Causal relationship between epigenetic markers and type 2 diabetes in West African populations: a Mendelian randomisation analysis

Aims/hypothesis: Evidence for a causal role of DNA methylation sites (CpGs) in type 2 diabetes and glycaemic traits is limited due to the cross-sectional nature of many epigenome-wide association studies (EWAS). In addition, epigenetic studies in West African populations are particularly sparse, despite the high and rising burden of type 2 diabetes in these populations. Hence, we aimed to identify CpGs causally associated with type 2 diabetes among West Africans by leveraging Mendelian randomisation (MR) analysis and longitudinal data. Methods: We used the Illumina EPIC DNA methylation array to profile the methylation of DNA extracted from white blood cells collected from 879 Ghanaian individuals (the Research on Obesity and Diabetes among African Migrants [RODAM] study) and 332 Nigerian individuals (the Africa America Diabetes Mellitus [AADM] study) who were not on glucose-lowering medication. We carried forwards CpGs identified in EWAS for type 2 diabetes and meta-analysed EWAS for HbA_1c and homeostatic model assessment estimates of insulin sensitivity (HOMA-S) as exposures to two-sample MR analysis. Independent cis methylation quantitative trait loci (meQTLs) were calculated using methylation data from blood and primary hepatocytes and subsequently used as instrumental variables (SNP–exposure associations). Genome-wide association analyses for type 2 diabetes on 4120 participants from the AADM study were used to derive the SNP–outcome associations. Longitudinal trait data (n=138) and RNA-seq data (n=77 blood, 49 adipose, 55 muscle) available for a subset of Nigerians were used for follow-up analyses. Results: We identified 28 CpGs associated with type 2 diabetes, 26 with HbA_1c and three with HOMA-S (total CpGs: 57), of which 49 had meQTLs in blood (AADM study data) and four had meQTLs in primary hepatocytes from African American individuals. MR analysis provided evidence for causality for cg00036588 and cg16759041 in type 2 diabetes using blood and hepatocyte meQTLs, respectively. Longitudinal analyses showed an association between baseline methylation of these CpGs with HbA_1c at follow-up. RNA-seq data revealed a cis correlation of cg00036588 with FAM83C (false discovery rate [FDR]=3.3 × 10^–4) and EIF6 (FDR=0.13) in skeletal muscle. Conclusions/interpretation: Our study identified two epigenetic markers as likely to be causal for type 2 diabetes in West African populations. In addition to enhancing our understanding of disease mechanisms, these CpGs with evidence of causal associations could be prioritised as potential biomarkers for early detection of disease or as drug development targets.