Pitx2 modulates a Tbx5-dependent gene regulatory network to maintain atrial rhythm

See allHide authors and affiliations

Science Translational Medicine  31 Aug 2016:
Vol. 8, Issue 354, pp. 354ra115
DOI: 10.1126/scitranslmed.aaf4891

The genetic underpinnings of atrial fibrillation

The irregular heartbeat of atrial fibrillation puts people in danger of stroke and heart disease; genomic studies have identified gene variants that increase the risk for this abnormality. Nadadur et al. now reveal how these genes influence the beat of the heart’s atrium. In a mouse model of atrial fibrillation, which lacks one of these genes, Tbx5, the authors describe a multitiered transcriptional network that links seven of these atrial fibrillation risk loci. Organized as an incoherent feed-forward loop, this network tightly controls expression of atrial rhythm genes, and its perturbation by the risk loci causes susceptibility to atrial fibrillation.


Cardiac rhythm is extremely robust, generating 2 billion contraction cycles during the average human life span. Transcriptional control of cardiac rhythm is poorly understood. We found that removal of the transcription factor gene Tbx5 from the adult mouse caused primary spontaneous and sustained atrial fibrillation (AF). Atrial cardiomyocytes from the Tbx5-mutant mice exhibited action potential abnormalities, including spontaneous depolarizations, which were rescued by chelating free calcium. We identified a multitiered transcriptional network that linked seven previously defined AF risk loci: TBX5 directly activated PITX2, and TBX5 and PITX2 antagonistically regulated membrane effector genes Scn5a, Gja1, Ryr2, Dsp, and Atp2a2. In addition, reduced Tbx5 dose by adult-specific haploinsufficiency caused decreased target gene expression, myocardial automaticity, and AF inducibility, which were all rescued by Pitx2 haploinsufficiency in mice. These results defined a transcriptional architecture for atrial rhythm control organized as an incoherent feed-forward loop, driven by TBX5 and modulated by PITX2. TBX5/PITX2 interplay provides tight control of atrial rhythm effector gene expression, and perturbation of the co-regulated network caused AF susceptibility. This work provides a model for the molecular mechanisms underpinning the genetic implication of multiple AF genome-wide association studies loci and will contribute to future efforts to stratify patients for AF risk by genotype.


The transcriptional architecture that confers robustness to cardiac rhythm must tightly control cardiac channel gene expression, because increased or decreased channel expression can cause cardiac arrhythmias (1). Atrial fibrillation (AF), the most common human arrhythmia, is an irregularly irregular pattern of atrial depolarization resulting in uncoordinated atrial contraction. AF affects more than 33 million people worldwide and represents a growing cause of morbidity and mortality (2). Although most AF presents in the context of concomitant cardiac pathology, such as hypertension and heart failure, genome-wide association studies (GWAS) have identified a genetic predisposition underlying AF.

Human AF GWAS have implicated multiple transcription factors, including TBX5 and PITX2 in AF, raising the possibility that perturbations of a gene regulatory network for atrial rhythm control may underlie some AF susceptibility (3). Dominant mutations in the T-box transcription factor TBX5 cause Holt-Oram syndrome, characterized by disrupted heart and limb development (46) and increased AF risk (6). GWAS has linked common sequence variants close to or intronic within TBX5 to prolongation of the PR interval—the time interval between the electrocardiogram (ECG) P wave and QRS, an AF risk factor—and, more recently, to increased AF risk itself (710). PITX2, a paired-like homeodomain transcription factor, plays a critical role in heart development and adult rhythm control. PITX2 is the most frequently reported AF-susceptibility locus (3). Adult-specific Pitx2 deletion in mice causes AF susceptibility and increased expression of ion channels linked to AF (11).

AF pathophysiology requires a cellular trigger event, which initiates inappropriate depolarization, and a fibrillogenic substrate, or abnormal myocardium that propagates the trigger. Most genetic loci implicated in AF encode ion channels that affect trigger frequency or substrate, but not both (3). Animal models based on disruption of these channels do not exhibit spontaneous AF in the absence of concomitant cardiac pathophysiology, and no mouse model of primary AF has been reported (12). We tested the hypothesis that adult-specific removal of Tbx5 in the mouse may cause atrial gene regulatory network dysfunction and AF. This removal generated primary, spontaneous, and sustained AF in the absence of other cardiac pathologies. We identified a regulatory element at PITX2 that was modulated by a canonical TBX5 binding site. Tbx5 haploinsufficiency–induced transcriptional and myocyte abnormalities and AF susceptibility were rescued by Pitx2 haploinsufficiency. We have therefore uncovered a multilevel gene regulatory network for atrial rhythm homeostasis, driven by TBX5 and repressed by PITX2, organized as a type 1 incoherent feed-forward loop (13). This transcriptional architecture links 7 AF GWAS loci in a single network, providing a molecular model for their genetic implication in AF risk.


Adult-specific Tbx5 deletion causes rapid-onset AF

Tbx5 was deleted from the adult mouse by combining a Tbx5 floxed allele (Tbx5fl) (5) with a tamoxifen (TM)–inducible Cre recombinase allele at the Rosa26 (R26) locus (14). Mice homozygous for both alleles (Tbx5fl/fl;R26CreERt2) were treated with TM at 6 weeks of age, removing the T-box DNA binding region and ablating TBX5 expression (4, 5, 15). Tbx5-deleted mice developed an irregularly irregular heart rate 1 to 2 weeks after TM treatment. ECGs of ambulatory TM-treated Tbx5fl/fl;R26CreERt2 mice showed a disorganized pattern of atrial activity compared to TM-treated R26CreERt2 controls (Fig. 1A). The absence of P waves in Tbx5fl/fl;R26CreERt2 mice but not in controls was observed by signal-averaging ~1000 heartbeats (Fig. 1B). Poincaré analysis of the heartbeat, comparing successive beat lengths (using RR intervals, the time interval between sequential ECG “R” peaks), demonstrated stable intervals in controls but tremendous instability in Tbx5fl/fl;R26CreERt2 mice, indicative of an irregularly irregular heartbeat and AF (Fig. 1C). Furthermore, a single atrial depolarization overlapped the surface P wave in R26CreERt2 controls, but multifocal irregular depolarizations were observed in Tbx5fl/fl;R26CreERt2 mice by in vivo intracardiac electrograms 2 weeks after TM treatment (Fig. 1D).

Fig. 1. Removal of Tbx5 from the adult mouse results in spontaneous, sustained AF.

