Methyl-CpG/MBD2 Interaction Requires Minimum Separation and Exhibits Minimal Sequence Specificity

Abstract

Determining the pattern of methylation at CpG dinucleotides in a cell remains an essential component of epigenetic profiling. The correlations among methylation, gene expression, and accompanying disease have just begun to be explored. Many experiments for sensing methylation use a relatively inexpensive, high-throughput approach with a methyl-binding domain (MBD) protein that preferentially binds to methylated CpGs. Here, we characterize the cooperativity and sequence specificity of MBD2-DNA binding in a pulldown experiment revealing three potential biases in such experiments. The first is caused by steric clashes between two MBD2 proteins at mCpGs separated by 2 bp or less, which suggests that simultaneous binding at these sites is inhibited. This is confirmed by comparing input versus pulldown high-throughput sequencing data on M.SssI-treated samples, from which we also find that pulldown efficiency sharply increases for DNA fragments with four or more mCpGs. Analysis of these two data sets was again employed to investigate MBD2's sequence preferences surrounding a methylated CpG (mCpG). In comparing the distributions of bases at positions with respect to an mCpG, statistically significant preferences for certain bases were found, although the corresponding biases in pulldown efficiency were all <5%. While this suggests that mCpG sequence context can mostly be ignored in MBD2 binding, the statistical certainty exhibited by our highthroughput approach bodes well for future applications.

A model of pulldown alignments from SssI-treated DNA improves DNA methylation prediction

“…Three ingredients are used to calculate the expected pulldown of a particular fragment of length ℓ to a location x : (i) The number of accessible mCpGs on the fragment, (ii) the relative enrichment of that fragment due to this number of mCpGs, and (iii) the probability of a fragment of length ℓ being sequenced in the pulldown library. Ingredient (i) depends on the minimum separation of consecutive mCpGs in order for the two to be bound by two separate MBD2 domains, set here to be 3 bp [13]. Ingredient (ii) depends on the pulldown efficiencies as a function of accessible CpGs.…”

“…Ingredient (ii) depends on the pulldown efficiencies as a function of accessible CpGs. These pulldown efficiencies were calculated in [13] from the same MethylMiner kit as used here, and we thus use the pulldown efficiencies E ( n mCpGs) for n well-separated CpGs from [13]. Then, the coefficients for (ii) are derived as C n = E ( n mCpGs)/ E (0 mCpGs) (Table 1).…”

“…Since pulldown reads are sequenced and aligned without knowing which of the CpGs on the DNA fragment were methylated, MBD-seq data are often processed around the resolution of the DNA fragment length, typically in 100-500 bp windows. The interpretation of MBD pulldown reads is also affected by the density and arrangement of mCpGs on the fragment, which is known to influence the efficiency of capture by MBD pulldown [11–13]. Thus, statistical approaches must be used to quantify methylation levels from MBD pulldown alignments and to increase its resolution to make it competitive with bisulfite sequencing techniques.…”

“…Given our previous characterizations of methylated DNA and MBD2 interactions [13], we built a model of MBD pulldown alignments from SssI-treated DNA that we tested the efficacy of through substitution for the SssI control data set utilized by the BayMeth model. Our model incorporates the fragment length distribution in the MBD pulldown library, the minimum separation between neighboring mCpGs needed for optimal pulldown efficiency, and the relative representation of DNA fragments with n mCpGs to those with 0 mCpGs, and generates an expected MBD pulldown for every site in the human genome from SssI-treated DNA (Fig.…”

A methylation study of long-term depression risk

“…To assay the methylome at baseline, we used an optimized protocol for methyl-CG binding domain sequencing (MBD-seq) [11][12][13][14] . The optimizations involved the choice of the the MBD protein 15 and adaptations of the enrichment and sequencing protocol 16; 17 .…”