Epigenetic regulation of cardiomyocyte maturation by arginine methyltransferase CARM1

Abstract

Background:

During the neonatal stage, the cardiomyocyte undergoes a constellation of molecular, cytoarchitectural, and functional changes known collectively as cardiomyocyte maturation to increase myocardial contractility and cardiac output. Despite the importance of cardiomyocyte maturation, the molecular mechanisms governing this critical process remain largely unexplored.

Methods:

We leveraged an in vivo mosaic knockout system to characterize the role of Carm1, the founding member of protein arginine methyltransferase, in cardiomyocyte maturation. Using a battery of assays including immunohistochemistry, immuno-electron microscopy imaging, and action potential recording, we assessed the effect of loss of Carm1 function on cardiomyocyte cell growth, myofibril expansion, T-tubule formation, and electrophysiological maturation. Genome-wide transcriptome profiling, H3R17me2a ChIP-seq, and ATAC-seq were used to investigate the mechanisms by which CARM1 regulates cardiomyocyte maturation. Lastly, we interrogated the human syntenic region to the H3R17me2a ChIP-seq peaks for single nucleotide polymorphisms associated with human heart diseases.

Results:

We report that mosaic ablation of Carm1 disrupts multiple aspects of cardiomyocyte maturation cell autonomously, leading to reduced cardiomyocyte size and sarcomere thickness, severe loss and disorganization of T-tubules, and compromised electrophysiological maturation. Genomics study demonstrates that CARM1 directly activates genes that underlie cardiomyocyte cytoarchitectural and electrophysiological maturation. Moreover, our study reveals significant enrichment of human heart disease-associated single nucleotide polymorphisms in the human genomic region syntenic to the H3R17me2a ChIP-seq peaks.

Conclusion:

This study establishes a critical and multifaceted role for CARM1 in regulating cardiomyocyte maturation and demonstrates that deregulation of CARM1-dependent cardiomyocyte maturation gene expression may contribute to human heart diseases.

Introduction

During the first weeks of extrauterine life, the murine cardiomyocyte switches from a fetal state to an adult phenotype, enhancing its ability to generate and withstand increased contractile force^1,2. One of the early manifestations of cardiomyocyte maturation is the virtually complete loss of cardiomyocyte proliferative capacity shortly after birth^3–5. As such, postnatal cardiac growth is primarily driven by physiological cardiomyocyte hypertrophy⁶. Meanwhile, as the cardiomyocytes mature, the originally round-shaped cell undergoes substantial shape change and becomes rod-like in shape^7,8. Such cell size and shape changes allow for myofibril maturation characterized by increased sarcomere length and thickness and enhanced sarcomere organization and alignment^2,9. Another major hallmark of cardiomyocyte maturation is the formation of the transverse-tubules (T-tubules)^10,11, the membrane invaginations that play a critical role in synchronizing calcium release and facilitating efficient propagation of electrical activity^11,12. Cardiomyocyte maturation also features a metabolic switch from glycolysis to fatty acid β-oxidation as the primary source of ATP generation¹³. This metabolic switch is supported by enhanced mitochondrial biogenesis and fusion, enabling the originally small, fragmented organelle to form large, interconnected networks with high oxidative capacity¹⁴. In sum, cardiomyocyte maturation is an intricate process involving changes in cardiomyocyte structure, metabolism and function that prepare the heart for efficient and forceful contraction and relaxation cycle throughout postnatal life.

Despite the acknowledged importance of cardiomyocyte maturation for establishing and maintaining heart function, the regulators of this critical process remain largely unidentified. Given the diverse cytoarchitectural and physiological changes involved in cardiomyocyte maturation, we hypothesize that this process could be regulated by a broad-acting epigenetic regulator with the capacity to control multiple gene activation pathways. Protein arginine methyltransferases (PRMTs) are a family of epigenetic regulatory enzymes that catalyze arginine methylation on their protein substrates^15,16. As one of the most abundant post-translational modifications, protein arginine methylation has been linked to the regulation of a broad range of biological processes^15–17. Yet, the roles of PRMTs in cardiomyocyte maturation remains largely unexplored. In this study, we focus on coactivator-associated arginine methyltransferase 1 (CARM1), the founding member of PRMTs also known as Prmt4¹⁸. We uncovered a multifaceted cell-autonomous role of CARM1 in controlling cardiomyocyte maturation. Specifically, we found that Carm1 mutant cardiomyocytes exhibit multiple maturation defects, including reduced cell and myofibril size, perturbed mitochondrial fusion, disrupted T-tubules formation, and compromised electrophysiological maturation. Mechanistically, genome wide transcriptomic analysis and chromatin immunoprecipitation followed by sequencing (ChIP-seq) demonstrated that CARM1 directly activates genes that underlie cardiomyocyte cytoarchitectural and electrophysiological maturation. Furthermore, a significant enrichment of cardiomyopathy (CP)-, QT interval (QT)-, and arrhythmia (AR)-associated single nucleotide polymorphisms (SNPs) are found to be located at the human region syntenic to the mouse ChIP-seq peaks, providing evidence that deregulation of CARM1-dependent cardiomyocyte maturation gene expression may contribute to human heart diseases.

Methods

The data, analytic methods, and study materials will be made available on reasonable request. The detailed methods are provided in the online-only Data Supplement.

Animal studies

All experiments involving animals were performed in accordance with the University of North Carolina at Chapel Hill Institutional Animal Care and Use Committee (IACUC) approved protocols.

Statistical Analysis

Two-tailed t-tests were performed for all the pairwise comparisons, and for those involving multiple groups, Bonferroni correction was performed. NS, no significance; *P<0.05; **P<0.01; ***P<0.001. The genomic data that support the findings of this study are available in the Gene Expression Omnibus (GEO) under the accession number GSE197197.