(A) Representative ambulatory telemetry ECG of R26CreERt2 control mice and Tbx5fl/fl;R26CreERt2 10 days after receiving TM. ECGs are representative of n = 24 Tbx5fl/fl;R26CreERt2 and n = 10 R26CreERt2 mice. (B and C) Signal-averaged ECG waveforms of ~1000 consecutive beats (B) and Poincaré plots of RR interval against the subsequent RR + 1 interval (C) in R26CreERt2 (n = 10) and Tbx5fl/fl;R26CreERt2 (n = 24) mice. P wave is present in control but absent in Tbx5fl/fl;R26CreERt2 mice (arrows). Number of mice with absent P wave and irregularly irregular heart rhythm described in A and B, in Tbx5fl/fl;R26CreERt2 versus R26CreERt2. P = 7.6 × 10−9 by two-tailed Fisher’s exact test. (D) Intracardiac atrial electrogram recordings and surface ECG in R26CreERt2 and Tbx5fl/fl;R26CreERt2 mice. A, atrial electrical signal; V, far field ventricular electrical signal. Data are representative of n = 6 Tbx5fl/fl;R26CreERt2 and n = 3 R26CreERt2 mice. Number of mice with irregular atrial electrogram in Tbx5fl/fl;R26CreERt2 versus R26CreERt2. P = 0.012 by two-tailed Fisher’s exact test. (E) Pulsed-wave Doppler across mitral valve alongside surface ECG. “A” wave is present in R26CreERt2 and absent in Tbx5fl/fl;R26CreERt2 recordings, indicative of a lack of coordinated atrial contraction. Data are representative of n = 6 Tbx5fl/fl;R26CreERt2 and n = 3 R26CreERt2 mice. Number of mice with absent “A” wave in Tbx5fl/fl;R26CreERt2 versus R26CreERt2. P = 0.012 by two-tailed Fisher’s exact test. (F) Representative atrial voltage activation maps from R26CreERt2 and Tbx5fl/fl;R26CreERt2 mice 7, 12, and 14 days after completion of TM treatment. Atrial activation maps demonstrate that conduction waves traverse the atria in ~12 ms at 7 days, ~20 ms at 12 days, and ~40 ms at 14 days after TM treatment. n = 2 for each of the three groups. SVC, superior vena cava; RA, right atrium; LA, left atrium; RV, right ventricle; LV, left ventricle. Trans-atrial conduction speed in Tbx5fl/fl;R26CreERt2 versus R26CreERt2. P = 0.02 by analysis of variance (ANOVA). (G) Atrial macro-reentrant pathways observed in Tbx5fl/fl;R26CreERt2 mice 14 days after TM treatment. Reentrant pathway travels right to left within the posterior atrial wall and left to right through the anterior atrial wall. Images are representative of n = 2 mice at 14 days.

We further interrogated the atrial arrhythmia by examining transmitral valve blood flow by pulsed-wave Doppler echocardiography. Transmitral valve blood flow demonstrated two phases: an “E wave” followed by an “A wave” in control R26CreERt2 mice, but the A wave was absent in Tbx5fl/fl;R26CreERt2 mice (Fig. 1E). The absence of A-wave flow across the mitral valve is indicative of a lack of coordinated atrial contraction, a characteristic of AF. We observed AF by these metrics in 24 of 24 Tbx5fl/fl;R26CreERt2 mice and in 0 of 10 R26CreERt2 mice (P = 7.6 × 10−9, Fisher’s exact test). These findings demonstrate spontaneous sustained AF in adult-specific Tbx5-mutant mice.

Cardiac function of Tbx5fl/fl;R26CreERt2 mice at the onset of AF was unchanged compared to controls, with no difference in left ventricular ejection fraction observed in Tbx5fl/fl;R26CreERt2 compared to R26CreERt2 mice 14 days after TM treatment (fig. S1). Ventricular conduction abnormalities, including monomorphic ventricular tachycardia (fig. S2A), polymorphic ventricular tachycardia (fig. S2B), and prolonged sinus pauses (fig. S2C), were observed; however, these arrhythmias were not present at the time of AF onset in Tbx5fl/fl;R26CreERt2 mice and only manifested well after AF was observed in these animals, beginning ~3 weeks after TM treatment. Removal of Tbx5 from the adult mouse therefore causes primary AF in the absence of left ventricular dysfunction or ventricular conduction abnormalities.

To assess atrial action potential (AP) propagation, we performed ex vivo optical mapping of intact adult mouse hearts (16). Adult-specific Tbx5-mutant mouse hearts demonstrated progressively slowed atrial conduction before AF (Fig. 1F). Atrial activation time was 13 ms at day 7, 20 ms at day 12, and >40 ms at day 14 after TM treatment in Tbx5fl/fl;R26CreERt2 mice (Fig. 1F and movies S1 to S3). AP duration (APD) increased from 44 ± 4 ms to 53 ± 3 ms (120-ms pacing interval) at day 14 after TM treatment (figs. S3 and S4). Macro-reentrant arrhythmias were observed in Tbx5fl/fl;R26CreERt2 mice with paroxysmal and sustained fibrillating patterns by day 14 (Fig. 1G). In contrast, R26CreERt2 control mice showed a uniform atrial activation pattern with no conduction slowing 14 days after the TM regimen (Fig. 1F). No paroxysmal or sustained AF was observed. Thus, removal of Tbx5 slowed trans-atrial conduction fourfold between the first and second week after deletion, setting the stage for reentrant arrhythmias—an organ-level mechanism for AF.

Disrupted calcium flux causes abnormal cellular depolarizations in Tbx5-mutant atrial myocytes

Removal of Tbx5 caused prolongation of APs and abnormal autonomous depolarizations in murine atrial cardiomyocytes. APs of atrial myocytes isolated from Tbx5fl/fl;R26CreERt2 mice 7 days after TM treatment and paced at 0.5 Hz were significantly prolonged during phases 2 and 3 of the AP, compared to R26CreERt2 atrial myocytes (Fig. 2A); time to 90% repolarization (APD90) and 50% repolarization (APD50) were both prolonged. Early afterdepolarizations (EADs), delayed afterdepolarizations (DADs), and spontaneous triggered APs were all observed frequently in Tbx5fl/fl;R26CreERt2 myocytes but never in R26CreERt2 myocytes isolated 7 days after TM treatment and paced at 0.5 Hz (Fig. 2B). These triggers are consistent with the initiation of paroxysmal AF, the most common form of AF (2, 17, 18). The aberrant depolarizations observed in isolated atrial myocytes provide evidence for an arrhythmic trigger after Tbx5 removal.

