|Home | About | Journals | Submit | Contact Us | Français|
In patients with heart failure, reactivation of a fetal gene program, including atrial natriuretic peptide (ANP) and brain natriuretic peptide (BNP), is a hallmark for maladaptive remodeling of the LV. The mechanisms that regulate this reactivation are incompletely understood. Histone acetylation and methylation affect the conformation of chromatin, which in turn governs the accessibility of DNA for transcription factors. Using human LV myocardium, we found that, despite nuclear export of histone deacetylase 4 (HDAC4), upregulation of ANP and BNP in failing hearts did not require increased histone acetylation in the promoter regions of these genes. In contrast, di- and trimethylation of lysine 9 of histone 3 (H3K9) and binding of heterochromatin protein 1 (HP1) in the promoter regions of these genes were substantially reduced. In isolated working murine hearts, an acute increase of cardiac preload induced HDAC4 nuclear export, H3K9 demethylation, HP1 dissociation from the promoter region, and activation of the ANP gene. These processes were reversed in hearts with myocyte-specific deletion of Hdac4. We conclude that HDAC4 plays a central role for rapid modifications of histone methylation in response to variations in cardiac load and may represent a target for pharmacological interventions to prevent maladaptive remodeling in patients with heart failure.
Chronic heart failure affects approximately 5 million patients in the United States and carries a poor prognosis (1, 2). An important risk factor for development of heart failure is LV hypertrophy, which occurs in response to neurohormonal activation and increased filling pressures of the LV (1, 2). Deterioration of cardiac function during heart failure progression is associated with activation of genes that trigger structural, functional, and electrical remodeling of the heart. In particular, the reactivation of a set of fetal genes, including atrial natriuretic peptide (ANP; also known as Nppa) and brain natriuretic peptide (BNP; also known as Nppb), correlates well with the clinical severity and prognosis of the disease (3–5). The mechanisms that regulate the reactivation of this fetal gene program, however, are incompletely understood.
In the nucleus, DNA is packed into chromatin. The basic building block of chromatin is the nucleosome, consisting of an octamer of 4 core histone proteins: H2a, H2b, H3, and H4. In a condensed chromatin formation, the DNA is hardly accessible for transcription factors; thus, gene expression requires nucleosome unfolding (6). A key role in regulating chromatin structure is played by histone acetylation and methylation. While acetylation of histones — for example, at lysines 9 and 27 of histone 3 (H3K9 and H3K27, respectively) or at lysine 91 of histone 4 (H4K91) — induces relaxation of chromatin, methylation can either facilitate or repress gene expression (7–10). For instance, methylation of H3K4 is associated with active genes, whereas di- and trimethylation of H3K9 (H3K9me2 and H3K9me3, respectively) occur primarily in silenced genes by creating a binding site for heterochromatin protein 1 (HP1) (9–11). It was proposed that by these and other chromatin modifications, a multifactorial “histone code” is established that governs the accessibility of DNA for transcription factors and, thus, gene expression (12).
Histone acetylation is regulated by histone acetyltransferases (HATs) and deacetylases (HDACs) (8, 13). Class II HDACs (HDAC4, HDAC5, HDAC7, and HDAC9) are highly expressed in the heart; are signal responsive; limit cardiomyocyte growth and hypertrophy; and suppress transition to heart failure (8, 14, 15). In response to GPCR activation, Ca2+/calmodulin-dependent kinase II (CaMKII), protein kinase D, and reactive oxygen species mediate the export of HDAC4, HDAC5, and HDAC9 from the nucleus to the cytosol by targeted phosphorylation and/or oxidation (16–19). Nuclear class II HDACs repress gene transcription by interacting with myocyte enhancer factor 2 (MEF2), and it is perceived that at least part of its repressive function is related to keeping the respective promoter region in a deacetylated state. Nucleo-cytoplasmic shuttling relieves this repressive function (8, 16, 20, 21). Despite evidence that HATs and HDACs regulate cardiac hypertrophy, little is known about the actual epigenetic modifications induced by these enzymes, especially in the human failing heart or in response to changes in hemodynamic load.
In contrast to acetylation, histone methylation — governed by histone methyltransferases SUV39H1, G9a, and others (9, 22) — has long been considered to be a more permanent epigenetic mark (12). However, the discovery of histone demethylases — i.e., lysine-specific demethylase 1A (LSD1; ref. 23) and the family of JMJC domain–containing proteins (24) — substantially shifted this paradigm (9, 25). In fact, recent reports indicate that histone methylation is dynamically regulated in inflammatory and metabolic disorders (26–28). Furthermore, differential methylation patterns for H3K4 and H3K9 occur in the vicinity of various gene clusters of failing human hearts (29). Recently, an important role for JMJD2A (also known as KDM4A) in regulating cardiac hypertrophy in response to long-term pressure overload was revealed, governing H3K9 methylation status in the promoter region of the four-and-a-half LIM domains 1 (FHL1) gene (30). Although in that study, deletion of JMJD2A also reduced the expression of ANP and BNP after pressure overload in vivo, it is unclear whether this was the result of direct epigenetic modifications in the promoter regions of these fetal genes or an indirect consequence of the reduction of cardiac hypertrophy through FHL1-related signaling and/or hemodynamics.
