Research ArticleCardiology Open Access | 10.1172/jci.insight.161788

Xianghai Liao,1 Peter J. Kennel,1 Bohao Liu,2 Trevor R. Nash,2 Richard Z. Zhuang,2 Amandine F. Godier-Furnemont,2 Chenyi Xue,1 Rong Lu,1 Paolo C. Colombo,1 Nir Uriel,1 Muredach P. Reilly,1 Steven O. Marx,1 Gordana Vunjak-Novakovic,2 and Veli K. Topkara1

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Liao, X. in: JCI | PubMed | Google Scholar

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Kennel, P. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Liu, B. in: JCI | PubMed | Google Scholar

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Nash, T. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Zhuang, R. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Godier-Furnemont, A. in: JCI | PubMed | Google Scholar

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Xue, C. in: JCI | PubMed | Google Scholar

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Lu, R. in: JCI | PubMed | Google Scholar

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Colombo, P. in: JCI | PubMed | Google Scholar

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Uriel, N. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Reilly, M. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Marx, S. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Vunjak-Novakovic, G. in: JCI | PubMed | Google Scholar |

1Division of Cardiology, Columbia University Irving Medical Center – New York Presbyterian, New York, New York, USA.

2Department of Biomedical Engineering, Columbia University, New York, New York, USA.

Address correspondence to: Veli K. Topkara, Center for Advanced Cardiac Care, Columbia University Medical Center – New York Presbyterian, 622 West 168th St, PH10-203A, New York, New York 10032, USA. Phone: 212.305.4600; Email: vt2113@cumc.columbia.edu.

Authorship note: XL and PJK contributed equally to this work.

Find articles by Topkara, V. in: JCI | PubMed | Google Scholar |

Authorship note: XL and PJK contributed equally to this work.

Published January 19, 2023 - More info

Published in Volume 8, Issue 4 on February 22, 2023
JCI Insight. 2023;8(4):e161788. https://doi.org/10.1172/jci.insight.161788.
© 2023 Liao et al. This work is licensed under the Creative Commons Attribution 4.0 International License. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Published January 19, 2023 - Version history
Received: May 11, 2022; Accepted: January 11, 2023 View PDF Abstract

Heart failure (HF) is characterized by global alterations in myocardial DNA methylation, yet little is known about the epigenetic regulation of the noncoding genome and potential reversibility of DNA methylation with left ventricular assist device (LVAD) therapy. Genome-wide mapping of myocardial DNA methylation in 36 patients with HF at LVAD implantation, 8 patients at LVAD explantation, and 7 nonfailing (NF) donors using a high-density bead array platform identified 2,079 differentially methylated positions (DMPs) in ischemic cardiomyopathy (ICM) and 261 DMPs in nonischemic cardiomyopathy (NICM). LVAD support resulted in normalization of 3.2% of HF-associated DMPs. Methylation-expression correlation analysis yielded several protein-coding genes that are hypomethylated and upregulated (HTRA1, FBXO16, EFCAB13, and AKAP13) or hypermethylated and downregulated (TBX3) in HF. A potentially novel cardiac-specific super-enhancer long noncoding RNA (lncRNA) (LINC00881) is hypermethylated and downregulated in human HF. LINC00881 is an upstream regulator of sarcomere and calcium channel gene expression including MYH6, CACNA1C, and RYR2. LINC00881 knockdown reduces peak calcium amplitude in the beating human induced pluripotent stem cell–derived cardiomyocytes (hiPSC-CMs). These data suggest that HF-associated changes in myocardial DNA methylation within coding and noncoding genomes are minimally reversible with mechanical unloading. Epigenetic reprogramming strategies may be necessary to achieve sustained clinical recovery from heart failure.