Fig. 2. AP abnormalities in Tbx5fl/fl;R26CreERt2 atrial cardiomyocytes are mediated by disrupted calcium handling.

(A) Representative AP from atrial cardiomyocytes isolated from R26CreERt2 and Tbx5fl/fl;R26CreERt2 mice and the corresponding properties of the APs. Resting membrane potential (RMP), AP amplitude (APA), and APD at 50% (APD50) and at 90% (APD90) repolarization. Data are means ± SEM (Tbx5fl/fl;R26CreERt2, n = 30 cardiomyocytes; R26CreERt2, n = 14 cardiomyocytes; n > 5 animals per group). P values were determined by two-tailed t test. (B) Representative abnormal depolarization events—EAD, DAD, and spontaneous phase 4 depolarization—observed in atrial cardiomyocytes. Total numbers of abnormal spontaneous events were recorded in R26CreERt2 (n = 14) and Tbx5fl/fl;R26CreERt2 (n = 30) atrial cardiomyocytes from n > 5 animals per group. Data are means ± SEM. P values were determined by one-tailed Fisher’s exact test. (C) Representative tracings of calcium release in isolated cardiomyocytes from R26CreERt2 and Tbx5fl/fl;R26CreERt2 mice. Calcium was imaged with Fluo-4 dye, and myocytes were paced at 1 Hz. Properties of calcium transient spikes, including decay constant (τ), time to 50% decay, and time to peak, were recorded across R26CreERt2 (n = 11) and Tbx5fl/fl;R26CreERt2 (n = 54) cardiomyocytes from n > 5 animals in each group. Data are means ± SEM. P values were determined by two-tailed t test. (D) Representative APs of atrial cardiomyocytes isolated from R26CreERt2 and Tbx5fl/fl;R26CreERt2 adult mutant mice treated with calcium-chelating agent BAPTA (5 mM). AP properties of these cardiomyocytes were determined, including RMP, APD50, and APD90. Data are means ± SD from Tbx5fl/fl;R26CreERt2 (n = 11) and R26CreERt2 (n = 7) cardiomyocytes across n > 3 animals per group. P values were determined by either two-tailed t test for continuous measurements or two-tailed Fisher’s exact test for count-based measurements (EADs, DADs, spontaneous depolarizations, and total abnormal depolarization events).

Calcium-mediated inappropriate depolarizations of atrial myocardium have been implicated as an AF mechanism (1). Cytosolic calcium transients were prolonged in atrial myocytes isolated from Tbx5fl/fl;R26CreERt2 mice but not R26CreERt2 controls 7 days after TM treatment (Fig. 2C). The decay constant, time to 50% decay, and time to peak were significantly prolonged in Tbx5fl/fl;R26CreERt2 compared with R26CreERt2 atrial myocytes. To assess whether prolonged calcium transients were the cause of prolonged APs in Tbx5fl/fl;R26CreERt2 atrial myocytes, we current-clamped isolated atrial myocytes after including the calcium chelator 1,2-bis(2-aminophenoxy)ethane-N,N,N′,N′-tetraacetic acid (BAPTA) (5 mM) in the intracellular solution. BAPTA rescued AP prolongation in Tbx5fl/fl;R26CreERt2 atrial myocytes at APD50 and APD90 (Fig. 2D). Calcium chelation by BAPTA also rescued abnormal depolarizations in Tbx5fl/fl;R26CreERt2 atrial myocytes (P = 1.0 versus R26CreERt2, P = 1 × 10−4 versus Tbx5fl/fl;R26CreERt2, without BAPTA, by two-tailed Fisher’s exact test) (Fig. 2D). Thus, disrupted calcium handling causes AP abnormalities in adult-specific Tbx5-mutant mice.

Adult specific Tbx5-deletion disrupts expression of AF-susceptibility genes

To explore the molecular mechanism underlying AF, we identified TBX5-dependent transcripts by RNA sequencing (RNA-seq) transcriptional profiling left atria isolated from Tbx5fl/fl;R26CreERt2 and R26CreERt2 1 week after TM treatment, before the onset of AF. Numerous genes critical to atrial rhythm and about half of the genes previously linked to AF (3) were significantly down-regulated in Tbx5fl/fl;R26CreERt2 left atria by RNA-seq (Fig. 3, A to C, and Table 1). Using RNA-seq, we noted significant down-regulation of numerous genes required for calcium handling, including Ryr2, Atp2a2, and Sln (Table 1). There was a modest and significant up-regulation of calmodulins Calm2 and Calm3, but no change in the expression of CaV1.2 (Cacna1c), CamkII (Camk2a/b/d/g), or NCX (Slc8a1/2/3) (Table 1).

Fig. 3. TBX5-PITX2 gene regulatory network for atrial rhythm control.

(A) Volcano plot of relative transcript expression from the left atria of Tbx5fl/fl;R26CreERt2 versus R26CreERt2 mice. All significantly misregulated genes (q < 0.05) are labeled blue, and all nonsignificant transcripts are in red. TBX5 and PITX2 shared targets labeled. (B) Heatmap of all significantly misregulated (q < 0.05) transcripts in left atria of Tbx5fl/fl;R26CreERt2 (n = 6) and R26CreERt2 (n = 4) mice. Cladogram shows clustering of biological replicates. (C) Relative gene expression by quantitative polymerase chain reaction (qPCR) of known AF ion channels from left atria. Data are means normalized to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) and relative to R26CreERt2 expression. Data are means ± SEM (n = 4 R26CreERt2 and n = 7 Tbx5fl/fl;R26CreERt2). Experiments were performed in technical triplicate. *P < 0.05 versus R26CreERt2 controls by two-tailed t test. See Table 1 for complete list of genes analyzed. (D) PITX2 genomic locus (hg19) aligned with ATAC-seq of iPS cell–derived cardiomyocytes and 4C with the PITX2c promoter as viewpoint. The region cloned for enhancer activity is boxed in red. Below, a smaller scale view of assayed enhancer showing ATAC-seq, Encyclopedia of DNA Elements (ENCODE) deoxyribonuclease (DNase) hypersensitivity (DHS) (darker to lighter gray showing relative strength of DNase signal, darker being stronger signal), and vertebrate conservation (Cons.). The TBX5-binding motif, which should occur by chance every ~1000 nucleotides, is labeled. (E) Fold enrichment after TBX5 chromatin immunoprecipitation (ChIP) by qPCR from a control locus and PITX2 enhancer in the human left atrial appendage. Data are means ± SEM (n = 3). (F and G) In vitro luciferase response assay of the candidate PITX2 regulatory element in human embryonic kidney (HEK) 293T cells cotransfected with TBX5 expression vector or HL-1 atrial cardiomyocytes. Wild-type (WT) enhancer, full mutant variant enhancer lacking the TBX5 binding motif mutant enhancer, and single-nucleotide polymorphism (SNP) rs1906595 variant enhancer are shown. Data are means ± SEM normalized to empty pGl4.23 vector with TBX5 expression (n = 4 WT enhancer, n = 4 empty, n = 3 T-box mutant, n = 4 SNP variant). (H) In vitro luciferase response assay of candidate TBX5/PITX2 co-regulated elements at Gja1, Ryr2, Dsp, Atp2a2, and Scn5a in HEK cells cotransfected with TBX5 or TBX5 and PITX2 or HL-1 cardiomyocytes and corresponding T-box mutant enhancers. Data are means ± SEM, normalized to blank vector with corresponding overexpression (n = 5 for Atp2a2 enhancer + TBX5 and Atp2a2 enhancer + TBX5/PITX2; n = 3 for all other groups). Experiments in (E) to (H) were performed in technical triplicate. P values in (E) to (H) were determined by two-tailed t test.

