Evaluation of Methyl-Binding Domain Based Enrichment Approaches Revisited

Abstract

Methyl-binding domain (MBD) enrichment followed by deep sequencing (MBD-seq), is a robust and cost efficient approach for methylome-wide association studies (MWAS). MBD-seq has been demonstrated to be capable of identifying differentially methylated regions, detecting previously reported robust associations and producing findings that replicate with other technologies such as targeted pyrosequencing of bisulfite converted DNA. There are several kits commercially available that can be used for MBD enrichment. Our previous work has involved MethylMiner (Life Technologies, Foster City, CA, USA) that we chose after careful investigation of its properties. However, in a recent evaluation of five commercially available MBD-enrichment kits the performance of the MethylMiner was deemed poor. Given our positive experience with MethylMiner, we were surprised by this report. In an attempt to reproduce these findings we here have performed a direct comparison of MethylMiner with MethylCap (Diagenode Inc, Denville, NJ, USA), the best performing kit in that study. We find that both MethylMiner and MethylCap are two well performing MBD-enrichment kits. However, MethylMiner shows somewhat better enrichment efficiency and lower levels of background "noise". In addition, for the purpose of MWAS where we want to investigate the majority of CpGs, we find MethylMiner to be superior as it allows tailoring the enrichment to the regions where most CpGs are located. Using targeted bisulfite sequencing we confirmed that sites where methylation was detected by either MethylMiner or by MethylCap indeed were methylated.

Fig 1. Methylation signals and levels of background noise.

The mean coverage of CpGs and non-CpGs from MethylMiner and MethylCap, respectively are shown. The horizontal lines indicate the threshold for background noise as determined by the 95^th percentile of the estimated coverage of the non-CpGs for MethylMiner and MethylCap, respectively.

Fig 2. The overlap of MethylMiner and MethylCap with local CpG density.

The local CpG density is plotted against the percentage CpG coverage. The distribution of the local CpG density in the reference sequence is the highest outside of CpG-rich regions. Better coverage of the regions where the majority of CpGs occur (in the range of 1–7 CpGs) is obtained with MethylMiner than with MethylCap.

Fig 3. The effect of altered salt concentrations for elutions in MethylMiner.

The shown data was generated as part of our recent methylome-wide investigations in humans[5]. The local CpG density is plotted against the percentage CpG coverage for two randomly selected individuals in the upper and lower graph, respectively. As expected, the unenriched genomic DNA closely follows the distribution of the CpG density in the reference genome. Best coverage of the regions where the majority of CpGs occur (in the range of 1–7 CpGs) is obtained by a single MethylMiner elution using a low salt (0.5M NaCl) buffer. Using a single elution with a higher salt (2.0M NaCl) concentration moves the distribution to regions with higher CpG density. If performing a double elution where the low salt elution is discarded and only the 2^nd fraction eluted with the high salt buffer is investigated, the distribution is shifted even further to the right excluding regions with low local CpG density and focusing mainly in regions with high CpG density.

Fig 4. The effect of altered salt concentrations for elutions in MethylCap.

The local CpG density is plotted against and the percentage CpG coverage. The distribution of the local CpG density in the reference genome is shown. The distribution of CpG coverage for MethylCap is shown when using a low salt elution buffer and when using a high salt elution buffer.