Big data-based identification of methylated genes associated with drug resistance and prognosis in ovarian cancer.

It is imperative to further the understanding of the drug resistance mechanisms of ovarian cancer (OC) and to identify useful biological markers for prognosis prediction.Cormine, cBioportal, and The Cancer Genome Atlas databases were used to search microarray data of gene methylation related to OC, drug resistance in OC, and prognosis, and to analyze methylated genes potentially inducing the drug resistance in OC. Fifty-five DNA-methylated genes significantly associated with drug resistance in OC were screened, and the regulatory mechanisms underlying changes in methylation levels of these genes were systematically integrated.Enrichment and annotation of biological processes indicated that most of the above DNA-methylated genes were significantly associated with cell proliferation and cell cycle. In addition, pathway enrichment demonstrated that the above DNA-methylated genes were significantly associated with PI3K-AKT and P53 signaling pathways. Among the 55 genes, 4 were significantly associated with OC prognostic disease-free survival, namely bromodomain containing 4, PDZ domain containing 1 (PDZK1), phosphatase and tensin homolog, and TNF receptor superfamily member 10c; 5 were significantly related to overall survival, namely bromodomain containing 4, PDZK1, PIK3C2B, Rh associated glycoprotein, and DYRK; among them, the degree of methylation of TNF receptor superfamily member 10c, PDZK1, and Rh associated glycoprotein genes was significantly correlated with mRNA expression. Furthermore, PDZK1, Rh associated glycoprotein, and TNF receptor superfamily member 10c genes showed significant hypomethylation in drug-resistance tissues of OC, and their mRNAs had significantly high expression.The association between the methylation of these 55 genes and OC and drug resistance in OC, in addition to bioinformatics analyses clarify the important mechanisms of gene methylation in the development, progression, and drug resistance of OC.

Introduction

Ovarian cancer (OC) is a common malignant tumor that poses a serious threat to women's health. The main type of OC is epithelial malignant tumor, which accounts for 85% to 90% of all ovarian malignancies and has the highest mortality rates among female malignant tumors of genital tract.[ Over the past decade, significant advances have been made in the research of risk factors and molecular pathways of OC, resulting in numerous opportunities for prevention, detection, early diagnosis, prognostic prediction, and treatment of OC. Nevertheless, the 5-year survival rate of patients with OC remains around 50%.[ In 2018, the Clinical Practice Guidelines by the American Cancer Society pointed out that there were approximately 22,240 newly diagnosed patients with OC in the United States each year, among whom approximately 14070 would die within 5 years.[ Under these circumstances, it is of utmost importance to further the understanding of the mechanisms of progression, development, and drug resistance in OC, and to explore effective biomarkers and therapeutic targets.

Over recent years, the genome sequencing and biotechnology have flourished, greatly enhancing the potential of cancer genome research. The biomedical studies have reported lots of information, including mRNA expression, DNA methylation, microRNA expression, etc., thus providing a reliable basis for a variety of new studies. Gene function prediction based on bioinformatics analysis is a potential, feasible, and valuable method for exploring gene functions, and related networks.[ The Cancer Genome Atlas (TCGA) is an unprecedented, data-disclosed multi-digit data set of cancer genome co-founded by the National Cancer Institute and the National Human Genome Research Institute of National Institutes of Health[ that provides researchers with new opportunities to study cancer development and progression. Data from TCGA can be used to improve diagnostic methods, develop new treatment standards and novel targeted therapies.

In the present study, a group of genes associated with the progression and prognosis of OC were identified using DNA methylation, mRNA expression, and clinical data extracted from the TCGA cohort.

Methods

Ethical approval and informed consent is not applicable in this study.

The data in this study were obtained from the Coremine Medical (http://www.cormine.com), cBioportal (http://cbioportal.org), and TCGA (The Cancer Genome Altas) databases, including microarray data on DNA-methylated genes of specimens obtained during operation from 97 platinum-resistant patients with OC and 190 platinum-sensitive patients with OC, mRNA expression data, and clinical data related to the prognosis of OC. The cBioPortal platform was employed for its suitability to analyze complex cancer genomics and clinical profiles such as the TCGA database.[ Before inclusion in the cBioPortal platform, all methylation data were submitted to checksum standardization.[

SPSS 22.0 software was used for data analyses. Independent samples t-test with Benjamini-Hochberg correction, with a threshold false discovery rate (FDR) of 0.1, was used to analyze the methylation degree of DNA-methylated genes in drug-sensitive and drug resistant patient groups, while 55 methylated genes significantly associated with the drug resistance in OC were screened.