Results

Cardiomyocyte Specific Mosaic Depletion of Carm1 Suppresses Cardiomyocyte Maturational Hypertrophy

An increase in cardiomyocyte cell size is one of the most notable hallmarks of cardiomyocyte maturation. However, to our knowledge there are no published resources that quantify cardiomyocyte cell size at each step in the maturation process. Cardiomyocyte maturation takes place perinatally and during early postnatal stage¹⁹. To establish a baseline or control cell size for each week in the maturation process, we performed wheat germ agglutinin (WGA) staining on myocardial sections for hearts harvested from postnatal day 1 (P1) to P28 wildtype mice (Figure S1A–S1E and Figure S1A’–S1E’). A quantifiable significant difference in cardiomyocyte cross-sectional area was observed between each time-point (Figure S1F), demonstrating that substantial increase in cardiomyocyte size occurs within just one week during cardiomyocyte maturation.

To uncover cell-autonomous gene function and avoid potential early lethality caused by organ-wide loss of essential genes, we studied cardiomyocyte maturation by mosaically ablating gene function in vivo. We took advantage of an in vivo mosaic knockout system²⁰ by injecting AAV9-cTnT-Cre-IRES-GFP (abbreviated as AAV-Cre) to P1 Carm1^fl/fl mice at a dose that depleted Carm1 in about 5–10% of the cardiomyocyte population (Figure 1A and 1B). The injected mice were allowed to further develop until 4 weeks of age for phenotypic analysis unless otherwise indicated (Figure 1A). The combined use of AAV stereotype 9 (AAV9), which has been shown to preferentially target the heart²¹, and cardiac troponin t (cTnT) promoter allows for efficient and targeted Cre expression in cardiomyocytes²². GFP expression was used to indicate successful AAV9-mediated Cre delivery, and consequent cardiomyocyte specific mosaic loss of Carm1 function (Figure 1A). Indeed, in contrast to the control cardiomyocytes with robust nuclear CARM1 expression, about 90% of the GFP+ cardiomyocytes isolated from AAV-Cre infected Carm1^fl/fl hearts had undetectable level of CARM1 expression, and are hereafter referred to as Carm1^KO cardiomyocytes (Figure 1C). This Cre-mediated recombination efficiency is comparable to what was recently reported²³. Notably, we found that the Carm1^KO cardiomyocytes in the mosaic mutant hearts had substantially smaller cross-sectional area than age-matched wildtype (WT) and control AAV-GFP+ Carm1^fl/fl cardiomyocytes, while no significant difference in cross-sectional area was observed between the cardiomyocytes of the latter two genotypes (Figure 1D and 1E). To further evaluate the effect of Carm1 loss on cardiomyocyte growth, we performed anti-α-Actinin immunostaining on dissociated cardiomyocytes from digested AAV-Cre infected Carm1^fl/fl hearts (Figure 1F). Quantification of longitudinal area and length indicated that the GFP+ Carm1^KO cardiomyocytes displayed substantial reduction in cell size in the longitudinal orientation compared to the isolated GFP- control cardiomyocytes (Figure 1G). Collectively, these data demonstrated a critical role of CARM1 in controlling cardiomyocyte maturational hypertrophy in both cross-sectional and longitudinal orientations.

Figure 1. Carm1 loss suppresses cardiomyocyte maturational hypertrophy.

A, Left: schematic of the experimental procedure. Carm1^fl/fl mice were injected subcutaneously with AAV9 viruses at P1. The infected hearts were then harvested for phenotypic analysis. Right: schematic showing that Cre inactivates Carm1 in GFP+ but not GFP- cardiomyocytes of the AAV9-Cre infected Carm1^fl/fl hearts. B, Mosaic infection of 5–10% cardiomyocyte population by AAV9 viruses. WGA labels cardiac cells and GFP expression indicates successful AAV infection. C, CARM1 and α-Actinin double staining of control cardiomyocytes isolated from uninfected Carm1^fl/fl hearts (left panels) and GFP+ cardiomyocytes isolated from AAV-Cre infected Carm1^fl/fl hearts (right panels). Scale bars, 10 μm. D, Cardiac sections from control AAV-GFP or AAV-Cre infected Carm1^fl/fl hearts. The sections were stained with WGA in white and GFP in green. Scale bar, 20 μm. Scale bar in the zoomed-in images, 10 μm. E, Quantification of the cross-sectional area of WT, AAV-GFP or AAV-Cre infected cardiomyocytes. For each cardiomyocyte type, at least 200 cardiomyocytes from 4 hearts were quantified. F, Representative images of isolated Carm1^fl/fl and Carm1^KO cardiomyocytes stained for α-Actinin, GFP and DAPI. Scale bars, 10 μm. G, Quantification of the longitudinal area and length of the isolated Carm1^fl/fl and Carm1^KO cardiomyocytes. NS, no significance; *P<0.05; **P<0.01; ***P<0.001.

Carm1 Loss Disrupts Cardiomyocyte Cytoarchitectural Maturation

Next, we evaluated the effect of Carm1 loss on myofibril maturation. Myofibrils are made up of thick and thin filaments that are organized into repeating units called sarcomeres. Like those in their neighboring cardiomyocytes, anti-cTnT antibody labeled sarcomeres in the GFP+ Carm1^KO cells exhibited a regular striated pattern, suggesting that mosaic ablation of Carm1 didn’t cause cell autonomous sarcomere disarray (Figure 2A). This is not unusual as sarcomere assembly initiates at cardiac differentiation rather than during cardiomyocyte maturation^1,8,24. However, quantification of sarcomere length, which is defined as the distance between Z-lines, indicated that the sarcomeres in the mutant cardiomyocytes were significantly shorter compared to those in the neighboring GFP- controls (Figure 2A and 2B). We also found that, consistent with their drastic cell size reduction, Carm1^KO cardiomyocytes harbored substantially fewer sarcomere elements (Figure 2C). We then evaluated myofibril maturation by high-resolution imaging of cardiomyocyte subcellular structures with immuno-electron microscopy (IEM)²⁵. Under IEM, Carm1^KO cardiomyocytes in the mosaic mutant hearts were distinguishable from their neighboring GFP- cardiomyocytes by GFP immunogold labeling with colloidal gold particles (Figure S1G). In supportive of our observation with cTnT immunohistochemistry, IEM imaging revealed that Carm1^KO cardiomyocytes exhibited a similar significant reduction in sarcomere length (Figure 2D and 2E). Myofibril maturation also features an increase in sarcomere thickness^1,2. Quantification of sarcomere thickness for the cardiomyocytes on IEM images revealed that Carm1^KO cardiomyocytes demonstrated reduced sarcomere thickness compared to the control cardiomyocytes (Figure 2D and 2F). Taken together, our data indicated that loss of Carm1 function compromises myofibril maturation but not its initial assembly. In addition to myofibril maturation defects, IEM imaging of cardiomyocyte mitochondria also uncovered a significant reduction in mitochondrial size in the mutant cardiomyocytes (Figure 2G and 2H), suggesting that ablation of Carm1 function causes defective mitochondrial fusion.

