Skip main navigation

Metastable Atrial State Underlies the Primary Genetic Substrate for MYL4 Mutation-Associated Atrial Fibrillation

Originally published 2020;141:301–312



Atrial fibrillation (AF) is the most common clinical arrhythmia and is associated with heart failure, stroke, and increased mortality. The myocardial substrate for AF is poorly understood because of limited access to primary human tissue and mechanistic questions around existing in vitro or in vivo models.


Using an MYH6:mCherry knock-in reporter line, we developed a protocol to generate and highly purify human pluripotent stem cell–derived cardiomyocytes displaying physiological and molecular characteristics of atrial cells. We modeled human MYL4 mutants, one of the few definitive genetic causes of AF. To explore non–cell-autonomous components of AF substrate, we also created a zebrafish Myl4 knockout model, which exhibited molecular, cellular, and physiologic abnormalities that parallel those in humans bearing the cognate mutations.


There was evidence of increased retinoic acid signaling in both human embryonic stem cells and zebrafish mutant models, as well as abnormal expression and localization of cytoskeletal proteins, and loss of intracellular nicotinamide adenine dinucleotide and nicotinamide adenine dinucleotide + hydrogen. To identify potentially druggable proximate mechanisms, we performed a chemical suppressor screen integrating multiple human cellular and zebrafish in vivo endpoints. This screen identified Cx43 (connexin 43) hemichannel blockade as a robust suppressor of the abnormal phenotypes in both models of MYL4 (myosin light chain 4)–related atrial cardiomyopathy. Immunofluorescence and coimmunoprecipitation studies revealed an interaction between MYL4 and Cx43 with altered localization of Cx43 hemichannels to the lateral membrane in MYL4 mutants, as well as in atrial biopsies from unselected forms of human AF. The membrane fraction from MYL4-/- human embryonic stem cell derived atrial cells demonstrated increased phospho-Cx43, which was further accentuated by retinoic acid treatment and by the presence of risk alleles at the Pitx2 locus. PKC (protein kinase C) was induced by retinoic acid, and PKC inhibition also rescued the abnormal phenotypes in the atrial cardiomyopathy models.


These data establish a mechanistic link between the transcriptional, metabolic and electrical pathways previously implicated in AF substrate and suggest novel avenues for the prevention or therapy of this common arrhythmia.

Clinical Perspective

What Is New?

  • A rare subset of atrial fibrillation as a result of mutations in MYL4 (myosin light chain 4) appears to be dependent on pathways previously implicated in cardiac development.

  • There are interactions between these developmental pathways and the genes predisposing to more common forms of atrial fibrillation vulnerability.

  • The biology identified highlights a fundamental link between atrial patterning, excitability, metabolism, and mechanotransduction integrating multiple triggers for atrial fibrillation.

What Are the Clinical Implications?

  • Multifaceted atrial cardiomyopathies underlie many discrete forms of atrial fibrillation.

  • The identification of molecular targets which discriminate between normal and abnormal atrial substrate offers the potential for the discovery of novel antiarrhythmics.


Atrial fibrillation (AF) is the most common cardiac arrhythmia and is an independent risk factor for the development of heart failure, stroke, and overall cardiovascular mortality.1 AF management typically includes rate control, suppression of episodes of atrial arrhythmia, and anticoagulation to reduce the incidence of stroke and other thromboembolic events observed in a subset of AF cases.2 Many drugs have been developed to suppress AF, but few exhibit significant advantage over placebo controls, and the majority have narrow therapeutic windows or significant side effects.3 Increased excitability of pulmonary venous smooth muscle cells is known to be a driver in a subset of AF patients and has led to the development of methods to electrically isolate the pulmonary veins from the left atrium.4 The thromboembolic complications of AF have been attributed to rheological stasis in the setting of loss of atrial contractility, but the risk of stroke is heterogenous, and there is also an increased baseline risk for hemorrhagic events, which suggests more complex mechanisms.

The limited success of pharmacological approaches to AF is at least in part a consequence of the inaccessibility of AF mechanisms, as human left atrial tissue is rarely available before the onset of the arrhythmia or data are confounded by chronic left atrial abnormalities.5,6 The most common animal models exploit chronic rapid atrial pacing, typically in goats or sheep, to mimic and then induce AF, which persists only when pacing is continued.7 These studies have demonstrated that rapid atrial rates predispose to subsequent AF, they have reproduced the molecular findings in human tissue collected after the onset of AF and they have rigorously reinforced the clinical concept that “AF begets AF.” However, these models do not reliably predict human pharmacological responses and also do not offer access to the earliest stages of AF vulnerability, before any arrhythmia.8

Genetic approaches to the mechanisms of AF have been slowed by reduced penetrance, but there is a substantial recurrence risk within families, suggestive of significant large effect size alleles.9 Genomewide association studies have identified multiple loci, although the mechanisms remain unknown and effect sizes mitigate against efficient drug discovery.8 Rare Mendelian forms of AF exist but are often confounded by ventricular dysfunction with secondary effects on the atrial phenotype.10 Many AF loci overlap with known loci for nonischemic cardiomyopathy and AF is a prominent feature of many of these inherited forms of heart failure, sometimes before overt ventricular disease.11 Heritable lone AF (without any echocardiographic evidence of structural abnormality or ventricular dysfunction) has been described in multiple unusual kindreds, including a variant of the long QT syndrome with specific gain of function mutations in KCNQ1 and a congenital form of AF associated with childhood sudden death and mutations in the nuclear membrane pore protein NUP155 (nucleoporin 155).12 Recently, genetic studies in Icelandic populations led to the discovery of an inherited form of more typical AF caused by mutations in the MYL4 (myosin light chain 4) gene.13 This form of AF is characterized by autosomal dominant or codominant inheritance, early onset of AF, subsequent sick sinus syndrome, stroke, and the absence of ventricular dysfunction in atrial cardiomyopathy patients harboring discrete MYL4 frameshift mutations.14

Given the overlap in phenotypes between MYL4-associated AF and a substantial subpopulation within typical AF cohorts, we explored the mechanism of this form of the arrhythmia and its links to more common forms of AF. We established and optimized a protocol for deriving human atrial lineages from human pluripotent stem cells and characterized the resultant lineages at single cell level. A truncation mutant in MYL4 that recreates the human mutant peptide was generated using Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) cas9 in human embryonic stem cell (hESC)–derived atrial cells to enable the assessment of cell-autonomous components of the pathogenesis of this form of AF. To assess non–cell-autonomous components of MYL4-associated AF substrate, we created the cognate germline mutations in a zebrafish model. Together, these human cellular and animal models were chosen to undertake chemical or genetic modifier screens, which were performed in parallel to increase the specificity of the findings. Initial characterization of these models revealed the emergence of a discrete population of retinoic acid (RA) synthesizing cells, as well as distinctive cytoskeletal and metabolic defects. In addition, we were able to establish a genetic interaction between MYL4 effects and the most common risk allele for AF at the PITX2 (paired like homeodomain 2) locus. An exploratory chemical suppressor screen identified blockade of the Cx43 (connexin 43) hemichannel as a robust mechanism for rescue of the core phenotypes in both cellular and zebrafish models, and this was subsequently validated using multiple independent approaches. Last, we were also able to identify perturbations of PKC (protein kinase C) signaling, which, when reversed, could rescue cellular and cardiac abnormalities. These data define subtle perturbations of cardiomyocyte differentiation and coupling, via classical epithelial-to-mesenchymal transition pathways, as a mechanistic link between the transcriptional, metabolic, and electrical pathways previously implicated in AF. This work also identifies novel targets for the suppression of components of this complex multifaceted disease.


