Skip to main content
. 2022 May 23;11:e66909. doi: 10.7554/eLife.66909

Figure 3. Loss of Dnmt3a leaves thousands of genomic regions in a fetal-like demethylated state.

(A) 70 genes were differentially expressed (DE) (false discovery rate [FDR] < 0.05) in P39 pyramidal neurons in Dnmt3a conditional knockout (cKO) vs. control. Top, fold-change (FC) between cKO and control; bottom, heatmap showing normalized expression of the DE genes in each sample. Z-scores were computed using mRNA counts per million (CPM) for each DE gene. (B) Differential gene expression in control vs. Dnmt3a cKO excitatory neurons at P39. Significant up-regulated and down-regulated DE genes are shown in red and blue, respectively. Differentially expressed (DE) genes associated with dendrite morphogenesis (Elavl4, Hecw2, Ptprd), and regulation of Na+ (Hecw2, Scn3b) and Ca2+ levels (Cacnb3) are labeled. (C) Non-CG DNA methylation (mCH) is eliminated, and mCG is reduced, in P39 Dnmt3a cKO pyramidal cells, while mCG and mCH levels are not changed in P0 (t-test: *, p<0.05; n.s., not significant). P0 and P39, postnatal days 0 and 39, respectively. Each bar represents the methylation level in one replicate. (D) Non-CG DNA methylation (mCH) in P39 pyramidal cells in control samples, in 1 kb bins in the flanking region around the transcription start (TSS) and end site (TES) of DE genes and non-DE genes with matched expression levels. The lines denote the means across genes in each gene set, and the shared areas represent the 95% confidence intervals of the means. (E) The difference in gene body methylation vs. fold-change of gene expression between P39 Dnmt3a cKO and control. The plots show mean ± SEM gene expression fold-change for genes in 10 non-overlapping bins (deciles of mC difference). (F) The Nedd4 promoter locus contains five differentially methylated regions (DMRs, yellow horizontal rectangles and shaded in blue boxes) with naive, fetal-like mCG in P39 Dnmt3a cKO. Ticks show mCG at CG sites. Four out of the five P39 Dnmt3a cKO DMRs overlapping developmental gain-of-methylation DMRs (red horizontal rectangles) are marked with arrows. CGI, CpG island. R1 and R2, replicates 1 and 2. (G) Overlap of P39 Dnmt3a cKO DMRs and developmental DMRs. (H) P39 Dnmt3a cKO hypo-DMRs are significantly enriched (depleted) in DMRs that normally gain (lose) methylation during development (Fisher’s test, p<0.05).

Figure 3.

Figure 3—figure supplement 1. RNA-seq data showed transcriptomic disruption in P39 Dnmt3a conditional knockout (cKO) pyramidal neurons.

Figure 3—figure supplement 1.

(A) Correlation matrix of gene expression (log10TPM) in the biological replicates of the control and Dnmt3a cKO mouse excitatory neurons. The color bar represents the Spearman correlation coefficients. (B) Volcano plot shows the gene expression fold-change of P39 Dnmt3a cKO vs. control samples and their significance. Significant up-regulated and down-regulated differentially expressed genes (DE genes, false discovery rate [FDR] < 0.05) are colored in red and blue, respectively. DE genes associated with dendrite morphogenesis (Elavl4, Hecw2, Ptprd), and regulation of Na+ (Hecw2, Scn3b) and Ca2+ levels (Cacnb3) are labeled.
Figure 3—figure supplement 2. The expression of transposable elements (TEs) was not affected by Dnmt3a conditional knockout (cKO).

Figure 3—figure supplement 2.

TE subfamily expression was estimated in fragments per kilobase per million (FPKM) from P39 control and Dnmt3a cKO samples grouped by TE class (A) and TE family (B).
Figure 3—figure supplement 3. Genome-wide reduction of DNA methylation was observed in Dnmt3a conditional knockout (cKO).

