The human EOC A2780 cell line and the A2780-Taxol-resistant cell line were purchased from ImmoCell Biotechnology Co., Ltd. (provided by American Tissue Culture Collection). A2780 and A2780/Taxol cells were cultured in DMEM (Gibco; Thermo Fisher Scientific, Inc.) supplemented with 10% FBS (Zhejiang Tianhang Biotechnology Co., Ltd.) and 1% penicillin-streptomycin solution (Biosharp Life Sciences), at 37°C in a humidified incubator with 5% CO₂. The medium for the A2780 cell line was additionally supplemented with 1% L-glutamine (Procell Life Science & Technology Co., Ltd.). The medium for the A2780/Taxol cell line was also supplemented with 60 ng/ml Taxol (cat. no. H20203702; Sichuan Huiyu Pharmaceutical Co., Ltd.).

Total RNA was extracted from ~1×10⁶ cells using Trizol (Beyotime Institute of Biotechnology), followed by RNA-seq analysis performed by BGI [Platform: DNBSEQ (Homo sapiens)]. The RNA-seq data were filtered with SOAPnuke (version, 1.5.2). Reads were mapped to the reference genome (version, GCF_000001405.39_GRCh38.p13) and aligned using HISAT2 (version, 2.0.4) and Bowtie2 (version, 2.2.5) for comparison with the human reference genome.

Gene expression levels were calculated for both the A2780 and the A2780/Taxol cell lines in the present study using RSEM (version 1.2.8) to screen DEGs1 with |log2(Fold Change)|≥1 and FDR≤0.001. The gene expression level file GSE159791 (https://www.ncbi.nlm.nih.gov/geo/) of A2780/Taxol cells was downloaded from the GEO database (16). Combination with DEGs1, common genes were selected and analyzed using the R software package DESeq2 (version 1.42.1) with the A2780 cell as the control. DEGs were screened from common genes with |log2(Fold Change)|≥1 and P_adj<0.01. Results were visualized using the R packages pheatmap (version 1.0.12) and ggplot2 (version 3.4.3).

GO terms and pathways were obtained from the GO (ftp://ftp.ncbi.nih.gov/gene/DATA/gene2go.gz) and KEGG databases (version 101.0). The P-value was calculated using hypergeometric test in the function phyper of R, and the package qvalue (version 2.4.2) was used to perform a multiple positive test on the P-value. Q-value (corrected P-value) of <0.05 as the threshold to select the significantly enriched GO term and pathway, visualized with R language package GOplot (version 1.0.2) and ggplot2. The files of protein-protein interaction (PPI) analysis for DEGs in the pathway were downloaded from the STRING database (https://cn.string-db.org/), and visualized with Cytoscape (version 3.8.0).

A2780 and A2780/Taxol cells (6×10³ cells/well) were cultured into 96-well plates, treated with varying concentrations of Taxol solution (0, 30, 60, 120, 240, or 480 ng/ml), with five replication wells per group. After 48 h, 10 µl MTT (Biofroxx; NeoFroxx) was added to each well and incubated at 37°C for 4 h. The supernatant was discarded, and 150 µl DMSO solution was added to measure absorbance at 490 nm and calculate the IC₅₀.

The A2780 and A2780/Taxol cells were lysed to extract total protein with RIPA and PMSF solution (100:1 ratio; Beyotime Institute of Biotechnology). Total protein concentration was quantified with the BCA Protein Concentration Assay Kit (Beyotime Institute of Biotechnology) according to the manufacturer's instructions. The protein samples (20 µg per lane) were separated on 10% gels using SDS-PAGE. PVDF membranes (Beyotime Institute of Biotechnology) were blocked with non-fat powder milk for 2 h at room temperature and incubated overnight at 4°C with rabbit anti-MMP1 (cat. no. AF0231; Beyotime Institute of Biotechnology), rabbit anti-ZYX (cat. no. 38377; SAB Biotherapeutics, Inc.) and rabbit anti-UNC5C (cat. no. 44671; SAB Biotherapeutics, Inc.) diluted at a ratio of 1:1,000. Membranes were washed with TBST (contain 0.1% Tween) and incubated with HRP-conjugated affinipure goat anti-rabbit IgG (H+L) (1:1,000 dilution; cat. no. A0208; Beyotime Institute of Biotechnology) at room temperature for 2 h. Bands were visualized using a chemiluminescence kit (Beyotime Institute of Biotechnology). Protein hybridization results were observed using an automated chemiluminescence instrument (Tanon-5200; Ewell Biotechnology), and abundance was processed using Fiji (−win64; ImageJ, National Institutes of Health).

