A multi-omics Mendelian randomization framework identifies PM2.5-annotated candidate regulatory genes and methylation-linked pathways in diabetic nephropathy

Fine particulate matter (PM2.5) is an established environmental risk factor for chronic kidney disease, yet the molecular pathways linking PM2.5 exposure to diabetic nephropathy (DN) progression remain to be elucidated. This study employs a multi-omics causal inference framework to identify and characterize candidate genetic and epigenetic mediators that may underlie this association.

Utilizing a curated set of PM2.5-annotated genes, we performed summary-data-based Mendelian randomization (SMR) analyses, integrating expression, protein, and methylation quantitative trait loci (eQTLs, pQTLs, mQTLs) with DN genome-wide association study (GWAS) summary statistics. Bayesian colocalization was used to distinguish shared from distinct causal variants. The expression of prioritized genes was validated in an independent DN transcriptomic dataset (GSE96804) and through in vitro PM2.5 exposure experiments. A predictive nomogram for DN risk was subsequently developed and evaluated.

Transcript-level SMR identified ten PM2.5-associated genes (e.g., TNF, NOS2, DUSP1) with genetically predicted causal effects on DN susceptibility. Bayesian colocalization revealed strong evidence for a shared causal variant only at EIF2AK3 (PP.H4 = 0.90). Methylation analysis revealed 48 CpG sites significantly associated with DN risk, mapping to 28 genes, with TNF methylation showing the strongest effect. Multi-omics integration and protein-protein interaction analysis highlighted TNF as a central hub within an enriched inflammatory network. Furthermore, mQTL-eQTL SMR showed that hypermethylation at cg11484872 suppressed TNF, whereas methylation at cg14968860 and cg18091275 upregulated DUSP1 and NOS2, respectively, establishing bidirectional epigenetic regulation of these genes. In DN kidney tissue, TNF and NOS2 were upregulated while DUSP1 was downregulated, and these changes were recapitulated in cellular models following PM2.5 exposure. A nomogram based on NOS2 and TNF yielded an AUC of 0.766 in the training set and 0.726 in an independent external cohort (GSE142153).

Our integrative analysis identifies a PM2.5-responsive epigenetic-gene regulatory network involving TNF, NOS2, and DUSP1 that may represent candidate mediators of PM2.5-related DN progression. These findings nominate molecular candidates for environmental kidney injury and highlight potential biomarkers for risk stratification.