Figure 3—figure supplement 3.

(A) Heatmap to show the Spearman correlations across the control and Dnmt3a cKO samples from P39 and P0 animals. The correlations coefficients were computed using CG methylation levels in 10 kb genomic bins (left) and CH methylation levels in 10 kb genomic bins (right). (B) The reduction of genome-wide DNA methylation level in P39 is observed in all three non-CG contexts (CA, CC, and CT). t-test significance, **, CA, p=0.009745; ***, CC, p=0.0009546; *, CT, p=0.01323. (C) DNA methylation at CG sites is reduced across most functional genomic compartments in P39. UTR, untranslated region. t-test significance, *, p<0.05; n.s., not significant. (D) Reduced mCG is strongly correlated with the reduction in mCH in P39 in 10 kb tiling genomic bins (259,718 bins with at least 10 reads covered in each sample). r, Spearman correlation coefficient.
Figure 3—figure supplement 4. Reduction of DNA methylation cannot fully explain the disruption in the transcriptome after Dnmt3a conditional knockout (cKO).

Figure 3—figure supplement 4.

(A) Correlation of gene expression and gene body mCH level for up-regulated genes (red), down-regulated genes (blue), and non-differentially expressed (DE) genes (black, false discovery rate [FDR] ≥ 0.05 and fold-change <1.1) in the P39 control samples. For each gene group, genes are stratified by their expression in the control sample by 15 bins, and the mean gene body mCH levels are plotted. The shaded ribbon areas indicate the standard error of the mean. TPM, transcripts per million. (B) Violin plots to show the expression levels of non-DE genes (gray) selected to match baseline expression as those in significantly (FDR < 0.05) up-regulated genes (left, red) and down-regulated genes (right, blue). (C) CG DNA methylation (mCG) in P39 pyramidal cells in 1 kb bins in the region around the transcription start (TSS) and end site (TES) of DE genes and non-DE genes with matched expression levels. The lines denote the means across genes in each gene set, and the shared areas represent the 95% confidence intervals of the means. (D) Density scatter plots show the relationship between changes of gene body methylation (delta mCG or mCH) and the gene expression fold-changes for expressed genes (14,754 genes) between P39 Dnmt3a cKO and control samples. The linear regression fits, p-values, and variances explained by ∆%mC (R2) are shown. (E) Gene length distribution of P39 DE genes (FDR < 0.1). As a comparison, non-DE genes were selected with FDR ≥ 0.1 and fold-change <1.1 (see Supplementary file 2). The down-regulated genes are generally shorter than the up-regulated genes or the non-DE genes. kb, kilobases. Wilcoxon test, ****, p<10–4; **, p<0.01; n.s. not significant.
Figure 3—figure supplement 5. Differentially methylated regions (DMRs) in P39 Dnmt3a conditional knockout (cKO) were associated with regulatory regions in enhancers and repressed chromatin.

Figure 3—figure supplement 5.

(A) 2D distribution of mCG levels in P39 control vs. Dnmt3a cKO samples at DMRs. DMR density is estimated through a Gaussian smoothed kernel. (B) The density of P39 Dnmt3a cKO hypo-DMRs around significant differentially expressed (DE) genes (false discovery rate [FDR] < 0.05) and non-DE genes with matched expression levels. The lines denote the means across genes in each gene set, and the shared areas represent the 95% confidence intervals of the means. TSS, transcription start site; TES, transcription end site. (C) Enrichment (red) or depletion (blue) of P39 cKO DMRs in GENCODE annotated gene features (top), in CpG island-related features (middle), and in the chromatin states map in mouse embryonic stem cell (bottom). All enrichments and depletions shown are significant (Fisher’s test p<0.05). (D) Number of known transcription factor binding motifs within P39 Dnmt3a cKO hypo-DMRs and their fold enrichment. Significant motifs (FDR < 0.05) are colored in red.