Figure 2. Carm1 loss disrupts cardiomyocyte cytoarchitectural maturation.

A, Left: Representative images of cardiac sections from AAV-Cre infected Carm1^fl/fl hearts. Right, the top two panels are the zoomed-in images for the white-boxed area, and the bottom right two panels are the zoomed-in images for the yellow-boxed area. The cardiac sections were immunostained with WGA in white, cTnT in red and GFP in green. Scale bar, 10 μm. Quantification of (B) sarcomere length and (C) % total sarcomeric elements for the GFP- and GFP+ cardiomyocytes from AAV-Cre infected Carm1^fl/fl hearts. For each cardiomyocyte type, at least 10 cardiomyocytes from 2 hearts were qualified. D, Representative EM images of the sarcomeres of the control and Carm1^KO cardiomyocytes in the AAV-Cre infected Carm1^fl/fl hearts. Scale bars, 200 nm. Quantification of (E) the sarcomere length and (F) sarcomere thickness for the cardiomyocytes on the EM images. G, Representative EM images of the mitochondria of control and Carm1^KO cardiomyocytes for AAV-Cre infected Carm1^fl/fl heart. Scale bars, 200nm. H, Quantification of mitochondria size for the cardiomyocytes. N= 173 and 327 for control and mutant cardiomyocytes, respectively. I, Representative images of cardiac sections from control AAV-GFP (left two images) or AAV-Cre (right two images) infected Carm1^fl/fl mice. The hearts were stained with FM4–64 in white and GFP in green. Scale bars, 10 μm. J, Quantification of T-tubule elements and spacing of the AAV-GFP infected cardiomyocytes, and the GFP- and GFP+ cardiomyocytes from AAV-Cre infected Carm1^fl/fl hearts. For each cardiomyocyte type, at least 5 cardiomyocytes from 2 hearts were qualified. Abbreviation: TT for T-tubule.

In contrast to sarcomere, T-tubule formation is initiated during cardiomyocyte maturation^10,11. T-tubules are formed when the plasma membrane invaginates deeply into the cells to form an interconnected network critical for excitation-contraction coupling²⁶. Compared to control GFP+ cardiomyocytes in the AAV9-GFP infected Carm1^fl/fl hearts or control GFP- cardiomyocytes in the AAV9-Cre infected Carm1^fl/fl hearts, Carm1^KO cardiomyocytes exhibited notable loss of T-tubules (Figure 2I and 2J). Yet, the spacing of the residual T-tubules was significantly increased. Notably, no significant difference in T-tubule formation and spacing was observed between the above mentioned GFP+ and GFP- control cardiomyocytes (Figure 2I and 2J). Together, our data demonstrated that Carm1 is required for T-tubule formation and organization during cardiomyocyte maturation.

Carm1 Loss Compromises Electrophysiological Maturation

One of the hallmarks of ventricular cardiomyocyte maturation is shortening of the action potential duration (APD) and lowering of the resting membrane potential (RMP)²⁷. Therefore, to test whether Carm1^KO cardiomyocytes exhibited delay in electrophysiological maturation, we measured RMP and APD in isolated GFP+ basal left ventricular Carm1^KO cardiomyocytes isolated from 2 mosaic mutant hearts at 8 weeks of age (n=7 cells) and age-, sex-, and strain-matched wildtype control cells (n=10 cells) (Figure 3A). We found that while mutant cells and wildtype cells did not exhibit significantly different RMP or AP amplitude (Figure 3B and 3C), both APD50 and APD90 were markedly prolonged in mutant cells (Figure 3D and 3E), lending support to our hypothesis of a delay in electrophysiological maturation.

Figure 3. Carm1 loss compromises electrophysiological maturation.

A, Representative Action potentials (APs) recorded from a Carm1 mutant cell (green tracing) and a WT cell (black tracing) in response to stimulation in current clamp configuration at 1.5 times threshold at 1 Hz. Comparisons of (B) resting membrane potential, (C) AP amplitude, (D) APD50, and (E) APD90 between 7 mutant cells and 10 WT control cells. F, Left: Representative image of AAV9-Cre or AAV9-GFP infected hearts. Right top: representative zoomed-in image of uninfected control heart. Right bottom: zoomed-in image for the boxed area in the panel to the left. Scale bar for the image to the left: 300 μM. Scale bar for the image to the right: 30 μM. G, Representative images of Carm1^fl/fl hearts infected with AAV-GFP (left) or with AAV-Cre (right). H, Quantification of heart weight/tibia length of Carm1^fl/fl hearts infected with AAV-GFP (left) or with AAV-Cre. I, Quantification of the cross-sectional area for the control and mutant cardiomyocytes in 2 weeks old WT, AAV9-GFP infected, and AAV9-Cre infected Carm1^fl/fl hearts. J, Survival curve of Carm1^fl/fl mice infected with low and high dose of AAV9-Cre at P1. *P<0.05, **P<0.01, ***P<0.001.