Table 1. Gene expression changes in Tbx5fl/fl;R26CreERt2 mouse left atria.

Gene expression of ion channels implicated in AF and of Tbx5/Pitx2 cotargets in left atrial tissue from Tbx5f/f;R26CreERt2 adult mutant mice normalized to R26CreERt2 by RNA-seq or qRT-PCR, in independent cohorts. Data are means ± SEM (n = 4 R26CreERt2 and n = 6 Tbx5f/f;R26CreERt2 for RNA-seq; n = 4 R26CreERt2 and n = 7 Tbx5fl/fl;R26CreERt2 for qRT-PCR). Nonsignificant changes in the RNA-seq were not validated by real time, with the exception of Pitx2c, which was just above significance by RNA-seq.

View this table:

We validated the dysregulated genes critical to atrial conduction by reverse transcription PCR (RT-PCR), including the sarcoplasmic reticulum (SR) calcium handling genes Ryr2, Sln, and Atp2a2 (SERCA2); connexins Gja1 and Gja5; sodium channels Scn5a and Hcn4; potassium channel subunits Kcnj3, Kcnd3, Kcnn2, Kcnj8, Kcna5, Kcnj2, Kcnj5, and Abcc9; and desmoplakin (Dsp) (Fig. 3, A to C, and Table 1). Expression of some channels previously linked to AF, including Kcnq1, Kcne1, Kcnh2, and Scn1b, were unchanged, indicating specific gene expression changes after Tbx5 deletion. These data establish a TBX5-directed gene regulatory network for atrial rhythm.

A PITX2 cis-regulatory element is modulated by TBX5

PITX2 is the most frequently reported human AF GWAS locus and was significantly down-regulated in the left atrium of Tbx5fl/fl;R26CreERt2 mice (Fig. 3C and Table 1). We hypothesized that the genomic region comprising the AF GWAS signal at PITX2, 100 kilo–base pairs (kbp) upstream of the PITX2 promoter (3, 7, 19, 20), may harbor TBX5-dependent cis-regulatory elements (CREs). To identify candidate regulatory elements, we performed assay for transposase accessible chromatin sequencing (ATAC-seq) (21), which reveals open chromatin, indicative of CRE activity. ATAC-seq was performed on cardiac cells derived from human induced pluripotent stem (iPS) cells. A strong signal indicative of open chromatin was located in a single region upstream of PITX2, defining a candidate CRE (hg19 chr4:111,711,915–111,716,751) (Fig. 3D). This region, located 150 to 155 kb 5′ to the PITX2 transcriptional start site, also harbored marks for DNase I hypersensitivity (22), sequence conservation (22) from ENCODE data, and a canonical TBX5 motif, AGGTG (hg19 chr4: 111,712,182–111,712,186) (Fig. 3D) (23).

We performed high-resolution circular chromosome conformation capture (4C) in murine left atrial tissue to identify genomic regions that interacted with the Pitx2c promoter on a genome-wide scale, using the Pitx2c promoter as a viewpoint (Fig. 3D). The region syntenic to the candidate CRE made contacts with the Pitx2c promoter, indicating a cis-interaction (Fig. 3D). This topology, indicating proximity between the CRE and the Pitx2c promoter, supports the region as a candidate Pitx2 CRE. We then assayed the candidate CRE for TBX5 occupancy by ChIP-qPCR in human left atrial tissue. The CRE was significantly enriched 145-fold by TBX5 ChIP compared with a control locus (Fig. 3E), indicating TBX5 occupancy at this CRE in human left atria.

The CRE significantly activated transcription in response to TBX5 expression in HEK293T cells and in the HL-1 atrial cardiomyocyte cell line, which expresses endogenous Tbx5 (Fig. 3, F and G). A mutant CRE with the T-box binding site ablated (AGGTG to TTTTT) failed to activate expression in response to TBX5 in HEK293T cells (P = 0.8 versus blank vector control; t test) or in HL-1 cells (P = 0.53 versus blank vector control; t test) (Fig. 3, F and G). The CRE harbors a common SNP, rs1906595, residing in the T-box binding site. The minor allele (G) (29%) is in perfect linkage disequilibrium with the AF signal tagged by SNP rs2200733 (24). The minor allele (G) completes the canonical T-box binding element (AGGTG), whereas the major SNP allele (T) disrupts a central nucleotide of the T-box binding motif (AGTGT). We found that the major allele completely abolished CRE activity in response to TBX5 expression in HEK and HL-1 cells (Fig. 3, F and G). Together, these findings identified a TBX5-dependent CRE at PITX2 and suggested that TBX5 and PITX2 may co-regulate a common atrial gene regulatory network.

Scn5a, Gja1, Dsp, Ryr2, and Atp2a2 are co-regulated by TBX5 and PITX2

We hypothesized that TBX5 and PITX2 co-regulate a gene regulatory network essential for atrial rhythm. We compared atrial TBX5-dependent transcripts (Fig. 3, A to C) with PITX2-dependent cardiac transcripts (11). Scn5a, Gja1, Dsp, Ryr2, and Atp2a2 were significantly dysregulated genes in both Tbx5 and Pitx2 deletion models. These genes are all critical to atrial conduction, have been previously linked to AF (1, 25, 26), and were shown to be down-regulated in Tbx5-mutant adult mouse hearts and up-regulated in Pitx2-mutant adult mouse hearts (11).

