Epigenetic Mechanisms in Heart Failure Pathogenesis

I. Introduction

While neurohormonal antagonist and device therapies have improved outcomes in patients with systolic heart failure (HF), residual morbidity and mortality remain high ^{1, 2}. Novel HF therapeutic approaches thus remain an unmet clinical need of pressing urgency. Such approaches depend, in turn, on keener understanding of the molecular pathways underlying HF pathogenesis. While a number of novel myocardial signaling effectors have been implicated as drivers of HF pathogenesis ^{1, 3}, translating these findings into human therapies has remained extremely challenging.

Gene expression profiling studies in animal HF models ⁴ and in human failing hearts ⁵ consistently demonstrate aberrant gene control in HF. The term “epigenetics” - a fusion of “epigenesis” and “genetics” - was coined over fifty years ago to describe the process of cell fate commitment during development ⁶. Today, the “epigenome” denotes the totality of sequence independent processes that modulate cell-state specific gene expression, e.g., post-translational histone or DNA modifications and non-coding RNA/protein complex interactions with chromatin ^7-10. The epigenome may differ between cell types, drive local formation of higher-order chromatin structures, modulate transcription factor access to DNA and preserve memory of past transcriptional activities ¹⁰.

This review focuses on the chromatin-specific epigenetic regulatory mechanisms that may inform novel therapeutic targets in HF. We specifically highlight examples of chromatin remodeling, biochemical modifications to histones and integrated features of chromatin-dependent signal transduction that are pertinent to cardiac biology. Other epigenetic pathways in HF, including miRNAs, have been extensively reviewed elsewhere ^{11, 12}.

II. Current concepts in eukaryotic gene control

Eukaryotic cell identity or more broadly, “cellular state”, is largely governed by precise spatio-temporal coordination of gene expression ⁹. Pathologic transformation from a normal to a diseased cardiomyocyte (e.g. hypertrophied and hypocontractile) represents a cell state transition driven by defined transcriptional events. Dynamic interplay between accessible DNA sequences, chromatin-binding transcription factors and associated DNA/RNA-binding proteins alter local chromatin structure to orchestrate gene expression programs. To provide a necessary framework for this review, we first briefly summarize some fundamental features of eukaryotic gene regulation ^{9, 13} (Supplemental Table 1).

Chromatin

In the nucleus of every cell, over 1 meter of linear DNA is densely compacted into chromatin ¹⁴, a dynamic macromolecular complex of DNA, RNA and diverse proteins (Figure 1). The fundamental primary unit of chromatin is the nucleosome core particle, comprised of approximately 147 base pairs of double-stranded DNA wrapped in 1.67 left-handed superhelical turns around a histone octamer consisting of 2 copies each of the core histones H2A, H2B, H3 and H4 ¹⁵. The histones within nucleosomes can be dynamically modified and/or exchanged with variants to confer plasticity to primary chromatin structure ¹⁶. Primary chromatin, in turn, is arrayed into dynamic, three-dimensional, higher-order configurations to permit state-specific accessibility of the genome and allow for efficient DNA recombination and DNA repair ^17-19. By vastly expanding the signaling repertoire of the primary DNA template, higher order chromatin structures endow metazoans the remarkable ability to generate diverse, highly specialized cell types from a single genome ^{9, 20}.

Figure 1. Transcriptional Regulation and Histone Marks.

A. Pre-initiation complex formation. Transcription factors bind to specific DNA elements (enhancers) and to coactivators, which bind to RNA polymerase II (Pol I), which in turn binds to general transcription factors at the transcription start site (arrow). The DNA loop formed between the enhancer and the start site is stabilized by cofactors such as the Mediator complex and cohesin.

B. Initiation and pausing by Pol II. Pol II begins transcription from the initiation site, but pause control factors cause it to stall some tens of base pairs downstream.

C. Pause release and elongation. Various transcription factors and cofactors recruit elongation factors such as P-TEFb, which phosphorylate the pause release factors and Pol II, allowing elongation to proceed.

D. Chromatin structure is regulated by ATP-dependent remodeling complexes that can mobilize the nucleosome, allowing regulators and the transcription apparatus increased access to DNA sequences.

E. Transcriptional activity is influenced by proteins that modify and bind the histone components of nucleosomes. Some proteins add modifications (writers), some remove modifications (erasers) and other bind via these modifications (readers). Modifications include acetylation (Ac), methylation (Me), phosphorylation (P), sumoylation (Su), and ubiquitylation (Ub).

F. Histone modifications occur in characteristic patterns associated with different transcriptional activities. Characteristic patterns at actively transcribed genes include H3K27 acetylation (H3K27Ac), H3K4 trimethylation (H3K4me3), H3K79 dimethylation (H3K79me2) and H3K36 trimethylation (H3K36me3).

Figure reproduced with permission from reference ⁹. Legend adapted from figure legend from reference ⁹.

Transcription factors and cis-regulatory elements

As convergence points for multiple pathologic signals in the myocardium, DNA-binding transcription factors (TFs) have been the subject of intense interest. Accumulating evidence implicates a defined set of “master” TFs as capable of controlling the selective transcription of genes by Pol II, thereby governing any given cell state ⁹. Gene targeting and transgenic models have clearly demonstrated that activation of specific DNA-binding TFs (e.g. NFAT, MEF2, NFκB, GATA4, C-MYC) is critical for pathological cardiac remodeling in vivo ^{3, 21-25}. In addition, a number of the TFs hyper-activated in HF are also key determinants of cardiomyocyte lineage (e.g. GATA4, MEF2) ^26-28.

TFs typically regulate gene expression by binding regulatory DNA elements called enhancers, an event which recruits cofactors, the general transcriptional machinery and Pol II complexes to target genes (Figure 1) ^29-31. An active enhancer typically binds multiple transcription factors in a cooperative fashion and regulates transcription from core promoters often via long-range genomic interactions that involve looping of DNA ^{32, 33}. In addition, transcription factors can bind directly to core promoter elements in proximity to transcriptional start sites to recruit transcriptional machinery and regulate gene expression ³⁴.