DAVID and String online software were used for enrichment analysis and gene/protein interaction analysis of biological process for differentially methylated genes in drug-resistant tissues in 55 OC genes. The common biological pathways and related genes were screened by a corrected P < .05. In addition, the Benjamini method was applied for P value adjustment.

Based on the TCGA database, the genes corresponding to methylation changes and mRNA expression in drug-resistant tissues of OC were screened via integration analysis.

Kaplan-Meier survival analysis was performed for prognostic analysis of the 55 genes, and survival curves had been drawn. Four genes related to prognosis of ovarian cancer PFS were screened, and 5 significantly associated with OS were selected. There were two genes that were related to both PFS and OS.

Results

Screening of methylated genes related to drug resistance in ovarian cancer (OC)

Genes were derived from the Coremine Medical database, where OC, ovarian carcinoma, methylation, resistant or resistance or chemoresistance were used as keywords. The advanced nested searches were performed to retrieve genes associated with OC, gene methylation, and chemotherapeutic resistance in OC, resulting in a total of 1407 methylated genes significantly associated with OC and drug resistance in OC. The selected genes were put into cBioportal for examination, revealing that a total of 885 genes had data related to DNA methylation, mRNA expression and clinical data in OC tissues.

Microarray data of DNA methylation in 490 patients with OC from the TCGA database, related data on mRNA expression, and clinical prognoses were downloaded, including data from 90 platinum-resistant patients with OC and 197 platinum-sensitive patients with OC, which included 885 methylated genes screened. The t test was used to analyze the data, and a total of 55 differentially methylated genes markedly associated with drug resistance in OC were screened (Table 1), among which there were 27 genes with hypermethylation in drug-resistant tissues, and 28 with hypomethylation in drug-resistant tissues. Among these 55 genes, previous studies have reported that hypermethylation changes of genes were clearly associated with drug resistance in OC, including death associated protein[ and estrogen receptor 1 (ESR1),[ midkine (MDK),[ myelin and lymphocyte,[ phosphatase and tensin homolog (PTEN),[ Ras association domain family member 1 (RASSF1),[ and homeobox A11 (HOXA11);[ hypomethylation changes of genes were clearly related to drug resistance in OC, including carbonic anhydrase 9,[ cyclin dependent kinase 2 (CDK2),[ and ETS proto-oncogene 1 (ETS1).[

Table 1

Screening of 55 significantly differentially methylated genes in drug resistance of ovarian cancer.

Bioinformatics analysis of differentially methylated genes for drug resistance in OC

Pathway enrichment analysis of differentially methylated genes for drug resistance in OC

DAVID online software was used to perform a biological process analysis of the above 55 differential DNA-methylated genes.[ The results are shown in Table 2. In the cluster of 2 biological processes annotated to the highest score, cluster1 was mainly a biological process related to cell proliferation, covering 7 genes, while cluster2 was mainly related to the cell cycle process, covering a total of 3 genes. These results suggested that the above 55 differential DNA-methylated genes may be involved in the development, progression, and drug resistance in OC mainly via regulation of cell proliferation and cell cycle. There were 4 enriched pathways, ie, PI3K-Akt signaling pathway, p53 signaling pathway, FoxO signaling pathway, and Cell cycle, which were clearly related to drug resistance in OC. Previous studies have reported the relevance of PI3K-Akt signaling pathway,[ p53 signaling pathway,[ Cell cycle,[ and FoxO signaling pathway in other cancers, as well as tumor resistance.[

Table 2

Screening of 55 significantly differentially methylated genes in drug resistance of ovarian cancer.

Protein interaction analysis of the 55 differentially methylated genes

In order to further analyze the association between differential DNA-methylated genes and drug resistance in OC, a comprehensive analysis of the interaction of 55 gene proteins by the String tool was performed as shown in Figure 1. In addition to the transmembrane protein 158, kinesin family member 1C, membrane bound transcription factor peptidase, site 1, DNA topoisomerase III beta, PDZ domain containing 1 (PDZK1), solute carrier family 19 member 1, CDC42 binding protein kinase beta (CDC42BP) and lysosomal protein transmembrane 4 beta genes, the direct or indirect interactions were observed among the remaining 47 protein genes, which indicated that these genes might be involved in the regulation of drug resistance in OC. Among these protein genes, ESR1 interacted directly with other 21 genes, and PTEN interacted directly with other 17 genes, while MDM2 proto-oncogene (MDM2) and cluster of differentiation 44 (CD44) interacted directly with 14 genes. As a classical gene, PTEN has been widely studied in drug resistance in OC. Numerous studies have shown that PTEN, as a key link, is involved in the PI3K/AKT/PTEN drug-resistance pathway and participates in drug resistance in OC by affecting biological processes, such as DNA methylation, cell cycle, apoptosis, and cell proliferation.[ In addition, it had been reported that changes in gene methylation of ESR1[ are involved in the development and drug resistance in OC and are associated with prognosis. With reference to MDM2 and CD44, although no previous studies have shown that changes in methylation are associated with OC and drug resistance in OC, numerous studies have indicated that both MDM2 and CD44 genes are involved in drug resistance in OC, and are related to the prognosis of OC. Based on these and our findings, it can be concluded that MDM2 and CD44 genes may be involved in the development of OC and drug resistance in OC through changes in gene methylation.

