DNA methylation on N(6)-adenine in mammalian embryonic stem cells

Abstract

It has been widely accepted that 5-methylcytosine is the only form of DNA methylation in mammalian genomes. Here we identify N(6)-methyladenine as another form of DNA modification in mouse embryonic stem cells. Alkbh1 encodes a demethylase for N(6)-methyladenine. An increase of N(6)-methyladenine levels in Alkbh1-deficient cells leads to transcriptional silencing. N(6)-methyladenine deposition is inversely correlated with the evolutionary age of LINE-1 transposons; its deposition is strongly enriched at young (<1.5 million years old) but not old (>6 million years old) L1 elements. The deposition of N(6)-methyladenine correlates with epigenetic silencing of such LINE-1 transposons, together with their neighbouring enhancers and genes, thereby resisting the gene activation signals during embryonic stem cell differentiation. As young full-length LINE-1 transposons are strongly enriched on the X chromosome, genes located on the X chromosome are also silenced. Thus, N(6)-methyladenine developed a new role in epigenetic silencing in mammalian evolution distinct from its role in gene activation in other organisms. Our results demonstrate that N(6)-methyladenine constitutes a crucial component of the epigenetic regulation repertoire in mammalian genomes.

Extended Data Figure 1. Low N6-mA levels in adult tissues and the lack of DNA alkylation adducts in ES cells

a, A majority of N6-mA peaks identified by SMRT-ChIP is located in H2A.X deposition region in ESCs determined by native ChIP. b, Number of SMRT-ChIP N6-mA sites at different coverage and QV cut-off. c, Top: A DNA motif of H2A.X deposition region determine with standard ChIP-seq. Bottom: sequence motifs for N6-mA peaks at H2A.X deposition regions determined with SMRT-ChIP. d, Distribution of N6-mA peaks at H2A.X deposition regions (P value determined by binomial test).

Extended Data Figure 2. LC-MS/MS data of N6-mA

a, Experimental workflow for determining N6-mA level with LC-MS/MS. [N5]N6-mA was used as the internal standard. b, N6-mA levels are ultralow in adult tissues. c, No detection of DNA alkylation adducts, such as N1-mA, N3-mA or N3-mC in mouse ES cells or Alkbh1 knockout cells by MS. d, LC-MS/MS analysis of N1-mA or N6-mA digested from synthetic oligonucleotides (top) and ES cell DNA samples (bottom). e, ESI-QTOF-MS/MS spectra of analytical standard of N6-mA nucleosides (top) and N6-mA containing HPLC fraction from ES cells.

Extended Data Figure 3. Alkbh1 is a specific N6-mA demethylase in vivo and in vitro

a, Top: schematic of the CRISPR–Cas9 approach. Alkbh1 KO alleles don’t contain the XmaI site at exon 3. Bottom left: PCR-DNA digestion approach indicating the homozygosity of the knockout alleles, which are resistant to Xma1 digestion. Bottom right: western blotting did not detect any ALKBH1 proteins in the KO cells. b, Three additional Alkbh1 knockout ES cell clones show similar levels of N6-mA upregulation. Shown are dot blot results. c, Validating the specificity of anti-N6-mA antibodies with synthetic oligonucleotides. d, Validating the specificity of anti-N6-mA antibodies with DNA samples of different N6-mA/dA ratio. 125 ng of genomic DNA (MEFs) which does not contain any endogenous N6-mA was spiked with N6-mA containing oligonucleotides at the indicated concentration. e, Tandem mass spectrometric analysis shows the lack of H2AK118/119 methylation in wild-type or Alkbh1 knockout ES cells. Spectral counts for H2A peptides containing K118/119 revealed that H2AK118/119 is predominately non-methylated at similar levels between wild-type and Alkbh1 knockout ES cells. Spectral counts are reported as an average with standard deviation from biological triplicate analyses. K118/119: no methylation; K118/119me1: K118/119 monomethylation. f, MS analysis showed that the co-purified factors with recombinant ALKBH1 proteins are mainly heat shock proteins. g, ALKBH1 proteins don’t have noticeable activities towards to dual- or hemi-methylated double-stranded oligonucleotide substrates. h, ALKBH1 activities are dependent on Fe²⁺ and α-KG. Error bars: standard deviation of triplicates. i, Ectopic expression of wild-type, but not mutant, Alkbh1 (D233A) at the catalytic motif, can rescue the aberrant increase of N6-mA level in Alkbh1 knockout ES cells. The wild-type and mutant Alkbh1 were expressed at similar levels. j, Quantification of three independent rescue experiments in i. P value as labelled, determined by t-test; error bars, s.d. for three biological replicates. k, The demethylation activity of N6-mA by recombinant D233A mutant protein is much reduced in comparison with the wild-type counterpart. l, No significant activities were detected with increasing concentrations of recombinant D233A mutant proteins in demethylation reaction. Error bars, s.d. of triplicates.

