Integrative bioinformatics and machine learning reveal hub genes and immune signatures bridging type 2 diabetes mellitus, fracture susceptibility, and osteoblast differentiation dysfunction

Abstract

To investigate the causal relationship between type 2 diabetes mellitus (T2DM) and fracture and identify molecular targets linking T2DM to impaired osteogenesis, we performed two-sample Mendelian randomization (MR) using large-scale GWAS data, followed by integrated bioinformatic and machine learning analyses of T2DM-, fracture-, and osteoblast differentiation-related datasets. Hub genes were identified through WGCNA, differential expression analysis, PPI network analysis, and multiple machine learning algorithms, and further evaluated by GSEA, cross-dataset validation, and immune cell infiltration analysis. In vitro validation was performed in MC3T3-E1 preosteoblasts cultured under normal-glucose, high-glucose, and osmotic-control conditions using Alizarin Red S staining, qRT-PCR, and Western blotting. MR analysis confirmed a positive causal effect of T2DM on fracture risk without evidence of horizontal pleiotropy. A total of 280 overlapping genes were identified, mainly enriched in extracellular matrix organization, ossification, bone remodeling, osteogenic differentiation, and pathways related to the cell cycle and AGE-RAGE signaling. PLK1, CCNA2, and CHEK1 were identified as hub genes and were associated with cell cycle regulation. Cross-dataset validation confirmed their upregulation in T2DM patients, while immune infiltration analysis indicated a predominance of neutrophils. High-glucose exposure markedly impaired osteogenic mineralization in MC3T3-E1 cells and dysregulated PLK1, CCNA2, and CHEK1 expression at both the mRNA and protein levels. Collectively, T2DM may increase fracture risk and impair bone regeneration through dysregulated cell cycle and AGE-RAGE signaling, with PLK1, CCNA2, and CHEK1 representing potential therapeutic targets.