Druggable Genome-Wide Mendelian Randomization Identifies Therapeutic Targets and DNA Methylation-Mediated Regulation in Urological Malignancies

Abstract

Introduction

Urological cancers, including prostate, renal, and bladder cancers, present significant challenges to human health, particularly for patients in advanced stages due to limited targeted therapy options, high drug resistance, and poor prognosis. This study seeks to systematically identify new therapeutic targets by integrating multi-omics data and elucidating their molecular mechanisms.

Methods

Our methodology involved a multi-phase integrative approach to identify therapeutic targets in urological cancers. We conducted a transcriptome-wide Mendelian randomization (TWMR) analysis using the druggable genome. We used cis-expression quantitative trait loci (cis-eQTL) data from the eQTLGen consortium (N = 31,684) as exposures and GWAS summary statistics from FinnGen R11 and Open GWAS as outcomes. Significant genes were identified through Bonferroni correction, followed by colocalization analysis to assess shared genetic effects between risk SNPs and gene expression. Summary-data-based MR (SMR) analysis was carried out using eQTLGen data. Validation of potential targets included drug prediction using DSigDB and molecular docking with AutoDock Vina. Furthermore, we integrated DNA methylation quantitative trait loci (mQTL) data from GoDMC to explore methylation-mediated regulatory mechanisms through mediation MR analysis.

Results

Our transcriptome-wide MR analysis identified 11 high-confidence therapeutic targets in urological cancers: 4 in prostate cancer (BNIP2, KAT5, MMP24, SIK2), 2 in renal cancer (GRB10, GPR17), and 5 in bladder cancer (CELSR1, GPS1, PROC, GSTM1, KLC3). Molecular docking confirmed favorable interactions between these targets and existing drugs. Epigenetic analyses showed DNA methylation-mediated regulation of key targets: PROC (cg09747827) and CELSR1 (cg00784671) in bladder cancer, and SIK2 (cg11344533) in prostate cancer.

Conclusions

This study integrates multi-omics data to identify 11 high-confidence therapeutic targets in urological cancers and is the first to elucidate DNA methylation-mediated regulatory mechanisms for targets like PROC and SIK2. The findings introduce a “DNA methylation-gene expression” axis-based strategy for combined targeted-epigenetic therapy, providing crucial evidence for optimizing precision treatment in urological oncology.

Keywords

druggable genome Mendelian randomization (MR)DNA methylation urological cancers therapeutic targets multi-omics analysis

1. Introduction

Urological cancers, such as prostate cancer, renal cell carcinoma (RCC), and bladder cancer, present a significant global health challenge despite advancements in research. Androgen deprivation therapy (ADT) is a common treatment for prostate cancer, yet resistance often leads to castration-resistant prostate cancer (CRPC) or metastatic CRPC (mCRPC) with limited survival benefits from current therapies.^1,2 RCC, characterized by high immunogenicity and immune cell infiltration, has seen a shift in systemic treatment with the introduction of immune checkpoint inhibitors (ICIs), notably anti-PD-1 antibodies, for metastatic RCC (mRCC). However, primary resistance and acquired resistance to immunotherapy are common in mRCC cases.³ Early detection challenges contribute to the diagnosis of up to 30% of advanced RCC cases.⁴ Non-muscle-invasive bladder cancer (NMIBC) exhibits a high recurrence rate, necessitating frequent intravesical instillations with Bacillus Calmette-Guérin (BCG) therapy, limited by side effects during long-term maintenance.⁵ Antibody-drug conjugates (ADCs) have emerged as a promising therapy for bladder cancer, yet challenges persist in their clinical application, including treatment response durability, off-target toxicity management, and antigen expression variability among patient subtypes.⁶ Targeted therapy is crucial for enhancing cancer patient survival, but efficacy limitations, adverse effects, and resistance leading to relapse hinder effective targeted anticancer treatments. Consequently, exploring diverse therapeutic targets is imperative to address these challenges in cancer treatment.

Identifying effective drugs for specific diseases and confirming their ability to alter disease progression is crucial for successful drug development. This task is still difficult, time-consuming, and costly within traditional frameworks. The low yield of approved drugs, largely due to late-stage failures in randomized clinical trials (RCTs), poses significant obstacles. The lack of economically viable alternatives to RCTs exacerbates this issue.⁷ Integrating genetics into drug development offers a promising solution. Extensive human genetic research provides distinct chances for advancing therapies for complex diseases, as drug targets validated genetically exhibit increased success rates in the drug discovery pipeline.^8,9 Although genome-wide association studies (GWAS) have pinpointed single-nucleotide polymorphisms (SNPs) associated with the risk of urological cancer, they are limited in their capacity to directly identify causative genes or specific drug targets.

Many identified SNPs reside in non-coding regions or between genes, complicating the linkage between these variants and their functional relevance to drug discovery.^10,11 Proteins encoded by druggable genes either serve as drug targets or have potential for targeting by small molecules or monoclonal antibodies. Integrating expression quantitative trait loci (eQTLs) with GWAS data through Mendelian randomization (MR) analysis aids in identification of target genes linked to risk variants through causal inference.¹² Druggable genes, which usually encode proteins with drug-binding sites, offer crucial clues for pinpointing actionable drug targets, as many genes are not amenable to drug targeting. eQTLs located in the genomic regions of druggable genes often serve as proxies for potential drug targets.¹³

Mendelian randomization (MR) is a robust genetic statistical method that mimics randomized controlled trials to evaluate drug efficacy.¹⁴ It employs single nucleotide polymorphisms (SNPs) linked to changes in gene expression, known as expression quantitative trait loci (eQTLs), to simulate the effects of long-term drug exposure on proteins encoded by specific genes. This technique integrates data from genome-wide association studies (GWAS) by connecting the same SNPs to disease outcomes. MR enables researchers to determine causal relationships between gene expression levels and disease outcomes by merging SNP-gene expression and SNP-disease associations.¹⁵ Robust Mendelian randomization (MR) studies can leverage publicly available data from large-scale genome-wide association studies (GWAS) to evaluate exposures and outcomes across independent cohorts. Genetic variants influencing the expression of druggable genes are randomly distributed at conception, and since individuals typically remain unaware of their genotypes, MR studies effectively mimic blinded trials. Integrating genetics into drug development could greatly improve the process, as therapies supported by genetic evidence have a higher likelihood of success in clinical trials. Previous Mendelian randomization (MR) analyses have successfully utilized druggable genome data to pinpoint drug targets, as demonstrated in prostate cancer research.¹⁶ However, these studies lacked thorough colocalization and summary-data-based MR (SMR) validation. In the context of bladder cancer, only proteome-wide MR studies have been conducted,^17,18 and no equivalent research has been performed for renal cancer, indicating a gap in MR-based investigations for specific urological cancers.

This study sought to identify promising therapeutic targets for three major urological cancers by integrating multiple data layers. We performed a comprehensive transcriptome-wide and epigenome-wide Mendelian randomization (MR) analysis, including mediation MR. This involved combining transcriptomic data from druggable genes, methylomic epigenome data from potential target genes, and genome-wide association study (GWAS) summary statistics for the cancers. The analysis aimed to identify potential therapeutic targets at the mRNA level. To bolster our findings’ robustness, we performed colocalization and summary-data-based Mendelian randomization (SMR) analyses, offering more reliable validation of causal links between genetic variants and urological cancers’ outcomes. Tissue-specific MR analysis, single-cell expression analysis, protein-protein interaction studies, and GO and KEGG enrichment analysis were performed on the identified significant genes to elucidate their biological functions. We further investigated upstream regulatory mechanisms of key genes by pinpointing critical DNA methylation sites, elucidating how epigenetic changes may drive cancer development through druggable genes. Additionally, candidate drug prediction, molecular docking analyses, and phenome-wide association studies have identified potential therapeutic agents targeting these genes and evaluated off-target effects, offering insights for more effective and targeted treatments for the three major urological tumors. This comprehensive approach enhances our understanding of molecular mechanisms and establishes a foundation for developing personalized treatment options for these cancers.

2. Materials and Methods

2.1. Study Design

The study design is depicted in Figure 1. This research adhered to the STROBE-MR guidelines¹⁹ and was predicated on three core assumptions: (1) Relevance assumption: SNPs must exhibit a strong association with the exposure; (2) Independence assumption: SNPs should not be correlated with confounding factors; (3) Exclusion assumption: SNPs must influence the outcome solely through the exposure.²⁰ This study utilized publicly available data from prior investigations, all of which had received ethical approval and informed consent, thus requiring no further approval.

Figure 1.

Outline of the study strategy

2.2. Druggable Genes Identification

Druggable genes were identified by integrating data from the Drug-Gene Interaction Database (DGIdb v4.2, https://www.dgidb.org/downloads) with findings from the study by Finan et al. This approach allowed for the compilation of 5,012 medically relevant genes for Mendelian Randomization (MR) analysis (see Tables S1 and S2). The DGIdb database was chosen because it systematically aggregates drug-gene interactions from various sources, creating a comprehensive repository of potential therapeutic targets. We enhanced our dataset with pharmacogenomic insights from the Finan et al study, which refined the identification of druggable genes relevant to urological cancers. By linking genome-wide association study (GWAS)-identified genes associated with urological cancers to pharmacologically actionable targets, our approach improves the accuracy of therapeutic candidate selection.

