An RNA-dependent and phase-separated active subnuclear compartment safeguards repressive chromatin domains

Abstract

The nucleus is composed of functionally distinct membraneless compartments that undergo phase separation (PS). However, whether different subnuclear compartments are connected remains elusive. We identified a type of nuclear body with PS features composed of BAZ2A that associates with active chromatin. BAZ2A bodies depend on RNA transcription and BAZ2A non-disordered RNA-binding TAM domain. Although BAZ2A and H3K27me3 occupancies anticorrelate in the linear genome, in the nuclear space, BAZ2A bodies contact H3K27me3 bodies. BAZ2A-body disruption promotes BAZ2A invasion into H3K27me3 domains, causing H3K27me3-body loss and gene upregulation. Weak BAZ2A-RNA interactions, such as with nascent transcripts, promote BAZ2A bodies, whereas the strong binder long non-coding RNA (lncRNA) Malat1 impairs them while mediating BAZ2A association to chromatin at nuclear speckles. In addition to unraveling a direct connection between nuclear active and repressive compartments through PS mechanisms, the results also showed that the strength of RNA-protein interactions regulates this process, contributing to nuclear organization and the regulation of chromatin and gene expression.

Keywords: BAZ2A; H3K27me3; Malat1; RNA; chromatin; ground-state pluripotency; nuclear condensates; nuclear speckles; phase separation.

Graphical abstract

Figure 1

BAZ2A forms bodies that depend on its RNA-binding domain TAM (A) Representative immunofluorescence image showing BAZ2A localization in ESC + 2i using antibodies against BAZ2A and the nucleolar marker NPM1. Scale bars, left panel: 10 μm; right panel: 2 μm. (B) Live-cell image of ESCs expressing endogenous BAZ2A tagged with mGFP (mGFP-BAZ2A_end). Scale bars: 10 μm. (C) Images of ESCs treated with RNase A. Boxplot shows the number of BAZ2A bodies/cell. Error bars represent SD. Statistical significance for three independent experiments was calculated using Mann-Whitney test (^∗∗∗p < 0.001). (D) Western blot showing the expression levels of endogenous BAZ2A and mGFP-BAZ2A_WT and -BAZ2A_ΔTAM transgenes in the corresponding ESC lines. BAZ2A signal was detected with BAZ2A antibodies. SNF2H serves as loading control. (E) GFP-trap immunoprecipitation (IP) in parental ESCs, ESC + mGFP-BAZ2A_WT, and ESC + mGFP- BAZ2A_ΔTAM. BAZ2A signal was detected with BAZ2A antibodies. SNF2H is a known BAZ2A-interacting protein. (F) Live-cell image of ESC + mGFP-BAZ2A_WT and ESC + mGFP- BAZ2A_ΔTAM. The domain composition of BAZ2A_WT (1,889 aa) and BAZ2A_ΔTAM is shown. Yellow lanes represent AT-hook domains. (G) Anti-HA immunoprecipitation from HEK293T cells transfected with plasmids expressing F/H-BAZ2A_N and mGFP-BAZ2A_N/WT or -BAZ2A_N/ΔTAM. (H) GST pull-down of recombinant GST-BAZ2A_N/WT and mGFP-BAZ2A_N/WT or -BAZ2A_N/ΔTAM. GST and GFP antibodies were used to visualize the corresponding proteins. (I) Anti-HA immunoprecipitation from HEK293T cells transfected with plasmids expressing F/H-BAZ2A_C and mGFP-BAZ2A_N/WT or -BAZ2A_N/ΔTAM. (J) Profile of size exclusion chromatography of recombinant BAZ2A_N and BAZ2A_N/ΔTAM. Fractions were measured by absorbance at 280 nm. (K) Negative staining of BAZ2A_N fractionated sample corresponding to >600 kDa. Scale bars: 50 nm.

