FTO-Dependent m6A Regulates Cardiac Function During Remodeling and Repair

Abstract

Background:

Despite its functional importance in various fundamental bioprocesses, the studies of N6-methyladenosine (m6A) in the heart are lacking. Here we show that, fat mass and obesity-associated (FTO), an m6A demethylase, plays a critical role in cardiac contractile function during homeostasis, remodeling and regeneration.

Methods:

We used clinical human samples, preclinical pig and mouse models and primary cardiomyocyte cell cultures to study the functional role of m6A and FTO in the heart and in cardiomyocytes. We modulated expression of FTO using AAV9 (in vivo), adenovirus (both in vivo and in vitro) and siRNAs (in vitro) to study its function in regulating cardiomyocyte m6A, calcium dynamics and contractility and cardiac function post-ischemia. We performed methylated (m6A) RNA immunoprecipitation sequencing (MeRIP-seq) to map transcriptome-wide m6A, and MeRIP qPCR assays to map and validate m6A in individual transcripts, in healthy and failing hearts and myocytes.

Results:

We discovered that FTO has decreased expression in failing mammalian hearts and hypoxic cardiomyocytes, thereby increasing m6A in RNA and decreasing cardiomyocyte contractile function. Improving expression of FTO in failing mouse hearts attenuated the ischemia-induced increase in m6A and decrease in cardiac contractile function. This is carried out by the demethylation activity of FTO, which selectively demethylates cardiac contractile transcripts, thus preventing their degradation and improving their protein expression under ischemia. Additionally, we demonstrate that FTO overexpression in mouse models of MI decreased fibrosis and enhanced angiogenesis.

Conclusion:

Collectively, our study demonstrates the functional importance of FTO-dependent cardiac m6A methylome in cardiac contraction during heart failure and provides a novel mechanistic insight into the therapeutic mechanisms of FTO.

INTRODUCTION

Current therapeutic approaches have resulted in limited success in treating ischemic heart disease and mitigating post-ischemic adverse cardiac remodeling. Therefore, new concepts for myocardial repair and regeneration that improves cardiac function have to be developed. While several transcription factors and transcriptional co-activators have been studied in the context of heart failure, post-transcriptional regulations of cardiac mRNAs that can affect expression of key proteins and cardiac function remain largely unexplored. Earlier studies have shown that i) protein expression in failing human hearts does not correlate with the corresponding mRNA abundance ^{1, 2}, and ii) nuclear and cytosolic mRNA levels do not correlate in cardiac myocytes ³, suggesting a role for altered post-transcriptional RNA modifications in regulating protein expression in failing hearts. Therefore, we investigated epitranscriptomic regulations underlying cardiac remodeling and found a predominant but previously unidentified mechanism contributing to heart failure via dysregulated RNA modifications.

Even though epitranscriptomic mechanisms are under intense study, our knowledge about its biological function, especially in healthy tissues, organs and under pathological conditions are limited. Recent findings suggest that the most abundant internal chemical modification in RNA, N6-methyladenosine (m6A) is a critical regulator of mRNA stability, protein expression and several other cellular processes ^4–7. While dysregulated m6A has been linked to various types of cancers ^{8, 9} and brain diseases ¹⁰, its role in cardiac homeostasis and failure has not been studied.

Several recent discoveries suggest that epitranscriptomic mRNA modifications are reversible and dynamically regulated with dedicated writers (methyltransferases) that catalyze addition of m6A (METTL3, METTL4, METTL14 and WTAP) and dedicated erasers (demethylases) that catalyze removal of m6A (FTO, ALKBH5) from mRNA ¹¹. The fat mass and obesity-associated (FTO) protein has been found to have a key role in regulating transcriptome-wide m6A modification in mRNA ^{4, 12} and is one of the m6A regulators that has been associated with metabolic disorders such as diabetes and obesity ^{13, 14}. Although FTO is expressed ubiquitously, cardiac ventricular levels are high in addition to brain and liver tissues in human embryos ¹⁵. Interestingly, FTO has been implicated in cardiac defects including hypertrophic cardiomyopathy, ventricular septal and atrio-ventricular defects ¹⁵, arrhythmias ¹⁶ and coronary heart disease ¹⁷ suggesting an important role for FTO in ischemic heart failure.

So far, m6A function in physiological and biological processes have been investigated. However, studies of m6A under pathological conditions, especially in tissues and organs are limited. Moreover, FTO-dependent m6A demethylation and its role in cardiac protein expression and contractility in healthy and failing myocardium have not been addressed. Using clinical human samples, preclinical pig and mouse models of heart failure and primary cardiomyocyte cell culture, we have investigated the physiological and pathological role of FTO-dependent m6A epitranscriptome in cardiac homeostasis, remodeling and repair. We present evidence that m6A in RNA is dysregulated in failing hearts and that FTO-dependent m6A plays a significant role in pathomechanisms of heart failure at the molecular (mRNA degradation and protein expression), cellular and organ (cardiomyocyte and cardiac function) levels. Moreover, we show that FTO gene delivery attenuates the ischemia-induced cardiac remodeling demonstrating the therapeutic potential of FTO in the treatment of heart failure. Mechanistically, FTO selectively demethylates calcium handling and contractile transcripts, preventing their degradation and regulating their protein expression in the failing heart. Our results have uncovered a novel function for FTO-regulated m6A mechanisms in cardiac remodeling and repair. We also establish that targeting cardiac epitranscriptome via FTO can be an effective therapeutic strategy for heart failure.

