Dynamic Chromatin Targeting of BRD4 Stimulates Cardiac Fibroblast Activation

Abstract

Rationale: Small molecule inhibitors of the acetyl-histone binding protein BRD4 have been shown to block cardiac fibrosis in preclinical models of heart failure (HF). However, since the inhibitors target BRD4 ubiquitously, it is unclear whether this chromatin reader protein functions in cell type-specific manner to control pathological myocardial fibrosis. Furthermore, the molecular mechanisms by which BRD4 stimulates the transcriptional program for cardiac fibrosis remain unknown.

Objective: We sought to test the hypothesis that BRD4 functions in a cell-autonomous and signal-responsive manner to control activation of cardiac fibroblasts, which are the major extracellular matrix-producing cells of the heart.

Methods and results: RNA-sequencing, mass spectrometry, and cell-based assays employing primary adult rat ventricular fibroblasts demonstrated that BRD4 functions as an effector of TGF-β (transforming growth factor-β) signaling to stimulate conversion of quiescent cardiac fibroblasts into Periostin (Postn)-positive cells that express high levels of extracellular matrix. These findings were confirmed in vivo through whole-transcriptome analysis of cardiac fibroblasts from mice subjected to transverse aortic constriction and treated with the small molecule BRD4 inhibitor, JQ1. Chromatin immunoprecipitation-sequencing revealed that BRD4 undergoes stimulus-dependent, genome-wide redistribution in cardiac fibroblasts, becoming enriched on a subset of enhancers and super-enhancers, and leading to RNA polymerase II activation and expression of downstream target genes. Employing the Sertad4 (SERTA domain-containing protein 4) locus as a prototype, we demonstrate that dynamic chromatin targeting of BRD4 is controlled, in part, by p38 MAPK (mitogen-activated protein kinase) and provide evidence of a critical function for Sertad4 in TGF-β-mediated cardiac fibroblast activation.

Conclusions: These findings define BRD4 as a central regulator of the pro-fibrotic cardiac fibroblast phenotype, establish a p38-dependent signaling circuit for epigenetic reprogramming in heart failure, and uncover a novel role for Sertad4. The work provides a mechanistic foundation for the development of BRD4 inhibitors as targeted anti-fibrotic therapies for the heart.

Keywords: chromatin; fibroblast; heart failure; mass spectrometry; phenotype; signaling.

Figure 1.. JQ1 suppresses TGF-β-induced cardiac fibroblast activation.