Figure 2

BAZ2A bodies in ESCs show physicochemical LLPS properties in vitro (A) Representative images of recombinant mGFP-BAZ2A_N under the indicated conditions. Scale bars: 5 μm. (B and C) Quantification of the number (B) and circularity and area (C) of recombinant BAZ2A_N droplets from three independent experiments. (D and E) Representative images (D) and corresponding quantifications of droplets number (E) of mGFP-BAZ2A_N/WT and mGFP-BAZ2A_N/ΔTAM. Scale bars: 5 μm. (F) Left panel, representative images of droplets formed with recombinant mCherry-BAZ2A_N/WT and mGFP-BAZ2A_N/WT or mGFP-BAZ2A_N/ΔTAM. Scale bars: 2 μm. Right panel, quantifications of mGFP- and mCherry-positive droplets from two independent experiments. (G) Tracks displaying BAZ2A_WT and BAZ2A_ΔTAM occupancy in ESC + F/H-BAZ2Ar_WT and ESC + F/H-BAZ2Ar_ΔTAM. Eigenvector values of A and B compartments in ESC + 2i are from Dalcher et al. (H) Proportional Venn diagram showing common and specific genome occupancy of BAZ2A_WT and BAZ2A_ΔTAM. (I) Boxplots showing levels of BAZ2A_WT and BAZ2A_ΔTAM occupancy at domains both bound by BAZ2A_WT and BAZ2A_ΔTAM. Error bars represent SD. Statistical significance was calculated using Mann-Whitney test (^∗∗∗∗p < 0.0001). (J) Model showing BAZ2A_WT forming bodies through the TAM domain and the association with chromatin. The lack of the TAM domain impairs body formation and promotes BAZ2A association with chromatin.

Figure 3

BAZ2A bodies associate and regulate H3K27me3 bodies (A) Representative immunofluorescence images showing BAZ2A and H3K27me3 distribution in ESC + 2i. Right images represent the magnification of the section labeled with a rectangle. Scale bars, left images: 10 μm; right images: 2 μm. (B) Proportion of ESCs with H3K27me3 bodies having BAZ2A bodies (left) and with BAZ2A bodies having H3K27me3 bodies (right panel). Data were from 20 ESC colonies. (C) 3D image of one ESC showing BAZ2A and H3K27me3 bodies and DAPI-stained chromocenters. (D and E) Quantification of overlapping volume of BAZ2A bodies over H3K27me3 bodies and chromocenters (D) and of H3K27me3 bodies over BAZ2A bodies and chromocenters (E). Data are from 13 nuclei. Statistical significance was calculated with Mann-Whitney test (^∗∗∗∗p < 0.0001). (F) Representative immunofluorescence images showing mGFP-BAZ2A_end and H3K27me3 distribution in ESC depleted of BAZ2A by siRNA. Scale bars: 10 μm. (G) Quantification of ESCs with H3K27me3 bodies and number and mean intensity of H3K27me3 bodies upon treatment with siRNA-Control and siRNA-Baz2a. Error bars represent SD. (H) Representative immunofluorescence images showing BAZ2A and H3K27me3 distribution in ESC + mGFP-BAZ2A_WT and ESC + mGFP-BAZ2A_ΔTAM. Scale bars: 10 μm. (I) Quantification of ESCs with H3K27me3 bodies, and the number and mean intensity of H3K27me3 bodies in ESCs expressing mGFP-BAZ2A_WT and mGFP-BAZ2A_ΔTAM. Data are from three independent experiments. (J) Left panel, heatmap showing H3K27me3 peaks detected in ESC + F/H-BAZ2Ar_WT and the corresponding signal in ESC + F/H-BAZ2Ar_ΔTAM. Right panel, average density plots of H3K27me3-ChIP-seq read counts at ±1 kb from H3K27me3 peak summits in the corresponding ESC lines. (K) Levels of H3K27me3 at the 10% top or bottom H3K27me3 regions in ESC + BAZ2A_WT and the corresponding levels in ESC + BAZ2A_ΔTAM. Values are shown as average reads per kilobase per million (RPKM) of a 10-kb bin size region. (L) Tracks displaying H3K27me3 BAZ2A_WT and BAZ2A_ΔTAM occupancy in ESC + F/H-BAZ2Ar_WT and ESC + F/H-BAZ2Ar_ΔTAM at Hoxd gene cluster. (M) H3K27me3 levels at BAZ2A_WT- and BAZ2A_ΔTAM-bound regions in parental ESC + 2i. (N) Spearman’s correlation heatmap for BAZ2A_WT, BAZ2A_ΔTAM, H3K27ac, and H3K27me3. H3K27me3 and H3K27ac ChIP-seq in ESC + 2i were from Dalcher et al. (O) H3K27me3 levels in ESC + F/H-BAZ2Ar_WT and ESC + F/H-BAZ2Ar_ΔTAM at the 10% top H3K27me3 regions bound by BAZ2A_ΔTAM. (P) Model showing BAZ2A_WT forming bodies through the TAM domain and the association with chromatin depleted of H3K27me3. The lack of the TAM domain impairs body formation and promotes BAZ2A invasion into H3K27me3 domains and the loss of this repressive signature. Statistical significance in boxplots (G), (I), (K), (M), and (O) was calculated with Mann-Whitney test (^∗p < 0.05, ^∗∗∗∗p < 0.0001). Error bars represent SD.