2.3. Exposure Data

The cis-eQTLs data were acquired from the eQTLGen Consortium (https://eqtlgen.org/). The eQTLGen dataset included 16,987 genes and 31,684 cis-eQTLs from blood samples of mainly healthy European individuals.²¹ Tissue-specific cis-eQTLs for causal genes were sourced from the Genotype-Tissue Expression Project V10 (GTEx V10) for validation. DNA methylation mQTL data were sourced from the GoDMC database, encompassing results from genome-wide scans of cis- and trans-eQTL meta-analyses involving 420,509 DNA methylation sites (https://mqtldb.godmc.org.uk/downloads). Information on DNA methylation sites linked to druggable genes can be downloaded at https://ngdc.cncb.ac.cn/ewas/datahub/exploration. All data utilized in this research received ethical approval and informed consent from the original study participants.

2.4. Outcome Data

GWAS data from discovery cohorts of prostate cancer (PCa), bladder cancer (BCa), and renal cell carcinoma (RCC), encompassing subtypes such as ccRCC, pRCC, and chRCC, were obtained from the FinnGen (R11) study. The replication cohort for PCa was obtained from Open GWAS_ieu-b-85, for RCa from Open GWAS_ukb-b-1316, and for BCa from Open GWAS_ukb-b-8193. Supplementary Table S3 provide detailed data information.

2.5. Statistical Analysis

2.5.1. MR Analysis

The R package TwoSampleMR (version .6.7) was utilized to summarize the MR analysis, employing blood and tissue-specific cis-eQTLs and mQTLs data as exposures and GWAS data from three major urological tumors as outcomes.²² The package’s harmonise_data function was used to import and align the exposure and outcome data. The genetic instrumental variables used in the MR analysis must satisfy the MR assumptions that SNPs are directly associated with the exposure and are highly correlated with at least one gene expression.²³ Ultimately, 8,600 cis-eQTLs corresponding to 3,887 druggable genes were retained for further analysis (Table S4).

The MR estimate for a single SNP was derived using the Wald ratio, while the inverse-variance weighted (IVW) average was applied to multiple SNPs. The IVW method, assuming all genetic instruments are valid, offers the highest statistical power when this condition is met.²⁴ The weighted median approach assigns weights to SNPs based on their MR estimates, producing an overall MR estimate via the median with bootstrapped standard errors, accommodating up to 50% invalid instruments.²⁵ MR-Egger permits horizontal pleiotropy, where SNPs may impact the outcome through pathways unrelated to the exposure, although this reduces statistical power.²⁶ Consistency in directional estimates across these methods implies that pleiotropy does not bias the IVW estimate.

Heterogeneity and horizontal pleiotropy were evaluated to identify differences among instrumental variables and assess potential bias. Heterogeneity was assessed using Cochran’s Q test via the mr_heterogeneity() function in the TwoSampleMR package, where a Q-statistic with p < 0.05 indicates heterogeneity. Horizontal pleiotropy was evaluated using MR-Egger regression through the mr_pleiotropy_test() function in TwoSampleMR, focusing on the intercept term to identify directional pleiotropy. A non-significant intercept p > 0.05 indicates no pleiotropic bias.²⁷ To account for multiple testing, the Bonferroni adjustment was utilized to decrease the likelihood of false positives. Significant findings in the discovery cohort were identified with P-values less than 1.28E-5 (P = 0.05/3887). Due to the scarcity of positive findings for renal and bladder cancer under Bonferroni correction, a more permissive threshold of 1E-3 was applied to these phenotypes. Genes with P-values above this threshold but below 0.05 were classified as nominally significant causal associations. For significant genes, quality control included verifying effect direction consistency across three methods and ensuring no horizontal pleiotropy via the MR-Egger test. Heterogeneity and pleiotropy tests were also conducted. Genes meeting quality control underwent further analysis in a replication cohort, with significance assigned to associations having P-values below 0.05.

2.5.2. Colocalization Analysis

To identify genes with significant causal associations, we performed a colocalization analysis of urological cancer risk using the R package coloc (version 5.2.3). We analyzed SNPs that were harmonized using the TwoSampleMR package, applying the default parameters of the “coloc.abf()” function, setting prior probabilities as p1 = 1E-4, p2 = 1E-4, and p12 = 1E-5. These parameters denote the probabilities that SNPs in the test region are strongly linked to gene expression, urological cancer risk, or both. These default values are standard in GWAS-eQTL colocalization studies and have consistently produced reliable and reproducible results in prior research.²⁷ The posterior probabilities from colocalization analysis represent five hypotheses: PPH0 indicates no association of the SNP with any trait; PPH1 suggests the SNP is linked to gene expression but not urological cancer risk; PPH2 implies the SNP is related to urological cancer risk but not gene expression; PPH3 indicates the SNP is associated with both urological cancer risk and gene expression but by different SNPs; PPH4 suggests the SNP is associated with both urological cancer risk and gene expression, driven by a shared SNP. High colocalization support was defined as PPH4 > 0.75, and moderate support as PPH4 > 0.5. Genes identified through colocalization may serve as potential targets for drug development.

2.5.3. Summary-Data-Based MR (SMR) Analysis

Summary-data-based Mendelian randomization (SMR) analysis was employed as a supplementary approach to validate the causal relationships between candidate genes and urological cancers.²⁸ The heterogeneity in dependent instruments (HEIDI) test was used to identify genes linked to urological cancer risks that arise from shared genetic variants rather than genetic linkage. This was done by analyzing multiple SNPs within specific regions. Both the SMR and HEIDI tests were conducted using the SMR software (version smr-1.3.1). A P-value greater than 0.05 in the HEIDI test suggested that the association between the gene and urological cancers was not influenced by linkage disequilibrium. Subsequently, the evidence levels for the gene-disease relationships were classified.

2.6. Single-Cell Type Expression Analysis: Cellular Heterogeneity and Microenvironmental Localization of Target Genes

We analyzed single-cell RNA-seq data from the Gene Expression Omnibus (GEO) to assess cell type-specific gene expression in urological cancers. The datasets included: Prostate cancer (GSE193337) with 8 samples from 4 patients, consisting of 4 cancerous and 4 benign tissues; Clear cell renal cell carcinoma (GSE135337) with 8 tumor and 6 benign kidney specimens; and Bladder cancer (GSE159115) with 7 primary tumor samples and 1 adjacent non-tumor tissue. Utilizing the “Seurat” package,²⁹ we initiated data preprocessing and transformation on raw single-cell RNA-seq data. Quality control steps included filtering out genes with less than 3 counts per cell and cells with less than 200 unique features. RNA TPM was subsequently normalized and scaled using the NormalizeData and ScaleData functions. Dimensionality reduction and visualization were accomplished through RunUMAP. We performed a comprehensive search for cell type markers to aid in classification and manual annotation. The DotPlot function visualized target gene expression patterns in urologic tumors and benign control tissues.

2.7. Enrichment Analysis and Protein-Protein Interaction (PPI)

Enrichment analyses were performed on significant causal genes in urological cancers using the R packages clusterProfiler³⁰ and org.Hs.eg.db, focusing on Gene Ontology(GO) and Kyoto Encyclopedia of Genes and Genomes(KEGG). The analysis elucidated the genes’ functional roles in biological processes, molecular functions, and cellular components, shedding light on associated metabolic pathways.

To investigate interactions between candidate genes and established drug targets for urological cancers, we sourced therapeutic drugs from the literature and identified their targets via the DrugBank database (https://go.drugbank.com/).³¹ The urological cancers’ known targets are detailed in the Supplementary Table S5. We then constructed a PPI network using STRING (https://string-db.org/), setting a minimum interaction threshold confidence score of 0.4, while keeping all other settings at their default values.³²

2.8. DNA Methylation-Mediated Mendelian Randomization

In urological tumors, notable alterations in gene methylation have been documented.^33-35 However, the interplay between DNA methylation changes, candidate target gene expression, and disease progression remains unclear. We conducted a two-sample Mendelian randomization and mediation analysis to investigate the role of target genes in mediating the relationship between DNA methylation and urological tumors. This approach separates the total effect into an indirect effect (mediated) and a direct effect (non-mediated). The impact of DNA methylation on urological tumors is twofold: (1) a direct effect of DNA methylation on urological tumors and (2) an indirect effect mediated by candidate target genes. To quantify the mediating role of target genes, we calculated the proportion of the indirect effect relative to the total effect. Since mediation analysis requires that there are no unmeasured confounding factors, we used genetic instruments to reduce confounding. Additionally, we conducted sensitivity analyses to evaluate the robustness of our findings.

2.9. Phenome-Wide Association Analysis (Phewas)

We utilized phenotype data from 1,431 traits in the Neale Lab consortium, accessible via the Open GWAS database, to conduct Phewas Mendelian randomization (MR-PheWAS) analyses. This approach aimed to investigate the off-target effects of specific genes and to determine whether the identified potential drug target genes showed additional pleiotropy that was not detected by the MR-Egger intercept test. The MR-PheWAS analysis utilized the Wald ratio and inverse-variance weighted (IVW) methods, ensuring that the parameters and criteria were consistent with those of the initial MR analysis.

