BRD4 bimodal binding at promoters and drug-induced displacement at Pol II pause sites associates with I-BET sensitivity

Abstract

Background: Deregulated transcription is a major driver of diseases such as cancer. Bromodomain and extra-terminal (BET) proteins (BRD2, BRD3, BRD4 and BRDT) are chromatin readers essential for maintaining proper gene transcription by specifically binding acetylated lysine residues. Targeted displacement of BET proteins from chromatin, using BET inhibitors (I-BETs), is a promising therapy, especially for acute myeloid leukemia (AML), and evaluation of resistance mechanisms is necessary to optimize the clinical efficacy of these drugs.

Results: To uncover mechanisms of intrinsic I-BET resistance, we quantified chromatin binding and displacement for BRD2, BRD3 and BRD4 after dose response treatment with I-BET151, in sensitive and resistant in vitro models of leukemia, and mapped BET proteins/I-BET interactions genome wide using antibody- and compound-affinity capture methods followed by deep sequencing. The genome-wide map of BET proteins sensitivity to I-BET revealed a bimodal pattern of binding flanking transcription start sites (TSSs), in which drug-mediated displacement from chromatin primarily affects BRD4 downstream of the TSS and prolongs the pausing of RNA Pol II. Correlation of BRD4 binding and drug-mediated displacement at RNA Pol II pause sites with gene expression revealed a differential behavior of sensitive and resistant tumor cells to I-BET and identified a BRD4 signature at promoters of sensitive coding and non-coding genes.

Conclusions: We provide evidence that I-BET-induced shift of Pol II pausing at promoters via displacement of BRD4 is a determinant of intrinsic I-BET sensitivity. This finding may guide pharmacological treatment to enhance the clinical utility of such targeted therapies in AML and potentially other BET proteins-driven diseases.

Keywords: Bromodomain proteins; Leukemia; Promoters; Regulatory regions; Sensitivity and resistance to drug treatment; TSS.

Fig. 1

BET proteins displacement from chromatin is compound dose dependent. a Proliferation assays in sensitive (MV4;11) and resistant (K562) cells treated with increasing concentrations of I-BET151. Compound IC50 (M) or compound concentration which induces 50% inhibition of cell viability as well as the response area (see “Materials and methods” section for the calculation) is shown below the graph. b Experimental design followed in the study and I-BET151 structure. c Bar plot showing the ChIP-qPCR enrichment relative to DMSO for the ZC3H4 locus in K562 and MV4;11 for cells treated with three concentrations of I-BET151 and DMSO as a control. d Barplot showing the number of peaks called for BRD4 at TSS for cells treated with three concentrations of I-BET151 and DMSO as a control and for both cell lines

Fig. 2

Genome-wide I-BET151 dose response of BET proteins binding profiles. a Genome-wide ChIP-seq profiles on TSS −/+ 4 Kb for BRD2, BRD3 and BRD4. All ChIP-seq profiles are RPGC (Reads Per Genomic Content) normalized followed by input subtraction. b ChIP-seq profiles obtained as in “a” for H3K27ac, H4K5ac and Pol II in MV4;11 from previously published data (52). c Genome browser visualization of two typical loci in both cell lines: KMT2E locus in K562 (left panel) and BCL2 locus in MV4;11 (right panel) showing the enrichment and the associated gradual decrease in ChIP-seq signal for all three BET proteins in DMSO and compound-treated samples. The same color codes are used for the genome browser and genome-wide profiles. The Y-scale is the same for all conditions and both cell lines for comparison purposes. Scale is indicated in the lower left corner

