The GWAS data for prostatic diseases were extracted from FinnGen study and UK Biobank (Table 1). Summary statistics of GWAS data for PCa were retrieved from the UK Biobank, which involved 183,888 participants, and from the FinnGen study, which included 146,465 individuals. For BPH, the data were obtained from UK Biobank (166,988 individuals) and FinnGen study (177,901 individuals). Prostatitis data were derived from the UK Biobank and FinnGen consortium, with sample sizes of 183,888 and 146,043 individuals, respectively. All participants were of European ancestry. A genome-wide meta-analysis was conducted on the independent GWASs of prostatic diseases using a fixed effects model implemented via METAL software (Willer et al., 2010).

Mitochondrial genes were obtained using MitoCarta3.0, with a library containing 1,244 mitochondrial genes (Rath et al., 2021). Quantitative trait loci (QTL) can uncover the relationships between single nucleotide polymorphisms (SNPs) and the levels of transcription, protein, and methylation. We extracted datasets of expression quantitative trait loci (eQTLs) from the eQTLGen (Võsa et al., 2021) and Genotype Tissue Expression (GTEx) (Consortium, 2020), protein quantitative trait loci (pQTLs) form the deCODE consortium (Ferkingstad et al., 2021), and methylation quantitative trait loci (mQTLs), from the LBC_BSGS_meta consortium (McRae et al., 2018) (Table 1). By screening mitochondrial genes, we identified 1,004 expressed genes from the eQTLGen consortium, 1,003 expressed genes from the GTEx consortium in prostate and whole blood tissues, 708 methylated genes with a total of 2,878 methylation sites from the LBC_BSGS_meta consortium, and 278 proteins from deCODE consortium with a threshold of P < 5 × 10⁻⁵.

We estimate the correlation between mitochondrial genes and prostatic diseases at the gene expression, methylation, and protein abundance levels based on top-associated cis-QTLs using SMR analysis (Zhu et al., 2016; Liu et al., 2024). Windows centered on the respective genes (±1,000 kb) were considered, and SNPs were excluded at the threshold of P-value <5 × 10⁻⁵. To address genetic confounders, we incorporated the heterogeneity in dependent instruments (HEIDI) test to distinguish linkage disequilibrium from pleiotropy, where results exceeding the pleiotropy threshold (P-HEIDI <0.01) underwent rigorous quality control exclusion. Associations with the P-value <0.05 and P-HEIDI >0.01 were undertaken for further prioritization. Furthermore, we conducted colocalization analysis to detect shared causal variants between mitochondrial genes and prostatic diseases (Giambartolomei et al., 2014). This approach utilizes Bayesian statistical methods to estimate the probability of colocalization, thereby assessing the likelihood that the same genetic variants influence both phenotypes (Yoshiji et al., 2023; Battle et al., 2017). A posterior probability of H4 (PPH4) > 0.70 or H3 (PPH3) + H4 (PPH4) > 0.70 was regarded as evidence of colocalization (Chen et al., 2024).

We integrated these results from distinct molecular levels. To address the interpretational constraints of single-omic approaches, cross-validation through at least two omics layers was mandatory for each candidate gene in our tri-layered classification. We classified the candidate genes into three tiers (Figure 1): Tier 1) Tier 1 genes are defined as those associated with prostatic diseases at the cross-validation of transcriptomics both in eQTLgen consortium and GTEx consortium (P-value <0.05), with positive posterior probability in colocalization analysis, and showing associations with prostatic diseases at both methylation and protein levels (P-value <0.05). Tier 2) Tier 2 genes are regarded as those associated with prostatic diseases at the level of gene expression both in eQTLgen consortium and GTEx tissues (P-value <0.05), with evidence of colocalization, and presenting positive evidence either the protein or methylation level (P-value <0.05). Tier 3) Tier 3 genes are defined as those associated with prostatic diseases at the gene expression level in eQTLgen (P-value <0.05), with co-localized evidence, and positive evidence at either the protein or methylation levels (P-value <0.05).

To further validate the prioritized genes, we explored the expression patterns between normal and tumor, hyperplasia, and inflammation samples of prostate through RNA-sequencing data. We extracted two independent datasets for BPH (GSE7307 and GSE132714) and PCa (GSE46602 and GSE70768) from Gene Expression Omnibus (GEO) database. Due to the unavailability of human prostate tissue sequencing data from prostatitis patients, we utilized the murine prostate tissue RNA-sequencing data from established prostatitis mouse models at Anhui Medical University (Zhang et al., 2022).