All materials, datasets, and protocols used in this study will be made available to other researchers upon request for purposes of reproducing the results or replicating the procedure. Zebrafish husbandry and procedures performed in this study have been approved by the Harvard Medical Area Standing Committee on Animals, the local Institutional Animal Care and Use Committee.

Cell Culture

hESC line H9 and the derivative lines were grown on Matrigel-coated plates and maintained in mTeSR medium (Stem Cell Technologies).

Atrial Lineage Induction

Differentiation was initiated 72 hours after plating when the culture was approximately 80% confluent. Cells were first differentiated toward a mesoderm phenotype subsequent to treatment with 1.5 μM CHIR 99021 (StemRD), 20 ng/mL BMP4, and 20 ng/mL Activin A in RPMI (Cellgro) supplemented with B27 minus insulin, 2 mM GlutaMAX, 1x nonessential amino acid, and 1x Normocin (InvivoGen) for 3 days. Cells were caudalized toward an atrial phenotype by treatment with 1 μM RA (Sigma Aldrich) from days 3 to 6 in RPMI (Cellgro) supplemented with B27 minus insulin, 2 mM GlutaMAX, 1x nonessential amino acid, and 1x Normocin (InvivoGen). From days 4 to 6, 5 µM XAV939 was added. From day 6 onward, differentiation of cells was carried out in RPMI supplemented with B27, 2 mM GlutaMAX, 1x nonessential amino acid, and 1x Normocin.

Chemical Screening Platform

A library consisting of U.S. Food and Drug Administration–approved drugs (Prestwick, Harvard Medical School, ICCB-Longwood facility) and ion channel library (Selleckem) was used for high throughput screening.

Zebrafish Screening

We used a previously published approach to automated physiologic screening in the zebrafish exploiting an ”off the shelf” automated microscopy platform (DiscoveryOne) and in house scripts to obtain and analyze high-speed videos from each well in a multiwell plate. Zebrafish were lysed in luciferase buffer afterwards and light intensity associated with NPPB (natriuretic peptide B) expression was quantified for each well. For every criterion a binary score was given to each compound based on its positive or negative effect on the fish. None of the dimethylsulfoxide-treated mutant fish were able to attain a score of 3 or 4, while majority of the wild type (WT) controls score 4. Compounds scoring 4 out of 4 on the initial zebrafish screening were selected for further testing.

Sulfonylureas, angiotensin receptor blockers, angiotensin-converting enzyme inhibitors, β-blockers, alcohol dehydrogenase inhibitors, and calcium channel blockers were among the initial screen hits, each of which has been found to be partially effective in AF management or targets a molecule already implicated in AF pathophysiology.15–20 Out of 35, 4 normalized all 4 endpoints (2 amlodipine derivatives, praziquantel, and carbenoxolone), while 20 compounds normalized only 3 of 4 endpoints in subsequent cherry picking experiment. Again, only compounds scoring 4 of 4 in cherry picking were picked for additional confirmation (Table I in the online-only Data Supplement).

Cell-Based Screening

D25 hESC-atrial cells derived from MYL4-/- were replated into 384 well plates. After 5 days, cells were treated with the same library of chemicals that was used for zebrafish assay for a total of 7 days. Cells were fixed and labeled with an antibody that targeted RDH10 (retinol dehydrogenase 10) 14 days after starting the treatment. Using a high-content imaging system (Harvard Medical School, ICCB-Longwood screening facility), the percentage of cells positive for RDH10 was calculated for each treatment group. About 20% of vehicle treated D45 MYL4-/- hESC-atrial cells were positive for RDH10, whereas only 4% of the D45 MYL4+/+ hESC-atrial cells were positive for RDH10. Compounds that resulted in RDH10 percentages fell between 1 SD above/below WT and mutant lines, respectively, and were selected to narrow-down the zebrafish screening hits that were reconfirmed in the cherry picking process (Table I in the online-only Data Supplement).

Statistical Analyses

A minimum of 3 biological replicates were performed for each experiment. Error bars in figures represent standard error of the mean and 2-tailed Student t test was performed for analyses unless otherwise indicated in the Methods section or figure legends. For experiments involving more than 2 treatment groups, 1-way ANOVA with Fisher post hoc analysis was used to estimate the significance of the differences.

Refer to Methods and Tables (Tables II and III in the online-only Data Supplement) for more detailed descriptions of procedures.


Single Cell Characterization of Human Atrial hESCs

To assess the differentiation efficiency and facilitate prospective purification of atrial cardiomyocytes, a MYH6:mCherry SHOX2:GFP knock-in reporter line was created in the parental H9 hESC background using CRISPR/Cas9-based gene-targeting techniques. The mCherry reporter signal colocalized with cardiac structural proteins, including cTnT (cardiac troponin T) and cardiac actin (Figure IA and IB in the online-only Data Supplement), and enabled the development of protocols to optimize the generation of human embryonic stem cell–derived atrial cells (hESC-atrial cells). On the basis of previous studies,21–23 we used a stepwise differentiation protocol, with initial mesoderm induction followed by direction of the identity of cardiac progenitors toward a more caudal phenotype (Figure 1A). To assess the phenotypic fidelity of the resultant hESC-atrial cells, we sorted mCherry+ cells and measured action potentials in current clamp mode, revealing features consistent with mature atrial electrophysiology (Figure 1B; Figure IC in the online-only Data Supplement). At day 60, hESC-derived cells stained positive by immunofluorescence for MLC2a (myosin regulatory light chain 2, atrial isoform), cTnT, cardiac actin, and MYL4, and negative for MLC2v (myosin regulatory light chain 2, ventricular isoform) (Figure 1C). No GFP+ (green fluorescent protein positive) cells were detected in the fluorescence-activated cell sorting analysis, excluding a nodal identity for these cells (Figure ID in the online-only Data Supplement).

Figure 1.

Figure 1. Generation of human atrial cells from pluripotent stem cells. (A) Schematic representation of the differentiation protocol. (B) Representative image of averaged action potentials recorded from an individual hESC-atrial cell. Amplitude 101 mV, Resting potential –81mV and liquid junction potential 11.8mV. (C) Fluorescence images of the day 60 atrial cells derived from the MYH6:mCherry H9 line using the corresponding differentiation protocol. Scale bar=75 µM (MLC2a), 100 µM (cTnT, MYL4 and cardiac actin), and 200 µM (MLC2v). (D) fluorescence-activated cell sorting plot of the day 60 hES-atrial cells sorting based on mCherry expression. (E) t-SNE plot of single-cell cluster in hES-atrial cells. (F) Heat map showing key atrial lineage and (G) differentially expressed markers in top 3 clusters. Heat map showing key atrial lineage and (G) differentially expressed markers in top 3 clusters of hES-atrial population. AA indicates activin A; B, bone morphogenetic protein 4; Ch, CHIR 99021; cTnT, cardiac muscle troponin T; DAPI, 4′,6-diamidino-2-phenylindole; hES, human embryonic stem cells; MLC2v, MYL4, myosin light chain 4; MLC2v, myosin regulatory light chain 2, ventricular isoform; RA, retinoic acid; t-SNE, t-distributed stochastic neighbor embedding; and Wi, XAV (Wnt inhibitor).