Graphical Abstract Introduction

Heart failure (HF) is a major public health problem with more than 6 million patients affected in the United States alone (1). Despite significant progress achieved in the diagnosis and treatment of HF, the majority of patients progress to advanced stages of the disease, leading to unacceptably high rates of morbidity and mortality, even exceeding most cancers. Pharmacological management of HF has traditionally focused on targeting endogenous neurohormonal signaling cascades that are associated with disease progression as well as diuretic therapy for symptom relief (2). However, our limited mechanistic understanding of the complex HF pathophysiology does not fully explain the wide variation in disease progression and treatment response observed in this population, highlighting the need for novel molecular diagnostic and therapeutic strategies.

The failing human heart undergoes structural and functional remodeling that is accompanied by profound alterations in the myocardial transcriptome including recapitulation of the fetal gene expression program and downregulation of the genes involved in the oxidative phosphorylation pathway (3, 4). While some of these changes are common to all forms of HF, gene expression could be etiology specific and help distinguish patients with different types of HF (5, 6). Transcriptional profiling of paired myocardial samples obtained from patients with advanced HF before and after left ventricular assist device (LVAD) support demonstrate that only a small percentage of genes that are dysregulated in HF normalize with mechanical unloading of the failing human heart (7–9). While the mechanisms responsible for the persistent dysregulation of myocardial gene expression in the end-stage human HF remains unknown, emerging evidence suggests that epigenetic regulation may play important roles in transcriptional reprogramming by altering gene accessibility and TF binding to gene promoters or enhancers (10, 11).

DNA methylation is an essential epigenetic modification involving transfer of a methyl group onto the fifth position of the cytosine catalyzed by a family of DNA methyltransferases (DNMTs) and generally signals for transcription repression at the promoter sites (12). Several pathological conditions, in particular malignant transformation, have been associated with hypermethylation of specific target gene promoters as well as global hypomethylation leading to genomic instability (13). DNA methylation assays have been FDA approved for early detection of cancer and DNA demethylation agents have become standard-of-care therapy for high-risk patients with myelodysplastic syndrome (14–16). While the data is lagging for HF, several studies to date investigated genome-wide DNA methylation in the failing human heart using targeted bisulphite sequencing or chip-based approaches (17–23). However, none of these studies have examined whether or not, and if so to what extent, aberrant DNA methylation is reversible in the failing human heart. Moreover, HF-associated changes in DNA methylation of noncoding genomic elements including long noncoding RNAs (lncRNAs) remain largely unknown. Hence, in the current study, we characterize the myocardial DNA methylation profile of patients with end-stage cardiomyopathy before and after mechanical unloading with LVAD support. Furthermore, we identify what we believe to be a novel cardiac-specific– and super-enhancer–associated lncRNA (LINC00881) that is hypermethylated and downregulated in the failing human heart as a key regulator of calcium handling in the cardiomyocyte.

Results

Myocardial DNA methylation in ischemic and nonischemic cardiomyopathy. A total of 36 patients with end-stage HF, i.e., 12 with ischemic cardiomyopathy (ICM) and 24 with nonischemic dilated cardiomyopathy (NICM), who underwent LVAD implantation at Columbia University Irving Medical Center and 7 nonfailing (NF) controls were included for genome-wide DNA methylation analysis. The analysis pipeline has been summarized in Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.161788DS1 Patients with myocarditis, amyloid cardiomyopathy, restrictive/hypertrophic cardiomyopathy, and previous cardiac transplantation were excluded from the study. Demographics and clinical characteristics of patients with HF with myocardial genome-wide DNA methylation profiling at the time of LVAD implantation were summarized in Table 1. Paired post-LVAD cardiac samples were obtained from 8 patients who were bridged to heart transplantation. LVAD support resulted in significant reductions in left ventricular (LV) end-diastolic diameter, and a trend toward increased left ventricular ejection fraction (LVEF) and decreased serum B-type natriuretic peptide (BNP) levels in 8 patients (Supplemental Table 1A). Clinical information of NF cardiac donors has been provided in Supplemental Table 1B. Histopathological examination of human heart tissue samples demonstrated cardiomyocyte hypertrophy and myocardial fibrosis in pre-LVAD patients compared with NF controls consistent with HF phenotype (Figure 1A). Cardiomyocyte hypertrophy significantly regressed with LVAD support while myocardial fibrosis did not. Quantitative PCR (qPCR) analysis demonstrated upregulation of NPPA, MYH7, and COL1A1 genes in HF compared with NF controls. LVAD support resulted in significant downregulation of NPPA and MYH7 genes but not the COL1A1 gene (Figure 1B).

