FGF signaling inhibition in ESCs drives rapid genome-wide demethylation to the epigenetic ground state of pluripotency

Abstract

Genome-wide erasure of DNA methylation takes place in primordial germ cells (PGCs) and early embryos and is linked with pluripotency. Inhibition of Erk1/2 and Gsk3β signaling in mouse embryonic stem cells (ESCs) by small-molecule inhibitors (called 2i) has recently been shown to induce hypomethylation. We show by whole-genome bisulphite sequencing that 2i induces rapid and genome-wide demethylation on a scale and pattern similar to that in migratory PGCs and early embryos. Major satellites, intracisternal A particles (IAPs), and imprinted genes remain relatively resistant to erasure. Demethylation involves oxidation of 5-methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC), impaired maintenance of 5mC and 5hmC, and repression of the de novo methyltransferases (Dnmt3a and Dnmt3b) and Dnmt3L. We identify a Prdm14- and Nanog-binding cis-acting regulatory region in Dnmt3b that is highly responsive to signaling. These insights provide a framework for understanding how signaling pathways regulate reprogramming to an epigenetic ground state of pluripotency.

Figure 1

Erk1/2 and Gsk3β Signal Inhibition Induces Global DNA Demethylation (A) Schematic of the signaling pathways inhibited by the “2i” small molecule inhibitors. (B) Global CpG methylation measured by whole-genome BS-seq in serum, 2i, and E9.5 PGCs (data from Seisenberger et al., 2012). Error bars represent the standard deviation in three replicates. (C) Immunofluorescence staining of E14 ESCs with an antibody against 5mC shows reduced euchromatic methylation in 2i while pericentromeric heterochromatic regions maintain high 5mC levels. (D) Example of BS-seq profile in serum (black bars) and 2i (purple bars) ESCs with the Nanog locus being strongly demethylated in 2i while the ICR methylation at Kcnq1ot1 is maintained. (E) Heatmap methylation levels in 500 randomly selected elements (CpG islands (CGIs), Exons, Introns, LINE1, SINE) in Day 0, Day 24 Serum, and Day 24 2i. (F) Confirming IF data, methylation at pericentromeric major satellites remains high in 2i as measured by BS-seq (error bars represent the standard deviation between CpGs). (G) Heatmap methylation levels in 500 randomly selected IAP elements and 15 ICRs. See also Figure S1.

Figure 2

Signaling and Transcriptional Regulation in 2i ESCs (A) Comparison of promoter methylation levels with corresponding gene expression shows that in serum ESCs highly methylated genes are mostly underexpressed (blue group of gene promoters, blue histogram). In 2i (lower panel), there is no global change in the distribution of gene expression (see histograms on the right hand side). (B) Substantial downregulation of Dnmt3a, Dnmt3b, and Dnmt3L in ESCs cultured in 2i for 24 days as measured by RNA-seq (error bars represent the range of values in two biological replicates). Transcription of maintenance methylation components Dnmt1 and Uhrf1 is not reduced, while Tet2 and Prdm14 expression are upregulated in 2i ESCs. (C) Timecourse of downregulation of Dnmt3b in 2i versus serum ESCs. Note the rapid downregulation of cMyc, a direct target of the Erk1/2 signaling pathway (error bars represent the range of values in two biological replicates). (D) DNMT3B protein levels at different time points after 2i addition with visible decrease at 24 hr after 2i. (E) Dnmt3b upstream promoter genomic region shows PRDM14 and NANOG binding peaks overlapping putative enhancer elements. Shaded horizontal bars indicate the constructs used in the luciferase assay in (F). (F) Luciferase reporter assay with no promoter (pGL3-Basic) or the Dnmt3b promoter p3b −8615/+93 showing strong inhibition of luciferase expression in 2i. Deletion of Prdm14 and Nanog binding sites (p3b −8615/+93Δ) reduces repression in 2i by 50% (error bars represent the standard deviation of four independent transfections for each plasmid in each condition, except the p3b−500/+500 with two independent transfections).

Figure 3

Demethylation in 2i Involves Hydroxylation (A) Mass spectrometric measurement of global 5mC and 5hmC levels in E14 (Day 24 Serum and 2i) and Tet1wt ESCs at 0, 24, and 72 hr after 2i addition showing a stepwise decline of 5mC and a substantial transient increase in 5hmC. (B) Absolute 5mC (blue) and 5hmC (red) levels in individual CpGs measured by GlucMS-qPCR validate the BS-seq data and show that demethylating CpGs are significantly hydroxymethylated after 2i addition with a peak at 72 hr. CpGs were selected according to the extent of demethylation: not demethylated (IAP1) and demethylated in 2i (D1, D3, and D5). (C) Hairpin bisulphite and oxidative bisulphite sequencing in LINE1Tf 5′UTR at 24 hr, 72 hr, and 7 days after 2i addition showing the extent of hemimethylation through early demethylation phase with a 5hmC peak at 72 hr in LINE1Tf, but not IAP long terminal repeat 1 (IAPLTR1). (D) GlucMS-qPCR in Tet2 KO with Tet1 knockdown ESCs shows increased 5mC levels 48 hr after 2i addition (boxplot, blue bars, paired t test on 13 demethylating targets) and virtually no 5hmC acquisition compared to the control (boxplot, red bars). Individual examples are shown to the left. (E) GlucMS-qPCR in Tet1wt versus Tet1^−/− and Tet2wt versus Tet2^−/− shows demethylation in some CpG sites but significantly delayed demethylation in others. (F) Gene expression measurement in NanogGFP sorted ESCs and similarities with 2i versus serum. (G) GlucMS-qPCR in NanogGFP sorted ESCs. Error bars represent the range of values in two biological replicates throughout. See also Figures S2–S4.

Figure 4

Global Hypomethylation in 2i ESCs Resembles that of ICM Cells and Migratory PGCs (A) Global methylation levels in ESCs, ICM cells, and PGCs from whole-genome BS-seq data and reduced representation BS-seq (RRBS) data sets (RRBS data are labeled with asterisk) (Seisenberger et al., 2012; Smith et al., 2012). (B) Number of highly methylated (75%–100%) or moderately methylated (50%–75%) CGIs of the same samples in (A). (C) Expression of de novo methyltransferases Dnmt3a, Dnmt3b, Dnmt3L, and Myc in PGC E11.5 compared to E6.5 epiblast cells and ICM versus ESCs (Tang et al., 2010; Seisenberger et al., 2012). (D) E3.5 blastocysts stained for NANOG (red), DNMT3B (green), and TET1 (white). Mouse E3.5 blastocysts were collected from CD1 × CD1 and (C57BL/6 × CBA/Ca) × 129/Sv crosses and were classified as early, mid, or late E3.5, depending on their size and extent of cavitation. Late blastocysts show higher DNMT3B expression as well as a greater proportion of ICM cells expressing both NANOG and DNMT3B and slightly fewer NANOG- and TET1-expressing cells (the number of Nanog-positive cells counted from the total number of embryos in parenthesis is shown below the barplot). RGB profiles for representative ICM (NANOG high) and outer trophectoderm (TE) cells are shown for each stage. (E) Model for 2i-mediated demethylation including a peak of oxidation and reduced efficiency of maintenance methylation and disruption of de novo methylation.