Figure 1

Protein interaction network analysis of 55 methylated genes related to drug resistance in ovarian cancer.

Analysis of the negative correlation between methylation level of differentially methylated genes for drug resistance in OC and mRNA expression

Based on the TCGA database, 10 genes corresponding to the methylation degree of differential methylated genes and mRNA expression from the above 55 genes were further screened. Among these, 5 were hypermethylated in drug-resistant tissues, including calpain 5, cathepsin L (CTSL), ETS1, membrane bound transcription factor peptidase, and syndecan 1. The mRNA expressions of these 5 genes were decreased in drug-resistant tissues; coagulation factor II, PDZK1, Rh associated glycoprotein (RHAG), TNF receptor superfamily member 10c (TNFRSF10C), and orosomucoid 2 showed hypomethylation in drug-resistant tissues, and mRNA expressions of these 5 genes were up-regulated in drug-resistant tissues, as shown in Figure 2. In addition, CTSL has been previously reported to be clearly associated with drug resistance in OC.[

Figure 2

Analysis of gene methylation level and mRNA expression in drug-sensitive and resistant tissues of ovarian cancer.

Association between methylation level of differentially methylated genes for drug resistance in OC and prognosis of OC

Univariate Kaplan-Meier analysis was performed for the association between the changes in the above 55 gene methylation and clinical prognosis of ovarian epithelioma (including disease-free survival [DFS] and overall survival [OS]). The results are shown in Table 3 and Figures 3–5. The methylation of PTEN and TNFRSF10C genes had an effect on the prognostic DFS in OC (P < .05), and the 3 genes, phosphatidylinositol-4-phosphate 3-kinase catalytic subunit type 2 beta (PIK3C2B), RHAG, and dual specificity tyrosine phosphorylation regulated kinase 2 (DYRK2), had effects on the prognostic OS for OC (P < .05). The methylation of bromodomain containing 4 (BRD4) and PDZK1 genes had effects on the prognostic DFS and OS of OC (P < .05). Among them, the hypomethylation or demethylation of 4 genes, BRD4, PDZK1, PTEN, and TNFRSF10C, which had effects on the prognostic DFS of ovarian cancer, shortened the median survival time of DFS by 3 to 5 months in OC. The hypomethylation or demethylation of the 5 genes, BRD4, PDZK1, PIK3C2B, RHAG, and DYRK2, which had effects on prognostic OS of OC, resulted in a 7 to 10 month reduction in the median survival time of OS in ovarian cancer. The methylation changes of these 7 genes in the methylation degree and mRNA expression in drug-resistant and sensitive tissues of OC are shown in Figure 6. Among them, methylation of PDZK1, RHAG, and TNFRSF10C corresponded to mRNA expression, and these 3 genes showed hypomethylation and high mRNA expression in drug-resistant tissues of OC.

Table 3

Methylated genes associated with the prognosis of ovarian cancer.

Figure 3

DNA-methylated genes related to the disease free survival of ovarian cancer.

Figure 5

DNA-methylated genes related to the disease free survival and disease free survival of ovarian cancer.

Figure 6

Analysis of gene methylation level and mRNA expression related to the prognosis of ovarian cancer.

DNA-methylated genes related to the overall survival of ovarian cancer.

Discussion

At present, OC has high fatality rates, and the predominant cause of chemotherapy failure is the formation of multidrug resistance (MDR). DNA methylation is 1 of the epigenetic mechanisms that has an important regulatory role in the development, progression, and drug resistance of OC. Numerous studies have reported on the association between DNA methylation and OC and drug resistance in OC are available, showing that drug resistance in OC may occur through cell cycle, cell proliferation, and apoptosis. In the current study, we used TCGA database to systematically integrate the association of the 55 methylated genes with OC, MDR, and clinical prognosis. This study can be used as sort of a guide for the clinical selection of chemotherapy drugs, early detection, and prognosis. In addition, we performed bioinformatics analysis on these 55 genes, including annotation of biological processes and protein interaction, thus providing a basis for the overall study of DNA-methylated genes involved in the OC and drug resistance.

