Identification of methylated deoxyadenosines in vertebrates reveals diversity in DNA modifications

Abstract

Methylation of cytosine deoxynucleotides generates 5-methylcytosine (m(5)dC), a well-established epigenetic mark. However, in higher eukaryotes much less is known about modifications affecting other deoxynucleotides. Here, we report the detection of N(6)-methyldeoxyadenosine (m(6)dA) in vertebrate DNA, specifically in Xenopus laevis but also in other species including mouse and human. Our methylome analysis reveals that m(6)dA is widely distributed across the eukaryotic genome and is present in different cell types but is commonly depleted from gene exons. Thus, direct DNA modifications might be more widespread than previously thought.

Figure 1. Identification of dA^6m in the genome of higher eukaryotes.

a, Illustration of dA^6m identification. dA^6m was identified using dot blots, UHPLC-MS/MS and dA^6m Ab DIP sequencing (DIP-Seq). b, Dot blot with dA^6m Ab on DNA templates. c-d, Representative (1 out of 4) UHPLC-MS/MS extracted ion chromatogram (EIC) (left) and fragmentation spectrum (right) monitoring presence of dA^6m in genomic DNA from X. laevis. * indicates parent ion, AU=arbitrary units, m/z=mass to charge ratio, n=4 biological replicates (tissues from different animals). e, Percentage of dA^6m versus total deoxyadenosines in DNA from different samples following dA^6m Ab DIP enrichment. Error bars, s.e.m. n=4 tissues from different animals or from independent bacterial cell cultures, **P≤0.005, *P≤0.05, two-sided t-test. f, Percentage of dA^6m versus total deoxyadenosines in DNA from different samples without dA^6m Ab DIP enrichment. Error bars, s.e.m., n=4 tissues from different animals or from independent bacterial cell cultures, **P≤0.005, *P≤0.05, two-sided t-test, but when one sample was zero the one-sided t-test was applied. g, dA^6m dot-blot of DNA from different higher eukaryotic sources, n=3 technical replicates.

Figure 2. Genome wide identification of dA^6m marks in X. laevis fat, oviduct, testes and M. musculus kidney.

a, dA^6m peak signal tracks. One gene region for each tissue type is shown. The y-axis represents the amount of immunoprecipitated DNA at each position normalized by the total number of reads. RefSeq gene annotations are shown. n=2 for each X. laevis tissue and n=3 for M. musculus, biological replicates from different animals, AU=arbitrary units. b, Heat map of Pearson correlation coefficient values from comparisons between X. laevis and M. musculus dA^6m DIP-Seq and input-Seq samples. Correlation was calculated pairwise for all samples, excluding windows where both samples in the pair had zero depth. n=2 for each X. laevis tissue and n=3 for M. musculus, biological replicates from different animals, **P<1×10⁻¹⁶, Pearson Correlation test on mapped reads. c, Scatter plots comparing individual samples pairwise for overlapping peaks. The enrichment score ranging from 0–8 is plotted on both axes. The color in each plot reflects the correlation value (Pearson correlation coefficient), which is also shown in the top right corner of each plot. n=2 for each X. laevis tissue and n=3 for M. musculus, biological replicates from different animals, **P<1×10⁻¹⁶, two sided t-test on dA^6m peaks. d, Number of unique and overlapping dA^6m peaks identified in X. laevis and in M. musculus tissues. n=2 for each X. laevis tissue and n=3 for M. musculus, biological replicates from different animals, **P<1×10⁻¹⁶, χ2-test on dA^6m peaks.

Figure 3. Genome wide distribution of dA^6m in the vicinity of genes.

a, Distribution of dA^6m peaks around TSS, from 20kb 5’ to 20kb 3’, identified in X. laevis fat, oviduct, testes and in M. musculus kidney. One biological replicate from one animal is shown in each graph. b, Density of dC^5m versus unmethylated dC in distinct areas of the M. musculus testes genome. Only the regions shown in dark grey are statistically significant. One biological replicate from one animal is shown, **P<0.007, binomial test on dA^6m peaks. c, Density of dA^6m versus unmethylated dA in distinct areas of the genome of X. laevis testes. Only the regions shown in dark grey are statistically significant. One biological replicate from one animal is shown, *P<0.03, binomial test on dA^6m peaks.

Figure 4. Genome wide distribution of dA^6m peaks in X. laevis testes samples as determined with three different dA^6m recognizing antibodies in DIP-Seq and comparison to input-Seq or IgG-Seq controls.

a, dA^6m peak signal tracks. Tracks obtained from dA^6m Ab DIP-Seq, dA^6m Ab* DIP-Seq, dA^6m Ab** DIP-Seq and control input-Seq and IgG-Seq are shown. One gene region for each biological replicate is shown. The y-axis of each profile represents the amount of reads at each position normalized by the total number of reads in a given dataset. RefSeq gene annotations are shown. n=2, biological replicates from different animals, AU=arbitrary units. b, Heat map of Pearson correlation coefficient values from comparisons between X. laevis testes dA^6m Ab DIP-Seq, dA^6m Ab* DIP-Seq, dA^6m Ab** DIP-Seq, IgG DIP-Seq and input-Seq samples. Correlation was calculated pairwise for all samples, excluding windows where both samples in the pair had zero depth. n=2, biological replicates from different animals, **P<1×10⁻¹⁶, Pearson Correlation test on mapped reads. c, Scatter plots comparing individual samples pairwise for overlapping peaks. The enrichment score ranging from 0–8 is plotted on both axes. The color in each plot reflects the correlation value (Pearson correlation coefficient), which is also shown in the top right corner of each plot. n=2, biological replicates from different animals, **P<1×10⁻¹⁶, two sided t-test on dA^6m peaks.

Figure 5. dA^6m motif identification in bacteria and X. laevis.

a, Abundance of 4bp motif in dA^6m peaks. 256 potential 4bp motifs, each representing one column along x-axis, were ranked for their abundance in Dam+ and Dam– bacterial dA^6m peaks. The Dam recognition motif GATC is shown in red. n=1 biological replicate from one bacterial culture for Dam+ and Dam- bacteria, Spearman rank correlation coefficient=0.94 between Dam+ and Dam– bacteria, Spearman’s **ρ=2.2×10⁻¹⁶. b, Ratio between Dam+ and Dam– bacteria enriched 4bp motifs. Only motifs that are at least 5% enriched are illustrated. c, Bacterial dA^6m motif identified by MEME. In Dam+ bacteria, the motif GATC has been identified (73 out of 73 gene regions identified by MEME). n=1 biological replicate from one bacterial culture for Dam+ and Dam- bacteria, AU=arbitrary units. d, X. laevis dA^6m motif identification by MEME. Overlaps between biological replicates from different animals were used for analysis. Same tissue n=2 biological replicates from different animals, different tissue overlaps n=6 biological replicates for different animals, **E-value<1.2×10⁻⁸, statistics by MEME, AU=arbitrary units.