METHODS

The data, analytic methods, and study materials will be made available to other researchers for purposes of reproducing the results or replicating the procedure.

Study design

Human left ventricular (LV) tissues were obtained from deidentified, postmortem failing (with diagnosed cardiomyopathy) and non-failing (with no history of cardiomyopathy and non-cardiac related death) hearts from National Disease Research Interchange (NDRI). No human subjects were involved. Mount Sinai Institutional Review Board approved procurement of human tissue samples from NDRI and all usage was done as per Mount Sinai approved guidelines. Both male and female tissues were used from human failing and non-failing hearts. Similarly, for all animal experiments, both male and female mice (c57Bl6) and rats (Sprague-Dawley rats for cardiomyocyte isolation) were used. All the experimental protocols are in compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and standards of United States regulatory agencies. All treatments and measurements of cardiomyocyte (Ionoptix) and cardiac (echocardiography) function were performed in a blinded manner. To investigate m6A in cardiac transcriptome, we developed novel methods such as m6A quantification in total or polyA⁺ RNA and methylated (m6A) RNA immunoprecipitation and sequencing (MeRIP-seq) in RNA isolated from mouse and human cardiac tissues. m6A-related new protocols are described here. Rest of the methods are described in detail in the Online-only Data Supplement.

PolyA⁺ RNA isolation from human and mouse LV

For human polyA⁺ RNA, ~100mg of LV tissue was homogenized using multiple RNase/DNase free 2 ml lysing matrix tubes (MP Biomedicals) using tissue homogenizer. For mouse polyA⁺ RNA isolation, 3–4 sham or 5–6 MI (infarct) LV tissues were pooled to achieve ~100mg tissue and homogenized as described for human samples. RNA was extracted using miRNAeasy kit (Qiagen) as recommended by manufacturer (more detail in Online Supplement). For polyA⁺ RNA isolation, ~150ug of total RNA in 500ml RNAse-free water from each LV extract was then pre-warmed in a 65^oC-heating block for 10 minutes before adding Biotinylated-Oligo (dT) probe for hybridization (Promega). After allowing Oligo-(dT) hybridization to poly(A) tail for 10 minutes in room temperature, Streptavidin-Paramagnetic Particles (Promega) were added and polyA⁺ containing RNA was isolated using magnetic capture of biotin-streptavidin conjugates. PolyA⁺ RNA was then eluted in RNase-free water as recommended by the manufacturer’s protocol. The isolation of polyA⁺ RNA fraction was ensured by analyzing polyA⁺ RNA in a bioanalyzer. PolyA⁺ RNA was immediately quantified using nanodrop spectrophotometer and RNA samples were stored as aliquots at −80^oC.

Quantification of m6A in total and polyA⁺ RNA

Column purified total RNA (miRNeasy, Qiagen) or Oligo-dT enriched polyA⁺ RNA (Promega) was used for quantification of m6A modification. For quantification of m6A in human, pig and mouse failing or non-failing ischemic LV, we used previously described antibody-based m6A capture and colorimetric quantification method (P-9005, EpiGentek)^{18, 19}. For both total or polyA⁺ RNA, 200ng of RNA per sample was transferred to corresponding wells in a 96-well plate. All samples were performed in duplicates. Both negative and positive controls as well as a standard curve in the range of 0.02ng to 1ng of m6A were included as recommended by the manufacturer. Briefly, after binding of RNA to wells, anti-m6A antibody was added and washed three to four times as recommended by manufacturer’s protocol. Antibody bound m6A in RNA was detected by adding developer solution and colorimetric quantification was carried out subsequently using SpectraMax plus microplate reader. Percentage of m6A in RNA was calculated and compared to corresponding control samples and established as % m6A in total RNA.

Methylated RNA immunoprecipitation and next generation sequencing (MeRIP-seq)

Briefly, 200ug of human or mouse total RNA was used. For mouse, at least 4 LVs per sham group and at least 6 LVs per MI group were pooled to achieve 200ug of RNA (RIN>8). Both human and mouse total RNA were then isolated for polyA⁺ RNA (Promega) and RNA quantified. PolyA⁺ RNA was then fragmented to ~100nt long fragments using RNA fragmentation buffer (Millipore Sigma). RNA fragmentation was ensured by bioanalyzer before proceeding to m6A-IP. Non-IP RNA was stored from fragmented RNA for bioinformatics analysis. The remaining fragmented polyA⁺ RNA was subjected to m6A-IP using 5ug of anti-M6A (17–10499, Millipore Sigma)²⁰. Antibody bound m6A-modified RNAs were eluted in RNase-free water as recommended by the manufacturer’s protocol. Immunopurified RNA was then precipitated overnight at 4^oC and samples were sent to Weill Cornell Medical College Genomic Core facility (New York) for library construction. The usual steps involving polyA⁺ enrichment and fragmentation during library construction were omitted and libraries were constructed using RNA-seq TruSeq-straded mRNA type. Libraries were sequenced by the Illumina HiSeq 2500 platform (Illumina) as 50bp single reads. Unaligned reads were mapped to the mm10 reference genome using STAR v2.5.3a at default settings. For each condition, total MeRIP reads per transcript were calculated for reads that fully overlapped with a given feature, and were normalized to total non-IP reads per corresponding transcript. This ratio was normalized to corresponding Sham values, and was defined as fold change (FC). Differential peak analysis of m6A MeRIP-Seq datasets were carried out using a modification of exomePeak R/Bioconductor package to compare the ratio of the absolute number of MeRIP reads to non-IP reads at a given peak between two conditions. To remove false positive signaling in our differential methylation analysis, we have used the same input read threshold as recommended²¹.