Epigenetic writers, erasers, readers and their marks

A critical mechanism by which enhancer-bound TFs set the stage for gene control is via the recruitment of co-factors which alter chromatin structure. Two major categories of co-factors are those that mobilize nucleosomes (e.g. the ATP-dependent chromatin remodeling complexes or CRCs) ¹⁴ and those that enzymatically modify histones (histone modifiers) via post-translational modifications (e.g. acetylation, methylation, phosphorylation, and ubiquitylation) ³⁵. Modifiers that add posttranslational modifications (PTMs) or “marks” to histones (e.g. acetyltransferases, methyltransferases, kinases, ubiquitin ligases) have been dubbed “epigenetic writers” ^36-41 (Table 1). Conversely, histone modifiers that remove histone PTMs (e.g. deacetylases, demethylases, phosphatases, de-ubiquitinases) have been dubbed “epigenetic erasers”. Proteins harboring recognition motifs for specific histone PTMs have been dubbed “epigenetic readers”, which generally facilitate locus-specific protein complex formation and signal propagation.

Table 1.

Common Chromatin Writers, Erasers, Readers

Writers* (enzymes)	Members	Substrates	Comments
Histone acetyltransferases	5 HAT classes	Histone lysine residues. Can acetylate numerous non-histone proteins.	Includes p300 and CREB-binding protein (see text)
Histone methyltransferases	KMT1-6 and other members	Histone lysine (e.g., H3K9, H3K36) and arginine residues; Can de-acetylate numerous non-histone proteins.	Site-dependent variable effects on transcription - H3K4me3 and H3K36me3 typically promote transcriptional activation; H3K9 and H3K27 methylations typically promote transcription repression (see text)
DNA methyltrasnferases	DNMT 1, 3A, 3B	Cytosine carbon 5, typically at symmetrical CG dinucleotide residues (CpG)	Critical epigenetic regulatory mechanism including expression silencing via CpG promoter and/or TSS methylation (see text)
Erasers (enzymes)
Histone deacetylases	18 HDACs divided into four HDAC classes I-IV	Acetylated lysine residues	Appealing targets for epigenetic modulators (see text)
Histone demethylases	KDM 1-8 and other members	Histone methylated lysine residues	Site-dependent variable effects on transcription (see Histone methylases above and text)
DNA demethylases	TET family dioxygenases, redox-dependent DNA 5-dehydroxymethylases	Methylated cytosine carbon 5 at symmetrical DNA CG dinucleotide residues (CpG)	Critical epigenetic regulatory mechanism including facilitating transcriptional activation via demethylating promoter and/or TSS (see text)
Reader (domains)
Bromodomains (BRDs)	46 diverse types of human BRD containing chromatin-interacting proteins	Acetyl-lysine (e.g. H3Kac, H4Kac, H2AKac, H2BKac). May also function via recognizing Kac on non-histone proteins.	BRDs are found in HATs, CRCs, helicases, methyltransferases, transcriptional coactivators (TRIM/TIFI, TAF), transcriptional mediators, nuclear scaffolding proteins, BET family “reader” proteins
PHD finger domains	Diverse types of human chromatin regulatory proteins	Histone lysine methylation motifs (e.g., H3K4me3, HkK4me2, HkK9me3), unmodified histones (H3un)	e.g., DMT2A-E subfamily contains 24 PHD fingers
Chromodomains (CRDs)	Diverse types of human chromatin regulatory proteins	Histone lysine methylation motifs (e.g., H3K9me3, H3K9me2, H3K27me3, H3K27me2)	CRDs are found in Heterochromatin protein 1, Polycomb repressive complexes
MBT domains	Human genes L3MBTL1-4, SCMH1-2, SFMBT1-2 encode proteins with PHD finger domains	Histone lysine methylation motifs (e.g., H3Kme1, H3Kme2, H4Kme1, H4Kme2)	Linked to E2F/Rb- and PRC mediated repression
HTD domains	HTDs are found in Jmj family demethylases (e.g., JMJD2A), p53-binding protein 1	Histone lysine methylation motifs (e.g., H3K4me3, H4K20me2, h4K20me3, H3L36me3)	HTD containing proteins play roles in Piwi-RNA pathways, RNA metabolism, DNA damage repair, chromatin remodeling
PWWP domains	Diverse types of human chromatin regulatory proteins	Histone lysine methylation motifs (e.g., H3K36me3, H4K20me1, H4K20me2, H3K79me3)	PWWP domain found in DMT3a

^*Additional histone “marks” include acylation, phosphorylation, ADP ribosylation, β-N-acetylglucosamine glycosylation, ubiquitylation and sumoylation.

Table adapted from references ^36-41.

In aggregate, histone writers, erasers and readers modify local chromatin structure in a stereotypic fashion ¹⁶ that is associated with differential transcriptional activity (Supplemental Table 2) ^{9, 35}. For example, H3K27ac marks active enhancers, H3K36me3 marks actively transcribed gene bodies, and H3K27me3 marks heterochromatic or transcriptionally repressed regions ¹⁰. In addition to histone modifications, DNA itself can be covalently modified via methylation of cytosine, a mark generally associated with transcriptional repression ¹⁰. Collectively, interplay between TFs and dynamic alterations in local chromatin structure allow for genomic signal transduction with spatiotemporal precision ^{9, 20}.

Control of RNA Polymerase II dynamics

TFs and recruited co-regulatory proteins signal via chromatin to control the initiation, elongation and termination activities of Pol II ^42-44. Once recruited to promoters, Pol II forms an initiation complex at the transcriptional start site, transcribes a short distance generating a nascent mRNA of 20-50 nucleotides, and then stalls in the proximal gene body (termed promoter-proximal pausing). Physical association of Pol II with pausing factors such as NELF and DSIF inhibit productive transcriptional elongation ⁴². Release of this paused state is mediated by recruitment of protein complexes such as the positive transcription elongation factor b (P-TEFb, which consists of the kinase CDK9 and the regulatory protein Cyclin T1) and Mediator ⁴⁵. P-TEFb triggers pause release and transcriptional elongation by phosphorylating Pol II (on conserved serine residues in its C-terminal heptapeptide repeat domain) and associated pausing factors ¹³. Control of Pol II pause release and elongation dynamics is increasingly appreciated as a major epigenetic regulatory mechanism of eukaryotic gene control in physiology and disease. P-TEFb activation and Pol II pause release are critically involved in cardiac hypertrophy and heart failure pathogenesis ^{46, 47} as will be discussed in subsequent sections.