Figure 4

BAZ2A-TAM domain regulates gene expression (A) Volcano plot showing log₂ fold change transcript levels of ESC + F/H-BAZ2Ar_ΔTAM vs. ESC + F/H-BAZ2Ar_WT. (B) Proportion of BAZ2A-regulated genes significantly regulated in ESC + BAZ2A_ΔTAM. (C) Heatmap showing fold changes of upregulated and downregulated transcripts in ESC + Baz2a-siRNA and corresponding fold changes in ESC + F/H-BAZ2Ar_ΔTAM vs. ESC + F/H-BAZ2Ar_WT. (D) Fold changes of H3K27me3 levels in ESC + F/H-BAZ2Ar_ΔTAM vs. ESC + F/H-BAZ2Ar_WT at genes up- or downregulates in ESC + BAZ2Ar_ΔTAM. Error bars represent SD. Statistical significance was calculated with unpaired two-tailed t test (^∗∗∗∗p < 0.0001; ns, nonsignificant). (E) Proportion of genes upregulated and bound by BAZ2A_ΔTAM with decreased H3K27me3 levels in ESC + F/H-BAZ2Ar_ΔTAM. (F) Model showing how the lack of the BAZ2A-TAM domain impairs body formation and promotes BAZ2A invasion into H3K27me3 domains, the loss of H3K27me3, and upregulation of gene expression. (G) Quantifications of up- and downregulated genes in ESC + F/H-BAZ2Ar_ΔTAM significantly up- or downregulated in ESC + 2i relative to ESC + serum.

Figure 5

BAZ2A associates with RNA (A) Pie chart showing genome annotation of BAZ2A-iCLIP sites. (B) Scatterplot showing the correlation between gene expression and BAZ2A-iCLIP read density. (C) Boxplot showing the expression level of genes containing or depleted of BAZ2A-iCLIP sites. Values are shown as log₁₀ average RPKM. (D) Pie chart showing the percentage of BAZ2A-bound genes containing BAZ2A-iCLIP sites among all genes with BAZ2A-iCLIP sites. (E) Boxplot showing expression level of BAZ2A-bound genes containing or depleted of BAZ2A-iCLIP sites. Values are shown as log₁₀ average RPKM. (F) Representative live-cell images showing ESC + mGFP-BAZ2A_end treated with triptolide (TPL) for 4 h. Scale bars: 10 μm. (G–I) Boxplots showing amounts of cells with BAZ2A bodies (G) and the number (H) and area (I) of BAZ2A bodies in ESCs treated with TPL. (J) Tracks of BAZ2A-ChIP-seq displaying BAZ2A occupancy in ESCs treated with TPL. (K) Bar diagram showing the coverage of BAZ2A-bound sites in ESCs treated with TPL. (L) Boxplots showing the levels of BAZ2A occupancy in ESCs treated with DMSO or TPL for 4 h. (M) Representative immunofluorescence images showing H3K27me3 in ESC treated with triptolide (TPL) for 4 h. Scale bar represents 10 μm. Quantifications of the number of H3K27me3 bodies/cell are shown. Statistical significance in boxplots in (C), (E), (G)–(I), (L), and (M) was calculated with Mann-Whitney test (^∗∗p < 0.01 ^∗∗∗∗p < 0.0001).