To explore the biological mechanisms underlying these genetic associations, we conducted a mediation analysis with 1,400 metabolomics, and 731 immunomics. A two-step mediation analysis was conducted to evaluate the potential mediating effects of prioritized mitochondrial gene-related protein, metabolite, and immune characteristics on the causal associations between mitochondrial genes and prostatic diseases, with the basis of TwoSampleMR R package. Specifically, we initially estimated the causal effects of mitochondrial genes on potential mediators (metabolomics and immunomics), and we subsequently measured the effects of mediators on prostatic diseases. The mediation effects were calculated through the “product of coefficients” method, namely, mediation effect = mediation effect A * mediation effect B, the mediation effect A refers to the estimates in the first step, and the mediation effect B refers to the estimates in the second step. Total effect meant the effect of mitochondrial genes on prostatic diseases. Only mediators whose effect directions of total effect and mediation effect were consistent were included in the mediation analysis.

We conducted a genome-wide meta-analysis to aggregate the GWAS data of BPH from UK Biobank and FinnGen. Multivariate GWAS identified 61 loci and 7,756 SNPs (P-value <5 × 10⁻⁸) associated with the genic architecture of BPH, and the Manhattan plot of the GWAS was shown in Figure 2A. We conducted multi-omics analysis using the eQTLs, mQTLs, and pQTLs databases. Initially, based on the eQTLGen database, we identified 69 significant genes (Supplementary Figure S1A; Supplementary Table S1), 12 of which were validated by colocalization (Figures 2B,C). Furthermore, through the GTEx database, we identified 40 positive genes in blood tissue and 20 positive genes in prostate tissue ((Supplementary Figure S1B,C; Supplementary Table S2,3). Among the 12 genes with colocalization evidence, only LYPLAL1, SPRYD4, DCXR, ELAC2, and NOA1 in blood tissue, and ELAC2 in prostate tissue showed significant evidence (Figure 2C). In addition, we identified 12 proteins related to BPH via the deCODE database ((Supplementary Figure S1D; Supplementary Table S4), of which only the DCXR presented the colocalization evidence (Figure 2C). Using the mQTL database, we identified 216 methylation sites ((Supplementary Figure S1E; Supplementary Table S5), 10 of them showed the evidence of colocalization (ACAT1: cg14994056, cg08152564, cg04873221, cg19829446; NOA1: cg14719990, cg00922110, cg08905166; DCXR: cg07073120, cg27226927; ELAC2: cg13723217) (Figure 2C). After consolidating the multi-omics analysis results, we identified one tier 1 gene (DCXR), two tier 2 genes (NOA1 and ELAC2), and one tier 3 gene (ACAT1) (Figure 2D).

Multivariate GWAS identified 6 loci and 44 SNPs (P-value <5 × 10⁻⁸) associated with the genetic inheritance of prostatitis (Figure 3A). We further analyzed mitochondrial genes related to prostatitis. Initially, using the eQTLGen database, we identified 52 significant genes (Supplementary Figure S2A; Supplementary Table S6), 2 of which were validated by colocalization (Figure 3B). Furthermore, through the GTEX database, we identified 26 positive genes in blood tissue and 6 positive genes in prostate tissue (Supplementary Figure S2B,C; Supplementary Table S7,8). Among them, only TRMU in blood tissue presented evidence of colocalization (Figure 3C). In addition, we identified 10 proteins related to prostatitis via the deCODE database (Supplementary Figure S2D; Supplementary Table S9), although none showed colocalized evidence (Figure 3C). Using the mQTL database, we identified 118 methylation sites (Supplementary Figure S2E; Supplementary Table S10), of which only 3 sites presented colocalization evidence (SFXN5: cg03344820, cg23482839; TRMU: cg20376123). After consolidating the multi-omics analysis results, we identified two tier 3 genes (TRMU and SFXN5) (Figure 3C).

For PCa, genome-wide meta-analysis identified 64 loci and 5,011 SNPs (P-value <5 × 10⁻⁸) (Figure 4A). For mitochondrial genes associated with PCa, we identified 86 significant genes using the eQTLGen database (Supplementary Figure S3A; Supplementary Table S11), 12 of which passed colocalization validation (Figure 4B). Furthermore, through the GTEX database, we identified 41 positive genes in blood and 17 positive genes in prostate tissue (Supplementary Figure S3B,C; Supplementary Tables S12, 13). Among them, only NBR1 in blood tissue presented colocalized evidence (Figure 4C). Using the deCODE database, we identified 16 proteins related to PCa (Supplementary Figure S2D; Supplementary Table S14), of which NBR1 and PUS1 showed the evidence of colocalization (Figure 4C). Through the mQTL database, we identified 270 methylation sites, (Supplementary Figure S2E; Supplementary Table S15). Among them, 11 sites were selected (NDUFS6: cg17290868, cg00561739, cg24139843, cg15671317, cg00232265; PUS1: cg13382100; NBR1: cg05368731; GLOD4: cg17267398; PCBD2: cg05713859) (Figure 4C). After summarizing the multi-omics analysis results, we identified 6 tier 3 genes: MRPL24, NDUFS6, PUS1, NBR1, GLOD4, and PCBD2 (Figure 4D).

