Histone 3 K4 trimethylation (depositing H3K4me3 marks) is typically associated with active promoters yet paradoxically occurs at untranscribed domains. Research to delineate the mechanisms of targeting H3K4 methyltransferases is ongoing. The oocyte provides an attractive system to investigate these mechanisms, because extensive H3K4me3 acquisition occurs in nondividing cells. We developed low-input chromatin immunoprecipitation to interrogate H3K4me3, H3K27ac and H3K27me3 marks throughout oogenesis. In nongrowing oocytes, H3K4me3 was restricted to active promoters, but as oogenesis progressed, H3K4me3 accumulated in a transcription-independent manner and was targeted to intergenic regions, putative enhancers and silent H3K27me3-marked promoters. Ablation of the H3K4 methyltransferase gene Mll2 resulted in loss of transcription-independent H3K4 trimethylation but had limited effects on transcription-coupled H3K4 trimethylation or gene expression. Deletion of Dnmt3a and Dnmt3b showed that DNA methylation protects regions from acquiring H3K4me3. Our findings reveal two independent mechanisms of targeting H3K4me3 to genomic elements, with MLL2 recruited to unmethylated CpG-rich regions independently of transcription.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank K. Tabbada and C. Impey of the Babraham Next Generation Sequencing Facility, Babraham Institute, Cambridge, and A. Dahl and S. Reinhardt of the Deep Sequencing Group SFB 655, BIOTEC, Dresden, for their contribution to data generation. We thank H. Demond for help with sample collection; D. Spenserberger for the cultured ESCs; C. Novo and J. Ahringer for providing manuscript feedback; A. Segonds-Pichon for input into the statistical approach; and F. Krueger for contribution to sequencing QC and mapping at the Babraham Institute, Cambridge. C.W.H. and G.K. were supported by the UK Medical Research Council and Biotechnology and Biological Sciences Research Council (MR/K011332/1 and BB/J004499/1); A.F.S. and A.K. were supported by the Deutsche Forschungsgemeinschaft (KR2154/3-1 to A.K. and STE903/4-1 to A.F.S.).
Integrated supplementary information
A) A snapshot of the SeqMonk browser that shows a chromosome view of enrichment for H3K4me3 in ESCs using the ULI-nChIP method compared to ENCODE data. Probes were 2kb running windows with 500bp step, quantitated as RPKM. B) A snapshot of the SeqMonk browser that shows a chromosome view of enrichment for titrated antibody for H3K27me3 and H3K27ac in GV oocytes. Probes were 5kb running windows with 500bp step, quantitated as RPKM. C) Cumulative distribution plot for titrated antibody for H3K27me3 and H3K27ac in GV oocytes. Probes were 5kb running windows with 500bp step, quantitated as RPKM. D) A snapshot from the SeqMonk browser showing enrichment for biological replicates of H3K4me3, H3K27me3 and H3K27ac in pools of 250 GV oocytes from 25-day old C57BL/6Babr mice. Probes were 500bp running windows with 100bp step, quantitated as RPKM. Chromatin states were called using chromstaR combinatorial peak calling approach with 1kb bin size (FDR<0.0001).
A) Cumulative distribution plot for H3K4me3 ChIP-seq in d5 NGOs, d10 GOs, d15 GV and d25 GV oocytes compared to pooled input controls (left). Cumulative distribution plot for publically available H3K4me3 ChIP-seq in d7, d10 and d14 growing oocytes, 8-week GV oocytes1 compared to pooled input controls from this study (right). Probes were 5kb running windows, quantitated as RPKM. B) A snapshot of the SeqMonk browser that shows a chromosome view of enrichment for titrated antibody for H3K4me3 in d5 NGOs compared to published d7 GOs1. Probes were 500bp running windows with 100bp step, quantitated as RPKM. 1. Zhang, B. et al. Allelic reprogramming of the histone modification H3K4me3 in early mammalian development. Nature 537, 553-557 (2016).
A) Venn diagram shows the number of bivalent domains that are common between d25 GV oocytes, d5 NGOs and E11.5 PGCs. B) Pie charts show the proportion of bivalent peaks that overlap with CGI promoters, non-CGI promoter, orphan CGIs and distal regions called in PGCs, GV oocytes or that were common between PGCs and GV oocytes (p<0.0001, Chi-Square). C) SeqMonk screenshot shows enrichment for H3K4me3 and H3K27me3 in d25 GV oocytes. Chromatin states were called using chromstaR combinatorial peak calling approach with 1kb bin size (FDR<0.0001). Bivalent CGI promoters are highlighted in the grey shaded box. D) The top 30 results from gene ontology analysis for common PGC and GV oocyte bivalent domains, based on the closest TSS (within 5kb). E) Histogram of the distribution of GV oocyte transcript expression observed among genes with an unmethylated CGI (N=16,663) or non-CGI (N=44,324) promoter based on overlap with GV oocyte combinatorial states. Source data for panel d, e are available online.
