Brachyury and SMAD signalling collaboratively orchestrate distinct mesoderm and endoderm gene regulatory networks in differentiating human embryonic stem cells

Abstract

The transcription factor brachyury (T, BRA) is one of the first markers of gastrulation and lineage specification in vertebrates. Despite its wide use and importance in stem cell and developmental biology, its functional genomic targets in human cells are largely unknown. Here, we use differentiating human embryonic stem cells to study the role of BRA in activin A-induced endoderm and BMP4-induced mesoderm progenitors. We show that BRA has distinct genome-wide binding landscapes in these two cell populations, and that BRA interacts and collaborates with SMAD1 or SMAD2/3 signalling to regulate the expression of its target genes in a cell-specific manner. Importantly, by manipulating the levels of BRA in cells exposed to different signalling environments, we demonstrate that BRA is essential for mesoderm but not for endoderm formation. Together, our data illuminate the function of BRA in the context of human embryonic development and show that the regulatory role of BRA is context dependent. Our study reinforces the importance of analysing the functions of a transcription factor in different cellular and signalling environments.

Keywords: Embryonic stem cells; Gastrulation; Gene regulatory networks; Human; SMAD; T-BOX.

Fig. 1.

An in vitro differentiation system to study the role of BRA in human gastrulation. (A) hESCs differentiated in FLyA (blue) or FLyB (red) media for 36 h resemble the anterior (endoderm progenitors) or posterior (mesoderm progenitors) regions of the early primitive streak. (B) Western blots showing the expression of BRA, EOMES, phospho-SMAD1, total SMAD1, phospho-SMAD2/3, total SMAD2/3 and β-actin during pluripotency (FA), and in FLyA and FLyB conditions. (C) Flow cytometry analysis of hESCs differentiated in FLyA or FLyB media for 36 h. FLyA-treated cells were co-immunostained for BRA and EOMES (upper left), or for BRA and SOX17 (lower left). FLyB-treated cells were co-immunostained for BRA and EOMES (upper right), or for BRA and CDX2 (lower right).

Fig. 2.

BRA exhibits distinct genomic binding profiles in FLyA- and in FLyB-differentiated hESCs. (A) hESCs treated with FLyA (BRA^low) or FLyB (BRA^high) media for 36 h were used to analyse and compare the genome-wide binding of BRA (ChIP-seq). (B) Venn diagram showing the detectable overlap between BRA binding (ChIP-seq peaks) in FLyA- and FLyB-treated hESCs. (C) Dot plot of ChIP-seq fold enrichment values (normalised to Input samples) of common BRA peaks in FLyA- and FLyB-treated hESCs. R, correlation coefficient. (D) Examples of ChIP-seq peaks depicting BRA-binding profiles in hESCs treated with FLyA (blue track) or FLyB (red track) media: stronger peaks in FLyA (left); peaks detected in both FLyA and FLyB (centre); stronger peaks in FLyB (right). Tracks under ChIP-seq peaks: gene locus (exons depicted as full rectangles, introns depicted as lines with chevrons), DNase I-hypersensitive clusters (ENCODE project) and mammalian conservation profiles (UCSC genome browser). The y axis shows the number of normalised unique reads.

Fig. 3.

BRA has different sets of target genes in FLyA- and FLyB-treated hESCs. (A) Venn diagram showing the overlap of BRA putative target genes between FLyA-treated hESCs (blue), FLyB-treated hESCs (red) and WNT3A/activin- treated hESCs (grey; Tsankov et al., 2015). (B-D) Gene ontology analyses (GREAT algorithm; McLean et al., 2010) of BRA-binding regions detected only in FLyA (B), only in FLyB (C) and in both FLyA and FLyB (D). Ontology terms are ranked according to their enrichment P-values: ‘Gene family’ terms (P-value <1×10⁻⁵), all other terms (P-value <1×10⁻⁹). TS, Theiler stage of mouse development.

Fig. 4.

BRA in the context of activin A signalling. (A) Comparison of DNA recognition sites of five protein families (row above) and DNA motifs enriched at BRA FLyA ChIP-seq peaks (row below). (B) Co-immunoprecipitation of SMAD2/3 (pulldown) with BRA and EOMES (WB, western blot) in FLyA-treated hESCs; IgG (negative control immunoglobulin). (C,D) Histograms showing the distance between BRA-binding peaks in FLyA-treated hESCs and EOMES binding (Teo et al., 2011) or SMAD2/3 binding (Brown et al., 2011) in FLyAB-treated hESCs. (E) Venn diagram showing the overlap of putative target genes between BRA in FLyA-treated hESCs (FLyA, blue), EOMES (green) and SMAD2/3 (orange). (F) Wild-type (control) and BRA knockdown hESCs were differentiated for 36 h in FLyA and profiled for transcriptome-wide (microarray) differential expression analysis. (G) Venn diagram showing the overlap between BRA putative target genes (FLyA, dark blue) and genes that were either up- or downregulated (FDR <0.05) in BRA knockdown hESCs when compared with wild-type hESCs. (H) Microarray gene expression heat-map of wild-type versus BRA knockdown (KD) hESCs grown in FLyA for 36 h. Green indicates downregulation and red indicates upregulation. Symbols after gene names indicate expression pattern in vivo (Mouse Genome Informatics; Alev et al., 2010). (I) ChIP-seq peaks depicting BRA binding in hESCs treated with FLyA (blue) or FLyB (red), and EOMES binding (green) or SMAD2/3 binding (orange). Tracks under ChIP-seq peaks: gene locus (exons depicted as full rectangles, introns depicted as lines with chevrons), DNase I-hypersensitive clusters (ENCODE project) and mammalian conservation profiles (UCSC genome browser). The y axis shows the number of normalised unique reads. Blue boxes highlight FLyA-specific BRA binding peaks.

