American Society of Hematology

Figure 2.

$Figure 2. PMDs contain mostly silent genes, whereas HMDs contain a mixture of expressed and silent genes. (A) GENCODE gene models were divided into 4 expression quartiles based on RNA-seq and CAGE data, and the number of genes within HMDs, s-PMDs, and l-PMDs were plotted for each cell type. Genes overlapping both HMDs and PMDs were excluded from the analysis. Q1 represents unexpressed genes, whereas Q2 through Q4 represent tertiles of increasing expression. Almost all active genes are in the HMD compartment, except for transformed cell lines where most of the genome is in PMDs. (B) Bar graph illustrating the fraction of all active and silent promoters that are unmethylated, separated by CpG density of their promoters into CpG-rich (top) and CpG-poor (bottom). y-axis represents the fraction of promoter that are expressed or silent. Unmethylated promoters were defined as promoters overlapping either an UMR or LMR. Silent promoters were defined as promoters associated with genes exhibiting no RNA-seq (TPM < 1) and no CAGE signal; expressed promoters as promoters associated with genes exhibiting both an RNA-seq (TPM > 1) and a CAGE signal (see “Methods”). In both HMDs and PMDs, almost all active and a significant fraction of silent promoters are unmethylated. (C) ChromHMM analysis based on the 15-state core model from the Roadmap Epigenome Project. Top: bar graphs illustrating the average fractional chromatin composition within HMDs, s-PMDs, and l-PMDs. For each cell line, the percentage of the genome in each of the 3 methylation domains that is covered by the 5 chromatin states was calculated. Results were then averaged per each cell category (SPC, differentiated or transformed). Error bars indicate the standard error of the mean for the respective groups of cell types analyzed. For clarity, the 15 ChromHMM classes were summarized into 5 larger categories: active (red: TssA, TssAFlnk, TxFlnk, Tx, TxWk, EnhG, Enh, ZNF.Rpts), bivalent (green: TssBiv, BivFlnk, EnhBiv), repressed (blue: RepPC, ReprPCWk), heterochromatin (purple), and quiescent (gray). The y-axis represents the percentage of HMDs or PMDs that overlap with a chromatin states as defined previously. HMDs of differentiated cells contain most of the active chromatin, whereas PMDs are composed mostly of repressed and quiescent chromatin. Bottom, graphs are organized as previously, but the y-axis represents the enrichment, calculated as (# of bases in HMDs, s-PMDs, or l-PMDs covered with a particular ChromHMM state/# of bases in HMDs, s-PMDs or l-PMDs)/(# of bases in the genome covered with a particular ChromHMM state/# number of bases in the genome). The bar represents the average log2 of the enrichment (± standard deviation) for each cell category and each ChromHMM category. Some of the error bars are very large because the fraction of the genome covered by some of the cell categories was very low and therefore highly variable in different cell types.$

PMDs contain mostly silent genes, whereas HMDs contain a mixture of expressed and silent genes. (A) GENCODE gene models were divided into 4 expression quartiles based on RNA-seq and CAGE data, and the number of genes within HMDs, s-PMDs, and l-PMDs were plotted for each cell type. Genes overlapping both HMDs and PMDs were excluded from the analysis. Q1 represents unexpressed genes, whereas Q2 through Q4 represent tertiles of increasing expression. Almost all active genes are in the HMD compartment, except for transformed cell lines where most of the genome is in PMDs. (B) Bar graph illustrating the fraction of all active and silent promoters that are unmethylated, separated by CpG density of their promoters into CpG-rich (top) and CpG-poor (bottom). y-axis represents the fraction of promoter that are expressed or silent. Unmethylated promoters were defined as promoters overlapping either an UMR or LMR. Silent promoters were defined as promoters associated with genes exhibiting no RNA-seq (TPM < 1) and no CAGE signal; expressed promoters as promoters associated with genes exhibiting both an RNA-seq (TPM > 1) and a CAGE signal (see “Methods”). In both HMDs and PMDs, almost all active and a significant fraction of silent promoters are unmethylated. (C) ChromHMM analysis based on the 15-state core model from the Roadmap Epigenome Project. Top: bar graphs illustrating the average fractional chromatin composition within HMDs, s-PMDs, and l-PMDs. For each cell line, the percentage of the genome in each of the 3 methylation domains that is covered by the 5 chromatin states was calculated. Results were then averaged per each cell category (SPC, differentiated or transformed). Error bars indicate the standard error of the mean for the respective groups of cell types analyzed. For clarity, the 15 ChromHMM classes were summarized into 5 larger categories: active (red: TssA, TssAFlnk, TxFlnk, Tx, TxWk, EnhG, Enh, ZNF.Rpts), bivalent (green: TssBiv, BivFlnk, EnhBiv), repressed (blue: RepPC, ReprPCWk), heterochromatin (purple), and quiescent (gray). The y-axis represents the percentage of HMDs or PMDs that overlap with a chromatin states as defined previously. HMDs of differentiated cells contain most of the active chromatin, whereas PMDs are composed mostly of repressed and quiescent chromatin. Bottom, graphs are organized as previously, but the y-axis represents the enrichment, calculated as (# of bases in HMDs, s-PMDs, or l-PMDs covered with a particular ChromHMM state/# of bases in HMDs, s-PMDs or l-PMDs)/(# of bases in the genome covered with a particular ChromHMM state/# number of bases in the genome). The bar represents the average log2 of the enrichment (± standard deviation) for each cell category and each ChromHMM category. Some of the error bars are very large because the fraction of the genome covered by some of the cell categories was very low and therefore highly variable in different cell types.

This Feature Is Available To Subscribers Only