Preventing CpG hypermethylation in oocytes safeguards mouse development

Abstract

Except for regulatory CpG-island sequences, genomes of most mammalian cells are widely DNA-methylated. In oocytes, though, DNA methylation (DNAme) is largely confined to transcribed regions. The mechanisms restricting de novo DNAme in oocytes and their relevance thereof for zygotic genome activation and embryonic development are largely unknown. Here we show that KDM2A and KDM2B, two histone demethylases, prevent genome-wide accumulation of histone H3 lysine 36 di-methylation, thereby impeding DNMT3A-catalyzed DNAme. We demonstrate that aberrant DNAme at CpG islands inherited from Kdm2a/Kdm2b double-mutant oocytes represses gene transcription in two-cell embryos. Aberrant maternal DNAme impairs pre-implantation embryonic development, which is suppressed by Dnmt3a deficiency during oogenesis. Hence, KDM2A/KDM2B are essential for confining the oocyte methylome, thereby conferring competence for early embryonic development. Our research implies that the reprogramming capacity eminent to early embryos is insufficient for erasing aberrant DNAme from maternal chromatin, and that early development is susceptible to gene dosage haplo-insufficiency effects.

Keywords: CpG island; DNA methylation; KDM2A; KDM2B; PRC1; Polycomb; embryogenesis; maternal epigenetic inheritance; oocyte; reprogramming.

Graphical abstract

Figure 1

KDM2A and KDM2B function in oocytes to regulate embryonic development (A) Schematic overview of Kdm2a and Kdm2b genes expressed in oocytes of ctrl and mutant conditions. Positions of JmjC- and CxxC-encoding domains flanked by loxP sites are indicated.^, (B) Developmental progression rates of ctrl, single, and double/compound mutant pre-implantation embryos at indicated days of in vitro embryonic development. p values according to Fisher’s exact test. (C–E) Immunofluorescence staining and quantification of KDM2A, cytosolic KDM2B, H2AK119u1, and H3K27me3 in GOs (KDM2A) or FGOs (others) of indicated genotypes. p values according to two-sided Student’s t test. Scale bars, 20 μm. For panels (B)–(E), numbers of embryos or oocytes analyzed are indicated.

Figure 2

KDM2A/KDM2B regulate H2AK119u1 deposition and gene expression during oogenesis (A) MA plots showing differential expression of indicated mutant FGOs over respective ctrl FGOs (log2 fold change [log2FC]) as a function of expression in respective ctrl FGOs (log2RPKM). #UP and #DN refer to numbers of more highly or lowly expressed genes in mutant versus ctrl FGOs (log2FC > 1.0; adj p value < 0.05). Ratio refers to #UP genes over #DN genes. (B) Venn diagram showing numbers of genes upregulated in indicated mutant FGOs relative to respective ctrl FGOs. (C) Scatterplots showing expression log2FC of indicated mutant FGOs over ctrl FGOs versus indicated mutant FGOs over ctrl FGOs. H2AK119u1 occupancy (log2) at promoters (−1,500/+500 bps of TSS) is indicated by color scale. R indicates Pearson’s correlation coefficient. (D) Heatmap displaying sequence composition, transcriptional, and chromatin variables within CGI-promoter genes (5 kb upstream, TSS, gene body, TES, and 5 kb downstream) grouped into 8 clusters by k-means clustering. From left to right: gene numbers per cluster; CpG coverage; GC percentage; oocyte-specific sense (green) and antisense (red) transcripts; absolute RNA (scaled RPKM) in ctrl and mutant FGOs; log2FC expression in mutant versus ctrl FGOs (delta); H2AK119u1 occupancy in ctrl and mutant FGOs; and H3K4me3, H3K36me3, H3K27me3, and H2AK119u1 occupancies in WT GOs and FGOs.^,, All chromatin data are shown as Z scores. (E) Boxplot presenting RNA expression levels per gene cluster in ctrl FGOs with gene numbers per cluster indicated. (F) Boxplot presenting log2FC in expression per gene cluster measured in various mutant FGOs relative to respective ctrl FGOs.