The present study is the first in-depth analysis of epigenetic modifications in the promoter regions of ANP and BNP in human failing and nonfailing myocardium. Despite pronounced nuclear export of HDAC4, histone acetylation was not required for ANP and BNP gene activation. In contrast, H3K9 demethylation was closely associated with reactivation of fetal genes and nuclear export of HDAC4. Experiments in genetically modified mice revealed that nuclear HDAC4 export was required for H3K9 demethylation, HP1 dissociation, and ANP gene activation in response to increased hemodynamic load. Protein-protein interaction of HDAC4 with the histone methyltransferase SUV39H1 was disrupted upon nuclear export of HDAC4 in response to targeted phosphorylation by CaMKIIδB, representing a potential underlying mechanism. Furthermore, upregulation of H3K9-specific JMJC domain–containing demethylases may sustain these epigenetic modifications in the terminally failing human heart.
Human LV myocardium was obtained from patients with terminal heart failure due to ischemic cardiomyopathy (ICM) or dilated cardiomyopathy (DCM) and compared with nonfailing donor hearts (see Supplemental Table 1; supplemental material available online with this article; doi: 10.1172/JCI61084DS1). Since no differences between ICM and DCM were detected in most subsequent analyses, unless indicated otherwise, these groups were merged into a single group referred to herein as failing myocardium. As described previously (3–5), ANP and BNP were upregulated in failing versus nonfailing myocardium, with a close correlation between both natriuretic peptides (Figure (Figure1,1, A–C). Furthermore, mRNA expressions of α–myosin heavy chain (α-MHC; also known as MYH6) and sarcoplasmic reticulum Ca2+ ATPase (SERCA2a; also known as ATP2A2) were downregulated (Figure (Figure1,1, D and E), both of which are characteristic hallmarks of human heart failure (3–5). In agreement with a previous study (5), an inverse correlation was observed between BNP and SERCA2a expression (Figure (Figure11F).
Since ANP and BNP expression increase in response to elevated hemodynamic load, predicting clinical severity and prognosis in patients with heart failure, we analyzed histone acetylation and methylation patterns in the promoter regions of these genes by ChIP (Figure (Figure2A2A and Supplemental Figure 1, A and B). The promoter regions of the housekeeping (and thus active) gene GAPDH served as an internal control. In nonfailing myocardium, the promoter regions of ANP and BNP showed less acetylation at H3K9 and H3K27 (H3K9ac and H3K27ac, respectively) than those of GAPDH (Figure (Figure2,2, A–D). While H3K9ac was unchanged (Figure (Figure2,2, A and B), H3K27ac was modestly increased in the BNP (but not ANP) promoter region of failing versus nonfailing myocardium (Figure (Figure2,2, C and D), in favor of a slightly more relaxed chromatin structure. H4K91ac, however, was similar in GAPDH, ANP, and BNP promoter regions and was not affected in failing myocardium (Supplemental Figure 2A).
Another epigenetic mark of active genes is methylation of H3K4, whereas H3K9me2 and H3K9me3 induce gene silencing (10). In nonfailing myocardium, H3K9me2 and H3K9me3 were more pronounced in the promoter regions of ANP and BNP compared with GAPDH (Figure (Figure3,3, A–C). In failing myocardium, however, H3K9me2 and H3K9me3 were reduced in ANP and BNP promoter regions, favoring an open chromatin formation (Figure (Figure3,3, A–C). In contrast, H3K4me2 and H3K4me3 did not change in failing versus nonfailing myocardium (Supplemental Figure 2, B–D). Since H3K9me3 is a binding site for HP1 (11), we analyzed HP1 binding to the promoter regions of ANP and BNP by ChIP assays. In agreement with demethylation of H3K9, HP1 binding was reduced in failing versus nonfailing myocardium (Figure (Figure3,3, D–G).
The changes of H3K9 methylation status alone did not correlate to the variations in ANP and BNP expression. However, since the combination of various epigenetic marks may determine a histone code governing accessibility of the DNA for transcription factors (12), we determined the ratio of those marks that were (at least partially) modified in human failing myocardium: namely, H3K27ac (Figure (Figure2D)2D) divided by H3K9me2 (Figure (Figure3B).3B). This ratio was increased in both ANP and BNP promoter regions of failing myocardium and correlated with increased ANP and BNP expression (Figure (Figure3,3, H–K).
