DOP010. Epigenome-wide analysis in childhood-onset Crohn's disease implicates MIR21 in pathogenesis and identifies multiple methylation-based diagnostic biomarkers
N.A. Kennedy1,2, A.T. Adams1, R. Hansen3, N.T. Ventham1, K.R. O'Leary1, H. Drummond1, C.L. Noble1,2, E. El-Omar3, R.K. Russell4, D.C. Wilson5, E.R. Nimmo1, G.L. Hold3, J. Satsangi1,2, 1University of Edinburgh, Gastrointestinal Unit, Centre for Genomic and Experimental Medicine, Edinburgh, United Kingdom, 2Western General Hospital, Gastrointestinal Unit, Edinburgh, United Kingdom, 3University of Aberdeen, Gastrointestinal Research Group, Division of Applied Medicine, Aberdeen, United Kingdom, 4Royal Hospital for Sick Children, Paediatric Gastroenterology, Glasgow, United Kingdom, 5University of Edinburgh, Paediatric Gastroenterology and Nutrition, Child Life and Health, Edinburgh, United Kingdom
DNA methylation influences transcriptional activity and marks sites of active transcription. Technological developments allow the rapid assessment of methylation state at 450,000 sites across the genome. We have performed a three-stage retrospective case–control study with epigenome-wide analysis of leucocyte DNA methylation in paediatric discovery and replication cohorts and targeted replication in adults.
Using the Illumina 450k platform we performed epigenome-wide analysis in symptomatic children presenting for diagnostic colonoscopy, half of whom were diagnosed with Crohn's Disease (CD) and half never developed any pathology. Replication was performed in children with established CD vs. symptomatic non-disease controls. Further targeted replication by pyrosequencing was performed in adults with CD vs. healthy controls.
Meta-analysis of the combined paediatric datasets (n = 66) with Bonferroni correction for multiple testing identified 165 CpGs with epigenome-wide significance. 138 differentially methylated regions (DMR) were identified - defined as stretches of probes where all had false discovery rate corrected p values <0.05 and a shared direction of change in disease. Thirdly, we demonstrate an enrichment (p < 0.0001) of significant methylation changes in proximity to loci implicated by genome-wide association studies (GWAS). The strongest result by each approach was MicroRNA 21, within the autophagy gene VMP1 (p = 1.2×10−14), 48kb from GWAS SNP rs1292053 - also the strongest result in the initial discovery/replication analysis.
We have replicated methylation changes at this region in adults (p = 6.6×10−5, n = 172), and shown altered expression in blood (p < 0.005, n = 66) and in inflamed vs. non-inflamed intestinal biopsies in CD (p = 1.4×10−6, n = 99) but not controls (n = 73). Alternative lines of evidence support further study of MIR21 in IBD - e.g. dysregulation in dysplasia and colorectal cancer in IBD, an established role in T-cell differentiation, and protection from DSS-induced fatal colitis by MIR21 knockout.
Of immediate clinical interest, linear discriminant analysis of methylation at combinations of two probes consistently predicted disease state with high accuracy (94% sensitivity, 100% specificity) in symptomatic children presenting for diagnostic colonoscopy. This bypasses the complications of establishing causation and the impact of shifting subpopulations of cells on measured methylation.
These data demonstrate a novel approach for identifying biological variations associated with germ-line variants identified by GWAS and demonstrate translational potential for biomarker development and therapeutic target discovery.