Comparative analysis of the Illumina Mouse Methylation BeadChip and Reduced Representation Bisulphite Sequencing for routine DNA methylation analysis of murine samples

Abstract

BackgroundResearching the murine epigenome in disease models has been hampered by the lack of an appropriate and cost-effective DNA methylation array. Until recently, investigators have been limited to the relatively expensive and analysis intensive bisulphite sequencing methods. Here, we performed a comprehensive, comparative analysis between the new Mouse Methylation BeadChip (MMB) and reduced representation bisulphite sequencing (RRBS) in two murine models of colorectal carcinogenesis, providing insight into the utility to each platforms in a real world environment. ResultsWe captured 1.47x106 CpGs by RRBS and 2.64x105 CpGs by MMB, mapping to 13,778 and 13,365 CpG islands, respectively. RRBS captured significantly more CpGs per island (median 41 for RRBS versus 2 for MMB). We found that 64.4% of intra-island CpG methylation variability can be captured by measuring approximately one quarter of CpG island (CGI) CpGs. MMB was more precise in measuring DNA methylation, especially at sites that had low RRBS coverage. This impacted differential methylation analysis, with more statistically significantly differentially methylated CpG sites identified by MMB in all experimental conditions, however the difference was minute when appropriate thresholding for the magnitude of methylation change (0.2 beta value difference) was applied, providing confidence that both techniques can identify similar differential DNA methylation. Gene ontology enrichment analysis of differentially hypermethylated gene promoters identified similar biological processes and pathways by both RRBS and MMB across two murine model systems. ConclusionMMB is an effective tool for profiling the murine methylome that performs comparably to RRBS, identifying similar differentially methylated pathways. Although MMB captures a similar proportion of CpG islands, it does so with fewer CpGs per island. We show that subsampling informative CpGs from CpG islands is an appropriate strategy to capture whole island variation. Choice of technology is experiment dependent and will be predicated on the underlying biology being probed.