Figure 1

Impact of mechanical unloading on histopathology and gene expression of the failing human heart. (A) Cardiomyocyte CSA by H&E staining and myocardial collagen deposition by Masson’s trichrome staining (*P < 0.05, Kruskal-Wallis test with pairwise BH-corrected Wilcoxon test, n = 7 controls, 28 pre-LVAD, and 6 post-LVAD patients). Original magnification fields, ×20. (B) Myocardial NPPA, MYH7, and COL1A1 mRNA levels by qPCR (*P < 0.05, Kruskal-Wallis test with pairwise BH-corrected Wilcoxon test, n = 7 controls, 17 pre-LVAD, and 7 post-LVAD patients).

Table 1

Clinical characteristics of study subjects with genome-wide DNA methylation profiling at the time of LVAD implantation

Genome-wide DNA methylation profiling identified 2,079 differentially methylated positions (DMPs) in the myocardium of patients with ICM (ICM versus NF, q < 0.05; Supplemental Table 2). Of those, 625 DMPs were hypermethylated and 1,454 DMPs were hypomethylated. A total of 261 DMPs were differentially methylated in the myocardium of patients with NICM (NICM versus NF, q < 0.05; Supplemental Table 3). Of those, 117 DMPs were hypermethylated and 144 DMPs were hypomethylated in NICM. A total of 192 DMPs (Supplemental Table 4) were common to both patients with ICM and NICM (Figure 2A). All of these “common HF DMPs” were either concordantly hypomethylated (n = 125) or concordantly hypermethylated (n = 67) in ICM and NICM.

Figure 2

Genome-wide changes in myocardial DNA methylation based on HF etiology. (A) Venn diagram depicting the number of DMPs in patients with ICM versus NICM (q < 0.05, |Δ β |  > 10%). (B) Proportion of ICM and NICM DMPs by genomic location and feature. (C) PCA of myocardial DNA methylation grouped by etiology of HF (NF, green; ICM, red; and NICM, blue). (D) Heatmap clustering of subjects by methylation profile of 192 common HF DMPs.

Characterizing the location of DMPs within gene regions, for ICM and NICM similarly, a higher proportion of probes located within the transcription start site were hypermethylated than hypomethylated, whereas intergenic region probes were more likely to be hypomethylated than hypermethylated (Figure 2B). DMPs within open sea regions were more likely to be hypomethylated whereas DMPs within CpG islands and associated shores were more likely to be hypermethylated in both ICM and NICM (Figure 2B). A total of 17.9% of DMPs mapped to promoter regions (TSS200, TSS1500, and 5′UTR) in patients with ICM compared with 22.2% of DMPs in patients with NICM. Principal component analysis (PCA) of genome-wide methylation levels demonstrates clustering of the NF versus failing samples, however, did not clearly separate between subjects with ischemic versus nonischemic etiology (Figure 2C). Heatmap of methylation level Z scores of 192 common HF DMPs across NF healthy controls and HF samples with unbiased hierarchical heatmap clustering separated the NF controls from the failing heart DMPs but did not separate ICM from NICM (Figure 2D).

