Barcoded-bisulfite-amplicon sequencing

Bisulfite-converted DNA was used for a nested PCR using the PyroMark PCR kit (Qiagen; primers are provided in Supplemental Table S5). The second PCR added barcoded Illumina adapters that allowed to distinguish donors and passages, as described before⁸. Amplicons were pooled and sequenced on an Illumina MiSeq lane with the v2 nano reagents (Illumina) in 250 PE mode. Bisulfite converted sequencing data were analyzed using TrimGalore, Bismark⁵³ and bowtie2⁵⁴. Mean sequencing coverage of amplicons was ~3900 reads per amplicon (Supplemental Data 3). Further pattern analysis and visualization was performed with custom perl and R scripts or with R package ggplot2 for area plots. Pearson correlation of neighboring CpGs and the corresponding heatmap were produced with the python packages scipy and seaborn, respectively. Single read predictions were performed as described before³¹. In short, single reads of the BBA-Seq amplicons were assigned to their most likely passage number (from 0 to 50) based on their binary sequel of methylated and unmethylated CpG sites. Probabilities of passage numbers were based on linear regression models at each CpG site, retrieved from the training dataset. Finally, we calculated the mean passage number for each sample based on all sequencing reads. Further details on the rational and derivation of the mathematical model are provided in our previous work³¹.