Extended Data Figure 4. RNA-seq analysis in Alkbh1 knockout ES cells

a, RT-qPCR validation of the RNA-seq analysis. Unchanged genes (gene names labelled in black) identified by RNA-seq were unaltered in RT-qPCR analysis. Highly repressed (red), or modestly repressed (green) genes identified by RNA-seq also showed expected levels of repression in RT-qPCR analyses. Of note, the genes (blue) identified as upregulated in RNA-seq; however, they don’t show differential expression (no significance) in RT-qPCR analysis, which further confirmed the suppression function of ALKBH1. Error bars, s.d. of triplicates. b, MA plot of RNA-seq analysed by DESeq2, which shows the similar pattern to that of CuffDiff2 (see Fig. 3a and Methods). c, Gene ontology analysis demonstrated that lineage specifying factors involved in embryonic development are greatly downregulated by Alkbh1 deficiency. d, RNA-seq transcripts of the representative subfamilies in three major retrotransposon superfamilies (LINE, SINE and LTR) in Alkbh1 knockout ES cells (Methods).

Extended Data Figure 5. Validation of N6-mA DIP-seq approach

a, ‘Spike-in’ experiments for determining the threshold and linear response range of N6-mA DIP. Genomic DNAs were spiked with N6-mA containing oligonucleotides at indicated concentration (x axis). After N6-mA DIP, the relative enrichment of N6-mA over input control was determined by a RT-qPCR approach. Blue line: linear regression based on data points between 20–130 p.p.m. The threshold (the red line) is the background signals detected by RT-qPCR in which unmodified (control) oligonucleotides were spiked in. b, The track of different sequencing method showed N6-mA sites overlapped between SMRT-ChIP and DIP-Seq in Alkbh1 knockout ES cells. c, Number of SMRT-ChIP N6-mA sites in Alkbh1 knockout cells at different coverage and QV cut-off. With rising coverage and QV cut-off, overlap between SMRT-ChIP N6-mA sites and DIP-Seq N6-mA sites also increases. d, The biological replicates of Alkbh1 knockout ES cells N6-mA-DIP peaks show 87.4% overlap. e, A large majority of N6-mA peaks are in the intergenic regions at the whole-genome level or on the X chromosome. f, In Alkbh1 knockout ES cells, N6-mA peaks are mainly targeted to LINE-1 transposons on the X chromosome or genome-wide. g, N6-mA peaks are significantly enriched on full-length, but not on truncated L1 elements (P < 1.0 × 10⁻⁵, chi-squared test). h, Enrichment of N6-mA in each full length L1 subfamily. Lx, L1_Mus1-4: >6 million years; L1VL1, L1MdF1-4: 1.5–6 million years; L1MdGf, L1MdA, L1mdT: <1.5 million years.

Extended Data Figure 6. N6-mA enrichment on 5′-end of young full-length L1 elements

a, Aggregation plot shows that signal intensity of N6-mA at young full-length L1 is enriched at the 5′ UTR and ORF1. b, qPCR analysis of N6-mA DIP samples confirmed the enrichment at the 5′ UTR and ORF1 regions of L1 that are retained in the young full-length L1 elements, but not the 3′ UTR or Nanog promoter.

Extended Data Figure 7. The correlation between N6-mA deposition on young full-length L1 elements and epigenetic silencing

a, Violin diagram of the density distribution of the distance between L1 and downregulated genes in Alkbh1 knockout cells. b, The distances between ES cells expressing genes in Alkbh1 knockout ES cells and young full-length L1 elements were plotted for indicated chromosomes. c, The distances between downregulated genes in Alkbh1 knockout ES cells and young full-length L1 elements were plotted for indicated chromosomes.

Extended Data Figure 8. N6-mA accumulation correlates with epigenetic silencing

a, Normalized 5mC levels on gene bodies or promoters in wild-type or Alkbh1 knockout ES cells. b, Histone marks (H2A.X or H3K27Me3) or 5 mC levels on young full-length L1 elements, SINE or LTR transposons. c, Representative sequencing tracks of decommissioned enhancers. H3K27Ac and H3K4me1 levels at this locus are greatly downregulated in Alkbh1 knockout ES cells. See Supplementary Table 2 for all decommissioned enhancers in Alkbh1 knockout ES cells. d, Violin diagram shows the density distribution of the distance between L1 and decommissioned enhancers in Alkbh1 knockout cells. e, ChIP-qPCR approach showed that H3K4me3 levels are decreased at the transcription start sites (TSS) of LINE-1 or Dax1, an X chromosome gene, while unchanged at the control gene TSS. *P < 0.01, t-test; error bars, ± s.e.m. of three technical triplicates.

