BET Bromodomain Proteins Regulate Transcriptional Reprogramming in Genetic Dilated Cardiomyopathy

The bromodomain and extraterminal (BET) family of epigenetic reader proteins are key regulators of pathologic gene expression in the heart. Using mice carrying a human mutation in phospholamban (PLNR9C) that develop progressive dilated cardiomyopathy (DCM), we previously identified the activation of inflammatory gene networks as a key early driver of DCM. We reasoned that BETs control this inflammatory process, representing a key node in the progression of genetic DCM. Using a chemical genetic strategy, PLNR9C or age-matched wild type mice were treated longitudinally with the BET inhibitor JQ1 or vehicle. JQ1 abrogated DCM, reduced cardiac fibrosis, and prolonged survival in PLNR9C mice by inhibiting inflammatory gene network expression at all disease stages. Cardiac fibroblast proliferation was also substantially reduced by JQ1. Interestingly, JQ1 had profound effects on pathologic gene network expression in cardiac fibroblasts, while having little effect on transcription in cardiomyocytes. Using co-immunoprecipitation, we identified BRD4 as a direct and essential regulator of NFκB-mediated inflammatory gene transcription in cardiac fibroblasts. In this this model of chronic, heritable DCM, BETs activate inflammatory gene networks in cardiac fibroblasts via an NFκB-dependent mechanism, marking them as critical effectors of pathologic gene expression.


INTRODUCTION
Heart failure (HF) is a common and morbid disease. Existing pharmacotherapies have improved clinical outcomes for HF patients, but residual mortality remains exceedingly high with an estimated 5-year survival of approximately 50%.(1) HF often results from dilated cardiomyopathy (DCM), a heterogeneous disease with a strong genetic basis. (2) Characteristic phenotypic features of DCM include ventricular wall thinning, cardiac chamber dilatation, eccentric myocyte hypertrophy, diffuse interstitial fibrosis and reduced systolic function. Concomitant with this negative cardiac remodeling is the activation of a suite of cardiac signaling cascades that result in pathologic gene expression. These dynamic changes in gene expression are coordinately controlled in temporal fashion at different stages of disease. (3) To study the regulation of pathologic gene expression in DCM, we utilized a transgenic mouse model expressing a missense mutation (p.Arg9Cys) in phospholamban (PLN R9C ), a dynamically regulated protein that controls calcium cycling via regulation of the cardiac sarcoplasmic/endoplasmic reticulum calcium adenosine triphosphate (SERCA2a) pump.(4) Abnormal cardiomyocyte calcium homeostasis is a common and proximal mediator of stress-induced cardiac remodeling.(5, 6) PLN R9C mice display altered calcium handling prior to onset of phenotypic disease (preDCM) and ultimately develop progressive DCM with fulminant heart failure culminating in premature death. (4) This inexorable process mimics the disease course of individuals carrying this mutation (4,7,8) and recapitulates the natural history of DCM.
Thus, PLN R9C is a powerful model by which to study the molecular changes that underlie DCM.
We previously demonstrated temporal changes in gene expression in PLN R9C hearts.(3) These findings were similar to those seen in human induced pluripotent stem cells engineered to carry the PLN R9C mutation, (9) and highlight the coordinate integration of multiple signaling networks into a concentrated and temporal stress-response program. Pathologic gene networks activated in HF are associated with dynamic remodeling of chromatin (10,11), including changes in lysine acetylation of histone amino-terminal tails and other chromatin-associated proteins. These acetylated lysine residues can be recognized by epigenetic reader proteins harboring acetyl-lysine recognition domains (bromodomains), which in turn regulate gene transcription. The bromodomain and extraterminal (BET) family of epigenetic reader proteins (BRD2, BRD3 and BRD4) bind to acetylated chromatin via their dual N-terminal bromodomains and co-activate gene transcription by assembling protein complexes that promote pause release of RNA-polymerase. (12,13) BETs have been identified as critical coactivators of pathologic gene expression in rodent models of pressure-overload and ischemia-mediated heart failure (13)(14)(15). Gene expression profiling in these models suggest that BETs exert therapeutic effects via preferential suppression of inflammatory and pro-fibrotic transcriptional programs.