A, ARVFs seeded on compressible collagen gel matrices were assayed for gel contraction following treatment with TGF-β₁ (10ng/mL) and/or JQ1 (500nM) for 72 hours. B, Quantification of gel contraction images, reported as percentage contraction; (n=4 plates per condition). Data are presented as mean ± SEM. *P< 0.05 vs. TGF-β alone. Additional two-way ANOVA analysis at the 72hr time point indicated that the TGF-β Factor accounted for 78.04% of variation (P<0.0001), the JQ1 Factor 18.25% of variation (P<0.0001), and the interaction 1.877% of variation (P<0.05). C, ARVFs were treated with TGF-β₁ in the absence or presence of JQ1 for 72 hours prior to fixation for indirect immunofluorescence detection of fibronectin (red) and α-SMA (green). Nuclei were stained with Hoechst 33342 (blue); scale bar = 100μm. D, ARVFs were treated with TGF-β₁ in the presence of a concentration response of JQ1 (10μM to 5nM final concentration) for 72 hours prior to fixation for indirect immunofluorescence detection α-SMA. Each concentration was tested in duplicate wells and the data for α-SMA were normalized to the control wells of the plate to determine percent inhibition of the TGF-β₁ stimulated α−SMA signal. The IC₅₀ for inhibition of α-SMA was determined by non-linear regression of the normalized data. E and F, ARVFs were transfected with negative control siRNA (siCtrl) or siRNAs targeting BRD2, BRD3 or BRD4 and maintained in low serum medium for 48 hours. BET family member mRNA expression was determined by qRT-PCR (E) or immunoblotting (F); n=4 plates of cells per condition, *P < 0.05 vs siCtrl. α-tubulin served as a loading control. G, ARVFs were transfected with the indicated siRNAs and maintained in low serum medium or were treated with TGF-β₁ for 48 hours prior to determination of α−SMA mRNA expression levels by qRT-PCR; n=4 plates of cells per condition, P<0.05 vs. unstimulated siControl. H, ARVFs were transfected with the indicated siRNAs and maintained in low serum medium or were treated with TGF-β₁ for 48 hours prior to determination of BRD4 mRNA expression levels by qRT-PCR; n=4 plates of cells per condition, P<0.05 vs. unstimulated siControl. Additional two-way ANOVA showed the TGF-β Factor accounted for 3.35% of variation (P= 0.0117), the BRD4 Factor 93.24% of variation (P<0.0001), and the interaction <1% of variation (P<0.01). I, ARVFs were transfected with the indicated siRNAs and maintained in low serum medium or were treated with TGF-β₁ for 48 hours prior to indirect immunofluorescence staining of α-SMA; nuclei are marked by Hoechst 33342 staining; scale bar = 10μm. J, Quantification of α-SMA staining; *P<0.05, ***P<0.0001. Additional Two-way ANOVA of all depicted groups showed the TGF-β Factor accounted for 51.08% of variation (P< 0.0001), the BET Factor 11.28 of variation (P<0.0001), and the interaction 13.7% of variation (P<0.0001). When two-way ANOVA was run individually (siCtrl vs siBET, one BET at a time), only the siBRD4 analysis showed that variance was attributed to BET targeting (16.4%, P<0.0001). All data are presented as mean +SEM. Statistical analysis was performed by unpaired t-test (E) or one-way ANOVA with Tukey’s post-hoc test (B, G, H, J) (J, siCtrl group not normally distributed, Kruskal-Wallis test P<0.05).

Figure 2.. JQ1 globally suppresses pro-fibrotic gene expression in cardiac fibroblasts.

ARVFs were treated with TGF-β1 in the absence or presence of JQ1 or vehicle control (DMSO) for 24 hours prior to extraction of RNA for RNA-seq analysis. A, Heat map of genes significantly altered by TGF-β1. Each column represents data from a distinct plate of cells. Color indicates row-normalized expression from +2 (red) to −2 (blue). B, Ingenuity pathway analysis (IPA) was used to determine canonical pathways significantly altered by TGF-β stimulation relative to vehicle control or versus TGF-β + JQ1 treatment. The top five affected pathways for each analysis are reported; three redundant cholesterol biosynthesis pathways were removed from the lower table. C, IPA analysis also indicated TGF-β as the strongest upstream effector molecule in the dataset. Induced genes that led to this determination are displayed (red indicates increased expression following TGF-β treatment). D, Comparison of expression of these same genes in TGF-β-treated versus TGF-β + JQ1 treated cells revealed that JQ1 blocked induction of the majority of TGF-β downstream target genes (blue indicates decreased expression following JQ1 treatment, grey indicates no significant change). Connecting lines in C and D represent IPA prediction of direct TGF-β effects on gene expression in the TGF-β vs unstimulated analysis (C); orange = predicted and measured activation, grey = predicted effect without predicted directionality, and yellow = predicted effect opposite of measured effect in the dataset. Shape key: =cytokine, =enzyme, =G-protein couple receptor,=kinase,=peptidase, =transcription regulator, =transporter, =transmembrane receptor, =other.

Figure 3.. JQ1 suppresses TGF-β-induced expression of protein markers of cardiac fibroblast activation.

