A screening system to identify transcription factors that induce binding site-directed DNA demethylation

Abstract

Background: DNA methylation is a fundamental epigenetic modification that is involved in many biological systems such as differentiation and disease. We and others recently showed that some transcription factors (TFs) are involved in the site-specific determination of DNA demethylation in a binding site-directed manner, although the reports of such TFs are limited.

Results: Here, we develop a screening system to identify TFs that induce binding site-directed DNA methylation changes. The system involves the ectopic expression of target TFs in model cells followed by DNA methylome analysis and overrepresentation analysis of the corresponding TF binding motif at differentially methylated regions. It successfully identified binding site-directed demethylation of SPI1, which is known to promote DNA demethylation in a binding site-directed manner. We extended our screening system to 15 master TFs involved in cellular differentiation and identified eight novel binding site-directed DNA demethylation-inducing TFs (RUNX3, GATA2, CEBPB, MAFB, NR4A2, MYOD1, CEBPA, and TBX5). Gene ontology and tissue enrichment analysis revealed that these TFs demethylate genomic regions associated with corresponding biological roles. We also describe the characteristics of binding site-directed DNA demethylation induced by these TFs, including the targeting of highly methylated CpGs, local DNA demethylation, and the overlap of demethylated regions between TFs of the same family.

Conclusions: Our results show the usefulness of the developed screening system for the identification of TFs that induce DNA demethylation in a site-directed manner.

Keywords: Binding site; DNA demethylation; Transcription factor.

Fig. 1

Feasibility of the approach to identify TFs that induce binding site-directed DNA methylation. a Flowchart of the approach. The TF of interest is overexpressed in a lentivirus vector that co-expresses a puromycin-resistant marker in 293T cells. Puromycin selection is applied for 7 days; then, TF-overexpressing 293T cells are subjected to DNA methylation analysis to identify differentially methylated CpGs (DMCs). Corresponding TFBMs of overexpressed TF are identified and then assessed to determine whether they are overrepresented at DMC regions. b Scatter plot showing DMCs caused by SPI1 overexpression in 293T cells. X- and Y-axes show M-values for 293T-mock and SPI1-overexpressing 293T (293T-SPI1) cells, respectively. Dashed lines represent ΔM borders of > 2. Green, purple, and gray dots represent significantly methylated, demethylated, and insignificant probes, respectively. Numbers of DMCs are shown upper left (methylated) and bottom right (demethylated). c Distribution of enrichment scores for SPI1 binding motifs within ± 5000 bp of demethylated CpGs (left) and methylated CpGs (right) in SPI1-overexpressing 293T cells. X- and Y-axes show distance from probe CpG position and enrichment score, respectively. Horizontal line represents enrichment score = 0

Fig. 2

Identification of DNA-demethylating TFs. a Number of differentially methylated CpGs by TF overexpression. X- and Y-axes show overexpressed TFs and number of differentially methylated CpGs, respectively. Purple and green bars represent demethylated CpGs and methylated CpGs, respectively. b Distribution of enrichment score for TFBMs within ± 5000 bp of demethylated CpG probes in TF-overexpressing 293T cells. X- and Y-axes show distance from probe CpG position and enrichment score, respectively. Horizontal lines are enrichment score = 0. c Boxplot showing ratio of number of methylated and demethylated CpGs for DNA-demethylating TFs (demethyl TFs) and non-DNA-demethylating TFs (non-demethyl TFs). Medians are indicated by central black horizontal lines, upper quartiles are indicated by upper edges of the box, and lower quartiles are indicated by lower edges of the box. Maximum and minimum values are marked as lines extending from the boxes. The p value is shown above the plots (Wilcoxon rank sum test)

Fig. 3

Gene ontology and specifically expressing tissue enrichment analysis. Bar plots showing enrichment of gene ontology (black labels) and specifically expressing tissues (red labels). X-axis represents −log₁₀ P of Fisher’s exact test. Specifically expressed tissues were analyzed using Affymetrix GNF_U133A tissue expression data

Fig. 4

Original DNA methylation level of TF-mediated DNA demethylation targets. Violin plots showing kernel density distribution plot of original M-values for TF-mediated DNA demethylation targets. M-value distribution of all probes, of demethylated probes by DNA-demethylating TFs, and of demethylated probes by non-DNA-demethylating TFs is shown as magenta, cyan, and green, respectively. Interquartile ranges are shown as boxes in the overlaid boxplots (gray), and medians are shown as horizontal bars. Maximums and minimums are the upper and lower ends of the black line. Outliers (> Q3 + interquartile range × 1.5) are shown as gray dots

Fig. 5

Local effect of TF-mediated DNA demethylation. a DNA methylation patterns of demethylated regions, selected from methylation array results, were analyzed by bisulfite-PCR sequencing. Each circle represents a cytosine of CpG. Black and white circles indicate methylated and unmethylated cytosine, respectively. Horizontal lines represent the sequencing result of each sub-clone. Arrows represent the position of demethylated CpGs identified by methylation array analysis (*p < 0.05; **p < 0.01; Fisher’s exact test). b Gaussian distribution model fitting of enrichment score peaks. X-axis shows the distance from demethylated CpGs. Y-axis represents the enrichment score (left) and probability of the Gaussian distribution model (right). Gray plots are enrichment scores from Figs. 1c and 2b, and red dotted plots are fitted Gaussian distributions

Fig. 6

Family DNA-demethylating TFs share demethylation targets. a Position weight matrix-based sequence logos of TFBMs for RUNX1 (top left), RUNX3 (bottom left), CEBPA (top right), and CEBPB (bottom right). The height of each letter represents the probability of TF appearance at binding sites. b Overlap of demethylated CpGs between RUNX1-overexpressing (red) and RUNX3-overexpressing (blue) cells and between CEBPA-overexpressing (red) and CEBPB-overexpressing (blue) cells. The number of overlapping CpGs is shown at the intersection of each circle. The total number of demethylated CpGs is depicted as a particular circle size and is shown above the circles