Identification of tissue-restricted TF motifs in human developmental enhancers.

A. Human embryo showing tissue/organs used in this study. Samples collected ranged from Carnegie Stage 14 to 21. Black lines connect each organ/tissue to its corresponding number of H3K27ac-positive bins; 1kb bins replicated only in one tissue were used as putative tissue-specific enhancers as described in (20). Each tissue is labelled with the same colour code across the manuscript. B. HOMER known motif enrichment analysis of heart ventricle-specific H3K27ac bins. In this representative example, the top 10 motifs of the list are shown. For analysis, the top 20 known enriched motifs in each set of tissue-specific H3K27ac bins were used. C. Heatmap showing patterns of expression of tissue-restricted TFs across embryonic tissues. All tissues were duplicated except for pituitary, muscle and skin. Raw counts were down sampled based on the 75th percentile and only genes with total expression value over 75 were considered. Expression of each TF was divided by its median expression across all tissues and log median-normalized expression was used to plot the heatmap. TFs whose expression remained constant across tissues were removed. The heatmap of all TFs present in the list obtained from (5) and represented in the RNA-seq data from is shown in figure S1A. D. Heatmap showing the expression of ventricle-restricted TFs. TFs with expression higher than approximately 150 times their median expression in one tissue, were considered restricted to that tissue. Those ventricle-restricted TFs, whose recognition motifs are within the top 20 motif enrichment analysis results shown in 1B were considered ‘First Search TFs’. GATA4, HAND1, MYOG and TBX20 are ventricle-restricted TFs that bind a motif contained in the top 20 enriched motif in ventricle-specific enhancers (1B), namely GATA6, HAND2, AP4 and TBX20 motifs. E. Pipeline to identify ‘Second Search’ TFs co-occurring with ‘First Search’ TFs. Motif enrichment analysis was performed using ‘First Search’ TF motif coordinates within tissue-specific regulatory regions as foreground, and ‘First Search’ TF motif coordinates within random regions as background. Motif enrichment analysis was performed 100nt either side of the ‘First Search’ TF’s motif to identify TF motifs that co-occur with the ‘First Search’ TF motif in tissue-specific regulatory regions. Co-occurring motifs are labelled as ‘Second Search’ TFs.

Global network of TF co-binding at tissue-specific developmental enhancers.

Network of co-occurring TF motifs at active tissue-specific enhancers in human embryonic development. A cut-off of p-value < 1×10-5 and presence in > 5% of foreground regions was used in HOMER known motif enrichment analysis results. On the y-axis are the ‘First Search’ TFs. For each ‘First Search’ TF, its corresponding motif cluster is shown on the left. On the x-axis are TF motif clusters enriched within 100nt either side of the ‘First Search’ TFs (‘Second Search’ TF), clustered by motif similarity (as seen in Fig S1C and table S3). Colours correspond to the ones assigned in Fig 1B. The size of the bubbles matches the negative log p-value: larger bubbles represent smaller p-values.

Motifs belonging to ubiquitous families of TFs co-occur with tissue-restricted TF motifs.

A. Heatmap showing number of co-occurring events for each ‘Second Search’ TF cluster. Clustering by number of co-occurrences assigns ‘Second Search’ clusters to three groups: high co-occurring, medium co-occurring and low co-occurring. bHLH clades A and B have the highest number of co-occurring events. ‘Second Search’ clusters of motifs recognised by ubiquitous TFs are abundant (found in the high/medium co-occurring groups). For instance, TEAD cluster presents high levels of co-occurring events, whilst TALE and ETS are found in the medium co-occurring cluster. B. Heatmap showing patterns of expression of TFs belonging to TEAD, STAT, ETS and TALE TF families. All tissues were duplicated except for pituitary, muscle and skin. Raw counts were down sampled based on the 75th percentile. Expression of each TF was divided by its median expression across all tissues and log median-normalized expression was used to plot the heatmap. Expression of ubiquitous TFs belonging to the TEAD, TALE, STAT and ETS families do not deviate from their median expression across tissues (with the exception of STAT4, ETV2, ELF3, ERG, ELF5 and EHF). C. ‘First Search’ TFs (y-axis) co-occurring with TEAD binding sites in the tissues shown (x-axis). Colours correspond to the ones assigned in Fig 1B. The size of the bubbles matches the negative log p-value: larger bubbles represented smaller p-values.