ARVFs were treated with DMSO vehicle or TGF-β in the absence or presence of JQ1 for 48 hours. Total protein was subjected to LC-MS analysis, as described in the Methods section. A, Label-free quantification of Periostin protein expression. Box: interquartile range; whiskers: min to max, with n=3 plates of cells per condition; Normalized Spectral Abundance Factor (NSAF). B, Principal component analysis of relative protein abundance revealed clear segregation of each treatment group. Each box represents data from an independent plate of ARVFs. C, Heat map of standardized protein values depicts proteins that were increased in expression upon TGF-β stimulation, and impact of JQ1 treatment on this induction (Benjamini & Hochberg adjusted P-value <0.1). Color indicates row-normalized expression from +2 (red) to −2 (blue).

Figure 4.. JQ1 suppresses pressure overload-induced pro-fibrotic gene expression in cardiac fibroblasts in vivo.

A, Schematic representation of the experiment. B, Heat map of genes significantly upregulated by transverse aortic constriction (TAC) and suppressed by JQ1. Color indicates row-normalized expression from +2 (red) to −2 (blue). C, IPA was used to determine which gene expression pathways were significantly altered in cardiac fibroblasts isolated from mice subjected to TAC isolated versus Sham controls, or in cardiac fibroblasts from TAC +JQ1 treated mice versus TAC alone. The top five affected pathways for each analysis are shown. D, The diagram depicts genes from the IPA ‘hepatic fibrosis pathway’ that were upregulated in cardiac fibroblasts from mice subjected to TAC. E, Comparison of expression of these same genes in TAC versus TAC + JQ1-treated mice revealed that JQ1 blocked induction of the majority of the target genes in this pathway (D and E, blue indicates decreased expression, red indicates increased expression, and gray indicates no change; shape key: =cytokine, =enzyme, =kinase, =transmembrane receptor, =other).

Figure 5.. Chromatin targeting of BRD4 in cardiac fibroblasts correlates with RNA pol II elongation and downstream target gene expression.

A, Schematic representation of the experiment. B, Heat map of BRD4- bound enhancers in TGF-β and unstimulated ARVFs covering 10kb upstream and downstream of the enhancer summit, where enhancers are grouped by increased, decreased or unchanged BRD4 binding in TGF-β stimulated ARVFs compared to unstimulated cells; mean reads per million mapped reads per base pair (RPM/bp). C, Top – a histogram of cumulative BRD4 binding to enhancer/promoters of 4,674 active genes ranked by cumulative BRD4 enhancer abundance in response to TGF-β treatment. Bottom - a second histogram depicts the relative mRNA expression of genes proximal to the BRD4 bound enhancers/promoters. The downstream target gene mRNAs were clustered into bins of 100, ranked left-to-right based on cumulative BRD4 abundance at the enhancer/promoter (above); the data are presented as means ±SEM. D, RNA Pol II binding to gene bodies (top) and promoters (bottom) of the corresponding genes is also shown as a histogram. RNA Pol II-bound genes were clustered into bins of 100, ranked left-to-right based on log2 fold-change of cumulative BRD4 abundance at the enhancer/promoter (above); the data are presented as means ±SEM. E and F, RNA Pol II rpm/bp plotted for 400 genes where TGF-β treatment led to increased BRD4 binding (E), or decreased BRD4 binding (F). G and H, BRD4 rpm/bp plotted for 400 genes where TGF-β treatment led to increased BRD4 binding (G), or decreased BRD4 binding to proximal enhancers (H). The insets magnify gene bodies, highlighting RNA Pol II elongation behavior and BRD4 at intragenic enhancers.

Figure 6.. TGF-β stimulates binding of BRD4 to enhancers and super-enhancers associated with the Sertad4 gene.