Since changes in histone methylation were more prominent than changes in histone acetylation, we set out to analyze the underlying mechanisms for H3K9 demethylation in failing hearts. Methylation status is balanced by activities of histone methyltransferases (G9a and SUV39H1) and histone demethylases (LSD1; JMJD1, also known as KDM3A; and JMJD2) (9, 22–25). While mRNA expression of G9a, SUV39H1, and LSD1 was unchanged (Supplemental Figure 3), expression of JMJD1A, JMJD2A, and JMJD2B was 2- to 2.5-fold upregulated in failing versus nonfailing myocardium (Figure (Figure4,4, A–C). Upregulation of JMJD1A correlated positively with increased ANP and BNP expression and inversely with H3K9me2 in the promoter region of ANP (Figure (Figure4,4, D and E). ChIP assays revealed that both JMJD1A and JMJD2A bound to the promoter regions of ANP in nonfailing and failing myocardium of patients with ICM or DCM (Figure (Figure4F).4F). An increased recruitment to the ANP promoter region was detected for JMJD2A in myocardium of patients with ICM, but not patients with DCM, while JMJD1A recruitment was unchanged (Figure (Figure4,4, F–H). ChIP assays with LSD1, SUV39H1, and G9a indicated that direct binding of these histone-modifying enzymes to the promoter regions was weak and not differentially regulated between failing and nonfailing myocardium (data not shown).
Recruitment of G9a, SUV39H1, and LSD1 to the DNA requires other regulatory proteins, such as REST (for G9a, SUV39H1, and LSD1) and/or HDAC4 (for HP1 and possibly SUV39H1), both of which have been associated with regulation of fetal genes in the heart (8, 14, 15, 31, 32). However, we did not observe any regulatory role for REST in human failing myocardium (see Supplemental Results and Discussion and Supplemental Figure 4). Thus, we focused on a possible role of HDAC4 for histone methylation. Under in vitro conditions in cellular expression systems, H3K9 methylation is governed by a corepressor complex consisting of HDAC4 and HP1 (33), and HP1 was previously shown to interact with SUV39H1 (34). Since HDAC4 is selectively targeted by CaMKIIδB (17), and CaMKII activity is upregulated in human failing hearts (35), we sought to determine whether H3K9 demethylation (Figure (Figure3,3, A–C), HP1 dissociation from ANP and BNP promoter regions (Figure (Figure3,3, D–G), and ANP and BNP upregulation (Figure (Figure1,1, A and B) could be related to nuclear export of HDAC4. While similar amounts of HDAC4 were detected in the cytosolic and nuclear protein fractions of nonfailing myocardium, a 4-fold shift of the cytosolic/nuclear ratio of HDAC4 protein expression was observed in failing myocardium (Figure (Figure5,5, A–C). This nucleo-cytoplasmic shuttling of HDAC4 correlated inversely with H3K9me2 levels in the promoter regions of ANP and BNP and positively with ANP and BNP expression (Figure (Figure5,5, D and E).
Taken together, these findings suggest that in human failing myocardium, H3K9 demethylation, rather than increased histone acetylation, in the promoter regions of ANP and BNP is associated with upregulation of these fetal cardiac genes. Furthermore, our data suggest that nuclear export of HDAC4 and/or upregulation of JMJD1A, JMJD2A, and JMJD2B could be potential underlying mechanisms for H3K9 demethylation.
To determine whether changes in JMJD1A and JMJD2A expression affect gene activation of ANP and BNP, we performed experiments in isolated rat neonatal and adult cardiac myocytes. As expected, mRNA expression of ANP and BNP was 80- and 7-fold higher, respectively, in neonatal versus adult rat cardiac myocytes (Figure (Figure6A).6A). Interestingly, JMJD1A and JMJD2A expression was also increased 4- and 9-fold, respectively, in neonatal versus adult cardiac myocytes. Accordingly, expression of a known target gene of JMJD2A, FHL1 (30), was also increased 7-fold, while SERCA2a expression was reduced 3-fold. In agreement with elevated JMJD1A and JMJD2A expression, H3K9me2 and H3K9me3 were reduced in the ANP and BNP promoter regions in neonatal versus adult cardiomyocytes (Figure (Figure66B).