After integrating the multi-omics evidence, we prioritized one gene with tier 1 evidence, two genes with tier 2 evidence, and nine genes with tier 3 evidence associated with prostatic diseases (Table 2). Notably, all significant mitochondrial signatures presented a consistent effect direction both in gene expression (eQTLGen and GTEx datasets) and protein levels, while some methylation sites showed different directions. Among them, DCXR was identified as a tier 1 gene related to BPH, which was validated in transcriptomics, proteomics, and methylomics across three dimensions. In the results from eQTLGen, it was identified as a protective gene [OR (95% CI) = 0.99 (0.98, 1.00), P-value = 0.03] and was validated in the colocalization analysis (PP.H3 + PP.H4 = 0.951). It was also found to be a protective effect in both GTEx [OR (95% CI) = 0.96 (0.93, 1.00), P-value = 0.03] and pQTL [OR (95% CI) = 0.907 (0.835, 0.984), P-value = 0.02]. Moreover, we identified two key methylation sites for DXCR, cg07073120 [OR (95% CI) = 1.04 (1.01, 1.08), P-value = 0.03] and cg27226927 [OR (95% CI) = 1.01 (1.00, 1.02), P-value = 0.03], despite the inconsistency in their effect directions compared to the previous analysis. NOA1 and ELAC2 were identified as a tier 2 genes related to BPH, exhibiting protective effects both in eQTL and pQTL levels. For prostatitis, TRMU and SFXN5 were identified as tier 3 genes, with TRMU emerging as a risk gene [eQTLGen: OR (95% CI) = 1.67 (1.29, 2.15), P-value = 8.96 × 10⁻⁵], and SFXN5 as a protective factor [eQTLGen: OR (95% CI) = 0.89 (0.80, 0.98), P-value = 0.02]. We further prioritized six tier 3 genes associated with PCa, of which MRPL24, NDUFS6, and GLOD4 were related to decreased risk of PCa, and PUS1, NBR1, and PCBD2 were related to increased risk.

We validated theses prioritized genes through RNA-sequencing data. For those 4 genes in BPH, only DCXR was cross-validated in two datasets (GSE132714: P-value <0.05; GSE7307: P-value <0.05), while ELAC2 was validated in GSE7307 (P-value <0.05) (Figure 2E). Neither of the two candidate genes in prostatitis presented supporting evidence from RNA-sequencing (Figure 3D). For 6 prioritized genes in PCa, GLOD4 (GSE46602: P-value <0.001; GSE70768: P-value <0.05), MRPL24 (GSE46602: P-value <0.001; GSE70768: P-value <0.001), and PUS1 (GSE46602: P-value <0.001; GSE70768: P-value <0.001) were cross-validated in two RNA-sequencing datasets (Figure 4E).

We performed mediation analyses to identify metabolomic, and immunologic mediators linking mitochondrial dysfunction to prostatic diseases. Screening 731 immune traits, and 1,400 circulating metabolites, we identified 25 immune features and 66 metabolites associated with BPH (Figure 5A; Supplementary Table S16), and, 23 immune features, and 57 metabolites linked to prostatitis (Figure 5B; Supplementary Table S17). Besides, we identified 23 immune features and 77 metabolites associated with PCa (Figure 5C; Supplementary Table S18). We also explored the causal effects of mitochondrial genes (within the 3-tiers evidence) on significant immune and metabolic features (Supplementary Table S19–21), and conducted a two-step analysis to estimate the mediation effects (Supplementary Table S22–24). We found 8, 4, and 13 metabolites mediating the interaction between mitochondrial genes and BPH, prostatitis, and PCa, respectively, without the involvement of immune characters (Figure 5D). Among them, 2,3-dihydroxy-2-methyl-butyrate levels exhibited a significant mediation effect linking the DCXR gene and BPH with a mediation proportion of 22.0%, and the level of 3-methyl-2-oxovalerate linking PUS1 and PCa with mediation proportion of 15.5%, sulfate levels mediated association between TRMU and prostatitis with 5.9% (Supplementary Table S22–24).