We performed single-cell RNA-seq analysis on differentiated hESC-atrial cells. To minimize batch-to-batch variation, we sorted the cells according to mCherry expression and mixed positive and negative fractions in fixed ratios before barcoding to minimize intersample variation (Figure 1D; Figure IE in the online-only Data Supplement). The MYH6+ population also expressed atrial specific markers including NPPA, MYL4, MYL7 (MLC2a) while the entire atrial differentiation population (including the MYH6+ [myosin heavy chain 6 positive cells] majority) lacked ventricular, conduction system, or epicardial markers such as MYL2, NPPB, CNTN1, IRX4, HEY2, ISL1, TBX18, and MYH11 (Figure 1E through 1G; Figure IF and IG in the online-only Data Supplement). The top 20 transcripts in the MYH6+ population included discrete markers proposed for human atrial cell annotation such as NR2F224,25 (Figure IF in the online-only Data Supplement), among which MYH6 exhibited the most consistent overlap with other known atrial-specific genes (such as MYL4 and NPPA [natriuretic peptide A]) and therefore was chosen for lineage marking.

MYL4 Mutant Cell Lines Reveal Increased RA Synthesis and Actin Disorganization

We generated disease-associated mutations in MYL4 using CRISPR-Cas9 technology in hESCs (Figure 2A; Figure IIA and IIB in the online-only Data Supplement) to study this inherited form of AF substrate. Single-cell RNA sequencing (scRNA-seq) on MYL4+/- and MYL4-/- derived hESC-atrial cells at day 60 was compared with that from MYL4+/+ controls (Figure 2C). Cells derived from both WT and mutant lines included populations expressing atrial lineage markers (Figure IIIA in the online-only Data Supplement). However, the transcriptional data also identified a discrete population of cells in mutant lines significantly enriched in RA synthesis and RA response pathway members, including ALDH1A2, RDH10, EDN3, and KLK5 (Figure 2D and 2E; Figure IIIB through IIID in the online-only Data Supplement). This transcriptional phenotype was further amplified in the context of the known common PITX2 AF risk allele rs220073326,27 (Figure 2E; Figure IIIE in the online-only Data Supplement), which has been postulated as a general modifier of AF substrates.28 Fluorescence-activated cell sorting analysis revealed increased percentages of cells expressing ALDH1A2 (aldehyde dehydrogenase 1 family member A2) and RDH10 in the MYL4-/- compared with MYL4+/+ with additional increases observed in MYL4-/- bearing the rs2200733 allele. The changes observed at cellular level in the hESC-atrial model were also observed in Myl4 knock-out zebrafish lines generated on a previously published NPPB:Luc background (Figure 2F; Figure IIIF in the online-only Data Supplement). Specifically, immunofluorescence revealed increases in ALDH1A2 expression at 5dpf and RNA-seq, and ATACseq of 3dpf hearts further confirmed in the zebrafish model the activation of RA synthesis genes observed in vitro (Figure 2G and 2H; Figure IIIG in the online-only Data Supplement).

Figure 2.

Figure 2. Single cell RNA-seq reveals a distinct cell population in MYL4 mutants. (A) Schematic and (B) IF analysis for MYL4 in hES-atrial cells showing missing or decreased expression of MYL4 in the mutant clone. Scale bar =150 µM. (C) t-SNE plot of single-cell clusters in combined WT and mutant populations color coded for discrete clusters (left) and experimental condition (right; KO shown in red vs. WT shown in blue). The mutant unique cluster (#0) is circled in red. (D) Violin plots of top differentially expressed genes between WT and mutant lines in cluster #0. (E) Intracellular fluorescence-activated cell sorting staining of cardiac structural markers (Cx43 and alpha actinin) with retinoic acid synthesizing enzymes (ALDH1A2 and RDH10) in day 60 hES-atrial cells from WT, MYL4-/- and MYL4-/- harboring SNP rs2200733 C-T. (F) Schematic illustration of generation of Myl4 mutant zebrafish lines on Nppb:Luc background (G) Immunofluorescent co-staining of ALDH1A2 and S46 (atrial myosin heavy chain) in WT and Myl4-/- 5dpf zebrafish hearts. Magnification:30x (H) Sequencing tracks of open chromatin accessibility for the Rdh10b in 3dpf zebrafish hearts shows distinct ATAC-seq peaks at the promoter region, more frequently in Myl4-/- fish. (I) t-SNE overlay of selected atrial lineage makers expressed in cardiomyocyte fraction. (J) Violin plots of selected top differentially expressed genes between WT and mutant lines in cardiomyocyte population. (K) Fluorescence actin polymerization assay in the presence of either buffer, MYL4 recombinant protein or ARP2/3+VCA. (L) Schematic diagram showing patients enrolled in this study for obtaining tissue biopsy that included individuals with normal sinus rhythm, atrial cardiomyopathy defined as PR interval >120 ms and permanent AF. (M) Immunofluorescence staining and (N) quantification of biopsy samples for Phalloidin (marks F-actin; grey), and MYL4 (red; merged staining shown here; magnification 40×). A higher magnification is shown in the right column. *P<0.05 aCMP indicates atrial cardiomyopathy; AF, atrial fibrillation; ALDH1A2, aldehyde dehydrogenase 1 family member A2; Cx43, connexin 43; IF, immunofluorescence; hES, human embryonic stem cells; KO, knockout; MLC2v, myosin regulatory light chain 2, ventricular isoform; MYL4, myosin light chain 4; RDH10, retinol dehydrogenase 10; SNP, single nucleotide polymorphisms; SR, sinus rhythm; t-SNE, t-distributed stochastic neighbor embedding; VCA, verprolin, cofilin, acidic; and WT, wild type.

Next, we focused on the differentially expressed genes between WT and mutant lines within the MYH6+ hESC-atrial cells. The mutant cells expressed key atrial lineage markers (Figure 2I; Figure IV in the online-only Data Supplement). However, a number of genes associated with familial forms of cardiomyopathies that are linked with actin cytoskeleton remodeling or organization were differentially expressed (Figure 2J; Figure IV in the online-only Data Supplement). We confirmed the MYL4-actin interaction by demonstrating that recombinant MYL4 protein could activate pyrene actin polymerization as efficiently as 2 known potent activators, Arp2/3 and Verprolin, Cofilin, Acidic (Figure 2K). We tested MYL4 and cytoskeletal actin distribution in heart biopsies from individuals in normal sinus rhythm, demonstrating significant colocalization. This colocalization was perturbed, with evidence of increased membrane localization of MYL4, in left atrial biopsies from those undergoing cardiac surgery who developed AF postoperatively and from those with known AF, but not in control subjects who did not develop AF (Figure 2L through 2N; Figure V in the online-only Data Supplement).

Chemical Screening to Identify Proximate Pathway Modifiers

There are numerous potential pathways by which MYL4 mutation might predispose to AF through alterations in RA signaling, cytoskeletal remodeling, or some other unknown means, so to prioritize subsequent investigation, we performed a chemical suppressor screen with a library of known bioactive molecules, using a series of physiologic, cellular, and molecular endpoints in both hESC-atrial mutant cell lines and zebrafish Myl4 mutant models. Our goal was to identify accessible proximate pathways in the complex pathophysiologic cascade.

Homozygous Myl4 mutant fish exhibited significantly lower heart rates and lower Nppb::luciferase levels (as well as consistent ATAC-seq; Figure 3A; Figure VIA in the online-only Data Supplement), with conserved ventricular function consistent with clinical data from Myl4 mutant kindreds. Physiologic screening endpoints in fish included heart rate, contractility, blood flow and NPPB::luciferase levels (Figure IIIA and VIB in the online-only Data Supplement), as these were representative of the human disease and appeared to reproducibly discriminate between mutant and WT animals. In hESC-atrial mutant cells the endpoint used was the percentage of RDH10+ cells (Figure 3B). While there were multiple partial hits in each screen, parallel unbiased screening in duplicate using the same chemical libraries in human mutant hESC-atrial cells and zebrafish identified only one shared high fidelity hit: carbenoxelone (Figure 2C through 2E; Table I and Figure VIB through VID in the online-only Data Supplement).