Given the multifaceted role of CARM1 in cardiomyocyte maturation, we ascertained the effect of complete or near complete depletion of Carm1. We infected P1 Carm1^fl/fl mice with control AAV-GFP and AAV-Cre, respectively, at a dose that transduced about 90% of cardiomyocyte population (Figure 3F). We found that, compared to control hearts, the mutant hearts were significantly smaller in size at 2-weeks of age (Figure 3G). These mutant mice also demonstrated substantially decreased heart weight to tibia length ratio (HW/TL) (Figure 3H). Consistently, at the cellular level, the mutant hearts exhibited substantially decreased cardiomyocyte cross-sectional area, suggesting that the reduced cardiac size was probably attributable to the diminished cardiomyocyte size (Figure 3I). Likely due to a combination of cardiomyocyte cytoarchitectural and electrophysiological defects at whole organ level, the mutant mice succumbed to death before four weeks of age (Figure 3J),

Carm1 Loss Causes Widespread Transcriptomic Alterations that Underlies Cardiomyocyte Maturation Cytoarchitectural and Functional Defects

To determine the molecular mechanisms underlying CARM1-mediated cardiomyocyte maturation, we conducted comparative transcriptome analysis between control and Carm1^KO hearts at 2-weeks of age, when substantial cardiomyocyte cytoarchitectural maturation has been occurring in wildtype heart. Sample distribution analysis confirmed the consistency between the biological repeats for each condition (Figure S2A). The difference between Carm1^KO and control heart accounts for 87% of the variance in principle component analysis (PCA) analysis (Figure 4A). In total, we identified 1667 differentially expressed genes (DEGs) between control and Carm1^KO hearts, of which 1121 genes were downregulated in the mutant hearts (Figure 4B and 4C). Gene Ontology (GO) enrichment analysis^28,29 of the DEGs revealed that the upregulated genes are related to cell-substrate adhesion and microtubule cytoskeleton organization, suggesting that mutant cardiomyocytes may undergo cytoskeleton reorganization and remodeling of their adhesion network to compensate for maturation defects (Figure 4D). The downregulated genes in Carm1^KO mutant, however, were mostly involved in mitochondrial organization, muscle hypertrophy, heart contraction, and calcium/potassium ion transport, among others (Figure 4D). To further validate this observation, we performed gene expression profiling on isolated GFP+ control and Carm1^KO cardiomyocytes. We generated mosaic hearts that contained AAV9-Cre infected Carm1^KO and control AAV9-GFP infected control cardiomyocytes, respectively. These hearts were then dissected and dissociated into single cells, which were then subjected to FACS to enrich the viral infected GFP+ cardiomyocytes for next generation RNA sequencing (Figure S2B–S2D). Consistently, we found that the downregulated genes in isolated Carm1^KO cardiomyocytes are associated with a similar set of GO terms such as muscle hypertrophy, heart contraction, calcium/potassium ion transport and regulation of fatty acid metabolism (Figure S2E).

Figure 4. Carm1 loss led to dysregulation of cardiomyocyte maturation-associated genes.

A, PCA plot of RNA-seq data for the control and Carm1^KO cardiac samples. B, Volcano plot for the up- and down-regulated genes between the control and Carm1^KO mutant. C, Heatmap for the up- and down-regulated genes between the control and Carm1^KO mutant. D, GO terms associated with the up- and down-regulated genes between the control and Carm1^KO mutant. E, Differential expression of the genes associated with cardiac cytoarchitectural and insulin pathway. *P<0.05, **P<0.01. F, Heatmap for the RNA binding genes showing decreased expression between control and Carm1^KO mutant. G, Schematic of the rescue experiment. H, Expression ratio of N2BA/N2B isoforms in Carm1 knockdown cardiomyocytes that overexpressed Rbfox2 (dark green, right) or GFP (light green, mid). Biological replication of samples, n=3. Error bars represent standard deviation. *, P< 0.05; NS, not significant. I, Expression ratio of sarcomere genes (isoforms) associated with cardiomyocyte maturation. Biological replication of samples, n=3. Error bars represent standard deviation. *, P< 0.05; **, P< 0.01; ***, P< 0.001; NS, not significant. J, Quantification of the size of Carm1 knockdown cardiomyocytes that overexpressed Insr (light green, mid) or GFP. Biological replication of samples, n>=30 for all groups; Error bars represent standard deviation. **, P< 0.01; NS, not significant. K, Representative images of cardiomyocytes from the rescue group (left, siCarm1 Insr OE), knockdown group (mid, siCarm1 GFP OE), and control group (right, siNT GFP OE. The cells were stained with DAPI in blue and αActinin in green. Scale bars, 50 μm.

One of key molecular events that occurs during myofibril maturation is the sarcomeric protein isoform switching^1,2. Notably, Carm1 loss significantly downregulated the expression of the mature myofibril isoforms Tnni3 and Myh6, while resulting in an upregulation of the corresponding fetal isoforms Tnni1 and Myh7 (Figure 4E). Likewise, Titin N2BA/N2B isoform ratio was significantly increased upon Carm1 ablation, underscoring a critical role of CARM1 in myofibrillar isoform switching (Figure 4E). To identify the splicing regulators that mediate the effect of CARM1 on N2BA/N2B isoform switching, we compared the expression of the RNA splicing factors between Carm1^KO and control samples. Our analysis revealed that 46 RNA binding proteins showed decreased expression upon Carm1 loss, including several RNA splicing factors Rbfox1, Rbfox2, Rbpms, and Rbm24 (Figure 4F). To investigate the role of splicing factors in CARM1-mediated N2BA/N2B isoform switching, we induced maturation of cardiomyocytes differentiated in vitro from embryonic stem cells following the previously published protocol³⁰. (Figure 4G). Consistent with our in vivo observation, knockdown of Carm1 resulted in an increased N2BA/N2B ratio, which was partially rescued by overexpressing Rbfox2 (Figure 4H). These findings indicate that CARM1 regulates the expression of RNA splicing factors, such as Rbfox2, among others, which in turn control Titin isoform switch during cardiomyocyte maturation.

