Abstract
Background/Aim: Prostate cancer remains a major global health burden, with treatment resistance posing a significant challenge. Enhancer of zeste 2 polycomb repressive complex 2 subunit (EZH2), a histone methyltransferase, is frequently overexpressed in prostate cancer, contributing to tumor progression and castration resistance. Clinical trials of EZH2 inhibitors may have therapeutic benefits. This study aimed to evaluate the impact of genetic variants in EZH2-related genes on survival outcomes in prostate cancer.
Patients and Methods: We conducted a genetic association study evaluating 76 single nucleotide polymorphisms (SNPs) across 10 EZH2-related genes in 630 patients with prostate cancer undergoing androgen deprivation therapy (ADT). Functional analyses, including gene ontology and pathway enrichment assessments, were performed to elucidate the biological significance of key genes across multiple datasets.
Results: DNMT3A rs77993651 was significantly associated with both cancer-specific survival [hazard ratio (HR)=0.82, p=0.042] and overall survival (HR=0.80, p=0.011). Functional annotation indicated that rs77993651 resides within enhancer histone marks, potentially regulating DNMT3A expression. Elevated DNMT3A expression was observed in prostate tumor tissues and correlated with more aggressive features and shorter progression-free survival. Gene set enrichment analysis revealed that DNMT3A expression was strongly associated with cell cycle G2/M checkpoint regulation, implicating a role in prostate cancer progression.
Conclusion: The prognostic significance of DNMT3A and its genetic variant rs77993651 in prostate cancer is herein highlighted. Targeting DNMT3A-mediated pathways may offer novel therapeutic strategies for prostate cancer management.
Introduction
Prostate cancer is a significant health issue, particularly among older men, with an estimated 313,780 new cases and 35,770 deaths in the United States by 2025 (1). The disease disproportionately impacts African-American men and individuals of Caribbean descent. Treatment options vary and include watchful waiting, surgery, radiotherapy, androgen deprivation therapy (ADT), chemotherapy, radiopharmaceuticals, and proton beam radiation. The choice of treatment depends on the cancer stage, patient’s overall health, and personal preferences. Predicting the progression of prostate cancer involves several clinical factors, such as age, prostate-specific antigen (PSA) level, family history, smoking status, alcohol consumption, and cholesterol level (2). Elevated PSA levels are significant, and combining these with other risk factors can improve prediction accuracy. However, the management of prostate cancer presents challenges, including difficulties in early detection, complex treatment decisions, treatment resistance, and the impact of individual patient factors. Genetic factors are important, with approximately 60% of prostate cancer risk attributed to inherited factors (3). Germline mutations in genes, such as BRCA1 and BRCA2, are associated with a higher risk and more aggressive forms of cancer (4). Genetic testing can help identify individuals at high risk and guide personalized treatment strategies (5, 6). Despite advancements, the slow progression of prostate cancer and potential for treatment resistance highlight the need for further research and novel therapeutic approaches.
Enhancer of zeste 2 polycomb repressive complex 2 subunit (EZH2), a histone methyltransferase, aids in prostate cancer by catalyzing the trimethylation of histone H3 at lysine 27, leading to gene silencing (7). Overexpression of EZH2 is frequently observed in prostate cancer and associated with aggressive tumor behavior, metastasis, and poor prognosis (8). Mechanistically, EZH2 drives cancer progression by silencing tumor suppressor genes, interacting with androgen receptors (AR) to promote castration-resistant prostate cancer, and methylating nonhistone proteins involved in DNA damage repair and apoptosis (9, 10). DNA methyltransferases (DNMTs), which add methyl groups to DNA, contribute to tumor progression by cooperating with EZH2 to silence tumor-suppressor genes (11, 12). Clinical trials of EZH2 inhibitors, such as mevrometostat, have demonstrated promising results, particularly combined with enzalutamide, leading to improved median radiographic progression-free survival and a manageable safety profile (13). Single nucleotide polymorphisms (SNPs) in EZH2 and DNMT may significantly influence cancer risk and prognosis (14). For example, EZH2 rs2302427 is associated with susceptibility to cancer, including prostate cancer (15), whereas DNMT3B rs1569686 has been implicated in the risk of gastric cancer (16).
Given the emerging role of EZH2-related genes in cancer progression, we hypothesized that genetic variants of these genes may influence survival outcomes in patients with prostate cancer undergoing ADT. Therefore, this comprehensive genetic association study aimed to evaluate 76 SNPs across 10 EZH2-related genes in a cohort of 630 patients with prostate cancer. To further elucidate the biological mechanisms underlying these associations, we performed functional analyses, including gene ontology and pathway enrichment assessments, to explore the role of EZH2-responsive genes in prostate cancer progression.
