reChIP-seq reveals widespread bivalency of H3K4me3 and H3K27me3 in CD4(+) memory T cells

Abstract

The combinatorial action of co-localizing chromatin modifications and regulators determines chromatin structure and function. However, identifying co-localizing chromatin features in a high-throughput manner remains a technical challenge. Here we describe a novel reChIP-seq approach and tailored bioinformatic analysis tool, normR that allows for the sequential enrichment and detection of co-localizing DNA-associated proteins in an unbiased and genome-wide manner. We illustrate the utility of the reChIP-seq method and normR by identifying H3K4me3 or H3K27me3 bivalently modified nucleosomes in primary human CD4(+) memory T cells. We unravel widespread bivalency at hypomethylated CpG-islands coinciding with inactive promoters of developmental regulators. reChIP-seq additionally uncovered heterogeneous bivalency in the population, which was undetectable by intersecting H3K4me3 and H3K27me3 ChIP-seq tracks. Finally, we provide evidence that bivalency is established and stabilized by an interplay between the genome and epigenome. Our reChIP-seq approach augments conventional ChIP-seq and is broadly applicable to unravel combinatorial modes of action.

Figure 1. The reChIP-seq method.

(a) Experimental design. Black, purple, red and grey circles denote chromatin containing A and B antigens, only A antigens, only B antigens, or neither A nor B antigens, respectively. (b) ChIP- and reChIP-seq at the human HOXD locus. The colours of the boxes in the TSS state track indicate the co-occupancy patterns as described in Fig. 2.

Figure 2. reChIP-seq and normR analysis reveals bivalent promoters in primary human CD4⁺ central memory T cells.

(a) Anticipated outcomes after analysis. (X)/(Y), where X denotes (re)ChIP-seq track (A ChIP, B ChIP, BA reChIP or AB reChIP) normalized by the control Y (I=Input, A or B). Yellow, blue and blue/yellow boxes indicate enrichment, no enrichment or borderline enrichment, respectively. The colours next to the (re)ChIP enrichment patterns denote: full co-occupancy (black); A antigen partial co-occupancy (beige); B antigen partial co-occupancy (blue); low co-occupancy (yellow); pseudo co-occupancy (orange); only A antigen (purple); only B antigen (red); or no occupancy (grey). (b) TSS clustered by their enrichment pattern for H3K4me3, H3K27me3, reChIP H3K4me3 and reChIP H3K27me3 over respective controls resulting in seven classes (see a): black, full bivalent; beige, H3K4me3 partial bivalent; blue, H3K27me3 partial bivalent; orange, pseudo bivalent; purple, H3K4me3-only; red, H3K27me3-only and grey; unmodified. Within a class, the TSS are ordered from low (bottom) to high DNA methylation. The gene names on the right denote the gene state of critical regulators of T Helper subtype differentiation. Enrichment was calculated by normR (see the ‘Methods' section) (c) Enrichment/Depletion analysis of the classes in the top 1,000 most variably expressed genes. Stars indicate statistical significance (***) P value <0.0001 (Two-sided Fisher's exact test). (d) Boxplots of gene expression values of the top 1,000 most variably expressed genes contingent on the class. (e) Enrichment/Depletion analysis of the classified TSSs on 23 Human chromosomes.

Figure 3. Functional characterization of genes driven by bivalent promoters.

(a) Distribution of the seven TSS classes in High CpG content promoters (HCPs). (b) GO-term enrichment of the seven classes for GO:0032502 (developmental process), GO:0008152 (metabolic process) and GO:0002376 (immune system process). Shown are the calculated log₂ fold enrichment in the respective terms indicated by the colour (see colour key). Top, for all genes; bottom for genes annotated to be bivalent by BGDB. Stars indicate statistical significance: (***) P value <0.0001, (**) P value <0.001 and (*) P value <0.05 (Two-sided Fisher's exact test). (c) Boxplots of DNA methylation levels contingent on the seven promoter classes. (***) P value <2.2e−16 (Wilcoxon signed-rank test). (d) Top, enrichment/depletion analysis of BGDB bivalent genes in the seven classes. Stars indicate statistical significance as in b; bottom, percentage of TSS in the different classes. Dark bars denote TSS overlapping with the BGDB bivalent genes; fair bars denote the genomic distribution. (e) Boxplots of gene expression for the different classes. Dark coloured, gene expression measured in the sample used for (re)ChIP; fair coloured, gene expression average over five independent central memory T-cell samples.

Figure 4. Sequence characteristics of bivalent promoters.

(a) Boxplots of the G+C content in percentage, (b) the CpG content in percentage, (c) the CpG odds, that is, observed CpG count over the expected CpG count and (d) the CpA/TpG (deamination products of methylated CpGs) odds.

Figure 5. Bivalency is the default response to unmethylated CpG-rich DNA sequences.

CpG islands clustered by their enrichment pattern for H3K4me3, H3K27me3, reChIP H3K4me3 and reChIP H3K27me3 over respective controls. Fragment coverage for the tracks indicated above each column (1 to 4) for each CpG island (rows, sorted by CpG island length) ±2,000 base pairs centred around the centre of the CpG island. The fifth column represents DNA methylation (0 denotes no DNA methylation and 1 denotes full DNA methylation). The last two columns represent the CpG content and the CpGA/TpG content in percentage.

Figure 6. Cross-talk between genome and epigenome.

(a–c) The filled circles denote the relative fraction of TSSs in the indicated state. Purple indicates H3K4me3-only, red H3K27me3-only, black bivalent and grey the unmodified state. Continuous lines indicate possible transitions, while broken lines indicate a low probability of a transition between the states. (d,e) Possible models promoting either a bivalent (d) or H3K4me3-only state (e) as depicted in a. H3K4me3_X and H3K27me3_Y indicate that they reside on two different H3-tails, while H3K4me3_XY indicates that H3K4me3 is on both H3-tails of a nucleosome.