Genome-wide assays that identify and quantify modified cytosines in human disease studies

Abstract

The number of different assays that has been published to study DNA methylation is extensive, complemented by recently described assays that test modifications of cytosine other than the most abundant 5-methylcytosine (5mC) variant. In this review, we describe the considerations involved in choosing how to study 5mC throughout the genome, with an emphasis on the common application of testing for epigenetic dysregulation in human disease. While microarray studies of 5mC continue to be commonly used, these lack the additional qualitative information from sequencing-based approaches that is increasingly recognized to be valuable. When we test the representation of functional elements in the human genome by several current assay types, we find that no survey approach interrogates anything more than a small minority of the nonpromoter cis-regulatory sites where DNA methylation variability is now appreciated to influence gene expression and to be associated with human disease. However, whole-genome bisulphite sequencing (WGBS) adds a substantial representation of loci at which DNA methylation changes are unlikely to be occurring with transcriptional consequences. Our assessment is that the most effective approach to DNA methylation studies in human diseases is to use targeted bisulphite sequencing of the cis-regulatory loci in a cell type of interest, using a capture-based or comparable system, and that no single design of a survey approach will be suitable for all cell types.

Keywords: 5-methylcytosine; Assay; CpG island; DNA methylation; Enhancer; Epigenomic; microarray.

Figure 1

Cytosine variants and their production. We show how cytosine within DNA can be acted upon by DNA methyltransferases (DNMT) to generate 5-methylcytosine (5mC), which can subsequently be oxidized by TET enzymes through the 5-hydroxymethylation (5hmC), 5-formylcytosine (5fC) and 5-carboxylcytosine (5caC) variants, favoring the activity of thymine DNA glycosylase to create an abasic site that can then be repaired to add back an unmethylated cytosine to complement the guanine on the other strand. As well as alpha-ketoglutarate (α-KG), another known TET enzyme cofactor is ascorbic acid (vitamin C).

Figure 2

The effects of bisulphite conversion on different cytosine variants. Bisulphite conversion deaminates not only unmodified cytosine but also 5fC and 5caC [27] to uracil, which is then amplified and sequenced as a thymine. Both 5mC and 5hmC are resistant to bisulphite conversion and are sequenced as cytosines. While the output of bisulphite sequencing has been generally described to be the ratio of 5mC/(5mC + C), more correctly it should be described as (5mC + 5hmC)/(5mC + 5hmC + 5fC + 5caC + C), which only approximates 5mC/(5mC + C) when the other cytosine variants are in very low proportions.

Figure 3

The proportional representation of functional genomic elements by different DNA methylation assays. Using annotations of human embryonic stem cells by Segway (a) and ChromHMM (b), we calculated how four types of genome-wide DNA methylation assays represent each type of functional element. When the proportion of functional genomic elements tested by each assay is calculated, all of the assays work best to represent the candidate promoters annotated by Segway (a) or ChromHMM (b), but generally only represent a minority of candidate cis-regulatory elements. Even WGBS does not represent 100% of each functional genomic element, so no assay achieves complete comprehensiveness, and the penalty of the WGBS approach is that it sequences equally deeply a number of genomic contexts that are not as likely to be informative, an inefficient use of costly sequencing resources.

Figure 4

Assays for cytosine modifications and a targeted sequencing strategy. We show regular bisulphite sequencing, which reports both 5mC and 5hmC, and two assays detecting 5hmC alone, as examples of the broader group of assays detecting cytosine variants. Any system that captures bisulphite converted DNA can be used downstream of these sample preparation approaches to target deep sequencing to loci of interest, here indicated by cis-regulatory elements (a). In (b) we show the steps and the nucleotide conversions involved in bisulphite sequencing, TAB-seq and oxBS-seq.