Context and aims: Studies investigating the association between DNA methylation and Type 2 Diabetes Mellitus (T2DM) have focused on either prevalent or incident T2DM. We aimed to examine, using a longitudinal study design, blood DNA methylation associated with both prevalent and incident T2DM cases.

Methods: Participants in the Melbourne Collaborative Cohort Study self-reported their diabetic status at baseline (1990-1994, N=106 with T2DM), wave 1 (1995-1998, N=76 incident cases), and wave 2 (2003-2007, N=143 incident cases). We conducted an epigenome-wide association study of DNA methylation, measured using the Illumina HM450K assay in baseline blood samples, and T2DM. Mixed-effects linear regression models were used to compare, at each wave, methylation levels between cases and non-cases of T2DM, adjusting for age, sex, ethnicity, smoking, alcohol drinking, socio-economic status, physical activity, diet, and BMI.

Findings: The strongest signal was observed at a CpG site in the U’3TR region of TXNIP, which is involved in the regulation of glucose metabolism. At baseline, the regression coefficient comparing prevalent cases to non-cases was β= -0.26 [95% CI: -0.36 to -0.16]. This association was attenuated for incident disease before wave 1, β= -0.14 [-0.25 to -0.03], and for incident disease between wave 1 and wave 2, β= -0.10 [-0.18 to -0.02]. Other strong associations were observed for example at CPT1A (baseline: β= -0.17 [-0.28 to -0.07], wave 2: β= -0.21 [-0.32 to -0.10]) and ABCG1 (baseline: β=0.09 [0.03 to 0.15], wave 2: β=0.11 [0.05 to 0.16]), which are involved in fatty acid oxidation and the regulation of blood lipids, respectively.

Innovative contribution to policy, practice and/or research: Methylation markers at genes relevant to the pathophysiology of T2DM, measured in peripheral blood DNA, are associated with both prevalent and incident disease and thus might be useful in the risk stratification and management of TD2M.