The BAF chromatin remodeler synergizes with RNA polymerase II and transcription factors to evict nucleosomes

Abstract

Chromatin accessibility is a hallmark of active transcription and entails ATP-dependent nucleosome remodeling, which is carried out by complexes such as Brahma-associated factor (BAF). However, the mechanistic links between transcription, nucleosome remodeling and chromatin accessibility are unclear. Here, we used a chemical-genetic approach coupled with time-resolved chromatin profiling to dissect the interplay between RNA Polymerase II (RNAPII), BAF and DNA-sequence-specific transcription factors in mouse embryonic stem cells. We show that BAF dynamically unwraps and evicts nucleosomes at accessible chromatin regions, while RNAPII promoter-proximal pausing stabilizes BAF chromatin occupancy and enhances ATP-dependent nucleosome eviction by BAF. We find that although RNAPII and BAF dynamically probe both transcriptionally active and Polycomb-repressed genomic regions, pluripotency transcription factor chromatin binding confers locus specificity for productive chromatin remodeling and nucleosome eviction by BAF. Our study suggests a paradigm for how functional synergy between dynamically acting chromatin factors regulates locus-specific nucleosome organization and chromatin accessibility.

Fig. 1. RNAPII promoter-proximal pausing promotes BAF chromatin occupancy.

a, Heatmaps (bottom) and average plots (top) comparing chromatin accessibility assayed by S5P CUTAC, with RNAPII-S5P and BRG1 occupancy (CUT&Tag), relative to the primary peaks (summits) of S5P CUTAC and sorted by decreasing accessibility (CUTAC signal). b, Schematic showing distinct stages in RNAPII transcription that are inhibited by drugs used in this study. c,e,g, Violin plots of spike-in calibrated CUT&Tag signal distribution comparing RNAPII-S5P and BRG1 occupancy over S5P CUTAC peaks at time points after drug treatments for transcription inhibition: Triptolide (c), Flavopiridol (e) and Actinomycin D (g). Median values (solid lines), upper and lower quartiles (broken lines) and outliers were calculated using the Tukey method; n = 9,700. d,f,h, Fold changes in mean RNAPII-S5P and BRG1 occupancy (spike-in calibrated CUT&Tag) over S5P CUTAC peaks at time points after drug treatments: Triptolide (d), Flavopiridol (f) and Actinomycin D (h). All datasets are representative of at least two biological replicates. The RNAPII illustration was created with BioRender.com. Nuc, nucleosome; PIC, pre-initiation complex.

Fig. 2. Enrichment of BAF increases nucleosome eviction.

a, Comparison of chromatin structure and chromatin accessibility by means of S5P CUTAC fragment size distribution over peaks (promoter and enhancer NDR spaces) in cells treated with DMSO (control) and Flavopiridol. Peaks were called with DMSO control. b, Heatmaps comparing nucleosomal (≥150-bp reads) and subnucleosomal (≤120-bp reads) protection by BAF (BRG1 CUT&RUN) and BAF-associated histones (BRG1 CUT&RUN.ChIP) in untreated (DMSO control) cells. Heatmaps were plotted relative to S5P CUTAC summits and sorted by decreasing accessibility (CUTAC signal). CUT&RUN.ChIP heatmaps show enrichment over IgG isotype control (for ChIP). c,d, Enrichment of nucleosomal (≥150-bp, solid lines) and subnucleosomal (≤120-bp, broken lines) reads from BRG1 CUT&RUN and CUT&RUN.ChIP experiments, relative to S5P CUTAC summits, in DMSO- (c) or Flavopiridol-treated (d) cells. e, Flavopiridol treatment causes eviction of partially unwrapped nucleosomes through enrichment of BAF, leading to NDR persistence. All datasets are representative of at least two biological replicates. DMSO, dimethylsulfoxide.

Fig. 3. Nucleosome eviction by BAF is ATP-dependent.

a,b, Heatmaps (bottom) and average plots (top) comparing BRG1 occupancy by CUT&RUN upon 8 h of BRM014 treatment versus DMSO control (a), and dual inhibition with BRM014 plus Flavopiridol versus Flavopiridol only (b). Data are plotted relative to the primary peaks (summits) of S5P CUTAC and sorted by decreasing BRG1 CUT&RUN in DMSO. c,d, Enrichment of nucleosomal (≥150-bp, solid lines) and subnucleosomal (≤120-bp, broken lines) reads from BRG1 CUT&RUN.ChIP experiments relative to S5P CUTAC summits, in BRM014 over DMSO control (c) and upon dual inhibition with Flavopiridol and BRM014 over Flavopiridol only (d). Normalized counts of reads in DMSO were subtracted from BRM014 (c), and normalized counts of reads in Flavopiridol only were subtracted from the dual inhibitor treatment (d). All datasets are representative of two biological replicates. FLV, Flavopiridol.