Extended Data Figure 9. N6-mA accumulation results in imbalanced cell fate decisions during ESC differentiation

Wild-type or Alkbh1 knockout ES cells were subject to embryoid body differentiation (Methods). mRNA samples were collected at day 1 or day 9. Gene expression levels were quantified by RT–qPCR approaches. *P < 0.01, t-test; error bars, ± s.e.m. of technical triplicates. a, At day 9, Nanog expression is reduced significantly in wild-type ES-cell-derived embryoid bodies as expected, while its level in Alkbh1 knockout ES-cell-derived embryoid bodies is still high. b, Lefty-1 and Lefty-2 are repressed at day 1 or day 9 in Alkbh1 knockout ES-cell-derived embryoid bodies. c, Activation of Cdx2, is insufficient in Alkbh1 knockout ES-cell-derived embryoid bodies. d, However, expressions of other endoderm markers, Foxa2, Gata4, Gata6, are significantly higher in Alkbh1 knockout ES-cell-derived embryoid bodies than wild-type ES-cell-derived embroid bodies. e, Ectoderm markers, Fgf5 and Pax6 are transiently (day 1) overexpressed in Alkbh1 knockout ES-cell-derived embryoid bodies. f, Mesoderm marker, T/Brachyury is similarly expressed in wild-type or Alkbh1 knockout ES-cell-derived embryoid bodies during differentiation.

Figure 1. A SMRT-ChIP approach identified N6-mA in mammalian genomes

a, Schematic of SMRT-ChIP. b, Sequencing tracks of N6-mA in ES cells. IPD ratio, inter-pulse distance ratio. c, Top: LC-mass spectrometry analysis of N6-mA (m/z = 266.1 to m/z = 150.1). Bottom: stable isotope labelled N6-mA (m/z = 271.1 to m/z = 155.1), internal standard. MRM, multiple reaction monitoring. d, Quantification of the LC-MS/MS results. P < 0.01, t-test; error bars, ± s.e.m. of three biological replicates.

Figure 2. Alkbh1 is a demethylase for N6-mA in ES cells

a, Mass spectrometry analysis of N6-mA in Alkbh1 knockout (KO) ES cells (P value determined by t-tests). b, Dot blotting of N6-mA in Alkbh1 knockout or wild-type (WT) ES cells (in triplicates). c, In vitro demethylation reaction with recombinant ALKBH1 proteins monitored by dot blotting (Methods). d, Quantification of demethylation activity in three independent demethylase assays in c (P value <5.0 × 10⁻⁵, t-test). e, In vitro demethylation reaction monitored by mass spectrometry (P value <0.01, t-test). Error bars, s.d. for three biological replicates.

Figure 3. Alkbh1 deficiency silences genes on the X chromosome and young full-length L1 elements

a, RNA-seq analysis of Alkbh1 knockout ES cells versus wild-type controls. Blue: most highly downregulated genes, red: upregulated genes (false positives, see main text). b, Downregulated genes were most enriched on X chromosome (P < 0.01, binomial test) and Chr13 to a lesser extent (P < 0.05, binomial test). c, qRT-PCR analysis of downregulated genes (*P < 0.05, t-test). d, RT-qPCR of transposon expression (*P < 0.01, t-test). L1Md-Gf-X: a young full-length L1 on Chr-X L1Md-Gf-17: a young full-length L1 on Chr17. Error bars, ± s.e.m. of three technical replicates.

Figure 4. N6-mA is enriched at young full-length L1 elements, which are located in the vicinity of the downregulated genes in Alkbh1 knockout ES cells

a, Enrichment of N6-mA on full-length L1 elements (P value determined by t-test). b, Left: relative enrichment of N6-mA peaks on each chromosome (P = 1.4 × 10⁻³²², binomial test). Right: relative enrichment of young full-length L1 s on each chromosome. c, Normalized frequency of full-length L1 elements was plotted as a function of their genomic distance to downregulated genes (red, N6-mA enriched, median: 424 kb; grey, non-enriched, median: 1.6 Mb). d, The Dax1 gene locus.

Figure 5. N6-mA upregulation induced transcriptional silencing on the X chromosome, which is persistent during differentiation

a, Aggregation of 5mC. b, Aggregation of H3K9Me3 signals. c, Normalized frequency of decommissioned enhancers was plotted as a function of their genomic distance to full-length L1 elements red, N6-mA enriched, median: 484 kb; grey, non-enriched, median: 2 Mb. d, RT-qPCR analysis of the Gm8817 and Rhox6 genes (on the X chromosome) during embryoid body differentiation. *P < 0.05, t-test; error bars, ± s.e.m. of three biological replicates. e, Schematics of Alkbh1 and N6-mA functions (see main text).