Molecular features driving condensate formation and gene expression by the BRD4-NUT fusion oncoprotein are overlapping but distinct

Abstract

Aberrant formation of biomolecular condensates has been proposed to play a role in several cancers. The oncogenic fusion protein BRD4-NUT forms condensates and drives changes in gene expression in Nut Carcinoma (NC). Here we sought to understand the molecular elements of BRD4-NUT and its associated histone acetyltransferase (HAT), p300, that promote these activities. We determined that a minimal fragment of NUT (MIN) in fusion with BRD4 is necessary and sufficient to bind p300 and form condensates. Furthermore, a BRD4-p300 fusion protein also forms condensates and drives gene expression similarly to BRD4-NUT(MIN), suggesting the p300 fusion may mimic certain features of BRD4-NUT. The intrinsically disordered regions, transcription factor-binding domains, and HAT activity of p300 all collectively contribute to condensate formation by BRD4-p300, suggesting that these elements might contribute to condensate formation by BRD4-NUT. Conversely, only the HAT activity of BRD4-p300 appears necessary to mimic the transcriptional profile of cells expressing BRD4-NUT. Our results suggest a model for condensate formation by the BRD4-NUT:p300 complex involving a combination of positive feedback and phase separation, and show that multiple overlapping, yet distinct, regions of p300 contribute to condensate formation and transcriptional regulation.

Fig. 1:. BRD4-NUT fusion protein forms nuclear condensates in human cells.

a) Cartoon of BRD4, NUT and BRD4-NUT fusion proteins. Bromodomains 1 and 2 and extraterminal domain of BRD4 as well as predicted α-helices are indicated. b) Line profiles across cells expressing mNeonGreen-tagged BRD4, NUT and BRD4-NUT fusion; scale bars = 10μm. c) Micrographs of cells expressing mNeonGreen-tagged BRD4, NUT and BRD4-NUT; scale bars = 10μm. d) Quantification of percentage of cells forming condensates (at least 2 condensates larger than 1.25μm in diameter).

Fig. 2:. BRD4-NUT condensates recruit p300 histone acetyltransferase and are heavily acetylated at H3K27.

a) Cartoon of BRD4-NUT and p300 interaction; known interaction motifs are indicated (Shiota et al., Cell Rep., 2018; Reynoird et al., EMBO, 2010; Ibrahim et al., Nat Comm, 2022) b) Micrographs of HCC2429 Nut Carcinoma cells co-stained with BRD4 and p300 or BRD4 and Histone H3K27Ac antibodies; scale bar = 10μm. c) Immunoprecipitation against mNeonGreen; western blot with a p300 antibody showing that BRD4-NUT pulls down p300.

Fig. 3:. Minimal p300-interaction fragment of NUT in BRD4-NUT fusion is necessary and sufficient for condensate formation.

a) Schematic of all BRD4-NUT – based constructs. b) Representative micrographs showing condensate formation in stable cell lines expressing different constructs. Staining against mNeonGreen shown in red, overlay with DAPI to indicate the nucleus. Scale bars = 10μm. c) Quantification of the micrographs represented in b. d) Representative micrographs comparing expression of constructs with or without p300-interaction motif in cells. Cells stained with anti-BRD4 antibody (magenta) and anti-p300 antibody (green). Scale bars = 10μm. e) and f) Quantification of the overlap between condensates via co-staining shown as Pearson correlation. Cells co-stained with NUT and p300 antibodies (left) or mNeonGreen and p300 antibodies (right); Nut antibody epitope was removed in BRD4-NUT(ΔMIN), seefig.S2c for more details.

Fig. 4:. p300 binding and enzymatic activity are necessary for gene expression changes induced by BRD4-NUT.

a) Quantification of cells capacity to form condensates, comparing cells that are untreated (gray) and cells treated with C646 inhibitor (yellow) or JQ1 inhibitor (pink). Cells compared here include HCC2429 Nut Carcinoma cell line and stable cell lines expressing BRD4-NUT(FL) and BRD4-NUT(MIN). b) Local acetylation measured as average fluorescence intensity across a condensate. Data shown for cells untreated (gray) or treated with C646 (yellow), in cells expressing BRD4-NUT(FL), BRD4-NUT(MIN) and BRD4-NUT(ΔMIN). c) Graph showing logFC for SOX2 and TP63 genes upon expression of different BRD4-p300 constructs. d) Venn diagram of gene occupancy by BRD4-NUT(FL) (pink) and BRD4-NUT(MIN) (blue), found via ChIPseq with @NUT antibody: all annotated genes occupied by either protein. Values in the diagram show the number of unique genes annotated; numbers in parentheses represent the mean overlap in 20 iterations with a randomly generated gene pool of the same size, from the human genome. e) Venn diagram of gene occupancy by BRD4-NUT(FL) (pink) and BRD4-NUT(MIN) (blue), found via ChIPseq with @NUT antibody: genes annotated at promoter-TSS regions of genes, with normalized signal value of 0.3–1. Values in the diagram as in d). f) Example ChIPseq tracks comparing gene occupancy by BRD4-NUT(FL) and BRD4-NUT(MIN). Data scale: 0–5. Inset shows more closely an example of the same genes being occupied by both proteins, but BRD4-NUT(MIN) occupying more loci. g) Venn diagram of genes upregulated upon expression of BRD4-NUT(FL) (purple) or BRD4-NUT(MIN) (green), found via RNAseq. Genes shown when fold change in expression was greater than two and p<0.05. Numbers in parentheses represent the mean overlap in 20 iterations with a randomly generated gene pool of the same size, from the human genome. h) Venn diagram of genes downregulated upon expression of BRD4-NUT(FL) (purple) or BRD4-NUT(MIN) (green), found via RNAseq. Statistical analyses as in g). i) RNAseq-ChIPseq data integration: heatmap showing up- and down-regulated genes found via RNAseq, bound by both BRD4-NUT(FL) and BRD4-NUT(MIN). Top 200 genes shown in the heatmap, based on p value.