In other words, differential methylation between two conditions at a given peak was calculated as:

$$\text{FC}=\left({\text{Bound}}_1/\text{non-}{\text{IP}}_1\right)/\left({\text{Bound}}_2/\text{non-}{\text{IP}}_2\right)$$

where FC denotes fold change; Bound₁ and Bound₂ are the number of reads within a peak for the MeRIP samples for conditions 1 and 2, respectively; and non-IP₁ and non-IP₂ are the number of reads within a peak for the non-IP samples for conditions 1 and 2, respectively. In our differential methylation analysis, fold changes of ±2 between two conditions at a given peak and a formal test with proper FDR (false discovery rate) control (FDR<0.05) were included and considered significant. The data have been deposited in the GEO repository with the accession number GSE112789. MeRIP tracks were visualized with Integrative Genomics Viewer (IGV), using filtered (see above for filtering methods) .bam files from each group at select loci.

For MeRIP-qPCR, immunopurified RNA was purified and first strand cDNA synthesis was carried out as described earlier. Enrichment of mRNA in m6A-immunopurified samples was expressed relative to 18s rRNA in bound samples and expressed as fold change between groups. Taqman primers were used for testing m6A enrichment within transcripts; for RYR2 (Hs00181461_m1), ATP2A2/SERCA2A (Hs00544877_m1).

The DAVID bioinformatics database was used for gene ontology (GO) analysis on significant differentially methylated MeRIP peaks (defined above). GO classification was performed at default settings.

Statistical methods:

Data are shown as mean +/− SEM unless otherwise stated. One-way analysis of variance (ANOVA) was used to determine statistical significance for experiments with more than two groups followed by Bonferroni’s post hoc tests. Figures with ANOVA analysis where applicable are indicated in corresponding figure legends. Comparison between two groups were carried out using GraphPad software with an unpaired Student’s t-test. P-values <0.05 were considered statistically significant and assigned in individual figures.

RESULTS

Increased m6A in RNA in human, pig and mouse failing hearts

We quantified m6A levels in RNA extracted from failing human (both ischemic and non-ischemic), pig and mouse (post-myocardial infarction ischemic) hearts and compared them to m6A in non-failing human and sham surgical controls respectively. We detected significantly elevated levels of m6A in both total and polyA⁺ RNA extracted from human, pig and mouse failing left ventricular (LV) explants compared to non-failing or sham (Figure 1A-C). The increase in m6A in total RNA was observed as early as one-week post-MI in mice (Figure 1C) and two weeks post-MI in pigs (Figure 1B). We also observed a sustained increase in m6A in both total and polyA⁺ RNA in chronic phases of MI-induced heart failure in both mouse and pig ischemic LVs measured at four weeks and twenty weeks respectively (Figure 1B and 1C). Interestingly, the increase in m6A appeared to be confined only to the infarct and peri-infarct regions, as it was not detected in the non-infarct (remote) LV tissues in both pigs and mice in all time points investigated (Figure 1D and 1E). These results provide strong evidence of increases in m6A in RNA in chronic heart failure conditions in humans, which were conserved across species in swine and mouse.

Figure 1. Increased m6A in RNA in failing human (both ischemic and non-ischemic), pig and mouse (post-myocardial infarction ischemic) hearts.

Quantification of m6A in total or polyA⁺ RNA in LV of A, human, n=6–11; from infarct/peri-infarct areain B, pig, n=3–6, C, mouse, n=5–11; from non-infarct area in D, pig, n=3 and E, mouse, n=3–6. Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001, compared with non-failing or sham. n.s., non significant.

Decreased FTO expression in human and mouse failing hearts

To identify regulators of elevated m6A in failing hearts, we measured expression levels of several known RNAs and proteins associated with m6A methylation (writers, such as METTL3, METTL4 and METTL14 and their regulatory subunit, WTAP) and demethylation (erasers, such as FTO and ALKBH5) ⁵ in human and mouse hearts. Western blot and qRT-PCR data revealed that expression of FTO (for human; for mouse/rat: Fto) was significantly decreased in failing LV explants from human and mouse, both at the RNA and the protein levels as compared to their respective controls (Figure 2A-2F). The post-ischemic decrease of Fto was detected as early as four hours post-MI in mice hearts and the decrease in Fto mRNA and protein levels consistently correlated with increased m6A at one week and four weeks post-MI in mouse (Figure 2D-2F). Moreover, the transient, inconsistent increase or decrease of other m6A writers and erasers (Figure 2G and Supplemental Figure 1A-1D) could not fully explain the aberrant and sustained increase in m6A in the failing hearts. Interestingly, among all the m6A writers and erasers studied, FTO had the highest baseline expression both at the RNA and protein levels in healthy human (Figure 2A and Supplemental Figure 2A-2B) and no-surgery mouse (Supplemental Figure 2C-2E) LV tissues. Collectively, these data established that ischemia-induced loss of FTO could be an important molecular hallmark that may explain the increase in m6A in human and mouse failing hearts.

