Whole-Genome Bisulfite Sequencing on Illumina

All sequencing runs were single-ended and 100 nucleotides (nt) in length, and performed on the Illumina HiSeq 2000 platform. Based on the qPCR quantification, 3×10⁸ copies of double-stranded DNA from the PBAT library were sequenced per lane on HiSeq 2000, as previously described [15] (link). Cluster generation and sequencing were performed in single-read mode using the TruSeq SR Cluster Kit v3-cBot-HS (Illumina) and the TruSeq SBS Kit v3-HS (Illumina) according to the manufacturer's protocols. Sequenced reads were processed using the standard Illumina base caller (v.1.8.2). We truncated raw reads to 92 nt to remove lower quality bases near the end of the reads and any remaining adapter sequences incorporated in the read. The resulting reads were aligned to the reference genome (mouse mm9) using Bismark alignment software v.0.6.3 [49] (link) with a maximum of two mismatches, and only uniquely aligned reads were retained. We estimated bisulfite conversion rates using reads that uniquely aligned to the lambda phage genome. For strand-independent analysis of CG methylation, counts from the two cytosines in a CG and its reverse complement were combined. We subsequently evaluated only CG sites with at least 6× coverage and non-CG sites with at least 4× coverage, and discarded cytosines with more than 100× coverage. CG islands, RefSeq genes, and repeat sequences for the mm9 genome were downloaded from the UCSC Genome Browser [50] (link). The logo plot images were generated with WebLogo [51] (link).