To test the hypothesis that TBX5 and PITX2 directly co-regulated the expression of these genes, we defined regions with overlapping chromosomal occupancy of both PITX2 and TBX5 from published ChIP data sets (11, 23). Candidate elements were refined by sequence conservation and colocalization of active enhancer marks p300 and H3k27ac (23). The candidate enhancers at all five loci demonstrated significant activation in response to TBX5 expression in HEK293T cells: Gja1 (mm9 chr10: 88816980–88817471), Ryr2 (mm9 chr13: 12153077–12153901), Atp2a2 (mm9 chr5: 122923655–122924226), Scn5a (mm9 chr9: 119562160–119563164), and Dsp (mm9 chr13: 38244307–38245159) (Fig. 3H). All of the enhancers also showed significant positive responsiveness in HL-1 atrial cardiomyocytes, which was abrogated by mutation of the T-box binding motifs in each regulatory element. PITX2 coexpression blunted TBX5-dependent activation in all cases (Fig. 3H). These findings demonstrate that TBX5 and PITX2 directly and antagonistically co-regulate an atrial rhythm gene regulatory network and describe an incoherent feed-forward loop transcriptional architecture, driven by TBX5 and repressed by PITX2.

Reduced Pitx2 rescued atrial gene expression abnormalities and atrial arrhythmias caused by reduced Tbx5

The TBX5/PITX2 incoherent feed-forward loop model predicts that the effects of decreased Tbx5 dose may be mitigated by decreased Pitx2 dose. Adult-specific Tbx5 heterozygous mice (Tbx5fl/+;R26CreERt2) but not Pitx2c heterozygotes (Pitx2cfl/+;R26CreERt2) showed diminished left atrial expression of Tbx5 compared with R26CreERt2 mice (Fig. 4A and Table 2). Tbx5fl/+;R26CreERt2 left atria also showed significantly diminished expression of four of the five shared Tbx5/Pitx2 targets versus control animals: Scn5a, Ryr2, Gja1, and Atp2a2. Pitx2fl/+;R26CreERt2 heterozygotes showed reduced left atrial Pitx2 and increased Scn5a, but no change of Tbx5, Ryr2, Gja1, Dsp, or Atp2a2 expression versus R26CreERt2 mice (Fig. 4A and Table 2). Scn5a, Ryr2, Gja1, and Dsp expression were all significantly increased in adult-specific compound Tbx5;Pitx2 heterozygote (Tbx5fl/+;Pitx2fl/+;R26CreERt2) mice compared to Tbx5 heterozygote mice. Expression of these Tbx5/Pitx2 cotargets was normalized in compound Tbx5;Pitx2 heterozygotes to their expression in control R26CreERt2 mice. Thus, reduced Tbx5 dose caused decreased expression of Tbx5/Pitx2 cotargets that was rescued by reduced Pitx2 dose.

Fig. 4. Pitx2 haploinsufficiency rescues Tbx5 haploinsufficiency in mice.

(A) Relative transcript expression by qPCR in the left atria from Pitx2 and Tbx5 heterozygotes and Tbx5;Pitx2 compound heterozygotes. Data are means ± SEM normalized to R26CreERt2 mice (set as 1) (n = 11 R26CreERt2; n = 15 Tbx5fl/+;R26CreERt2; n = 4 Pitx2fl/+;R26CreERt2; n = 8 Tbx5fl/+;Pitx2fl/+;R26CreERt2). *P < 0.05, two-tailed t test. Experiments were performed in technical triplicate. (B) Representative Poincaré plot of RR interval versus the subsequent RR + 1 interval (n = 7 R26CreERt2, n = 6 Tbx5fl/+;R26CreERt2, n = 9 Pitx2fl/+;R26CreERt2, and n = 7 Tbx5fl/+;Pitx2fl/+;R26CreERt2). (C) P-wave duration and PR interval calculated from ambulatory telemetry ECG recordings from mice in (B). P values were determined by two-tailed t test. Data are means ± SEM. (D) Representative AP recordings from atrial myocytes isolated from R26CreERt2, Tbx5fl/+;R26CreERt2, Pitx2fl/+;R26CreERt2, and Tbx5fl/+;Pitx2fl/+;R26CreERt2 mice. RMP, APA, APD50, and APD90 repolarization were determined from n > 3 animals per group (n = 14 R26CreERt2 cells, n = 15 Tbx5fl/+;R26CreERt2, n = 14 Pitx2fl/+;R26CreERt2, and n = 13 Tbx5fl/+;Pitx2fl/+;R26CreERt2). P values were determined versus R26CreERT2 controls. Inappropriate depolarization P values were measured by two-tailed Fisher’s exact test; APD90 and APD50 were determined by two-tailed t test. (E) Pacing induction by intra-atrial pacing of R26CreERt2 (n = 10), Tbx5fl/+;R26CreERt2 (n = 7), Pitx2fl/+;R26CreERt2 (n = 5), and Tbx5fl/+;Pitx2fl/+;R26CreERt2 (n = 7) mice. P values were determined by two-tailed Fisher’s exact test.

Table 2. Gene expression changes in Tbx5fl/+;R26CreERt2, Pitx2fl/+;R26CreERt2 , and Tbx5fl/+;Pitx2fl/+; R26CreERt2 mouse left atria.

Gene expression TBX5/PITX2 shared targets in left atrial tissue normalized to R26CreERt2 by qRT-PCR, in independent cohorts. Data are means ± SEM (n = 4 R26CreERt2 and n = 6 Tbx5f/f;R26CreERt2 for RNA-seq; n = 4 R26CreERt2 and n = 7 Tbx5fl/fl;R26CreERt2 for qRT-PCR). *P < 0.05 versus R26CreERt2, P < 0.05 versus Tbx5fl/+;R26CreERt2. n = 11 R26CreERt2; n = 15 Tbx5fl/+;R26CreERt2; n = 4 Pitx2fl/+;R26CreERt2; n = 8 Tbx5fl/+;Pitx2fl/+;R26CreERt2. For Tbx5fl/+;R26CreERt2 versus R26CreERt2 Tbx5: P = 1.5 × 10−5; Pitx2c: P = 0.86; Scn5a: P = 0.028; Ryr2: P = 3.98 × 10 −5; Gja1: P = 6.98 × 10−3 and Atp2a2: P = 5.5 × 10−3. For Pitx2fl/+;R26CreERt2 versus R26CreERt2 Pitx2c: P = 0.05; Scn5a: P = 0.03; Tbx5: P = 0.47; Ryr2: P = 0.44; Gja1: P = 0.52; Atp2a2: P = 0.60. For Tbx5fl/+;Pitx2fl/+;R26CreERt2 versus Tbx5fl/+;R26CreERt2 Tbx5: P = 0.56; Scn5a: P = 7.5 × 10−3; Ryr2: P = 0.01; Gja1: P = 0.02; Atp2a2: P = 0.19; Dsp: P = 0.05. Tbx5fl/+;Pitx2fl/+;R26CreERt2 versus R26CreERt2 Scn5a: P = 0.58; Ryr2: P = 0.07; Gja1: P = 0.99; Atp2a2: P = 0.19; Dsp: P = 0.31. By two-tailed t test. Experiments performed in technical triplicate.