Downregulation of either JMJD1A or JMJD2A by siRNA in neonatal cardiac myocytes led to compensatory upregulation of the other JMJD isoform and had no effect on ANP or BNP expression (data not shown). In light of this finding, and because both JMJD1A and JMJD2A were upregulated in concert in human heart failure, we applied simultaneous JMJD1A and JMJD2A siRNA (Figure (Figure6C).6C). This JMJD1A and JMJD2A double knockdown (by 20% and 45%, respectively; Figure Figure6D)6D) resulted in moderate but consistent downregulation of ANP and BNP expression (by 13% and 15%, respectively; Figure Figure6E).6E). The degree of ANP and BNP downregulation correlated with the degree of JMJD1A knockdown (r = 0.39 and r = 0.54; P < 0.05 and P < 0.005, respectively) and JMJD2A knockdown (r = 0.43 and r = 0.67; P < 0.05 and P < 0.0005, respectively). Accordingly, H3K9me2 and H3K9me3 in the promoter regions of ANP and BNP increased after JMJD1A and JMJD2A double knockdown (Figure (Figure6,6, F–H). These data indicate that varying expression of JMJD1A and JMJD2A is sufficient to affect ANP and BNP gene expression through variations of H3K9me2 and H3K9me3 status in these promoter regions in cardiac myocytes.
To determine whether JMJD1A and JMJD2A upregulation is not only sufficient, but also necessary, for ANP upregulation during cardiac hypertrophy, we performed thoracoaortic constriction (TAC) in C57BL/6NCrl mice for 6 weeks. Despite an increase in heart weight/BW ratio and robust upregulation of ANP after TAC, expression of JMJD1A and JMJD2A was unaltered (Supplemental Figure 5, A–D). In contrast, nucleo-cytoplasmic shuttling of HDAC4 was observed after TAC (Supplemental Figure 5E). Together with the data in cardiac myocytes (Figure (Figure6),6), these data indicate that although variations in JMJD1A and JMJD2A expression are sufficient, they are not necessary for ANP regulation in the heart.
In the heart, expression of ANP and BNP correlates with LV wall stress, and cardiac pre- and afterload are both elevated in patients with heart failure. To determine whether hemodynamic changes directly induce epigenetic remodeling, and to further elucidate the underlying regulatory mechanisms, isolated mouse hearts were exposed to physiological pre- and afterload (10 and 80 mmHg, respectively) in a working heart apparatus. After equilibration, either preload was elevated from 10 to 30 mmHg or afterload was elevated from 80 to 120 mmHg for up to 90 minutes (Figure (Figure7,7, A–C). These hearts were compared with a control group with maintained physiological load conditions.
Elevation of either preload or afterload increased expression of ANP after 30 minutes, with sustained expression after 60 and 90 minutes, respectively (Figure (Figure7,7, D and E). This was paralleled by nucleo-cytoplasmic shuttling of HDAC4, with a correlation between ANP upregulation and nuclear HDAC4 export (Figure (Figure7,7, F–I). In contrast, neither mRNA expression of JMJD2A nor binding of JMJD2A to the ANP promoter region changed after elevation of preload (Supplemental Figure 6). Despite nuclear export of HDAC4, H3K9ac and H3K27ac did not change in response to elevated pre- or afterload even after 90 minutes (Supplemental Figure 7). In contrast, H3K9me2 and H3K9me3, but not H3K27me3, decreased in response to elevated preload after 60 minutes (Figure (Figure8,8, A and C–E, and data not shown). Accordingly, HP1 dissociated from the ANP promoter region after an increase of preload (Figure (Figure8,8, B and F). Similar changes were observed after an increase of afterload, although these changes were slightly less pronounced (Supplemental Figure 8). These data indicate that an increase in cardiac preload induces rapid nucleo-cytoplasmic shuttling of HDAC4 and H3K9 demethylation in the promoter region of ANP, with subsequent HP1 dissociation and gene activation.
To test whether HDAC4 is causally involved in H3K9 demethylation in the ANP and BNP promoter regions in response to variations of workload, we used mice with cardiomyocyte-specific deletion of Hdac4 (referred to herein as HDAC4-KO mice) and WT littermate controls. Western blot experiments confirmed the absence of HDAC4 in LV myocardium of HDAC4-KO mice (Figure (Figure9A).9A). Under physiological load conditions, these mice had similar LV systolic and diastolic function during working heart analysis (Figure (Figure9,9, B and C). In WT mice, an increase in preload induced nucleo-cytoplasmic shuttling of HDAC4 (Figure (Figure9D).9D). In HDAC4-KO mice, baseline H3K9 and H3K27 methylation in the ANP promoter region was reduced compared with WT mice, while H3K9ac and H3K27ac were unchanged (Figure (Figure9E9E and Supplemental Figure 9). After elevating preload, H3K9me2 and H3K9me3 paradoxically increased in HDAC4-KO mice, while they decreased in WT mice (Figure (Figure9,9, E–G). There was also a trend toward increased H3K27me3 in HDAC4-KO mice (Figure (Figure9H).9H). In contrast, H3K9ac and H3K27ac were not upregulated in WT or HDAC4-KO mice after an increase in preload (Supplemental Figure 9). In agreement with the divergent preload-induced effects on H3K9 methylation, HP1 binding to the promoter region of ANP decreased in WT mice, but increased in HDAC4-KO mice (Figure (Figure9,9, I and J). Under basal conditions, expression of ANP was 6-fold elevated in HDAC4-KO versus WT hearts, and preload-induced upregulation of ANP was completely blunted in HDAC4-KO compared with WT mice (Figure (Figure9K).9K). Taken together, these data indicate that HDAC4 plays a causal role in regulating the H3K9 methylation status of ANP promoter and gene expression in an HP1-dependent manner.
