Longitudinal association between DNA methylation and type 2 diabetes: findings from the KORA F4/FF4 study

Abstract

Background: Type 2 diabetes (T2D) has been linked to changes in DNA methylation levels, which can, in turn, alter transcriptional activity. However, most studies for epigenome-wide associations between T2D and DNA methylation comes from cross-sectional design. Few large-scale investigations have explored these associations longitudinally over multiple time-points.

Methods: In this longitudinal study, we examined data from the Cooperative Health Research in the Region of Augsburg (KORA) F4 and FF4 studies, conducted approximately seven years apart. Leucocyte DNA methylation was assessed using the Illumina EPIC and 450K arrays. Linear mixed-effects models were employed to identify significant associations between methylation sites and diabetes status, as well as with fasting plasma glucose (FPG), hemoglobin A1c (HbA1c), homoeostasis model assessment of beta cell function (HOMA-B), and homoeostasis model assessment of insulin resistance (HOMA-IR). Interaction effects between diabetes status and follow-up time were also examined. Additionally, we explored CpG sites associated with persistent prediabetes or T2D, as well as the progression from normal glucose tolerance (NGT) to prediabetes or T2D. Finally, we assessed the associations between the identified CpG sites and their corresponding gene expression levels.

Results: A total of 3,501 observations from 2,556 participants, with methylation measured at least once across two visits, were included in the analyses. We identified 64 sites associated with T2D including 15 novel sites as well as known associations like those with the thioredoxin-interacting protein (TXNIP) and ATP-binding cassette sub-family G member 1 (ABCG1) genes. Of these, eight CpG sites exhibited different rates of annual methylation change between the NGT and T2D groups, and seven CpG sites were linked to the progression from NGT to prediabetes or T2D, including those annotated to mannosidase alpha class 2a member 2 (MAN2A2) and carnitine palmitoyl transferase 1 A (CPT1A). Longitudinal analysis revealed significant associations between methylation and FPG at 128 sites, HbA1c at 41 sites, and HOMA-IR at 57 sites. Additionally, we identified 104 CpG-transcript pairs in whole blood, comprising 40 unique CpG sites and 96 unique gene transcripts.

Conclusions: Our study identified novel differentially methylated loci linked to T2D as well as to changes in diabetes status through a longitudinal approach. We report CpG sites with different rates of annual methylation change and demonstrate that DNA methylation associated with T2D is linked to following transcriptional differences. These findings provide new insights into the molecular mechanisms of diabetes development.

Keywords: DNA methylation; Diabetes progression; Gene expression; Glycemic traits; Type 2 diabetes.

Fig. 1

Illustration of the selection criteria for study participants and CpG sites included in the analysis

Fig. 2

Miami plot illustrating EWAS results associated with T2D. The x axis indicates the chromosome location, and the y-axis represents the −log10 (p-value). The Bonferroni threshold of 1.34×10−7 is marked by a red dashed line, while the Benjamini–Hochberg (FDR) threshold (p_FDR < 0.05) is indicated by a blue solid line. The upper side represents the positive estimates, and the lower side represents the negative estimates

Fig. 3

Volcano plots illustrating the results for glycemic traits. The x axis indicates the effect size, and the y-axis represents the −log10 (p-value). The Bonferroni threshold of p=1.34×10−7 is marked by a red dashed line, while the Benjamini–Hochberg (FDR) threshold (p_FDR < 0.05) is indicated by a blue dashed line. (A) Volcano plot for FPG. (B) Volcano plot for HbA1c. (C) Volcano plot for HOMA-B. (D) Volcano plot for HOMA-IR.

Fig. 4

Line plots illustrating the rate of methylation change over time for the NGT and T2D groups. The red and blue line represents the individuals with NGT and T2D, respectively. (A) cg19693031 (TXNIP); (B) cg00574958 (CPT1A); (C) cg15418499 (IL18); (D) cg20507228 (MAN2A2).

Fig. 5

Venn diagram illustrating the overlap of CpG sites (with annotated gene names) in the sensitivity analysis. The light cyan colour represents the number of significant CpG sites associated with T2D in the main analysis with all individuals. The greyish-yellow colour represents the number of significant CpG sites associated with T2D in the extended models with all individuals. The light pink colour represents the number of significant CpG sites associated with T2D from individuals with repeated methylation measurements at two time points.

Fig. 6

Volcano plots illustrating the association between DNA methylation and changing diabetes status over time. The x axis indicates the effect size, and the y-axis represents the −log10 (p-value). The Bonferroni threshold of 2.27×10−3 is marked by a red dashed line, while the Benjamini–Hochberg (FDR) threshold (p_FDR < 0.05) is indicated by a blue dashed line. (A) Volcano plot for the persistent prediabetes or T2D. (B) Volcano plot for the progression of diabetes.