View this table:

We tested whether atrial rhythm was sensitive to Tbx5 dosage by examining adult-specific Tbx5 heterozygotes. Tbx5fl/+;R26CreERt2 mice demonstrated an irregularly irregular rhythm by surface ECG 2 weeks after TM administration (Fig. 4B and fig. S5A). AF was reproducibly induced in 6 of 7 Tbx5fl/+;R26CreERt2 mice using catheter-directed intracardiac pacing with either an S1/S2 coupling interval or burst pacing, whereas 0 of 10 R26CreERt2 mice experienced AF (Fig. 4E and fig. S5B). Atrial myocytes isolated from Tbx5fl/+;R26CreERt2 heterozygotes demonstrated inappropriate depolarizations and prolonged atrial APs at both 50% and 90% repolarization (Fig. 4D).

Atrial rhythm was also sensitive to Pitx2 dosage. Adult-specific Pitx2 haploinsufficiency (Pitx2fl/+;R26CreERt2) caused no abnormalities of P-wave duration or PR interval, the time interval between the ECG P wave and QRS, by surface ECG (Fig. 4C). However, Pitx2fl/+;R26CreERt2 atrial myocytes showed shortened APDs compared to R26CreERt2 mice as well as increased APA (Fig. 4D). Furthermore, Pitx2fl/+;R26CreERt2 mice were susceptible to pacing-induced AF (Fig. 4E and fig. S5B).

Remarkably, the atrial rhythm and cellular electrophysiology abnormalities caused by reduced Tbx5 dose were all rescued by reduced Pitx2 dose. Compound adult-specific Tbx5fl/+;Pitx2fl/+;R26CreERt2 mice showed no atrial rhythm instability or other ECG abnormalities, by surface ECG or Poincaré analysis 2 weeks after TM treatment (Fig. 4B and fig. S5A). P-wave duration, prolonged in Tbx5fl/+;R26CreERt2 mice 1 week after TM administration, was rescued in Tbx5fl/+;Pitx2fl/+;R26CreERt2 mice (Fig. 4C and fig. S5A). No discernible PR interval difference was observed between any of these groups (Fig. 4C).

Furthermore, decreased Pitx2 dose rescued the cellular electrophysiology abnormalities observed in Tbx5fl/+;R26CreERt2 mice. APD90 prolongation was rescued in Tbx5fl/+;Pitx2fl/+;R26CreERt2 atrial myocytes (Fig. 4D). Additionally, the ectopic depolarizations observed in Tbx5fl/+;R26CreERt2 were also rescued in Tbx5fl/+;Pitx2fl/+;R26CreERt2 mice (Fig. 4D). Finally, Tbx5fl/+;Pitx2fl/+;R26CreERt2 mice were not susceptible to AF induction by intracardiac pacing (Fig. 4E and fig. S5B). We conclude that the defects in atrial rhythm, AF susceptibility, and cellular electrophysiology caused by reduced Tbx5 dose were all rescued by reduced Pitx2 dose.


This study links two cardiac transcription factors, TBX5 and PITX2, in a regulatory circuit that controls atrial rhythm. Removal of Tbx5 profoundly disrupts cardiac channel gene expression and causes primary, spontaneous AF in mice. Despite the high incidence and healthcare burden of AF, there is a limited understanding of the mechanisms underlying its genetic susceptibility. These data provide mechanistic support for a TBX5-driven gene regulatory network in AF risk and link human AF GWAS loci in atrial rhythm control.

We generated a model of primary spontaneous AF in the absence of other cardiac pathology or developmental defects by removing Tbx5 from the adult mouse. Tbx5 deletion was sufficient to produce the two components of AF pathophysiology: a “trigger” (ectopic depolarizations) and “substrate” that propagates the trigger (1). Atrial myocytes demonstrated triggered activity, including EADs, DADs, and phase 4 depolarizations. Ex vivo optical mapping revealed slowing of atrial conduction velocity, a substrate allowing arrhythmia propagation through reentry. Trigger and substrate after Tbx5 removal may explain the resulting primary spontaneous AF, previously elusive in mice.

We then defined a Tbx5-dependent network that accounts for both ectopic depolarizations and prolonged APs in adult-specific Tbx5-mutant mice. Calcium cycling genes can cause ectopic depolarizations (1), and we observed direct regulation of Ryr2 and Atp2a2 by TBX5. We hypothesize that the primary mechanism for the prolonged calcium transient and prolonged AP is the decrement in SERCA2 (Atp2a2) expression. Decreased SR Ca2+ reuptake resulted in prolongation of cytosolic calcium transient, which in turn may contribute to both prolonged APs and propensity for afterdepolarizations. However, we cannot discount the role of other calcium handling genes, such as Ryr2 and Sln, in the arrhythmogenic cellular electrophysiologic phenotype. The adult-specific Tbx5-mutant atrial cardiomyocyte phenotype was rescued by calcium chelation, demonstrating the importance of altered calcium cycling in this setting.

Decreased expression of potassium channel subunits that modulate IK1, IKr, and Ito1 currents have been implicated in prolonged APs and abnormal triggering propensity (1). In this vein, expression of Kcna7, Kcnd2, Kcnd3, Kcng2, Kcnj11, Kcnj2, Kcnj3, Kcnj5, Kcnj8, Kcnk3, and Kcnv2 were all significantly decreased after Tbx5 deletion. Slowed conduction was observed by ex vivo mapping, likely owing to decreased expression of sodium channel gene Scn5a and high conductance gap junction connexins Gja5 and Gja1. Consistent with our previous work (27), we did not observe significant expression of Scn10a in the adult mouse atria. Thus, Tbx5 is a driver of rhythm control genes with specific roles, including those that control AP fidelity and myocardial conduction speed. Decreasing the output of this network causes both arrhythmogenic trigger and substrate, providing a mechanism for the observed arrhythmias (Fig. 5).

