Combinatorial binding in human and mouse embryonic stem cells identifies conserved enhancers active in early embryonic development

Abstract

Transcription factors are proteins that regulate gene expression by binding to cis-regulatory sequences such as promoters and enhancers. In embryonic stem (ES) cells, binding of the transcription factors OCT4, SOX2 and NANOG is essential to maintain the capacity of the cells to differentiate into any cell type of the developing embryo. It is known that transcription factors interact to regulate gene expression. In this study we show that combinatorial binding is strongly associated with co-localization of the transcriptional co-activator Mediator, H3K27ac and increased expression of nearby genes in embryonic stem cells. We observe that the same loci bound by Oct4, Nanog and Sox2 in ES cells frequently drive expression in early embryonic development. Comparison of mouse and human ES cells shows that less than 5% of individual binding events for OCT4, SOX2 and NANOG are shared between species. In contrast, about 15% of combinatorial binding events and even between 53% and 63% of combinatorial binding events at enhancers active in early development are conserved. Our analysis suggests that the combination of OCT4, SOX2 and NANOG binding is critical for transcription in ES cells and likely plays an important role for embryogenesis by binding at conserved early developmental enhancers. Our data suggests that the fast evolutionary rewiring of regulatory networks mainly affects individual binding events, whereas "gene regulatory hotspots" which are bound by multiple factors and active in multiple tissues throughout early development are under stronger evolutionary constraints.

Figure 1. Overview of genome-wide binding data in human and mouse embryonic stem cells and embryonal carcinoma cells.

Shown is the locus of the SOX2 gene in the human genome (top), along with mapped reads for OCT4, SOX2, NANOG and p300. Individual experiments are shown separately. The orthologous locus in the mouse genome is aligned at the bottom along with mapped reads from the individual experiments.

Figure 2. Co-localization within the genome identifies known protein interaction.

(A) Clustering of genome-wide binding profiles from mES cells based on the number of shared binding events identifies three main classes: Enhancer binding (Oct4, Sox2, Nanog), Insulator binding (CTCF, Smc1, Smc3) and Mediator associated binding (Med1, Med12, Nipbl). (B) Protein interaction network inferred from genome-wide binding data. Edges represent the pairwise similarities with a z-score above a threshold. (C–D) The number of overlapping experiments is much higher than expected by chance, both for mouse binding data (mm9, D) and human binding data together with the aligned mouse data (hg19, C). Randomized data sets show only very few cases where more than five experiments overlap (black line). The data used in this study shows a much stronger overlap with many loci where binding was detected in more than five experiments (red line).

Figure 3. Mediator co-localizes with Oct4, Sox2 and Nanog at combinatorially bound enhancers.

(A) Bars indicate the fraction of loci where Med1, Med12 and CTCF binding can be observed, depending on the combination of Oct4, Sox2 and Nanog, indicated by boxes below. Dark boxes indicate binding, white boxes indicate no binding (“AND” relation), light grey boxes with “v” indicate binding of at least one factor (“OR” relation). Both Med1 and Med12 preferentially co-localize at loci bound by Oct4, Sox2 and Nanog simultaneously. Combinatorial binding is more sensitive than a stringent control of false positives: The 10% most significant peaks are significantly associated with Med1 and Med12, however, the overall fraction is much lower compared to combinatorially bound loci. CTCF serves as a control to estimate unspecific binding. (B) The majority of loci bound by Oct4, Sox2 and Nanog are more than 1000 bp away from the nearest transcription start sites for all possible combinations (indicated by boxes above). Mediator co-localization mainly occurs at distant regulatory sites, showing that the increased overlap of Med1/Med12 at combinatorially bound loci is not caused by promoter specific co-localization.

Figure 4. Average H3K27ac ChIP-Seq signal in mES cells around combinatorially bound loci.

Loci bound by Oct4, Sox2 and Nanog together with Mediator are enriched in H3K27ac, a mark associated with active enhancers (black line). In contrast, loci without Mediator co-localization show a much weaker enrichment (red line) suggesting that Mediator associates with active enhancers.

Figure 5. Gene Set Enrichment Analysis (GSEA) of genes near combinatorial binding events.