Patients and Methods
Patient and response assessment. This study included 630 patients with prostate cancer who underwent ADT at the National Taiwan University Hospital, Kaohsiung Medical University Hospital, and Kaohsiung Veterans General Hospital (17, 18). The study was approved by the Institutional Review Board of the Kaohsiung Medical University Hospital (KMU-HIRB-2013132) and conducted in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice guidelines. Written informed consent was obtained from all participants. Clinicopathological data were extracted from hospital medical records, and the endpoints of the study were cancer-specific survival (CSS) and overall survival (OS), defined as the time from the start of ADT to death from prostate cancer or any other cause. Over a median follow-up period of 165.8 months, 414 patients died, with 314 succumbing to prostate cancer (19). Clinical factors, such as age, PSA level at ADT initiation, clinical stage, Gleason score at diagnosis, PSA nadir, and time to PSA nadir, were significantly associated with OS and CSS (p<0.05).
SNP selection and genotyping. Haplotype-tagging SNPs (htSNPs) were selected to capture the majority of genetic variability across 10 EZH2-related genes, including BMI1 polycomb ring finger proto-oncogene (BMI1), DNA methyltransferase 1 (DNMT1), 3A (DNMT3A), 3B (DNMT3B), embryonic ectoderm development (EED), enhancer of zeste 2 polycomb repressive complex 2 subunit (EZH2), histone deacetylase 1 (HDAC1), jumonji and AT-rich interaction domain containing 2 (JARID2), RB binding protein 4, chromatin remodeling factor (RBBP4), RB binding protein 7 (RBBP7), and SUZ12 polycomb repressive complex 2 subunit (SUZ12). Haploview 4.2 and the tagger algorithm (20) were used for htSNP selection, based on genotype data from the 1000 Genomes Project for Han Chinese in Beijing and Southern Han Chinese populations. htSNPs were identified with a minor allele frequency (MAF) >0.05 and a pairwise linkage disequilibrium threshold of r2>0.8, ensuring efficient coverage of common variants with a minimal set of SNPs. Genomic DNA was extracted from whole blood using the QIAamp DNA blood kit (Qiagen, Germantown, MD, USA) and genotyped at the National Center for Genome Medicine using the Affymetrix Axiom Genotyping Array system (Thermo Fisher Scientific, Waltham, MA, USA) (21). SNPs with MAF <0.03, genotyping call rates <0.94, and significant deviations from Hardy–Weinberg equilibrium (p<0.0001) were excluded. The final analysis included 76 htSNPs.
Bioinformatic analyses. To functionally annotate rs77993651 and investigate its potential regulatory roles, we integrated data from HaploReg v4.2 and FIVEx databases (22, 23). HaploReg was used to assess the impact of rs77993651 on regulatory elements, transcription factor binding, and evolutionary conservation. Expression quantitative trait loci (eQTL) analysis was performed via FIVEx, using linear regression models to associate the rs77993651 genotype with DNMT3A expression. To explore DNMT3A expression levels and their correlation with prostate cancer outcomes, publicly available datasets, including PCaDB (24), the Gene Expression Database of Normal and Tumor Tissues 2 (25), and The Cancer Genome Atlas Prostate Adenocarcinoma (TCGA PRAD), were used to examine the clinical significance of DNMT3A in prostate cancer. The molecular mechanisms linked to DNMT3A were explored using LinkedOmics via gene ontology and hallmark pathway enrichment analyses employing gene set enrichment analysis (GSEA) (26). GSEA, using Pearson correlation coefficients to rank genes based on their correlation with DNMT3A expression, identified enriched gene sets by a weighted Kolmogorov–Smirnov-like running sum statistic (27). Statistical significance was determined using 1000 gene set label permutations and Benjamini–Hochberg false discovery rate (FDR) correction. The prognostic significance of DNMT3A and G2/M checkpoint genes, including aurora kinase B (AURKB), cyclin A2 (CCNA2), and cyclin-dependent kinase 1 (CDK1), was assessed using TCGA PRAD data.
Statistical analyses. Statistical analyses were conducted using Statistical Product and Service Solutions version 19.0.0 (IBM, Armonk, NY, USA), with statistical significance set at two-sided p<0.05. Kaplan–Meier analysis and log-rank tests were used to compare survival curves, whereas univariate and multivariate Cox regression analyses were performed to assess the associations between clinico-pathological features and patient prognosis, with hazard ratios (HRs) and 95% confidence intervals (CIs). The correlation between DNMT3A expression and tumor characteristics was evaluated using Spearman’s and Pearson’s correlations. A pooled analysis of DNMT3A expression in prostate cancer and normal tissues was conducted using Review Manager version 5.4.1 (Cochrane, London, UK) by applying a random-effects model to account for study heterogeneity, with standardized mean differences (SMDs) and corresponding 95% CIs.
