Fig. 1

Levels of HMR specificity recapitulate developmental relationships through accumulation and maintenance. A Multiple alignment of WGBS methylation and HMR tracks across 10 cell types: H1 ESC, fetal spinal cord, fetal heart, adrenal gland, liver, hematopoietic stem and progenitor cells, neutrophil, macrophage, B cell, and T cell. Methylation tracks are represented by orange vertical bars showing methylation value per CpG site. Methylation fraction is calculated as the fraction of reads containing a methyl-C over the total number reads covering a CpG site. HMR calls are shown by dark blue horizontal bars. Developmentally constitutive, lineage-specific, and cell specific HMRs are highlighted by blue and green dotted bars, respectively. The plotgardener R package was used to generate the genome browser snapshot [29]. B Heatmap of average methylation per HMR across cell types. Non-coding HMRs were k-means clustered based on their average CpG methylation values across 10 cell types represented in (A). A k-means of 10, assessed by the elbow method, was used to cluster HMRs into groups that are consistent with the biological relationships of their cell types. Groups are manually labeled to reflect their biological relationships. C The transcription factor (TF) motif enrichment of each k-means group reflects biological relationships captured in (B). Representative TFs were selected from the top significant hits ranked by natural log adjusted p-value for each k-means group. The top ranked TFs are shown unless the top TF(s) for that group were redundant; the second top ranked TF is shown for the group, “Myeloid + HSPC,” and the third ranked TF is shown for the group, “Differentiated.” Fold enrichment values are normalized from 0 to 1 across TFs. The background comparison file comprises HMRs across all represented cell types. D Bar graph of the total number HMRs for each cell type, arranged by developmental progression. E Bar graph measuring the presence of HMRs established in either H1 ESCs (top; green) or HSPCs (bottom; grey) in developmentally progressive cell types. The software Bedtools intersect was used to determine overlap between cell type HMR datasets using default settings [30]. Overlap was defined as a 1 bp minimum