Integrative plasma proteomic network analysis identifies physical activity-associated protein modules potentially mediating type 2 diabetes risk

Moderate-to-vigorous physical activity (MVPA) lowers the risk of type 2 diabetes (T2D). Identifying protein modules that mediate this association may help dissect these biological underpinnings, but they remain incompletely characterized. We studied 28,842 diabetes-free UK Biobank participants with baseline plasma proteomics. Guideline-adherent MVPA was defined as ≥ 600 metabolic equivalent of task-minutes per week. We first evaluated associations of MVPA with incident T2D and 2910 plasma proteins, followed by mediation analyses of individual proteins. We then identified protein co-expression modules using multiscale embedded gene co-expression network analysis (MEGENA), tested module-level associations with MVPA, and mediation of incident T2D. Protein modules were further evaluated using module-level Mendelian randomization, whole-genome-sequencing genome-wide association studies, human pancreatic islet single-cell RNA sequencing, leave-one-out sensitivity analyses, and pathway enrichment analyses. Over a median 13.60 years (IQR: [12.82, 14.35]), 1225 participants developed incident T2D. Guideline-adherent MVPA was associated with lower T2D risk and with 650 circulating proteins, of which 451 were Bonferroni-significant mediators. MEGENA analysis identified 171 robust modules, with 67 mediating the MVPA-T2D association. Twelve terminal modules, representing the finest-resolution functions, were selected for further analysis based on their significant mediated effects and enrichment in potential individual mediators. Genetic and functional analyses highlighted two modules: CKB-IGFBP2-centric module was positively associated with MVPA ( β [95% CI] 0.196 [0.168, 0.224]) and genetically associated with lower T2D risk (OR [95% CI] 0.57 [0.43, 0.77]). Conversely, RIDA-PCBD1-centric module was inversely associated with MVPA (− 0.134 [− 0.162, − 0.108]) and genetically associated with higher T2D risk (1.54 [1.13, 2.09]). Compared to individual protein analyses, these modules captured unique pathways, supporting the CKB-IGFBP2-centered and RIDA-PCBD1-centered modules as antigen-presentation-enriched and complement/apoptic-cell-clearance-enriched modules, respectively. Finally, the constituent genes of both modules were enriched in T2D-related gene changes in pancreatic β-cells, with the antigen-presentation-enriched module showing additional enrichment across α-, δ-, and ductal cells. MVPA may reduce T2D risk through two coordinated plasma protein networks rather than only through individual circulating proteins. Evaluating coordinated protein modules may provide a system-level biological framework for understanding how exercise protects against T2D.