To further investigate the underlying mechanism by which HDAC4 regulates H3K9 methylation, we cotransfected COS cells with GFP-tagged HDAC4 and FLAG-tagged SUV39H1, using a secondary antibody coupled to Texas Red, in the absence and presence of a constitutively active mutant of CaMKIIδB (CaMKIIδB-T287D), which is known to selectively target and phosphorylate HDAC4 (17). Under control conditions, both HDAC4 and SUV39H1 were located primarily in the nucleus (Figure (Figure10,10, A–C). The yellow staining visible in the merged image (Figure (Figure10A)10A) suggests colocalization of both proteins. In response to CaMKIIδB-T287D, HDAC4 was exported from the nucleus to the cytosol and appeared in a characteristic punctuate pattern (Figure (Figure10,10, A and B). In contrast, SUV39H1 remained in the nucleus, but its distribution changed from a homogeneous pattern under control conditions to a more punctuate one in response to CaMKIIδB-T287D (Figure (Figure10,10, A and C). Accordingly, colocalization between HDAC4 and SUV39H1 was disrupted by CaMKIIδB-T287D (Figure (Figure1010A).
To test whether HDAC4 physically interacts with SUV39H1, we cotransfected HEK293T cells with Myc-tagged HDAC4 and FLAG-tagged SUV39H1 and performed co-IP experiments. Under control conditions, we observed protein-protein interaction between HDAC4 and SUV39H1, which was reduced in response to CaMKIIδB-T287D (Figure (Figure10,10, D and E). These data indicate that HDAC4 physically interacts with SUV39H1 in the nucleus and that this interaction is disrupted after CaMKIIδB-induced phosphorylation of HDAC4. Together with our findings in HDAC4-KO mice (Figure (Figure9),9), these data suggest that preload-induced nucleo-cytoplasmic shuttling of HDAC4 induces H3K9 demethylation in the ANP and BNP promoter regions, possibly through disruption of the interaction between HDAC4 and the histone methyltransferase SUV39H1.
Despite a plethora of knowledge on the roles of histone acetyltransferases and deacetylases in regulating cardiac hypertrophy and failure (8, 14–21), surprisingly little information is available on their actual ensuing epigenetic modifications. This is the first study on human myocardium that comprehensively analyzes histone modifications in the promoter regions of genes that are activated during cardiac remodeling in patients with heart failure, i.e., ANP and BNP (3–5). Unexpectedly, despite pronounced nuclear export of HDAC4 in failing hearts, no major changes in histone acetylation were observed in the promoter regions of these genes. In contrast, substantial H3K9 demethylation and HP1 dissociation occurred, predicting an open chromatin formation (7–10, 12). Our data in human myocardium and genetically modified mouse hearts and cell systems indicated that as a central mechanism, nucleo-cytoplasmic shuttling of HDAC4 occurs early after an increase of hemodynamic load and accounts for H3K9 demethylation, HP1 dissociation from the promoter region, and ANP gene activation. Under basal conditions, HDAC4 was associated with the histone methyltransferase SUV39H1, and this association was disrupted after CaMKIIδB-induced phosphorylation of HDAC4 with subsequent nucleo-cytoplasmic shuttling. Although this does not prove a causal relationship, we propose that HDAC4 may function to recruit SUV39H1 to the promoter regions of ANP and BNP, increasing H3K9 methylation status, and that this recruitment is lost with signal-responsive nuclear export of HDAC4. Furthermore, increased expression of H3K9-specific JMJC domain–containing demethylases may sustain these epigenetic modifications in the terminally failing human heart. These findings are summarized in Figure Figure1111.
Although histone methylation was long considered to be an irreversible marker for silenced genes (12), recent reports revealed that histone demethylation can dynamically regulate gene expression. For instance, at 4 hours after LPS treatment, activation of proinflammatory genes in dendritic cells was associated with demethylation of H3K9 and recruitment of RNA polymerase II to the respective promoter regions (26). H3K9 demethylation correlated better with this recruitment than did acetylation of H3 and H4 (26), consistent with our present observation that demethylation of H3K9 was more prominent than H3K9ac, H3K27ac, or H4K91ac in human failing myocardium. Furthermore, our experiments using isolated working hearts indicated that H3K9 was rapidly demethylated within 30 minutes in response to elevated cardiac preload, supporting the dynamic nature of H3K9 methylation as an epigenetic control mechanism of ANP and BNP in the heart. In contrast, H3K4 methylation was unchanged in failing hearts, in agreement with a recent study on adult cardiac myocytes in which reduction of overall H3K4 levels by deletion of PAX interacting protein 1, a key component of the H3K4 complex, did not affect BNP expression and had only a small — and, paradoxically, activating — effect on ANP expression (36).