Figure 2. Decreased FTO mRNA and protein expression in human and mouse failing hearts.

A,quantification of mRNA, n=3–6, B, representative immunoblots, C, densitometry quantification of protein, n=5–8 for m6A regulators in human non-failing and failing hearts. D, quantification of mRNA, n=3–7, E, representative immunoblots and F, densitometry quantification of protein, n=3–4 at different time points in mouse LV. G, qRT-PCR quantification of selected mRNA expressions in mouse LV, n=4–8. mRNA/protein data represented as F/MI normalized to NF/sham. Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001, compared with non-failing or sham.

Fto-dependent m6A demethylation regulates intracellular Ca²⁺ and sarcomere dynamics in cardiomyocytes

To investigate the role of Fto in cardiomyocytes and to determine if Fto is a direct regulator of m6A, we established cell culture models with loss of Fto using siRNA to Fto (siFto) and gain of Fto using adenovirus carrying Fto (adFto) in isolated adult rat primary cardiomyocytes and compared to siCtrl and adnull controls, respectively (Figure 3A and Supplemental Figure 3A-3B). Interestingly, expression of Fto inversely correlated with the level of m6A in total RNA in primary myocytes (Figure 3A and 3B). Similar to ischemic mouse and human hearts, primary cardiomyocytes subjected to hypoxia had decreased Fto expression (Figure 3A) and increased m6A in RNA (Figure 3B). Overexpressing Fto in cardiomyocytes cultured under hypoxia reversed the hypoxia-induced aberrant increase in m6A in RNA (Figure 3B) suggesting that Fto is a key regulator of m6A in cardiomyocytes. On the other hand, m6A and Fto levels in adult rat primary non-myocytes subjected to hypoxia remained unchanged compared to normoxic non-myocytes (Supplemental Figure 3C and 3D) suggesting that myocytes are more responsive to Fto-dependent m6A dysregulation than non-myocytes. In the same line, we found that primary cardiomyocytes isolated from healthy human LV has significantly higher expression of FTO compared to primary non-myocytes from the same hearts (Supplemental Figure 4). In addition, human ventricular myocyte cell line, but not human cardiac fibroblasts or endothelial cells in culture had significantly lower FTO mRNA expression under hypoxia stress (Supplemental Figure 4). Together, these data indicate that myocytes regulate FTO expression under hypoxic stress.

Figure 3. Fto-mediated m6A demethylation regulates intracellular Ca²⁺ recycling and contractile dynamics in isolated adult rat primary cardiomyocytes.

Quantification of A, Fto mRNA, n=3–4, B, m6A in total RNA, n=3–5, C, cells with arrhythmic events, n=47–90 cells per group from 4–12 rats. D, Representative Ca²⁺ transients obtained from pacing-induced myocytes. Measurements of E, maximal Ca²⁺ amplitude, F, time to 50% decay, G, Tau, H, cell shortening, n=23–74 cells per group from 4–12 rats. Sarcomere and Ca²⁺ transients were recorded at 1Hz pacing stimulation frequency with MyoPacer Field Stimulator (IonOptix MA, USA). Abbreviations, Unt: untreated; siCtrl: siRNA control, siFto: siRNA-mediated Fto knockdown, adnull: adenovirus with empty CMV promoter, adFto: adenovirus with full length Fto. Error bars represent SEM. *P<0.05, **P<0.01, ***P<0.001by one-way ANOVA.

To investigate the functional significance of Fto-modulated m6A in cardiomyocyte contractile function in vitro, we measured pacing-induced calcium (Ca²⁺) dynamics and contractile properties in Fto-modulated primary cardiomyocytes cultured under normoxic and hypoxic conditions. Remarkably, siFto treated myocytes with higher m6A levels exhibited significantly increased number of arrhythmic events as compared to siCtrl cardiomyocytes (Figure 3C and 3D), supporting an earlier observation that a decrease in Fto-demethylase activity leads to proarrhythmic remodeling and altered ventricular repolarization in Fto-deficient mice hearts¹⁶. On the other hand, hypoxia-induced cardiomyocyte dysfunction was significantly improved by overexpressing Fto in hypoxic cardiomyocytes which resulted in improved Ca²⁺ amplitude (Figure 3E, and Supplemental Figure 5A), accelerated Ca²⁺ decay (Figure 3F, 3G and Supplemental Figure 5B), increased sarcomere shortening (Figure 3H) and maximum and minimum shortening velocities (Supplemental Figure 5C and 5D) as compared to adnull treated cardiomyocytes. Collectively, these data demonstrate that Fto attenuated hypoxia-induced cardiomyocyte dysfunction in vitro.