Results
To explore the relationship between EZH2-related genes and prostate cancer progression, we examined the association of 76 htSNPs within 10 functional partner genes of EZH2 with CSS and OS in patients with prostate cancer. Among these SNPs, two (rs77993651 in DNMT3A and rs794782 in JARID2) were significantly associated with CSS, whereas four (rs77993651 in DNMT3A, rs17576436 and rs17576233 in JARID2, and rs2288937 in DNMT1) were significantly associated with OS (p<0.05, Figure 1). Notably, rs77993651 in DNMT3A was the only SNP that showed a significant correlation with both CSS and OS. Specifically, each additional minor A allele of rs77993651 corresponded to a 18% reduction in cancer-specific mortality risk (HR=0.82, 95% CI=0.68-0.99, p=0.042, Table I) and 20% decrease in all-cause mortality risk (HR=0.80, 95% CI=0.68-0.95, p=0.011). Multivariate analysis further confirmed that rs77993651 in DNMT3A was an independent prognostic factor for OS after adjusting for clinical covariates (p=0.025, Table I).
Manhattan plots depicting the association between 76 single-nucleotide polymorphisms (SNPs) across 10 EZH2-related genes and (A) cancer-specific survival and (B) overall survival in patients with prostate cancer undergoing androgen deprivation therapy. The Y-axis represents −log10(p) values, whereas the X-axis indicates chromosomal positions of the SNPs. The nominal significance threshold (p=0.05) (blue horizontal line marks). SNPs meeting the significance criteria (red circles).
The association of DNMT3A rs77993651 with cancer-specific and overall survival in patients with prostate cancer receiving androgen deprivation therapy.
To assess the potential functional role of rs77993651, we used HaploReg, which suggested that this SNP is located within enhancer histone marks across various tissues, indicating a possible regulatory function (Figure 2A). The FIVEx database provided additional support, revealing that the protective A allele of rs77993651 was linked to decreased DNMT3A expression in memory and naïve regulatory T cells (p≤0.039, Figure 2B). However, direct eQTL analysis of the prostate tissue remains unavailable.
Functional impact of DNMT3A rs77993651. (A) Regulatory annotations for rs77993651 derived from HaploReg. (B) Expression quantitative trait loci analysis demonstrates the relationship between rs77993651 and DNMT3A expression levels across various human tissues. The nominal significance threshold (p=0.05) (blue horizontal line marks). The negative (inverted triangle) and non-significant (circles) effects of rs77993651 on DNMT3A expression. Sample sizes for each subgroup (numbers in brackets).
To determine the clinical significance of DNMT3A expression in prostate cancer, we analyzed 2,461 tumor samples and 986 normal prostate tissues across 35 public datasets. DNMT3A expression was significantly elevated in tumor tissues compared to normal tissues (SMD=0.21, 95% CI=0.01-0.40, p=0.04, Figure 3). Furthermore, analysis of TCGA PRAD data revealed a strong correlation between increased DNMT3A expression and higher Gleason scores and advanced tumor stage (p<0.001, Figure 4, left and middle). Moreover, patients with higher DNMT3A expression exhibited significantly shorter progression-free survival in both the GSE21032 and TCGA PRAD datasets (p≤0.045, Figure 4, right), suggesting that DNMT3A may contribute to tumor progression.
Pooled analysis compares DNMT3A expression in normal versus prostate cancer tissues across 35 independent studies. DNMT3A levels are significantly elevated in prostate cancer samples. SD, standard deviation. IV, inverse variance. CI, confidence interval. Std, standardized. TCGA PRAD, The Cancer Genome Atlas Prostate Adenocarcinoma. df, Degrees of freedom.
Clinical relevance of DNMT3A expression in prostate cancer. Elevated DNMT3A expression is linked to poorer progression-free survival in datasets (A) GSE21032 and (B) The Cancer Genome Atlas prostate adenocarcinoma (TCGA PRAD). Additionally, DNMT3A levels are significantly higher in tumors with increased Gleason scores and advanced staging within the TCGA PRAD cohort. Sample sizes for each subgroup (numbers in brackets).
