Activin/nodal signaling and NANOG orchestrate human embryonic stem cell fate decisions by controlling the H3K4me3 chromatin mark

Abstract

Stem cells can self-renew and differentiate into multiple cell types. These characteristics are maintained by the combination of specific signaling pathways and transcription factors that cooperate to establish a unique epigenetic state. Despite the broad interest of these mechanisms, the precise molecular controls by which extracellular signals organize epigenetic marks to confer multipotency remain to be uncovered. Here, we use human embryonic stem cells (hESCs) to show that the Activin-SMAD2/3 signaling pathway cooperates with the core pluripotency factor NANOG to recruit the DPY30-COMPASS histone modifiers onto key developmental genes. Functional studies demonstrate the importance of these interactions for correct histone 3 Lys4 trimethylation and also self-renewal and differentiation. Finally, genetic studies in mice show that Dpy30 is also necessary to maintain pluripotency in the pregastrulation embryo, thereby confirming the existence of similar regulations in vivo during early embryonic development. Our results reveal the mechanisms by which extracellular factors coordinate chromatin status and cell fate decisions in hESCs.

Keywords: Activin/Nodal; DPY30; H3K4me3; NANOG; SMAD2/3; hESCs.

Figure 1.

Activin/Nodal regulates H3K4me3 of a subset of genes in hESCs. (A) Schematic of the experimental approach. (B) ChIP-qPCR for SMAD2/3 on its binding sites associated to the indicated genes (see also E) before and after inhibition of Activin/Nodal with SB for 2 h. (C) Results of the statistical analysis of ChIP-seq data. (D) Annotation of H3K4me3 peaks to genomic features. (E) ChIP-seq results for H3K4me3 and H3K27me3 on selected SMAD2/3 target genes before and after inhibition of Activin/Nodal with SB for 2 h. Lines represent read enrichments normalized by million mapped reads and the size of the library. SMAD2/3-binding sites in hESCs are reported (Brown et al. 2011). (F) Colocalization of H3K27me3 peaks with H3K4me3 peaks centered in a range of ±5 kb from the closest transcription start site (promoters). (G) Colocalization of H3K4me1 and H3K27ac peaks with H3K4me3 peaks centered outside a range of ±5 kb from the closest transcription start site (intergenic).

Figure 2.

Dynamics of the transcriptional response to Activin/Nodal inhibition and their relationship with epigenetic changes. (A) Schematics of the experimental approach. (B) Euclidean hierarchical clustering of differentially expressed microarray probes across a time course of Activin/Nodal inhibition in hESCs (top 10% ranked by Hotelling T² statistic). Z-scores indicate the differential expression measured in number of standard deviations from the average level across all the time points. The three major probe clusters are indicated. (C) Expression profiles of probes in the clusters indicated in B. Selected results of gene enrichment analysis and representative genes for each cluster are reported (Supplemental Table S2 contains the complete set of results). (D) As in B, but only selected representative genes that showed decreased H3K4me3 upon 2 h of inhibition of Activin/Nodal signaling are reported. (E) Gene set enrichment analysis (GSEA) for genes whose expression was decreased after 48 h of SB in the list of H3K4me3-associated genes ranked by the differential H3K4me3 enrichment before or after 2 h of SB. (F) Expression of the genes closest to the H3K4me3 peaks decreased after 2 h of SB. The blue lines indicate no expression change. The significance of expression differences after 2 h or 48 h of SB treatment versus Activin as calculated by Dunn's multiple comparisons tests is shown (see Supplemental Fig. S2B for results for other time points).

Figure 3.

DPY30 is required for H3K4me3 and expression of SMAD2/3 target genes. (A) Western blots of SMAD2/3 or control (IgG) immunoprecipitations (IP) from nuclear extracts of hESCs. Input is 5% of the material used for immunoprecipitation. (B) Western blots in stable DPY30 knockdown (KD) hESC lines or controls (cells expressing a scramble [SCR] shRNA). (C) Phase-contrast images of the same DPY30 knockdown or control hESC colonies after the indicated number of days from the cell split (day 0). Bars, 200 μm. (D) Gene expression qPCR in DPY30 knockdown or control hESCs. Note that SOX2 is both a pluripotency and a neuroectoderm marker. For each gene, significant differences versus both SCR sh1 and SCR sh2 (only the highest P-value is shown) as calculated by one-way ANOVA are reported. (E) Immunofluorescences for the indicated proteins (green) or nuclear staining (DAPI, blue) in DPY30 knockdown or control hESCs. Bars, 100 μm. (F) Rank–rank hypergeometric overlap analysis (RRHO) for genes ranked by their differential expression after DPY30 knockdown or inhibition of Activin/Nodal for 48 h with SB. Color-coded log₁₀ P-values indicate the significance of the overlap between genes in the two conditions, considering hypergeometric tests. (G) Heat map showing changes in gene expression of selected SMAD2/3 target genes after DPY30 knockdown or 48 h of SB. Z-scores were separately calculated for each experiment. (H) ChIP-qPCR for H3K4me3 on SMAD2/3 target genes in hESCs transiently transfected for 48 h with a DPY30 shRNA or a control shRNA. For each gene, significant differences versus SCR as calculated by t-test are reported.