The PLN R9C mouse model features robust activation of a broad array of inflammatory gene networks very early in the disease course, prior to the onset of LV dilation or systolic dysfunction. (3) However, it remains unknown whether activation of this transcriptional program plays a causal role in DCM pathogenesis. Interestingly, the genes and networks upregulated in PLN R9C mice were remarkably similar to those suppressed by JQ1 treatment in other mouse cardiomyopathy models. Therefore, we tested the hypothesis that inhibition of BETs with the small molecule JQ1 could directly suppress pathologic inflammatory gene expression and thereby exert salutary effects in this model of chronic, genetic DCM. Herein, we demonstrate the effectiveness of JQ1 in suppressing inflammatory gene network activation, thus establishing a causal relationship between early pathologic gene expression and DCM progression. We identify cardiac fibroblasts as the chief driver of this process, and substantiate the critical mechanistic link between BRD4 and NFB, establishing BETs as nodal activators of cardiac fibroblasts in DCM.  Histochemistry: Ventricular tissue was fixed in 4% paraformaldehyde, paraffin-embedded and sections cut from apex to base to cover all regions of the LV (see Supplementary Methods for additional details on all histochemical protocols). Nuclei were stained with DAPI (Sigma, D9542), and the extracellular matrix was stained with anti-wheat germ agglutinin (WGA; Molecular Probes, W32464). To quantify fibrosis, sections were stained with Masson trichrome and the ratio of fibrotic area to total ventricular area was measured using a Keyence BZ-X700 microscope.

Mouse
To assess cell proliferation mice were injected intraperitoneally on 2 consecutive days prior to sacrifice with 5-ethynyl-2'-deoxyuracil (EdU) at 25mg/kg body weight in 10% DMSO-PBS at 14-weeks of age (Supplementary Figure 1). Tissue was stained using the Click-iT assay (Thermo-Fisher) and confocal microscopy performed to quantify EdU-positive cells. To identify the non-myocyte cells that were EdU-positive, additional slides were co-stained with the Click-iT assay and either anti-vimentin (Abcam, ab92547), anti-CD31 (Abcam, ab124432) or anti-CD45 (Abcam, ab10558).
Cell Isolation: Myocyte and non-myocyte cell populations were isolated from PLN R9C/+ and age-matched wild type mice treated with either JQ1 or vehicle at 16-weeks of age (n=3/treatment group). Hearts were excised and retrograde coronary perfusion was established via aortic cannulation. (16) The heart was perfused with enzyme buffer (see Supplementary Methods) for 10 minutes. The atria and right ventricle were removed and the LV was minced into small pieces in transfer buffer (perfusion solution + 5mg/mL BSA) and then passed several times through a sterile pipette. The resulting cell suspension was passed through a mesh filter into a 50mL centrifuge tube and incubated for 15 minutes at RT to allow myocytes to pellet by gravity. The pellet was collected as a cell fraction enriched in myocytes. The supernatant from the filtered cell solution was centrifuged at 1500rpm for 5 minutes at 4°C, and then plated in 75mm tissue culture dishes. After 2 hours, dead cells were washed off with 1x PBS and plated cells collected, resulting in a non-myocyte cell population that is enriched for cardiac fibroblasts (Supplementary Figure 2). RNA-seq: RNA was extracted from LV tissue or pooled cardiac fibroblast or cardiomyocyte fractions using TRIzol. Prior to cDNA synthesis, 2 rounds of poly-A selection were performed using oligo dT Dynabeads (Invitrogen). cDNA libraries were constructed from individual mouse samples and RNA-seq libraries were constructed using the Nextera XT DNA Library Preparation Kit (Illumina). Libraries were sequenced on the Illumina platform, then aligned to the mouse reference sequence mm10 using the STAR aligner software(17) and a custom data processing pipeline as previously described. (3) Bioinformatics Pipeline for Pathway Analysis: Genes were defined as differentially expressed if they met ALL of the following 3 criteria: (1) normalized read count >1 copy/million reads in the RNA-seq library, and (2) fold-change >1.33 (up-regulated) or < -1.33 (down-regulated), and (3) p-value <0.001 for the comparison of normalized read counts between experimental samples of interest. To increase stringency, all genes identified as differentially expressed had to meet all of these parameters for all comparisons between n=3 mice in each group (total of 9 comparisons per gene). Differentially expressed genes meeting these parameters were identified using a custom pipeline in R (version 3.0.1). Culled lists of differentially expressed genes were then subjected to bioinformatics analyses using Ingenuity Pathway Analysis (IPA) as described in detail in the Supplementary Methods.