To elucidate the biological role of DNMT3A in prostate cancer, we identified genes correlated with DNMT3A expression in TCGA PRAD data. In total, 5,136 genes were positively correlated, whereas 4,459 genes were negatively correlated with DNMT3A (Pearson’s correlation FDR<0.01). GSEA using the ranked gene list indicated the enrichment of positively correlated genes in cellular components, such as the chromosomal region, nuclear chromosome, and condensed chromosome (Figure 5A). Additionally, these genes are involved in biological processes, such as heterochromatin organization, chromosome segregation, and epigenetic regulation of gene expression (Figure 5B), and molecular functions, including helicase activity, histone binding, and catalytic activity acting on DNA (Figure 5C). Hallmark pathway analysis revealed enrichment of pathways associated with cell cycle progression, particularly at the G2/M checkpoint, E2F targets, and mitotic spindle assembly (Figure 5D).
Gene ontology (GO) and pathway enrichment analyses of genes associated with DNMT3A expression. Top 10 GO terms are shown for (A) cellular components, (B) biological processes, and (C) molecular functions. (D) The most enriched Hallmark pathways. The ratio of core enrichment genes (bubble size) and its statistical significance (color scale).
GSEA identified G2/M checkpoint regulation as the most significantly enriched pathway (normalized enrichment score=2.404, FDR<2.2×10−16), implying a potential role for DNMT3A in this checkpoint. Therefore, we examined the correlation between DNMT3A expression and key hub genes within the protein-protein interaction network of the G2/M checkpoint, including AURKB, CCNA2, and CDK1. Given the central role of DNMT3Ain DNA methylation, we investigated its potential influence on the G2/M checkpoint gene expression by modulating DNA methylation. Our findings showed a negative correlation between DNMT3A expression and the DNA methylation status of AURKB, CCNA2, and CDK1 in TCGA PRAD dataset (Figure 6, left). This pattern was consistent with the positive correlation observed between DNMT3A expression and the expression levels of these hub genes (Figure 6, middle left). Additionally, these genes were significantly upregulated during prostate cancer progression (Figure 6, middle right). Moreover, the elevated expression of AURKB, CCNA2, and CDK1 was associated with worse survival outcomes, reinforcing the hypothesis that DNMT3A exerts an oncogenic effect by activating G2/M checkpoint genes during prostate cancer progression.
Correlation between DNMT3A expression and key regulators of the G2/M checkpoint pathway in prostate cancer. The inverse relationship between DNMT3A and the methylation status of three hub genes: (A) AURKB, (B) CCNA2, and (C) CDK1 (left panel). The positive correlation between DNMT3A and the expression of these genes (middle-left panel). Elevated expressions of AURKB, CCNA2, and CDK1 in prostate cancer tissues relative to normal samples (middle-right panel). Higher expression levels of these genes are associated with worse survival outcomes in prostate cancer (right panel). Expression data were log2(x+1) transformed RNA sequencing by expectation-maximization normalized count. Patient groups were dichotomized into low- and high-expression categories based on the median expression value. Sample sizes for each subgroup (numbers in brackets).
Discussion
This study identified DNMT3A and its genetic variant, rs77993651, as critical factors influencing prostate cancer survival. The rs77993651 A allele is associated with improved survival, likely because of its role in reducing DNMT3A expression. Functional analyses revealed that DNMT3A exerts oncogenic effects as its elevated expression is significantly associated with prostate cancer progression and poor survival outcomes. Notably, DNMT3A expression positively correlated with key regulators of the G2/M cell cycle checkpoint, highlighting its role in cell cycle dysregulation. These findings highlight DNMT3A as a promising prognostic biomarker and potential therapeutic target for prostate cancer.
SNP rs77993651 is located within an intronic region of DNMT3A that exhibits enhancer-like chromatin modification patterns, suggesting a potential regulatory role in DNMT3A expression. Our eQTL analysis revealed that the protective allele A was associated with decreased DNMT3A expression in regulatory T cells. However, the potential effect of rs77993651 on DNMT3A expression in prostate tissues remains unexplored. DNMT3A, a key DNA methyltransferase responsible for de novo DNA methylation, aids in gene regulation and has been implicated in various cancers, including prostate cancer (28, 29). By establishing new DNA methylation patterns, DNMT3A often silences genes involved in cell growth and differentiation (30). It cooperates with EZH2, a component of polycomb repressive complex 2, to achieve robust gene silencing (11, 12). Elevated DNMT3A expression is frequently observed in cancers, leading to oncogene activation, such as CDK1 in acute myeloid leukemia (31), and silencing of tumor suppressor genes, such as the DAB2 interactive protein in colorectal cancer (32). These alterations contribute to uncontrolled cell growth and tumor progression. Additionally, DNMT3A regulates oncogene activation and the epithelial-to-mesenchymal transition by modulating the microRNA-200 family, thereby promoting cancer progression and metastasis (33).