Evolving epigenetic regulatory role of noncoding RNAs

Only 1.22% of human genome DNA encodes for protein-coding exons (i.e., coding mRNA) to comprise the known 20,687 protein-coding genes ⁴⁸. Fully 80.4% of the human genome, however, is transcribed at some point in at least one cell type ⁴⁹. Hence the majority of the human genome codes for an ever-expanding array of non-coding RNAs (ncRNAs) including at least 8,801 small RNAs (tRNA, miRNA, snRNA, snoRNA), 9,640 long non-coding RNAs (lncRNAs) and 11,224 pseudogenes ⁵⁰. There are at least 399,124 enhancer-like regions and 70,292 promoter-like regions in the human genome that support dynamic “networks” of TFs and other regulatory elements including ncRNAs. Some ncRNA regulatory mechanisms reported to date include: 1) transcriptional silencing or augmentation by interaction with chromatin associated proteins, 2) interaction with actively transcribed mRNA, 3) direct interaction with DNA itself, 4) nuclear siRNA generation via ncRNA “parent” fragmentation, 5) cytoplasmic miRNA decoy via acting as competing endogenous RNA and 6) interference with cytoplasmic mRNA stability by competition for mRNA stabilization factors ⁵¹. It has been suggested that the ability of ncRNA to form complex chemical structures may confer upon ncRNA a functional repertoire exceeding that of the proteome ⁵². The ever-expanding roles of ncRNAs, particularly lncRNAs and miRNAs, are informing a more complex, finely nuanced model of epigenetic regulation ^{53, 54}.

III. ATP-dependent Chromatin Remodeling Complexes in Cardiac Biology

CRCs harness the energy liberated by ATP hydrolysis to disrupt the DNA-nucleosome electrostatic association thereby “unwrapping” DNA from nucleosomes, shuttling nucleosomes along DNA and/or exchanging or removing nucleosomes ^{14, 55}. CRCs are large macromolecular aggregates, consisting of 9-12 catalytic and regulatory subunits that assemble combinatorially in a cell-state specific manner ⁵⁶. In addition to the ATPase catalytic domains, other CRC domains may “read” covalently modified histones, modulate ATPase activity, and interact with chromatin-associated proteins ⁵⁶. There are four major human CRC families (the SWI/SNF, ISWI, CHD and INO80) which are in turn sub-divided into 24 subfamilies, that can catalyze a broad range of chromatin transformations ^{55, 57}.

The role of the SWI/SNF family of CRCs in cardiac development and hypertrophy has been the focus of detailed investigation ^56-61. The Brg1 (Smarca4) component of SWI/SNF promotes myocyte proliferation by suppressing p57^kip2 expression and preserves the fetal cardiomyocyte state by interacting with histone deacetylase (HDAC) and poly-ADP ribose polymerase (PARP) to repress α-MHC (adult) and activate β-MHC (fetal) ⁶². Brg1 activity is suppressed in adult cardiomyocytes and reactivated during stress to trigger fetal gene induction ⁶². In selected patients with hypertrophic cardiomyopathy, Brg1 remains activated and Brg1 levels correlate with disease severity ⁶². During cardiac development, differential spatiotemporal expression of Brg1-associated subunits occurs and is required for normal heart formation ⁶³. BAF250a, a regulatory subunit of SWI/SNF, plays a key role in cardiac progenitor cell differentiation, and Baf250a ablation in mice results in right ventricular trabeculation defects, ventricular septal defect, persistent truncus arteriosus, reduced cardiomyocyte proliferation and embryonic lethality ⁶⁴. De novo mutations in four different SWI/SWF subunits have been associated with human congenital syndromes that typically display cardiac defects ⁵⁸. As SWI/SNF components interact functionally with cardiogenic TFs such as TBX5, GATA-4 and NKX2-5 ⁵⁸, it is possible that similar interactions occur in the stressed adult myocardium and contribute to HF pathogenesis.

IV. Histone PTMs - Lysine Acetylation

Reversible lysine acetylation was the first histone PTM discovered ^65-67 and has been the most extensively studied. Lysine side chains within the unstructured amino terminal tail of histone proteins are sites of local N-ε-acetylation (Kac) and local histone hyperacetylation at regulatory regions of chromatin is generally associated with transcriptional activation ²⁰. Histone acetylation favors an open chromatin configuration, increases chromatin accessibility to TFs, promotes protein complex assembly and facilitates downstream signal transduction to Pol II ²⁰. Dynamic positioning of Kac is mediated by lysine acetyltransferases (KATs/HATs), which function as epigenetic “writers” and lysine deacetylases (KDACs/HDACs) which function as epigenetic “erasers”. Proteins harboring acetyl-lysine recognition modules, or bromodomains, bind to acetylated histones in a context-specific manner at regions of actively transcribed euchromatin and thus serve as epigenetic “readers” ^{39, 68}. Molecular recognition of acetylated histone by reader proteins promotes assembly of macromolecular complexes that remodel chromatin and regulate transcriptional initiation and elongation ⁶⁹. In addition to histones, lysine acetylation affects proteins of multiple classes including mitochondrial proteins, cytoskeletal proteins, and transcription factors ^{70, 71} and the acetylation of these non-histone targets may also play important roles in cardiac biology ^{72, 73}.