2.10. Druggability Assessment

Assessing gene-drug interactions is crucial for identifying viable drug targets. Initially, we examined the DrugBank database to check if candidate genes are targeted by existing drugs. For genes without known drugs, we utilized the Drug Signatures Database (DSigDB) to predict potential drug candidates.³⁶ DSigDB contains 22,527 gene sets and 17,389 compounds linked to 19,531 genes, connecting drugs with their respective target genes. The target genes identified were submitted to DSigDB for drug prediction and assessment of their pharmacological capabilities.

2.11. Molecular Docking

Molecular docking simulations were performed to assess the binding energy and interaction mode of potential drugs with target proteins. AutodockVina 1.5.7 was used for receptor-ligand docking analyses on drugs that significantly influenced protective or risk genes.³⁷ This facilitated the evaluation of binding affinities and interaction profiles, guiding the prioritization of drug targets for subsequent experimental validation and optimization of potential drug candidates. Drug structures were sourced from the PubChem Compound Database (https://pubchem.ncbi.nlm.nih.gov/).³⁸ Protein structures were retrieved from the Protein Data Bank (https://www.rcsb.org/). Initially, water molecules were removed from both protein and ligand files, and polar hydrogen atoms were added. The protein and ligand PDBQT files were then imported into Autodock. Grid boxes were positioned to enclose the structural domains of each protein, allowing for unrestricted molecular mobility. Subsequently, the docking range of the protein and ligand was defined, and molecular docking was executed using AutodockVina 1.5.7. Visualization was conducted with Pymol v3.10 (https://www.pymol.org). The binding activity of potential compounds with the target was evaluated through docking energy values, highlighting the interaction between the ligand and protein within the docking pocket.

Results

3.1. Transcriptome Screening for Causal Druggable Genes in Urological Cancers

In the discovery cohort for prostate cancer, the expression of 16 genes showed a causal association with PCa risk (Figure 2). Of these, 15 risk genes were confirmed in the replication cohort.

Figure 2.

Manhattan plot of MR results for candidate genes associated with prostate cancer in the discovery phase. Significant genes are labeled, and the gray dashed line denotes the Bonferroni-corrected threshold (1.28E-5)

In the discovery cohort for overall renal cancer, only GRB10 surpassed the significance threshold. To identify additional potential risk genes, we adjusted the threshold to 1e-3, resulting in seven genes meeting the criteria (Figure 3). However, these genes did not validate in the replication stage. For renal cancer subgroups, significant genes are illustrated in the Figure 4. Lacking external cohort data for further validation, we intersected findings from overall renal cancer and three renal cell carcinoma subtypes to identify validated causal genes for further analysis (Figures 5 and 6). In the bladder cancer discovery cohort, no genes met the initial significance threshold. Using the adjusted threshold of 1e-3, seven genes were identified, but these also failed validation in the replication cohort (Figure 7). Detailed results of these analyses are presented in Figure 8.

Figure 3.

Manhattan plot of MR results for candidate genes associated with renal cancer in the discovery phase. Significant genes are labeled, and the gray dashed line denotes the threshold (1E-3)

Figure 4.

Volcano plots depicting associations between druggable genes and renal cell carcinoma subtypes: (A) ccRCC, (B) pRCC, and (C) chRCC. Note. Abbreviations: RCa: Renal Cancer, ccRCC: Clear cell renal cell carcinoma, pRCC: Papillary renal cell carcinoma, chRCC: Chromophobe renal cell carcinoma

Figure 5.

UpSet plot of nominally significant druggable genes across renal cancer subtypes (RCa, ccRCC, pRCC, chRCC). Horizontal bar on left represents number of set size in each covariates. Dots and lines represent subsets of set size. Vertical histogram represents number of set size in each subset. Note. Abbreviations: RCa: Renal Cancer, ccRCC: Clear cell renal cell carcinoma, pRCC: Papillary renal cell carcinoma, chRCC: Chromophobe renal cell carcinoma

Figure 6.

Forest plots illustrate the intersection of nominally significant druggable genes (p < 0.05) between unspecified RCa and three renal cell carcinoma subtypes. ccRCC: 85 total genes (top 8 most significant shown, p < 1e-3); pRCC: 21 genes; chRCC: 12 genes

Figure 7.

Manhattan plot of MR results for candidate genes associated with bladder cancer in the discovery phase. Significant genes are labeled, and the gray dashed line denotes the threshold (1E-3)

Figure 8.

Forest plots illustrate significant gene associations identified in discovery analyses using the FinnGen R11 database, including 16 prostate cancer-associated genes and 7 genes each for renal cell carcinoma and bladder cancer. In replication analyses, distinct line patterns denote associations obtained from different outcome-specific GWAS datasets

In the sensitivity analysis of prostate cancer, the Wald ratio method is employed due to the presence of a single SNP per risk gene, thus excluding heterogeneity and pleiotropy interference. In bladder cancer, eight SNPs represent GSTM1, allowing for pleiotropy and heterogeneity testing, which reveals no such effects. The consistency in effect direction across methods and an MR Egger p-value of 0.213 indicate no horizontal pleiotropy. Similarly, for renal cancer, three SNPs represent RPL14, enabling pleiotropy and heterogeneity testing. The results show no such effects, with consistent effect directions and an MR Egger p-value of 0.472, indicating no horizontal pleiotropy. In the renal cancer subgroup, clear cell carcinoma exhibited neither heterogeneity nor pleiotropy. However, LTBR, PMS2, and SENP7 were not considered for further analysis because of inconsistent effect directions. In papillary cell carcinoma, no pleiotropy or heterogeneity was observed, but BLMH was excluded for the same reason. Chromophobe cell carcinoma showed no pleiotropy or heterogeneity, and all risk genes had consistent effect directions.These risk genes are thus included in subsequent analyses.The complete results are provided in Tables S6-S32.

3.2. Colocalization Analysis

Previous studies have shown that significant Mendelian Randomization (MR) outcomes may arise from loci where SNPs are in tight linkage disequilibrium. In such cases, the associations between SNP exposure and SNP outcomes may stem from two distinct causal SNPs, potentially leading to false-positive results in inference.³⁹ To address this limitation, colocalization analysis can identify if the same causal SNPs are shared between the exposure and outcome when SNPs are linked to both.³⁴

Evidence suggests that genes passing both MR and colocalization tests are more likely to become drug targets and have a higher chance of approval.⁴⁰ Therefore, this study aimed to perform colocalization analysis on the causal genes associated with three major urological tumors and three subtypes of renal cell carcinoma, as identified in the previous analysis.

As detailed in Figure 9A, in prostate cancer, 6 out of 15 genes replicated in the external cohort (BNIP2, CFAP44, KAT5, MMP24, SIK2, SUFU) showed evidence of colocalization with PCa (strong PP.H4 > 0.75, moderate PP.H4 > 0.5). In the case of renal cancer, as none of the genes from the overall renal cancer cohort were replicated in the external cohort, 7 genes that met the significance threshold of 1e-3 were selected. Of these, 5 genes (GRB10, MAPK11, RPL14, GPR17, PAPLN) passed the colocalization analysis. In bladder cancer, 6 out of 7 genes (CELSR1, GPS1, GSTM5, PROC, GSTM1, KLC3) passed the colocalization analysis. For the renal cancer subgroups, we selected genes that overlapped with the overall renal cancer data and had consistent effect directions for colocalization. Since there were many overlapping genes in ccRCC, we took the intersection of genes passing the significance threshold of 1e-3 and 85 replicated genes, and included 8 significant causal genes for colocalization. Four of these genes (GPR17, CDKN1A, GRB10, PPP1R13L) passed the colocalization test. For pRCC, we excluded ZNF816, which had an inconsistent effect direction with the overall renal cancer, as well as BLMH, which showed inconsistent directions across the three methods. We included 21 risk genes for colocalization, among which 2 genes (ACOT7, GAB1) passed the test. In chRCC, 12 overlapping genes were included, and only SULT1B1 passed the colocalization test (Figure 9B).

Figure 9.

Heatmap displays colocalization results of significant druggable genes with urological cancers: (A) significant druggable genes with three major urological cancers, and (B) significant druggable genes with three renal cell carcinoma subtypes

Genes that passed the colocalization analysis may serve as potential drug target genes. Further details are presented in supplementary Table S33.

3.3. SMR and HEIDI Tests

To validate our findings, we performed SMR and HEIDI tests on candidate target genes for the three major cancer types identified by MR, utilizing comprehensive summary-level data of gene mRNA levels in blood from eqtlgen. In prostate cancer (PCa), 6 out of 15 genes passed both the SMR (P < 0.05) and HEIDI (P > 0.05) tests. In renal cancer (RCa), 3 out of 7 genes met these criteria, while in bladder cancer (BCa), 5 out of 7 genes did. Within renal cancer subtypes, 6 genes in ccRCC, 9 in chRCC, and 16 in pRCC passed both tests (supplementary Table S34). Figure 10 displays the SMR locus maps, effect maps, and colocalization locus maps for each target gene.

Figure 10.