Figure 3.

Figure 3. A high throughput combined in vitro and in vivo screening platform to identify disease modifiers. (A) Cardiac function quantification of myl4 mutant zebrafish lines on Nppb: luc background (left) and the schematic illustration of in vivo screening platform design (right). (B) %RDH10 in day 45 hES-atrial cells derived from wild type vs. mutant lines (right) and schematic representation of the cell based assay (right). Results of primary screening of 1280 U.S. Food and Drug Administration–approved chemicals for (C) zebrafish heart rate, contractility, tail blood flow (56 hpf), luciferase level (72 hpf) and (D) %RDH10 in Day 60 hES-atrial cells. Red boxes show the compounds that could met the individual criteria, but only the one that fulfilled all the criteria was selected for further confirmation. (E) Chemical structure of Carbenoxolone, a connexin 43 hemichannel blocker. *P<0.05, **P<0.01, *** P<0.001, ****P<0.0001; t test. Data are mean ± SEM. BF indicates blood flow; Cont, contractility; hES, human embryonic stem cell; HR, heart rate; Luc, luciferase; HT CV, high throughput cardiovascular; IF, immunofluorescence; and RDH10 Ab, retinol dehydrogenase 10 antibody.

While the molecular targets of carbenoxolone are not fully understood, there are well-described effects blocking connexin hemichannels (HCs) at high affinity.29,30 Given these findings, we explored a potential primary role for HC physiology in the atrial cardiomyopathy caused by MYL4. Both immunofluorescence and electron microscopy revealed evidence of abnormal trafficking of Cx43 in human and zebrafish MYL4 mutants (Figure 4A and 4B; Figure VIIA in the online-only Data Supplement). Furthermore, biochemical pull-down confirmed that MYL4 is associated with Cx43 and actin in both WT and mutant D60 hESC-atrial cells (Figure 4C, full blots in Figure VIIB in the online-only Data Supplement). Both actin filaments and microtubules are known to be involved in Cx43 trafficking in cardiomyocytes.31 Cx43 localization abnormalities were identified in the preoperative atrial biopsies from patients who developed postoperative AF and from those with permanent AF but not in control subjects with sinus rhythm. Immunostaining of the human biopsies for Cx43 and actin, revealed relocalization of Cx43 to the lateral membrane and reduced colocalization with actin filaments. (Figure 4D and 4E; Figure VIIC in the online-only Data Supplement).

Figure 4.

Figure 4. Connexin 43 HC blocker prevents membrane leak and reverses electrophysiological changes in atrial fibrillation. (A) Immunofluorescent costaining of Cx43 with MYL4 in D60 hES atrial cells derived from WT and mutant lines. Scale bar=50 µM. (B) Immunofluorescent costaining of Cx43 with S46 in 5dpf Myl4+/+ and Myl4+/- fish lines. Magnification=30×. (C) Immunoglobulin pulldown of MYL4 in D60 MYL4+/+ and MYL4-/- hES-atrial cells and Western blot analysis of MYL4, Cx43, and actin. Dashed arrow denotes the band corresponding to the MYL4 protein. (D) Immunofluorescence images and (E) quantification of Cx43 (green) colocalization with intracellular actin (red; higher magnification in the right column) in biopsy samples from individuals with normal SR, CMP, and AF. (F) Polar metabolite abundance analysis from D60 WT and mutant hES-atrial cells as determined by liquid chromatography–mass spectrometry. (G, H) Relative NAD level in MYL4-/- D60 hES-atrial cells treated with vehicle, carb (5 uM), or GAP19 (5 uM). (I) Absolute action potential voltage amplitude and dv/dt in D60 hES-atrial cells derived from MYL4+/+, MYL4-/-, and MYL4-/- after 3 days of treatment with carb (5 uM) in a current clamp setting. (J) Averaged action potential duration and (K) calcium amplitude recordings from atrium of 3 dpf WT or mutant fish (with or without 2-day drug pretreatment). Significance was assessed using 1-way ANOVA with Fisher post hoc tests. *P<0.05, **P<0.01, ***P<0.001. Data are mean ± SEM. AF indicates atrial fibrillation; Carb, carbenoxolone; CMP, atrial cardiomyopathy; Co-IP, co-immunoprecipitation; Cx43, connexin 43; dpf, days past fertilization; G19, GAP19; G26, GAP26; hES, human embryonic stem cell; MYL4, myosin light chain 4; NAD, nicotinamide adenine dinucleotide; SR, sinus rhythm; WT, wild type.

Cx43 HC Blockade Reverses Metabolic and Electrical Changes in Mutant Lines

Cx43 HCs are present in nonjunctional membrane and play a role in transmembrane signaling that is dependent on their phosphorylation status.32,33 Conducting HCs allow ions, as well as metabolites, to transit, though gating mechanisms must exist to prevent the dissipation of transmembrane electrical gradients. We compared the cellular metabolite profiles of D60 hESC-atrial cells derived from MYL4+/+ or MYL4-/- lines and observed significant differences in the levels of multiple metabolites, including nicotinamide adenine dinucleotide (NAD) and nicotinamide adenine dinucleotide + hydrogen (Figure 4F; Figure VIIIA in the online-only Data Supplement). There was no difference in the level of nicotinamide riboside, the precursor of NAD in these cells, suggesting that the differences observed are not a consequence of precursor bioavailability. Quantitative NAD measurement revealed higher levels of NAD in rescued mutant atrial cells than in untreated mutant controls (Figure 4G). The specific Cx43 HC blocker Gap19, which does not exhibit gap junction blocking effects, restored intracellular NAD concentrations in mutant cells (Figure 4H).

Tight control of Cx43 HC permeability is necessary for normal electrophysiologic function.34 Current-clamp measurement of action potentials from hESC-atrial cells demonstrated significant changes in voltage amplitude and duration, both reversible with carbenoxolone treatment (Figure 4I). No changes were observed in maximum diastolic potential (Figure VIIIB in the online-only Data Supplement). We tested whether HC blockade was able to rescue whole-heart electrophysiological changes in Myl4 mutant fish. First, 3dpf zebrafish hearts were isolated and labeled with voltage sensitive dye (Fluovolt) for optical mapping (Figure 4J). MYL4-/- atria showed significantly prolonged action potential duration compared to WTs. By contrast, there was no difference observed in conduction velocity (Figure IXA in the online-only Data Supplement), which is in part determined by gap junction–dependent intercellular coupling. The effects of MYL4 mutation on action potential duration were reversible with 2 specific Cx43 HC blocking peptides, Gap19 and Gap26, which also did not affect conduction velocities (Figure IXA in the online-only Data Supplement).

