Figure 1.
5-hmC- and 5-mC-specific immunostaining of metaphase chromosomes from human ES cells.
Metaphase chromosomes from human embryonic stem cells were immunostained with antibodies specific to 5-hmC and 5-mC. A–D, Both 5-hmC and 5-mC could be observed along the chromosome arms. Strong 5-mC signal (yellow open arrow) but distinct exclusion of 5-hmC (white open arrow) from the heterochromatin of pericentromeres and the Y chromosome was observed. 5-hmC is enriched in some regions of multiple chromosomes (solid arrow). E–H, Representative images showing 5-hmC is strongly depleted from 5-mC-enriched pericentromeric region of chromosome 1 as indicated by arrow.
Figure 2.
Summary of genome-wide distribution of 5-hmC in human H1 ES cells.
A. Chromosome-wide distributions of 5-hmC-enriched and -unenriched input genomic DNA reads (reads/million), compared to the distribution expected by chance if reads were randomly distributed amongst chromosomes (106/hg18 length X chromosome length, with expected values divided by 2 for chromosome X). B. Association between 82,221 5-hmC enriched regions (p-value threshold of 1e-8) and annotated genomic features (obtained through UCSC Tables, NCBI36/hg18 and [15]). Values are represented as the fold change in percentage of peaks overlapping a defined feature over the percent expected by chance give the genomic base coverage of that feature. Shown to the right is the fraction of total peaks corresponding to each genomic feature for reference. C. Observed-to-Expected (O/E) ratios of all possible dinucleotides in 5-hmC-enriched regions (n = 82,221, p-value threshold of 1e-8), CpG Islands (NCBI36/hg18, n = 28,226), and randomly selected genomic regions (n>100,000). D. The cumulative fraction of genomic regions with a given GC content plotted for 5-hmC-enriched regions (n = 82,221, p-value threshold of 1e-8), CpG Islands (NCBI36/hg18, n = 28,226), and randomly selected genomic regions (n>100,000).
Figure 3.
Genome-wide correlation among 5-hmC, 5-mC, and 11 distinct histone modifications in human H1 ES cells.
(A–L) MeDIP and ChIP-Seq data for 11 diverse histone modifications and input from the same experiments were obtained from the previous publication and GSM456941 [15], binned at 10 kb, and normalized to the total number of aligned reads in millions. MeDIP and histone-specific values were then input-normalized (IP-Input). Data from 5-hmC-enriched and -unenriched input DNA were binned and normalized identically. Input-normalized 5-hmC signals were then plotted against histone-input normalized values to obtain correlations (r2).
Figure 4.
Gene expression level-dependent 5-hmC and H3K36me3 distributions in human ES cells.
A. Gene expression values (RPKM) obtained from [16] were ranked in descending order, and 5-hmC and H3K36me3 read densities were measured in 100-bp bins ±5 kb of the TSS or TES. Values were normalized to the total number of aligned reads in millions. The scale of 5-hmC and H3K36me3 signals are indicated at the below each respective heatmap. B. 5-hmC and input read densities at the top 25% of genes based on RPKM expression level. C. 5-hmC and input read densities among genes expressed within the 25–50% range of all genes based on RPKM expression level. D. 5-hmC and input read densities among genes expressed within the 50–75% range of all genes based on RPKM expression level. E. 5-hmC and input read densities at the bottom 25% of genes based on RPKM expression level. For B–E, reads were summed in 10-bp windows 2.5 kb upstream and downstream of TSS and TES.
Figure 5.
5-hmC marks distinct subtypes of promoters in human H1 ES cells.
A. Distribution of 5-hmC at P1-type, bimodal promoters. 5-hmC reads were summed in 100-bp windows and the immediate 4-kb region, as well as 4 kb upstream and downstream of the immediate 4-kb region, centered on each H1 hES cell promoter type defined on the basis of their chromatin signature [15], [27]. Counts were normalized to the total number of aligned reads in millions and input reads counted and normalized in the same manner were subtracted. B. Heatmap representations of 5-hmC and 11 histone modifications at P1-type bimodal promoters. Heatmap scale is indicated below, and maximum values per mark are indicated to the right of each representative heatmap. C. Genomic view of the DNMT3A locus with read distributions for 5-hmC and 11 histone modifications exemplifying the bimodal distributions observed. Highlighted in red are the promoter regions. D. Distribution of 5-hmC at P2-type promoters. 5-hmC reads were summed in 100-bp windows in the immediate 4-kb region, as well as 4 kb upstream and downstream of the immediate 4-kb region, centered on each H1 hES cell promoter type. Counts were normalized to the total number of aligned reads in millions and input reads counted and normalized in the same manner were subtracted. E. Heatmap representations of 5-hmC and 11 histone modifications at P2-type bimodal promoters. Heatmap scale is indicated below, and maximum values per mark are indicated to the right of each representative heatmap. F. Genomic view of the CENPB locus with read distributions for 5-hmC and 11 histone modifications exemplifying P2-type promoters. Highlighted in red is the promoter region.
Figure 6.
Enrichment of 5-hmC at specific enhancers in human H1 ES cells.
A. 5-hmC reads were summed in 100-bp windows in the immediate 4-kb region, as well as 4 kb upstream and downstream of the immediate 4-kb region, centered on each of 12 predicted H1 hES cell enhancers defined on the basis of their chromatin signature [15], [27], [28]. Counts were normalized to the total number of aligned reads in millions and input reads counted and normalized in the same manner were subtracted [15], [27], [28]. B. Heatmap representations of read distributions for 5-hmC and 11 histone modifications at 5 predicted enhancer subtypes found to have significant enrichment in H1 hES cells. Heatmap scale is constant for all marks and indicated in the lower left-hand corner. C. Genomic view of an E8-type enhancer overlapping a region identified as significantly enriched for 5-hmC by peak identification upstream of the PRDM14 gene.
Figure 7.
5-hmC at pluripotency-associated core transcription factor binding sites in human H1 ES cells.
A. 5-hmC read distributions at KLF4 (n = 3,795), OCT4 (n = 3,890), SOX2 (n = 5,684), NANOG (n = 25,076), p300 (n = 3,094), and TAF1 (n = 12,363) binding sites in H1 hES cells. 5-hmC reads were counted in 40 equally portions within, upstream, and downstream of each binding site and values were normalized to the total number of aligned reads. Input reads counted and normalized in the same manner were then subtracted. Shown to the right of each plot is a scaled Venn diagram of the number of transcription factor binding sites (in gray) directly overlapping 5-hmC enriched regions (in black, 82,221) for reference. Overlaps are defined as ≥1 bp. B. Genomic view of 5-hmC coverage and 11 histone modifications at the DNMT3B locus, showing overlap of a 5-hmC peak, NANOG site, and the associated histone modifications. C. Scaled Venn diagram depicting the overlap between 5-hmC enriched regions (black, 82,221), p300 binding sites (gray, 3,094), and predicted enhancers reported in [15] (blue, 58,023). Shown in red is the overlap of all three (25 regions). Overlaps are defined as ≥1 bp.