Insulin signaling is one of the major regulatory pathways that controls neonatal cardiac growth³¹. Myocardial conditional knockout of insulin receptor gene Insr leads to substantial reduction in cardiomyocyte size, resulting in cardiac growth retardation³². Interestingly, we found that cardiomyocyte specific ablation of Carm1 significantly reduced the expression of Insr and its downstream component insulin receptor substrate 1 (Irs1) (Figure 4E), suggesting that diminished cardiomyocyte size caused by Carm1 loss could be at least partly attributable to attenuated insulin signaling. Notably, utilizing the in vitro cardiomyocyte maturation platform described above, we demonstrated that overexpression of Insr, when compared to GFP overexpression, not only partially rescued the sarcomeric gene isoform switching defect caused by Carm1 depletion (Figure 4I), but also restored the size of the cardiomyocytes (Figure 4J and 4K), indicating that attenuated insulin signaling is a contributing factor to the reduced cardiomyocyte size caused by Carm1 loss.

In line with the reduced mitochondrial size (Figure 2H), cardiac depletion of Carm1 caused significant downregulation of both Mfn1 and Mfn2, both of which are essential for mitochondrial fusion^33,34 (Figure S3). We also noted that Carm1 loss caused significantly reduced expression of mitochondrial transcription factor Tfam³⁵ and the mitochondrial biogenesis regulator Ppargc1a, (Figure S3). We also observed downregulation of Kcnh2 and Kcnq1 (Figure S3), both of which encode potassium channel required for cardiac action potential repolarization^36–38. Additionally, we found that the downregulated genes associated with the GO term calcium ion transport were mostly those involved in calcium handling, such as Atp2a2 and Ryr2 (Figure S3). These genes are known to be upregulated during cardiomyocyte maturation. In sum, loss of cardiac Carm1 function leads to profound transcriptomic changes that underlie cardiomyocyte maturation cytoarchitectural and functional defects.

CARM1 Directly Activates Genes that Underlie Cardiomyocyte Cytoarchitectural and Electrophysiological Maturation

CARM1 is known to regulate transcription through promoting asymmetric dimethylation of histone H3 at arginine 17 (H3R17me2a)³⁹. To further unravel the molecular mechanism of CARM1-dependent cardiomyocyte maturation, we performed H3R17me2a ChIP-Seq to identify the direct transcriptional targets of CARM1 in hearts. To ensure the specificity of anti-H3R17me2a antibody, we performed pilot ChIP-seq experiment with anti-H3R17me2a and anti-H3R17me2aK18Ac antibodies. Since Histone H3R17 methylation by CARM1 is linked to K18 acetylation⁴⁰, we reasoned that, if both antibodies are of high quality, they could immunoprecipitate significant number of common target DNA fragments. We compared the binding profiles of both H3R17me2a antibodies and observed around 60% of overlapping ChIP peaks (Figure S4A). The genome browser tracks for H3R17me2a(K18Ac) ChIP peaks also indicate that H3R17me2a antibodies showed very similar peaks (Figure S4B), further suggesting the specificity of the H3R17me2a(K18Ac) antibodies. To further evaluate the specificity of the H3R17me2a chromatin immunoprecipitation (IP), specifically the strength of enrichment in the IP samples compared to the input, we plotted the enrichment of IP versus input samples using the deepTools plotFingerprint⁴¹. The cumulative sum for the H3R17me2a samples displays a pronounced and steep rise towards the highest rank (Figure S4C), indicative of strong and specific ChIP enrichment.

We first profiled H3R17me2a on WT mouse heart. The identified H3R17me2a ChIP-seq peaks were mostly located in the promoters and intronic regions of the mouse genome (Figure 5A). We then performed GO analysis for the ChIP-seq peaks associated genes and found a significant enrichment of GO terms associated with cardiomyocyte maturation, including mitochondrial transport, fatty acid catabolic process, muscle hypertrophy, and regulation of cation channel activity (Figure S4D). These GO terms are also enriched for the genes downregulated upon loss of Carm1 function in cardiomyocytes (Figure 4D). Altogether, these observations suggest that deposition of H3R17me2a is involved in the regulation of cardiomyocyte maturation related genes’ expression. To gain further insight into the regulatory mechanism underlying such transcriptional regulation of cardiomyocyte maturation, we performed HOMER analysis, a discovery algorithm for identifying enriched transcription factor (TF) binding motifs⁴², on the H3R17me2a ChIP-seq peaks. This analysis revealed an over-representation of binding motifs for cardiac TFs, including MEF2, GATA factors, NKX2.5 and T-box TFs (Figure 5B). The peaks were also enriched for binding motifs for YY1, zinc finger TFs SP1-SP3, bHLH TFs USF1-USF2, and nuclear hormone receptors including GR, PPAR/RXR and ERRs (Figure 5B). We next sought to evaluate the effect of Carm1 loss on H3R17me2a deposition on cardiac loci. To this end, we generated mosaic mutant hearts that contained AAV9-Cre infected Carm1^KO cardiomyocytes, and AAV9-GFP infected mosaic control hearts. These hearts were then dissected and dissociated into single nuclei, which were then subjected to FACS to enrich the live DRAQ5+ nuclei and then PCM1+ cardiomyocyte nuclei^43,44 (Figure S5A–C). We performed H3R17me2a ChIP on both the control and Carm1^KO cardiomyocytes nuclei. We detected 1661 H3R17me2a peaks that were enriched in the AAV-GFP control cardiomyocytes lost in Carm1^KO cardiomyocytes (Figure 5C and 5D). Conversely, 1123 H3R17me2a peaks that were enriched in AAV-CRE were found to be gained in the Carm1^KO cardiomyocytes (Figure 5C and Figure S6A). Such gain of H3R17me2a peaks may be due to secondary effects resulting from Carm1 loss but caused by other arginine methyltransferases, such as Mettl23⁴⁵, whose expression is increased in Carm1^KO cardiomyocytes. Interestingly, we found that the H3R17me2a peaks lost in the mutant cardiomyocyte nuclei are mostly associated with the genes significantly upregulated in cardiomyocytes during maturation, such as Ryr2 and Ppargc1a (Figure 5E and Figure S6B). Additionally, GO analysis of the genes associated with the differentially lost H3R17me2a peaks revealed enrichment of the cardiac muscle tissue development, suggesting that Carm1 loss leads to the reduction of H3R17me2a modification related to the cardiac muscle maturation program (Figure S6C).

