Pioneer factor ASCL1 cooperates with the mSWI/SNF complex at distal regulatory elements to regulate human neural differentiation

Abstract

Pioneer transcription factors are thought to play pivotal roles in developmental processes by binding nucleosomal DNA to activate gene expression, though mechanisms through which pioneer transcription factors remodel chromatin remain unclear. Here, using single-cell transcriptomics, we show that endogenous expression of neurogenic transcription factor ASCL1, considered a classical pioneer factor, defines a transient population of progenitors in human neural differentiation. Testing ASCL1's pioneer function using a knockout model to define the unbound state, we found that endogenous expression of ASCL1 drives progenitor differentiation by cis-regulation both as a classical pioneer factor and as a nonpioneer remodeler, where ASCL1 binds permissive chromatin to induce chromatin conformation changes. ASCL1 interacts with BAF SWI/SNF chromatin remodeling complexes, primarily at targets where it acts as a nonpioneer factor, and we provide evidence for codependent DNA binding and remodeling at a subset of ASCL1 and SWI/SNF cotargets. Our findings provide new insights into ASCL1 function regulating activation of long-range regulatory elements in human neurogenesis and uncover a novel mechanism of its chromatin remodeling function codependent on partner ATPase activity.

Keywords: ASCL1; ATAC-seq; ChIP-seq; chromatin regulation; mSWI/SNF; neural stem cell; neurogenesis; pioneer transcription factor; scRNA-seq.

Figure 1.

ASCL1 mRNA expression marks a transitional cell population bridging progenitors and postmitotic neurons. (A) qRT-PCR analysis of ASCL1 expression at multiple time points during neural differentiation of human iPSCs; mRNA expression relative to DIV0 (D0). ASCL1 shows a transient spike in expression at DIV24 (D24), following Notch inhibition (via DAPT addition, indicated by a red dashed line) at DIV23 (D23). Error bars represent mean ± SEM for three biological replicates. (B) Western blotting shows transient ASCL1 protein increase during neural differentiation. Notch inhibition is indicated by a red dashed line. CTNNB1 loading control is included. (C) Uniform manifold approximation and projection (UMAP) plot and unsupervised clustering of single-cell transcriptomes from 28,316 cells in DIV24 neural cultures treated with γ-secretase inhibitor DAPT, collected from three independent cultures. Dots represent single cells. Colors represent the different clusters identified at the right. (D) Clusters from C were grouped into three broad cellular state clusters based on gene expression of canonical markers (see F). A large cluster of VIM⁺NES⁺ cells uniquely enriched for ASCL1 and its transcription targets, lacking in neuronal markers and positioned between a cluster of cycling progenitors (CPs) and a cluster of neurons (Nrs), was termed “transitional progenitors” (TPs). (E) Predicted cell cycle phases of cells based on their expression of cell cycle-related genes. Transitional progenitors are found in the G1/G0 phase. (F) Dot plot representation of the expression of genes used for post-hoc annotation of unsupervised clusters (shown in C) to classify cell state identities represented in D. Dot size indicates proportion of cells in each cluster expressing a gene, and shading indicates the relative level of gene expression according to the key at the right. (G) Single-gene expression overlaid onto UMAP plot defined in C shows ASCL1 expression is enriched in transitional progenitors. (H) RNA velocity vectors projected onto the UMAP plot show differentiation directionality from cycling progenitors to neurons through transitional progenitors. A second direction was identified that reflected the cell cycle phases. (I) A partition-based graph abstraction (PAGA) velocity graph, with PAGA connectivities (dashed) and transitions (solid/arrows), shows a similar differentiation trajectory. The size of a node reflects the number of cells belonging to the corresponding cluster.

Figure 2.

