NF-κB directs dynamic super enhancer formation in inflammation and atherogenesis

Abstract

Proinflammatory stimuli elicit rapid transcriptional responses via transduced signals to master regulatory transcription factors. To explore the role of chromatin-dependent signal transduction in the atherogenic inflammatory response, we characterized the dynamics, structure, and function of regulatory elements in the activated endothelial cell epigenome. Stimulation with tumor necrosis factor alpha prompted a dramatic and rapid global redistribution of chromatin activators to massive de novo clustered enhancer domains. Inflammatory super enhancers formed by nuclear factor-kappa B accumulate at the expense of immediately decommissioned, basal endothelial super enhancers, despite persistent histone hyperacetylation. Mass action of enhancer factor redistribution causes momentous swings in transcriptional initiation and elongation. A chemical genetic approach reveals a requirement for BET bromodomains in communicating enhancer remodeling to RNA Polymerase II and orchestrating the transition to the inflammatory cell state, demonstrated in activated endothelium and macrophages. BET bromodomain inhibition abrogates super enhancer-mediated inflammatory transcription, atherogenic endothelial responses, and atherosclerosis in vivo.

Figures

Figure 1. p65 and BRD4 Genome Binding During Proinflammatory Activation in ECs

(A) Images of ECs ± TNFα stained for p65 (red) or DAPI (blue) (25 ng/mL, 1 hr) cells. (B) Western blot for p65, Ku-70, and Tubulin in cytosolic (left) and nuclear (right) protein fraction lysates in ECs±TNFα. (C) Images showing adhesion of calcein-labeled THP-1 monocytes to ECs ± TNFα (25 ng/mL, 3 hrs). (D) Bar plots showing cell count normalized expression levels of SELE and VCAM1 in ECs±TNFα (25 ng/mL, 3 hrs). Error bars are standard error of the mean (SEM). (E) Pie chart of p65 binding site distribution in EC genome in TNFα(+). (F) Heatmap of p65 (blue), BRD4 (red) and H3K27ac (yellow) levels in resting ECs and after TNFα (25 ng/mL, 1 hr). Each row shows ± 5kb centered on p65 peak. Rows are ordered by max p65 in each region. ChIP-Seq signal (rpm/bp) is depicted by color scaled intensities. (G,H) Gene tracks of ChIP-Seq signal for p65, BRD4, and H3K27ac at the VCAM1 and TEK gene loci in untreated (top) or TNFα(+) (bottom) ECs. Y-axis shows ChIP-Seq signal (rpm/bp). The x-axis depicts genomic position with TNFα gained typical enhancers (TE, gray) and SEs (SE, red) and promoter regions (white) marked. See also Figure S1.

Figure 2. p65 and BRD4 Establish Super Enhancers During Proinflammatory Stimulation

(A) Ranked plots of enhancers defined in resting (top) or TNFα(+) (bottom) ECs ranked by increasing BRD4 signal (units rpm). Enhancers are defined as regions of BRD4 ChIP-Seq binding not contained in promoters. The cutoff discriminating TEs from SEs is shown as a dashed line. Genes associated with enhancers that are considered typical or super are colored gray and red respectively. (B) Pie charts displaying characteristics of TE and SE regions including number of loci, size and BRD4 signal. (C) Boxplots of median enhancer length (kb), signal (rpm) and density (rpm/bp) in TNFα-gained enhancers. Significance of the difference between distributions determined using a two-tailed t test. ** p < 1e-5, *** p < 1e-10. (D) Boxplot of absolute change in BRD4 signal in response to TNFα measured at all enhancers in TNFα(-) and TNFα(+). Significance of the difference between distributions determined using a two-tailed t test. ** p < 1e-5, *** p < 1e-10. (E) Boxplot of p65 binding signal (rpm) at all active gene promoters (TSS), TEs and SEs in TNFα treated ECs. Significance of the difference between distributions determined using a two-tailed t test. ** p < 1e-5, *** p < 1e-10. (F,G) Schematic of transcription factor motif binding sites at the VCAM1 SE (red box) (F) or TEK TE (grey box) (G) loci in ECs treated with TNFα. (H) Line plots of kinetic ChIP-PCR showing enrichment (% input normalized to time 0) of p65 and BRD4 at an NF-κB binding site in the VCAM1 (left) SE and TEK TE (right) in ECs treated with TNFα (25 ng/mL; 0, 5, 15, 30, 60 min). The effect of co-treatment with vehicle (top), BAY (NF-kB inhibitor, middle) and JQ1 (bottom) are shown. See also Figure S2.