Figures (a) to (k) depict a comprehensive analysis of 11 candidate genes identified based on consistent evidence from colocalization and summary-data-based Mendelian randomization (SMR). The figures integrate SMR locus plots showing regional association signals and patterns of linkage disequilibrium, SMR effect plots quantifying potential causal relationships between instrumental variants and urological cancer outcomes, and LocusCompare plots illustrating the genomic colocalization between expression quantitative trait loci (eQTLs) of candidate genes and susceptibility loci for urological cancer

3.4. Classification of the Evidence

To systematically present the strength of the evidence obtained in this study, we categorized the candidate target genes into three tiers (Tier 1, Tier 2, and Tier 3) based on the results of Mendelian randomization (MR), colocalization analysis, summary-data-based Mendelian randomization (SMR), and the HEIDI test. The classification criteria were as follows:

Tier 1 (Primary Targets): Genes that met all three criteria: colocalization (posterior probability PP.H4 > 0.5), SMR (P < 0.05), and the HEIDI test (P > 0.05). These targets represent the highest level of evidence, suggesting their association with the corresponding cancer is unlikely to be driven by confounding factors, linkage disequilibrium, or pleiotropy.

Tier 2 (Secondary Targets): Genes that met any of the following conditions: (1) identified through colocalization analysis alone; (2) identified through SMR and HEIDI tests alone; or (3) could not be fully assessed due to the unavailability of required summary data. These targets have supportive evidence, though it is less robust than that for Tier 1.

Tier 3 (Preliminary Targets): Genes identified solely through MR analysis (P < 0.05) that either were not subjected to or did not pass subsequent colocalization or SMR tests. These findings are considered preliminary and warrant further validation.

The detailed classification results, including the number of genes identified for each cancer type, are summarized in Table 1.

Table 1.

Prioritised Therapeutic Targets Across Urologic Cancers by Evidence Tier

Cancer type	Tier 1	Tier 2	Tier 3
PCa¹	BNIP2, KAT5, MMP24, SIK2	SUFU, CFAP44, TET2, GP9	CRTC2, ZFP36L2, TPM3, STK25, DTYMK, KRT18, KRT8
RCa²	GRB10, GPR17	MAPK11, RPL14, PAPLN, CDKN1A	CAV1
BCa³	CELSR1, GPS1, PROC, GSTM1, KLC3	GSTM5	HLA-DRB1
ccRCC⁴	CDKN1A, GRB10, GPR17	PPP1R13L, PTH2R, LIPH, HLA-DQA1	CBLN3
pRCC⁵	ACOT7, GAB1	FSIP1, TTN, NDUFB5, MAG, COL7A1, APEH, CD14, ANTXR2, TMCO6, MMEL1, GPX1, SP3, FZD2, MLH4	DSC1, TUBB2A, CD300LF, SRPK1, CHKB
chRCC⁶	SULT1B1	IL1RAP, RSPO4, HLA-F, SIGLEC6, SIK2, CLEC11A, MOCOS, PDHX	ATP5F1E, CDKN1C, CST7

¹PCa: Prostate Cancer.

²RCa: Renal Cancer (unspecified).

³BCa: Bladder Cancer.

⁴Clear Cell Renal Cell Carcinoma.

⁵pRCC: Papillary Renal Cell Carcinoma.

⁶chRCC: Chromophobe Renal Cell Carcinoma.

3.5. Validation of Causal Genes Using Single-Tissue eQTL From the GTEx V10 Dataset

We extracted eQTL data for candidate target genes from prostate, kidney, and bladder tissues using the GTEx V10 dataset and performed Mendelian randomization analysis to assess tissue-specific causal associations with related urological tumors. From prostate tissue, we identified “SUFU, KRT18, DTYMK, and KRT8.” Notably, “CFAP44” demonstrated a causal association with prostate cancer in prostate tissue. GPR17 was analyzed in renal cortex tissue but showed no causal link to renal cancer. In contrast, HLA-DRB1 from bladder tissue maintained a causal association with bladder cancer (Figure 11, Table S35).

Figure 11.

Forest plots present tissue-specific transcriptome-wide association study (TWAS) results validating putative causal genes for urological cancers using GTEx version 10 data

3.6. Cell-Type Specific Expression in Urologic Tumors and Normal Tissue

We analyzed single-cell RNA-seq data from GEO to evaluate the enrichment of urologic tumor target genes in tumor versus normal tissues, specifically focusing on tier 1 and tier 2 targets.

The GSE193337 dataset comprised 8 paired prostate cancer samples from 4 patients, consisting of 4 cancerous and 4 benign tissues. Post-quality control, 27,916 cells were analyzed. These cells were organized into 23 clusters and categorized into 9 distinct cell types. All 8 target genes were expressed in prostate tissues. Specifically, BNIP2 and SUFU were predominantly found in endothelial cells, KAT5 in smooth muscle cells, SIK2 and CFAP44 in fibroblasts, MMP24 in basal epithelial cells, TET2 in granulocytes, and GP9 in luminal cell, as depicted in Figure 12.

Figure 12.

Single-cell type expression of candidate genes identified through transcriptome-wide Mendelian randomization was examined in prostate tumor tissue and benign tissue. (A) A total of 23 cell clusters representing 9 cell types were identified. (B) Illustrates the cell type composition in both prostate benign and tumor tissue, while (C) and (D) display the expression levels of the candidate genes across different cell types

The study analyzed 7 ccRCC tumor samples and 6 benign renal specimens from the GSE159115 dataset. Post-quality control, 24,926 cells were identified. These cells were categorized into 17 clusters and subsequently classified into 6 distinct cell types (Figure 12). Of the 6 target genes, expression data were available for 5 within ccRCC tissues, with no data for GPR17. GRB10 and MAPK11 were predominantly expressed in endothelial cells, RPL14 in CD4+ T cells, PAPLN in epithelial cells, and CDKN1A in mesangial cells, as depicted in Figure 13.

Figure 13.

Single-cell type expression of candidate genes identified through transcriptome-wide Mendelian randomization was examined in renal tumor tissue and adjacent normal tissue. (A) A total of 17 cell clusters representing 6 cell types were identified. (B) Illustrates the cell type composition in both renal tumor and adjacent normal tissue, while (C) and (D) display the expression levels of the candidate genes across different cell types

The GSE135337 dataset included 7 tumor samples and 1 adjacent tissue sample from 7 patients with bladder cancer. After quality control, 42,290 cells were analyzed. These cells were grouped into 13 clusters and identified as 8 distinct cell types. Of the 6 target genes, 4 were expressed in bladder tumor tissues. GSTM5 and GSTM1 were expressed only in normal bladder tissues. CELSR1 and PROC were predominantly expressed in Basal cells within tumor tissues, GSTM5 and GPS1 in fibroblasts, and KLC3 in neutrophils of tumor tissues. The primary cell types expressing the remaining target genes are depicted in Figure 14.

Figure 14.

Single-cell type expression of candidate genes identified through transcriptome-wide Mendelian randomization was examined in bladder tumor tissue and adjacent normal tissue. (A) A total of 13 cell clusters representing 8 cell types were identified. (B) Illustrates the cell type composition in both bladder tumor and adjacent normal tissue, while (C) and (D) display the expression levels of the candidate genes across different cell types

3.7. Enrichment Analysis and PPI Networks

GO and KEGG enrichment analyses are powerful tools for investigating gene functions and their associated pathways.⁴¹ In prostate cancer, GO analysis reveals that the most prominent pathways in the Biological Process (BP) category are primarily linked to metabolic processes, particularly the regulation of carbohydrate metabolism.⁴² In the Cellular Component (CC) category, drug target genes are enriched in focal adhesions and cell-substrate junctions, consistent with previous studies,⁴³ while in the Molecular Function (MF) category, these genes are involved in protein kinase activity. KEGG enrichment analysis further identifies apoptosis and the cAMP signaling pathway as the top pathways in prostate cancer. In renal cancer, the key BP pathways are associated with gland development, while the CC category highlights the enrichment of drug target genes in membrane rafts and microdomains, corroborating prior research.⁴⁴ These genes are also implicated in protein kinase activity in the MF category. The top KEGG pathways in renal cancer include the MAPK signaling pathway, cellular senescence, and neurotrophin signaling pathway. In bladder cancer, the significant BP pathways are related to peptidyl-serine modification and antigen processing and presentation (MHC Class II). The CC analysis reveals enrichment of drug target genes in MHC protein complexes, consistent with earlier studies,⁴⁵ while the MF category shows involvement in protein kinase activity. The top KEGG pathways for bladder cancer include phagosome, Th17 cell differentiation, and Th1/Th2 cell differentiation.

The target genes of tier 1 and tier 2 across the three major urological cancers were loaded into the STRING database to construct protein-protein interaction (PPI) networks. Specifically, eight drug targets in prostate cancer, six in renal cell carcinoma, and six in bladder cancer were found to interact with known protein targets of these malignancies. The detailed results of GO and KEGG enrichment analyses, together with the PPI network information, are presented in the Supplementary Materials.