Western blotting and immunoprecipitation: Total protein was extracted using RIPA buffer, quantified, separated by SDS-polyacrylamide gel electrophoresis (PAGE) on 4-12% precast bis-tris gels and transferred to PVDF membranes. Western blotting was performed using antibodies to BRD2, BRD3, BRD4, vimentin, RELA and acetyl lysine-310 (aK310) RELA as described in the Supplementary

Methods.
Immunoprecipitation (IP) was performed targeting BRD4 and the NFB subunit p65/RELA.
Protein G magnetic beads (50μl) were added to the protein slurry and incubated for 3 hours with rotation.
Statistics: Data are presented as mean  standard deviation (SD). For echocardiography data, timedependent between-groups variance was calculated using ANOVA with a model-based fixed-effects standard error method. To estimate sample size, a power calculation was performed using the "pwr" package in R based on our prior echo data in PLN R9C mice.(3) To have 80% power to detect a 50% improvement in FS and LVEDD (a moderate effect size) with 4 groups of mice at a significance level of 0.05, we needed n=12 mice/group. Since PLN R9C mice can die suddenly at any point, we used n=14 PLN R9C mice in each arm. As JQ1 was not expected to alter echocardiographic parameters in WT mice, we reduced these arms to n=7. Tests of the 4 a priori echocardiographic hypotheses were conducted using a Bonferroni corrected p-value for significance (i.e., false discovery rate; FDR) of <0.0125.
Survival data were analyzed using the log-rank test; p<0.05 was considered significant. Betweengroups differences were calculated using a 2-tailed student's t-test for single paired comparisons where indicated; p-values <0.05 were considered significant. A Bayesian p-value was calculated for RNA-seq data as described previously.(3) Fold-change p-values of <0.001 were considered significant for RNA-seq data. For bioinformatics analyses, nominal p-values of <0.05 were considered significant for specific pathways or molecules of interest; p-values were also corrected for multiple hypothesis testing using the IPA software and a FDR <0.01 was considered significant. Please see the Supplementary Methods for more details on the statistical analyses provided by IPA software.

BET inhibition blunts negative cardiac remodeling and improves survival in PLN R9C
PLN R9C mice display robust inflammatory and pro-fibrotic gene network activation before any phenotypic evidence of DCM (preDCM), but it is not clear if this plays a causal role in DCM progression.(3) BET inhibition has been recently shown to suppress this inflammatory program in other cardiomyopathy models. (13,15) BET expression was robust in PLN R9C hearts (Supplementary Figure   3). Therefore, we chose to target BETs in order to test if inhibition of this pathologic inflammatory gene expression program was a primary driver of DCM progression. We injected PLN R9C or age-matched WT mice with the BET inhibitor JQ1 ( Figure 1A). We first assessed the effects of BET inhibition on cardiac structure and function.
Treatment of WT mice with JQ1 had no effect on cardiac structure or function (Figure 1). A separate cohort of mice (n=16) was followed for survival (Figure 2). Mice were treated with JQ1 or vehicle beginning at 8-weeks of age and monitored for mortality (as defined by our IACUC protocol). JQ1 treatment increased lifespan by 10% in PLN R9C mice (p<0.001). These data demonstrate that JQ1 substantially delays disease progression and conveys a survival advantage in a robust model of chronic, progressive DCM.

BETs control a pathologic gene expression program
Next, we assessed the primary transcriptional effects of BET inhibition at a very early stage of disease. We performed RNA-seq on whole LV tissue from PLN R9C mice at the preDCM stage and age-  Table 2).