Figure 5. Carm1 loss led to dysregulation of H3R17me2a deposition on cardiomyocyte maturation-associated loci.

A, Distribution of H3R17me2a ChIP-seq peaks in the genome. B, A partial list of the representative motifs enriched in the ChIP-seq peaks. C, Number of differentially enriched H3R17me3a peaks between the control and Carm1^KO cardiomyocyte nuclei. D, Average enrichment plot and heatmap of H3R17me2a peaks lost in Carm1^KO cardiomyocyte nuclei. E, Representative USCS genome browser view of RNA-seq, H3R17me2a ChIP-seq, and ATAC-seq on distal genomic regions annotated to Ppargc1a. A zoomed-in view is presented for each highlighted zone (red dashed box) on the genome browser track. CRE is the Carm1^KO sample (green). GFP is the wildtype control (grey). F, Venn diagram for the overlapping of the downregulated genes (green) and the regions that lost (blue)/gained (red) accessibility in Carm1^KO heart. P-value was calculated with exact hypergeometric probability. G, Motif footprint plot of GATA4 in the open chromatin regions of WT heart.

To further investigate the effect of Carm1 depletion on cardiac epigenetic landscape, we performed Assay for Transposase-Accessible Chromatin with high-throughput sequencing (ATAC-Seq)⁴⁶. which allows for the identification of differentially accessible regions (DARs) between Carm1^KO and control hearts. In total, 2960 DARs lost accessibility were detected in Carm1^KO cardiomyocytes, while 343 DARs gained accessibility. This difference in the number of gained versus lost DARs highlights the role of CARM1 as transcriptional co-activator. To correlate DNA accessibility landscape with H3R17me2a histone modification, we overlapped the DNA accessibility profile with H3R17me2a deposition. Overall, 6925 out of 20878 H3R17me2a peaks showed overlapping with the open chromatin regions from ATAC-seq (Figure S7A). 385 out of 2960 DARs that lost accessibility in Carm1^KO heart are decorated with H3R17me2a modification in control heart, suggesting that the loss of DNA accessibility is partially caused by a direct loss of Carm1 function (Figure S7B). We then associated gene expression level changes with alterations in chromatin accessibility. To avoid bias introduced by a direct association of DNA accessibility change on gene’s promoter and transcription start site (TSS) with gene expression, we selected distal DARs located within 50Kb to the nearest genes. By comparing the downregulated genes to the nearest genes that are associated with distal DARs lost accessibility in Carm1^KO heart, we observed a significant overlap between the downregulated genes to the DARs. In contrast, we didn’t observe any significant association of the downregulated genes with the DARs gained accessibility in Carm1^KO (Figure 5F). Together, our data suggest that the downregulation in gene expression in the mutant heart is associated with a loss of accessibility in distal cis-regulatory elements. TFs work by recruiting cofactors and often work collaboratively with other TFs. To identify which transcription factor CARM1 might function cooperatively to regulate cardiomyocyte maturation, we performed HOMER motif analysis on the DARs lost accessibility upon Carm1 loss (Figure S7C). Consistent with the H3R17me2a ChIP, we observed a significant over-representation of the binding motifs for cardiac TFs. Among them, the binding motif for GATA family members is the most enriched. We then performed TF footprint analysis using HINT-ATAC⁴⁷, and demonstrated that GATA4 showed a strong footprint on these Carm1 dependent open chromatin regions (Figure 5G). In contrast, other cardiac TFs, such as SRF and NKX2–5, did not show such strong footprint (Figure S7D), suggesting that CARM1 could cooperate with cardiac TFs, specifically GATA4, to activate or maintain the cardiac muscle maturation program during cardiomyocyte maturation through H3R17me2a deposition.

Human Disease Relevance of CARM1-mediated Transcriptional Regulation of Cardiomyocyte Maturation

Noncoding variants are widely believed to be important genetic contributors to mendelian and complex human diseases⁴⁸. Given the importance of cardiomyocyte maturation for establishing and maintaining postnatal heart function, we postulated that SNPs in the region surrounding H3R17me2a ChIP-seq peaks might exhibit associations with heart diseases. To test this possibility, we first identified the human genomic regions that are syntenic to the murine H3R17me2a ChIP-seq peaks⁴⁹, and then interrogated the human genomic regions for cardiomyopathy (CP)-, QT interval (QT)-, and arrhythmia (AR)-associated SNPs⁵⁰ (Figure 6A). A significant enrichment of SNPs was found to be in the ChIP-seq peaks compared to genomic background (CP, p=2.55e-06; QT, p=9.00e-71; arrhythmia, p=1.26e-48) (Figure 6B). This analysis identified 41, 69 and 125 SNPs associated with CP, QT and AR, respectively, including associations near genes that have been previously implicated in these disorders, such as BAG3 and ATP2A2^51,52 (Figure 6C). To investigate the potential function of the SNPs in regulating gene expression, we overlapped the GWAS SNPs with GTEx eQTL data of left ventricle⁵³, and found examples of GWAS SNPs being identified as eQTLs as well. For example, the GWAS SNP chr1:162,112,958–162,112,959 overlapped with our ChIP-seq peak at NOS1AP locus was reported to function as an eQTL for NOS1AP. This result indicates that, although not all GWAS SNPs have been characterized at the functional level, some of which affect gene expression (Figure 6D). With the knowledge of the enriched TF binding motifs in the peaks (Figure 5B), we determined how the SNPs could deregulate the CARM1 target genes in human patients. This effort was inspired by the fact that a growing number of noncoding SNPs have been implicated in human heart diseases, yet only few of them have been mechanistically interrogated. Interestingly, 54.5% of our identified SNPs appear to disrupt the binding motifs of TFs, such as NKX2.5 and SP2 by CP-related SNPs, and GATA4 and PPAR/RXR by QT-related SNPs (Figure 6E), at the regulatory regions of the CARM1 target genes identified in our animal model.