EP300 histone acetyltransferase (Kac writers)

Discovered as a target of the adenoviral E1A oncoprotein, the transcriptional coactivator p300 ⁷⁴ plays broad roles in cellular differentiation, homeostasis, and growth ⁷⁵. p300 contains a lysine acetyltransferase domain capable of acetylating histones and non-histone proteins including transcription factors (e.g. GATA4, MEF2, p53 and p65) ⁷⁶. In cultured cardiomyocytes, p300 protein abundance and activity are elevated after neurohormonal stimulation and are required for GATA4 acetylation and cellular hypertrophy in vitro ^{77, 78}. Genetic studies in mice revealed that the p300 acetyltransferase domain is essential for cardiac development in vivo ⁷⁵. Mice harboring Ep300 germline deletion die between E9-11.5 with multi-organ developmental defects including cardiac abnormalities ⁷⁹. p300 abundance is increased in hypertrophied mouse hearts and in failing human LV tissue and augmented p300 activity is both necessary and sufficient for the development of pathologic cardiac hypertrophy in vivo ⁷³. Specifically, Ep300 haploinsufficient mice that survive into adulthood are protected from pressure-overload mediated hypertrophy while mice with cardiomyocyte-specific p300 overexpression develop dose-dependent pathologic hypertrophy ⁷³. Cardiac p300 acetyltransferase activity is also required for post-infarct LV remodeling in vivo ⁸⁰. Finally, the polyphenol compound curcumin, which has inhibitory activity against p300 acetyltransferase, attenuates pathologic cardiac hypertrophy in vitro and in vivo ⁸¹. Recent epigenomic analyses show that p300 is enriched at cardiac enhancers genomewide in the adult mouse heart ^{46, 82}, suggesting that p300 drives hypertrophic growth via its ability to locally hyperacetylate chromatin and non-histone proteins such as GATA4, MEF2, and CDK9 ^{73, 81, 83, 84}. Taken together, these studies support that excessive activation of p300 acetyltransferase is involved in pathologic cardiac hypertrophy and inhibition of p300 might be a potential therapeutic strategy in HF. While p300 has been the most extensively studied, other proteins with Kac ”writing” capabilities such as CBP and PCAF are also emerging as important regulators of cardiac remodeling ⁸⁵.

HDACs/KDACs (Kac erasers)

There are 18 known mammalian HDACs divided into four major classes: class I (HDAC1, 2, 3, 8), Class IIa (HDAC4, 5, 7, 9), Class IIb (HDACs 6, 10), Class III (the Sirtuin family of NAD-dependent deacetylases Sirt1-7), and class IV (HDAC11) ⁸⁶. A number of these HDACs have been studied in the context of cardiac remodeling and HF pathogenesis using genetically modified mouse models and pharmacological inhibitors ^{46, 85, 87-111} (Table 2). Of these, members of HDAC classes I, IIa, and III have been most extensively studied while very little is known about class IIb and IV HDACs. Here we will focus on class I and II HDACs in the adult heart, noting that the role of class III HDACs (Sirtuins) has been reviewed in detail elsewhere ¹¹².

Table 2.

Therapeutic potential of targeting acetyl-lysine dependent pathways in heart failure

Potential	Comments	Direction for Future Studies Targeting Drug Potential
Inhibition of hypertrophy	Class I HDAC inhibition suppresses pathologic cardiac hypertrophy in vivo; mechanism likely involves both chromatin and non-chromatin dependent effects 87-90.	Creation of more isoform specific HDAC inhibtors. Elucidating the acetylated proteins that are critical targets of various HDACs.
Inhibition of autophagy	The HDAC inhibitor Trichostatin A (TSA) attenuates both load- and agonist-induced hypertrophic growth in vivo and attenuates excessive autophagic flux during pathologic cardiac stress 91.	Defining which HDAC family members play a dominant role in regulation of autophagy. Understanding the extent to which the salutary effects of HDAC are autophagy-dependent. Assessing whether HDAC inhibition affects adaptive autophagy under physiological conditions.
Inhibition of apoptosis	TSA potentiates ischemic pre-conditioning and inhibits apoptosis in rodent infarction models. Activation of certain sirtuin family members reduces oxidative stress-induced apoptosis 92-97.	Understanding the role of HDACs in regulating the mitochondrial acetylome and mitochondrial function (particularly in metabolism, ROS production, and cell death/survival signals).
Inhibition of fibrosis	HDAC inhibition reduces fibrosis during pathologic stress. Mechanisms include suppression of cardiac fibroblast proliferation, cardiac myofibroblast (MF) transformation, and ECM deposition 98-105.	Use of evolving genetic models to study the role of various HDAC genes in fibroblasts. Assessing the ability of HDAC inhibtors to halt or suppress pre-established cardiac fibrosis in vivo.
Inhibition of inflammation	A subset of HDAC inhibitors has anti-inflammatory effects. Some of these effects result from inhibition of innate immune responses within the myocardium as well as effects on primary immune cells. Activation of certain Sirtuins has anti-inflammatory effects 92-95, 106, 107	Use of genetic models to study the role of various HDAC genes in non-cardiomyocytes (e.g. myeloid cells) in experimental models of heart disease.
Improvement in contractility	HDAC4 (class IIa HDAC) has been observed to be localized to sarcomeres where their specific role if currently not well understood. HDAC4 may decrease myofilament calcium sensitivity by promoting deacetylation of muscle LIM protein 85, 108.	Understanding the interactions of HDACs with non-histone proteins and the role of HDACs in regulating myofilament function.
Inhibition of ischemic and reperfusion injury	HDAC inhibitiors such as TSA decrease myocyte hypertrophy, limit collagen deposition and improve cardiac performance in aminal models of myocardial infarction 98, 109, 110.	Defining the appropriate timing and dose of HDAC inhibition post ischemia/MI and whether these compounds can predispose to rupture on aneurysm formation under certain experimental circumstances.
BET Bromodomain inhibition	The small molecule JQ1, when administrered early after stress, blocks pathologic hypertrophy in mice during pressure overload and phenylephrine infusion 46, 111.	Assess whether BET inhibition protects against pathologic remodeling in an expanded set of clinically relevant experimental models (post-MI, established HF, HF with preserved EF, large animal models of HF). Understand the therapeutic window of BET inhibitor drugs in humans (many of which are in early phase cancer trials). Elucidate the gene-specific and tissue specific effects of BRD2, 3, and 4 in cardiovascular biology using genetic models.

Class I HDACs

This subfamily is generally thought to exert pro-hypertrophic effects in adult myocardium. Early insights were gleaned from pharmacologic studies using the hydroxamic acid-based compounds trichostatin-A (TSA) and valproic acid, inhibitors of both class I and II HDAC acetyltrasnferase activity. Since class IIa HDACs are generally thought to repress MEF2 function in a deacetylase-independent manner ^{113, 114} and possess significantly weaker deacetylase activity ¹¹⁵, the dominant effect of such compounds is likely via inhibition of class I HDACs. Hydroxamic acid based HDAC inhibitors attenuate pathologic cardiac remodeling in response to numerous stressors, including pressure overload, neurohormonal excess, genetic abnormalities and myocardial infarction ^{100, 116-118}. An important therapeutic effect of TSA may involve suppression of excess autophagic flux in the hypertrophied heart, although the precise mechanisms underlying this effect are just beginning to be understood ^{91, 119}. The apicidin derivative API-D, which inhibits HDACs 1-3 (albeit with modest in vitro activity against HDAC6) ¹¹⁵ has been shown to have beneficial effects in rodent models of pressure overload ¹²⁰. Future work using next-generation inhibitors with better target-specificity will be required ¹¹⁹ to elucidate the therapeutic potential of HDAC modulation in the heart ^{46, 85, 87-111} (Table 2).

