minimizing the artefacts and improving the significance and resolution of results.
EWAS are designed to identify a specific epigenetic mark [44] associated with a
particular phenotype through the use of array and sequencing-based technologies.
One of the most studied epigenetic marks is DNA methylation of 5-methylcytosine
(5mC) on CpGislands [45] since methylated DNA is the key component of epigenetic
information in mammals [46]. A first step in studying DNA methylation is to quantify
all global levels of DNA methylation. Massive parallel sequencing (MPS) [47]
quantifies results after DNA pretreatment. Pretreatment may be affinity-based,
which enriches methylated or unmethylated sites of DNA. Alternatively, pretreatment
may be performed using methylation-sensitive restriction enzymes or bisulfite conversion of DNA. In the latter case, the gold standard approach is whole-genome
bisulfite sequencing [48], which provides quantitative information about cytosines
throughout the genome. However, this is a very expensive assay.
MPS approaches can be divided into two different groups: (1) reduced representation bisulfite sequencing (RRBS), which uses a measuring range of generated
fragments from digestion with restriction enzymes for targeting deep sequencing at
specific loci, and (2) HELP-tagging assays using digestion with methylationsensitive restriction enzymes in order to generate tags at these sites, which represent
the DNA methylation level. MPS provides data that cannot be obtained by microarray such as single-nucleotide polymorphisms. As microarrays are designed to
recognize 5mC, in the context of CG, in some cases, levels of CHG and CHH
(where H equals A, T or C) may not be detected or might interfere with restriction
enzyme digestion. In bisulfite sequencing (BS-seq), cytosine that is not converted by
bisulfite can be both 5mC and 5-hydroxymethylcytosine (5hmC). In order to overcome this issue, two different arrays are therefore necessary. Four different modifications including 5hmC, 5mC, 5-formylcytosine (5fC) and 5-carboxylcytosine
(5caC) are read by bisulfite as unmethylated cytosine, which could require more
specific assays [49], described in the following subsections. Although these assays
have been improved to be specific for each type of modification, considering the
huge amount of (epi)-genetic modifications underpinning different areas of the
genome, they inevitably present both advantages and disadvantages (Table 4).
2.1.4 Tet-Assisted Bisulfite Sequencing (TAB-Seq)
TAB-Seq is a novel method using bisulfite conversion and Tet proteins to study
5hmC [50]. Specifically, genomic DNA is first treated consecutively with T4 phage
β-glucosyltransferase (β-GT) and recombinant mouse Tet1 to convert 5hmC to
5gmC and 5mC to 5caC. β-GT catalyses the attachment of β-D-glucosyl residues
from uridine diphosphoglucose to the hydroxyl group of 5hmC. DNA bisulfite
treatment leaves 5hmC unchanged, while 5mC and unmethylated cytosines are
converted to uracil. This assay provides an accurate representation of 5hmC localization in the genome and clearly differentiates between 5hmC and 5mC, specifically
540
C. Dell’Aversana et al.
EWAS are designed to identify a specific epigenetic mark [44] associated with a
particular phenotype through the use of array and sequencing-based technologies.
One of the most studied epigenetic marks is DNA methylation of 5-methylcytosine
(5mC) on CpGislands [45] since methylated DNA is the key component of epigenetic
information in mammals [46]. A first step in studying DNA methylation is to quantify
all global levels of DNA methylation. Massive parallel sequencing (MPS) [47]
quantifies results after DNA pretreatment. Pretreatment may be affinity-based,
which enriches methylated or unmethylated sites of DNA. Alternatively, pretreatment
may be performed using methylation-sensitive restriction enzymes or bisulfite conversion of DNA. In the latter case, the gold standard approach is whole-genome
bisulfite sequencing [48], which provides quantitative information about cytosines
throughout the genome. However, this is a very expensive assay.
MPS approaches can be divided into two different groups: (1) reduced representation bisulfite sequencing (RRBS), which uses a measuring range of generated
fragments from digestion with restriction enzymes for targeting deep sequencing at
specific loci, and (2) HELP-tagging assays using digestion with methylationsensitive restriction enzymes in order to generate tags at these sites, which represent
the DNA methylation level. MPS provides data that cannot be obtained by microarray such as single-nucleotide polymorphisms. As microarrays are designed to
recognize 5mC, in the context of CG, in some cases, levels of CHG and CHH
(where H equals A, T or C) may not be detected or might interfere with restriction
enzyme digestion. In bisulfite sequencing (BS-seq), cytosine that is not converted by
bisulfite can be both 5mC and 5-hydroxymethylcytosine (5hmC). In order to overcome this issue, two different arrays are therefore necessary. Four different modifications including 5hmC, 5mC, 5-formylcytosine (5fC) and 5-carboxylcytosine
(5caC) are read by bisulfite as unmethylated cytosine, which could require more
specific assays [49], described in the following subsections. Although these assays
have been improved to be specific for each type of modification, considering the
huge amount of (epi)-genetic modifications underpinning different areas of the
genome, they inevitably present both advantages and disadvantages (Table 4).
2.1.4 Tet-Assisted Bisulfite Sequencing (TAB-Seq)
TAB-Seq is a novel method using bisulfite conversion and Tet proteins to study
5hmC [50]. Specifically, genomic DNA is first treated consecutively with T4 phage
β-glucosyltransferase (β-GT) and recombinant mouse Tet1 to convert 5hmC to
5gmC and 5mC to 5caC. β-GT catalyses the attachment of β-D-glucosyl residues
from uridine diphosphoglucose to the hydroxyl group of 5hmC. DNA bisulfite
treatment leaves 5hmC unchanged, while 5mC and unmethylated cytosines are
converted to uracil. This assay provides an accurate representation of 5hmC localization in the genome and clearly differentiates between 5hmC and 5mC, specifically
540
C. Dell’Aversana et al.