Epigenome-wide association study of incident type 2 diabetes in Black and White participants from the Atherosclerosis Risk in Communities Study

Abstract

Aims/hypothesis

DNA methylation studies of incident type 2 diabetes in US populations are limited and to our knowledge none include individuals of African descent. We aimed to fill this gap by identifying methylation sites (CpG sites) and regions likely influencing the development of type 2 diabetes using data from Black and White individuals from the USA.

Methods

We prospectively followed 2091 Black and 1029 White individuals without type 2 diabetes from the Atherosclerosis Risk in Communities study over a median follow-up period of 17 years, and performed an epigenome-wide association analysis of blood-based methylation levels with incident type 2 diabetes using Cox regression. We assessed whether significant CpG sites were associated with incident type 2 diabetes independently of BMI or fasting glucose at baseline. We estimated variation in incident type 2 diabetes accounted for by the major non-genetic risk factors and the significant CpG sites. We also examined groups of methylation sites that were differentially methylated. We performed replication of previously discovered CpG sites associated with prevalent and/or incident type 2 diabetes. All analyses were adjusted for batch effects, cell-type proportions and relevant confounders.

Results

At an epigenome-wide threshold (10⁻⁷), we detected seven novel diabetes-associated CpG sites, of which the sites at MICOS10 (cg05380846: HR 0.89, p=8.4 × 10⁻¹²), ZNF2 (cg01585592: HR 0.88, p=1.6 × 10⁻⁹), JPH3 (cg16696007: HR 0.87, p=7.8 × 10⁻⁹) and GPX6 (cg02793507: HR 0.85, p=2.7 × 10⁻⁸; cg00647063: HR 1.20, p=2.5 × 10⁻⁸) were identified in Black adults; chr17q25 (cg16865890: HR 0.8, p=6.9 × 10⁻⁸) in White adults; and chr11p15 (cg13738793: HR 1.11, p=7.7 × 10⁻⁸) in the meta-analysed group. The JPH3 and GPX6 sites remained epigenome-wide significant on adjustment for BMI, while only the JPH3 site retained significance after adjusting for fasting glucose. We replicated known type 2 diabetes-associated CpG sites, including cg19693031 at TXNIP, cg00574958 at CPT1A, cg16567056 at PLCB2, cg11024682 at SREBF1, cg08857797 at VPS25 and cg06500161 at ABCG1, three of which were replicated in Black adults at the epigenome-wide threshold and all of which had directionally consistent effects. We observed a modest increase in type 2 diabetes variance explained by the significantly associated CpG sites over and above traditional type 2 diabetes risk factors and fasting glucose (26.2% vs 30.5% in Black adults; 36.9% vs 39.4% in White adults). At the Šidák-corrected significance threshold of 5%, our differentially methylated region (DMR) analyses revealed several clusters of significant CpG sites, including a DMR consisting of a previously discovered CpG site at ADCY7 (p_Black=1.8 × 10⁻⁴, p_White=3.6 × 10⁻³, p_All=1.6 × 10⁻⁹) and a DMR consisting of the promoter region of TP63 (p_Black=7.4 × 10⁻⁴, p_White=3.9 × 10⁻³, p_All=1.4 × 10⁻⁵), which were differentially methylated across all racial and ethnic groups.

Conclusions/interpretation

This study illustrates improved discovery of CpG sites and regions by leveraging both individual CpG site analysis and DMR analyses in an unexplored population. Our findings include genes linked to diabetes in experimental studies (e.g. GPX6, JPH3 and TP63). The JPH3 and GPX6 sites were likely associated with incident type 2 diabetes independently of BMI. All the CpG sites except that at JPH3 were likely consequences of elevated glucose. Replication in African-descent individuals of CpG sites previously discovered mostly in individuals of European descent indicates that some of these methylation–type 2 diabetes associations are robust across racial and ethnic groups. This study is a first step towards understanding the influence of methylation on the incidence of type 2 diabetes and its disparity in two major racial and ethnic groups in the USA. It paves the way for future studies to investigate causal relationships between type 2 diabetes and the CpG sites and potentially elucidate molecular targets for intervention.

Graphical Abstract