Genetically modified mouse models have provided critical insights into gene-specific effects of class I HDACs in the heart. Cardiomyocyte-specific overexpression of HDAC2 using the Mhy6 (αMHC) promoter results in spontaneous pathologic hypertrophy and decreased activity of the anti-hypertrophic kinase GSK3β ¹²¹. Although mice harboring systemic germline deletion of Hdac2 have perinatal death due to a spectrum of cardiac abnormalities, this phenotype is not cell-autonomous as CM-specific deletion of either Hdac2 or Hdac1 using a Myh6-Cre driver results in minimal developmental defects ¹²². However, cardiac-specific deletion of both Hdac2 and Hdac1 results in a rapid onset postnatal cardiomyopathy, suggesting functional redundancy between these two class I HDACs ¹²². Importantly, mice with cardiac deficiency of either Hdac2 or Hdac1 have comparable hypertrophic responses to pressure-overload or isoproterenol as do control mice, suggesting either of these genes alone is dispensable for pathologic cardiac growth in vivo ¹²². Thus, while HDAC2 overexpression is sufficient to drive pathological hypertrophy, the above genetic studies support that the net effect of hydroxamic acid-based compounds is likely due to inhibition of multiple HDACs, although “off-target” chemical interactions with non-HDAC proteins remain possible.

Mice with cardiac specific deletion of Hdac3 develop severe cardiomegaly associated with early mortality, myocardial lipid accumulation, induction of genes regulating lipid flux and excess activity of the nuclear receptor PPARα, a central regulator of myocardial lipid metablism ¹²². Interestingly, recent studies of HDAC3 in the liver have demonstrated that many of its metabolic effects are independent of its deacetylase activity ¹²³. Transgenic overexpression of HDAC3 in the myocardium increases cardiomyocyte hyperplasia without significant augmentation of cardiac mass ¹²⁴. Germline deletion of HDAC8 in mice leads to perinatal lethality due to abnormalities in skull development ¹²⁵, however a specific role for HDAC8 in the heart has not been described.

Class II HDACs

Studies in genetically modified mouse models have revealed that class IIa HDACs generally act as negative regulators of cardiomyocyte hypertrophy. A major mechanism by which class IIa HDACs function is via repression of MEF2 function, an effect that is independent of deacetylase activity and mediated by physical interaction between the HDAC N-terminal domain and MEF2 ^{113, 114}. Overexpression of HDAC4, 5, or 9 suppresses MEF2 dependent transcription and agonist-induced hypertrophy in cultured cardiomyocytes ^{114, 126, 127}. In response to hypertrophic stimuli, class IIa HDACs are phosphorylated by stress responsive kinases (e.g. CaMK, PKD) on conserved N-terminal serine residues ^{114, 126, 128}. HDAC phosphorylation promotes dissociation of the HDAC-MEF2 inhibitory complex and triggers HDAC nuclear export. These events release tonic repression of MEF2, allowing it to recruit coactivator proteins (e.g. HATs), interact with other TFs (e.g. NFAT, GATA4), and transactivate a pro-hypertrophic gene program ¹²⁹. In this regard, class IIa HDACs function as signal-responsive repressors of MEF2 function and pathologic hypertrophic growth ¹¹⁴. In support of this general mechanism, mice deficient in either Hdac5 or Hdac9 develop exaggerated pathologic hypertrophy in response to pressure overload ^{114, 127}. The relationships between particular stress-activated kinases and downstream HDAC targets can be specific as demonstrated by the selective interaction CaMKII with HDAC4 ^{129, 130}. In addition to regulation by upstream kinases, phosphorylation-independent HDAC4 nuclear export is triggered by stress-activated oxidation of conserved cysteine thiols (Cys-667 and Cys-669) in the C-terminal deacetylase domain, an event which is inhibited by Thioredoxin 1 ¹³¹.

Very little is known about class IIb HDACs in the heart. HDAC6 activity has been shown to be increased in rodent myocardium after pathologic stress ^{132, 133}. A recent study found that the partially-selective HDAC6 inhibitor tubastatin-A was protective effects in a canine model of tachypacing-induced atrial dysfunction ¹³⁴. The role of HDAC10 in the heart is unknown. Further studies of class IIb HDAC function using more selective compounds and genetic models are required.

BET bromodomains (Kac readers)