Fig. 5.

A TBX5-PITX2 regulatory loop regulates atrial rhythm. (A) Loss of TBX5 in the adult atrium leads to slowed atrial conduction, prolonged AP, and disrupted calcium handling, leading to AF trigger and substrate. (B) A TBX5-PITX2 incoherent feed-forward loop regulates atrial conduction genes Scn5a, Gja1, Ryr2, Atp2a2, and Dsp. TBX5 drives PITX2 expression. TBX5 and PITX2 positively and negatively regulate downstream targets. Misregulation of this loop disrupts atrial conduction.

The intersection of Tbx5- and Pitx2-dependent transcripts identified a subset of essential atrial rhythm genes oppositely regulated by TBX5 and PITX2. We showed here that decreased transcriptional activator TBX5 or decreased transcriptional repressor PITX2 causes opposite effects on downstream effector gene regulatory network expression, yet both cause AF susceptibility (11). Thus, AF can occur as a common phenotypic endpoint of opposite perturbations of atrial rhythm gene expression. This observation suggests that substratification of AF patients based on underlying molecular mechanism may afford personalized approaches to this common disorder.

The TBX5-PITX2 loop links critical AF loci and anticipates the identification of genetic variation functionally responsible for AF risk and future genotype-based AF risk stratification. We identified common genetic variation that associates with AF and affects a CRE at PITX2. SNP rs1906595 resides within a T-box binding site in the TBX5-driven enhancer identified at PITX2. AF risk associates with the minor allele and canonical T-box site, and therefore correlates with increased TBX5-driven enhancer activity. We speculate that this SNP may increase PITX2 expression and thereby decrease downstream cardiac channel gene regulatory network expression. This model is consistent with previously published work, in which AF risk SNPs at PITX2 are associated with increased PITX2 expression (28).

We defined the architecture of an atrial gene regulatory network as an incoherent feed-forward loop, driven by TBX5 and repressed by PITX2 (Fig. 5B). A TBX5-dependent CRE was identified at PITX2, and TBX5- and PITX2-dependent CREs were identified at shared target loci, genes critical to calcium handling (Ryr2 and Atp2a2), rapid depolarization (Scn5a), and intracellular communication (Gja1 and Dsp). TBX5 and PITX2 acted directly and antagonistically at the terminal atrial channels of the atrial rhythm gene regulatory network. TBX5/PITX2 antagonism was also observed genetically, in which the physiologic effects of reduced Tbx5 dose, including decreased expression of critical AF genes, atrial rhythm instability, cellular AP abnormalities, and AF susceptibility, were all rescued by reduced Pitx2 dose.

The robustness of atrial contraction cycles requires the coordinated expression of genes governing AP fidelity, cell-cell conduction, and free calcium. Variation in these properties is arrhythmogenic (1). Incoherent feed-forward loops impose molecular buffering on terminal regulatory network transcriptional output in response to upstream variance in gene expression (2931). The Tbx5/Pitx2 transcriptional architecture is therefore ideal for an atrial gene regulatory network, which must maintain uniform channel gene expression and cellular electrophysiological behavior for normal atrial rhythm for billions of iterations. The TBX5-driven PITX2-modulated incoherent feed-forward loop provides a molecular logic for atrial rhythm homeostasis.

The transcriptional architecture we describe is based on in vitro studies and in vivo genetic experiments in mice. Although the CREs we tested were human elements or conserved in humans, we have yet to formally observe the incoherent feed-forward loop transcriptional architecture in human atria. Essential future steps include an evaluation of the transcriptional logic of atrial rhythm in humans, the identification of genetic variation in the essential regulatory elements comprising the network, and testing of the functional impact of variants on atrial gene expression and atrial rhythm control. TBX5 and PITX2 do not act in isolation, but are undoubtedly part of a much larger atrial transcriptional complex. How the complete compendium of transcriptional components interact to effect atrial rhythm homeostasis and how genetic variation in the broader atrial gene regulatory network affects atrial rhythm provide opportunities for future studies.

An essential goal of the functional genetics and genomics approaches applied here is to transition from genetic implication of GWAS to molecular mechanism underlying the genetics of AF risk. This work supports a model in which AF is a common endpoint disease process resulting from opposite perturbations of an atrial rhythm control network. This model suggests that a genotype-based personalized approach to the treatment of AF may ultimately be possible. We expect that these and analogous efforts will contribute to improved platforms for disease risk determination and therapeutic stratification.


Study design

The objective of our study was to investigate the role of TBX5 in atrial rhythm. We used murine models of conditional Tbx5 deletion, conditional Pitx2 deletion, and in vitro assays in murine HL-1 cells and human iPS cell–derived cardiomyocytes. Mouse models were chosen for their similarity in cardiac physiology to humans and for the availability of specific genetic tools. Human iPS cell–derived cardiomyocytes were used as an in vitro model of the regulatory loci. Sample sizes were chosen based on power calculations after pilot studies to estimate population mean and SD. All recordings and analyses were conducted in a blinded fashion. Genotype controls were used, and mice randomized into groups when appropriate. Endpoints for studies were selected based on the progression of phenotype. Samples were excluded if replicates were >2 SDs away from the population mean. Replication for specific experiments is in the figure legends.

Generation of conditional Tbx5-deleted mice

Tbx5 was deleted from the adult mouse using the published Tbx5 floxed allele (Tbx5fl) (5), as described in Supplementary Materials and Methods.

ECG recordings

Mice were implanted with subcutaneous telemetry transmitters (ETA-F10; Data Science International) (15). Arrhythmia analysis was carried out using Ponemah Physiology Platform software. Signal average and Poincaré plots were generated with a custom Python script. Catheter-based intracardiac electrophysiology is described in Supplementary Materials and Methods.

Myocyte isolation and AP recordings

To isolate mouse cardiomyocytes, hearts were excised and mounted on a Langendorff apparatus and retrogradely perfused with collagenase type 2 (Worthington Biochemical) in perfusion buffer. Membrane potential was recorded with a ruptured patch current clamp at 37°C using an Axon Axopatch 200B amplifier. During recording, cells were perfused in normal Tyrode buffer. APs were triggered using 0.5 nA × 2 ms current clamp pulses. When noted, 5 mM BAPTA-tetrapotassium was included in the pipette solution to abolish cytosolic calcium transients. In all cases, liquid junction potentials were corrected before recording APs. Additional details are in Supplementary Materials and Methods.

