Epigenomic Landscape of Human Fetal Brain, Heart, and Liver

Abstract

The epigenetic regulation of spatiotemporal gene expression is crucial for human development. Here, we present whole-genome chromatin immunoprecipitation followed by high throughput DNA sequencing (ChIP-seq) analyses of a wide variety of histone markers in the brain, heart, and liver of early human embryos shortly after their formation. We identified 40,181 active enhancers, with a large portion showing tissue-specific and developmental stage-specific patterns, pointing to their roles in controlling the ordered spatiotemporal expression of the developmental genes in early human embryos. Moreover, using sequential ChIP-seq, we showed that all three organs have hundreds to thousands of bivalent domains that are marked by both H3K4me3 and H3K27me3, probably to keep the progenitor cells in these organs ready for immediate differentiation into diverse cell types during subsequent developmental processes. Our work illustrates the potentially critical roles of tissue-specific and developmental stage-specific epigenomes in regulating the spatiotemporal expression of developmental genes during early human embryonic development.

Keywords: development; epigenetics; gene regulation; genomics; high throughput screening.

FIGURE 1.

Dynamic changes of the K27ac-positive DNA segments in the human fetal and adult tissues. A, Venn diagrams showing the K27ac-positive DNA segment (2-kb tiles, potential active enhancers) dynamics across stages (left panel), and across different fetal tissues (right panel). hESCs were used as a control. B, representative examples of K27ac-positive regions exhibiting tissue-specific and developmental stage-specific patterns.

FIGURE 2.

Dynamic changes of the K27ac+K4me1-positive regions (potential active enhancers) in the human fetal and adult tissues. A, heat map displaying the H3K27ac (left panel) and H3K4me1 (middle panel) enrichment signals in the different enhancer clusters, together with the expression levels of their associated genes (right panel) in the human 12-week-old fetal and adult tissues. The color keys that change from white to dark green in the left panel, from white to dark red in the middle panel, and from blue to yellow in the right panel indicate the relative ChIP-seq signal intensities of H3K27ac and H3K4me1, and the relative expression of the enhancer-associated genes (z-scores of the FPKM value), respectively. Some of the representative genes for these enhancer clusters are listed to the right side of the panel. B, MGI mouse phenotypic ontology of the tissue-specific and developmental stage-specific enhancer clusters. The color key that changes from black to yellow indicates low to high enrichment scores, respectively.

FIGURE 3.

Identification and characterization of the super-enhancers. A, average H3K27ac ChIP-seq signal intensities of the regular enhancers (black lines) and the super-enhancers (red lines), in addition to their 10-kb up- and downstream flanking regions. B, fold-difference values of the H3K27ac ChIP-seq signals (including the total signal and the average signal intensities) between the regular enhancers and the super-enhancers in the human fetal and adult tissues. C, histograms of the numbers of enhancers per RefSeq gene in the three fetal tissues, which indicate that the majority of the genes have at least one putative enhancer, whereas some harbor more than eight enhancers. The blue fitted curves represent the probability densities of the proportion of enhancers per RefSeq gene. The boxplots of the average lengths (D) and the average expression levels of the associated genes (E) of the putative regular enhancers and the super-enhancers identified across the H3K27ac datasets are shown. The bottoms and tops of the boxes are the first and third quartiles, and the red lines inside each box are the median values (the numbers are indicated). A two-tailed Student's t test was applied to calculate the p values of the pairwise comparisons.

FIGURE 4.

Dynamic changes of the super-enhancers in human fetal and adult tissues. A, heat map displaying the H3K27ac (left panel) and H3K4me1 (middle panel) enrichment signals in the different super-enhancer clusters, together with the expression levels of their associated genes (right panel) in human 12-week-old fetal and adult tissues. The color keys that change from white to dark green in the left panel, from white to dark red in the middle panel, and from blue to yellow in the right panel indicate the relative ChIP-seq signal intensities of H3K27ac and H3K4me1, and the relative expression of the enhancer-associated genes (z-scores of the FPKM value), respectively. B, representative examples of putative super-enhancers exhibiting tissue-specific and developmental stage-specific patterns.

FIGURE 5.

Characterization of the broad H3K4me3 domain-associated genes. Venn diagrams showing that the broad H3K4me3 domain-associated genes (the numbers are indicated) generally do not overlap between different fetal tissues (A) or with the super-enhancer-associated genes (B). GO analysis of the broad H3K4me3 domain-associated genes in the three fetal tissues was shown in C–E. Ten non-overlapping GO terms are shown, with the terms with higher enrichment p values at the top and proceeding down to those with lower enrichment p values in each fetal tissue.

FIGURE 6.

Enhancer comparisons between human and mouse. A, principal component analysis of the H3K27ac signal intensities between human and mouse in three different organs. B, Pearson correlation clustering analysis of these three organs. The color key from green to red indicates the correlation from low to high. Notably, the mouse genome coordinates (mm9) are converted to the human assembly (hg19) using a lift-over tool, and only the one-to-one orthologous regions were taken into considerations.

FIGURE 7.

Representative examples of bivalent H3K4me3/H3K27me3 domains in the human 12-week-old fetal tissues. The ChIP-seq tracks showing the bivalent H3K4me3/H3K27me3 domains in the promoter regions of the representative genes that were analyzed using sequential ChIP-seq (red tracks). The enrichment of the H3K4me3 (green tracks) and H3K27me3 (blue tracks) peaks using regular ChIP-seq in the same loci across two replicates are also shown as controls. Overall, three fetal brain-specific (A), three fetal heart-specific (B), and three fetal liver-specific (C) bivalent H3K4me3/H3K27me3 domains are shown.

FIGURE 8.

Gene ontology analysis of the H3K4me3 broad domain-associated genes and characterization of the bivalent H3K4me3/H3K27me3 genes in human fetal tissues. A, MGI phenotype enrichment for the broad H3K4me3 domain-associated genes in the human fetal and adult tissues. B, Venn diagrams showing the identification and characterization of the bivalent H3K4me3/H3K27me3 genes in the human fetal tissues. The majority of the genes (numbers indicated) with the bivalent H3K4me3/H3K27me3 state in the sequential ChIP-seq datasets also have H3K4me3 and H3K27me3 enrichments in the regular ChIP-seq datasets.