Myocardial Fto gene transfer attenuates ischemia-induced loss of cardiac function in mouse failing hearts

To investigate the beneficial function of Fto during cardiac remodeling in vivo, we overexpressed Fto using AAV9 (aavFto) for sustained Fto overexpression pre-MI (Figure 4A, 4B and Supplemental Figure 6A-6C) or using adenovirus (adFto) for transient overexpression post-MI (Figure 5A and 5B) and compared them with the control ischemic (aavgnp-MI or adnull-MI) hearts. Both AAV- and adeno-mediated overexpression of Fto significantly decreased the MI-induced increase of m6A in RNA (Figure 4C and 5C and Supplemental Figure 6D) but did not affect the expression of other m6A writers and erasers (Supplemental Figures 6E-6G, 7A-7D and 8A-8D). Body weight measurements throughout the study period indicated no significant changes between control and Fto overexpressing mice (Supplemental Figure 9A-D) suggesting that either sustained or transient overexpression of FTO in the myocardium may not have direct effect on body weight. Interestingly, Fto overexpression significantly improved cardiac function at the chronic stages of post-MI as indicated by higher ejection fraction (Figure 4D and 5D, Supplemental Figure 10A, 10D and 10G), fractional shortening (Figure 4E and 5E, Supplemental Figure 10B, 10E and 10G)and improved wall motion (Figure 4F and 5F, Supplemental Figure 10C and 10F) both at two and four weeks post-MI. To address the functional mechanisms of FTO-induced cardiac repair, we quantified fibrosis and angiogenesis in FTO overexpressing (AAV or adenovirus) mice and compared them to AAV/adenonull mice post-MI. Interestingly, fibrosis, as determined by scar size (%), was significantly reduced in FTO overexpressing (aavFto-MI or adFto-MI) mice compared to respective controls (aavgnp-MI or adnull-MI) quantified at four weeks (Figure 4G, 4H and 5G, 5H). Similarly we determined angiogenic response by quantifying CD31-positive endothelial cells in the murine hearts at the infarct border zone at four weeks post-MI. Both aavFto and adFto overexpressing hearts had significantly higher number of CD31-positive cells compared to aavgnp and adnull control hearts (Supplemental Figure 11). These data suggest that both sustained and transient overexpression of Fto resulting in lower m6A is beneficial for the heart and can improve MI-induced cardiac dysfunction. Moreover, transient overexpression of Fto starting after few days of adenovirus injection post-MI significantly improved cardiac function suggesting potential therapeutic application of FTO or FTO mimics in the treatment of heart failure.

Figure 4. AAV9-mediated myocardial FTO gene transfer rescues cardiac function in mouse models of MI.

A, Design of aavFto study. B, representative immunoblots showing Fto protein expression at week 0 of the aavFto study. C, m6A quantification in total RNA at week 4 of aavFto study, n=4–7. Echocardiographic assessments of LV function showing D, ejection fraction (EF) and E, fractional shortening (FS) at 4w post-MI, n=10–13. F, representative M-Mode echocardiograms showing anterior and posterior LV wall motion at 4w post-MI in aavFto mice. Error bars represent SEM. G, Representative Masson trichrome staining images of histological cross-sections taken in bright-field mode processed from sham, aavgnp-MI or aavFto-MI. H,Scar Size was measured as percentage of total LV area after Masson trichrome staining in sham, aavgnp-MI and aavFto-MI at four weeks post-MI surgeries. Error bars represent SD. *P<0.05, ***P<0.001by one-way ANOVA.

Figure 5. Adenovirus-mediated myocardial FTO gene transfer rescues cardiac function in mouse models of MI.

A, Design of adFto study. B, representative immunoblots showing Fto protein expression at week 4 of the adFto study. C, m6A quantification in total RNA at week 4 of adFto study, n=3–5. Echocardiographic assessments of LV function showing D, ejection fraction (EF) and E, fractional shortening (FS) at 4w post-MI, n=6–9. F, representative M-Mode echocardiograms showing anterior and posterior LV wall motion at 4w post-MI in adFto mice. Error bars represent SEM. G, Representative Masson trichrome staining images of histological cross-sections taken in bright-field mode processed from sham, adnull-MI or adFto-MI. H,Scar Size was measured as percentage of total LV area after Masson trichrome staining in sham, adnull-MI and adFto-MI at four weeks post-MI surgeries. Error bars represent SD. *P<0.05, ***P<0.001 by one-way ANOVA.

Contractile transcripts hypermethylated in failing hearts are demethylated by Fto overexpression

To identify the m6A demethylating targets of Fto in the transcriptome, we performed immunoprecipitation of m6A-modified RNA (MeRIP) followed by sequencing from sham, MI, aavgnp-MI and aavFto-MI mouse LV tissues from peri-infarct area (Supplemental Figure12) ^{4, 22}. Differential expression of individual transcripts across conditions can affect perceived m6A enrichment in MeRIP; therefore, to identify differential m6A modifications across conditions independent of varying transcription ^{4, 21}, we normalized MeRIP reads to corresponding non-IP reads within each condition using a modified ExomePeak analysis (Supplemental Figure12).