Figure 3

KDM2A/KDM2B restrict H3K36me2 and DNAme at promoters and along genes in oocytes (A) Immunofluorescence staining and quantification of H3K36me2 and H3K36me3 in GOs and 5mC in FGOs of indicated mutant and respective ctrl genotypes. Numbers of analyzed oocytes are indicated. p values according to the two-sided Student’s t test. Scale bars, 20 μm. (B) 2D-density plots displaying distributions of CpGs according to their mean methylation levels (mCpG/CpG in %) in mutant FGOs versus ctrl FGOs. (C) Violin plot showing distribution of mCpG/CpG (%) values at different genome elements in FGOs of indicated genotypes. (D) Heatmap displaying sequence composition and chromatin variables within 8 CGI-promoter gene clusters previously described in Figure 2D. From left to right: CpG coverage; GC percentage; H3K36me2, H3K36me3, and DNAme in FGOs of indicated genotypes; differential DNAme at CGI promoters in mutant versus ctrl FGOs. (E) Genomics snapshot of the Hoxa–Evx1 gene cluster illustrating gain in DNAme and H3K36me2 at CGI promoters (in orange) and gene bodies in mutant versus ctrl FGOs. (F) Boxplots displaying differential H3K36me2, DNAme, and H3K36me3 at CGI promoters (−1,500/+500 bps of TSS) or gene bodies (+500 bps of TSS to TES) of genes assigned to gene clusters in which expression is either upregulated, downregulated, not changed, or not detected in Kdm2a^KOKdm2b^KO FGOs relative to ctrl FGOs. Numbers of genes per condition are indicated. (G) Boxplots displaying differential H3K36me2 and DNAme at promoters of all genes belonging to the 8 CGI-promoter gene clusters in Kdm2a^KOKdm2b^KO and Kdm2a^KOKdm2b^ΔCxxC FGOs relative to ctrl FGOs, as indicated. Numbers of genes per cluster are indicated.

Figure 4

H3K36me2 and DNAme accumulate in Kdm2a/Kdm2b-deficient oocytes, independently of transcription (A) Heatmap displaying chromatin and transcriptional variables within 8 clusters of 10 kb intergenic regions in oocytes. Data are displayed in 20 neighboring 500 bp bins. From left to right: number of regions per cluster; CpG coverage; GC percentage; H2AK119u1, H3K36me2, and DNAme in ctrl and mutant FGOs; presence of annotated TTS in 20 kb flanking regions that could be compatible with run-through transcription through the window; log2FC in expression between indicated mutant and ctrl genotypes in day 9 and day 14 GOs; in random primed (total) day 14 GOs and in FGOs. RNA expression data are based on poly(A)-primed RNA capture and Smart-seq2 library generation, if not indicated otherwise. (B–D) Genomic interval^,, snapshots illustrating gain of DNAme and H3K36me2 at large intergenic regions in Kdm2a^KOKdm2b^KO FGOs that are marked by H2AK119u1 and H3K27me3 in ctrl FGOs.

Figure 5

Low H3K4me3 in GOs is permissive for H3K36me2 and DNAme acquisition during oocyte growth (A) Scatterplot showing correlations between H3K36me2 and H3K4me3 occupancies and DNAme (%) at promoter and intragenic CGIs in ctrl and WT FGOs. (B) Scatterplot showing differential DNAme (%) at promoter and intragenic CGIs between Kdm2a^KOKdm2b^KO over ctrl FGOs in relation to H3K36me2 occupancy in Kdm2a^KOKdm2b^KO FGOs and H3K4me3 occupancy in WT GOs. (C) Bar plot showing linear regression coefficients of chromatin marks, genomic location, and triplet nucleotide sequences contributing to predicting differential H3K36me2 occupancy at CGIs in Kdm2a^KOKdm2b^KO over ctrl FGOs (R² = 0.338). (D) Boxplots representing frequencies of CCG/CGG trinucleotides per 100 bp and enrichments of H3K36me2 in ctrl and Kdm2a^KOKdm2b^KO FGOs at 533 bp regions surrounding CGI centers for bins of H3K4me3 occupancies in GOs having low or high H2AK119u1 levels. (E) Cartoon illustrating dependencies between DNA sequences, histone modifying enzymes, and DNA methyltransferases in regulating H3K36 di-/tri-methylation and de novo DNA methylation.

