Histone H3K27ac separates active from poised enhancers and predicts developmental state

Abstract

Developmental programs are controlled by transcription factors and chromatin regulators, which maintain specific gene expression programs through epigenetic modification of the genome. These regulatory events at enhancers contribute to the specific gene expression programs that determine cell state and the potential for differentiation into new cell types. Although enhancer elements are known to be associated with certain histone modifications and transcription factors, the relationship of these modifications to gene expression and developmental state has not been clearly defined. Here we interrogate the epigenetic landscape of enhancer elements in embryonic stem cells and several adult tissues in the mouse. We find that histone H3K27ac distinguishes active enhancers from inactive/poised enhancer elements containing H3K4me1 alone. This indicates that the amount of actively used enhancers is lower than previously anticipated. Furthermore, poised enhancer networks provide clues to unrealized developmental programs. Finally, we show that enhancers are reset during nuclear reprogramming.

Conflict of interest statement

The authors declare no conflict of interest.

Figures

Fig. 1.

H3K27ac is a good candidate to distinguish active from inactive enhancer elements. (A) Heatmap of 25,036 putative enhancers based on H3K4me1 enrichment and H3K4me3 depletion in ES cells. Eight-kilobase pairs around the center of the enhancer region are displayed. (B) Gene expression of microarray data generated in duplicates in ES cells. Box-plots show all genes (All) and genes found associated with enhancers (H3K4me1+) either enriched (+) or unenriched (−) for Oct4 or (C) H3K27ac. Solid bars of boxes display the 25–75% of ranked genes with the mean indicated as an intersection. (D) Graph showing the percentage of enhancers identified in A that are called bound at P = 10−8 by the indicated factors or chromatin marks. (E) Heatmap showing a composite of distal H3K27ac- and H3K4me1-enriched regions ordered by presence and absence of these two histone marks. Left: Occurrence of short RNA reads based on this distribution within a −8 kb to +8 kb window (SI Materials and Methods).

Fig. 2.

Distal putative enhancers (n = 94,437) identified according to distal H3K4me1 enrichment in different tissue types reveals cell-specific distributions. (A) Heatmap displaying k-means clustering of ChIP-Seq datasets for H3K4me1 and the indicated transcription factors. Clusters are generated according to H3K4me1 enrichment at the enhancer regions. The different transcription factor data sets are arranged to match the order of enhancers found by clustering H3K4me1. Four-kilobase pairs around the enhancer peaks are displayed. (B) Gene expression of microarray data generated in duplicates from the different tissues as displayed (x axis). Box-plots show genes found specifically associated with enhancers in liver. The different boxes display how these genes behave in the different cell types tested as shown along the axis. Solid bars of boxes display the 25–75% of ranked genes, with the mean indicated as an intersection. (C) GO analysis displaying gene functions for genes specifically associated with enhancer activity in adult liver according to H3K4me1 enrichment but not associated with distal H3K4me1 enriched elements in the other cell types tested.

Fig. 3.

Enhanced proximal gene activity of distal H3K4me1-enriched regions is a function of H3K27ac. (A) Heatmap of H3K27ac enrichment at known TS sites (Upper) and 40,274 distal enhancers according to H3K27ac enrichment and H3K4me3 absence (Lower) in the four indicated cell/tissue types showing tissue-specific distribution. Four-kilobase pairs around the enhancer peaks are displayed. (B) Correlation analysis for the H3K27ac-enriched regions shown in A. Color intensities are a measure of correlation (Pearson), which is also indicated by numbers. Left: Correlation for H3K27ac at known transcriptional start sites. Right: Correlation for the 40,274 distal H3K27ac-enriched regions found in the four cell types combined. (C) Gene expression of microarray data generated in duplicate from adult liver. Box-plots show all genes (All) and genes found specifically associated with enhancers in liver either enriched (+) or unenriched (−) for H3K4me1 or H3K27ac. A hierarchical model corrects for multiple enhancers being associated with single genes (SI Materials and Methods). Solid bars of boxes display the 25–75% of ranked genes, with the mean indicated as an intersection.

Fig. 4.

Global enhancer networks are reset during nuclear reprogramming. (A) Heatmap of 118,935 distal enhancers based on H3K4me1 enrichment (green) and H3K4me3 depletion in the six indicated cell/tissue types showing tissue-specific enhancer distribution. Four-kilobase pairs around the center of the enhancer region are displayed. (B) Correlation analysis for the H3K4me1-enriched regions shown in A across the whole set of enhancer regions. Color intensities are a measure of correlation (Pearson), which is also indicated by numbers.

Fig. 5.

Poised and active enhancers can discriminate between current and future developmental states. (A) Gene expression of microarray data generated in duplicate from neural progenitors. Box-plots show all genes (All) and genes found specifically associated with enhancers in neural progenitors either enriched (+) or unenriched (−) for H3K4me1 or H3K27ac. A hierarchical model corrects for multiple enhancers being associated with single genes (SI Materials and Methods). Solid bars of boxes display the 25–75% of ranked genes, with the mean as an intersection. (B) Pie chart showing the distribution of H3K4me1 marked (green) and H3K27ac marked (red) enhancers in neural progenitors as indicated. In blue are unmarked enhancers. Total enhancers are based on combining distal H3K4me1- and H3K27ac-enriched regions totaling 136,397 regions enriched. Lines indicate the section of enhancers used to derive the genes for GO analysis in C and D. (C and D) GO analysis displaying gene functions for genes specifically associated with enhancer activity in neural progenitors split according to either H3K4me1 enrichment but absence of H3K27ac enrichment (C, green bars) or H3K27ac enrichment (D, red bars). Displayed is a representation of GO functions found.

Fig. 6.

Poised enhancers become activated during differentiation to activate key genes that drive the new developmental state. (A) Heatmap of 361 enhancers that are H3K4me1 enriched (Left, green) but negative for H3K27ac in ES cells and gain H3K27ac enrichment in neural progenitors (Right, orange) for the different cell types displayed. Four-kilobase pairs around the enhancer region are displayed. (B) Gene expression of microarray data generated in duplicate in the different cell types indicated (x axis) for the genes proximal to the 361 enhancers that become H3K27ac positive during differentiation from an ES cell to a neural progenitor. Box-plots show expression of these genes in the indicated cell types. Solid bars of boxes display the 25–75% of ranked genes, with an intersection as the mean. (C) GO analysis displaying gene functions for the genes in B that are specifically associated with enhancers that are poised in ES cells (H3K4me1+/H3K27ac-) and become active (H3K27ac+) in neural progenitors. (D) Gene track for H3K4me1 (green, Top), H3K27ac (red, Middle), and H3K4me3 (black, Bottom) for a 20,792-bp large region containing the Neuroglycan C gene. Tracks are shown in ES cells (Left) and NPCs (Right). Arrows indicate a poised enhancer that gains the H3K27ac mark. y axis shows number of reads.

Fig. 7.

Model for regulation of poised enhancers during differentiation to a new developmental state. While the cell is locked in a differentiation state it can activate poised enhancers in response to external stimuli. The repertoire of poised enhancers determines which responses are available to the cell.