Minimal reversibility of myocardial DNA methylation with LVAD support. Patterns of myocardial DNA methylation were analyzed in paired myocardial samples obtained from 8 patients before and after LVAD support. A total of 1,075 CpG sites were differentially methylated in the failing myocardium compared with NF (pre-LVAD versus NF, q < 0.05; Supplemental Table 5). In contrast, only 130 CpG sites were differentially methylated with LVAD support (post-LVAD versus pre-LVAD, q < 0.05; Supplemental Table 6). Only 35 CpG sites (Supplemental Table 7) were common in HF and reverse remodeling (RR) (Figure 3A), all of which were methylated in opposite directions.

Figure 3

Impact of mechanical unloading on myocardial DNA methylation profile. (A) Venn diagram depicting the number of DMPs in HF (pre-LVAD versus NF) versus RR (post-LVAD versus pre-LVAD) (q < 0.05, |Δ β|  > 10%). (B) Proportion of HF and RR DMPs by genomic location and feature. (C) PCA of myocardial DNA methylation before and after LVAD support (NF, red; pre-LVAD, blue; and post-LVAD, green). (D) Heatmap clustering of subjects by 35 LVAD responsive HF DMPs.

A higher proportion of DMPs located within gene bodies and transcription start sites were hypermethylated while DMPs located within intergenic regions (IGR) were hypomethylated in HF. Conversely, a higher proportion of probes located within gene bodies and transcription start sites were hypomethylated while DMPs located within IGRs were hypermethylated in RR, suggesting an opposite trend between HF and RR (Figure 3B). PCA of genome-wide methylation levels demonstrated clustering of the NF versus pre-LVAD samples, but not pre- versus post-LVAD samples, suggesting only minor changes in global DNA methylation with mechanical unloading (Figure 3C). Heatmap of methylation level Z scores of 35 LVAD-responsive HF DMPs across NF and paired LVAD samples with unbiased hierarchical clustering demonstrated grouping of post-LVAD samples with 8 out of 9 NF control samples (Figure 3D), confirming LVAD-induced changes in DNA methylation toward the NF state in these genomic positions.

Integrated analysis of DNA methylation with gene expression in the failing human heart. The classical paradigm of promoter DNA methylation as a transcriptional silencing mechanism has recently been challenged with growing lines of evidence suggesting that DNA hypermethylation could also result in transcriptional activation (24–26). Accordingly, we assessed the relationship between differentially methylated CpG sites and changes in gene expression levels using a large publicly available transcriptional data set obtained from patients with HF (5). Out of 192 common HF DMPs, 121 CpG sites map to 93 protein-coding genes (Figure 4A). Of those, 74 genes were expressed in the myocardium (RPKM > 1), and 34 were differentially expressed in human HF. When changes in DNA methylation were correlated with the changes in gene expression, 14 genes were hypomethylated and transcriptionally upregulated, 13 genes were hypomethylated and transcriptionally downregulated, 6 genes were hypermethylated and transcriptionally upregulated, and 1 gene was hypermethylated and transcriptionally downregulated (Figure 4B and Table 2). Using independent cardiac tissue samples obtained from patients with end-stage ICM and NICM, we validated upregulation of AKAP13 (fold change [FC] = 2.98), HTRA1 (FC = 1.52), EFCAB13 (FC = 2.86), and FBXO16 (FC = 1.92) as well as downregulation of TBX3 (FC = 0.39) in the failing human hearts by qPCR analysis, while changes in RPTOR and HDAC9 transcripts did not reach statistical significance (Figure 4C).

Figure 4

Correlation of DNA methylation with gene expression in human HF. (A) Differential methylation and gene expression of 192 common HF DMPs. (B) Methylation versus gene expression correlation plot. (C) qPCR validation of candidate genes that are epigenetically regulated in independent samples obtained from patients with HF. *P < 0.05, #P < 0.10; unpaired 2-tailed t test; n = 6 (control) and 23 (failing). Data shown as mean ± SEM.