Transcriptome-wide methylation represented by total number of MeRIP reads indicated global hypermethylation in MI compared to sham (Figure 6A). These data are consistent with our earlier observation of increase in m6A in failing LVs in vivo (Figure 1A-1C) and in adult primary myocytes under hypoxia in vitro (Figure 3B). While this hypermethylation effect was also evident in aavgnp-MI, overexpression of Fto (aavFto-MI) led to apparent global demethylation (Figure 6A). Analysis of total MeRIP reads (normalized to non-IP reads) for each transcript in sham, MI, aavgnp-MI and aavFto-MI suggests overexpressing Fto in MI hearts reversed the MI-induced global hypermethylation as aavFto-MI clustered with sham (Supplemental Figure13A). It also suggests that Fto demethylates a subset of transcripts (Figure 6B) that are largely associated with cardiac hypertrophy, muscle contraction, filament sliding and sarcomere organization (Figure 6C and Supplemental Tables S1-S8). MeRIP-seq map of individual transcripts show hypermethylated sites in MI compared to sham for several Ca²⁺ and contractile transcripts such as Nppa, Myh7, Serca2a, Ryr2, as well as myocardium-specific lncRNAs Mhrt and Chast, which were demethylated in aavFto-MI (Figure 6D and Supplemental Tables S3 and S4). In addition, our MeRIP-seq identified that FTO overexpression resulted in regulation of angiogenic and fibrotic (ECM) pathways (Supplemental Tables S4 and S8).Our results confirm prior observations of Serca2a methylation in 3T3L1 pre-adipocytes and in HeLa cells in response to heat shock ^23–25. PCR analyses of MeRIP enriched RNA from human LV tissues confirmed that Ca⁺² handling transcripts SERCA2a, and RYR2 were hypermethylated in the failing human hearts (Figure 7A). Consistent with this, SERCA2a mRNA was demethylated when Fto was overexpressed in human myocytes (Figure 7B and Supplemental Figure13B and 13C) resulting in increased SERCA2a mRNA expression (Figure 7C). Indeed, these effects were translated to mouse post-MI in FTO overexpressing LV tissues (Figure 7D and Figure 7E). Demethylation of SERCA2a mRNA and its association with increased mRNA levels could be a result of FTO increasing the stability of SERCA2a mRNA and possibly of other contractile mRNAs ^{26, 27} or a co-transcriptional regulation by FTO ¹⁹. Our in vivo data showing increased Serca2a protein expression in aavFto-MI and adFto-MI mice corroborated our findings (Figure 7F). Collectively, our MeRIP analysis confirmed transcriptome-wide hypermethylation post-MI, demethylation with Fto gene delivery and Fto targeting a subset of mRNAs involved in positive regulation of calcium and contractile function during cardiac remodeling.

Figure 6. Contractile transcripts hypermethylated in failing hearts are demethylated by Fto overexpression.

A, total MeRIP bound reads from mouse MeRIP-Seq. B, mouse heatmap showing total MeRIP reads normalized to total non-IP reads within each transcript relative to sham. For each condition, total MeRIP reads per transcript were calculated for reads that fully overlapped with a given feature, and were normalized to total non-IP reads per corresponding transcript. C, top 10 DAVID GO terms enriched (±2-fold + FDR<0.05) from mouse MeRIP-Seq. Differential peak analysis of m6A MeRIP-Seq datasets were carried out using a modification of exomePeak R/Bioconductor package to compare the ratio of the absolute number of MeRIP reads to non-IP reads at a given peak between two conditions. D, IGV plots of mouse MeRIP reads for selected transcripts, IGV numbers indicate scale for MeRIP reads for all conditions per transcript.

Figure 7. Serca2a mRNA is hypermethylated in human failing hearts and demethylation by FTO overexpression induces Serca2a mRNA and protein expression.

A, human MeRIP-qPCR showing m6A enrichment in mRNA. B,FTO mRNA and C, SERCA2A mRNA, n=3 expressions in AC16 human myocytes. Fto mRNA expression in D, aavFto-MI and E, adFto mice 4w post-MI, n=3–5. F, Representative immunoblots showing Serca2a protein expression at 4w post-MI. Error bars represent SEM. *P<0.05, ***P<0.001. A-C, student t-test; D and E, one-way ANOVA.

Our results have uncovered dysregulation of m6A as an important molecular hallmark of post-ischemic cardiac remodeling and a novel layer of gene regulation at the RNA level in heart failure.

DISCUSSION

In this study, we have found compelling in vitro, in vivo and translational evidence demonstrating an important role for FTO in cardiomyocyte and cardiac function under physiological, pathological and reparative conditions. We have identified FTO as a key myocardial demethylase that regulates cardiac m6A and provided a novel characterization of FTO-dependent m6A in cardiac contractile function. FTO expression is downregulated in heart failure, leading to aberrant increase in global cardiac m6A as well as m6A in selective contractile transcripts leading to their decreased protein expression. Loss of FTO resulted in anomalous calcium handling and sarcomere dynamics resulting in loss of contractile function in primary cardiomyocytes. Interestingly, forced expression of FTO in stressed hypoxic cardiomyocytes or failing murine myocardium attenuated ischemia-induced cardiac remodeling, loss of cardiac contractile protein expression and loss of cardiac contractile function, demonstrating therapeutic potential of FTO. By comparing the cardiac m6A maps of individual transcripts we discovered that FTO selectively demethylates cardiac contractile transcripts such as SERCA2A, MYH6/7, RYR2 and many others improving their mRNA and protein expression. We have further shown that the cardioprotective mechanism of FTO is mediated by selective demethylation of cardiac contractile transcripts under ischemia, which increases mRNA stability and protein expression^{10, 28, 29}. Our MeRIP-seq pathway analysis indicate that FTO acts on selective cardiac pathways such as those relevant to sarcomere organization, myofibril assembly, calcium handling and contractility (Figure 6C, Tables S3, S4, S7 and S8). In addition, it also acts on pathways related to angiogenesis and fibrosis (ECM) and on lncRNAs including Chast or Mhrt, which are implicated in fibrosis and hypertrophy (Figure 6D, Tables S4 and S8). In addition to restoring contractile protein expression such as SERCA2A, we provided experimental evidence that FTO overexpression decreased cardiac fibrosis and enhanced angiogenesis in the ischemic myocardium. Our study demonstrates functional importance of FTO-dependent cardiac m6A methylome in cardiac contraction, fibrosis and angiogenesis during heart failure and provides robust mechanistic insights into the therapeutic potential of FTO (Figure 8).