Fig. 5:. Brd4-p300 fusion forms condensates and recapitulates many of the transcriptional changes observed for Brd4-Nut fusion.

a) Schematic of Brd4-Nut and Brd4-p300. b) Micrographs of Brd4-Nut and Brd4-p300 – expressing stable cell lines. Cells are stained with mNeonGreen antibody; scale bars = 10μm. c) Quantification of the micrographs represented in b. d) Quantification of histone H3K27Ac staining in untreated vs. C646-treated cells. Local acetylation shown as average fluorescence intensity per condensate. e) Venn diagram of genes acetylated upon expression of BRD4-NUT(MIN) (pink) and BRD4-p300 (blue), found via ChIPseq with @H3K27Ac antibody, at promoter-TSS regions: all annotated genes. Values in the diagram show the number of unique genes annotated; numbers in parentheses represent the mean overlap in 20 iterations with a randomly generated gene pool of the same size, from the human genome. f) Venn diagram of genes acetylated upon expression of BRD4-NUT(MIN) (pink) and BRD4-p300 (blue), found via ChIPseq with @H3K27Ac antibody, at promoter-TSS regions: Diagram shows only genes after applying 0.3–1 normalized signal value cutoff. Values in the diagram as in e). g) Venn diagram of genes upregulated upon expression of BRD4-NUT(MIN) (green) or BRD4-p300 (gray), found via RNAseq. Genes shown when fold change in expression was greater than two and p<0.05. Numbers in parentheses represent the mean overlap in 20 iterations with a randomly generated gene pool of the same size, from the human genome. h) Venn diagram of genes downregulated upon expression of BRD4-NUT(MIN) (green) or BRD4-p300 (gray), found via RNAseq. Statistical analyses as in g). i) RNAseq-ChIPseq data integration: heatmap showing up- and down-regulated genes found via RNAseq, that are bound by both BRD4-NUT(MIN) and BRD4-p300. Top up- and downregulated genes shown in the heatmap, based on log fold change, where log fold change was greater than 0.5.

Fig. 6:. p300 IDRs are dispensable for condensate formation.

a) Disorder prediction of p300 IDRs as shown via IUPRED2 analysis. b) Schematics of BRD4-p300(FL), (IDR) and (ΔIDR) constructs; different domains and disordered regions are indicated. c) Micrographs of all the BRD4-p300 constructs from a); scale bars = 10μm. d) Quantification of the micrographs represented in c).

Fig. 7:. IDRs, HAT and TF binding domains of p300 collectively contribute to condensate formation.

a) Schematic of BRD4-p300 – based constructs, including FL, H*IT, HI and I. b) Micrographs of all the BRD4-p300 constructs from a); scale bars = 10μm. c) Quantification of the micrographs represented in b). d) Colocalization of BRD4-p300(FL), as visualized by α-mNeonGreen staining, with α-H3K27Ac staining, shown in micrographs and via line profiles. e) Colocalization of BRD4-p300(H*IT), as visualized by α-mNeonGreen staining, with α-H3K27Ac staining, shown in micrographs and via line profiles. f) Quantification of correlation between α-mNeonGreen and α-H3K27Ac condensates, as shown via Pearson Correlation.

Fig. 8:. Acetylation activity of p300 is necessary for transcriptional changes observed upon expression of BRD4-p300(FL).

a) Percent of genes affected upon expression of different BRD4-p300 constructs. b) Graph showing logFC for SOX2 and TP63 genes upon expression of different BRD4-p300 constructs. c) PCA plot, summarizing RNAseq data for cells expressing all constructs. Cell lines excluded in this plot: BRD4-NUT(FL) and BRD4-NUT(MIN) are included in the full PCA in Fig.S4b. d) Summary of molecular parts of p300 required for condensate formation, condensatelocalized acetylation and transcriptional changes in cells. e) Final model of condensate formation.