Preference of DNA methyltransferases for CpG islands in mouse embryonic stem cells

Abstract

Many CpG islands have tissue-dependent and differentially methylated regions (T-DMRs) in normal cells and tissues. To elucidate how DNA methyltransferases (Dnmts) participate in methylation of the genomic components, we investigated the genome-wide DNA methylation pattern of the T-DMRs with Dnmt1-, Dnmt3a-, and/or Dnmt3b-deficient ES cells by restriction landmark genomic scanning (RLGS). Approximately 1300 spots were detected in wild-type ES cells. In Dnmt1(-/-) ES cells, additional 236 spots emerged, indicating that the corresponding loci are methylated by Dnmt1 in wild-type ES cells. Intriguingly, in Dnmt3a(-/-)Dnmt3b(-/-) ES cells, the same 236 spots also emerged, and no additional spots appeared differentially. Therefore, Dnmt1 and Dnmt3a/3b share targets in CpG islands. Cloning and virtual image RLGS revealed that 81% of the RLGS spots were associated with genes, and 62% of the loci were in CpG islands. By contrast to the previous reports that demethylation at repeated sequences was severe in Dnmt1(-/-) cells compared with Dnmt3a(-/-)Dnmt3b(-/-) cells, a complete loss of methylation was observed at RLGS loci in Dnmt3a(-/-)Dnmt3b(-/-) cells, whereas methylation levels only decreased to 16% to 48% in the Dnmt1(-/-) cells. We concluded that there are CpG islands with T-DMR as targets shared by Dnmt1 and Dnmt3a/3b and that each Dnmt has target preferences depending on the genomic components.

Figure 1

DNA methylation status of NotI sites spread throughout the genome in wild-type and Dnmt-deficient ES cells. (A) RLGS profile of Dnmt3a^-/-Dnmt3b^-/- ES cells is shown as a representative; 236 spots indicated by arrowhead with number were invisible in the wild-type, Dnmt3a^-/-, and Dnmt3b^-/- single-mutant ES cells but emerged in Dnmt1^-/- and Dnmt3a^-/-Dnmt3b^-/- ES cells. Corresponding genomic loci of these spots were hypomethylated by the loss of Dnmt1 or Dnmt3a/3b. (B) Enlarged RLGS spots of wild-type and Dnmt-deficient ES cells. Areas with indicated numbers are represented in A. Spots with arrowhead exist in the profiles in Dnmt1^-/- and Dnmt3a^-/-Dnmt3b^-/- ES cells but are invisible in those of wild-type, Dnmt3a^-/-, and Dnmt3b^-/- ES cells. Note that intensity of all spots in Dnmt3a^-/-Dnmt3b^-/- ES cells (red arrowheads) is greater than those in Dnmt1^-/- (blue arrowheads). (C) Relative intensities of the representative RLGS spots. The intensity of each spot was digitized and averaged from three independent RLGS profiles and normalized by the average intensity of the surrounding invariant spots. Differences between Dnmt1^-/- and Dnmt3a^-/-Dnmt3b^-/- ES cells were compared by t test. ^*P < 0.1 ^**P < 0.01 (n = 3)

Figure 2

DNA methylation status at CpG sites around the NotI sites. Methylation status around ∼600-bp genomic region, including the NotI site, was assessed by sodium bisulfite sequencing method. Open circles and closed circles represent unmethylated and methylated cytosine residues, respectively. In wild-type ES cells, all genomic regions are hypermethylated, including the NotI sites, indicating that DNA methylation is maintained at these regions in the wild-type ES cells. In contrast, CpG sites in the same regions are demethylated in Dnmt1^-/- and Dnmt3a^-/-Dnmt3b^-/- ES cells. Although the demethylation is moderate in Dnmt1^-/- ES cells, almost all CpG sites are totally demethylated in Dnmt3a^-/-Dnmt3b^-/- ES cells, indicating that RLGS analysis reflects DNA methylation status not only at the NotI site but surrounding CpG dinucleotides of the NotI site. These data suggest that Dnmt3a/3b are more significant for DNA methylation in CpG islands than Dnmt1.

Figure 3

Evaluation of DNA methylation by methylation-sensitive quantitative PCR. (A) The amount of undigested genomic DNA after NotI treatment was estimated by real-time PCR using genomic DNA of wild-type, Dnmt1^-/-, and Dnmt3a^-/-Dnmt3b^-/- ES cells. Spots 97 and 231 are chosen as examples. (Top panels) Amplification plots of PCR products from each NotI-treated genomic DNA sample. (Bottom panels) Methylation levels at each corresponding genomic site, which are calculated from the ratio of amounts of NotI treated/untreated genomic DNA. For details, see Methods. (B) Evaluation of methylation levels at the corresponding genomic loci represented in Fig. 1B in the wild-type, Dnmt1^-/-, Dnmt3a^-/-, Dnmt3b^-/-, and Dnmt3a^-/-Dnmt3b^-/- ES cells. In the wild-type, Dnmt3a^-/-, and Dnmt3b^-/- ES cells, all NotI sites were hypermethylated and amplified by real-time PCR at the same levels as NotI-untreated genomic DNA. In contrast, NotI sites were only methylated from 16.1% to 47.5% in Dnmt1^-/- ES cells and from 0% to 8.5% in Dnmt3a^-/-Dnmt3b^-/- ES cells.

Figure 4

Summary of DNA methylation at CpG islands and genes, or repetitive elements such as interspersed repeats and centromeric satellites. Although Dnmt1 and Dnmt3a/3b are involved in DNA methylation at both repeated sequences (zigzag lines) and genes (closed boxes) with/without CpG islands (open boxes), Dnmt1 and Dnmt3a/3b act in a different fashion; Dnmt1 functions as a maintenance methyltransferase both in repeated sequences and CpG islands, whereas Dnmt3 primarily functions in CpG islands. Lower table represents the methylation statuses at various genomic areas in the wild-type and Dnmt-deficient ES cells reported here or elsewhere (Okano et al. 1999; Chen et al. 2003).