ASCL1 protein expression also marks a transitional progenitor population poised to differentiate. (A) Immunofluorescence images of DIV24 neural cultures showing cells colabeled for ASCL1 and the progenitor marker PAX6 (top; e.g., white arrowheads), but not for ASCL1 and the neuronal marker CTIP2 (BCL11B; bottom). Scale bar, 50 μm. (Right) Quantification of PAX versus ASCL1 (top) and CTIP2 versus ASCL1 (bottom) nuclear immunofluorescence intensity is shown. Yellow indicates coexpression: 99.8% of ASCL1⁺ coexpress PAX6, and 3.6% coexpress CTIP2. (B) Flow cytometry profiles showing a contour plot of intensities of labeling for ASCL1 and SOX2 in single cells analyzed at DIV24 (left) and a dot plot equivalent to the contour plot at the left, pseudocolored according to the level of TUBB3 expression (right; key shown at the bottom right). Insets indicate ASCL1-high and ASCL1-low populations, respectively. (C) DNA content histogram profiles obtained by flow cytometry after DAPI staining for the ASCL1-high (left) and ASCL1-low (right) progenitor populations defined in B, indicating the cell cycle distribution of cells in these populations. (D) Quantification of normalized SOX2 intensities in ASCL1-high and ASCL1-low populations from B. Error bars represent mean ± SEM for three biological replicates. Unpaired t-test: (****) P < 0.0001. (E) Quantification of the proportion of cells in different cell cycle phases as a percentage of the total cell population (cell cycle analysis by flow cytometry, as analyzed in C), showing significant accumulation of cells in S phase in the ASCL1-low population and accumulation of cells in G1 and G2/M in the ASCL1-high population. Error bars represent mean ± SEM for three biological replicates. Unpaired t-test: (*) P < 0.05, (**) P < 0.01, (****) P < 0.0001. (F) PCW 16 fetal brain slice coimmunolabeled for ASCL1, PAX6, and CTIP2, showing colabeling of ASCL1 with progenitor marker PAX6 (e.g., white arrowheads) but not with deep-layer marker CTIP2 (BCL11B). Scale bar, 50 μm. (VZ) Ventricular zone, (SVZ) subventricular zone.

Figure 3.

ASCL1 binds to distal regulatory elements of many target genes involved in neural development. (A) Single-cell transcriptomes from 35,755 differentiating ASCL1 KO neural cells at DIV24, collected from three independent cultures, projected onto the wild-type cells UMAP embedding represented in Figure 1D. (B) Relative proportion of major cell clusters from A in comparison with the cell clusters in Figure 1D. Loss of ASCL1 results in significantly reduced proportions of transitional progenitors (TPs) and neurons (Nrs) and a significant increase in cycling progenitors (CPs). Unpaired t-test: (***) P < 0.001. (C) Dot plot representation of the expression of biologically relevant genes in the major clusters from Figure 1D, showing expression differences between WT and ASCL1 KO cells mapped to those clusters. Dot size indicates proportion of cells in cluster expressing a gene, and shading indicates the relative level of expression. (D) Heat maps representing ChIP-seq coverage for ASCL1 binding sites at genomic regions predicted to be regulatory elements by the ABC algorithm (based on their chromatin accessibility and H3K27ac signature) (Fulco et al. 2019). (p.c.) Peak center. (E) ASCL1-regulated genes; i.e., genes selected after enhancer–gene regulatory relationship was predicted with the ABC algorithm and where enhancers were identified to be bound by ASCL1 via ChIP-seq, and subsequently the regulated genes were found to be differentially expressed in bulk RNA-seq analysis of ASCL1 KO versus wild-type cells (Supplemental Fig. S2). Color coding indicates differential gene expression fold change in ASCL1 KO versus wild-type cultures. Illustrative genes for each category are highlighted (listed in Supplemental Table S1). (F) Representative Integrative Genomic Viewer (IGV) tracks of ChIP-seq, ATAC-seq, and bulk RNA-seq data illustrating examples of ASCL1-occupied active regulatory elements (as predicted by ABC algorithm) targeting genes up-regulated (left) or down-regulated (right) in ASCL1 KO versus control cultures at DIV24. Bar plots show mean expression in FPKM for the depicted genes in wild-type and ASCL1 KO cultures. (****) P-adj < 0.0001. (G) Graphical representation of log transformed P-value and gene number for top enriched GO biological process terms ([BP] biological process, [MP] molecular process) and reactome pathways for the genes down-regulated (top; blue) and up-regulated (bottom; red) in ASCL1 KO cells at DIV24, as displayed in E (complete list in Supplemetal File S3).

Figure 4.

ASCL1 regulates chromatin accessibility through different pioneer and nonpioneer functions. (A) Heat maps showing ASCL1 binding sites identified by ChIP-seq in open (top; n = 38,377) and closed (bottom; n = 17,723) chromatin, as determined by ATAC-seq in DIV24 neural cultures. (B) Heat maps showing open chromatin sites in DIV24 (from A) where ASCL1 regulates accessibility (i.e., “ASCL1-dependent” sites) by opening (top; n = 4288) or closing (middle; n = 4896) chromatin, and where ASCL1 binds without regulating accessibility (bottom; n = 29,220). (C) Heat maps showing the differential gene expression for ABC-predicted genes associated with regulatory elements at which ASCL1 promotes (top) or represses (bottom) chromatin accessibility (from B), and dysregulated in ASCL1 KO versus wild-type DIV24 cultures. Decreased accessibility is associated with decreased expression in 79.3% of ASCL1-regulated genes; conversely, 74.9% of ASCL1-regulated genes showed increased accessibility of regulatory elements associated with increased gene expression. Color coding indicates differential gene expression fold change versus wild-type culture. Illustrative genes for each category are highlighted at the right. (D) Heat maps showing the changes in chromatin accessibility between wild-type DIV24 cultures (right) and wild-type DIV20 cultures (left) and ASCL1 mutant DIV24 cultures (middle) at ASCL1-dependent sites (from B) where ASCL1 opens chromatin, acting as a classical pioneer transcription factor (top, n = 2155), and where its activity changes accessibility at permissive sites (bottom, n = 2133). (E, top) Representative IGV tracks flanking a putative ASCL1-dependent gene showing ChIP-seq and ATAC-seq profiles at DIV20 and DIV24 (from D). (Bottom) Bar plots show mean expression in FPKM for the depicted genes in wild-type and ASCL1 KO cultures. (****) P-adj < 0.0001.