Based on the TCGA database, 55 differentially methylated genes in drug-resistant and sensitive tissues of OC were screened. Among the 55 genes, BRD4,[CD44,[CTSL,[DYRK2,[ MCL1 apoptosis regulator,[MDM2.[ proliferation and apoptosis adaptor protein,[PTEN,[RASSF1A,[TNFRSF10C,[MDK,[HOXA11,[ and lysosomal protein transmembrane 4 beta [ have been identified as clearly associated with drug resistance in OC. Furthermore, previous studies have reported that hypermethylation changes of death associated protein, ESR1, MDK, myelin and lymphocyte, PTEN, RASSF1, and HOXA11 genes are clearly related to drug resistance in OC, while hypomethylation changes of carbonic anhydrase 9, CDK2, and ETS1 are clearly related to drug resistance in ovarian cancer.

Bioinformatics analysis was performed for 55 genes that were significantly methylated in OC tissues. DAVID analysis revealed that the biological processes in which 55 genes had the highest score were cell proliferation and cell cycle processes (Table 2). The pathways with the highest score were PI3K-Akt signaling pathway and p53 signaling pathway (Table 2). These results suggested that these methylated genes may be involved in the regulation of drug resistance in OC mainly via regulation of cell proliferation and cell cycle processes as well as via PI3K-Akt and p53 signaling pathways. In addition, protein interaction analysis revealed that the PTEN gene directly interacted with other 17 genes, ESR1 gene directly interacted with other 21 genes, and that most of the 55 genes had direct or indirect interactions with PTEN or ESR1. It also suggested that many genes were involved in the drug resistance of OC via interaction with PTEN and ESR1 gene. These results were consistent with the existing studies.

As shown in Table 2, at least seven genes including ETS1 and PTEN were involved in the regulation of drug resistance in OC through the cell proliferation pathway, and at least 3 genes participated in the regulation of drug resistance in OC through the cell cycle pathway. Existing studies have shown that many genes are involved in the regulation of drug resistance in OC through cell proliferation and cell cycle processes, as well as via PI3K-Akt and p53 signaling pathways. For example, the proto-oncogene ETS1 methylation participates in the drug resistance in ovarian cancer through multiple physiological and pathological processes, such as phylogeny, stimulus response, vascular endothelial growth factor production, morphogenesis, cell proliferation, cell adhesion, and signal transduction.[MDM2, as a suppressor gene of P53 can lead to drug resistance by inhibiting the P53 pathway. The use of drugs to inhibit the binding of MDM2 to P53 can effectively activate P53, which leads to increased cell cycle and apoptosis, thereby reversing drug resistance.[ In addition, upregulation of PTEN increases P53 levels by inhibiting the AKT pathway and MDM2, leading to cell cycle G(2)/M arrest and apoptosis.[ Further evidence comes from the gene proliferation and apoptosis adaptor protein, which participates in the development and progression of OC by regulating cell proliferation, autophagy, apoptosis and glucose metabolism. Its overexpression can destroy the microtubule instable protein function, thereby promoting paclitaxel-induced mitosis and apoptosis and increasing the sensitivity of tumor cells to paclitaxel.[ Similarly, CDK2 is considered to have an important role in regulating cell cycle, particularly when changing from G1 to S, and it is known to be involved in cancer progression and cancer cell proliferation. In OC tissues, CDK2 methylation leads to significant decrease in the mRNA expression, thereby participating in the development and progression of OC via cell cycle pathway.[ The above findings indicate that cell proliferation and cell cycle processes might be important mechanisms through which DNA-methylated genes regulate the drug resistance in OC. Accordingly, an in-depth study of the regulating effect of DNA-methylated genes on cell proliferation and cell cycle could provide more possibilities to address MDR in ovarian cancer. After adjustment by the Benjamini method, P values were above 0.05 for the identified genes (Table 2). Since the actual P value represents an important parameter in the DAVID's statistical model, these genes might still have a certain biological significance in drug resistance. Studies have been initiated by our team to confirm these findings.