Fig. 5.

BRA in the context of BMP4 signalling. (A) Comparison of DNA recognition sites of five protein families (row above) and DNA motifs enriched in BRA FLyB ChIP-seq peaks (below). (B) Co-immunoprecipitation of SMAD1 (pulldown) with BRA and EOMES (western blot) in FLyB-treated hESCs; IgG (negative control immunoglobulin). (C) Histogram showing the distance between BRA-binding peaks in FLyB-treated hESCs and EOMES binding (Teo et al., 2011) in FLyAB-treated hESCs. (D) Venn diagram showing the overlap of putative target genes between BRA in FLyB-treated hESCs (FLyB, red) and EOMES (green). (E) Graph with fold enrichment values (ChIP over input) for EOMES binding (green), SMAD1 binding (yellow) and control IgG binding (grey) to BRA target regions in FLyB-treated hESCs (36 h). Error bars correspond to s.d. (n=3). ChIP-qPCR values were normalised to the highest control IgG value (PRDM14). (F) Wild-type (control) and BRA knockdown hESCs were differentiated for 36 h in FLyB and profiled for transcriptome-wide (microarray) differential expression analysis. (G) Venn diagram showing the overlap between BRA putative target genes (FLyB, red) and genes that were either up- or downregulated (FDR <0.05) in BRA knockdown hESCs when compared with wild-type hESCs. (H) Microarray gene expression heat-map of wild-type versus BRA knockdown (KD) hESCs grown in FLyB for 36 h. Green indicates downregulation and red indicates upregulation. Symbols after gene names indicate expression pattern in vivo (Mouse Genome Informatics; Alev et al., 2010). (I) ChIP-seq peaks depicting BRA binding in hESCs treated with FLyA (blue) or FLyB (red), and EOMES binding (green). Tracks under ChIP-seq peaks: gene locus (exons depicted as full rectangles, introns depicted as lines with chevrons), DNase I-hypersensitive clusters (ENCODE project), mammalian conservation profiles (UCSC genome browser). The y axis shows the number of normalised unique reads. Red boxes highlight FLyB-specific BRA binding peaks.

Fig. 6.

BRA depends on the signalling environment to regulate key developmental genes. (A,B) hESCs transfected with either an empty/control vector or a BRA over-expression vector were differentiated for 36 h, as indicated and samples were collected to perform comparative gene expression analysis by qRT- PCR. (C,D) qRT-PCR of BRA target genes expressed in endoderm/anterior primitive streak (blue) or in mesoderm/posterior primitive streak (red), respectively. hESCs transfected with either an empty vector (dark colours) or a BRA overexpression vector (light colours) were differentiated as indicated in A,B. For each group of germ layer markers, gene expression is presented as fold change over the wild type reference sample (control vector): FLyA for endoderm genes (C) and FLyB for mesoderm genes (D), as indicated on the y axis. Bars indicate s.d. (n=3) (Student's two-tailed t-test: *P<0.05; **P<0.01; n.s., not significant). F, FGF2; Ly, Ly294002; B, BMP4; A, activin A; N, noggin (N); S, SB431542.

Fig. 7.

BRA target gene expression in mouse embryos. (A) Venn diagram showing the overlap between BRA putative target genes identified in FLyA-treated hESCs (blue), FLyB-treated hESCs (red) and activin-treated mouse embryoid bodies (light brown, Lolas et al., 2014). (B) Venn diagrams showing the overlap between BRA putative target genes that are misregulated in FLyA- or FLyB-treated BRA knockdown hESCs at 36 h and genes misregulated in E7.5 Bra^−/− mouse embryos (Lolas et al., 2014). Left diagram, downregulated genes; right diagram, upregulated genes. (C) RNA-seq analysis of E7.5 wild-type and Bra^−/− mouse embryos (Lolas et al., 2014); coloured dots indicate BRA target genes identified only in human (FLyA and FLyB datasets, purple), only in mouse (Lolas et al., brown) or in both human and mouse (black). Scale represents log₂ FPKM (fragments per kilobase of exon per million fragments mapped). Only differentially expressed (fold change >2) genes are shown. (D) Confocal microscopy analysis (middle embryo stack) of mouse gastrulae (E7.0) immunostained for Bra (red), Foxa2 (green) and Cdx2 (blue). Nuclei were stained with DAPI. Upper row shows a wild-type embryo (Bra^+/+). Bottom rows show Bra mutant embryos (Bra^−/−). Spatial orientation of the embryos is shown in the lower right corner.

Fig. 8.

BRA, EOMES and SMAD signalling mediate mesoderm or endoderm cell fate choice during gastrulation. Simplified model of gene regulatory mechanisms operating in cells of the anterior (blue) or posterior (red) early primitive streak. Dashed arrows indicate activin A or NODAL as upstream activators of SMAD2/3, and BMP4 as the upstream activator of SMAD1. Crosses over white arrows indicate transcriptional silencing, whereas coloured arrows indicate transcriptional activation.