A potential mechanism that could explain demethylation of H3K9 in the ANP and BNP promoter regions in failing hearts is upregulation of the H3K9-specific demethylases JMJD1A, JMJD2A, and JMJD2B. Expression of JMJD1A correlated inversely with H3K9me2 levels and positively with upregulation of both ANP and BNP in failing hearts. These data are in line with previous results in smooth muscle cells, in which increased expression of JMJD1A was associated with decreased global levels of demethylated H3K9 and activation of several smooth muscle–specific genes (37).
Although knockdown of JMJD1A alone was compensated for by increased JMJD2A expression and had no effect on baseline ANP expression in neonatal cardiac myocytes (data not shown), double knockdown of both JMJD1A and JMJD2A reduced ANP and BNP expression by increasing H3K9me2 and H3K9me3. Since expression of JMJD1A and JMJD2A was 4- and 9-fold higher, respectively, and basal H3K9me2 and H3K9me3 were approximately 2-fold reduced in neonatal compared with adult cardiac myocytes, it may be possible that these JMJC domain–containing demethylases play a regulatory role in cardiac development by controlling the fetal gene program through control of H3K9 methylation. Furthermore, concerted upregulation of JMJD1A and JMJD2A could — to a certain extent — contribute to the reactivation of the fetal gene program in patients with heart failure.
In addition to JMJD1A, JMJD2A, and JMJD2B were also upregulated in human failing hearts, and recruitment of JMJD2A to the promoter region of ANP was increased in myocardium from patients with ICM, but not DCM. A recent study revealed that JMJD2A plays an important role in the development of cardiac hypertrophy and failure (30). A major target of JMJD2A demethylase activity was the promoter region of FHL1, a component of the mechanotransducer machinery in the heart that can induce cardiac hypertrophy via a pathway that involves serum response factor (SRF) and myocardin (30). Since ANP is also an SRF/myocardin target gene, JMJD2A augmented ANP expression in response to myocardin in an SRF-sensitive manner (30). These data support a regulatory role of JMJD2A for reactivation of ANP and BNP in the failing heart.
However, several observations suggest that JMJC domain–containing demethylases may play a modulatory rather than a primary role for the reactivation of fetal genes in the failing heart. First, while JMJD2A knockdown by 80% reduced baseline expression of FHL1 by approximately 75% in neonatal cardiac myocytes, ANP expression, which was 80-fold higher in neonatal versus adult cardiac myocytes (Figure (Figure6A),6A), was not reduced under baseline conditions in the study by Zhang et al. (30). Second, the effect of double knockdown of JMJD1A and JMJD2A on ANP expression in neonatal cardiac myocytes was rather small (13%–15%; Figure Figure6E).6E). Third, in response to pressure overload in mice, ANP expression increased 40-fold without a change of JMJD1A and JMJD2A expression (Figure (Figure7).7). Finally, in response to an acute increase of preload in mouse hearts, H3K9 was demethylated without an increase of JMJD2A expression or recruitment to the ANP promoter region (Figure (Figure88 and Supplemental Figure 6). Thus, we considered additional mechanisms that may account for H3K9 demethylation in human heart failure.
Methylation of histones is governed by G9a, SUV39H1, and several other methylases (9, 22). SUV39H1 selectively methylates H3K9, which generates a binding site for HP1 (11, 38) and favors the formation of densely packed heterochromatin. Although overall SUV39H1 expression was unchanged in human failing hearts, H3K9 demethylation was closely associated with nucleo-cytoplasmic shuttling of HDAC4 and HP1 dissociation from the promoter regions of ANP and BNP. Our findings in mice with cardiomyocyte-specific deletion of Hdac4 indicated that HDAC4 plays a causal role in controlling H3K9 demethylation in — and HP1 dissociation from — the ANP promoter region and, thus, ANP upregulation in response to increased hemodynamic load. As a potential underlying mechanism, the results in cell systems indicate that HDAC4 undergoes colocalization and protein-protein interaction with the histone methyltransferase SUV39H1 in the nucleus, providing a mechanism how HDAC4 could control H3K9 methylation. The isoform CaMKIIδB is abundantly expressed in the heart, located in the nucleus, and specifically docks to and phosphorylates HDAC4 (17). In the present study, we found that constitutively active CaMKIIδB disrupted the interaction between HDAC4 and SUV39H1 by triggering nucleo-cytosolic export of HDAC4, while SUV39H1 remained in the nucleus. Although these data do not prove causality, they nevertheless support the concept of a multiprotein complex consisting of HDAC4, SUV39H1, and HP1 (33, 34) that governs H3K9 methylation. Furthermore, in concert with previously reported results with CaMKI (33), our findings may extend this concept to signal responsiveness to CaMKIIδB, a CaMKII isoform that selectively targets HDAC4 (17) and is known to be causally involved in the development of cardiac hypertrophy and failure in response to pressure overload (39).