The body of work highlighted above demonstrates that the interplay between epigenetic “writers” (e.g. HATs) and “erasers” (e.g. HDACs), which dynamically position acetyl-lysine (Kac) on target proteins, is an important mechanism of gene control during pathologic cardiac growth. In contrast, the role of epigenetic “reader” proteins of any type in the heart was not known until very recently. Studies from our group ⁴⁶ and the laboratory of Timothy McKinsey ¹¹¹ have established the importance of BET (Bromodomain and Extra Terminal) family acetyl-lysine “reader” proteins as critical effectors of pathologic cardiac hypertrophy and HF pathogenesis. While pathologic hypertrophy has been associated with increased histone acetylation at regulatory genomic regions ^{135, 136} and increased activity of P-TEFb ⁴⁷, the signaling events linking local chromatin hyperacetylation to Pol II dynamics in the heart remained poorly understood. BETs are a conserved family of proteins consisting of the ubiquitously expressed Brd2, Brd3, Brd4 and the testis-specific BrdT, all of which have two tandem N-terminal Kac recognition domains (bromodomains) that bind acetylated histone. BRD4 has been shown to facilitate transcriptional activation in other cellular systems via interactions with Mediator ¹³⁷ and P-TEFb complexes ^{138, 139}. Furthermore, the CDK9 component of P-TEFb kinase is required for cardiomyocyte hypertrophy in vitro while excessive activation of P-TEFb (via transgenic overexpression of Cyclin-T1 in cardiomyocytes) is sufficient to cause cardiac hypertrophy in vivo ⁴⁷. Therefore, we hypothesized that BETs might be critical in signaling between hyper-acetylated enhancers and Pol II at transcriptional start sites in the stressed heart. Our ability to probe the role of BETs in the heart was enabled by the development of the small molecule JQ1, a first-in-class, potent and selective BET bromodomain inhibitor ⁶⁸. JQ1 competitively inhibits BET bromodomain binding to Kac ^{38, 46}, thereby displacing these reader proteins from acetylated chromatin and suppressing downstream signaling events to Pol II (Figure 2) ^{68, 69}. These studies demonstrated that BET inhibition with nanomolar doses of JQ1 or Brd4-specific siRNAs blocked agonist-stimulated hypertrophy in cultured cardiomyocytes. In adult mice, JQ1 potently inhibited cardinal features of pathologic hypertrophy and HF progression in vivo in response to pressure overload or chronic phenylephrine infusion. Transcriptomic analyses revealed that BETs co-activated several canonical prohypertrophic transcription factors such as NFAT, GATA4, and NFκB to regulate a broad, but specific gene expression program. Epigenomic studies revealed that a major mechanism by which BETs co-activated gene transcription was via recruiting P-TEFb activity to transcriptional start sites and promoting pause release of Pol II. These studies suggested a model in which BET co-activator proteins function as essential signal transducers between activated enhancers (which are bound by pro-hypertrophic TFs and undergo dynamic histone hyper-acetylation) and poised Pol II near transcriptional start sites. In this regard, BETs act as rheostats on stress-activated gene induction in the heart via regulation of P-TEFb and transcriptional elongation (Figure 2) ^{46, 111}. These studies provide an impetus for developing BET inhibitors as investigational therapeutic agents in heart disease. Conditional genetic models will be required to further annotate gene and cell-type specific roles for BETs in vivo.

Figure 2. Bromodomain inhibitors in heart failure.

A) Schematic of BET Bromodomain acetyl-lysine readers in the absence (left) and presence (right) of small molecule BET inhibitor. Panel reproduced from reference ³⁸.

B) Mice subject to cardiac pressure overload (TAC vs. sham, 4 weeks) exhibit Pol II pause release, pathologic gene induction, cardiomegaly, fibrosis and heart failure. BET bromodomain inhibition with JQ1 suppresses transcriptional elongation in the heart and potently blocks progression of pathologic remodeling in vivo. Panel reproduced from reference ⁴⁶.

V. Histone PTMs - Lysine Methylation

Methylation of histones on specific lysine residues is also a major regulator of chromatin state. Several specific histone methylations are each associated with characteristic activating or repressive transcriptional functions (e.g. H3K4me3 marks active promoters, H3K9me3 marks heterochromatin, and H3K4me1 marks poised enhancers) ¹⁰. Locus-specific histone methylation is dynamically regulated by a families of histone methyltransferase and demethylase enzymes exhibiting varying specificity for both histone targets (e.g. H3K4, H3K27) and the types of methyl marks catalyzed (me1, me2, me3) ¹⁰. Proteins harboring methyl-lysine recognition motifs (e.g. chromodomains) function in downstream chromatin-dependent signal transduction ⁴¹. While numerous studies have implicated histone methylation in several developmental and disease contexts ^140-142, our understanding of this mode of chromatin remodeling in cardiac hypertrophy and heart failure is nascent. ChiP-chip studies from Dahl salt-sensitive rat LVH models and failing human hearts revealed dynamic, genomewide changes in H3K4me3 and H3K9me3 marks ¹⁴³. More recently, ChiP-Seq and RNA-Seq studies in adult cardiomyocytes isolated from pressure-overloaded mouse hearts demonstrated dynamic and locus-specific alterations in several histone methyl marks that were correlated with changes in transcript abundance ¹³⁵. The histone demethylase JMJD2A/KDM4A, which catalyzes demethylation of H3K9me3 and H3K36me3 ¹⁴⁴, has been implicated as a positive regulator of pathologic cardiac hypertrophy ¹⁴⁵. JMJD2A protein abundance is increased approximately 7-8 fold in LV tissue from patients with hypertrophic cardiomyopathy and mice with cardiomyocyte-specific deletion of Jmjd2a are partially protected from TAC induced hypertrophy. Conversely, mice overexpressing JMJD2A in cardiomyocytes (at 8-fold higher than control) have preserved baseline LV structure/function but develop exaggerated pathologic hypertrophy after TAC. JMJD2A was shown to induce a specific subset of cardiac genes (such as the prohypertrophic target Fhl1) and was associated with altered H3K9me3 and SRF/myocardin enrichment at this locus. The extent to which JMJD2A effects are FHL1-dependent are unknown. The authors of this study note that JMJD2A manipulation produces a relatively limited profile of differentially expressed transcripts in the mouse heart when assessed by cDNA microarrays ¹⁴⁵. Further cardiac studies using ChIP-Seq for JMJD2A and key histone marks coupled with RNA-Seq in these mouse models will help elucidate such target specificity. This study suggests that inhibition of JMJD2A function in the heart might be a therapeutic strategy in HF.