Figure 4.

DPY30 is required for mesendoderm differentiation of hESCs. (A) Hematoxylin and eosin histological staining of normal testis tissue (sham control) or teratomas derived from DPY30 knockdown (KD) or control hESCs. (B) Gene expression qPCR in DPY30 knockdown or control hESCs (cells expressing a SCR shRNA) before (PLURI) or after in vitro directed differentiation toward endoderm or neuroectoderm. For each gene, significant differences versus both SCR sh1 and SCR sh2 in the same condition (only the highest P-value is shown) as calculated by two-way ANOVA are reported. (C) Gene expression qPCR as in A but either before (pluripotent [PL]) or after liver (LIV) or pancreas (PA) differentiation. For each gene, significant differences versus SCR in the same condition as calculated by two-way ANOVA are reported. (D) ChIP-qPCR for H3K4me3 in DPY30 knockdown or control hESCs before (PLURI) or after endoderm differentiation. Significant differences versus SCR sh1 in the same condition as calculated by two-way ANOVA are reported.

Figure 5.

SMAD2/3 and NANOG recruit DPY30 onto their genomic targets. (A) RRHO analysis for genes ranked by their differential expression after NANOG knockdown (KD), DPY30 knockdown, or inhibition of Activin/Nodal for 48 h with SB. (B) Western blots of DPY30 or control (IgG) immunoprecipitations (IP) from nuclear extracts of hESCs. Input is 5% of the material used for immunoprecipitation. (C) ChIP-qPCR for H3K4me3 in NANOG knockdown or control hESCs. For each gene, significant differences versus SCR as calculated by t-test are reported. (D) Sequential ChIP-qPCR for DPY30 followed by SMAD2/3, NANOG, or control (IgG) ChIP. (E) ChIP-qPCR for the indicated proteins in NANOG knockdown or control hESCs. (F,G) ChIP-qPCR for the indicated proteins in hESCs before and after inhibition of Activin/Nodal with SB for 2 h. ChIP-qPCR results in D–G are representative of three independent experiments, and the location of primers used is shown in Supplemental Figure S5G.

Figure 6.

SMAD2/3, NANOG, and DPY30 control H3K4me3 on a subset of shared genes. (A) Overlap between H3K4me3 peaks significantly down-regulated after DPY30 knockdown (KD) and NANOG knockdown. (B) The proportion of H3K4me3 peaks significantly down-regulated after 2 h of Activin/Nodal inhibition with SB that were also similarly affected by DPY30 knockdown, NANOG knockdown, or both treatments. Representative genes associated to peaks down-regulated in all conditions are reported. (C) Examples of ChIP-seq results for H3K4me3 on selected SMAD2/3 target genes before or after DPY30 knockdown, NANOG knockdown, or SB treatment for 2 h. Lines represent read enrichment normalized by million mapped reads and the size of the library. (D) Expression of the genes closest to H3K4me3 peaks significantly down-regulated by >50% after both DPY30 knockdown and NANOG knockdown but not decreased to the same extent after 2 h of SB. The significance of expression differences versus Activin as calculated by Welch's t-test is reported. (E) Average normalized H3K4me3 read enrichment in three biological replicates before or after 2 h of SB. Data refer only to H3K4me3 peaks described in D. The level of significant change versus Activin as calculated using Welch's t-test is reported. (F) ChIP-qPCR for H3K4me3 before and after inhibition SB for 2, 4, 8, or 48 h. The genes analyzed belong to the group described in D. For each gene, significant changes versus Activin as calculated by one-way ANOVA are reported. (G) Examples of ChIP-seq coverage (top) and peaks (bottom) for SMAD2/3, NANOG, and DPY30 in hESCs. (H) Heat maps of coverage for SMAD2/3, NANOG, and DPY30 ChIP-seq relative to the SMAD2/3 peaks located 100 kb upstream of/downstream from H3K4me3 regions decreased after 2 h of SB (530 peaks). (I) Mean coverage plots for the peaks considered in H.

Figure 7.

Dpy30 is required for mouse post-implantation embryonic development. (A) Genotyping results from Dpy30 mutant heterozygous crosses. Stages are embryonic (E) or postnatal (P) days. (B) Bright-field images of wild-type (+/+) or Dpy30 knockout (−/−) embryos at E6.5 or E7.5. The anterior–posterior (A/P) and proximal–distal (PR/D) axes are shown. Bars, 100 μm. (C) Gene expression qPCR in E6.5 embryos from Dpy30 mutant heterozygous crosses. Note that Sox2 is both an epiblast and a neuroectoderm marker. Significant differences versus Dpy30^+/+ as calculated by one-way ANOVA are reported. n = 6 for Dpy30^+/+ and Dpy30^−/−; n = 11 for Dpy30^+/−. (D) Schematic of the model that we propose for Activin/Nodal-dependent transcriptional regulation of pluripotency and mesendoderm genes in hESCs and mouse epiblast cells. Note that the interactions depicted in the model must be interpreted as functional ones rather than direct protein–protein interactions, as this aspect was not the focus of the present study.