TEAD1 binding occurs mainly at tissue-specific intronic or distal intergenic regions.

A. Intersection of adult mouse liver, embryonic mouse ventricle and human primary keratinocyte TEAD1 ChIP-seq shows a high percentage of tissue-specific binding. TEAD HOMER de novo motifs enriched in liver, ventricle and keratinocyte are shown. Each motif was used to centre the corresponding ChIP-seq 200nt regions around the peak summits (referred to as ‘TEAD1 peak’ in the result section). B. Annotation of TEAD1 specific peaks in mouse liver, ventricle and primary keratinocytes and overlapping TEAD1 peaks in all three ChIP-seq experiments. Non-tissue-specific peaks contain a higher percentage of promoters compared to tissue-or cell-type specific TEAD1 binding events. C-F. Top five mouse phenotype and biological process GOs associated with ventricle-specific (C), liver-specific (D), keratinocyte-specific (E) and non-tissue-specific (F) TEAD1 peaks, using the whole mouse genome as background.

TEAD1 binds regions co-occupied by GATA4 with CHD4 during heart development.

A. Overlap between TEAD1 and H3K27ac ChIP-seq peaks in the mouse embryonic heart. B. Sequence motif enrichment in TEAD1 peaks overlapping with H3K27ac (red and grey) and without H3K27ac (red only). The left bar chart shows the percentage of the target regions with motif; the right bar chart shows the neg log p-values of the motif, as identified by HOMER de novo motif enrichment analysis. Cardiac ‘First Search’ TFs and other known cardiac TFs are shown. Motifs bound by tissue restricted TFs HAND, BHLH (Ebox), TBOX and GATA are more enriched in non-H3K27ac TEAD peaks, whilst TF motifs MEF2 and MEIS are preferentially enriched in TEAD1-H3K27ac regions. C. Overlap between TEAD1 and CHD4 ChIP-seq peaks in the embryonic heart ventricle. Top 5 motifs, identified by HOMER de novo motif analysis, are shown for TEAD1 peaks with and without CHD4. GATA motifs are exclusively enriched in TEAD1 peaks overlapping CHD4 peaks. D. Overlap between GATA4, TEAD1 and CHD4 ChIP-seq peaks in the mouse embryonic heart. E. Percentage of TEAD1 peaks overlapping GATA4 peaks. GATA4 preferentially occupies TEAD1-CHD4 regions compared to TEAD1-only regions. Statistical significance was calculated using a two-proportion z test. F. Top ten biological process GOs associated with TEAD1, GATA4 and CHD4 co-occupied regions are related to cardiac cell differentiation. G. Top ten biological process GOs associated with TEAD1-GATA4 co-occupied regions (without CHD4) are involved with distinct biological processes, including vascular development.

TEAD1 binding attenuates enhancer activation by cardiac-specific TFs.

A. Workflow to identify high-confident candidate ventricle enhancers bound by TEAD1 and GATA4/GATA6. Mouse regions co-occupied by TEAD1, GATA4 and GATA6 were converted to human (hg38) coordinates. B. Luciferase activity driven by three ventricle-specific enhancers (enh_PDLIM5, enh_NKX2.6, enh_RAMP1) co-transfected with TEAD1 alone (T1), Gata6 alone (G6), YAP alone, Gata6 and TEAD1 (G6 + T1), TEAD1 and YAP (T1 + YAP), G6 and YAP, and G6, T1 and YAP in NIH3T3 cells, normalised against basal enhancer activity driven by an empty pcDNA3 vector (pcD). Co-expression of GATA6, YAP and TEAD1 significantly reduces enhancer activity. C. Luciferase activity driven by enh_MEIS2, enh_RAMP1 and enh_PDLIM5 co-transfected with GATA6 in NIH3T3 cells and treated with verteporfin (VP) or DMSO, normalised against basal enhancer level driven by an empty pcDNA3 vector (pcD). Inhibiting formation of the TEAD/YAP complex increases GATA6-dependent enhancer activation. Values are presented as the mean +/-SEM of two biological replicates, each obtained from the median of three experimental replicates. Asterisk denotes significant results. Significance was calculated using an unpaired t-test, * = P < 0.05. D. Diagram of the wild-type and mutant versions of enh_NKX2.6 in pGL4 plasmid (WT, G6-, T1– and G6-/T1-). E. Luciferase activity driven by wild-type enh_NKX2.6 and enh_NKX2.6 mutants. Each enh_NKX2.6 plasmids was co-transfected with TEAD1 alone (T1), with Gata6 alone (G6) and with Gata6 and TEAD1 (G6 + T1) in NIH3T3 cells. Luciferase values normalised against basal enhancer activity driven by an empty pcDNA3 vector (pcD). Compared to the wild-type enhancer, TEAD1 failed to significantly affect Gata6-induced enhancer activity upon mutation of TEAD1 binding site. Values (in B, E) are presented as the mean +/-SEM of three biological replicates, each obtained from the median of three experimental replicates. Asterisk denotes significant results. Significance was calculated using a one-way ANOVA with multiple comparisons for each of the enhancer constructs, **** = P < 0.0001, *** = P < 0.001, ** = P < 0.01, * = P < 0.05, ns = P > 0.05. H. Enrichment of CHD4 on_RAMP1 enhancer in the presence of GATA6 alone or GATA6, TEAD1 and YAP in NIH3T3 cells. All values are normalised against CHD4 enrichment on RAMP1 enhancer. Values are presented as the mean +/-SEM of two biological replicates, each obtained from the median of three experimental replicates. Significance was calculated using a one-way ANOVA with multiple comparisons for each of the enhancer constructs, **** = P < 0.0001, *** = P < 0.001, ** = P < 0.01, * = P < 0.05, ns = P > 0.05.