While H3K4me3 (eniched near promoters) and H3K36me2/3 (enriched within gene bodies and 3’ ends) are associated with actively transcribed genes, H3K27me3 marks are associated with gene and chromatin silencing ¹⁴⁶. The EZH2 subunit of Polycomb repressive complex 2 (PRC2), which catalyzes the H3K27 mono-, di- and trimethylations, has been subject of much interest ^{147, 148}. Human PRC2 consists of a minimum of four core subunits, EZH2, EED, SU(Z)12 and RbAp48 ¹⁴⁹. The C-terminal SET domain of EZH2 is essential for PRC2 transmethylase function and the EED and Su(Z)12 subunits are required for full catalytic activity. EZH2 activity is regulated by a number of mechanisms, including phosphorylation of PRC proteins, variation in subunit isoforms, and interaction with specific histone modifications ¹⁴⁶. The precise mechanisms by which PRC2 mediates chromatin silencing are just beginning to be understood ¹⁴⁹. PRC2-dependent gene silencing is critical to normal cardiac development and postnatal maturation. EZH2 deletion in mouse cardiac progenitors results in hypertrophied adult hearts characterized by fibrotic, non-compacted right ventricles, excessive activation of the homeodomain TF Six1, and derepression of a Six1-dependent skeletal muscle gene program ¹⁵⁰. Importantly, this study supports the contention that epigenetic perturbations during organogenesis can affect disease phenotype during adulthood. In a separate study, EZH2 inactivation in differentiated cardiomyocytes using Nkx2-5^Cre resulted in lethal congenital heart malformations due, in part, to inappropriate spatiotemporal gene control, including derepression of the cell cycle inhibitor INK4A/B and a number of potent TFs that drive non-cardiomyocyte gene expression programs ¹⁵¹. Beyond H3K27, EZH2 methylation substrates may also include key cardiac transcription factors such as GATA4. In cultured cardiomyocytes and fetal mouse hearts, EZH2 binds and methylates GATA4 at lysine 299 resulting in decreased GATA4 function ¹⁵². EZH2 can also bind lncRNAs (e.g., Xist, HOTAIR) which facilitate recruitment of PRC2 to specific chromatin sites ^153-155. Finally, interaction of EZH2 with primary microRNA-208b was shown to regulate aspects of myosin heavy chain isoform switching in the adult mouse heart during pressure overload ¹⁵⁶.

VI. DNA Methylation of Cytosine

Methylation of cytosine carbon 5 (5mC) at symmetrical DNA CG dinucleotide residues (CpG) is an important mechanism of gene regulation, particularly of transcriptional silencing ^{18, 157-159}. 5mC has been shown to play an important repressive role in imprinting, X chromosome inactivation and retrotransposon silencing ^{10, 159}. Reversible DNA methylation occurs at CpG sites by three highly conserved DNA methyltransferase enzymes (DNMTs), DNMT3A and DNTM3B (which catalyzes de novo methylation) and DNMT1 (which maintains 5mC through cell division) ^{159, 160}. The TET family of dioxygenase enzymes catalyzes the oxidation of methylcytosine to hydroxymethylcytosine, an intermediary for demethylation, and along with redox-dependent DNA 5-dehydroxymethylases, play key roles in cytosine carbon 5 demethylation ¹⁶¹. Overall, between 60 to 80% of the roughly 28 million CpG sites in the human genome are methylated ¹⁵⁹. The elucidation of complete human methylomes has established that that 5mC is not simply a stable repressive mark, but rather an epigenetic modification that is dynamically deposited and removed, can exist in non-CpG sequence contexts, and is enriched at bodies of actively transcribed genes ^{10, 159}. CpG methylation at promoters and TSSs is the best-studied methylation pattern to date and is thought to repress gene activity though multiple mechanisms including disruption of TF-DNA interactions, recruitment of repressive methyl-CpG binding proteins, and modulation of other chromatin binding proteins and histone modifications ^{160, 162}. Evidence for several additional non-canonical roles for DNA methylation is evolving ¹⁶⁰. For example, hypermethylation within gene bodies may repress gene expression via regulation of transcriptional elongation and co-transcriptional mRNA processing (e.g., alternative splicing) ¹⁶⁰. The importance of DNA methylation during vertebrate development is underscored by the finding that deletion of DNMT1, DNMT3a or DNMT3b is embryonically lethal in mice ^{163, 164}.

DNA methylation has not been widely studied to date in HF ¹⁶⁵ and available studies are summarized in Table 3 ^{143, 166-169}. Taken together, these studies suggest that: 1) differential DNA methylation occurs between failing and control hearts near selected genes, particularly in promoter regions, in a manner correlating with altered gene expression in HF, 2) these different methylation patterns may provide an opportunity for diagnostic discrimination between failing and control hearts and 3) differential methylation patterns may identify novel genomic loci involved in HF pathogenesis. Genome-wide profiling of the cardiac methylome in experimental systems and human hearts will be particularly informative.

Table 3.

Histone protein and DNA methylation in HF: Human studies

Subjects	Method	Findings	Reference
HF = 4 Control = 4	ChiP-Seq	Histone H3K4me3 and H3K9me3 marks markedly differed between HF vs. controls and mapped to HF-specific pathways	143
HF = 8 Control = 6	MeDIP	Differential promoter region DNA methylation for three angiogenesis-related genes (AMOTL2, ARHGAP24, PECAM1) observed and related to differential expression in HF vs. controls	168
HF = 4 Control = 4	MeDIP ChiP-Seq	Distinct genome-wide promoter, intragenic and gene body CpG island methylation, promoter-region DNA methylation and gene body H3K36me3 patterns observed in HF vs. controls	169
HF = 39 Control = 36	Infinium HumanMethylation 27 platform	CpG island DNA methylation patterns for 20 genes including LY75, ERBB3, HOXB13 and ADORA2A differed in DCM vs. controls	166
HF = 10 Control = 10	MeDIP	57 promoters exhibited differential DNA methylation with greatest differences in AURKB, BTNL9, CLDN5 and TK1 in HF vs. controls	167

VII. Lnc-RNAs

Although the role of lncRNAs in epigenetic regulation during development, differentiation and cancer is established ¹⁷⁰, our understanding of lncRNAs in cardiac development and disease is just beginning ¹⁷¹. A recent study that employed RNAseq in human heart tissue samples from patients with left ventricular assist devices reported that the dominant transcriptomic feature of mechanically unloaded left ventricles was a predominance of differentially expressed lncRNA species, as opposed to mRNAs or miRNAs ¹⁷². The lncRNA “Braveheart” (Bvht) is the best characterized lncRNA in the cardiomyocyte ¹⁷³. Bvht was shown to be required for cardiomyocyte lineage specification from mesoderm and maintenance of cardiac fate in neonatal mouse cardiomyocytes ¹⁷³. Bvht interacts with the SU(Z)12 component of Polycomb 2 repressive complex and Bvht deficiency in cardiac progenitor cells leads to increased enrichment of SU(Z)12 at promoters of genes involved in cardiac differentiation ¹⁷³. Thus, Bvht appears to function during cardiac differentiation via an epigenetic mechanism involving locus specific inhibition of PRC2 activity. In addition to the seminal discovery of Bvht, other lncRNA species have been implicated in cardiac biology. LncRNA natural antisense transcripts ¹⁷⁴ have been shown to modulate expression of troponin I ¹⁷⁵, myosin heavy and light chains ¹⁷⁶, ANP ¹⁷⁷ and ALC-1 ¹⁷⁸. Inappropriate amplification of a noncoding RNA derived from the DMPK 3'UTR CTG triplet has been implicated in the pathogenesis of myotonic dystrophy (DM1), possibly via inappropriate activation of the transcription factor Nkx2.5 ¹⁷⁹. Finally, in a genome wide association study in humans, polymorphisms in the locus encoding the lncRNA ANRIL conferred strong genetic susceptibility for coronary artery disease ¹⁸⁰. ANRIL, a natural antisense transcript of the protein coding INK4b/ARF/INK4a complex, has been shown to interact with PRC1 to modulate H3K27me and transcriptional repression in cancer cells, raising the possibility that related epigenetic mechanisms might be at play in cardiovascular disease ¹⁸¹.