Figure 6

Aberrant maternal DNA methylation impairs pre-implantation development (A) Immunofluorescence staining and quantification of 5mC in maternal (m) and paternal (p) pronuclei of ctrl and mutant late zygotes (numbers indicated). p values according to Tukey’s HSD test. Scale bars, 20 μm. (B) Developmental progression rates of ctrl and triple mutant pre-implantation embryos at indicated days of in vitro embryonic development. p values according to Fisher’s exact test. (C) Table summarizing post-implantation development of embryos with indicated genotypes. (D) MA plots showing differential expression in maternally mutant over respective ctrl two-cell embryos (log2FC) as a function of expression in ctrl two-cell embryos (log2RPKM) for all (total) and parental-specific sequencing reads. Genes upregulated or downregulated in mutants are indicated in red and blue (|log2FC| > 1.0; adj p value < 0.05). (E) Left: scatterplot showing log2FC expression of Kdm2a^matKOKdm2b^matKO over ctrl two-cell embryos versus log2FC expression of Kdm2a^KOKdm2b^KO over ctrl FGOs for parental-specific expression of CGI- and non-CGI-promoter-associated genes. Right: data as in the left panel for Kdm2a^matKOKdm2b^matΔCxxC over ctrl two-cell embryos versus Kdm2a^KOKdm2b^ΔCxxC over ctrl FGOs. Differential promoter methylation (%) in Kdm2a^KOKdm2b^KO over ctrl FGOs is indicated by color.

Figure 7

Genes marked by aberrant maternal promoter DNAme are repressed in early embryos (A) Heatmap showing clustering of UCSC-based CGI promoters (numbers indicated) according to absolute promoter methylation levels in ctrl and mutant FGOs and in WT sperm. (B) Level of enrichments and statistical significances in Gene Ontology enrichment analyses for genes associated with CGIs belonging to DNAme clusters 1–5 shown in (A). (C) Boxplot showing log2FC expression of clusters of CGI-promoter and all non-CGI-promoter genes in Kdm2a^KOKdm2b^KO over ctrl FGOs and in Kdm2a^matKOKdm2b^matKO over ctrl two-cell embryos according to all and parental-specific sequencing reads. (D) Bar plot showing over-/under-representation and statistical significance of CGI-promoter genes in DNAme clusters and being either significantly up- or down-expressed in Kdm2a^KOKdm2b^KO relative to ctrl FGOs or in Kdm2a^matKOKdm2b^matKO relative to ctrl two-cell embryos for all and parental-specific sequencing reads. Statistical significance (p value) is coded as follows: ^∗∗∗∗p ≤ 0.001%, ^∗∗p ≤ 0.1%, ^∗p ≤ 1%, p ≤ 5%. (E) Bar plot showing over-/under-representation and statistical significance of CGI-promoter genes in DNAme clusters that are significantly upregulated or downregulated in two-cell embryos treated with alpha-amanitin, an inhibitor of RNA polymerase II and III.p value: ^∗∗∗∗p ≤ 0.001%, ^∗∗p ≤ 0.1%, ^∗p ≤ 1%. (F) Heatmap showing clustering of UCSC-based CGI promoters according to absolute promoter DNAme levels in ctrl and Kdm2a^KOKdm2b^KO FGOs and ctrl and Kdm2a^matKOKdm2b^matKO four-cell embryos at maternal and paternal genomes. Mat- and pat-hypo refer to CGIs undergoing de novo methylation in ctrl embryos. “Genebodies” refers to CGIs having undergone transcription-coupled de novo DNAme in GOs. (G) Boxplot showing log2FC expression of different clusters of CGI- and non-CGI-promoter genes shown in (F) in Kdm2a^matKOKdm2b^matKO over ctrl two-cell embryos according to parental-specific sequencing reads. (H) Bar plot showing over-/under-representation and statistical significance of CGI-promoter genes in DNAme clusters as defined in (F) and being either significantly up- or down-expressed in Kdm2a^matKOKdm2b^matKO relative to ctrl two-cell embryos according to parental-specific sequencing reads. p value: ^∗∗∗∗p ≤ 0.001%, ^∗p ≤ 1%, p ≤ 5%. (I) Genomic snapshots of Gldc, Eomes, and Hes2 genes with aberrant DNAme at CGI promoters (in orange) in oocytes and four-cell embryos. RNA expression in FGOs and two-cell embryos (both alleles), DNA methylation in FGOs and four-cell embryos (maternal alleles), and chromatin marks in FGOs are indicated.