Figure 6

BAZ2A associates with Malat1 and regulates nuclear speckles (A) Tracks showing Malat1 signal in BAZ2A-iCLIP and RNA-seq. (B) Levels of Malat1 from HA-RIP in ESC + F/H-BAZ2Ar_WT and ESC + F/H-BAZ2Ar_ΔTAM. Data are normalized to BAZ2A_WT-bound Malat1 and are from two independent experiments. (C) Tracks showing SRRM2-TSA-seq Malat1-association to chromatin from SPRITE and BAZ2A-ChIP-seq. (D) Spearman’s correlation heatmap of BAZ2A, H3K27me3, and H3K27ac ChIP-seq, SRRM2 TSA-seq, and Malat1 SPRITE. BAZ2A ChIP-seq and SRRM2-TSA-seq are from this work. H3K27ac and H3K27me3 ChIP-seq and Malat1 SPRITE were from Dalcher et al. and Quinodoz et al. (E) Representative immunofluorescence images showing BAZ2A and SRRM2-marked nuclear speckles. Scale bars, top images: 5 μm; bottom images: 2 μm. (F) Quantification of overlapping volume of BAZ2A bodies over nuclear speckles. Data are from 25 cells and 184 BAZ2A bodies. (G) Representative immuno-RNA-FISH images of BAZ2A and Malat1. Scale bar is 10 μm, magnified image 5 μm.

Figure 7

Malat1 regulates BAZ2A association to chromatin and limits body formation (A) Tracks showing BAZ2A-ChIP-seq profile in ESC + gapmer-control or gapmer-Malat1. (B) BAZ2A levels at BAZ2A-bound regions in ESC control and corresponding levels in ESC + gapmer-Malat1. Values are shown as average RPKM. Statistical significance was calculated using the paired two-tailed t test (^∗∗∗∗p < 0.0001). (C) Anti-FLAG ChIP-qPCR of ESCs and ESC + F/H-BAZ2A_WT treated with gapmer-control or gapmer-Malat1. Data were normalized to input and to Atf7ip. Average values of three independent experiments. Error bars represent SD, and statistical significance was calculated using the paired two-tailed t test (^∗p < 0.05, ^∗∗p < 0.01; ns, nonsignificant). (D) Representative images showing ESC + mGFP-BAZ2A_end treated with gapmer-control or gapmer-Malat1. Scale bars: 10 μm. (E) Proportion of cells with BAZ2A bodies and area and number of BAZ2A bodies/cells in ESC + gapmer-control or gapmer-Malat1. (F) Representative images of ESC + mGFP-BAZ2A_end. When indicated, cells were treated with Triton X-100 prior fixation. Scale bars: 10 μm. (G) Quantification of area and number of BAZ2A bodies/cells in ESCs. (H) Representative images of droplets using 500 nM recombinant mGFP-BAZ2A_N, in the absence or presence of 50 nM RNA-control, pRNA, or Malat1. Scale bars: 2 μm. Right panel, quantifications of the number of droplets under the indicated RNA concentrations. Values are from two independent experiments. (I) BAZ2A binds to Malat1. Increasing equal moles of in vitro transcripts corresponding to RNA-Control and Malat1 sequences were used to compete for binding of recombinant BAZ2A_332–723 to radiolabeled control RNA. RNA/protein complexes were analyzed by EMSA. (J) Model showing the role of BAZ2A condensates and Malat1 in chromatin regulation of ESCs. BAZ2A condensates are close to H3K27me3 bodies and depend on BAZ2A-TAM domain and active transcription. BAZ2A bodies sequester BAZ2A and limit BAZ2A invasion into chromatin. The model shows how BAZ2A_ΔTAM occupies H3K27me3 chromatin with consequent H3K27me3 reduction and activation of gene expression. On the right, BAZ2A associates with Malat1 and chromatin contacting nuclear speckles. Malat1 is required for BAZ2A binding to chromatin and counteracts BAZ2A body formation, which do not contact nuclear speckles. Statistical significance in boxplots in (E) and (G) was calculated with Mann-Whitney test (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗∗p < 0.0001; ns, nonsignificant). Error bars represent SD.