Mapping the Global Chromatin Connectivity Network for Sox2 Function in Neural Stem Cell Maintenance

Abstract

The SOX2 transcription factor is critical for neural stem cell (NSC) maintenance and brain development. Through chromatin immunoprecipitation (ChIP) and chromatin interaction analysis (ChIA-PET), we determined genome-wide SOX2-bound regions and Pol II-mediated long-range chromatin interactions in brain-derived NSCs. SOX2-bound DNA was highly enriched in distal chromatin regions interacting with promoters and carrying epigenetic enhancer marks. Sox2 deletion caused widespread reduction of Pol II-mediated long-range interactions and decreased gene expression. Genes showing reduced expression in Sox2-deleted cells were significantly enriched in interactions between promoters and SOX2-bound distal enhancers. Expression of one such gene, Suppressor of Cytokine Signaling 3 (Socs3), rescued the self-renewal defect of Sox2-ablated NSCs. Our work identifies SOX2 as a major regulator of gene expression through connections to the enhancer network in NSCs. Through the definition of such a connectivity network, our study shows the way to the identification of genes and enhancers involved in NSC maintenance and neurodevelopmental disorders.

Keywords: ChIA-PET; SOX2; chromatin connectivity; neural stem cells; transcription factors.

Graphical abstract

Figure 1

Sox2 Ablation Causes Major Loss of Long-Range Interactions in Brain-Derived NSCs (A) Functional genomics analyses. (B) Top: “anchors” and “nodes” connected by long-range interactions; bottom: numbers of promoter/non-promoter nodes in WT and MUT NSCs, left: TR1 and right: TR2 and TR3 combined. (C–G) Connectivity diagrams in WT NSCs (WT interactions; red) and MUT NSCs (MUT interactions; blue), across 5 different chromosome regions, in the wTR1, wTR2, and mTR1, mTR2 analysis; regions coordinates are: chr8:87120161-87587163 (C), chr13:25372775-31004673 (D), chr11:117736788-117873172 (E), chr12:56459922-56634834 (F) and chr8:48254658-48486144 (G). Their genomic coordinates are indicated above each panel, and genes within each region shown below the panels. Pol II- and SOX2-binding peaks are shown. PET counts (Y axes); note different (log10) scales in some panels. In MUT NSCs, an overall decrease of “looping” is seen, but some interactions are lost, others are maintained. Note the persistence of Pol II binding in MUT NSCs and the frequent coincidence (in C) of SOX2 peaks with interaction anchors. See also Figures S1, S2, S3 and Table 1.

Figure 2

SOX2-Bound Regions Carrying Epigenetic Enhancer Marks (EM) Show Significantly Higher Overlap with Anchors Than SOX2-Negative EM-Positive Regions (A) Left: number of SOX2-bound sites in regions linked to annotated TSS (±1,000 nt) and in distal, non-P regions. Right (histograms): percentage of different enhancer marks (EMs)-positive regions within SOX2-bound TSS-linked (SOX2⁺ TSS) or distal (SOX2⁺ distal) regions (left histograms, peak calling; right histograms, chromHMM). (B) Fraction of SOX2-bound sites within EM-positive regions (H3K27Ac⁺) on TSS-linked or distal regions (peak calling). (C) Interaction types according to the nature of the connected regions, for wTR1, wTR2, wTR3, mTR1, mTR2, and mTR3. “Prom,” annotated TSS-containing region (i.e., promoter). (D) Numbers of P-P and P-nonP (P-E) SOX2-positive interactions in WT cells in wTR1, wTR2, and wTR3. See also Table S4. (E) Fraction of SOX2⁺ (left) versus SOX2⁻ (right) EM-positive regions that overlap with anchors in wTR1, wTR2, and wTR3. Top: distal epigenetically marked regions (H3K27Ac⁺ and H3K4me1⁺) overlap with distal anchors. Bottom: all epigenetically marked regions (H3K27Ac⁺ or H3K4me1⁺) overlap with all anchor types, chromHMM. See also Figure S4 and Tables S4 and S5.