Fig. 4. Transcription inhibition shows RNAPII and BAF occupancy at Polycomb (PRC2)-repressed gene promoters.

a, Representative genomic tracks comparing enrichment of histone PTMs, RNAPII-S5P and BRG1 by CUT&Tag at transcriptionally active and PRC2-repressed genes, and changes in RNAPII-S5P and BRG1 occupancy upon Flavopiridol (Flv.) and Actinomycin D (Act.) treatment. RNAPII-S5P and BRG1 CUT&Tag read counts were spike-in calibrated. b, Heatmaps (bottom) and average plots (top) comparing histone PTMs (CUT&Tag), chromatin structure (RNAPII-S5P CUTAC and ATAC-seq) and transcriptional activity (START-seq) at RNAPII-enriched (active) and H3K27me3-enriched (PRC2-repressed) promoters. Promoters were grouped based on k-means clustering of RNAPII-S5P and H3K27me3 CUT&Tag reads mapping to a 5-kb window around the TSSs of RefSeq-annotated mESC genes (Extended Data Fig. 4a). c,d, Violin plots of spike-in calibrated CUT&Tag signal distribution comparing RNAPII-S5P (c) and BRG1 (d) occupancy over PRC2-repressed promoter TSSs ± 1 kb at time points after drug treatments. e, Violin plot comparing spike-in calibrated H3K27me3 CUT&Tag at PRC2-repressed promoter TSSs ± 1 kb in cells treated with DMSO and Flavopiridol. Median values (solid lines), upper and lower quartiles (broken lines) and outliers were calculated using the Tukey method; n = 2,767. Numbers on top of the violin plots are mean values. f, S5P CUTAC fragment size distribution to compare chromatin accessibility at PRC2-repressed promoter TSSs ± 1 kb in cells treated with DMSO and Flavopiridol. All datasets are representative of at least two biological replicates. ATAC-seq, assay for transposase-accessible chromatin using sequencing; Mb, megabase.

Fig. 5. Upregulation of pluripotency TFs results in enhanced nucleosome eviction.

a, Heatmaps (bottom) and average plots (top) of pluripotency TF CUT&RUN reads relative to RNAPII-S5P CUTAC summits, showing TF binding at sites of DNA accessibility. Heatmaps were sorted by decreasing accessibility (CUTAC signal). b, Representative genomic tracks comparing occupancy of TFs (CUT&RUN), BRG1 (CUT&Tag) and RNAPII-S5P (CUT&Tag) in SL and 2i conditions. All datasets were spike-in calibrated. c, Violin plots of spike-in calibrated CUT&RUN (TF) and CUT&Tag (BRG1 and RNAPII-S5P) signal distribution comparing factor occupancy over RNAPII-S5P CUTAC peaks in 2i versus SL. Median values (solid lines), upper and lower quartiles (broken lines) and outliers were calculated using the Tukey method; n = 9,700. Numbers on top are mean values. d,e, Enrichment of nucleosomal (≥150-bp, solid lines) and subnucleosomal (≤120-bp, broken lines) reads from BRG1 CUT&RUN and CUT&RUN.ChIP experiments, relative to S5P CUTAC summits, in SL (d) and 2i (e). f, Heatmaps (bottom) and average plots (top) of RNAPII-S5P CUTAC (20–120-bp reads only) relative to S5P CUTAC summits, comparing chromatin accessibility in 2i versus SL. All datasets are representative of at least two biological replicates.

Fig. 6. RNAPII, BAF and DNA-sequence-specific TFs work synergistically in a dynamic cycle for productive chromatin remodeling and nucleosome eviction.