Future Directions and Conclusions

An era increasingly devoted to deciphering the fundamental epigenetic mechanisms in HF pathogenesis has dawned. Leveraging next generation sequencing technologies and contemporary modes of probing the epigenome (ChiP-Seq, Hi-C, TAB-Seq), detailed chromatin state maps and TF cistromes in both animal HF models and human HF myocardial tissue are now becoming possible. When coupled with gene expression profiles (e.g RNA-Seq), these analyses will be essential to understand how alterations in chromatin structure conspire with TFs and noncoding RNAs to drive HF pathogenesis. Such analyses may reveal genomic loci critical for chromatin dependent signaling (e.g. important enhancer regions) whose misregulation confer susceptibility to human heart HF. Use of gene editing technologies, patient-derived cells, and chemical biological approaches will be essential to delineate specific roles for chromatin regulators and critical cis-regulatory genomic elements (e.g. enhancers) in cardiac biology.

Evolving research consistently underscores the rich heterogeneity and functional interrelationship of the genome, epigenome and transcriptome in human HF ¹⁸². Acquired forms of human HF exhibit substantial phenotypic heterogeneity, reflecting complex interactions between environmental stressors, the genome, and the epigenome ¹⁸³. Even familial cardiomyopathies that are known to be caused by single-gene mutations exhibit substantial phenotypic diversity, as evidenced by variable intra- and interfamily expressivity and incomplete penetrance ¹⁸⁴. This clinical heterogeneity is undoubtedly influenced by epigenetic mechanisms that transduce environmental signals and gene-gene interactions that drive differential transcriptional responses in the myocardium. Regulation via ATP-dependent CRCs, histone modifications, DNA methylation and noncoding RNAs is a dynamic process that varies throughout development, physiology, and the various stages of HF pathogenesis. A deeper understanding of these dynamic epigenetic mechanisms may unmask novel HF risk factors and/or identify subclinical states along the disease spectrum, affording earlier diagnostic and therapeutic opportunities.

Deciphering fundamental epigenetic HF mechanisms may also usher in a new era of therapies for established HF. While targeting cardiac transcription has been an area of great therapeutic interest, direct pharmacological modulation of TFs has proven extremely difficult: many of these DNA-binding proteins reside within the nucleus in low abundance and lack structural features readily accessible to small molecules. In addition, the therapeutic window of putative TF modulating drugs is also narrow since many of the TFs hyper-activated in HF (e.g. GATA4, MEF2) are also key determinants of cardiomyocyte identity. The therapeutic manipulation of chromatin regulators, already a burgeoning area in cancer drug discovery and exemplified by the use of small-molecule BET bromodomain inhibitors in experimental HF ^{46, 111}, may afford novel opportunities to target pathological gene control in human HF.

List of Abbreviations

H2A, H3A, etc.

Histone 2A, histone 3A, etc.

5mC

DNA methylcytosine carbon 5

ac

Acetylation mark

ALC-1

Atrial myosin light chain 1

ANP

Atrial natriuretic peptide

BET

Bromodomain and extra terminal, a chromatin binding complex

Brd2-4

Bromodomain 2-4

Brg1

Transcriptional activator protein subunit of SWI/SNF

CaMK

Calcium-calmodulin dependent kinase

CBP

CREB binding protein, a transcriptional activator

ChiP-Seq

Chromatin immunoprecipitation followed by next generation DNA sequencing

C-MYC

Myelocytomatosis viral oncogene, a transcription factor

CpG

Symmetrical DNA CG dinucleotide residues

CRC

Catabolite repression control protein

DMPK

Dystrophia myotonica protein kinase

DNMT

DNA methylyransferase enzyme

DSIF

DRB sensitivity inducing factor

EZH2

Main methyltransferase subunit of Polycomb repressive complex 2

GATA4

GATA binding protein 4, a transcription factor with DNA GATA binding domain

HAT

Histone acetyl transferase enzyme

HDAC

Histone deacetylase enzyme

Hi-C

Chromatin conformation capture, detects chromatin conformational changes

JMJD2A

Histone demethylase enzyme

JQ1

A synthesized BET inhibitor

Kac

Histone acetylation mark

KAT

Histone acetyltransferase enzyme

KDAC

Histone deacetylase enzyme

lncRNA

Long non-coding RNA

me

Methylation mark

MEF2

Myocyte enhancer factor-2

ncRNA

Non-coding RNA

NELF

Negative elongation factor

NFAT

Nuclear factor of activated T-cells

NFkB

Nuclear factor kappa-light-chain-enhancer of activated B cells

p300

Histone acetyltransferase enzyme and transcriptional coactivator

PCAF

Histone acetyltransferase enzyme

PKD

Polycystin, one of the polycystic kidney disease proteins

Pol

II RNA polymerase II

PRC2

Polycomb repressive complex 2

P-TEFb

Positive transcription elongation factor b, a cyclin dependent kinase

PTM

Post-translational modification

SWI/SNF

Switch/sucrose nonfermentable complex, a nucleosome remodeling complex

TAB-Seq

Tet-Assisted Bisulfite sequencing, a modified bisulfite-sequencing DNA method

TBX5

T-box transcription factor 5, contains a DNA T-box binding domain

TF

Transcription factor

TSS

Transcriptional start site