3.8. MR-PheWAS

To investigate the impact of 11 candidate genes on various traits beyond the MR-Egger intercept test, we performed an MR-PheWAS analysis on these genes using 1,431 phenotype datasets from the UK Biobank. A significance threshold of 0.05/1431*11 was applied. The results are shown in Figure 15. KAT5 was negatively correlated with creatinine and Cystatin C, and positively correlated with urea and gout. SIK2 was positively correlated with allergic diseases and bilirubin, and negatively correlated with hypertension. MMP24 showed positive correlations with height and weight. GRB10 was positively correlated with both Cystatin C and creatinine. GSTM1 was negatively correlated with HDL cholesterol, and KLC3 was negatively correlated with platelet count. No significant pleiotropy was observed for the remaining target genes. These findings highlight the complex pleiotropic relationships of the target genes and their potential implications for various traits. Detailed MR-PheWAS results are available in Supplementary Tables S36-S46.

Figure 15.

(A)-(K) Causal effects of candidate urological cancer targets on 1,431 UK Biobank (UKB) phenotypes assessed through MR-PheWAS

3.9. Mediation Analysis of DNA Methylation, Candidate Target Genes, and Their Association With Urological Tumors

Methylation site data for candidate target genes were sourced using the “IlluminaHumanMethylation450kanno.ilmn12.hg19” R package, with corresponding mQTL data from the GoDMC database. Initially, we assessed the impact of DNA methylation sites on three urological tumors. Methylation sites were clumped using parameters: clump_kb = 10,000, clump_r2 = 0.1, and p-value < 5e-8. All selected SNPs had F-statistics exceeding 10, confirming their suitability as instrumental variables (IVs) for Mendelian randomization (MR) analysis. Subsequently, MR analysis evaluated the relationship between candidate gene methylation sites and urological tumors. The Wald ratio was applied when a single SNP was present, and the IVW method was used for two or more SNPs. In prostate cancer, two methylation sites across two candidate genes demonstrated a causal link to the disease. Similarly, in kidney cancer, one methylation site in a candidate gene showed a causal association. In bladder cancer, four methylation sites within three candidate genes were causally linked to the disease. No horizontal pleiotropy was observed in these findings. We conducted Mendelian randomization (MR) analyses on identified methylation sites and their target genes to assess their regulatory roles in disease development. In prostate cancer, cg11344533 accelerates progression by upregulating SIK2, mediating 88% of the effect (P = 6.98e-05). Similarly, cg12684822 influences cancer development through BNIP2, mediating 93.8% of the effect, though the P-value is not significant, warranting cautious interpretation. In renal cancer, cg10967178 advances papillary renal cell carcinoma via ACOT7, with a 39.6% mediation effect (P = 0.004). In bladder cancer, cg09747827 and cg00784671 promote progression by downregulating PROC and CELSR1, mediating 91.1% (P = 0.00067) and 89% (P = 0.00072) of their effects, respectively. Additionally, cg04382643 and cg25481705 enhance bladder cancer by upregulating KLC3, with mediation effects of 72.7% and 89.1% and P-values of 0.00138 and 0.00092, respectively (Table 2 and Figure 16)). Detailed MR results are available in Supplementary Tables S47 and S48.

Table 2.

Methylation-Mediated Druggable Gene Regulation in Urological Cancers

CpG site	Gene	Disease	βa (path A)	βb (path B)	βc (total)	βd (direct)	βe (indirect)	Proportion mediated (%)	P-value
cg09747827	PROC	BCa	−1.1791	−0.4145	0.5367	0.048	0.4887	91.10%	6.66e-4
cg00784671	CELSR1	BCa	−0.3414	−0.3246	0.1245	0.0137	0.1108	89.00%	7.21e-4
cg04382643	KLC3	BCa	0.3314	0.306	0.1394	0.038	0.1014	72.70%	1.38e-3
cg25481705	KLC3	BCa	0.4125	0.306	0.1416	0.0154	0.1262	89.10%	9.16e-4
cg11344533	SIK2	PCa	0.3694	0.2027	0.085	0.0102	0.0749	88%	6.98e-5
cg12684822	BNIP2	PCa	2.9896	0.2503	0.7977	0.0494	0.7483	93.80%	0.11
cg10967178	ACOT7	RCa	0.8241	0.7887	1.6407	0.9908	0.6499	39.60%	4.03e-3

Figure 16.

(A)-(F) Estimated proportion of the association between DNA methylation and urological malignancies mediated by target genes. (A) cg09747827 mediated by PROC. (B) cg00784671 mediated by CELSR1. (C) cg04382643 mediated by KLC3. (D) cg25481705 mediated by KLC3. (E) cg11344533 mediated by SIK2. (F) cg10967178 mediated by ACOT7

3.10. Candidate Drug Prediction

Using the DSigDB database, this study aimed to identify potential drugs for candidate genes without targeted therapies. We selected compounds based on the adjusted p-value significance and drug effect direction. Our findings revealed that doxorubicin, bexarotene, camptothecin, and alsterpaullone significantly target BNIP2 in prostate cancer (PCa), while pentamidine, equol, diarylpropionitrile, and C-75 target KAT5. In bladder cancer (BCa), L-sulforaphane and taxifolin were linked to GSTM1, and GW-851, camptothecin, and 0175029-000 to GPS1. Meanwhile, dienogest, 60282-87-3, enoxaparin, and norgestimate were associated with PROC. In renal cancer (RCa), ivermectin, minocycline, 5155877, pararosaniline, tribenoside, and naftifine were linked to GRB10. These associations are detailed in Supplementary Table S49.

3.11. Molecular Docking

Molecular docking was utilized to assess candidate drug affinity and target druggability. Autodock Vina version 1.2.2 was used to identify binding sites and interactions between proteins without existing drugs and the candidate compounds, calculating the binding energy for each interaction. Effective docking results are presented in Figure 17, showing each drug forming hydrogen bonds and strong electrostatic interactions with its protein target. Combinations with the lowest binding energy, indicating highly stable binding, were visualized using PyMOL (Figure 18, Table S50).

Figure 17.

The heatmap displays the molecular docking binding affinities between six candidate urological cancer targets (identified through Mendelian randomization) and predicted drug compounds

Figure 18.

Molecular docking results of protein-small molecule complexes with the lowest binding affinities, visualized using PyMOL. (A) BNIP2-doxorubicin complex, (B) KAT5-C75 complex, (C) GRB10-ivermectin complex, (D) PROC-gestodene complex, (E) GPS1-camptothecin complex, (F) GSTM1-taxifolin complex

4. Discussion

This study provides a comprehensive investigation into the causal relationship between 3,887 druggable genes and the risk of urological tumors. Using transcriptome-wide Mendelian randomization (TWMR), 11 candidate genes with strong evidence (Tier 1) were identified. Among these, four druggable genes (BNIP2, KAT5, MMP24, SIK2) with genetically determined higher transcription levels were linked to an increased susceptibility to prostate cancer (PCa). For bladder cancer (BCa), one druggable gene (KLC3) with higher transcription levels and four genes (CELSR1, GPS1, PROC, GSTM1) with lower transcription levels were associated with greater risk. In renal cancer (RCa), two druggable genes (GRB10, GPR17) with lower transcription levels were correlated with an elevated susceptibility. Within renal cancer subgroups, six candidate genes with high evidence were identified: reduced expression of three genes (GPR17, CDKN1A, GRB10) increased susceptibility to ccRCC; elevated expression of two genes (ACOT7, GAB1) raised the risk of pRCC; and in chRCC, only the elevated expression of one gene (SULT1B1) was linked to an increased susceptibility.

The genes with high evidence levels mentioned above not only establish causal associations with relevant urological cancers but are also validated through Bayesian colocalization, SMR and HEIDI tests. This study innovatively investigates the upstream regulatory mechanism of druggable target genes, specifically focusing on DNA methylation regulation, through mediation analysis. It is found that the expression of PROC is regulated by methylation sites cg09747827, CELSR1 by cg00784671, KLC3 by cg04382643 and cg25481705, SIK2 by cg11344533, and ACOT7 by cg10967178. An innovative DNA methylation - druggable gene - urological tumor axis is proposed. Epigenetic abnormalities have been recognized as a hallmark of cancer and are particularly prominent in urinary system tumors.³⁹ Currently known key genes characterized by promoter CpG island hypermethylation in urinary system cancers include BRCA1, VHL, hMLH1, p15 (CDKN2B), p16 (CDKN2A), and the androgen receptor (AR).⁴⁰ Of note, DNA methylation is a reversible process catalyzed by DNA methyltransferases (DNMTs), rendering it amenable to targeted intervention with DNMT inhibitors. Previous studies have demonstrated that in papillary renal cell carcinoma (RCC), the DNMT3A inhibitor ML-1508 restores tumor sensitivity to sunitinib. The combined treatment leads to reduced Ki-67 expression and increased cell necrosis, underscoring the antitumor potential of combining methylation inhibitors with TKI-based regimens.⁴⁶ In prostate cancer, AR promoter methylation may contribute to the development of hormone-resistant phenotypes. In a transgenic mouse model of prostate cancer, the combination of the DNMT inhibitor decitabine with castration therapy significantly delayed the onset of castration resistance compared with castration alone.⁴⁷ Gravina et al further showed that the combination of the DNMT inhibitor azacitidine and the anti-androgen drug bicalutamide induces apoptosis in both AR-positive (22RV1) and AR-negative (PC3) prostate cancer cell lines.⁴⁸ In urothelial carcinoma, hypomethylating agents have also been shown to possess antitumor activity and exhibit synergistic effects when combined with platinum-based drugs.^49,50 Collectively, these findings support the translational potential of therapeutic strategies centered on the “DNA methylation–druggable genes” axis.