A total of 4×10⁴ cells/ml were cultured in 12-well plates overnight. Cells were fixed with 4% paraformaldehyde (Wuhan Servicebio Technology Co., Ltd.) for 1 h at room temperature. Cells were then washed with PBS, and 50–100 µl film breaking solution was added to the plate for incubation at room temperature for 20 min. A total of 3% BSA was added for blocking for 30 min at room temperature, followed by incubation in a wet box at 4°C overnight with the primary antibody (anti-MMP1: 1:100 dilution; cat. no. AF0231; Beyotime Institute of Biotechnology; anti-ZYX: 1:100 dilution; cat. no. 38377; SAB Biotherapeutics, Inc; and anti-UNC5C: 1:100 dilution; cat. no. bs-11493R; Beijing Bioss Biotechnology Co., Ltd.). The plate was incubated for 50 min with HRP-labeled secondary antibody (1:100 dilution; cat. no. GB23303; Wuhan Servicebio Technology Co., Ltd.), and freshly prepared DAB color developing solution was added controlling the color developing time. Hematoxylin was used as a counterstain at room temperature for 3 min, and the hematoxylin fractionation solution fractionated for a few sec. Hematoxylin re-blueing solution (Wuhan Servicebio Technology Co., Ltd.) was used for staining and rinsed. The slivers were dehydrated and sealed with neutral gum. After light microscopic examination, images were collected for analysis.

For the IF assay, the medium was aspirated, and the cell crawls were washed three times with cold PBS. After being fixed with 4% paraformaldehyde for 30 min, the cells were penetrated with membrane breaking working solution for 20 min. Subsequently, the cells were blocked with 3% BSA for 30 min (Wuhan Servicebio Technology Co., Ltd.), washed three times with PBS and incubated with the primary antibody (same as aforementioned immunohistochemistry antibodies) overnight. After washing three times with PBS, the cells were incubated with the secondary antibody (1:100 dilution; cat. no. E032420; EarthOx Life Sciences) coupled with DyLight 594-TFP ester for 2 h. After washing three times with PBS, the cells were stained with DAPI dye solution, and incubated for 10 min in the dark. Finally, the cell crawls were washed three times with PBS, and blocked with an anti-fluorescence quencher. Images were captured using a fluorescence microscope (BX51-32FL; Olympus Corporation). The average fluorescence density was calculated using Fiji.

Clinical information of patients with EOC and other cancers was obtained from TCGA database using the R package TCGAbiolinks (version 3.14). RNA-seq data and survival-related files of patients from TCGA database were downloaded using the UCSC Xena online tool (https://xenabrowser.net/datapages/) for subsequent analysis.

Kaplan-Meier survival curves were created and visualized using the packages survminer (version 0.4.9), survival (version 3.5–7) and TSHRC (version 0.1–6; http://cran.r-project.org/web/packages/TSHRC/TSHRC.pdf) in R (version 4.3.1). Statistical analysis was carried out using the log-rank test for Kaplan-Meier survival curves. One-way ANOVA and least significant difference tests were used for the statistical analysis and were performed using SPSS (version 26.0; IBM Corp.); visualization was carried out using GraphPad Prism (version 5; Dotmatics). P<0.05 was considered to indicate a statistically significant difference. A total of three biologically independent repeats were carried out, and the data are presented as mean ± SD.

