related to biological processes of response to the virus, viral genome replication, response to lipopolysaccharide, and extracellular matrix organization. The signaling pathway categories included NOD-like receptor signaling pathway, IL-17 signaling pathway, hepatitis C, and cytokine–cytokine receptor interaction (Figure 3B).

Moreover, we analyzed the dysregulated gene sets in the shNTPCR group samples by using the GSEA algorithm. The results are shown in Figure 3C (enrichment score = 0.221325, p < 0.05). The enrichment score for upregulated genes is slightly higher than that of downregulated genes, indicating that the expression status of gene sets might exhibit a similar effect on EOC.

We constructed the PPI networks for these DEGs and screened the top 10 node genes with the higher scores based on the three topological properties (Figure 4A and Table 1). There were 348 nodes and 1,154 edges in the network. Several genes obtained higher degree values, such as STAT1 (degree = 43), OAS2 (degree = 36), OAS1 (degree = 36), IFIT3 (degree = 36), OASL (degree = 36), and MX1 (degree = 36).

Furthermore, we used MCODE plug to screen subnetwork modules related to PPI networks. By setting the score ≥ 10 and node ≥ 10 as cutoff criteria, two subnetwork modules were screened (Figure 4B). Several signaling pathways were identified as associated with these DEGs, including human papillomavirus infection, TNF signaling pathway, and Kaposi sarcoma-associated herpesvirus infection (Figure 4C).

The miRNAs and TFs correlated with DEGs were predicated by using Enrichr tools (Table 2). The top five miRNAs and TFs with the higher scores (combine score) were selected to establish a regulatory network of an miRNA–TF–target gene (Figure 4D). Hsa-miR-124-3p and EP300 obtained the highest count compared with other miRNAs or genes (Hsa-miR-124-3p, gene number = 195, p < 0.01; EP300, gene number = 27, p < 0.01), indicating the two genes were hub genes in the regulatory network.

Metabolism is the downstream of gene regulatory network and protein action network and is closely related to the phenotype of organisms. DEGs may ultimately affect the composition and level of intracellular metabolic spectrum. We used mass spectrometry to detect metabolite difference between the shNTPCR and shNC groups; by the threshold of FDR < 0.01, | log2FC|, and VIP > 1, we screened a total of 44 differential metabolites, including 26 metabolites in ESI + ion mode and 18 metabolites in the ESI− ion mode. Among these metabolites, there were 22 upregulated and 22 downregulated metabolites. Hierarchical cluster analysis and volcano plot results showed that these dysregulated metabolites can be significantly classified into two differential groups (Figures 5A,B).

Functional analysis results showed that a total of 12 signaling pathways were screened, including metabolic pathways, glycine, serine and threonine metabolism, glycerophospholipid metabolism, and ABC transporters (Figure 5C).

The integrated transcriptomic and metabolomic analysis showed that DEGs and differential metabolites in the NTPCR groups were associated with eight overlapped pathways, such as primary bile acid biosynthesis, protein digestion and absorption, pentose, and glucuronate interconversions (Figure 6). Of these pathways, pentose and glucuronate interconversions obtained a higher overlapped gene number than other pathways, including 10 overlapped genes (SLC7A7, SLC7A8, COL4A5, etc.) and one metabolite (L-threonine).