We validated the expression of potential target genes using single-cell data. In the druggability analysis for prostate cancer, targeted inhibitors are available for SIK2 and MMP24, namely Fostamatinib and Marimastat, respectively. In ccRCC, Arsenic trioxide, a targeted drug for CDKN1A, has been approved and acts as its activator. Several biochemical molecules targeting GPR17, such as Galactose-uridine-5′-diphosphate, Uridine diphosphate glucose, and Uridine-5′-diphosphate, are currently in the experimental phase. However, their roles as regulatory agents remain uncertain. For bladder cancer, drugs like 5-fluorotryptophan, targeting GSTM1, are also in the experimental phase, and their effects are similarly uncertain. In the druggability assessment, genes with existing targeted therapies were prioritized, as they are established targets for immune diseases and cancers and may be repurposed for urological tumors. For genes lacking targeted therapies, we conducted drug prediction and molecular docking analysis. Molecular compounds exhibiting high binding affinities to their biological targets are widely regarded as possessing enhanced druggability potential.

Our analysis involved candidate genes with reported evidence of urological tumors in terms of gene polymorphisms, mRNA levels, or protein levels from previous genetic or experimental studies. Genes including KAT5, MMP24, SIK2, GRB10, and GSTM1 were prioritized with the most compelling evidence. KAT5, a MYST family histone acetyltransferase, regulates chromatin remodeling and transcription by acetylating histone and non-histone proteins. It acts as a co-activator of androgen receptor (AR) signaling, influencing prostate cancer (PCa) tumorigenesis and progression. Overexpression of KAT5 in DU145 cells enhances proliferation, migration, invasion, and phosphorylated p38 and JNK expression, which can be reversed by si-KAT5 transfection or p38 and JNK signaling inhibition.⁵¹ MMP24, a member of the peptidase M10 family of matrix metalloproteinases (MMPs), participates in ECM breakdown in various physiological and pathological processes. MMPs facilitate cancer cell invasion by degrading the ECM, enabling cell migration. It has been reported that the MMP7 rs11568818 polymorphism is associated with a two-fold increase in the risk of prostate cancer (PCa).⁵² Salt-inducible kinase 2 (SIK2) is a multifunctional kinase in the AMPK family, active in the endoplasmic reticulum membrane. In prostate cancer cells, SIK2 regulates cell cycle progression positively and CREB1 activity negatively. SIK2 knockdown inhibits cell growth, delays cell cycle progression, induces cell death, and enhances CREB1 activity. Expression of a kinase-dead SIK2 mutant has similar effects. Treatment with the SIK2 inhibitor ARN-3236 enhances CREB1 activity in a dose- and time-dependent manner.⁵³ GRB10, an adaptor protein, interacts with various receptors, notably insulin and insulin-like growth factor receptors. Specific isoforms inhibit tyrosine kinase activity, thus impeding growth.Emerging evidence has implicated GRB10 as a putative tumor suppressor across various malignancies. In clear cell renal cell carcinoma (ccRCC), specific methylation events within the GRB10 locus have been identified; reduced promoter methylation, which correlates with increased gene expression, is significantly associated with improved patient outcomes. This clinical observation aligns with GRB10’s established role as a negative regulator of PI3K and insulin signaling. Given that aberrant activation of the PI3K/AKT pathway is a hallmark and a validated therapeutic target in ccRCC, these findings provide a compelling molecular rationale for the use of mTOR and related pathway inhibitors in this disease.⁵⁴ The tumor-suppressive function of GRB10 is further substantiated by studies in other cancer types. Research in small cell lung cancer (SCLC) demonstrates that GRB10 can inhibit oncogenic signaling by negatively regulating the VEGFR2/Akt/mTOR and Akt/GSK-3β/c-Myc axes.⁵⁵ Similarly, in osteosarcoma, GRB10 overexpression has been shown to attenuate the PI3K/AKT pathway activation induced by RILP knockdown, reinforcing its role as a key modulator of this central oncogenic driver.⁵⁶ Building upon these findings and our own Mendelian randomization results, which identify GRB10 as a protective factor for renal cancer, we propose a mechanistic model wherein GRB10 exerts its tumor-suppressive effects through the negative regulation of the VEGFR2/Akt/mTOR signaling cascade. By inhibiting this pathway, GRB10 could suppress renal cancer cell proliferation and survival, thereby counteracting a key driver of tumorigenesis. GSTM1 encodes a glutathione S-transferase that belongs to the mu class. Null mutations in this gene have been linked to an increased risk of several cancers, likely due to heightened susceptibility to environmental toxins and carcinogens. The deletion of the GSTM1 gene may be associated with an elevated risk of MIBC (a 30% increase in risk). The GSTM1-null genotype was more prevalent in the MIBC group than in the control group (OR = 1.30, 95% CI = 1.01–1.68, P = 0.037).⁵⁷

We also identified several novel candidate genes for urological tumors, including BNIP2 for prostate cancer, GPR17 for renal cancer, and CELSR1, PROC, KLC3, and GPS1 for bladder cancer.

BNIP2 is a BH3-only member of the BCL-2 family and plays a critical role in mitochondria-mediated apoptosis. Previous studies have shown that it exerts a protective role in various cancer types. It is negatively regulated by miR-20a, and its downregulation disrupts the balance between pro- and anti-apoptotic factors, thereby conferring survival advantages to colorectal adenocarcinoma cells during chemotherapy-induced apoptosis.⁵⁸ Similarly, in breast cancer, BNIP2 functions as a BCH domain-containing scaffold protein that links RhoA and GEF-H1. Upon microtubule depolymerization, BNIP2 facilitates proper RhoA activation, thereby maintaining normal cell motility and suppressing excessive migration of breast cancer cells.⁵⁹ However, the role of BNIP2 in prostate cancer remains poorly understood. Our study provides evidence that BNIP2 may increase the risk of prostate cancer. Importantly, the cellular context of BNIP2 function in prostate cancer appears distinct from that reported in colorectal and breast cancers. While previous studies focused on BNIP2 expression within cancer cells, our single-cell transcriptomic analysis revealed its predominant enrichment in tumor-associated endothelial cells, suggesting a non-cell-autonomous mechanism. We propose that BNIP2 may promote prostate cancer risk primarily through modulating endothelial function—potentially enhancing angiogenesis, vascular permeability, or endothelial-to-mesenchymal transition—rather than through direct effects on cancer cell apoptosis or migration. This endothelial-centric mechanism represents a paradigm shift from the established tumor cell-intrinsic roles of BNIP2 and highlights its context-dependent versatility across cancer types.

This study, utilizing Mendelian randomization, identifies GPR17 as a novel protective factor for renal cancer, a finding that resonates strongly with recent insights from glioblastoma research. GPR17 has emerged as a promising therapeutic target in glioblastoma and other neurological disorders. Notably, MDL29951, a specific GPR17 agonist, exerts anti-tumor effects by modulating intracellular cAMP levels⁶⁰; its high expression correlates with prolonged patient survival. Mechanistically, GPR17 activation induces apoptosis via the Caspase-3/7 pathway and suppresses core oncogenic signaling, including the PI3K-Akt and MAPK/ERK cascades, leading to G1 phase cell cycle arrest. Furthermore, recent evidence implicates GPR17 in the regulation of mitochondrial electron transport chain complex activity and reactive oxygen species levels, offering a novel dimension to its tumor-suppressive repertoire.⁶¹ Given that the PI3K-Akt-mTOR axis is a well-established driver of renal cancer pathogenesis, we hypothesize that GPR17 may exert its protective effects in this malignancy through analogous negative regulation of this pathway, potentially accompanied by the induction of apoptosis and cell cycle arrest. In addition, sustained activation of the VHL-HIF axis represents a core pathological hallmark of renal cell carcinoma. Core to this process is the extensive crosstalk between the HIF and MAPK signaling pathways. Notably, given this established interplay, GPR17 may modulate HIF activity in a MAPK-dependent manner. These findings position GPR17 not only as a potential prognostic biomarker for renal cancer but also suggest that the therapeutic application of its specific agonists could represent a promising strategy for renal cancer treatment.