Figure 8. Proposed working model based on our hypothesis.

A, Healthy heart with physiological FTO and m6A levels, B, Failing heart with decreased FTO, increased m6A, increased contractile mRNA degradation and dysfunctional myofilament, C, Fto rescued failing heart with attenuated m6A and restored contractile protein expression and myofilament.

FTO is a dioxygenase that oxidatively demethylates m6A-containing mRNAs ¹². Thus, the enzymatic activity of FTO may be expected to decrease under hypoxic or ischemic conditions independent of its expression levels. Interestingly, we also found that both hypoxia in primary cardiomyocytes and ischemia in mouse hearts have reduced expression of FTO, both at the protein and the RNA level. There could be several direct and indirect mechanisms for FTO downregulation under hypoxic/ischemic stress, including hypoxia-inducible miRNA binding. We identified binding sites in FTO mRNA for several hypoxia inducible cardiac miRs such as miR-21, miR-24, miR-488, miR-224, miR-489 and miR-199:miR-214 cluster^{30, 31}, (many of) which are known to be upregulated in heart failure. A direct role for these miRs could reveal upstream pathways of FTO regulation in heart failure. Both transcriptional and post-transcriptional suppression of FTO as well as hypoxia-induced reduction in FTO activity support our in vitro and in vivo observations of increase in m6A upon FTO silencing, in hypoxia and in heart failure.

By modulating FTO expression through silencing or overexpression in isolated primary cardiac myocytes, we have demonstrated that FTO is a key contributor of global m6A levels. Further, FTO expression inversely correlates with m6A and FTO-dependent m6A is a novel and positive regulator of cardiomyocyte and cardiac contractile function. In this line, a recent study investigated FTO in myoblast differentiation and found that FTO depletion interfered with myogenic differentiation highlighting that FTO is required for myogenesis ²⁸. These data implicate FTO as an important regulator of muscle physiology.

As m6A demethylation in a single stranded nuclear RNA is the only known primary function of FTO, we attribute the effects of FTO alteration directly to changes in m6A in the target transcripts. Nevertheless, we do not rule out FTO-dependent N⁶,2’-O-dimethyladenosine (m6A_m) demethylation ³², long noncoding RNA demethylation or other indirect effects resulting from FTO (m6A)-regulated transcriptional co-regulatory networks or miRNA expression. Moreover, in addition to FTO downregulation, we detected elevated levels of writer proteins such as METTL4/14 in human and Mettl14 in mouse failing hearts, suggesting that these writers possibly also contribute to the increased m6A in failing hearts. Whether these writers compete with FTO to target similar subsets and locations of transcripts or function in a mutually exclusive manner needs to be investigated. Interestingly, the increase in transcriptome-wide m6A was significant only at one-week post-MI although FTO expression was downregulated as early as four hours post-MI. This discrepancy in transcriptome-wide m6A increase could be explained by the upregulation of another closely relatedm6A demethylase, ALKBH5, at four hours and one-day post-MI. An increase in ALKBH5 could offset the increased m6A resulting from loss of FTO at both four hours and one-day time points. This claim was corroborated by the observation that ALKBH5 protein expression returned to normal levels comparable to control hearts at one week and four-week post-MI, effectively increasing global m6A levels at those time points- primarily resulting from loss of FTO and partially from the transient increase in m6A writers. As m6A is dynamically regulated, the interplay between writer, eraser and reader proteins could be important regulating the protein expression during cardiac remodeling and regeneration, and therefore, needs to be determined.

Transcriptome-wide m6A profiling of failing mouse hearts indicated hypermethylation of several transcripts encoding cardiac contractile proteins. Previous studies have shown that hypermethylated transcripts are less stable and negatively regulate translation ^{6, 33, 34}. Consistently, we detected hypermethylation in contractile transcripts including MYH6, RYR2, SERCA2a etc., whose expressions are known to be significantly decreased in human and mouse failing hearts. This finding corroborated our in vitro data with FTO depletion or hypoxic treatment increased the m6A that resulted in aberrant cardiomyocyte calcium handling and contractile function. On the other hand, forced expression of FTO in mouse failing hearts resulted in demethylation of contractile transcripts (SERCA2a), bringing their m6A to near normal level. This resulted in significant improvement in contractile protein (SERCA2a) expression and cardiac function as demonstrated by improved ejection fraction and fractional shortening. Our observation of positive regulation of contractile transcripts and proteins with FTO-dependent demethylation are consistent with previous observations, where m6A demethylation by FTO has been shown to promote mRNA and protein expression including expression of myosin heavy chain proteins ^{10, 28, 29}.

