Combinatorial transcription factor binding encodes cis-regulatory wiring of forebrain GABAergic neurogenesis

Abstract

Transcription factors (TFs) bind combinatorially to genomic cis-regulatory elements (cREs), orchestrating transcription programs. While studies of chromatin state and chromosomal interactions have revealed dynamic neurodevelopmental cRE landscapes, parallel understanding of the underlying TF binding lags. To elucidate the combinatorial TF-cRE interactions driving mouse basal ganglia development, we integrated ChIP-seq for twelve TFs, H3K4me3-associated enhancer-promoter interactions, chromatin and transcriptional state, and transgenic enhancer assays. We identified TF-cREs modules with distinct chromatin features and enhancer activity that have complementary roles driving GABAergic neurogenesis and suppressing other developmental fates. While the majority of distal cREs were bound by one or two TFs, a small proportion were extensively bound, and these enhancers also exhibited exceptional evolutionary conservation, motif density, and complex chromosomal interactions. Our results provide new insights into how modules of combinatorial TF-cRE interactions activate and repress developmental expression programs and demonstrate the value of TF binding data in modeling gene regulatory wiring.

Figure 1 –. TF Binding Profiles and Basic Genomic Features.

(A) Schematics of the scope of the present study, showing 3D structures combinatorially bound by TFs in transcriptionally active chromatin (H3K4me3-marked). The model was validated by enhancer transgenic mouse assays. (B) Distribution of combined bound loci between distal and proximal regions, segmented by the number of TFs sharing locus position. (C) Pie chart showing the distribution of loci in relation to the loops formed by H3K4me3-mediated PLAC-seq contacts (PSC). (D) Sp9 locus showing TF binding and PLAC-seq interactions with VISTA enhancers hs242, hs243, hs244, hs245, hs574, hs860, and hs953. PLAC-seq contacts are displayed as arcs and contact maps (adj. p < 0.01). (E) Doughnut plot showing individual TF number of binding loci, split into proximal and distal from gene TSS, with the associated core motifs and their average distribution around peak centers. In parentheses are the enrichments over background and percent of target motifs. See also Figure S1.

Figure 2 –. Organization of Bound Loci into Clusters of Similar Binding Neighborhood Profiles.

(A) Schematics showing the several binding patterns captured by investigating local neighborhood around ChIP-seq peak summits, as well the subsequent clustering and following genomic profile characterization. (B) and (C) Heatmaps representing each TF coverage around 1 kb of each called peak in proximal and distal regions, respectively. Row blocks and columns depict TFs and clusters, respectively. Within each row block, each line represents the coverage color-codes for intensity of ChIP-seq signal (intensity grows in the black-to-yellow-to-red direction). Within each cluster, lines with same position across the TFs represent the same genomic locus. Numbers (n = ) indicate the number of peaks called within each cluster. (D) Distribution of mean number of TFs bound to each locus across clusters. Mean distributions were calculated by sampling without replacement (N=1000). Random means was calculated by randomly sampling the genome (N=27398). (E) Distribution of mean number of TFs sharing loci across binding clusters, and compared to a random sample, calculated by sampling without replacement (N=1000). (F) Heatmap showing the relative enrichment of core binding motifs for each of the TFs across clusters. See also Figure S2.

Figure 3 –. Genomic and Functional Features of TF-Bound RE Clusters.

(A) Frequency of occurrence of peaks across chromatin states and binding clusters, split into distal and proximal. As reference, in between the two heatmaps is one derived from assigning chromatin states to a random loci sample. Color codes represent the percentage of peaks by cluster. (B) Gene ontology analysis of genes hitting interaction contact points (PSC) split into clusters, displaying select brain-specific or general terms. (C) Bar plot depicting the intersection of our binding clustering with clusters determined by single-nucleus ATAC-seq, showing putative neuronal cell differentiation states across binding clusters. (D) Dot plot showing the mean ensemble size across binding clusters in function of the percent of loci in the cluster colocalizing with PSCs. (E) Distribution of means of PhastCons scores across clusters for 60 vertebrates. Top panels are distal peaks, and proximal ones are on the bottom panels. References were random distal and proximal genomic regions of random widths. See also Figure S3.

Figure 4 –. VISTA and Novel Enhancer Activity across Bound Loci.