Calcium handling abnormalities are among the earliest features of AF and have been associated with the AF-driven myocardial remodeling that sustains the arrhythmia.35 Isolated hearts were stained with the calcium dye Fura2, and atrial [Ca2+] transients were measured using high-speed ratiometric calcium imaging (Figure 4K). Calcium transient amplitude, but not transient duration was increased in Myl4-/- zebrafish hearts (Figure IXB in the online-only Data Supplement) and rescued by Cx43 HC blockade. The Gap26 peptide, which binds the extracellular domain of the Cx43 HC (in contrast to Gap19 which binds intracellularly), only partially rescued this component of the mutant phenotype. These data suggest the increased calcium transient amplitude associated with perpetuation of AF is also a core component of the primary substrate in at least some inherited forms of the arrhythmia.36

PKC Mediates Cx43 Hemichannel Phosphorylation in MYL4 Mutants

Cx43 HC permeability is determined by phosphorylation state,33 and bulk RNA-seq revealed significant enrichment of the transcripts for multiple protein kinase genes in D60 hESC-atrial cells derived from mutants when compared with cells derived from WT controls (Figure 5A). The Cx43 in the membrane fraction of MYL4 mutant cells was hyperphosphorylated, compared with the cytoplasmic fraction (Figure 5B). Culturing cells on surfaces with higher surface tension to model the effects of chronic stretch or the combination of MYL4 mutation with the rs2200733 C-T PITX2 AF risk allele, each resulted in further increases in phosphorylated Cx43 in the membrane fraction (Figure 5C and D). Multiple phosphorylation sites in the C-terminal cytoplasmic domain of Cx43 regulate hemi-channel permeability, regulated by multiple kinases (Figure 5E).37–39 Previously, higher CAMKII (Ca2+/Calmodulin-dependent protein kinase II) activity has been demonstrated in atrial samples from chronic AF, but not from paroxysmal AF animal models.40 CAMKII inhibition in our models exaggerated the prolongation of the abnormal action potential duration in mutant cells, while both PKC and PKA (protein kinase A) inhibition partially normalized action potential duration (Figure 5F). While neither PKA nor CAMKII inhibition were able to significantly reduce calcium amplitude in mutants, PKC inhibition resulted in the reversal of calcium transient amplitude changes without affecting transient duration (Figure 5G; Figure IXC in the online-only Data Supplement). PKC inhibition also decreased Cx43 hyperphosphorylation and rescued intracellular NAD levels in mutant hESC-atrial cells (Figure 5H and 5I). In contrast, treating mutant hESC-atrial cells with RA accentuates Cx43 phosphorylation and PKC activity (Figure 5H and 5J). Taken together, these results suggest a role for PKC activity, in this form of AF, at least in part through Cx43 modulation.

Figure 5.

Figure 5. RA inducible PKC activity leads to Cx43 hemichannel phosphorylation and subsequently electrophysiological changes in AF. (A) Gene ontology analysis of WT vs. mutant transcriptional profiling in mCherry+ cells purified from D60 hES-atrial cells. (B) Phospho-Cx43 fraction in membrane assessed by Western blot in MYL4 mutant vs. WT hES-atrial cells (C) with increased underlying scaffold stiffness and (D) mutants harboring rs2200733 C-T risk allele. (E) Schematic representation of Cx43 c-term phosphorylation sites. Average (F) action potential duration and (G) calcium transient amplitude from atrium of 3 dpf WT or mutant fish (with or without 2-day drug pretreatment) heart fluorescence after incubation with cell membrane voltage dye or calcium dye, respectively. Significance was assessed using 1-way ANOVA with Fisher post hoc test. (H) Phospho-Cx43 fraction in membrane assessed by Western blot in MYL4 mutant treated with PKC inhibitor, PKA inhibitor and RA (1uM and 0.1uM) with or without AGN193109. (I) Relative NAD level in MYL4-/- D60 hES-atrial cells treated with vehicle, PKCi (5 uM) or CAMK2i (5 uM). (J) PKC activity measured in MYL4-/- derived D60 hES-atrial cells treated with vehicle, or RA or RA+AGN193109. (K) A summary of changes at the cellular level in a subset of the AF models studied here. *P<0.05. Data are mean ± SEM. AF, indicates atrial fibrillation; CAMK2i, Ca2+/Calmodulin-dependent protein kinase II inhibitor; Carb, Carbenoxolone; Cx43, connexin 43; dpf, days past fertilization; G19, GAP19; G26, GAP26; MYL4, myosin light chain 4; NAD, nicotinamide adenine dinucleotide; PKA, protein kinase A; PKC, protein kinase C; PKCi, protein kinase C inhibitor; RA, retinoic acid; WT, wild type


AF is observed in numerous disease syndromes ranging from primary myocardial disease to sepsis, but its characterization has been complicated by limited access to the primary atrial substrate before AF itself results in secondary findings.41 To identify proximate mechanisms for AF, we modeled both the cell-autonomous and non–cell-autonomous components of a genetic form of AF substrate (before AF) in systems suited to subsequent unbiased screening. Previous work studied a zebrafish line transgenic for a mutated cmlc1 (as the putative ortholog of human MYL4), which is expressed mainly in the ventricle in the fish.42 As a result of these observations, direct homology scores and the known atrial expression pattern of myl4, we elected to model a definitive human mutation (the original Icelandic codominant variant c.234delC) in the zebrafish myl4 gene on the basis of its relevance to the human trait of interest. Peng et al. studied N-terminal mutations in MYL4, which have clearly been associated with an atrial standstill phenotype.43 However, reported AF-associated variants do not appear to share this feature, so it is possible that the N-terminus of MYL4 is involved in discrete functions.

Focusing on mutations in the MYL4 gene, one of the few identified heritable causes of AF,13 we chose to characterize the features of this specific entity using mutant hESC-derived atrial cells and mutant zebrafish lines in parallel. In an effort to anchor our findings definitively in more common human AF mechanisms, we exploited a unique set of atrial biopsies obtained from patients during cardiac surgery before the emergence of postoperative AF in a discrete subset.

We first optimized protocols to enable the characterization of atrial lineages derived from human pluripotent stem cells at the single cell level in both health and disease. By introducing mutations in MYL4 we were able to study single cell structural, electrophysiologic and transcriptional features of AF substrate. Using a cognate approach in the zebrafish, we could also explore tissue and organ level physiology during similar stages of differentiation. Together these approaches offer efficient and complementary access to the biology of this genetic form of human atrial disorder.

The initial insight from these models was the emergence of a specific population of RA synthesizing cells in the setting of mutant MYL4. While scRNA-seq can result in spurious cell populations, we mitigated this risk using a controlled pooling strategy, and also used germline zebrafish MYL4 mutants to confirm these findings in vivo. Interestingly, RA is known not only to be a necessary signal for atrial development and patterning,22 but in multiple systems, it regulates regional membrane biology and cell–cell interaction through differential phosphorylation of a range of protein kinases.44 In the atrium these changes would be predicted to cause significant changes in cell physiology dependent on cell polarity.45 Multiple molecular pathways have been implicated in the maintenance of cellular polarity, including Wnt β-catenin, TGF-β, and calcium handling, and the cytoskeleton,46–48 but these pathways have diverse effects so we used unbiased approaches to understand the mechanisms of MYL4 mutation in AF and related myocardial diseases.

Single cell RNA-seq of mutant hESC-atrial cells also identified significant perturbation in cytoskeletal gene expression and we subsequently demonstrated that MYL4 interacts with cytoskeletal actin both in vitro and in vivo. In mutant cells from hESC-atrial lines and in human atrial biopsies from those who subsequently developed AF, we observed abnormal relocalization of MYL4 to the lateral membrane with relative loss of colocalization with actin. These findings, and mislocalization of actin and MYL4, further supported abnormal polarization of the cardiomyocyte cell membrane in both the cellular model and in human disease samples.