The drug resistance index (RI) of the A2780/Taxol cell line in the present study was 33.62, exhibiting highly resistant characteristics (Fig. S1B). The common gene expression profile was similar to that of other Taxol-resistant cell lines in the GEO database, but different from that of the A2780 cell line [Figs. 1A and S1B; other RI values from the data in the study by Szenajch et al (16)]. Through differential expression analysis, 6,226 DEGs (DEGs1) were identified in the transcriptome sequencing data (Fig. 1B). The A2780 cells were used in the present study as control to reprocess the drug resistance data of the GEO database; 498 DEGs (DEGs2) were finally obtained based on DEGs1 (Fig. 1C; screening method, Fig. S1C and D; gene names, Table SI).

As chemoresistance is associated with a decreased survival rate of patients, genes in DEGs2 that were both up or downregulated in the A2780/Taxol and drug resistance datasets were selected for survival curve analysis. A total of 27 genes were found to be significantly associated with the OS of patients with EOC (Figs. 1D, S2 and S3). Considering that the log-rank test might lose power when the survival curves crossed at a later stage (17), a two-stage test (TS) weighted analysis was carried out for POLR3GL, ZNF239, FLRT3 and ZFHX4. The TS P-values of POLR3GL, ZNF239, FLRT3 and ZFHX4 were 0.66, 0.35, 0.35 and 0.28, respectively (Figs. S2 and S3). Since these values are >0.05, they were excluded from subsequent analysis. Of these survival-related genes, increased levels of GDF15 were found to be associated with enzalutamide and EPI-001 resistance in prostate cancer cells (18). NR1D2 was shown to be associated with enzalutamide resistance in neuroendocrine prostate cancer, and PHLDA1 was found to be associated with Lewis(y) highly expressing chemoresistant ovarian cancer cell (19,20). Imiquimod facilitates chemoresistance via the upregulation of MMP1, and RNA interference targeting ZYX reduces tumor cell HN12 resistance to cisplatin (DDP) (21,22). The downregulation of NCAM2 was shown to be associated with resistance to the monoclonal antibody drug trastuzumab in HER2⁺ breast cancer, and FBXL7 knockdown affects DDP resistance in nasopharyngeal carcinoma (23,24). In summary, these 23 genes were intimately associated with resistance to chemotherapy in various types of cancer, demonstrating the feasibility and accuracy of the screening performed in the present study for these Taxol resistance-associated genes that were closely associated with the OS of patients with EOC.

To investigate the role of DEGs on the basis of cell integral changes, the DEGs1 genes were subjected to GO and KEGG enrichment analysis. A Q value of <0.05 was considered to indicate significant enrichment. As shown by the KEGG enrichment analysis, the most significant pathway was ‘focal adhesion’, while in cellular component and biological process of GO, ‘cell junction’ and ‘cell adhesion’ were the significant pathways, indicating that they were associated with Taxol resistance (Figs. 2A and S4A-C). In addition, the results of the enrichment analysis indicated the involvement of the cell migration process (25–27), suggesting that cell migration affects the occurrence of Taxol resistance in A2780 cells. The 23 survival-related genes were mapped into the significant enrichment pathways, and only MMP1, ZYX and UNC5C were enriched in the ‘PPAR signaling’, ‘focal adhesion’ and ‘axon guidance’ pathways, respectively (Fig. 2B). Therefore, these three genes were used as gene targets for follow-up experiments.