Figure 3. NF-κB Provokes Rapid, Global Redistribution of BRD4

(A,B) Gene tracks of ChIP-Seq signal (rpm/bp) for p65, BRD4, H3K27ac, H3K4me3, and RNA Pol II at the CCL2 gene (A) or SOX18 (B) locus in TNFα(-) (top) and TNFα(+) (bottom) ECs. (C) All genomic regions containing a SE in TNFα(-) and TNFα(+) ECs are shown ranked by log2 change in BRD4 signal (treated vs. untreated). X-axis shows the log2 fold change in BRD4 signal. Change in BRD4 levels at SEs are colored by intensity of change (green to red). (D) Line plot showing the median levels of p65 binding (rpm/bp) at SEs in either TNFα (-) light blue or TNFα (+) dark blue conditions. SEs were ranked by change in BRD4 and binned (50/bin). The median p65 level was calculated in each bin. Error bars represent 95% confidence intervals of the median determined by empirical resampling. (E) Horizontal bar plot showing the ratio of transcription factor motif density between TNFα-gained and TNFα-lost SEs. Twenty-one transcription factors are displayed whose motifs occur more frequently than expected based on dinucleotide background model. The transcription factor motifs are ranked by log2 fold change in density between TNFα-gained vs. TNFα-lost SEs. See also Figure S3.

Figure 4. NF-κB Formed Super Enhancers Drive Proinflammatory Transcription

(A,B) Bar plot of change in elongating RNA Pol II (A) or mRNA expression (B) at genes associated with TNFα-gained, TNFα-lost or TNFα-conserved SEs (red), TEs (gray) and no enhancers (black). Significance of the difference between distributions determined using a two-tailed t test. ** p < 1e-5, *** p < 1e-10. (C, D) Change in elongating RNA Pol II in the gene body region of genes (4C, y-axis) or change in mRNA levels (4D, y-axis) are plotted ranked by change in BRD4 at proximal SEs (x-axis). Dots represent median change sampled across 50 evenly distributed bins with a loess fitted line overlaid. Change in BRD4 levels at proximal SEs are colored by intensity of change (green to red). (E,F) Metagene representations of average RNA Pol II ChIP-Seq signal (grey untreated and black TNFα treated) in units of rpm/bp at a meta composite of target genes of SEs that are gained (E) or lost (F) in response to TNFα treatment. Boxplots of cell count normalized expression levels are shown to the right of each metagene in arbitrary units for genes with associated SEs (grey untreated and black TNFα treated) that are gained (E) or lost (F) in response to TNFα treatment. Significance of the difference between distributions determined using a two-tailed t test. ** p < 1e-5, *** p < 1e-10. (G, H) Table showing the functional categories of selected genes that are targets of SEs gained (G) or lost (H) in response to TNFα treatment. See also Figure S4.

Figure 5. NF-κB Formed Super Enhancers Drive Proinflammatory Gene Expression in a BET Bromodomain-Dependent Manner