In a systematic effort to identify proximate mechanisms in our 2 models of MYL4 mutant AF substrate, we performed unbiased screens of known bioactive small molecule libraries for compounds capable of suppressing the abnormal phenotypes. These parallel screens identified carbenoxelone as the only compound in this limited set capable of suppressing all of the phenotypes in both models. Given carbenoxelone’s previous annotation as a blocker of connexin hemichannels,29,30 we then tested this mechanism as the mode of suppression. We first characterized physiologic changes in both human cellular and zebrafish MYL4 mutants, demonstrating abnormalities in cellular electrophysiology, calcium handling, and cellular energy substrate leak. We found that not only did HC blockade explain the rescue we had observed, but also demonstrated that other abnormalities, including NAD/nicotinamide adenine dinucleotide + hydrogen leak, were also consistent with pathological increases in Cx43 HC density and conductance in the lateral membrane.32 Notably, these features were not associated with any electrical coupling abnormalities, further reinforcing the specificity of the hemichannel phenotype. We were able to confirm the presence of increased lateral membrane HC expression in human atrial tissue samples from those who were in sinus rhythm but subsequently developed AF, implicating a partial change of cellular organization as an upstream substrate for a subset of AF. We also established a significant genetic interaction between the MYL4 mutant allele and the largest effect-size common risk allele for AF at the PITX2 locus across multiple cellular outputs.

HC permeability is known to be regulated by both location and phosphorylation status. We found that among protein kinases with a potential target site on Cx43, PKC inhibition resulted in significant reduction in phosphorylation status as well inhibition of membrane permeability to nicotinamide adenine dinucleotide/nicotinamide adenine dinucleotide + hydrogen and perturbations in calcium and voltage homeostasis. RA is also able to directly increase PKC activity, a process which was reversible with RA antagonist. Thus, not only does the Cx43 HC appear to be an attractive candidate target for suppression of this form of AF substrate, but existing PKC inhibitors, already known modulators of cell coupling, may also be a viable adjunct.38

Together, these data support a graded and subtle form of cellular differentiation as the underlying mechanism for MYL4-related AF, and possibly a broader set of AF as manifest postoperatively. Indeed several observations are consistent with atrial myocardium occupying a metastable state at baseline which can be disrupted by factors which increase cardiomyocyte differentiation (exercise, stretch, RA) or decrease cardiomyocyte differentiation (MYL4 mutation, thyroid hormone, PITX2 locus, obesity). The pathways involved are prominent in epithelial mesenchymal transition during development, trabeculation, myocardial regeneration, and fibrosis, implicating directly core mechanisms integrating the patterning of cell coupling, membrane biology, mechanotransduction, and metabolism.49 This deviation from an intrinsically metastable state may reflect the balance between adaptability (eg, selected for mechanotransduction or volume sensing) and disease, and is consistent with the large number of inputs that result in this arrhythmia affecting up to 10% of all humans. Interestingly, using the Objective Prioritization for Enhanced Novelty algorithm to identify the most likely genes at genomewide association studies loci, those for the top AF genomewide association studies loci all participate in a large network which is dominated by genes involved in cell polarity or epithelial mesenchymal transition (data not shown50).

In summary, we identified perturbation of atrial myocardial cell polarity as a shared upstream mechanism integrating electrical, metabolic, and subcellular structural abnormalities in MYL4-related AF substrate (Figure 5K). Using an unbiased high-throughput screening platform that combined several physiologic and molecular endpoints, we identified the Cx43 HC as a key target that may offer an avenue for antiarrhythmic therapy that complements traditional ion channel blockade. While there is no direct evidence that the abnormalities we observed predispose to or lead to AF, these insights imply that many of the downstream features may be a consequence not just of the sustained burden of AF, but rather of the primary mechanism itself, and raise the possibility that endothelial cells, adipocytes, fibroblasts, and other cell types may each contribute through related perturbations of their basal differentiation. Additional work in a range of experimental and clinical settings will be required to test this hypothesis, but the possibility for precision interventions for some of the forms of AF is evident.


The computation in this article was run on the O2 cluster supported by the Harvard Medical School Research Computing Group.


*Dr Kiviniemi and and S.Olafsson contributed equally.

Sources of Funding, see page 311

The online-only Data Supplement is available with this article at

Zaniar Ghazizadeh, MD, Cardiovascular Medicine, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115. Email
Calum A. MacRae, MD, PhD, Cardiovascular Medicine, Brigham and Women’s Hospital, 75 Francis Street, Boston, MA 02115. Email