In addition to positively regulating contractile protein expression, Fto-dependent demethylation could also regulate non-contractile protein expression in cardiomyocytes, in line with a recent study reporting that FTO-dependent m6A demethylation negatively regulated mRNA and protein expression in AML cancer cells ⁹. In our study, we did not find any significant effect of hypoxia on FTO expression and on global m6A levels in non-myocytes in vitro. However, we cannot not rule out altered m6A sites within individual transcripts under hypoxia that may have functional roles. Interestingly, our MeRIP sequencing analysis revealed that in addition to contractile pathways, FTO-overexpression regulates important non-contractile pathways as well, including tissue morphogenesis, angiogenesis, extracellular matrix organization, fibrosis, and cell growth in murine MI hearts. These MeRIP data was supported by experimental observations of enhanced angiogenesis as well as reduced scar size in the FTO-overexpressed murine MI hearts. Collectively, these data suggest that FTO-dependent m6A regulates both contractile and non-contractile pathways in murine myocardium. Whether these non-contractile phenotypes are direct functional effects of m6A on non-contractile transcripts and pathways or indirect effects of m6A-dependent transcripts and pathways needs further investigation.

Our analysis of hypermethylated transcripts in failing hearts revealed pathways involving heart rate and muscle contraction, which have decreased expression at the mRNA and proteins level in failing hearts. Interestingly, our MeRIP analysis also detected several transcripts whose m6A levels were not affected by myocardial infarction surgery or by FTO overexpression, whereas, their protein expressions are known to be differentially regulated in failing mouse and human hearts. This observation possibly indicates site-specific demethylation of the transcript that is not reflected in the overall m6A quantification, downstream regulation of mRNA transport and processing by FTO or additional unknown mechanisms of FTO function ³⁵. Increased or decreased expression of proteins as a result of FTO-dependent mRNA demethylation could be regulated by mRNA stability ^{19, 24}, degradation ⁶, and the rate of translation ²⁶. In fact, m6A has been shown to have both stimulatory ^{24, 26} and inhibitory ^{19, 36} effects on the translation dynamics. The precise effect of m6A on translation may depend on the specific 5’ and 3’ location of m6A within the transcript ^{19, 24, 37}. The site-specific m6A maps of transcripts, and its role in translation warrants further investigation.

m6A has been shown to regulate stem cell renewal ³⁸, however a direct evidence for m6A in tissue regeneration and repair has not been established. Here, we identified a novel role for m6A and myocardial m6A demethylase, FTO, in regulating cardiac tissue repair and contractile function after myocardial ischemia. In addition to the sustained expression of FTO using AAV9 vector, we detected significant beneficial effects of transiently expressed FTO delivered post-ischemia using adFTO. This indicates that FTO is therapeutic and FTO-mediated m6A demethylation can induce homeostatic protein expression resulting in improved cardiac function. It is notable that Fto levels were downregulated as early as 4 hours post-MI in mouse. Adenovirus mediated gene delivery, which maximizes FTO expression in 2 days and provides stable FTO expression for about a week, could improve cardiac homeostasis and function following the initial remodeling phase induced by MI. This transient FTO overexpression following initial remodeling phase as well as sustained FTO overexpression using AAV9 suggest that FTO-mediated benefits could be a result of both protective and reparative mechanisms of FTO. However, further studies are warranted to specifically address the precise mechanisms of FTO-mediated beneficial action. This data strongly suggests a potential therapeutic application of FTO in the treatment of post-ischemic heart failure. Fto or FTO mimics can be an interesting next generation therapeutics that can target and simultaneously improve the expression of several key contractile proteins in the heart.

Furthermore, there is an ongoing debate for a role of FTO in obesity and body weight with several studies reporting either positive or negative association for FTO with body mass ^{39, 40}. We did not observe significant differences in body weights between the control and FTO-overexpressing mice. While we show cardioprotective function of FTO, a better understanding of effect of FTO in body mass is required to take advantage of its clinical potential for the treatment of human heart failure. Together, our data provides proof of principle that given the functional importance of FTO in cardiac homeostasis and myocardial repair, targeting FTO signaling may represent a promising therapeutic strategy to treat heart failure. Our findings on the dynamic nature of the cardiac m6A-epitranscriptome will lead to deeper understanding of the mechanism of cardiac remodeling on one hand and innovative therapeutic interventions on the other.

Clinical Perspective.

What Is New?

• We discovered that m6A, one of the most prevalent and functionally relevant RNA modifications, is increased in failing mammalian hearts and in hypoxic cardiomyocytes.

• Our research suggests that dysregulation of m6A is an important hallmark of mammalian heart failure.

• Expression of an m6A demethylase, FTO (Fat-mass and obesity associated protein) decreases in failing hearts leading to aberrant increase in transcriptome-wide m6A and decreasing cardiomyocyte contractile function.

• Improving expression of FTO in ischemic mouse hearts attenuates ischemia-induced increase in m6A and decrease in cardiac contractile function. FTO selectively demethylates cardiac contractile transcripts, thus preventing their degradation and improving their protein expression under ischemia.

What Are The Clinical Implications?

• Our study demonstrates the functional importance of FTO-dependent cardiac m6A methylome in cardiac contraction during heart failure and provides a novel mechanistic insight into the therapeutic mechanisms of FTO.

• Our data provides proof of principle that FTO plays important functional role in cardiac homeostasis and myocardial repair and that targeting FTO signaling may represent a promising therapeutic strategy to treat heart failure.

• Our findings on the dynamic nature of the cardiac m6A-epitranscriptome lead to deeper understanding of the mechanism of cardiac remodeling on one hand and innovative therapeutic interventions on the other.