Table 2

DNA methylation versus mRNA expression correlation of DMPs common to ICM and NICM

Super-enhancer associated LINC00881 is downregulated in the failing human heart. Common HF and LVAD-responsive HF DMPs located in intergenic regions were screened for the presence of overlapping noncoding RNAs using GENCODE and NONCODE data sets. We identified a novel long intergenic nonprotein coding RNA (LINC00881) located approximately 9 kb upstream of a significantly hypermethylated CpG site (cg01535205) (Supplemental Table 8). LINC00881 is highly and exclusively expressed in the human myocardium according to GTEx, which made it an interesting candidate for further exploration in the context of epigenetic modification of the failing heart (Supplemental Figure 2). Interestingly, LINC00881 and cg01535205 are located within a cardiac super-enhancer region (Supplemental Figure 3). RNA-Seq and CHIP-Seq tracks confirm high expression of LINC00881 in the adult NF human heart as well as the presence of active chromatin marks in this region including H3K27ac (Figure 5A). qPCR of LINC00881 in independent HF samples confirmed that this intergenic RNA is downregulated in ICM and NICM compared with myocardium from healthy controls (Figure 5B). LINC00881 is largely restricted to the nuclear compartment as opposed to cytoplasmic in hiPSC-CMs (Figure 5C). Inhibition of DNA methyltransferase using 5-Azacytidine (5-AZA) resulted in significant upregulation of LINC00881 transcript levels in the beating hiPSC-CMs, suggesting epigenetic regulation of LINC00881 in cardiomyocytes (Figure 5C). LINC00881 expression was significantly upregulated during differentiation of hiPSC-CMs, in parallel with upregulation of transcription factors including GATA4, HAND2, and TBX5, confirming its role as a cardiomyocyte lineage-specific super-enhancer lncRNA (Figure 5D).

Figure 5

LINC00881 is a cardiomyocyte lineage-specific super-enhancer lincRNA. (A) Position, expression, and epigenetic regulation of the LINC00881 locus in the NF human heart. (B) Expression of LINC00881 in patients with ICM and NICM cardiomyopathy by qPCR (*P < 0.05, 1-way ANOVA with Fisher’s post hoc test; n = 6 controls, 11 ICM, and 12 NICM). (C) Expression of LINC00881 in the nuclear versus cytoplasmic fractions (*P < 0.05, unpaired 2-tailed t test, n = 6/group), with or without treatment with 20 nM 5-AZA (*P < 0.05, unpaired 2-tailed t test, n = 6/group), in hiPSC-CMs (*P < 0.05, 1-way ANOVA with Fisher’s post hoc test compared with day 0, n = 5/group). (D) Expression levels of LINC00881 and cardiac transcription factors during hiPSC-CM differentiation (*P < 0.05, 1-way ANOVA with Fisher’s post hoc test compared with day 0, n = 5/group). Data shown as mean ± SEM.

To explore the mechanistic basis of LINC00881 dysregulation in human HF, we used plasmid-mediated overexpression and GapmeR-based knockdown of LINC00881 in the beating hiPSC-CMs (Supplemental Figures 4 and 5). RNA-Seq identified 1,545 genes that were differentially expressed with LINC00881 overexpression (Supplemental Table 9) and 2,268 genes that were differentially expressed with LINC00881 knockdown (P value cutoff < 0.05) (Supplemental Table 10). Among 199 common genes that were differentially regulated in both LINC00881 overexpression and LINC00881 knockdown models, 174 (87.4%) were regulated in opposite directions, including 73 genes that are positively regulated and 101 genes that are negatively regulated by LINC00881 (Figure 6A and Supplemental Table 11). Gene Ontology analysis of gene targets that are positively regulated by LINC00881 were enriched in sarcomere organization (MYH6, OBSCN, LDB3, MYOM2, and FHOD3), calcium ion transport (CACNA1C, CACNA1D, RYR2, CAMK2A, and ANXA6), ventricular tissue morphogenesis (FGFR2, MYH6, and TNNI1), and fatty acid β-oxidation (MECR, ACAD10, and PPARD) pathways (Figure 6B). Gene Ontology analysis of gene targets that are negatively regulated by LINC00881 were enriched in transcription from RNA II polymerase promoter (CEBPG, KLF5, KLF10, SMAD5, ATF1, ATF3, and ANKRD1) and regulation of apoptotic process (BCLAF1, DNAJA1, GNA13, MCL1, ANRKRD1, PHLDA1, and SIRT1) pathways. LINC00881 regulation of select sarcomere and calcium channel target genes were validated in hiPSC-CMs by qPCR with or without LINC00881 knockdown (Figure 6C). To determine whether LINC00881 regulation of sarcomere and calcium channel genes have functional relevance in the heart, we measured calcium transients in beating hiPSC-CMs treated with GapmeRs targeting LINC00881 versus scrambled control oligonucleotide. LINC00881 knockdown resulted in significant reductions in the peak calcium amplitude in hiPSC-CMs (Figure 6D).