Through KEGG pathway analysis, the enrichment pathways of MMP1, ZYX and UNC5C was identified. However, the mechanisms through which these genes play a role in the pathway remain unknown. Therefore, PPI network analysis was performed on the DEGs enriched in ‘PPAR signaling’, ‘focal adhesion’ and ‘axon guidance’ pathways. MMP1 interacted with angiopoietin-like 4 (ANGPTL4; Fig. 2C). It was hypothesized that MMP1 may affect cell migration through ANGPTL4, leading to the generation of drug resistance. The results from the study by Liao et al (28) confirmed that the downregulation of MMP1 hindered the migration and invasion of head and neck squamous cell carcinoma cells enhanced by EGF and recombinant ANGPTL4. ZYX interacted with paxillin (PXN) and vinculin (VCL) (Fig. 2D). It was hypothesized that ZYX may affect cell adhesion through PXN and VCL, thus mediating the generation of drug resistance. The study by Legerstee et al (29) on protein pairs related to the function of ‘focal adhesion’ demonstrated that the binding of ZYX and PXN, and that of VCL and vasodilator stimulated phosphoprotein affected cell adhesion and migration. In the PPI network of the ‘axon guidance pathway’ shown in Fig. 2E, FYN, the largest node, was the core of the network and interacted with UNC5C. According to literature, the knockdown of UNC5C enhances the phosphorylation of FAK and SRC (30), and FYN is a specific member of the SRC kinase family (31). Therefore, it was hypothesized that UNC5C mediated the generation of drug resistance by affecting Src kinase activity through FYN. Furthermore, these results confirmed the reliability of PPI network analysis to examine the interactions between proteins, and provided molecular information that the three targets may participate in the drug resistance mechanisms of EOC.

At the same time, in order to investigate the association of Taxol resistance of EOC with disease stage, the associations between MMP1, ZYX and UNC5C, and the disease stage of EOC (stage II, III and IV) were examined. The results indicated that MMP1, ZYX and UNC5C were not associated with the disease stage of patients with EOC (P>0.05, not statistically significant; Fig. S4D-F).

Chemotherapeutic drugs for EOC, in addition to Taxol, include CBP, DDP and DOX. CBP, DDP and DOX limit DNA replication and transcription via various mechanisms. Although DNA replication affected by CBP, DDP and DOX occurs during the S phase, and Taxol induces G2/M phase accumulation, these drugs ultimately lead to apoptosis, and they may share common signaling pathways and networks in the final stage (32–34). Therefore, it was hypothesized that Taxol resistance may cause resistance to other chemotherapeutic drugs as well. In the present study, CBP, DDP and DOX, as well as Taxol, were selected to analyze the association between MMP1, ZYX and UNC5C and the OS of patients with EOC in TCGA. The results revealed that MMP1, ZYX and UNC5C were significantly associated with the survival of patients treated with CBP, DDP and DOX, respectively, and this trend was consistent with Taxol, indicating that Taxol may share key resistance-related targets with CBP, DDP and DOX (Fig. 3). This finding preliminarily confirmed one of the ways that Taxol resistance causes resistance to other chemotherapeutic drugs.

Taxol is also used clinically in the treatment of patients with LUNG, HSSC and UCEC, as well as other types of cancer (35–37). Regarding HNSC, LUNG and UCEC, patients with a history of treatment with Taxol in TCGA, excluding patients with free tumor tissue, were selected for Kaplan-Meier survival curve analysis to investigate the association between MMP1, ZYX and UNC5C, and the OS of other patients with cancer treated with Taxol. The results revealed that only UNC5C was significantly associated with the survival of patients with HNSC treated with Taxol. MMP1, ZYX and UNC5C were not significantly associated with the survival of patients with LUNG and UCEC treated with Taxol (Fig. 4). The results indicated that MMP1 and ZYX may be specific in Taxol resistance in EOC, while the common resistance occurrence and mechanisms of UNC5C in EOC and HNSC remain to be explored.

The protein expression of MMP1 and ZYX was significantly increased in A2780/Taxol cells (P<0.05 and P<0.01, respectively), while the expression of UNC5C was significantly decreased (P<0.05; Fig. 5A). The same results were obtained by immunofluorescence and immunohistochemistry (Fig. 5B and C), and were consistent with the RNA-seq data. This demonstrated the reliability of the selection of Taxol resistance-associated gene targets through RNA-seq, and the stability and accuracy of the multiple validation work. These in vitro cellular results confirmed that the expression of MMP1 and ZYX was significantly upregulated, and the expression of UNC5C was significantly downregulated in A2780/Taxol cells, demonstrating that the three gene targets are potential and promising molecular markers of Taxol resistance in EOC.