Figure 6. Association of H3R17me2a ChIP-seq Peaks with adult heart disease associated SNPs.

A, Flowchart showing approach to GWAS for CP, QT, and AR on GWAS catalogue (Experimental Factor Ontology (EFO) accession number 0000318, 0000407, 0000538, 0004682 and 0004269, only variants with p-value < 5e-8 were considered.). B, Barchart showing significant enrichment of variants located in the H3R17me2a ChIP-seq peaks compared to genomic background. P-value of one-proportion test < 0.05 for all three traits. C, Manhattan plot showing a partial list of the CP-, QT-, and AR-associated SNPs in human genomic regions syntenic to murine H3R17me2a ChIP-seq peaks. D, Inferred eQTL regulatory interaction between GWAS variant chr1:162112958–162112959 and the promoter region of NOS1AP is highlighted in arc. H3R17m2a ChIP-seq peaks are shown in blue. E, Manhattan plot showing the TF binding motifs disrupted by CP-, QT- or AR-associated SNPs. F, Luciferase assay showing the H3R17me3a-associated SNP’s effect on the GATA4’s functionality to active gene expression. G, Schematic of the experiment workflow for mutating SNP regions. H, Sanger sequencing result showing the mutated SNP and the nuclear receptor motif proximal to CASQ2 locus. I, ChIP-qPCR showing enrichment (percentage input) of H3R17me2a in SNP knockout cardiomyocytes (KO) when normalized to wildtype cardiomyocytes. Biological replication of samples, n=3. Error bars represent standard deviation. **P< 0.01. NS, not significant. J, ChIP-qPCR showing enrichment (percentage input) of H3 in SNP knockout cardiomyocytes (KO) when normalized to wildtype cardiomyocytes. Biological replication of samples, n=3. Error bars represent standard deviation. NS, not significant.

As an initial step to assess the functionality of the SNPs, we evaluated the promoter or enhancer activity of the genomic regions around the SNPs. We performed luciferase assay using a 200 bp promoter fragment, wherein a QT-related SNP is identified within a putative GATA4 motif at the ANKRD9 locus in the human genome. The reporter vector containing the major allele (C, Global Frequency = 0.662) exhibited drastically enhanced luciferase activity upon co-transfection of GATA4. This enhancement of luciferase activity by co-expressing GATA4 was, however, significantly reduced by mutating the major allele to minor allele (T, Global Frequency = 0.338) (Figure 6F). Then we performed Electrophoretic Mobility Shift Assay to validate GATA4’s direct binding onto the SNP region. A reduction of GATA4 binding affinity was observed in the minor allele (T) group (Figure S7E–F). We also determined if SNPs could also affect H3R17me2a deposition on the genome. To this end, we selected 4 candidate SNPs that alter the core sequences of putative nuclear receptor binding sites. Utilizing the CRISPR/Cas9 technology, we generated a mouse embryonic stem cell (mESC) clone harboring a 6 bp deletion that removes the SNP nucleotide in the genomic loci for Casq2 (Figure 6G and 6H). We also identified mESC clones with 170 and 50 bp genomic deletion that remove the SNP nucleotide at the Fbxo32 and Cisd2 loci, respectively. We differentiated the mutant mESCs carrying the 6 bp deletion genomic deletion into cardiomyocytes with long-term culture to enhance maturation. Cardiomyocytes differentiated from wildtype mESCs were used as control. To assess the impact of the SNP knockout on H3R17me2a deposition, we performed ChIP-qPCR using a H3R17me2a antibody on the mutant and control cardiomyocytes. Our result revealed a significant reduction in H3R17me2a signal at the locus after removing the SNP nucleotides (Figure 6I), whereas no such trend is observed in the H3 ChIP-qPCR (Figure 6J). Together, our data demonstrate that deregulation of CARM1-dependent cardiomyocyte maturation gene expression may contribute to human heart diseases.

Discussion

Cardiomyocyte maturation is a critical developmental process that transforms cardiomyocyte from an immature fetal to a fully mature adult state. The acquisition of such mature adult phenotypes leads to substantial cytoarchitectural, electrophysiological and metabolic changes that endow cardiomyocyte with much enhanced ability to generate efficient and forceful muscle contractions. Despite the importance of cardiomyocyte maturation, little is known about the molecular underpinnings that regulate this key process. In this study, we found that CARM1 regulates multiple aspects of cardiomyocyte maturation, including cellular hypertrophic growth, myofibril expansion, mitochondrial fusion, T-tubules formation, and electrophysiological maturation. Genomics study indicates that CARM1 directly activates genes that underlie cardiomyocyte cytoarchitectural and electrophysiological maturation, demonstrating a critical and multifaceted role of CARM1 in controlling cardiomyocyte maturation. Furthermore, our interrogation of GWAS datasets indicated that dysregulation of CARM1-dependent cardiomyocyte maturation gene expression contributes to human heart diseases. Thus, defining key factors that regulate cardiomyocyte maturation will offer valuable insight into future diagnosis and treatment of heart disease patients.

