Multi-omics analysis reveals metabolic reprogramming and CTH downregulation in diabetic kidney disease.
文献摘要
Objective To characterize metabolic and transcriptional alterations in diabetic kidney disease (DKD) and identify diagnostic biomarkers using a multi-omics approach. Materials and methods Mendelian randomization (MR) was used to assess causal relationships between metabolites and DKD. Renal transcriptomic datasets validated pathway and gene alterations. Machine learning models evaluated diagnostic performance, with SHAP for interpretation. Single-cell and spatial transcriptomics defined cell-specific expression patterns. In silico gene perturbation and in vitro validation in high-glucose-treated HK-2 cells were conducted. Results MR analysis identified 60 plasma metabolites whose genetically predicted levels showed evidence of association with DKD. These metabolites were enriched in one-carbon metabolism, histidine metabolism, and glycine, serine, and threonine metabolism, which showed consistent alterations in renal transcriptomic analyses. Machine learning models showed strong diagnostic performance (mean AUC = 0.963), with MAOA, CTH, and ALDH1L1 as key contributors. CTH was specifically expressed in proximal tubular cells and downregulated in injury-associated subpopulations. Functional analyses indicated that CTH affected epithelial integrity, mitochondrial metabolism, and immune responses, and was suppressed under high-glucose conditions. Conclusion These findings provide insight into DKD pathogenesis and highlight CTH as a potential biomarker and therapeutic target.