Figure 3

Distal Anchors Connected by Sox2-Dependent Interactions to Genes Important in Neural Development and Disease (A and B) Sox2-dependent ChIA-PET interactions (TR1) between two different genes (Sox4, A; Zbtb18, B) and distal regions overlapping previously characterized “VISTA” enhancers (Visel et al., 2009); SOX2 ChIP-seq peaks (present paper), lacZ-stained transgenic embryos (from https://enhancer.lbl.gov), and evolutionary conservation (ECR browser). (C and D) Sox2-dependent interactions (wTR1, wTR2, wTR3, mTR1, mTR2, and mTR3) involving Gpr56 (C) and Arid1a (D), two genes whose human homologs are involved in neurodevelopmental brain disease. See also Figure S5 and Tables S5 and S6.

Figure 4

Distal Anchors in Sox2-Dependent Interactions Drive GFP Transgene Activity to Zebrafish Brain Top: enhancer-dependent GFP-reporter (ZED): distal anchors (DA) from Sox2-dependent interactions are cloned upstream to a minimal promoter and GFP. Bottom: first and second left columns, GFP expression in transgenic embryos and bright-field images (F1 of stable lines, except for c-fos, transient transgenics). Third column: expression (in situ hybridization from http://zfin.org) of the endogenous zebrafish gene corresponding to the gene connected, in mouse, to the tested anchor. Fourth column: forebrain lacZ staining driven by transgenes carrying the human enhancers corresponding to the anchor (from https://enhancer.lbl.gov) (Visel et al., 2009). See also Figure S6 and Table S6.

Figure 5

Reduced Gene Expression in Sox2 MUT NSCs Correlates with Loss of Long-Range Interactions (A) Distribution of expression values (TPM) of genes with TPM >0. Blue, WT NSCs; orange, MUT NSCs. From left to right: all genes; genes whose promoter is a node (P-P, P-E interactions); genes whose promoter is connected to an enhancer (P-E interactions); genes with SOX2-positive P-E interactions. (B) Distribution of expression values (y axis) of genes according to the number and type of element (enhancer or promoter anchors) interacting with the gene promoter (x axis) in wTR1, wTR2, and wTR3. Top: interactions with enhancers. Bottom: interactions only with promoters. The number of genes involved is shown in each diagram inside the box along the x axis. (C) Distribution of the fold ratio values for all genes with TPM >0 defined as log2 (TPM_WT/TPM_MUT). It confirms results shown in (B): the fold ratio is shifted toward positive values (i.e., a majority of genes have expression in WT higher than in MUT NSCs). (D) Coassociation scores (histograms, -log₁₀p) calculated for genes (TPM >5) showing either reduced gene expression in MUT NSCs (red), or no relevant change (green), and the indicated categories of interactions detected in wTR1, wTR2, and wTR3 (see STAR Methods). DOWN_MUT, genes showing statistically significant expression reduction; DOWN_MUT NO_SIGN, genes showing moderate expression reduction; OTHERS, all other genes. The types of interactions are shown on the right; data for “SOX2 promoter target” are not significant. See also Tables S5 and S8.

Figure 6

SOCS3 Re-expression in MUT NSCs Prevents NSC Exhaustion and Restores Self-Renewal (A) Top: Socs3 gene. ChIA-PET interactions, SOX2 peaks, and ChIA-PET reads in WT NSCs. Bottom: loss of interactions in Sox2-MUT NSCs. (B) Growth curves of MUT NSCs, not transduced (MUT) or transduced (MUT Socs3) with a Socs3-GFP-expressing lentivirus, and of WT controls (WT or WT Socs3). (C) Images (phase-contrast) of MUT or Socs3-transduced MUT NSCs 3 days after passage 12; neurospheres develop only from Socs3-transduced NSCs. For comparison, WT NSCs. (D) FACS analysis (GFP) of MUT, WT, and MUT Socs3 cells at the indicated passage number. With passaging, the fraction of GFP⁺ NSCs progressively increases in MUT Socs3 NSCs, eventually reaching 100%.