A, Top - gene track of the Sertad4 locus showing BRD4 binding to the promoter and six distinct proximal enhancers (E) in unstimulated ARVFs. Bottom – upon TGF-β stimulation for 24 hours, BRD4 binding to all of these sites was increased, reaching a threshold for definition of E1 and E2 as super-enhancers (SE). B, ARVFs were treated with TGF-β for 48 hours in the absence or presence of JQ1 or DMSO vehicle, and Sertad4 mRNA expression was determined by qRT-PCR. Data are presented as mean ± SEM; n=3 plates of cells per condition, P<0.05 vs. vehicle. C, Hockey-stick plots of BRD4-enriched enhancers in unstimulated ARVFs and ARVFs stimulated with TGF-β for 24 hours. SEs, based on a threshold level of BRD4 binding, are boxed. The abundance of BRD4 at E/SE1 and E/SE2 of the Sertad4 locus is indicated. D, Schematic representation of the ChIP-PCR experiment. E - K, BRD4-binding to the indicated Sertad4 gene regulatory elements was significantly increased by TGF-β stimulation, and these increases were blocked by JQ1. Data are presented as mean ± SEM; n=4 plates of cells per condition, P<0.05 vs. (−) TGF-β/(−) JQ1. For B and E - K, statistical analysis was performed by one-way ANOVA with Tukey’s post-hoc test (G, JQ1 group not normally distributed, Kruskal-Wallis test, P<0.05).

Figure 7.. p38 inhibition suppresses recruitment of BRD4 to enhancers and super-enhancers associated with the Sertad4 gene.

A, Schematic representation of mRNA expression experiment. B, Inhibition of p38 (SB203580), but not ERK (PD98059), JNK (SP600125) and calcineurin (CN; cyclosporin A), blunted TGF-β-induced Sertad4 mRNA expression in ARVFs. C, SB203580 also suppressed TGF-β-induced expression of α-SMA. Data for B and C are presented as mean +SEM; n=3–6 plates of cells per condition. *P < 0.05 vs unstimulated, ^#P < 0.05 vs TGF-β alone by one-way ANOVA with Tukey’s post-hoc test (B p38i and C TGF-β/p38i groups not normally distributed, Kruskal-Wallis/Mann-Whitney tests P<0.05). D, Schematic representation of the ChIP-PCR experiment. E, BRD4-binding to the Sertad4 gene regulatory elements was significantly reduced in TGF-β stimulated ARVFs treated with the p38 inhibitor. Values were normalized to input levels are presented as mean +SEM; n=7 per condition. *P < 0.05 vs TGF-β alone by unpaired t-test (E, E3, p38i group not normally distributed, Mann-Whitney test p<0.05).

Figure 8.. Sertad4 knockdown represses cardiac fibroblast activation.

ARVFs were infected with lentiviruses encoding control short-hairpin RNA (shRNA; shCtrl) or two independent shRNAs targeting Sertad4 (shSertad4 [#1] and [#2]). The cells were subsequently treated with TGF-β or DMSO control for 48 hours, and qRT-PCR analysis of Sertad4 (A), α-SMA (B) and periostin (C) mRNA expression was performed. Data are presented as mean +SEM; n=4 plates of cells per condition. *P < 0.05 vs shControl by one-way ANOVA with Tukey’s post-hoc test (A, shSertad4(#1)/TGF-β group not normally distributed, Kruskal-Wallis/Mann-Whitney tests P<0.05). To assess relative effects on α-SMA expression, an additional two-way ANOVA analysis was performed. The TGF-β Factor accounted for 57.27% of variation (P<0.0001), and the Sertad4 Factor accounted for 21.69% of variation (P<0.0001), while the interaction between TGF-β and Sertad4 accounted for 9.21% of variation (P<0.01). For periostin expression, the TGF-β Factor accounted for 34.95% of variation (P<0.0001), the Sertad4 Factor for 33.06% of variation (P<0.0001), and the interaction for 15.01% of variation (P<0.01). (D) A model for BRD4-mediated regulation of cardiac fibroblast activation. Cardiac stress signals, including TGF-β, stimulate p38 and likely other pathways to target BRD4 to gene regulatory elements, resulting in RNA Pol II elongation and expression of downstream targets, including Sertad4, which promote fibroblast activation.