Our GSEA of DNMT3A-associated expression networks revealed a positive relationship between DNMT3A expression and cell cycle G2/M checkpoint genes, including the key hub genes AURKB, CCNA2, and CDK1. AURKB is essential for chromosome condensation, spindle assembly checkpoints, and chromosome segregation during cell division. In prostate cancer, AURKB is overexpressed, which promotes tumor progression by enhancing proliferation and resistance to apoptosis. Its inhibition has shown promise in reducing tumor growth (34). CCNA2 is pivotal in regulating the cell cycle, particularly by facilitating the G1/S and G2/M transitions as a cyclin-dependent kinase regulator. In prostate cancer, CCNA2 exhibits oncogenic properties by promoting cancer cell proliferation, invasion, and metastasis, whereas its downregulation leads to cell cycle arrest (35). CDK1 is critical for the G2/M transition and mitotic progression, and its overexpression is linked to increased tumor proliferation and survival. Notably, CDK1 enhances AR activity by directly phosphorylating AR at Ser81, promoting prostate cancer growth, even under low-androgen conditions (36). Therapeutically, DNMT inhibitors, such as 5-aza-2′-deoxycytidine, have demonstrated efficacy in prostate cancer treatment by reducing tumor cell growth and enhancing sensitivity to AR inhibitors (37, 38). Furthermore, the combined inhibition of DNMT3A and EZH2 enhances antitumor effects by preventing compensatory epigenetic mechanisms and activating immune pathways, as observed in colon cancer (39). These findings highlight DNMT3A is a key driver of prostate cancer progression and a promising therapeutic target.
Conclusion
The study provides valuable insights into the role of EZH2-related genetic variants in prostate cancer progression and highlights DNMT3A rs77993651 as a potential prognostic biomarker. A major strength of this study is the use of a well-characterized patient cohort with comprehensive clinical data, which allowed for robust association analyses. Additionally, this study’s focus on epigenetic regulation offers a novel perspective on disease mechanisms. However, the limitations include the relatively small sample size, which may affect the statistical power, and lack of functional validation experiments to confirm the biological impact of the identified variants. Given the therapeutic relevance of EZH2 inhibitors, these insights may guide personalized treatment strategies. Future research should validate these associations in larger cohorts and explore the mechanistic link between DNMT3A dysregulation and progression of prostate cancer. Understanding these epigenetic interactions may enhance risk stratification and improve targeted therapies for patients with aggressive diseases.
Acknowledgements
The Authors thank Chao-Shih Chen for data analysis, and the National Centre for Genome Medicine, Taiwan, for technical support. The results published here are based in part on data generated by the 1000 Genomes and TCGA projects.
Footnotes
Authors’ Contributions
SPH contributed to project development, data collection, and funding acquisition. BYB performed data collection and analysis. CYH, CCY, and VCL performed data collection. THC and TLL performed data analysis. YTC contributed to project development, data analysis, and funding acquisition. All Authors prepared and agreed to the published version of the manuscript.
Conflicts of Interest
The Authors declare that they have no potential conflicts of interest in regard to this study.
Funding
This work was supported by the National Science and Technology Council of Taiwan (grant Nos: 110-2320-B-A49A-515, 110-2314-B-002-113, 111-2314-B-002-240-MY3, 111-2320-B-039-021-MY3, 111-2218-E-037-001, 112-2218-E-037-001, 113-2218-E-037-001, 112-2314-B-037-127, and 113-2314-B-037-016), the National Health Research Institute (grant no: NHRI-EX113-11313SI), the Kaohsiung Medical University (grant Nos: KMUH111-1R58, KMUH112-2R59, and KMUH113-3R52), and the China Medical University (grant Nos: CMU111-MF-66, CMU111-MF-09, CMU112-MF-10, and CMU113-MF-11). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Artificial Intelligence (AI) Disclosure
During the preparation of this manuscript, a large language model (Google Gemini) was used solely for language editing and stylistic improvements in select paragraphs. No sections involving the generation, analysis, or interpretation of research data were produced by generative AI. All scientific content was created and verified by the authors. Furthermore, no figures or visual data were generated or modified using generative AI or machine learning–based image enhancement tools.
- Received February 26, 2025.
- Revision received April 28, 2025.
- Accepted May 5, 2025.
- Copyright © 2025 The Author(s). Published by the International Institute of Anticancer Research.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