(A) Expression of genes in mES cells (V6.5) and differentiated cells after 14 days (14d), sorted by the signal-to-noise ratio obtained from the GSEA software . (B) The random walk that describes the gene set enrichment over genes sorted by their rank according to signal-to-noise ratio (similar sorting as in (A)). Set 1 (Oct4, Sox2, Nanog and Med1/Med12 in blue) is enriched in genes active in mES cells (enrichment score 0.43, p-value = 0.0), set 2 (Oct4, Sox2, Nanog without Med1/Med12 in yellow) is enriched in genes active in differentiated cells (enrichment score −0.3, p-value = 0.05). Combinatorial binding of Oct4, Sox2 and Nanog identifies active and poised enhancers; Mediator is associated with active gene expression.

Figure 6. The combination of OCT4, SOX2 and NANOG influences conservation of binding events.

(A) Bars indicate the fraction of loci where binding of Nanog, Sox2, Oct4 or CTCF can be observed at the orthologous locus in mouse ES cells for all combinations of OCT4, SOX2 and NANOG in human ES cells as indicated by the boxes below. Dark boxes indicate binding, white boxes indicate no binding (“AND” relation), light grey boxes with “v” indicate binding of at least one factor (“OR” relation). Combinatorial binding of OCT4, SOX2 and NANOG shows the largest fraction of conserved binding for Oct4, Sox2 and Nanog in mouse. Again, combinatorial binding is more sensitive than a stringent control of false positives, as is estimated by conservation at 10% most significant peaks. (B) The fractions of binding combinations in mES cells at conserved loci (for all combinations of binding in human cells as indicated by the boxes above). Combinatorial binding of Oct4, Sox2 and Nanog in mES cells is much higher at combinatorially bound loci in human, suggesting that combinatorial binding is conserved in evolution. (C) The fraction of proximal and distant binding sites for conserved and non-conserved binding events, split up according to the combinations of binding as indicated by the boxes above. The majority of conserved binding events are distant regulatory elements. Conserved binding events are more frequently in the proximal promoter than non-conserved binding events.

Figure 7. Average Fibroblast ChIP-Seq signal profile around loci bound by OCT4, SOX2 or NANOG in hES cells.

Enhancers bound by OCT4, SOX2 or NANOG which are active in mouse development (red line) are enriched in H3K27ac in human fibroblast cells (IMR90) supporting that many of these enhancers are developmentally active in human as well.

Figure 8. Binding conservation in embryonic stem cells is increased at developmental enhancers.

(A) Bars indicate the fraction of loci where binding of Nanog, Sox2, Oct4 and CTCF can be observed at the orthologous locus in mouse ES cells for all combinations of OCT4, SOX2 and NANOG in human ES cells discriminated by developmental activity as indicated by the boxes below. Dark boxes indicate “AND” relation, light grey boxes with “v” indicate “OR” relation, “?” indicates no restriction. Combinatorial binding events at developmentally active enhancers show the highest levels of binding conservation between mouse and human ES cells (>50%). (B) The fractions of binding combinations in mES cells at conserved loci (for all combinations indicated by the boxes above). The majority of conserved binding events at developmentally active enhancers where OCT4, SOX2 and NANOG bind simultaneously show combinatorial binding of Oct4, Sox2 and Nanog in mouse ES cells. (C) The fraction of proximal and distant binding sites for conserved and non-conserved binding events (split up according to the combinations of binding as indicated by the boxes above). The majority of conserved binding events are distant regulatory elements.

Figure 9. Model for “gene regulatory hotspots”.

(A) Enhancers are bound by OCT4, SOX2 and NANOG together with p300 in embryonic stem cells. These enhancers maintain pluripotency by activating gene expression in ES cells (top) or poisoning expression for activation after differentiation (bottom) (B) After differentiation of the cell, the same enhancers are bound by p300 in developmental tissues together with other transcription factors. The target gene is expressed. We propose that enhancers which recruit multiple transcription factors in different stages of development are gene regulatory hotspots which are crucial to connect the regulatory networks of pluripotency and development. These enhancers show higher sequence conservation compared to individual, isolated binding events which active in single cell types.