DNA methylation changes of DNA repetitive elements induced by DNMT isoforms. (A) DNA methylation changes of two interspersed DNA repeats, LINE1 and AluYb8. (B) DNA methylation changes of two tandem repeats, Sat-α and NBL-2. DNA methylation in HEK 293T cells treated with 5-aza-2′-deoxycytidine, a DNA methylation inhibitor (decreased methylation control), or M.SssI, a CpG methyltransferase (fully methylated control), is shown as 293T/DAC or 293T/M.SssI, respectively. Two independent transfection experiments are indicated as cell line 1 and cell line 2.

Distribution of CpG islands and histone marks at loci with changes in DNA methylation. (A) Distribution of CpG islands (black) and non-CpG islands (white) at hyper- and hypomethylated loci induced by each DNMT isoform. Hyper-and hypomethylated CpG sites induced by each DNMT were selected by an absolute difference in the beta value (|Δβ|-value > 0.1843). DNMT1 equally induced hyper and hypomethylated loci. Most isoforms induced exclusively hypermethylation at both CpG islands and non-CpG islands, except 3B3, 3B4 and 3B5, which are known to be catalytically inactive. (B) Unsupervised two-dimensional clustering analysis with all 514 de novo methylated loci induced by any DNMT isoform shows that de novo DNA methylation profiles induced by DNMT isoforms were clustered according to the structural similarity of the DNMT isoforms.

Changes in mRNA expression and histone modification by overexpression of DNMT3A1 and DNMT3B1. Number of downregulated (>2-fold) genes in DNMT3A1- and DNMT3B1-overexpressing cells. The number of downregulated genes in whole genome mRNA representing 8276 transcripts (A) and in the 808 genes studied for DNA methylation (B). (C) Changes in mRNA expression and histone modification. Real-time PCR to measure the mRNA expression of DNMT3A1 target genes (EYA4 and HOXA11) and DNMT3B1 target genes (IGF2AS and CDH11), and ChIP combined with real-time PCR for histone medications (H3K4me3, H3K27me3 and H3K9me3) in HEK 293T cells (white bar), DNMT3A1 cells (black bar) and DNMT3B1 cells (gray bar). Real-time RT-PCR and ChIP data represent the means ± SD. For mRNA expression, the primer efficiency for all four genes was ∼95%, and comparison of the fold changes of the standard curve was consistent with the ΔCt-calculated fold changes. (t-test; *P < 0.05 and ***P < 0.001 versus HEK 293T cells).

Generation of cell lines stably overexpressing DNA methyltransferases (DNMTs). (A) Schematic structure of DNMT isoforms and expression vector constructs. The conserved PWWP and PHD-like domains, DNMT catalytic motifs (I, IV, VI, IX and X), and alternative splicing sites are indicated. The coding regions of DNMT 3B4 and DNMT 3B5 in red represent frame shift mutations generated by alternative splicing. The expression vectors encode N-terminal myc-tagged DNMT isoforms. (B) mRNA expression of exogenous DNMTs. mRNA expression of each DNMT isoform was assessed by real time RT-PCR with primers designed for the IRES region located on a bicistronic transcript and normalized to the expression of GAPDH. (C) Protein expression levels of exogenous DNMT isoforms by western blotting using a myc antibody. The expression level of DNMTs shown in the right panel was lower than those in the left panel, such that a 10-fold higher amount of protein was loaded in the left panel, as shown by the GAPDH control. Note the 10-fold higher amount of DNMT 3A1 in the left panel was compared with 3A1 (1/10) in the right panel.

Gene-specific DNA methylation changes induced by DNMT isoforms. (A) Illumina DNA methylation assay for gene-specific DNA methylation. DNA methylation profiles were one-dimensionally clustered by methylation level at 1505 loci based on the β-value. The β-value, a continuous measure of DNA methylation levels, is represented in colors ranging from green (β-value 0, unmethylated sites) to red (β-value 1, methylated sites) (30). Each column represents 1 of 1505 CpG sites, and each row represents the DNA methylation profile from a stably transfected cell line that expresses each of the different DNMTs indicated on the left. The relatedness of the DNA methylation patterns is represented by a tree (top), with the lengths of the branches representing the degree of similarity based on the average linkage and Euclidean distance. 293T/M.SssI and 293T/WGA (whole genome amplification) were used as methylated and unmethylated controls. (B) Validation of the Illumina DNA methylation assay using bisulfite PCR pyrosequencing. Absolute DNA methylation at the EYA4_P794_F and HOXA5_P479_F loci was directly measured by bisulfite PCR pyrosequencing and compared with the percent methylation extrapolated from the β-value in Illumina DNA methylation assay. Two independent stable transfection experiments are indicated as cell line 1 and cell line 2.

Histone modifications of DNMT3A1- and DNMT3B1-specific target genes. (A) The Venn diagram depicts the number of specific and overlapping target gene promoter regions between DNMT3A1 and DNMT3B1. DNMT3A1 and DNMT3B1 have 195 overlapping target genes. DNMT3A1 has 71 specific target genes, and DNMT3B1 has 78 specific target genes. (B) The association between DNMT target genes and histone modifications. Cross-application analysis between DNA methylation measurement and ChIP-on-chip analysis of histone modifications (H3K4me3 and H3K27me3) showed that specific target genes of DNMT3A1 are associated with H3K4me3 modifications, whereas the specific target genes of DNMT3B1 are associated with H3K27me3. The total represents the percentage of histone modifications for all 808 genes. Overlapping target genes and specific target genes of DNMTs are shown Figure 5A. The association between de novo methylated sites and histone marks was calculated using a Chi-squared test. (C) Validation of the association between DNA methylation and histone modifications. Chromatin from HEK 293T cells was immunoprecipitated using antibodies against H3K4me3, H3K27me3 or H3K9me3 and analyzed by real-time PCR for 20 loci that showed the greatest increase in de novo DNA methylation changes specific for DNMT3A1 (left 10 loci) and DNMT3B1 (right 10 loci). The increase in DNA methylation was calculated by the difference in the β-value between mock vector-transfected cell lines and DNMT3A1- or DNMT3B1-transfected cell lines. ChIP-on-chip histone data are shown as K4: H3K4me3, K27: H3K27me3, Bi: H3K4me3/H3K27me3, and N: neither H3K4me3 nor H3K27me3. ChIP data represent the means ± SD. of three independent experiments. The white bar represents the IgG control, and the black bar represents enrichment of histone modifications. All data were converted to relative values based on a standard curve and normalized to 1% input values. Real-time PCR primers for 20 loci were designed within ±500 bp at the CpG sites analyzed in the Illumina DNA methylation assay, except the HS3ST2 genes (+1359 bp). See details for the primers in Supplementary Table S1.

Changes in DNA methylation target sites of DNMT isoforms in the absence of specific DNMT domains. DNA methylated loci were compared between DNMT isoforms that were identical, with the exception that they lacked specific domains. (A) DNMT3B exon10, (B) DNMT3B PWWP, (C) N-terminus of DNMT3A or DNMT3B and (D) both DNMT3B PWWP and the N-terminus. The association between de novo methylated loci and histone marks was calculated using a Chi-squared test; *P < 0.05 and **P < 0.01. All genes are listed in Supplementary Table S6.