PreDCM PLN R9C vehicle-treated hearts were heavily enriched for pro-inflammatory genes and pathways (Supplementary Tables 1, 2). Nearly all differentially expressed genes in PLN R9C vehicletreated hearts were upregulated (Figure 3B). Stress-response genes (growth-differentiation factor 15, the natriuretic peptides, fetal myosin heavy chain) were markedly increased in preDCM PLN R9C vehicletreated hearts as were genes involved in TGF signaling (Tgfb2, Tgfb3, Tgfbr2, Ctgf) and the innate immune response (Tlr2, Tlr4; Figure 3C). The pattern of differentially expressed genes in PLN R9C vehicle-treated mice predicted activation of key inflammatory signaling molecules and transcriptional regulators, implicating TGF, NFB, AKT and both the innate and adaptive immune response systems as major early effectors of DCM (Figure 3D, E, Supplementary Table 3).
Treatment with JQ1 strongly attenuated these gene expression changes (Figure 3B, C). Relative to vehicle-treated PLN R9C mice, JQ1 reduced the expression of 663 genes (80.1% of differentially expressed genes). Pathway analysis revealed near uniform suppression of inflammatory gene networks and pro-inflammatory upstream regulators (i.e., TGF, TLRs, and a variety of cytokines) by JQ1 in preDCM PLN R9C hearts (Figure 3E-G). These data demonstrate that BETs strongly and specifically regulate inflammatory gene network activation in preDCM LV tissue, a stage where changes in LV geometry, systolic function or fibrosis have yet to manifest.

Chronic BET inhibition dampens cardiac inflammatory signaling
To study the effects of long-term BET inhibition, mice were treated from 8-to 20-weeks of age with JQ1 or vehicle and RNA-seq performed ( Figure 4A). PCA again demonstrated excellent grouping of mice, and a shift towards WT gene expression in PLN R9C JQ1-treated hearts (Supplementary Figure 6). Longitudinal treatment with JQ1 continued to suppress pathologic inflammatory gene expression, though the effect on metabolic gene expression was only partial at this late stage of disease. Pathways suppressed ( Figure 4D) included fibrosis and Wnt-signaling networks. While many genes and pathways are still dysregulated in PLN R9C JQ1-treated mice, the magnitude of effect remained substantially lower than that observed in PLN R9C vehicle-treated mice, demonstrating the continued effect of JQ1 in late stage DCM (Figure 4E, F). Collectively, these data show that BET inhibition in preDCM mice potently and specifically suppresses pathologic inflammatory and pro-fibrotic gene networks, substantially reduced myocardial fibrosis and blunts negative cardiac remodeling, thus establishing a causal role for inflammatory gene networks in DCM progression.

BET inhibition suppresses fibroblast activation
Given the salutary effects on inflammatory gene network expression and fibrosis, we tested whether BET inhibition had differential effects in the different cardiac cell compartments. We previously showed excess non-myocyte proliferation in PLN R9C mice. (3) To test if BETs play a role in the proliferation of non-myocytes in this model, PLN R9C and WT mice treated with JQ1 or vehicle were treated with EdU prior to sacrifice ( Figure 5A). Proliferating non-myocytes were 3.7-fold more prevalent (p<0.001) in PLN R9C vehicle-treated mice compared to WT (Figure 5B, C). This was reduced 1.9-fold by JQ1 (p<0.001). Additionally, our RNA-seq data showed a strong increase in genes that are characteristic of myofibroblasts in DCM hearts; these were reduced by JQ1 (Supplementary Figure 7). Therefore, we hypothesized that these proliferating cells were cardiac fibroblasts.
Next, to test the cell compartment specific effects of JQ1 on gene expression, we performed RNA-seq on isolated pools of cardiac non-myocytes and cardiomyocytes (Figure 6). JQ1 had a substantial effect on the pattern of gene expression in non-myocytes (Figure 6B), but not cardiomyocytes ( Figure 6E). Similarly, JQ1 drove a significant change in gene expression in PLN R9C non-myocytes ( Figure 6D) but not cardiomyocytes (Figure 6F). Enriched pathways in PLN R9C non-myocytes predominantly consisted of inflammatory gene networks that were activated with progression to DCM, an effect completely eliminated by JQ1 (Figure 6C). By contrast, cellular metabolic pathways (the predominant program altered in DCM cardiomyocytes) all remained enriched in PLN R9C cardiomyocytes after JQ1 treatment ( Figure 6G). Collectively, these data demonstrate that BETs play a critical role in the expression of dynamically regulated inflammatory gene networks in cardiac non-myocytes, and suggest that BET inhibition does not exert its salutary effects on heart function in this model via modulation of adult cardiomyocyte gene expression programs.