  • 1. Staerk L, Sherer JA, Ko D, Benjamin EJ, Helm RH. Atrial fibrillation: epidemiology, pathophysiology, and clinical outcomes.Circ Res. 2017; 120:1501–1517. doi: 10.1161/CIRCRESAHA.117.309732LinkGoogle Scholar
  • 2. January CT, Wann LS, Alpert JS, Calkins H, Cigarroa JE, Cleveland JC, Conti JB, Ellinor PT, Ezekowitz MD, Field ME, et al; ACC/AHA Task Force Members. 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines and the Heart Rhythm Society.Circulation. 2014; 130:e199–e267. doi: 10.1161/CIR.0000000000000041LinkGoogle Scholar
  • 3. Laksman Z, Wauchop M, Lin E, Protze S, Lee J, Yang W, Izaddoustdar F, Shafaattalab S, Gepstein L, Tibbits GF, et al. Modeling atrial fibrillation using human embryonic stem cell-derived atrial tissue.Sci Rep. 2017; 7:5268. doi: 10.1038/s41598-017-05652-yCrossrefMedlineGoogle Scholar
  • 4. Weerasooriya R, Khairy P, Litalien J, Macle L, Hocini M, Sacher F, Lellouche N, Knecht S, Wright M, Nault I, et al. Catheter ablation for atrial fibrillation: are results maintained at 5 years of follow-up?J Am Coll Cardiol. 2011; 57:160–166. doi: 10.1016/j.jacc.2010.05.061CrossrefMedlineGoogle Scholar
  • 5. Barichello S, Roberts JD, Backx P, Boyle PM, Laksman Z. Personalizing therapy for atrial fibrillation: the role of stem cell and in silico disease models.Cardiovasc Res. 2018; 114:931–943. doi: 10.1093/cvr/cvy090CrossrefMedlineGoogle Scholar
  • 6. Yang C, Al-Aama J, Stojkovic M, Keavney B, Trafford A, Lako M, Armstrong L. Concise review: cardiac disease modeling using induced pluripotent stem cells.Stem Cells. 2015; 33:2643–2651. doi: 10.1002/stem.2070CrossrefMedlineGoogle Scholar
  • 7. Nattel S, Dobrev D. Controversies about atrial fibrillation mechanisms: aiming for order in chaos and whether it matters.Circ Res. 2017; 120:1396–1398. doi: 10.1161/CIRCRESAHA.116.310489LinkGoogle Scholar
  • 8. Heijman J, Algalarrondo V, Voigt N, Melka J, Wehrens XH, Dobrev D, Nattel S. The value of basic research insights into atrial fibrillation mechanisms as a guide to therapeutic innovation: a critical analysis.Cardiovasc Res. 2016; 109:467–479. doi: 10.1093/cvr/cvv275CrossrefMedlineGoogle Scholar
  • 9. Bapat A, Anderson CD, Ellinor PT, Lubitz SA. Genomic basis of atrial fibrillation.Heart. 2018; 104:201–206. doi: 10.1136/heartjnl-2016-311027CrossrefMedlineGoogle Scholar
  • 10. Ellinor PT, Macrae CA. The genetics of atrial fibrillation.J Cardiovasc Electrophysiol. 2003; 14:1007–1009. doi: 10.1046/j.1540-8167.2003.03307.xCrossrefMedlineGoogle Scholar
  • 11. Aragam KG, Chaffin M, Levinson RT, McDermott G, Choi SH, Shoemaker MB, Haas ME, Weng LC, Lindsay ME, Smith JG, et al; Genetic Risk Assessment of Defibrillator Events (GRADE) Investigators. Phenotypic refinement of heart failure in a national biobank facilitates genetic discovery.Circulation. 2018; 139:489–501.LinkGoogle Scholar
  • 12. Feghaly J, Zakka P, London B, MacRae CA, Refaat MM. Genetics of atrial fibrillation.J Am Heart Assoc. 2018; 7:e009884. doi: 10.1161/JAHA.118.009884LinkGoogle Scholar
  • 13. Gudbjartsson DF, Helgason H, Gudjonsson SA, Zink F, Oddson A, Gylfason A, Besenbacher S, Magnusson G, Halldorsson BV, Hjartarson E, et al. Large-scale whole-genome sequencing of the Icelandic population.Nat Genet. 2015; 47:435–444. doi: 10.1038/ng.3247CrossrefMedlineGoogle Scholar
  • 14. Gudbjartsson DF, Holm H, Sulem P, Masson G, Oddsson A, Magnusson OT, Saemundsdottir J, Helgadottir HT, Helgason H, Johannsdottir H, et al. A frameshift deletion in the sarcomere gene MYL4 causes early-onset familial atrial fibrillation.Eur Heart J. 2017; 38:27–34. doi: 10.1093/eurheartj/ehw379CrossrefMedlineGoogle Scholar
  • 15. Saygili E, Rana OR, Saygili E, Reuter H, Frank K, Schwinger RH, Müller-Ehmsen J, Zobel C. Losartan prevents stretch-induced electrical remodeling in cultured atrial neonatal myocytes.Am J Physiol Heart Circ Physiol. 2007; 292:H2898–H2905. doi: 10.1152/ajpheart.00546.2006CrossrefMedlineGoogle Scholar
  • 16. Kim SJ, Zhang H, Khaliulin I, Choisy SC, Bond R, Lin H, El Haou S, Milnes JT, Hancox JC, Suleiman MS, et al. Activation of glibenclamide-sensitive ATP-sensitive K+ channels during β-adrenergically induced metabolic stress produces a substrate for atrial tachyarrhythmia.Circ Arrhythm Electrophysiol. 2012; 5:1184–1192. doi: 10.1161/CIRCEP.112.975425LinkGoogle Scholar
  • 17. Khatib R, Joseph P, Briel M, Yusuf S, Healey J. Blockade of the renin-angiotensin-aldosterone system (RAAS) for primary prevention of non-valvular atrial fibrillation: a systematic review and meta analysis of randomized controlled trials.Int J Cardiol. 2013; 165:17–24. doi: 10.1016/j.ijcard.2012.02.009CrossrefMedlineGoogle Scholar
  • 18. Cadrin-Tourigny J, Shohoudi A, Roy D, Talajic M, Tadros R, Mondésert B, Dyrda K, Rivard L, Andrade JG, Macle L, et al. Decreased mortality with beta-blockers in patients with heart failure and coexisting atrial fibrillation: an AF-CHF substudy.JACC Heart Fail. 2017; 5:99–106. doi: 10.1016/j.jchf.2016.10.015CrossrefMedlineGoogle Scholar
  • 19. Nakano Y, Ochi H, Onohara Y, Sairaku A, Tokuyama T, Matsumura H, Tomomori S, Amioka M, Hironomobe N, Motoda C, et al. Genetic variations of aldehyde dehydrogenase 2 and alcohol dehydrogenase 1B are associated with the etiology of atrial fibrillation in Japanese.J Biomed Sci. 2016; 23:89. doi: 10.1186/s12929-016-0304-xCrossrefMedlineGoogle Scholar
  • 20. Yan J, Thomson JK, Zhao W, Gao X, Huang F, Chen B, Liang Q, Song LS, Fill M, Ai X. Role of stress kinase JNK in binge alcohol-evoked atrial arrhythmia.J Am Coll Cardiol. 2018; 71:1459–1470. doi: 10.1016/j.jacc.2018.01.060CrossrefMedlineGoogle Scholar
  • 21. Lee JH, Protze SI, Laksman Z, Backx PH, Keller GM. Human pluripotent stem cell-derived atrial and ventricular cardiomyocytes develop from distinct mesoderm populations.Cell Stem Cell. 2017; 21:179–194.e4. doi: 10.1016/j.stem.2017.07.003CrossrefMedlineGoogle Scholar
  • 22. Devalla HD, Schwach V, Ford JW, Milnes JT, El-Haou S, Jackson C, Gkatzis K, Elliott DA, Chuva de Sousa Lopes SM, Mummery CL, et al. Atrial-like cardiomyocytes from human pluripotent stem cells are a robust preclinical model for assessing atrial-selective pharmacology.EMBO Mol Med. 2015; 7:394–410. doi: 10.15252/emmm.201404757CrossrefMedlineGoogle Scholar
  • 23. van den Berg CW, Elliott DA, Braam SR, Mummery CL, Davis RP. Differentiation of human pluripotent stem cells to cardiomyocytes under defined conditions.Methods Mol Biol. 2016; 1353:163–180. doi: 10.1007/7651_2014_178CrossrefMedlineGoogle Scholar
  • 24. Wu SP, Cheng CM, Lanz RB, Wang T, Respress JL, Ather S, Chen W, Tsai SJ, Wehrens XH, Tsai MJ, et al. Atrial identity is determined by a COUP-TFII regulatory network.Dev Cell. 2013; 25:417–426. doi: 10.1016/j.devcel.2013.04.017CrossrefMedlineGoogle Scholar
  • 25. Schwach V, Verkerk AO, Mol M, Monshouwer-Kloots JJ, Devalla HD, Orlova VV, Anastassiadis K, Mummery CL, Davis RP, Passier R. A COUP-TFII human embryonic stem cell reporter line to identify and select atrial cardiomyocytes.Stem Cell Reports. 2017; 9:1765–1779. doi: 10.1016/j.stemcr.2017.10.024CrossrefMedlineGoogle Scholar
  • 26. Lubitz SA, Sinner MF, Lunetta KL, Makino S, Pfeufer A, Rahman R, Veltman CE, Barnard J, Bis JC, Danik SP, et al. Independent susceptibility markers for atrial fibrillation on chromosome 4q25.Circulation. 2010; 122:976–984. doi: 10.1161/CIRCULATIONAHA.109.886440LinkGoogle Scholar