Previous studies have reported the importance of DNA-methylated genes in prognosis of OC. In the current study, survival analysis of 55 differential DNA-methylated genes revealed that the methylation of PTEN and TNFRSF10C genes affected the prognostic DFS of ovarian cancer; PIK3C2B, RHAG, and DYRK2 genes had effect on the prognostic OS of OC. The methylation of 2 genes, BRD4 and PDZK1, had effects on the prognostic DFS and OS in ovarian cancer. Previous studies have shown that among these seven genes, DYRK2 has low expression in cisplatin-resistant tissues of OC and is associated with low postoperative overall survival rate in patients with OC.[ Other genes have also been studied in OC tissues. For example, as the transcriptional regulator of epigenetics, BRD4 is involved in the development and progression of ovarian epithelial cancer and drug resistance.[ The protein encoded by BRD4 has a BET structure, and administration of a BET inhibitor can effectively inhibit the proliferation of ovarian epithelial cancer cells and reverse the drug resistance by blocking the PIK3-AKT and ERK pathways. Previous studies have also shown that the low expression of TNFRSF10C is significantly related to the chemotherapy resistance of serous OC. In addition, there are also reports on the gene expression and prognosis of ovarian cancer for other genes. For example, the degree of methylation of ESR1 in normal ovarian tissues is higher than that in OC tissues,[ and patients with hypermethylation of ESR1 in ovarian cancer tissues have better OS,[ which is consistent with our results. The association between the degree of ESR1 methylation and the prognosis of OC is more significant in serous and mucinous tissues.[ However, some studies have also shown that in low-grade serous ovarian cancer, the degree of methylation of the ESR1 promoter is negatively related to survival rate.[ Another example is that ETS1 is a key regulator of the PARP1 gene in BRCA1 mutant ovarian cancer, and its degree of methylation is significantly associated with the prognosis of OC.[ In addition, CD44-positive cells of OC are associated with poor prognosis of OC.[ Some studies have shown that CD44+ phenotypic stem cells of OC are more invasive and have higher tendency to be resistant to paclitaxel and platinum.[ At the same time, the CD44+ phenotypic stem cells of ovarian cancer have higher autophagy, and when platinum drugs are used for anti-tumor, they can inhibit autophagy of cancer stem cells, improve the efficacy of chemotherapy, and effectively prevent drug resistance.[ Similarly, MLH1, RASSF1A, HOXA10, and HOXA11 are hypermethylated in OC tissues, and after the use of demethylating drug decitabine, PFS in patients with ovarian cancer is prolonged, and their prognosis is improved, while the sensitivity of tumor tissue to carboplatin chemotherapy is restored.[

Among the seven genes related to the prognosis of ovarian cancer, PDZK1, RHAG, and TNFRSF10C showed significant hypomethylation in drug-resistant tissues of OC, indicating that these 3 genes may be involved in the drug resistance of OC through changes in gene methylation. Some scholars have previously confirmed that the low expression of TNFRSF10C is associated with chemotherapy resistance of OC.[ Yet, there are only few reports on the association between the other 2 genes with ovarian cancer and drug resistance in OC. One such study has reported that for the TNFRSF10C in melanoma cells and tissues, hypermethylation of the promoter is involved in tumorigenesis and progression.[ Another example is PDZK1, a scaffold protein that binds to plasma membrane proteins and regulatory elements, regulating its surface expression in the apical region of epithelial cells, which is considered an oncogene of breast cancer.[ Overexpression of PDZK1 is associated with resistance to paclitaxel, 5-fluorouracil, and etoposide in breast cancer.[ In pheochromocytoma and paraganglioma, methylation of PDZK1 is involved in tumor formation through apoptosis and cell invasion pathways.[ Low expression of PDZK1 in renal clear cell carcinoma is related to recurrence, metastasis, and poor prognosis of patients.[PDZK1 expression is significantly down-regulated in pancreatic cancer tissues. Furthermore, low expression of PDZK1 is significantly associated with poor prognosis in pancreatic cancer. In vitro studies have shown the overexpression of PDZK1 in combination with PTEN can significantly reduce phosphorylation of PTEN and AKT, thus effectively reversing the proliferation and migration of pancreatic cancer cells.[ In addition, the protein expressed by RHAG is an erythrocyte-associated glycoprotein, and the silencing expression of RHAG has been previously identified in esophageal cancer tissues.[

Conclusions

PDZK1, RHAG, and TNFRSF10C genes are likely to be involved in the drug resistance in OC through changes in gene methylation, and may become markers for predicting the prognosis of OC.

Author contributions

Hua Bai contributed to the conception of the study; Bingbing Yan performed the data analyses and wrote the manuscript; Chunqiu Xiong contributed significantly to analysis and manuscript preparation; Feifeng Huang helped perform the analysis with constructive discussions; Mingming Zhang contributed significantly to Literature research; Yan Mo contributed to the date acquisition.