A) Hierarchical clustering of biological replicates for H3K4me3 ChIP-seq in d5, d10, d15, d25, Mll2 WT and Mll2 KO oocytes, based on quantitation of 5kb running windows (RPKM). Amount of antibody used for each ChIP-seq replicate is denoted in brackets. B) Bar chart shows the number of H3K4me3 peaks in the Mll2 KO and WT that overlap CGI promoters, non-CGI promoters, orphan CGIs, and distal genomic regions (p<0.0001). C) Aligned read plots show the density of reads for H3K4me3 across GV bivalent domains compared to active promoters, defined by H3K27ac enrichment, throughout oogenesis and in Mll2 KO and WT oocytes. D) Hierarchical clustering of biological replicates and pooled replicate sets for Mll2 KO and WT oocyte RNA-seq at annotated oocyte transcripts1. E) The heatmap shows the normalised RPKM for the significantly differentially expressed genes between Mll2 KO and WT oocytes. F) Correlation between gene expression in Mll2 KO and WT oocytes of genes associated with a promoters that lost an H3K4me3 peak in the Mll2 KO (p<2.2E-16). Gene annotation comprised of the oocyte transcripts1 and canonical genes, as the majority of associated genes are not in the oocyte transcriptome due to low expression levels. Source data for panel a, d are available online. 1. Veselovska, L. et al. Deep sequencing and de novo assembly of the mouse oocyte transcriptome define the contribution of transcription to the DNA methylation landscape. Genome Biol. 16, 209-015-0769-z (2015).
A) Hierarchical clustering of biological replicates for DNA methylation in Mll2 KO and WT GV oocytes, based on quantitation of 100-CpG running windows using probes with a minimum of 1 read covering 20 CpGs. B) The heatmap shows the percentage methylation for each 100-CpG probe that was differentially methylated between biological replicates of Mll2 KO and WT oocytes. C) Pie charts show the proportion of hypermethylated and hypomethylated 100-CpG probes that overlap promoters, gene bodies, and distal regions (p<0.0001, Chi-Square). D) Gene expression for transcripts overlapping hypermethylated and hypomethylated 100-CpG probes is shown for Mll2 KO and WT oocytes. Mean expression levels were compared using a Mann-Whitney U-test. E) Relative read distribution for H3K4me3 in Mll2 KO and WT oocytes compared to pooled input controls, across Mll2 KO hypermethylated (N=1828) and hypomethylated (N=3503) DMRs, comprising merged adjacent 100-CpG probes within 500bp. F) SeqMonk screenshot shows DNA methylation across a hypomethylated DMR (merged adjacent 100-CpG windows, within 500bp distance) between Mll2 KO and WT oocytes. Mean DNA methylation across biological replicates is shown, with error bars depicting the standard deviation. H3K4me3 enrichment is shown below for Mll2 KO and WT oocytes using running windows of 500bp with 100bp step and normalised RPKM. Grey shaded boxes depict the location of DMRs. Source data for panel a, d are available online.
A) Global CpG methylation and percent methylation at domains normally fully methylated in GV oocytes1 is shown for biological replicates of Dnmt3a/b DKO and WT oocytes. Methylation data was quantitated in 10% inputs of samples used for H3K4me3 ChIP-seq experiments. Methylation for GV methylated domains was quantitated if the probe had a minimum of 1 read depth for 20 CpGs. B) Hierarchical clustering of biological replicates for H3K4me3 ChIP-seq in Dnmt3a/b DKO and WT oocytes compared to pooled input controls, based on quantitation of 5kb running windows (RPKM). Source data for panel a, b are available online. 1. Veselovska, L. et al. Deep sequencing and de novo assembly of the mouse oocyte transcriptome define the contribution of transcription to the DNA methylation landscape. Genome Biol. 16, 209-015-0769-z (2015).
Supplementary Figure 7 Association between histone modifications and sequence composition in GV oocytes.
A) The scatterplots show the enrichment for H3K4me3 and H3K27me3 in d25 GV oocytes in 5kb running windows within the unmethylated regions of the oocyte genome. The blue and red boxes represent the subset of probes used to evaluate sequence composition between: (top) domains with high H3K4me3 and H3K27me3 (bivalent domains; blue) and those domains that are enriched for neither (red), and (bottom) domains enriched only for H3K4me3 (blue) or H3K27me3 (red). The corresponding heatmaps show the relative enrichment for dinucleotide composition, and how well it clusters the selected chromatin state domains. B) The boxplots show the distribution of H3K4me3 enrichment in Dnmt3a/b DKO oocytes by increasing CpG enrichment among 5kb probes within nascent chromatin domains (left), defined as probes that fell within GV methylated domains, and randomly throughout the genome (right).
Source Data, Fig. 1a,e
Source Data, Fig. 2a,b,d
Source Data, Fig. 3c,e
Source Data, Supplementary Fig. 3d,e
Source Data, Supplementary Fig. 4a,d
Source Data, Supplementary Fig. 5a,d
Source Data, Supplementary Fig. 6a,b