(A) Heatmap of combinatorial TF binding (percentage enrichment, 0-12 TFs) on VISTA enhancers that have subpallial (SP), pallial and subpallial (SP+P), pallial (P), non-telencephalic (non-tel), and no activity (inactive). (B) Stacked bar plot showing the percentage of newly identified regulatory sequences with high (8-12 TFs), intermediate (5-7 TFs), low (3-4 TFs), and very low (1 TF) binding in BG showing spatial regional activity. (C) Stacked bar plot depicting percentage of novel enhancers across binding clusters showing restricted spatial enhancer activity in the subpallium, pallium and shared among them. (D) Six pREs representing different clusters that were tested for activity in transgenic mouse assays. Clusters are shown in the left column, enhancer names are written in turquoise, and the success ratios are listed next to the name (i.e., 6/6 depicts 6 embryos with forebrain activity out of 6 embryos tested). Schemas predict the regulated genes by the tested enhancers (turquoise). The grey arrow depicts the orientation of the TSS. Green bars show the normalized binding of BG TFs, with color intensity proportional to ChIP-seq intensity. The specific TFs bound are shown above the top bar. Wholemounts (WM) and three sections representing the LacZ expression are shown. H3K27ac (green) and H3K27me3 (red) histone ChIP-seq results from the GE are shown to the right; the turquoise bars correspond to the tested genomic regions. Cx: Cortex; GE: Ganglionic Eminences; L: LGE; M: MGE; C: CGE. (E) Coronal brain section schematization showing: 1. the subregions of the primordial BG (LGE and MGE) as well as the cortex (top left hemisection) ; 2. the subregional laminae of the GEs (VZ, SVZ, and MZ; top right hemisection). Hemisections from 2 VISTA enhancers with specific sub regional activity are shown below with hs1056 showing activity in the VZ and SVZ of the MGE (bottom left hemisection) and hs566 showing activity in the mantle zones of the MGE and LGE (bottom right hemisection). (F) Bar plot depicting cluster classification of enhancers with VZ and non-VZ activity (n=99). See also Figure S4.

Figure 5 –. Arrayed TAATTA motifs anchor deeply-conserved GABAergic enhancers.

(A) Relative number of TAATTA motifs within each RE across clusters separated by symmetric, degenerate, and complex instances. (B) Histogram showing distribution of distance in base pairs between all motif pair occurrences within REs for selected clusters and motif pairs. (C) Average base-level sequence conservation (vertebrate PhyloP score) for TAATTA motif and flanking DNA for 1_D (top) and 16_D (bottom left) within 10bp of motif, and for 1_D out to 400bp of motif (bottom right). (D) TAATTA motifs exhibit the strongest base-level conservation across TF motif families enriched in 1_D REs. Enriched motifs with significant base-level conservation increase compared to 10bp flanking sequence labeled, primary motifs from BG TFs in bold. (E) Four representative 1_D REs with enhancer activity. Target gene, evolutionary conservation, BG TF binding, and enhancer activity in E12.5 mouse telencephalon. (F) Motif and evolutionary conservation landscape for enhancers in (d) showing motif clustering and overlap with conserved regions across core 500bp (top) and at single-base resolution (bottom) for selected intervals. Legend shows colors for BG TF primary motifs and all Homer motifs. See also Figure S5.

Figure 6 –. Pbx1 Genomic Locus and Associated Putative Enhancers.

(A) Representation of the Pbx1 locus, showing nearby genes, and the ensemble of PLAC-seq contacts creating a tridimensional structure made of multiple loops. Bound genomic loci are noted underneath in black, and enhancers are marked in turquoise and red (active and inactive, respectively). (B) Six enhancers around the Pbx1 locus exhibited activity in transgenic mouse assays. Of these, 1_D and 2_D enhancers were active in subpallium, while one 9_D enhancer was active in in pallium and the other 9_D and a 4_D enhancer showed non-telencephalic activity.

Figure 7 –. Cis-trans interactions underlying gene regulation driving GABAergic neurogenesis.

(A) Chromatin accessibility maps identify pREs, but TF binding is necessary to understand mechanisms and functional relevance of pRE activity. TF binding can direct either activation or repression of enhancer activity. Here we identify pRE-TF modules that drive specific regulatory activity in developing mouse BG, with representative examples depicted. Bold BG modules in (A) are highlighted in (B). (B) Combinatorial TF binding defines context-dependent patterns of enhancer activation and repression in embryonic BG. Three example cis-trans modules identified here are shown, with the enhancer activity and schematic of activity across VZ, SVZ, and MZ. (C) Developmental TF genes (i.e. Pbx1) relevant to embryonic BG have complex cis-regulatory landscapes and generally include multiple cis-trans regulatory modules. (D) Comparison of enhancers with simple versus complex TF binding identified in embryonic BG. Enhancers with exceptional TF binding also feature high density of TF binding motifs, complex chromosomal contacts, strong evolutionary conservation across the vertebrate tree (human, chicken, zebrafish conservation represented), and increased base pair size. Abbreviations: Imm. CIN: immature cortical interneurons, Imm. PN: immature projection neurons, BG NPC: basal ganglia neural progenitor cell, BG IPC: basal ganglia intermediate progenitor cell.