BETs interact with NFB in cardiac fibroblasts
The mechanisms by which BETs preferentially suppress inflammatory gene expression in this model of chronic DCM remain unknown. BRD4, the archetypal BET, is necessary for NFB-mediated inflammatory gene expression in models of atherosclerosis, (20) graft-versus-host disease,(21) and cancer. (22)(23)(24) Thus, we investigated if BETs exert their potent effects on non-myocyte gene expression via NFB. Our extensive RNA-seq studies revealed that NFB signaling was strongly upregulated in PLN R9C hearts and isolated cardiac non-myocytes at all disease stages, and genes activated by NFB were strongly upregulated in cardiac non-myocytes (Figure 7A, B).
Acetylation of the RELA subunit of NFB is a well-established marker of increased NFB activity. (25) We identified markedly increased aK310-RELA in PLN R9C hearts with DCM (Supplementary Figure 8). This finding was independent of JQ1 treatment (Figure 7C), suggesting NFB activation occurs upstream of BET action as was previously demonstrated in endothelial cells. (20) Next, we performed co-immunoprecipitation (IP) of BRD4 with RELA and aK310-RELA, demonstrating a clear physical association in PLN R9C mice with DCM, an association that was abolished by treatment with JQ1 ( Figure 7D). IP of RELA confirmed the binding of BRD4 with NFB (data not shown). These data reveal that in this model of DCM, BET-mediated gene expression occurs, at least in part, via coactivation of NFB.

DISCUSSION
We have demonstrated that BET bromodomain inhibition is an effective strategy for disrupting the progression of chronic non-ischemic HF. Our comprehensive transcriptomic analyses revealed that BETs control a pathologic gene expression program that is dominated by inflammatory and profibrotic signaling networks in the heart. Longitudinal treatment with JQ1 delayed the onset and progression of DCM, confirming that this inflammatory milieu plays a central role in DCM progression. Importantly, we discovered that BET inhibition has profound and specific effects on fibroblast activation and gene expression with minimal concomitant effects in cardiomyocytes. Finally, we identified an important mechanistic link between the BET family member BRD4 and NFB in cardiac fibroblasts. Collectively, these data highlight the central role of BETs in DCM pathobiology and define a critical role for BETs in cardiac fibroblast activation. These data confirm BET inhibition and transcriptional regulation as a potential novel therapeutic strategy for HF.
BET inhibition with JQ1 has been shown to reduce myocardial hypertrophy and fibrosis induced by phenylephrine or thoracic aortic banding, and limits infarct size and HF after myocardial infarction. Interestingly, this evolving inflammatory milieu precedes fibrosis, LV remodeling and overt cardiomyocyte pathology in spite of the fact that PLN R9C is a disease caused by a cardiomyocyte-specific gene mutation that alters cardiomyocyte calcium handling.(4) Thus, the inflammatory gene network activation in PLN R9C hearts is a secondary consequence of early cardiomyocyte changes that is triggered by an as-yet undefined mechanism. This left open the question of whether these secondary inflammatory changes were truly a driver of DCM progression or an epiphenomenon. By targeting BET bromodomains with JQ1, we were able to specifically and selectively block inflammatory gene network activation, which substantially inhibited DCM progression, thereby demonstrating a causal role for inflammatory gene network activation in DCM progression. This finding also highlights the potential therapeutic benefit of a strategy targeting transcription in HF.