  • 27. Kääb S, Darbar D, van Noord C, Dupuis J, Pfeufer A, Newton-Cheh C, Schnabel R, Makino S, Sinner MF, Kannankeril PJ, et al. Large scale replication and meta-analysis of variants on chromosome 4q25 associated with atrial fibrillation.Eur Heart J. 2009; 30:813–819. doi: 10.1093/eurheartj/ehn578CrossrefMedlineGoogle Scholar
  • 28. Gudbjartsson DF, Arnar DO, Helgadottir A, Gretarsdottir S, Holm H, Sigurdsson A, Jonasdottir A, Baker A, Thorleifsson G, Kristjansson K, et al. Variants conferring risk of atrial fibrillation on chromosome 4q25.Nature. 2007; 448:353–357. doi: 10.1038/nature06007CrossrefMedlineGoogle Scholar
  • 29. Tong D, Li TY, Naus KE, Bai D, Kidder GM. In vivo analysis of undocked connexin43 gap junction hemichannels in ovarian granulosa cells.J Cell Sci. 2007; 120(Pt 22):4016–4024. doi: 10.1242/jcs.011775CrossrefMedlineGoogle Scholar
  • 30. Moore AR, Zhou WL, Sirois CL, Belinsky GS, Zecevic N, Antic SD. Connexin hemichannels contribute to spontaneous electrical activity in the human fetal cortex.Proc Natl Acad Sci U S A. 2014; 111:E3919–E3928. doi: 10.1073/pnas.1405253111CrossrefMedlineGoogle Scholar
  • 31. Smyth JW, Vogan JM, Buch PJ, Zhang SS, Fong TS, Hong TT, Shaw RM. Actin cytoskeleton rest stops regulate anterograde traffic of connexin 43 vesicles to the plasma membrane.Circ Res. 2012; 110:978–989. doi: 10.1161/CIRCRESAHA.111.257964LinkGoogle Scholar
  • 32. Hershberger KA, Martin AS, Hirschey MD. Role of NAD+ and mitochondrial sirtuins in cardiac and renal diseases.Nat Rev Nephrol. 2017; 13:213–225. doi: 10.1038/nrneph.2017.5CrossrefMedlineGoogle Scholar
  • 33. Pogoda K, Kameritsch P, Retamal MA, Vega JL. Regulation of gap junction channels and hemichannels by phosphorylation and redox changes: a revision.BMC Cell Biol. 2016; 17 Suppl 1:11. doi: 10.1186/s12860-016-0099-3CrossrefMedlineGoogle Scholar
  • 34. De Bock M, Wang N, Bol M, Decrock E, Ponsaerts R, Bultynck G, Dupont G, Leybaert L. Connexin 43 hemichannels contribute to cytoplasmic Ca2+ oscillations by providing a bimodal Ca2+-dependent Ca2+ entry pathway.J Biol Chem. 2012; 287:12250–12266. doi: 10.1074/jbc.M111.299610CrossrefMedlineGoogle Scholar
  • 35. Nattel S. Molecular and Cellular Mechanisms of Atrial Fibrosis in Atrial Fibrillation.JACC Clin Electrophysiol. 2017; 3:425–435. doi: 10.1016/j.jacep.2017.03.002CrossrefMedlineGoogle Scholar
  • 36. Pattillo JM, Artim DE, Simples JE, Meriney SD. Variations in onset of action potential broadening: effects on calcium current studied in chick ciliary ganglion neurones.J Physiol. 1999; 514 (Pt 3):719–728. doi: 10.1111/j.1469-7793.1999.719ad.xCrossrefMedlineGoogle Scholar
  • 37. Palatinus JA, Rhett JM, Gourdie RG. Enhanced PKCε mediated phosphorylation of connexin43 at serine 368 by a carboxyl-terminal mimetic peptide is dependent on injury.Channels (Austin). 2011; 5:236–240. doi: 10.4161/chan.5.3.15834CrossrefMedlineGoogle Scholar
  • 38. Solan JL, Lampe PD. Connexin43 phosphorylation: structural changes and biological effects.Biochem J. 2009; 419:261–272. doi: 10.1042/BJ20082319CrossrefMedlineGoogle Scholar
  • 39. Axelsen LN, Calloe K, Holstein-Rathlou NH, Nielsen MS. Managing the complexity of communication: regulation of gap junctions by post-translational modification.Front Pharmacol. 2013; 4:130. doi: 10.3389/fphar.2013.00130CrossrefMedlineGoogle Scholar
  • 40. Voigt N, Heijman J, Wang Q, Chiang DY, Li N, Karck M, Wehrens XHT, Nattel S, Dobrev D. Cellular and molecular mechanisms of atrial arrhythmogenesis in patients with paroxysmal atrial fibrillation.Circulation. 2014; 129:145–156. doi: 10.1161/CIRCULATIONAHA.113.006641LinkGoogle Scholar
  • 41. Chugh SS, Havmoeller R, Narayanan K, Singh D, Rienstra M, Benjamin EJ, Gillum RF, Kim YH, McAnulty JH, Zheng ZJ, et al. Worldwide epidemiology of atrial fibrillation: a Global Burden of Disease 2010 Study.Circulation. 2014; 129:837–847. doi: 10.1161/CIRCULATIONAHA.113.005119LinkGoogle Scholar
  • 42. Orr N, Arnaout R, Gula LJ, Spears DA, Leong-Sit P, Li Q, Tarhuni W, Reischauer S, Chauhan VS, Borkovich M, et al. A mutation in the atrial-specific myosin light chain gene (MYL4) causes familial atrial fibrillation.Nat Commun. 2016; 7:11303. doi: 10.1038/ncomms11303CrossrefMedlineGoogle Scholar
  • 43. Peng W, Li M, Li H, Tang K, Zhuang J, Zhang J, Xiao J, Jiang H, Li D, Yu Y, et al. Dysfunction of myosin light-chain 4 (MYL4) leads to heritable atrial cardiomyopathy with electrical, contractile, and structural components: evidence from genetically-engineered rats.J Am Heart Assoc. 2017; 6:e007030. doi: 10.1161/JAHA.117.007030LinkGoogle Scholar
  • 44. Metzler MA, Raja S, Elliott KH, Friedl RM, Tran NQH, Brugmann SA, Larsen M, Sandell LL. RDH10-mediated retinol metabolism and RARalpha-mediated retinoic acid signaling are required for submandibular salivary gland initiation.Development. 2018; 145:dev164822. doi: 10.1242/dev.164822CrossrefMedlineGoogle Scholar
  • 45. De Jong AM, Maass AH, Oberdorf-Maass SU, Van Veldhuisen DJ, Van Gilst WH, Van Gelder IC. Mechanisms of atrial structural changes caused by stretch occurring before and during early atrial fibrillation.Cardiovasc Res. 2011; 89:754–765. doi: 10.1093/cvr/cvq357CrossrefMedlineGoogle Scholar
  • 46. Fan J, Lv Z, Yang G, Liao TT, Xu J, Wu F, Huang Q, Guo M, Hu G, Zhou M, et al. Retinoic acid receptor-related orphan receptors: critical roles in tumorigenesis.Front Immunol. 2018; 9:1187. doi: 10.3389/fimmu.2018.01187CrossrefMedlineGoogle Scholar
  • 47. Xiao Y, Hill MC, Zhang M, Martin TJ, Morikawa Y, Wang S, Moise AR, Wythe JD, Martin JF. Hippo signaling plays an essential role in cell state transitions during cardiac fibroblast development.Dev Cell. 2018; 45:153–169.e6. doi: 10.1016/j.devcel.2018.03.019CrossrefMedlineGoogle Scholar
  • 48. Shi G, Zheng X, Wu X, Wang S, Wang Y, Xing F. All-trans retinoic acid reverses epithelial-mesenchymal transition in paclitaxel-resistant cells by inhibiting nuclear factor kappa B and upregulating gap junctions.Cancer Sci. 2019; 110:379–388. doi: 10.1111/cas.13855CrossrefMedlineGoogle Scholar
  • 49. Nebigil CG, Désaubry L. The role of GPCR signaling in cardiac epithelial to mesenchymal transformation (EMT).Trends Cardiovasc Med. 2019; 29:200–204. doi: 10.1016/j.tcm.2018.08.007CrossrefMedlineGoogle Scholar
  • 50. Deo RC, Musso G, Tasan M, Tang P, Poon A, Yuan C, Felix JF, Vasan RS, Beroukhim R, De Marco T, et al. Prioritizing causal disease genes using unbiased genomic features.Genome Biol. 2014; 15:534. doi: 10.1186/s13059-014-0534-8CrossrefMedlineGoogle Scholar


eLetters should relate to an article recently published in the journal and are not a forum for providing unpublished data. Comments are reviewed for appropriate use of tone and language. Comments are not peer-reviewed. Acceptable comments are posted to the journal website only. Comments are not published in an issue and are not indexed in PubMed. Comments should be no longer than 500 words and will only be posted online. References are limited to 10. Authors of the article cited in the comment will be invited to reply, as appropriate.

Comments and feedback on AHA/ASA Scientific Statements and Guidelines should be directed to the AHA/ASA Manuscript Oversight Committee via its Correspondence page.