a, Model showing that RNAPII and BAF dynamically engage chromatin in an abortive manner (left-hand side) and require chromatin binding by DNA-sequence-specific TFs for productive chromatin remodeling and histone eviction to form/maintain an NDR (right-hand side). Relative thickness of arrows implies enrichment of factor-bound states in transcriptionally active chromatin as observed in steady-state bulk measurement. b, Steady-state promoter and enhancer chromatin structures can be explained by a dynamic cycle of nucleosome deposition and eviction and synergistic RNAPII, BAF and TF activity. We speculate that the cycle can start at any step: RNAPII loading at nucleosome-depleted regions (step 1) and transcription initiation (step 2), BAF binding to nucleosomes (step 3) or TF-binding nucleosomes that are partially unwrapped due to spontaneous thermal fluctuations in histone–DNA interactions or BAF binding and remodeling (step 4); and the cycle can continue as long as factor concentrations are high enough. The steps in the cycle facilitate each other, which we propose based on our observations that RNAPII promoter-proximal pausing promotes BAF occupancy and ATP-dependent nucleosome eviction, BAF is associated with partially unwrapped nucleosomes at pluripotency TF-binding sites and upregulated TF protein expression promotes nucleosome eviction by BAF leading to stable NDR formation, which facilitates new RNAPII loading. The RNAPII illustration was created with BioRender.com. Pi, inorganic phosphate.

Extended Data Fig. 1. CUT&Tag of chromatin epitopes and RNAPII-S5P CUTAC in mESCs.

a, Representative genomic tracks showing RNAPII, BRG1, histone PTM occupancy by CUT&Tag, chromatin accessibility (RNAPII-S5P CUTAC), and transcriptional activity (START-seq) at the Nanog promoter and enhancer cluster and flanking genes. Previously annotated enhancer regions are shown on top. b, c, Heatmaps (bottom) and average plots (top) comparing RNAPII, BRG1, and histone PTM occupancy by CUT&Tag, relative to the primary peaks (summits) of RNAPII-S5P CUTAC (b) and RefSeq annotated gene TSSs (c), sorted by decreasing RNAPII-S5P occupancy. d, Violin plots of CUT&Tag signal distribution comparing RNAPII-S5P, BRG1, and histone PTM occupancies at specific set of gene promoters (TSS) showing RNAPII-S5P enrichment versus promoter-distal S5P CUTAC and pluripotency TF-binding peaks (Distal). Median value (solid line), upper and lower quartiles (broken lines) and outliers were calculated using the Tukey method. Numbers on top show mean values. e, Scatterplots comparing BRG1 and RNAPII S5P, S2P, and RPB3 CUT&Tag reads in 1,000 bp genome-wide consecutive non-overlapping bins. f, Heatmaps (bottom) and average plots (top) comparing chromatin accessibility (RNAPII-S5P CUTAC and ATAC-seq), nucleosome positions (MNase seq), and transcriptional activity (START-seq), relative to the primary peaks (summits) of RNAPII-S5P CUTAC; and RNAPII-S5P CUTAC signal relative to RefSeq annotated gene TSSs (extreme right). All datasets are representative of at least two biological replicates.

Extended Data Fig. 2. CUT&Tag of RNAPII-S5P and BRG1 after inhibitor treatment.

a, b, d, e, Heatmaps (bottom) and average plots (top) comparing RNAPII-S5P (a, d) and BRG1 (b, e) occupancy by spike-in calibrated CUT&Tag relative to the primary peaks (summits) of RNAPII-S5P CUTAC in untreated cells (DMSO) versus cells treated with Triptolide (a, b), Flavopiridol, and Actinomycin D (d, e) at indicated time points post drug treatment. c, g, h, Comparison of fold changes in mean RNAPII-S5P and BRG1 occupancy (spike-in calibrated CUT&Tag) at gene promoters (TSS, squares) and promoter-distal regulatory regions (Distal, circles) at time points after drug treatments. f, Scatterplots comparing BRG1 and RNAPII-S5P CUT&Tag reads in 1000 bp genome-wide consecutive non-overlapping bins in cells treated with Flavopiridol and Actinomycin D. i, Violin plots of CUT&Tag signal distribution comparing histone PTM H3K27ac occupancy at S5P CUTAC peaks over time points after Actinomycin D treatment. Median value (solid line), upper and lower quartiles (broken lines) and outliers were calculated using the Tukey method. Numbers on top show mean values. All datasets are representative of at least two biological replicates.

Extended Data Fig. 3. CUT&RUN.ChIP of BRG1.

a, Heatmaps (bottom) and average plots (top) of RNAPII-S5P CUTAC separated by fragment size, relative to primary peaks (summits) of RNAPII-S5P CUTAC. b, Comparison of RNAPII-S5P CUTAC fragment size distribution over peaks (promoter and enhancer NDR spaces) in cells treated with DMSO (control) and Flavopiridol; same data as used for Fig. 2A, plotted differently. c-f, Enrichment of nucleosomal (≥150 bp, solid lines) and subnucleosomal (≤120 bp, broken lines) reads from BRG1 CUT&RUN and CUT&RUN.ChIP experiments, relative to gene promoter TSSs showing RNAPII-S5P enrichment (c, d) and distal regulatory sites (e, f), in DMSO (c, e) and Flavopiridol (d, f) treated cells. CUT&RUN.ChIP data were plotted as enrichment in histone ChIP over IgG isotype control. All datasets are representative of at least two biological replicates.