Our results indicate that reactivation of the fetal gene program in human heart failure is associated with epigenetic remodeling in the promoter regions of ANP and BNP. Despite nuclear export of HDAC4, ANP and BNP gene activation did not require increased histone acetylation in these promoter regions. In contrast, dynamic demethylation of H3K9 and dissociation of HP1 from the promoter regions were controlled by HDAC4, possibly by forming a transcriptional repressor complex with SUV39H1 that is disrupted by CaMKIIδB-induced phosphorylation of HDAC4 (Figure (Figure11).11). Furthermore, upregulation of JMJC domain–containing demethylases may sustain H3K9 demethylation in the course of heart failure. These findings improve our understanding of the dynamic epigenetic regulatory pathways in human heart failure and may help to identify novel targets to improve the treatment of patients with heart failure.
Human failing myocardium was obtained from patients with ICM (n = 8) or DCM (n = 8). 8 donor hearts of patients with no signs of heart disease that could not be used for transplantation due to ABO mismatch were used as nonfailing controls. See Supplemental Table 1 for clinical parameters and medication of patients with ICM and DCM.
Animals were housed under standard conditions with a 12-hour light/12-hour dark cycle and free access to water and chow. For working heart experiments, male CD1 mice (Charles River Laboratories) were used at 7 weeks of age. For TAC experiments, C57BL/6NCrl mice (Charles River) were used at 21–23 g BW. Conditional mutant mice with cardiac-specific Hdac4 deletion were generated by crossing existing floxed Hdac4 mice (Hdac4loxP/loxP; ref. 40) to transgenic mice expressing Cre recombinase under the control of the α-MHC promoter (Hdac4loxP/loxP;α-MHC-Cre; ref. 41) on a 129SvEv/C57BL/6 mixed background. From this intercross, male Hdac4loxP/loxP and Hdac4loxP/loxP;α-MHC-Cre mice were used for subsequent experiments.
Working heart experiments were performed as described previously (42). Briefly, hearts were perfused with Krebs-Henseleit solution at 37°C in a Langendorff apparatus, paced at 400 bpm, and exposed to effective preload of 10 mmHg and afterload of 80 mmHg. LV systolic and diastolic function was recorded via a high-fidelity conductance catheter (1.4 F SPR-839, 12 mm spacing; Millar) inserted into the LV cavity by apical puncture. Cardiac load conditions were varied by increasing either preload from 10 to 30 mmHg or afterload from 80 to 120 mmHg. Preload and afterload groups were compared with a third control group in which preload and afterload were maintained at physiological levels (10 mmHg preload, 80 mmHg afterload). After 30, 60, and 90 minutes, hearts were snap-frozen and stored for further analysis. See Supplemental Methods for details.
TAC was performed in male C57BL/6NCrl mice (21–23 g BW) as described previously (43). The aorta was constricted by 65%–70% for 6 weeks using a 27-gauge needle. Control animals underwent a sham operation without aortic ligation.
Western blot analysis was performed by standard methods on cytosolic and nuclear protein fractions of human LV and mouse myocardium as described previously (44) using antibodies against HDAC4, GAPDH, REST, and mRNA polymerase II. See Supplemental Methods for details.
See Supplemental Methods and Supplemental Tables 2, 3, and 6.
Briefly, after chromatin isolation, IP was performed with antibodies against H3K9ac, H3K27ac, H4K91ac, H3K4me2, H3K4me3, H3K9me2, H3K9me3, H3K27me3, HP1, JMJD1A, JMJD2A, G9a, LSD1, SUV39H1, and REST. The quantity of genomic DNA specifying the promoter regions of ANP, BNP, and GAPDH in the immunocomplexes were assessed by quantitative PCR. See Supplemental Tables 4 and 5 for primer sequences. For all experiments in human myocardium, tissues from patients with DCM and ICM were compared with nonfailing control tissues at the same time, under identical experimental conditions. PCR-products obtained from the ANP, BNP, or GAPDH promoter region were separated on 2% agarose gels. Similar conditions applied to experiments on mouse hearts and rat neonatal cardiac myocytes: all experimental conditions were handled under identical conditions as described above. See complete unedited blots in the supplemental material. Because no differences were found between ICM and DCM tissue in most analyses, unless otherwise indicated, the cumulative data of ICM and DCM were merged into the single failing myocardium group. See Supplemental Methods for details.