Fig. 3

BRD4 displacement at TSSs correlates with cell sensitivity. a Representation of a TSS and the surrounding − 1 Kb (core promoter) and + 1 Kb (pause site) used for the DB analysis. Right panel shows the count of total TSSs and the number (and percentage) of TSSs affected (at − 1 Kb or + 1 Kb or both) in MV4;11 and K562. b Table containing TSS counts for significant cases of DB at core promoter and pause site or both. The “Total affected” column contains the number (and percentage) of TSSs affected at any of the two sites. c Hierarchical clustering of ChIP signal (log2 fold change or LFC) for TSSs (rows) with DB in − 1 Kb or + 1 Kb as defined above. For each of the three treatment conditions (50 nM, 500 nM or 5000 nM), LFC of signal at TSS − 1 Kb and TSS + 1 Kb (columns) in K562 (left, total of 519 TSSs with DB) and MV4;11 (right, total of 3531 TSSs with DB) are shown (“Materials and methods” section). In each cluster, a random set of genes is shown in addition to the 5 most sensitive genes in Cluster 2 showing differential binding at 50 nM I-BET in MV4;11 (in red) in addition to BCL2 in Cluster 1. The category of each gene (protein coding or non-coding) is represented under the biotype column on the right to each heatmap. All four non-coding RNAs with a BRD4 DB at 50 nM I-BET151 in MV4;11 fall in cluster 2 for MV4;11 cells. For the K562 DB-based clustering (left panel heatmap), only the protein coding gene RBM38 showed a significant decrease in binding. The middle panel represents metagene profiles of BRD4 ChIP-seq data for the corresponding clusters. Inferred names to the identified clusters are indicated below the metagene profiles. d Genome browser visualization of the LINC-ROR locus showing the strong decrease in ChIP signal for all BET proteins, and specifically BRD4, when treated with 50 nM I-BET151. Note the marked decrease at TSS + 1 Kb (pause site). The green line and gray shadowed box depict the TSS −/+ 1 Kb area. The scale is shown in the lower left corner. e RNA expression profiling by RT-qPCR on the five genes showing DB in MV4;11 cells at 50 nM of I-BET151. The expression value of each transcript is referred to the expression of a reference gene (OTUD5), and the mean of the DMSO for each comparison group is set to 1. Stars represent level of significance with (*) for p -value < 0.05, and (**) for p value < 0.005. Only cases with detectable expression are shown

Fig. 4

Chem-seq shows cluster-dependent signal profiles of BRD4/I-BET151 interactions. a Method scheme featuring loci recovered by ChIP-seq alone and those recovered by ChIP-seq and Chem-seq. Using an antibody against a BET protein member, we can recover all genomic loci bound by the protein. Using a biotinylated derivative of I-BET121, we can pull down fragments where the BET protein exhibits an accessible bromodomain. b Normalized genome-wide cluster-specific Chem-seq profiles on TSS −/+ 4 Kb in the absence or presence of excess free I-BET151 as control

Fig. 5

Gene set enrichment analysis (GSEA) defines hallmarks of I-BET sensitivity. a Table summarizing counts of differentially expressed genes for all conditions in both cell lines. b Venn diagrams representing the counts of common or distinct sets between genes exhibiting BRD4 differential binding (DB) and genes exhibiting differential expression (DE) for each cell line and I-BET concentration. Counts and overlaps for the I-BET151 50 nM treatment were too small and thus not represented here. c Normalized enrichment scores (NES) of GSEA for hallmark gene sets v6.1. Listed are all hallmarks that show significant NES (− 1 < NES or NES > 1 with FDR < 0.1) in MV4;11 or K562 treated with I-BET 50 nM compared to DMSO. Log2 fold changes of gene expression were used to identify enriched sets. Red bars correspond to enriched gene sets for upregulated genes and blue for downregulated genes. The enrichment profiles for “Hallmark P53 Pathway” for both cell lines are shown on the right as an example

Fig. 6

Gene expression changes correlate with a significant loss of BRD4 binding at − 1 Kb and with increase in Pol II pausing. a Schematic representation of the analysis performed to correlate the signature of DB with DE. b Boxplots for RNA-seq log2 fold changes for each cluster identified in Fig. 3c as a function of cell type and condition: upper panel for K562-based clustering and lower panel for MV4;11-based clustering. P values correspond to Wilcoxon signed-rank test. c Traveling ratio (TR) based on Pol II ChIP-seq in MV4;11 cells: Each panel represents TR of genes belonging to clusters defined in Fig. 3c in DMSO condition (blue) and in response to 500 nM I-BET treatment (light blue)