Cross-sectional and longitudinal association of seven DNAm-based predictors with metabolic syndrome and type 2 diabetes.

Abstract

Background: To date, various epigenetic clocks have been constructed to estimate biological age, most commonly using DNA methylation (DNAm). These include "first-generation" clocks such as DNAmAgeHorvath and "second-generation" clocks such as DNAmPhenoAge and DNAmGrimAge. The divergence of one's predicted DNAm age from chronological age, termed DNAmAge acceleration (AA), has been linked to mortality and various aging-related conditions, albeit with varying findings. In metabolic syndrome (MetS) and type 2 diabetes (T2D), it remains inconclusive which DNAm-based predictor(s) is/are closely related to these two metabolic conditions. Therefore, we examined the cross-sectional associations between seven DNAm-based predictors and prevalent metabolic conditions in participants with methylation data from the KORA study. We also analyzed the longitudinal association with time-to-incident T2D and the relative prognostic value compared to clinical predictors from the Framingham 8-year T2D risk function in predicting incident disease over eight years.

Results: GrimAA and PhenoAA difference demonstrated consistently significant associations in the cross-sectional and longitudinal analyses. GrimAA difference reported a larger effect: with prevalent MetS at F4 (odds ratio = 1.09, 95% confidence interval = [1.06-1.13], p = 2.04E-08), with prevalent T2D at F4 (odds ratio = 1.09 [1.04-1.13], p = 1.38E-04) and with time-to-incident T2D (hazards ratio = 1.05 [1.01-1.10], p = 0.02) for each year increase in GrimAA difference. Mortality risk score was significantly associated with both prevalent metabolic conditions but not in the longitudinal analysis. The inclusion of DNAm-based predictor in the model with Framingham clinical predictors improved discriminative ability, albeit not significantly. Notably, the DNAm-based predictor, when fitted separately, showed a discriminative ability comparable to that of the model with clinical predictors. Overall, no clear pattern of significant associations was identified in the epigenetic measures from the "first-generation" clocks.

Conclusions: GrimAA, PhenoAA difference and mortality risk score, derived from the "second-generation" clocks, demonstrated significant associations with both MetS and T2D. These DNAm-based predictors may be useful biomarkers for risk stratification and disease prognosis in our study sample of European ancestry. Further research is warranted to investigate the generalizability of our findings across different ancestries and to examine the underlying shared biological mechanisms.