Hearts of 3- to 5-day-old Sprague-Dawley rats were removed, and the ventricles were dissected, digested in ADS buffer (116 mmol/l NaCl; 20 mmol/l HEPES; 0.8 mmol/l Na2HPO4; 5.6 mmol/l glucose; 5.4 mmol/l KCl; 0.8 mmol/l MgSO4 × 7 H2O, pH 7.35; 0.6 mg/ml pankreatin; 0.5 mg/ml collagenase type 2) at 37°C, and gently agitated in a bath shaker (86 rpm) in several steps. All supernatants were collected, and the cells were preplated in F10 medium (Gibco, Invitrogen) plus 10% horse serum, 5% FCS, and 1% penicillin/streptomycin to deplete contamination by nonmyocardial cells. After 45 minutes, the still-suspended myocardial cells were removed from the attached nonmyocardial cells, counted using a Neubauer counting chamber, and plated at a density of 1.65–1.75 × 106 cells per 60-mm culture dish (BD Primaria).
Neonatal rat cardiac myocytes were cultured under standard conditions (37°C, 5% CO2) in serum-containing F10 medium as described above. Transfection was performed with 20 nM predesigned siRNAs for JMJD1A and/or JMJD2A (Rn_Jmjd1a_3 and Rn_LOC313539_1; Qiagen) and custom-designed scrambled control siRNAs (scr1a, 5′-CUCACCGAUUACCGUACUATT-3′; scr2a, 5′-GCUCGAAACGCCUAUAGAATT-3′; Qiagen) using HiPerfect Transfection Reagent (7.5 μl per transfection; Qiagen) according to the manufacturer’s protocol.
Adult rat cardiomyocytes were isolated with a protocol similar to that previously described for guinea pig myocytes (45). After centrifugation, supernatant was discarded, and cells were applied for RNA isolation using peqGold TriFast following the manufacturer’s protocol (Peqlab).
Immunostaining was performed as described previously (17). Briefly, COS cells were grown on coverslips in DMEM plus 10% FCS and 1% penicillin/streptomycin; transfected with GFP-tagged HDAC4 (750 ng), FLAG-tagged SUV39H1 (750 ng), or myc-tagged CaMKIIδB-T287D (400 ng) using GeneJammer (Agilent) according to the manufacturer’s protocol; fixed after 48 hours; permeabilized; and blocked. Primary antibody against FLAG (rabbit anti-FLAG; Sigma-Aldrich) was used at a dilution of 1:200, secondary antibody (Texas Red anti-rabbit; vector) was used at a dilution of 1:400. All images were captured with a ×40 objective.
Co-IP was performed as described previously (17). Briefly, HEK293T cells were transfected with myc-tagged HDAC4 (1 μg), FLAG-tagged SUV39H1 (1 μg), or myc-tagged CaMKIIδB-T287D (0.8 μg); harvested after 48 hours; and disrupted with a 25-gauge needle. Cell debris was removed by centrifugation, after which FLAG-tagged proteins were subjected to IP with M2-agarose conjugate (Sigma-Aldrich) and washed. Bound proteins were resolved by SDS-PAGE, and IBs were performed as indicated with anti-Myc antibody (A-14; Santa Cruz Biotechnology Inc.) and anti-FLAG antibody (M2; Sigma-Aldrich). Proteins were visualized using a chemiluminescense system (GE Healthcare).
Data are presented as mean ± SEM. Differences between groups were calculated by unpaired Student’s 2-tailed t test or 1-way ANOVA where appropriate. Regression analysis was performed using GraphPad Prism (version 3.00 for Windows; GraphPad Software). A P value less than 0.05 was considered significant.
Human studies were conducted according to the Declaration of Helsinki and approved by the local ethics committee (Ärztekammer des Saarlandes no. 131/00). All patients gave informed consent. Animal studies conformed to NIH guidelines (Guide for the Care and Use of Laboratory Animals. NIH publication no. 85-23. Revised 1996) and were approved by the local ethics committee.
We thank Eric N. Olson for helpful advice on the manuscript and Michelle Gulentz, Jeannette Zimolong, and Lisa Lang for technical assistance. The study was supported by the DFG (Emmy Noether Programm to C. Maack and J. Backs; KFO-196 to C. Maack and M. Böhm; SFB-894 to C. Maack), Hans und Gertie Fischer-Stiftung (to M. Böhm, M. Hohl, and J.-C. Reil), and Kompetenznetz Herzinsuffizienz of the Bundesministerium für Bildung und Forschung (BMBF; to M. Böhm). J. Backs is supported by the DZHK and by the BMBF.
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2013;123(3):1359–1370. doi:10.1172/JCI61084.