Figure 5.

ASCL1 interacts with mSWI/SNF remodeling complexes predominantly at sites where it does not have classical pioneer activity. (A) Immunoprecipitation followed by Western blot analysis in DIV24 neural cultures showing reciprocal coimmunoprecipitation of ASCL1 and SMARCC1 or ARID1A. CTNNB1 loading control is included. (B,C) Representative immunofluorescence images of proximity ligation assay between ASCL1 and SMARCC2, SMARCB1, or ACTL6B in the human fetal cortex at PCW 16 (B) and in wild-type and ASCL1 KO DIV24 neural cultures (C). Cyan foci indicate PLA amplification signal. Nuclei are shown in magenta (DAPI). (C) Numbers of foci per nucleus were quantified in three nonoverlapping fields of view. Scale bars, 50 μm. The no antibody control is shown in Supplemental Figure S3E. Unpaired t-test: (****) P < 0.0001. (D) Heat maps of SMARCB1 binding from ChIP-seq analysis (right) at all ASCL1 binding sites (left; from Fig. 4A, redisplayed for comparison) found in open (top; n = 38,377) or closed (bottom; n = 17,723) chromatin (middle) in DIV24 neural cultures. Sites cobound by ASCL1 and SMARCB1 are mostly found in regions of open chromatin. (E,F) Heat map profiles of SMARCB1 binding (right) in DIV24 neural cultures at ASCL1-dependent sites (left; from Fig. 4, redisplayed for comparison) where ASCL1 acts as a classical pioneer factor binding closed chromatin (ATAC signal below threshold in ASCL1 KO cultures; E, middle right) and sites where ASCL1 binds permissive chromatin (nonpioneer chromatin remodeler; ATAC signal shows open signal in ASCL1 KO cultures, but ASCL1 causes changes in accessibility; F, middle right vs. middle left). (F, top and bottom) SMARCB1 is enriched at sites where ASCL1 binds open chromatin and changes accessibility.

Figure 6.

ASCL1 works in concert with mSWI/SNF complexes to remodel chromatin and regulate gene expression. (A,B) Heat maps showing the effect of BRM014 treatment on chromatin accessibility at ASCL1-bound sites where ASCL1 displays classical pioneer activity (A) or nonpioneer chromatin remodeling activity (B). (B) The ATPase activity of mSWI/SNF complexes is mainly required at sites of nonpioneer activity. ASCL1 ChIP-seq (Fig. 5E) is included for reference. (C, top) Heat map of differential gene expression for genes dysregulated in both ASCL1 KO and BRM014-treated cultures and whose ABC-predicted regulatory element(s) are associated with ASCL1–SMARCB1-cobound sites where both are required to increase accessibility and where ASCL1 acts as a classical pioneer factor. (Bottom) Heat map of differential gene expression for genes dysregulated in both ASCL1 KO and BRM014-treated cultures and whose ABC-predicted regulatory element(s) are associated with ASCL1–SMARCB1-cobound sites where both are required to increase accessibility and where ASCL1 acts as a nonpioneer chromatin remodeler. (D) Heat maps profiling ASCL1 and SMARCB1 binding at ASCL1–mSWI/SNF-dependent sites where the interaction is associated with open chromatin, showing that interfering with mSWI/SNF ATPase activity (BRM014 treatment) reduces ASCL1 binding and, reciprocally, that eliminating ASCL1 (ASCL1 KO) reduces SMARCB1 binding. (E) Diagram illustrating the proposed mechanisms for the ASCL1 transcription factor activity and ASCL1–mSWI/SNF recruitment dynamic. (Top) ASCL1 has nonpioneer chromatin remodeling activity, binding cooperatively with mSWI/SNF at sites of accessible chromatin. (Middle) ASCL1 has pioneer activity, binding cooperatively with mSWI/SNF at inaccessible regulatory elements, where they regulate accessibility. (Bottom) ASCL1 acts as a classical transcription factor (TF), binding accessible regulatory elements to regulate transcription of target genes. (Created with BioRender.com.)