Figure 6

LINC00881 is an essential regulator of cardiomyocyte calcium cycling. (A) Venn diagram depicting the number of DEGs in hiPSC-CMs with LINC00881 plasmid-based overexpression or GapmeR-mediated knockdown. (B) Gene Ontology analysis of gene targets that are positively or negatively regulated by LINC00881 for biological process. (C) qPCR validation of LINC00881 target genes (*P < 0.05, unpaired 2-tailed t test, n = 4–6/group). (D) Representative relative calcium traces in beating hiPSC-CMs with LINC00881 versus scrambled oligonucleotide knockdown with averaged normalized calcium amplitude (*P < 0.05, unpaired 2-tailed t test, n = 5–6/group). Data shown as mean ± SEM.

To gain further insights into LINC00881 regulation of cardiomyocyte gene expression, we performed in silico target prediction analysis using RNAct database and identified 54 putative LINC00881 interacting proteins (Supplemental Table 12). Candidate LINC00881 targets were significantly enriched for chromatin remodeling pathway proteins including BICRA, CECR2, ERCC6, SMARCA2, SMARCA4, BAZ1A, BAZ1B, and RSF1 (Figure 7A). RNA-Seq analysis in hiPSC-CMs suggested borderline significance for differential regulation of SMARCA4 by LINC00881 overexpression and knockdown, but not the other putative targets. LINC00881 regulation of SMARCA4 was validated in hiPSC-CMs by qPCR (Figure 7B). In vitro RNA IP in hiPSC-CMs showed that LINC00881 but not GAPDH was coprecipitated by SMARCA4 (Figure 7C). Chromatin accessibility experiment in hiPSC-CMs demonstrated a significant increase in RYR2 promoter accessibility with LINC00881 overexpression in hiPSC-CMs (Figure 7D).

Figure 7

LINC00881 regulates chromatin remodeling in human cardiomyocytes. (A) Gene Ontology analysis of putative LINC00881 protein targets for biological process (BP) and cellular component (CC) subontologies. (B) SMARCA4 mRNA levels with LINC00881 knockdown or LINC00881 overexpression in hiPSC-CMs by qPCR (*P < 0.05, unpaired 2-tailed t test, n = 5–6/group). (C) RNA IP of LINC00881 or GAPDH transcripts using SMARCA4 Ab versus IgG control by qPCR in hiPSC-CMs (*P < 0.05, unpaired 2-tailed t test, n = 3/group, data representative of 3 experiments). (D) Chromatin accessibility of MYH6, RYR2, and CACNA1C promoters with or without LINC00881 overexpression in hiPSC-CMs by qPCR (*P < 0.05, unpaired 2-tailed t test, n = 4/group).

Discussion

The present study utilized bead array technology for high-density genome-wide mapping of DNA methylation in the failing human heart before and after LVAD support. Our analysis identified myocardial DNA methylation patterns that are associated with reciprocal regulation of gene expression in patients with ICM and NICM. In addition to providing the most comprehensive mapping of etiology-specific changes in myocardial DNA methylation to date, we show, for what we believe to be the first time, that mechanical unloading with LVAD is associated with an incomplete normalization of the myocardial DNA methylation profile, suggesting that HF-related epigenetic alterations could be persistent. Moreover, our analysis identified a potentially novel cardiac-specific super-enhancer lncRNA gene (LINC00881), which is hypermethylated and downregulated in the failing human heart, as an essential regulator of cardiomyocyte calcium cycling.