CELSR1, an atypical cadherin, is crucial for epithelial planar cell polarity, kidney development, neural progenitor cell fate, and maintaining normal physiological processes. In LUAD, miR-629-5p enhances tumor cell motility by targeting PPWD1. Exosomal miR-629-5p from tumor cells reduces CELSR1 levels in endothelial cells, increasing permeability. Lower CELSR1 levels in invasive LUAD endothelial cells are observed. Restoring CELSR1 expression in endothelial cells inhibits miR-629-5p effects, while CELSR1 inhibition via siRNA increases permeability and tumor cell invasiveness.⁶² This supports the study’s conclusion linking reduced CELSR1 expression to increased bladder cancer risk. This Mendelian randomization analysis identifies high expression of KLC3 as a risk factor for bladder cancer pathogenesis, a finding consistent with its established oncogenic role across multiple malignancies. Recent studies have elucidated context-dependent mechanisms: in ovarian cancer, KLC3 activates the PI3K/AKT signaling pathway by upregulating COL3A1 expression, thereby promoting tumor cell proliferation, migration, and epithelial-mesenchymal transition (EMT)⁶³; in gastric cancer, KLC3 interacts with SLC2A5 and stabilizes its expression, subsequently activating the MAPK signaling pathway to drive tumor progression and EMT.⁶⁴ While the precise mechanism of KLC3 in bladder cancer remains to be fully elucidated, its biological function as a motor protein light chain involved in intracellular cargo transport offers mechanistic insights. GPS1, a gene that inhibits the G protein pathway, has been rarely studied in bladder cancer. In penile cancer, CSN1 (GPS1) acts as a novel tumor suppressor; mutations in this gene disrupt miRNA-mediated gene silencing, contributing to the development of penile cancer.⁶⁵ This study suggests that GPS1 may serve as a protective factor in bladder cancer, warranting further validation.

This study presents significant advantages as the first to systematically identify drug targets for three major urological tumors using Mendelian Randomization (MR) on the druggable genome, leveraging data from the largest GWAS on urological tumor risk. Validation through colocalization and SMR confirmed candidate targets with strong evidence. Single-cell expression analysis further elucidated the potential pathogenic roles of these genes in urological tumors. Enrichment analysis unveiled the functional traits of these causal genes. PPI revealed regulatory relationships among these drug target genes, suggesting alternative pathways for drug development in urological tumors. The final drug prediction emphasized the therapeutic potential of these genes, with molecular docking showing robust binding activity, highlighting their promise as drug targets. Moreover, PheWAS assessed the off-target effects of drugs. In addition to functional exploration, upstream regulatory mechanisms of candidate druggable genes were examined, identifying methylation sites associated with high-evidence targets that influence disease progression by regulating druggable gene expression. This study provides a comprehensive assessment—from gene identification to drug binding characteristics—and proposes 11 drug targets for urological tumors with compelling supporting evidence.

This study offers valuable insights, yet several limitations must be acknowledged. Primarily, the analysis was conducted solely on European populations, leaving the applicability of these findings to other ancestries unverified. Replication is uncommon in Mendelian randomization (MR) studies, yet we successfully replicated MR results for prostate cancer in two independent cohorts, highlighting a key strength of our research. To bolster the reliability of drug target identification, we mandated that genes achieve significance in both GWAS cohorts. This stringent criterion enhanced result robustness and significantly lowered false-positive rates, potentially boosting the clinical applicability of our findings. However, we could not replicate results for renal and bladder cancers in external cohorts, though we conducted subtype-specific analyses for renal cancer to partially mitigate this limitation. A key analytical consideration concerns our significance thresholds. For bladder and kidney cancers, the stringent Bonferroni correction yielded few significant genes, possibly obscuring true biological signals. Consequently, we applied a more lenient threshold (P < 1×10^-3), acknowledging the heightened risk of false positives. This trade-off is typical in genetic studies and warrants further investigation in future research. Second, our evaluation of blood mRNA in urological tumors was limited by data availability, restricting our analysis to mRNA levels of only 7 genes from other tissues. This underscores the necessity for future research that integrates mRNA and protein-level measurements across relevant tissues, particularly the prostate, kidney, and bladder, to enhance our understanding of disease mechanisms. Several methodological considerations merit attention. Our focus solely on cis-eQTLs may overlook significant trans-regulatory effects. Additionally, blood-based druggable genes might be affected by non-genetic factors, as indicated by the varying explanatory power of our genetic instruments (R² = .09-50.51%). These limitations highlight the need for future epidemiological studies to directly measure blood mRNA levels and their associations with urological cancer risk.

5. Conclusions

Several targetable genes linked to urological tumor susceptibility were pinpointed in our research, along with an exploration of the DNA methylation regulatory mechanisms governing these potential targets. These results provide new insights into the etiology of urological tumors and underscore potential avenues for the creation of diagnostic biomarkers and therapeutic drugs. Nevertheless, additional experimental and clinical investigations are necessary to evaluate the practicality and effectiveness of these candidate drugs and confirm the current findings.

Supplemental Material

Supplemental Material - Druggable Genome-Wide Mendelian Randomization Identifies Therapeutic Targets and DNA Methylation-Mediated Regulation in Urological Malignancies

Supplemental Material for Druggable Genome-Wide Mendelian Randomization Identifies Therapeutic Targets and DNA Methylation-Mediated Regulation in Urological Malignancies by Taoen Li, Qiang Wei in Technology in Cancer Research & Treatment.

Footnotes

Acknowledgments

The authors thank all the researchers for sharing the statistics included in this study.

ORCID iD

Taoen Li

Author Contributions

Taoen Li contributed to the study’s conceptualization, methodology design, software development, data validation, formal analysis, investigation, data curation, writing of the original draft, review and editing of the manuscript, and visualization. Qiang Wei provided resources, supervised the research, and managed the project. All authors reviewed and approved the final manuscript.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

All code and data have been uploaded to GitHub ().

Supplemental Material

Supplemental material for this article is available online. Detailed results are presented in .

Appendix

References

Siegel

O’Neil

Richards

Dowling

Weir

. Prostate Cancer Incidence and Survival, by Stage and Race/Ethnicity - United States, 2001-2017. MMWR Morbidity and mortality weekly report. 2020;69(41):1473-1480.

Khorasanchi

Hong

Yang

, et al. Overcoming drug resistance in castrate-resistant prostate cancer: current mechanisms and emerging therapeutic approaches. Cancer drug resistance (Alhambra, Calif). 2025;8:9.

Giudice

Beckermann

Siqueira Do Amaral

Rini

. Immunotherapy Strategies After Immune Checkpoint Inhibitor Exposure in Renal Cell Carcinoma: A Review. JAMA oncology. 2025;11(5):554-561.

Sánchez-Gastaldo

Kempf

González Del Alba

Duran

. Systemic treatment of renal cell cancer: A comprehensive review. Cancer treatment reviews. 2017;60:77-89.

Flaig

Spiess

Agarwal

, et al. Bladder Cancer, Version 3.2020, NCCN Clinical Practice Guidelines in Oncology. Journal of the National Comprehensive Cancer Network: JNCCN. 2020;18(3):329-354.

Shih

Lin

Luo

Sung

. Antibody-drug conjugates targeting HER2 for the treatment of urothelial carcinoma: potential therapies for HER2-positive urothelial carcinoma. Frontiers in pharmacology. 2024;15:1326296.

Arrowsmith

Miller

. Trial watch: phase II and phase III attrition rates 2011-2012. Nature reviews Drug discovery. 2013;12(8):569.

King

Davis

Degner

. Are drug targets with genetic support twice as likely to be approved? Revised estimates of the impact of genetic support for drug mechanisms on the probability of drug approval. PLoS genetics. 2019;15(12):e1008489.

Davitte

Stott-Miller

Ehm

Cunnington

Reynolds

. Integration of Real-World Data and Genetics to Support Target Identification and Validation. Clinical pharmacology and therapeutics. 2022;111(1):63-76.

10.

Abdellaoui

Yengo

Verweij

KJH

Visscher

. 15 years of GWAS discovery: Realizing the promise. American journal of human genetics. 2023;110(2):179-194.

11.

Tam

Patel

Turcotte

Bossé

Paré

Meyre

. Benefits and limitations of genome-wide association studies. Nature reviews Genetics. 2019;20(8):467-484.

12.

Zhu

Zhang

, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nature genetics. 2016;48(5):481-487.

13.

Finan

Gaulton

Kruger

, et al. The druggable genome and support for target identification and validation in drug development. Science translational medicine. 2017;9(383).

14.

Sekula

Del Greco

Pattaro

Köttgen

. Mendelian Randomization as an Approach to Assess Causality Using Observational Data. Journal of the American Society of Nephrology: JASN. 2016;27(11):3253-3265.

15.

Walker

Davey Smith

Davies

Martin

. Mendelian randomization: a novel approach for the prediction of adverse drug events and drug repurposing opportunities. International journal of epidemiology. 2017;46(6):2078-2089.

16.

Pan

Jing

. Exploring druggable targets and inflammation-mediated pathways in cancer: a Mendelian randomization analysis integrating transcriptomic and proteomic data. Inflammation research: official journal of the European Histamine Research Society [et al]. 2025;74(1):46.

17.

Cao

Chen

Wang

, et al. Integrating analysis of multi-omics summary data identifies novel plasma protein biomarkers and drug targets for bladder cancer. Discover oncology. 2025;16(1):660.

18.

Liu

, et al. Potential therapeutic targets for bladder cancer: a proteome-wide Mendelian randomization study. American journal of cancer research. 2025;15(3):1096-1108.

19.

Skrivankova

Richmond

Woolf

BAR

, et al. Strengthening the Reporting of Observational Studies in Epidemiology Using Mendelian Randomization: The STROBE-MR Statement. Jama. 2021;326(16):1614-1621.

20.

Skrivankova

Richmond

Woolf

BAR

, et al. Strengthening the reporting of observational studies in epidemiology using mendelian randomisation (STROBE-MR): explanation and elaboration. BMJ (Clinical research ed). 2021;375:n2233.