To extend these findings, we also performed RNA-seq on pooled cardiomyocytes. Interestingly, in WT mice, there were virtually no gene expression changes in response to JQ1. Further, while some gene expression changes were evident in cardiomyocytes from PLN R9C JQ1-treated mice, the vast majority of genes differentially expressed were not substantially affected by JQ1. By way of example, the The mechanism by which BETs exert their effect on inflammatory gene transcription in cardiac fibroblasts is steadily taking shape, though the picture remains incomplete. BETs bind to acetylated lysine residues on histone tails at promoters of active genes. The binding of BRD4, specifically, recruits the positive transcription elongation factor-b complex, which promotes pause release of RNA polymerase II, facilitating gene transcription. (13,27,28) Numerous studies have shown that BRD4 is found at enhancer sequences genome-wide, including in the heart. (13,29) However, under stress conditions, BRD4 is rapidly redistributed to super enhancers (SEs), driving disease-specific gene expression. SEs are cisregulatory elements that permit dense clustering of DNA-bound transcription factors that are associated with exceedingly high levels of transcriptional co-activators, including BRD4.(30, 31) SE activation alters the cellular transcriptome and can permit cell-state transition. (20,32) In the heart, this manifests as pathologic gene expression and the transdifferentiation of quiescent fibroblasts to myofibroblasts. (29) Precisely how this locus-specific flux of BRD4 to DCM-specific SEs occurs has not previously been elucidated in the heart. For the first time in cardiac fibroblasts, we have shown a direct mechanistic link between BRD4 and acetylated RELA, demonstrating that BRD4 exerts its effects on gene expression, at least in part, via NFB. BRD4 has been shown to bind acetylated lysine residues on the p65/RelA subunit of NFB, thus driving gene expression. (22,23,33) In endothelial cells, cytokine treatment triggers reorganization of chromatin activators from basal endothelial cell enhancers to inflammatory SEs, triggering inflammatory gene expression via NFB in a BRD4-dependent manner.(34) Importantly, these authors very elegantly demonstrated that NFB activation and SE binding appears to occur upstream of BRD4 as NFB was still bound to SEs in the absence of BRD4. Hence, JQ1 disrupts the ability of NFB to recruit BRD4 to stress-activated SEs, thus limiting changes in transcription. This likely explains why we did not see a significant decrease in aK310-RELA in PLN R9C JQ1-treated hearts relative to vehicle.
Future research should be aimed at characterizing the upstream activators of NFB and identifying additional cardiac transcription factors that may recruit BRD4 to SEs in activated cardiac fibroblasts.
Our data adds an important element to the BET story in the heart, yet many important questions remain. While this study contributes to mounting evidence that highlights the clear importance of BRD4 in pathologic cardiac gene expression, the roles of BRD2 and BRD3 in the heart have not been defined. In cultured rat cardiac fibroblasts, Brd2 inhibition did result in reduced expression of myofibroblast markers in response to cytokine stimulation, suggesting BRD2 has some role in fibroblast gene expression. (29) Further, in immune cells, BRD2 binds to the chromatin insulator CTCF; this facilitates SE formation, which in turn recruits BRD4, driving gene expression and cell-state transitions via coordinate BRD2- Fundamentally, HF is a disease of dysregulated gene expression, much like cancer. However, the mechanisms are distinct: in cancer, the progressive accumulation of mutations drives pathologic gene transcription that ultimately lead to dysregulated cell growth; whereas in HF, the continuous activation of stress signaling networks converges on the transcriptional machinery to alter the myocardial cellular phenotype, ultimately hampering cardiac function. Current HF treatments that target these systemic stress-signaling networks are effective, but morbidity remains exceedingly high. Consequently, there is an urgent unmet clinical need to directly target intramyocardial processes to improve outcomes. (36) Upstream stress signaling networks converge on BET bromodomain proteins, which integrate these signals into specific and distinct gene expression programs. As such, BET inhibition is a particularly appealing therapeutic strategy that could be both additive to neurohumoral blockade and widely applicable to HF of varying etiologies. These data have significantly extended our understanding of cardiac BET biology and proved the effectiveness of BET inhibition in a model of chronic DCM that closely mimics human HF. Concurrent with ongoing trials in cancer, this knowledge should serve as a springboard for advancing BET and other chromatin-based therapeutic strategies for the treatment of HF.