Extended Data Fig. 4. CUT&Tag of RNAPIIS5P and BRG1 at PcG-repressed promoters.

a, K-means clustering of RNAPII-S5P and H3K27me3 CUT&Tag reads relative to RefSeq annotated gene promoter TSSs to group promoters as active (I, RNAPII-S5P enriched) and PcG-repressed (II, H3K27me3 enriched), and not enriched for either (III). b, c, Heatmaps (bottom) and average plots (top) comparing RNAPII-S5P (b) and BRG1 (c) occupancy by spike-in calibrated CUT&Tag relative to PRC2-repressed promoter TSSs in untreated cells (DMSO) versus cells treated with Flavopiridol or Actinomycin D at indicated time points post drug treatment. d, Heatmaps (bottom) and average plots (top) comparing H3K27me3 histone PTM occupancy by spike-in calibrated CUT&Tag relative to PRC2-repressed promoter TSSs in untreated cells (DMSO) and cells treated with Flavopiridol. e, Heatmaps (bottom) and average plots (top) comparing H3K9me3 histone PTM occupancy (CUT&Tag) with RNAPII-S5P and BRG1 relative to H3K9me3 peaks in untreated cells (DMSO) versus cells treated with Flavopiridol or Actinomycin D. RNAPII-S5P and BRG1 CUT&Tag reads were spike-in calibrated and plotted to the same scales as in panels b and c, respectively. All datasets are representative of at least two biological replicates.

Extended Data Fig. 5. CUT&RUN of pluripotency TFs in SL versus 2i culture conditions.

a, Heatmaps (bottom) and average plots (top) comparing pluripotency TF occupancy by CUT&RUN at RNAPII-enriched (active) and H3K27me3-enriched (PRC2-repressed) promoters. Promoters were grouped based on K-means clustering of RNAPII-S5P and H3K27me3 CUT&Tag reads mapping to a 5 kb window around the TSSs of RefSeq-annotated mESC genes, see Extended Data Fig. 4a. b, c, Enrichment of nucleosomal (≥150 bp, solid lines) and subnucleosomal (≤120 bp, broken lines) reads from BRG1 CUT&RUN and CUT&RUN.ChIP experiments, relative NANOG foci (smallest fragment within primary peaks called in SL condition), in SL (b) and 2i (c) mESC culture conditions. CUT&RUN.ChIP data were plotted as enrichment in histone ChIP over IgG isotype control. d, Scatterplots comparing pluripotency TF CUT&RUN and RNAPII-S5P CUT&Tag reads over S5P CUTAC peaks in SL mESCs. e, Immunofluorescent staining comparing pluripotency TF and BRG1 expression in SL versus 2i culture conditions. Cy5-conjugated secondary antibodies were used in all experiment except for KLF4, where Rhodamine red-conjugated antibody was used. DAPI (blue) was used to stain the nucleus in cells. f, Western blot analysis comparing pluripotency TF and BRG1 expression in SL and 2i culture conditions. Equal amounts of extracted total proteins were loaded in each well of 4–20% gradient polyacrylamide SDS electrophoresis gel, and histone H3 signal is used as control to ensure equivalent protein loading. Bar-graph quantifications represent average of two biological replicates with individual data points shown as black dots. Data were normalized to values in SL. g, Heatmaps (bottom) and average plots (top) comparing pluripotency TF occupancy by spike-in calibrated CUT&RUN in SL versus 2i culture conditions. Heatmaps were plotted relative to S5P CUTAC summits showing TF binding at sites of DNA accessibility and sorted by decreasing TF occupancy in SL (CUT&RUN signal). All datasets are representative of at least two biological replicates. Source data

Extended Data Fig. 6. Labelled and uncropped Western blots of data shown in Extended Data Fig. 5f.

Western blots were dually stained with IRDye 800CW secondary antibody against primary antibodies targeting the TFs (panel a) and IRDye 680RD secondary antibody against primary antibodies targeting histone H3 (panel b) and imaged using respective filters on the Odyssey DLx Imaging System.