21.

Võsa

Claringbould

Westra

, et al. Large-scale cis- and trans-eQTL analyses identify thousands of genetic loci and polygenic scores that regulate blood gene expression. Nature genetics. 2021;53(9):1300-1310.

22.

Hemani

Zheng

Elsworth

, et al. The MR-Base platform supports systematic causal inference across the human phenome. eLife. 2018;7.

23.

Davies

Holmes

Davey Smith

. Reading Mendelian randomisation studies: a guide, glossary, and checklist for clinicians. BMJ (Clinical research ed). 2018;362:k601.

24.

Burgess

Davey Smith

Davies

, et al. Guidelines for performing Mendelian randomization investigations: update for summer 2023. Wellcome open research. 2019;4:186.

25.

Bowden

Davey Smith

Haycock

Burgess

. Consistent Estimation in Mendelian Randomization with Some Invalid Instruments Using a Weighted Median Estimator. Genetic epidemiology. 2016;40(4):304-314.

26.

Bowden

Davey Smith

Burgess

. Mendelian randomization with invalid instruments: effect estimation and bias detection through Egger regression. International journal of epidemiology. 2015;44(2):512-525.

27.

Greco

Minelli

Sheehan

Thompson

. Detecting pleiotropy in Mendelian randomisation studies with summary data and a continuous outcome. Statistics in medicine. 2015;34(21):2926-2940.

28.

Zeng

Zhang

, et al. Integrative analysis of omics summary data reveals putative mechanisms underlying complex traits. Nature communications. 2018;9(1):918.

29.

Butler

Hoffman

Smibert

Papalexi

Satija

. Integrating single-cell transcriptomic data across different conditions, technologies, and species. Nature biotechnology. 2018;36(5):411-420.

30.

Wang

Han

. clusterProfiler: an R package for comparing biological themes among gene clusters. Omics : a journal of integrative biology. 2012;16(5):284-287.

31.

Wishart

Feunang

Guo

, et al. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic acids research. 2018;46(D1):D1074-d1082.

32.

Szklarczyk

Kirsch

Koutrouli

, et al. The STRING database in 2023: protein-protein association networks and functional enrichment analyses for any sequenced genome of interest. Nucleic acids research. 2023;51(D1):D638-d646.

33.

Mizuno

Venkadakrishnan

, et al. Intraindividual epigenetic heterogeneity underlying phenotypic subtypes of advanced prostate cancer. Nature communications. 2025;16(1):5543.

34.

Goel

Tura

, et al. Towards Epigenetic-based Subtyping of Bladder Cancer: DNA Methylation Predicts Molecular Subtypes of Non-muscle-invasive Bladder Cancer. European urology. 2025;87(5):579-581.

35.

Shi

Chen

Wang

, et al. A urine DNA methylation assay for early detection of renal cancer. Oncogene. 2025;44(22):1709-1717.

36.

Yoo

Shin

Kim

, et al. DSigDB: drug signatures database for gene set analysis. Bioinformatics (Oxford, England). 2015;31(18):3069-3071.

37.

Morris

Huey

Olson

. Using AutoDock for ligand-receptor docking. Current protocols in bioinformatics. 2008;Chapter 8:Unit 8.14.

38.

Kim

Chen

Cheng

, et al. PubChem in 2021: new data content and improved web interfaces. Nucleic acids research. 2021;49(D1):D1388-d1395.

39.

Alva

Hahn

Aparicio

Singal

Yellapragada

Sonpavde

. Hypomethylating agents for urologic cancers. Future oncology (London, England). 2011;7(3):447-463.

40.

Christman

. 5-Azacytidine and 5-aza-2'-deoxycytidine as inhibitors of DNA methylation: mechanistic studies and their implications for cancer therapy. Oncogene. 2002;21(35):5483-5495.

41.

Chen

Zhang

Huang

Cai

. Analysis of cancer-related lncRNAs using gene ontology and KEGG pathways. Artificial intelligence in medicine. 2017;76:27-36.

42.

Yao

Zhang

, et al. Lon protease 1-mediated metabolic reprogramming promotes the progression of prostate cancer. Cell death & disease. 2025;16(1):116.

43.

Peng

Jiang

, et al. Altered glycosylation in cancer: molecular functions and therapeutic potential. Cancer communications (London, England). 2024;44(11):1316-1336.

44.

Tamaskar

Choueiri

Sercia

Rini

Bukowski

Zhou

. Differential expression of caveolin-1 in renal neoplasms. Cancer. 2007;110(4):776-782.

45.

Long

Wang

Yang

, et al. Identification of NOTCH4 mutation as a response biomarker for immune checkpoint inhibitor therapy. BMC medicine. 2021;19(1):154.

46.

Marques-Magalhães

Moreira-Silva

Graça

, et al. Combination of MLo-1508 with sunitinib for the experimental treatment of papillary renal cell carcinoma. Frontiers in oncology. 2025;15:1399956.

47.

Zorn

Wojno

McCabe

Kuefer

Gschwend

Day

. 5-aza-2'-deoxycytidine delays androgen-independent disease and improves survival in the transgenic adenocarcinoma of the mouse prostate mouse model of prostate cancer. Clinical cancer research : an official journal of the American Association for Cancer Research. 2007;13(7):2136-2143.

48.

Gravina

Marampon

Di Staso

, et al. 5-Azacitidine restores and amplifies the bicalutamide response on preclinical models of androgen receptor expressing or deficient prostate tumors. The Prostate. 2010;70(11):1166-1178.

49.

Cheng

Weisenberger

Gonzales

, et al. Continuous zebularine treatment effectively sustains demethylation in human bladder cancer cells. Molecular and cellular biology. 2004;24(3):1270-1278.

50.

Liu

Dai

. Study of 5-Aza-CdR on transcription regulation of RASSF1A gene in the BIU87 cell line. Urologia internationalis. 2009;82(1):108-112.

51.

Zhao

Dai

, et al. KAT5 promotes tumor growth in androgen-independent prostate cancer by facilitating aerobic glycolysis. Translational oncology. 2025;58:102436.

52.

Geng

Chen

Huang

Hou

. The prognostic value and potential mechanism of Matrix Metalloproteinases among Prostate Cancer. International journal of medical sciences. 2020;17(11):1550-1560.

53.

Bon

Wadhwa

Schreiner

, et al. Salt-inducible kinase 2 regulates mitotic progression and transcription in prostate cancer. Molecular cancer research: MCR. 2015;13(4):620-635.

54.

Cancer Genome Atlas Research Network . Comprehensive molecular characterization of clear cell renal cell carcinoma. Nature. 2013;499(7456):43-49.

55.

Shang

Shi

, et al. Combination of arsenic trioxide and apatinib synergistically inhibits small cell lung cancer by down-regulating VEGFR2/mTOR and Akt/c-Myc signaling pathway via GRB10. Hereditas. 2024;161(1):29.

56.

Wei

Xia

Zheng

Gong

Guo

. RILP inhibits tumor progression in osteosarcoma via Grb10-mediated inhibition of the PI3K/AKT/mTOR pathway. Molecular medicine (Cambridge, Mass). 2023;29(1):133.

57.

Kang

Song

, et al. Glutathione S-transferase M1 and T1 polymorphisms: susceptibility and outcomes in muscle invasive bladder cancer patients. European journal of cancer (Oxford, England: 1990). 2013;49(14):3010-3019.

58.

Chai

Liu

Tian

Tang

. miR-20a targets BNIP2 and contributes chemotherapeutic resistance in colorectal adenocarcinoma SW480 and SW620 cell lines. Acta biochimica et biophysica Sinica. 2011;43(3):217-225.

59.

Pan

Chew

Wong

DCP

, et al. BNIP-2 retards breast cancer cell migration by coupling microtubule-mediated GEF-H1 and RhoA activation. Science advances. 2020;6(31):eaaz1534.

60.

Saravanan

Palanivel

Yli-Harja

Kandhavelu

. Identification of novel GPR17-agonists by structural bioinformatics and signaling activation. International journal of biological macromolecules. 2018;106:901-907.

61.

Nguyen

Doan

Murugesan

, et al. GPR17 signaling activation by CHBC agonist induced cell death via modulation of MAPK pathway in glioblastoma. Life sciences. 2022;291:120307.

62.

Zhang

Fan

, et al. MiR-629-5p promotes the invasion of lung adenocarcinoma via increasing both tumor cell invasion and endothelial cell permeability. Oncogene. 2020;39(17):3473-3488.

63.

Yang

Zhang

, et al. KLC3 activates PI3K/AKT signaling and promotes ovarian cancer cell proliferation and migration through COL3A1. Oncology reports. 2025;53(6):67.

64.

Chen

, et al. KLC3 drives gastric cancer progression by stabilizing SLC2A5 to activate MAPK signaling and promote epithelial-mesenchymal transition. BMC cancer. 2025;25(1):1746.

65.

Feber

Worth

Chakravarthy

, et al. CSN1 Somatic Mutations in Penile Squamous Cell Carcinoma. Cancer research. 2016;76(16):4720-4727.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

10.96 MB

0.00 MB