(A) Gene tracks of ChIP-Seq signal (rpm/bp) for p65, BRD4, H3K27ac, H3K4me3, and RNA Pol II at the SELE locus in TNFα treated ECs co-treated with vehicle (top) or JQ1 (bottom). (B) The mean log2 fold change in H3K27ac (yellow), BRD4 (red) and p65 (blue) ChIP-Seq signal in TNFα treated cells ±JQ1 at either TEs or SEs gained in response to TNFα treatment. Error bars represent 95% confidence intervals of the mean determined by empirical resampling. (C) Metagene representations of average RNA Pol II ChIP-Seq signal (black TNFα treated and red JQ1 treated) in units of rpm/bp at a meta composite of target genes of SEs gained in response to TNFα treatment. Boxplots (right) show cell count normalized expression levels in TNFα (25 ng/mL, 3 hours) treated ECs ± JQ1. Significance of the difference between distributions determined using a two-tailed t test. ** p < 1e-5, *** p < 1e-10. (D) Line plots of mRNA levels (qRT-PCR) of 3 representative genes associated with TEs (LOX, TEK, NLRP1 in black) and SEs (FS3, CCL2, VCAM1 in red) in response TNFα and JQ1 (50, 100, 250, 500 nM). The mRNA levels from TNFα + VEH (10 ng/mL, 3 hrs) treated ECs were set to 100%. Results displayed as the %reduction from maximum. Error bars represent SEM. Representative results from 2 independent experiments are shown. (E) Gene tracks from Chem-Seq (JQ1) and ChIP-seq (BRD4, H3K27ac) datasets of the SELE SE locus (rpm/bp) for JQ1, BRD4 and H3K27ac from TNFα (-) or TNFα (+) stimulated ECs. (F,G) Scatter plot of JQ1 genome-wide binding levels on x-axis compared to the log2 change in BRD4 (F) or H3K27ac (G) ChIP-Seq signal on y-axis. The change in BRD4 and H3K27ac signal was determined comparing TNFα + JQ1 vs. TNFα. See also Figure S5 and S6.

Figure 6. Phenotypic Consequences of BET Bromodomain Inhibition in Endothelium

(A) Intravital microscopy image (left) and bar plot quantification (right) of leukocyte flux fraction in the cremaster post-capillary venule after TNFα (2 hr, n=7/group) in VEH or JQ1 treated samples. Error bars represent SEM. The statistical significance of the difference between JQ1 treated and VEH treated samples was determined using a two-tailed t test. * p < 0.05. (B) Velocity distribution of leukocytes measured in A. (C) Bar plot showing mean leukocyte velocity in cremaster post-capillary venule in TNFα(+) animals ± BET bromodomain inhibition. Error bars represent SEM using a two-tailed t test. * p < .05. (D,F) Representative fluorescence microscopy images showing adhesion of calcein-labeled THP1 cells to (D) ECs pretreated with JQ1 then activated with TNFα□□□□ as well as (F) TNFα treated ECs after siRNA knockdown of BRD4. (E,G) Bar plots showing quantification of fluorescence from D,F. (H) Bar plots showing quantification of transmigrating neutrophils on TNFα-activated EC monolayers. Results pooled from 3 independent experiments. Data represent mean ± SEM. The statistical significance of the difference between JQ1 and VEH treated samples was determined using a two-tailed t test. * p <.05. (I-L) Line plots of mRNA levels (qRT-PCR) for (I) SELE, (J) VCAM1, (K) CXCL8 and (L) CCL2 measured after stimulation of ECs with TNFα (12.5 ng/mL; 1, 3, 8, 24, 48 hrs) ± JQ1 (500 nM). The statistical significance of the difference in expression between vehicle (VEH) or JQ1 at each time point was determined using a two-tailed t test. *p < 0.05. Data represent mean ± SEM of fold change vs. 0hr. See also Figure S7.

Figure 7. BET Bromodomain Inhibition Suppresses Atherogenesis in Ldlr−/− Mice

(A-D) Photomicrographs of aortic root sections from Ldlr−/− animals treated with VEH or JQ1 stained for (A) oil red o, (B) Mac-3, (C) CD-4, or (D) VCAM1. Quantification of staining is shown below. Results represent mean ± SEM. The statistical significance of the difference between JQ1 and VEH treated samples was determined using a two-tailed t test. * p = .002 for (A); and * p < .05 for (B-D). (E) Oil red o staining of en face aortas prepared from cohort in A-D. (F) Quantification of lesion area (%) between VEH and JQ1 treated en face aortas. The statistical significance of the difference between JQ1 and VEH treated samples was determined using a two-tailed t test. * p < .05. See also Figure S7.