HOMER analysis of the H3R17me2a ChIP-seq data revealed an over-representation of binding motifs for cardiac TFs on the ChIP-seq peaks. Among them, SRF and the GATA factors GATA4 and GATA 6 (GATA4/6) were shown to be required for neonatal cardiomyocyte maturation^23,54. This raises important questions whether and how cardiac TFs and CARM1 function cooperatively to regulate cardiomyocyte maturation. HINT-ATAC analysis of the open chromatin regions indicates that GATA factors, instead of the other cardiac TFs such as SRF and NKX2–5, showed a strong footprint on the CARM1 dependent open chromatin regions, suggesting that CARM1 and GATA factors function cooperatively to orchestrate cardiomyocyte maturation. Future study using ChIP-grade antibodies to perform ChIP-seq for GATA4 in control and Carm1^KO cardiomyocytes will further address their relationship. CARM1-deposited H3R17me2a has been reported to mark active genes and open chromatin⁵⁵. However, based on the overlap of H3R17me2a with ATAC-seq (Figure S7A), there are still 60% of peaks that fall outside of these open chromatin structures, suggesting that, in addition to marking the active genes, H3R17me2a might have additional roles in the epigenetic regulation of the chromatin architecture. A more comprehensive comparison of the H3R17me2a profile with other histone modifications during cardiomyocyte maturation could provide insights on the function of these non-overlapped H3R17me2a regions.

Electrophysiological remodeling is a critical part of the maturation process in murine ventricular myocardium and is associated with marked shortening of the APD during the first few weeks of postnatal life^27,56. Here, we recorded a dramatic prolongation of the APD50 and APD90, consistent with a delay in maturation. While numerous changes in ion channel subunit expression occur in the postnatal heart, a large increase in expression of rapidly activating repolarizing currents is a key component of this process in murine ventricular myocardium⁵⁷. It is likely that failure of this coordinated gene expression program to occur accounts for the phenotype we observed. In addition, it is well established that APD varies systemically across the ventricular wall in mouse and human ventricular myocardium, allowing for ordered and sequential repolarization^58,59. Importantly, we were careful to sample similar regions of mouse ventricular myocardium in wildtype and mutant hearts, and the magnitude of changes we observed here in APD50 and APD90 are much greater than what has been previously reported for transmural differences in mouse myocardium, which together exclude regional sampling bias as a source of the differences we observed.

Noncoding variants have been increasingly recognized as critical contributors to Mendelian and complex genetic disorders in humans. While some of the non-coding variants are located in the untranslated regions and could affect post-transcriptional regulation of gene expression, others reside in functional non-coding regions such as cis-regulatory elements including promoters and enhancers, and could influence gene expression at a transcriptional level⁴⁸. In this study, we found that a significant enrichment of human heart diseases associated GWAS SNPs in the human genomic regions syntenic to the mouse cardiac H3R17me2a ChIP-seq peaks. Interestingly, a significant proportion of these GWAS SNPs are found to be in the putative cardiac TF binding sites. Our initial attempt to assess the functionality of the SNP suggests that at least some of the SNPs might negatively affect promoter/enhancer activity. Additionally, we also found that a small genomic deletion surrounding the SNP in the Casq2 locus significantly reduced the deposition of H3R17me2a mark in the cardiomyocytes. This approach doesn’t rule out the possibility that other deleted nucleotides affect H3R17me2a deposition, future study using prime editing to introduce the SNP nucleotide will further clarify the role of the GWAS SNPs.

In the current study, due to the limited access to human sample, we utilized a mouse H3R17me2a ChIP-seq dataset and mapped the resulting peaks to the human genome, identifying conserved H3R17me2a peaks that overlap with human GWAS SNPs. Moving forward, performing H3R17me2a ChIP-seq on human samples would significantly enhance our understanding of CARM1’s role in human cardiomyocyte maturation. The ATAC-seq presented in the current study is performed with the entire mutant heart tissue due to technical difficulties. It could detect chromatin accessible regions in non-myocytes. Thus, in the future, performing ATAC-seq with isolated cardiomyocytes could detect more cardiomyocyte-specific changes of open chromatin structures upon Carm1 loss.

Clinical Perspective.

What Is New?

• Ablation of the epigenetic regulator Carm1 during murine neonatal development impedes cardiomyocyte maturation, resulting in reduced cardiomyocyte size and sarcomere thickness, severe loss and disorganization of T-tubules, and compromised electrophysiological activity.

• CARM1 directly activates genes underlying cardiomyocyte cytoarchitectural, electrophysiological, and muscle maturation through methylating histone H3R17.

• Carm1 loss causes the inaccessibility of distal cis-regulatory elements, leading to downregulated gene expression.

What Are the Clinical Implications?

• SNPs in promoter/enhancer regions of the human genome syntenic to murine H3R17me2a deposition sites associated with cardiomyopathy, QT, and arrhythmia, suggesting that the deregulation of CARM1-dependent cardiomyocyte maturation gene expression may contribute to human heart diseases.

• Defining the regulatory mechanisms underlying cardiac maturation offers insights into the factors and pathways needed to produce fully functional mature cardiomyocytes, opening future treatment options and diagnosis parameters for patients with heart disease.

Non-standard Abbreviations and Acronyms

SNPs

Single Nucleotide Polymorphisms

T-tubules

Transverse-tubules

PRMTs

Protein arginine methyltransferases

ChIP-seq

Chromatin immunoprecipitation followed by sequencing

CP

Cardiomyopathy

QT

QT Interval

AR

Arrhythmia

WGA

Wheat Germ Agglutinin

AAV-Cre

AAV9-cTnT-Cre-IRES-GFP

WT

Wild Type

IEM

Immuno-Electron Microscopy

APD

Action Potential Duration

RMP

Resting Membrane Potential

HW/TL

Heart weight to Tibia Length ratio

PCA

Principal Component Analysis

DEGs

Differentially Expressed Genes

GO

Gene Ontology

IP

Immunoprecipitation

TF

Transcription Factor

FACS

Fluorescence-Activated Cell Sorting

ATAC-seq

Assay for Transposase-Accessible Chromatin with high-throughput sequencing

DARs

Differentially Accessible Regions

TSS

Transcription Start Site

GWAS

Genome-Wide Association Studies