Abstract
Large-scale proteomics enables the identification of biomarkers and undulations in metabolic aging. This study aimed to develop a metabolic age (MA) and identify proteomic biomarkers and their undulating changes during metabolic aging. Using UK Biobank data, MA was developed from mortality-associated metabolomic profiles (nuclear magnetic resonance platform) in 203,491 participants. Associations between 2,923 plasma proteins (Olink Explore 3072 platform) and metabolic aging phenotypes, including MA, telomere length, frailty index, incident type 2 diabetes, cardiovascular disease, and mortality, were examined in 24,920 participants via Cox proportional hazards or linear models. Differential expression-sliding window analysis captured protein waves during metabolic aging in 7,092 participants. MA improved the predictions of mortality, cardiovascular disease, and type 2 diabetes beyond conventional risk factors (C-index up to 0.786) and correlated strongly with chronological age (Spearman’s r: 0.876). Sixty proteins were associated with all metabolic aging phenotypes. Among them, growth differentiation factor 15 (GDF15), urokinase plasminogen activator surface receptor (PLAUR), tumor necrosis factor receptor superfamily member 10A (TNFRSF10A), tumor necrosis factor receptor superfamily member 10B (TNFRSF10B), gamma-interferon-inducible lysosomal thiol reductase (IFI30), hepatocyte growth factor (HGF), WAP 4-disulfide core domain protein 2 (WFDC2), collagen alpha-3(VI) chain (COL6A3), polymeric immunoglobulin receptor (PIGR), insulin-like growth factor-binding protein 4 (IGFBP4), and tumor necrosis factor receptor superfamily member 27 (EDA2R) ranked within the top 20 for at least 4 phenotypes based on P values. Pathway analysis highlighted symbiont entry into host cell and cytokine-cytokine receptor interaction in metabolic aging. Proteins showed undulating changes during metabolic aging, with 3 peaks at 44, 51, and 63 years. MA-protein trajectories clustered into 3 groups. Groups 1 and 3 exhibited linear increases with MA, whereas group 2 showed nonlinear increases. In conclusion, the identification of plasma proteomic biomarkers and their undulating changes in metabolic aging provides a critical foundation for developing clinical markers and precision interventions to prevent accelerated metabolic aging.