DNA methylation and 28 year incidence of two neuropathy phenotypes in type 1 diabetes: the Pittsburgh Epidemiology of Diabetes Complications cohort study

Abstract

Aims/hypothesis

Diabetic peripheral neuropathy (DPN) and neuropathic pain (NP) are common complications of type 1 diabetes that can greatly affect quality of life. Studying DNA methylation (DNAm) may help identify potential therapeutic targets; however, epigenome-wide association studies (EWAS) of DPN and NP are lacking. We thus performed prospective EWAS of 28 year DPN and NP incidence in the Pittsburgh Epidemiology of Diabetes Complications (EDC) study of childhood-onset (<17 years) type 1 diabetes.

Methods

DPN was defined as two or more of the following criteria: symptoms consistent with DPN; decreased tendon reflexes; or abnormal sensory examination. NP was reported as burning, aching or stabbing pain in the feet during an EDC examination or on the Michigan Neuropathy Screening Instrument (MNSI). The time of the first available blood-derived DNA specimen collected between 1988–1998 was considered the analytic ‘baseline’ (mean age 27 years; diabetes duration 19 years). After quality control, DNAm (EPIC array) at 683,597 CpGs was analysed in Cox models for time-to-DPN in 282 individuals free of DPN at baseline and time to NP in 365 individuals free of NP at baseline. False discovery rate (FDR) <0.05 was considered statistically significant. We also identified differentially methylated regions (DMRs), functional interaction networks and genetic variants associated with DNAm (methylation quantitative trait loci [meQTLs]), and performed Mendelian randomisation (MR) to assess evidence of causality.

Results

Over 28 years, 154 individuals (54.6%) developed DPN and 148 (40.5%) developed NP. Greater methylation at three CpGs was significantly associated (FDR≤0.05) with reduced hazard of DPN: cg06163904 (CHMP6); cg10835127 (CACNA1B); and cg18945945 (PKNOX1). CpG associations with DPN remained similar after adjustment for clinical risk factors. We identified 75 meQTLs for cg18945945 in the PKNOX1 region, 59 of which were validated in an external diabetes cohort. One-sample MR provided nominal evidence for a potentially causal association between cg18945945 and DPN (p=0.01). While no individual CpGs were significantly associated with NP, there were 49 NP-associated DMRs.

Conclusions/interpretation

Our study identified associations between DNAm and 28 year incidence of DPN and NP at several biologically plausible loci. Most notably, we identified a novel association between DNAm of PKNOX1 and future DPN, including evidence of a genetic influence on PKNOX1 methylation that was validated in an external diabetes cohort. PKNOX1 has previously been implicated in drug-induced neuropathy; our results provide strong evidence that epigenetic regulation of PKNOX1 may also play a functional role in the development of diabetic neuropathy. Our results suggest that epigenetic modification of the identified loci warrants further study to inform potential targets for prevention of DPN.

Graphical Abstract