Fig. 4. Contrasting CpG methylation dynamics at PU and DM subregions in individual ESCs.

a CpG methylation levels (mCpG/CpG) at PU, DM and INT SE subregions in 16 individual ESCs grown in 2i/LIF (2i, white) and 64 ESCs grown in serum/LIF³⁰ (serum, black). Each row represents one single cell, with blue–red colour gradient indicating proportion of methylated CpGs in the corresponding SE subregion class. Using hierarchical clustering, three cell clusters are identified: 2i (light blue), “naive-like” (blue) and “primed-like” (orange) serum-ESCs. b Violin plots representing the distribution of the average methylation (%) across SE subregions for five randomly selected single cells in each of the three single-cell clusters defined in a. c Profiles of CpG methylation levels throughout the Med13l- (class I) and Esrrb-associated (class II) SEs with indication of PU (light green), DM (pink) and INT (grey) subregions. Each line gives the lowest-smoothed average of each individual CpG methylation across all individual cells in the corresponding cluster: 2i cluster (light blue), “naive-like” cluster (blue) and “primed-like” cluster (orange). d Heatmaps showing the relative levels of enrichment (median log₂ fold change of ChIP-seq signal (RPKM + 1) over input signal (RPKM + 1)) for selected histone marks (left panel) and CpG methylation regulators (right panel) at PU, DM and INT subregions in bulk ESCs grown in serum/LIF. Enrichment scores for each feature are scaled by dividing by the mean enrichment score for that feature across all subregions. See Supplementary Data 1 for dataset accession numbers. e Box plots representing the distribution of methylation variances across PU (green) and DM (magenta) subregions for each individual cell analysed in different treatment conditions (2i and serum wild-type ESCs, and Dnmt3a/b DKO and Tet1-3 TKO ESCs grown in serum/LIF)^30,31. f Receiver operating characteristic (ROC) curves comparing the performance of naive (yellow), general (orange) and primed (red) pluripotency gene sets in discriminating “naive-like” and “primed-like” cell clusters based on the average of their normalised expression levels (left panel). ROC curves comparing the performance of individual naive pluripotency genes (discriminating “naive-like” and “primed-like” cell clusters based on normalised expression level), illustrating Esrrb, Klf2 and Zfp42 as best classifiers of “naive-like” and “primed-like” clusters (AUC values = 0.928, 0.963 and 0.948 respectively; p = 10⁻⁵; right panel).