RNA-seq and gene expression analysis

Left atrial free wall from Tbx5fl/fl;R26CreERt2 and R26CreERt2 mice was removed, and total RNA was prepared with a TRIzol-based extraction. Sequencing libraries were prepared with Ribo-Zero purification (Illumina) and sequenced on an Illumina HiScanSQ instrument. Samples were sequenced 50 bp single-ended at 10 million to 20 million reads per replicate. Library preparation and sequencing were performed at the University of Chicago Genomics Core Facility. Sequence was aligned to the mouse genome with TopHat2, and differential expression testing was performed with the DESeq2 pipeline. Heatmap and volcano plot were generated in R with the ggplot2 and gplot packages, respectively. Candidate targets were validated with quantitative real-time PCR using exon-spanning primers specific to the gene of interest in a distinct cohort of biological replicates.

ChIP-qPCR of human atrial tissue

Left atrial appendage tissue was collected from three patients undergoing robotic valvular surgery and left atrial appendage plication (Institutional Review Board 12797B). All patients were genotyped as heterozygous at SNP rs1906595. ChIP was performed as previously described using a ChIP-grade TBX5 antibody (sc-17866, Santa Cruz Biotechnology) (32). To determine fold enrichment, we performed qPCR using input controls compared with DNA bound to immunoprecipitated proteins, using primers specific to the site of interest and primers to a site not expected to be enriched. Control primers to the human GAPDH locus were as follows: 5′-TACTAGCGGTTTTACGGGCG-3′ and 5′-TCGAACAGGAGGAGCAGAGAGCGA-3′. Primers tiled across the candidate enhancer were 5′-AGTGGCATCAAGACAGCACA-3′ and 5′-CCCCGGATCACCAAATCCAAG-3′, 5′-GTGGGCTGGGTGACTGTATT-3′ and 5′-CCCTGCACTCATGCTGGTTA-3′, and 5′-GCTGCACAACTTAGCTGCAA-3′ and GCCAGAAACAACCTCAAAGCA-3′.

Statistical analysis

For comparison of quantitative metrics of APs, calcium transients, gene expression by quantitative RT-PCR (qRT-PCR), ChIP-qPCR, luciferase response, and ECG parameters, two-tailed t tests were used to test significance, with a Bonferroni correction for multiple comparisons when appropriate. For count-based analysis of cellular depolarization events and AF inducibility, two-tailed Fisher’s exact tests were used. RNA-seq differential expression was performed with the DE-Seq2 pipeline, with a false discovery rate of <0.05.


Materials and Methods

Fig. S1. No change in cardiac function after onset of AF in Tbx5fl/fl;R26CreERt2 mice.

Fig. S2. Tbx5fl/fl;R26CreERt2 develop more severe ventricular arrhythmias over time.

Fig. S3. Tbx5fl/fl;R26CreERt2 show paroxysmal AF.

Fig. S4. Optical AP from right atrium of Tbx5fl/fl;R26CreERt2 mice.

Fig. S5. Atrial rhythm instability and AF inducibility in Tbx5fl/+;R26CreERt2 is rescued by Pitx2 haploinsufficiency.

Table S1. qRT-PCR primers used.

Movie S1. Ex vivo optical mapping in Tbx5fl/fl;R26CreERt2 mice at 7 days after TM.

Movie S2. Ex vivo optical mapping in Tbx5fl/fl;R26CreERt2 mice at 12 days after TM.

Movie S3. Ex vivo optical mapping in Tbx5fl/fl;R26CreERt2 mice at 14 days after TM.

References (3339)


Acknowledgments: We thank H. Balkhy for assistance in obtaining tissue samples and C.-W. Wu for assistance in cellular electrophysiology studies. This research was supported in part by NIH through resources provided by the Computation Institute and the Biological Sciences Division of the University of Chicago and Argonne National Laboratory, under grant 1S10OD018495-01. We specifically acknowledge the assistance of L. Pesce. Funding: The study was funded by grants from the NIH (R01 HL114010 to I.P.M; R01HL118761, R01DE023177, and U54 HD083092 to J.F.M.; R01 HL114395 and R01 HL126802 to I.R.E.; and HL128075 to E.M.M.), The Leducq Foundation (FP058566-01-PR, to I.P.M., J.F.M., and V.C.), and the American Heart Association (Established Investigator Award 13EIA14690081 to I.P.M.). This research was supported in part by NIH through resources provided by the Computation Institute and the Biological Sciences Division of the University of Chicago and Argonne National Laboratory, under grant 1S10OD018495-01. Author contributions: R.D.N. was involved in design execution, analysis of all experiments, wrote the manuscript, and performed statistical analysis. M.T.B. performed and analyzed whole animal electrophysiology and echocardiography and wrote the manuscript. B.B. performed and analyzed ex vivo optical mapping experiments and wrote the manuscript. S.R.M. performed and analyzed single cell electrophysiology experiments and wrote the manuscript. X.Y. analyzed genomics experiments and performed statistical analysis. M.v.B. performed and analyzed chromosome conformation capture and wrote the manuscript. J.B. performed and analyzed gene expression, chromatin immunoprecipitation, and cellular experiments. M.G. performed and analyzed whole animal electrophysiology and chromatin immunoprecipitation experiments. Y.Q. performed and analyzed ex vivo optical mapping experiments. T.W. performed and analyzed iPS cardiomyocyte experiments and wrote the manuscript. M.Z. performed and analyzed enhancer identification and in vitro cellular experiments. J.F.M. analyzed enhancer identification and in vitro cellular experiments and wrote the manuscript. C.E.S. and J.S. analyzed iPS cardiomyocyte experiments and wrote the manuscript. V.C. analyzed chromosome conformation capture experiments and wrote the manuscript. I.R.E. analyzed ex vivo optical mapping experiments and wrote the manuscript. E.M.M. analyzed whole animal physiology, in vitro assays, and wrote the manuscript. C.R.W. performed and analyzed cellular electrophysiology experiments and wrote the manuscript. I.P.M. was involved in designing, performing, and analyzing all experiments; wrote the manuscript; and performed statistical analysis. All authors approved the final version. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The RNA and ATAC sequencing data from this study have been deposited in the Gene Expression Omnibus database.

Stay Connected to Science Translational Medicine

Navigate This Article