TEAD1 binding attenuates enhancer activation by RPE-specific TFs.

A. Workflow followed to identify high-confident candidate RPE enhancers bound by TEAD1 and CRX. Mouse regions occupied by CRX were converted to human (hg38) coordinates. B. Luciferase activity driven by three RPE-specific enhancers (enh_EYA2, enh_PCMTD2 and enh_PRDM16) co-transfected with TEAD1 alone (T1), CRX alone, YAP alone, CRX and TEAD1 (CRX + T1), TEAD1 and YAP (T1 + YAP), CRX and YAP, and CRX, TEAD1 and YAP (CRX + T1 + YAP) in NIH3T3 cells, normalised against basal enhancer activity driven by an empty pcDNA3 vector (pcD). Co-expression of CRX, YAP and TEAD1 reduces enhancer activity to almost pcDNA3 levels. C. Percentage of CRX-induced luciferase reporter activity when CRX was overexpressed with TEAD1 compared to CRX overexpression alone. Percentages were calculated using the mean of three biological replicates. D. Diagram of the wild-type and mutant versions of enh_EYA2 pGL4 plasmid constructs (WT, CRX-, T1– and CRX-/T1-). E. Luciferase activity driven by wild-type enh_EYA2 and three enh_EYA2 mutants: T1-, CRX/T1– and CRX. Each of the four _EYA2 plasmids was co-transfected with TEAD1 alone (T1), with Crx alone (CRX) and with CRX and TEAD1 (CRX + T1) in NIH3T3 cells, normalised against basal enhancer activity driven by an empty pcDNA3 vector (pcD). When TEAD1 binding site is mutated, there is no significant inhibition of Crx-induced enhancer activity. Values (in B, E) are presented as the mean +/-SEM of 3 biological replicates, each obtained from the median of 3 experimental replicates. Asterisk denotes significant results. Significance was calculated using a one-way ANOVA with Tukey test for multiple comparisons of the means calculated for each enhancer plasmid (significance results are shown in black). In addition, unpaired t-tests were used to compare two conditions across enhancers (significance results shown in red/cyan), **** = P < 0.0001, ** = P < 0.01, * = P < 0.05, ns = P > 0.05. F. Working model. In progenitor cells, TEAD alongside YAP binds tissue-specific enhancers nearby tissue-specific activators and attenuates enhancer activity, reducing the expression of tissue-specific genes and slowing down cell differentiation. We propose that attenuation of enhancer activity is due to recruitment of CHD4 by the TEAD/YAP complex and tissue-specific activators. Reduced activity of TEAD in differentiated cells, possibly due to either TEAD no longer binding to the enhancer, or to its inability to recruit CHD4, results in full enhancer activation and sustained expression of tissue-specific genes. This mechanism could be part of a wider role of TEAD/YAP to slow down differentiation and maintain appropriate pools of progenitor-like cells to achieve correct organ size during organogenesis.