DNA methylation in human HF. Consistent with previous reports, we found that the majority of DMPs in end-stage human HF had a reduction in DNA methylation levels, suggesting a global hypomethylation similar to what has been demonstrated in cancer biology (17, 18, 20, 27). Movassagh et al. demonstrated that HF-related differential methylation in CpG islands was predominantly located in gene promoters and gene bodies, but not in intergenic or 3′UTRs (17). Our analysis expands on this observation and demonstrates a large number of intergenic positions that are also differentially methylated, including a subset mapping to the noncoding genome. Our etiology-specific analysis identified 192 DMPs that were common to patients with ICM (2,079 DMPs) and NICM (261 DMPs), similar to a recent study by Glezeva et al., which identified 13 common HF DMPs from patients with ICM (51 DMPs) and NICM (118 DMPs) (20). Overlapping DNA methylation profiles in patients with ICM and NICM support a convergent “final common pathway” of gene expression that are found in all forms of HF irrespective of the inciting event. Our analysis showed that hypomethylated CpG sites, particularly in ICM, were more likely to be within the gene body or intergenic regions as opposed to transcription start sites, which is consistent with a prior study by Pepin et al. demonstrating a relative hypermethylation of promoter-associated CGIs in patients with ICM (28). Of the 93 common HF DMPs mapping to a protein-coding gene in the current study, 25 were also found to be differentially methylated in at least 1 previous study validating our analytical approach (18, 22, 23). Interestingly, we did not find differential DNA methylation for HF marker genes with the exception of NPPB promoter, which was significantly hypomethylated in NICM and ICM samples with a Δ β value below the detection cutoff level of 10%. Hypomethylation of natriuretic peptide gene promoters in patients with HF has been previously demonstrated by some (22, 23), but not all genome-wide DNA methylation studies (17, 18, 20). A plausible explanation for this discrepancy is that most studies to date used whole myocardium for DNA methylation profiling which could dilute cardiomyocyte-specific epigenetic signals. It is also possible that alternative epigenetic modifications such as histone modification may play a role in transcriptional regulation of HF marker genes (29).

Effect of mechanical unloading with LVAD on myocardial DNA methylation patterns. Mechanical unloading with LVAD is associated with favorable changes in the biology of the failing cardiomyocyte, including regression of myocyte hypertrophy, improvement in excitation-contraction coupling, and downregulation of the fetal gene expression, also termed RR, which results in normalization of cardiac function in a small number of patients, allowing for LVAD explantation (30–33). Transcriptional studies of paired myocardial samples obtained from patients with end-stage HF before and after LVAD support demonstrate that only less than 5% of HF-related transcripts normalize with LVAD support (7, 9). Consistent with these observations, we found that only 3.2% of HF-related DMPs were reverse methylated with LVAD support, which may in part explain persistent transcriptional dysregulation and the low incidence of myocardial recovery in patients supported by LVAD. Since DNA methylation is a reversible phenomenon, these findings raise the possibility that epigenetic modulation may be necessary to normalize the HF transcriptome and achieve myocardial recovery. In support of this hypothesis, inhibition of DNA methylation using 5-aza-2’-deoxycytidine rescued a HF phenotype in a rat model of norepinephrine-induced cardiac hypertrophy (34). Mice with cardiomyocyte-specific deletion of Dnmt3b develop cardiomyopathy with sarcomeric disarray and interstitial fibrosis (35). Similarly, differences in myocardial DNA methylation among mouse strains has been shown to determine susceptibility to cardiac hypertrophy following isoproterenol treatmen