Transcriptional factors in calcium mishandling and atrial fibrillation development

Abstract

Healthy cardiac conduction relies on the coordinated electrical activity of distinct populations of cardiomyocytes. Disruption of cell–cell conduction results in cardiac arrhythmias, a leading cause of morbidity and mortality worldwide. Recent genetic studies have highlighted a major heritable component and identified numerous loci associated with risk of atrial fibrillation, including transcription factor genes, particularly those important in cardiac development, microRNAs, and long noncoding RNAs. Identification of such genetic factors has prompted the search to understand the mechanisms that underlie the genetic component of AF. Recent studies have found several mechanisms by which genetic alterations can result in AF formation via disruption of calcium handling. Loss of developmental transcription factors in adult cardiomyocytes can result in disruption of SR calcium ATPase, sodium calcium exchanger, calcium channels, among other ion channels, which underlie action potential abnormalities and triggered activity that can contribute to AF. This review aims to summarize the complex network of transcription factors and their roles in calcium handling.

Introduction

Atrial fibrillation (AF), characterized by irregular electrical activity leading to incomplete and asynchronous contractions, is associated with numerous negative outcomes including stroke, myocardial infarction, and death [95, 143]. AF presents a significant challenge to the health care industry as it complicates overall patient management and can increase treatment cost fivefold [1]. On average, the cost per patient in the first year following AF diagnosis is on the order of 18,000 dollars and the total annual cost to treat AF patients in the USA is on the order of 26 billion dollars [30, 95]. Given the growing global impact of AF, many studies have aimed to uncover the mechanisms that underlie the electrophysiological abnormalities observed in AF patients. Our understanding of AF has grown tremendously over the past few decades, and in particular, there is now a body of work demonstrating that AF has a strong genetic component The advent of new technologies such as genome wide association studies (GWAS) has allowed for the identification of potential genes that regulate processes involved in arrhythmia development, which can then be further examined using transgenic animal models to determine the mechanistic underpinnings of genetic regulation [37]. However, there is tremendous complexity in the genetic regulatory networks and their consequences on cardiomyocyte action potential generation, calcium handling, and cell–cell communication. This review aims to summarize our current understanding of this complex network, the roles of individual players, and their interplay

Substrate and trigger in AF

Much work has been done to understand the cellular mechanisms that underlie atrial fibrillation, and one area of major interest is calcium handling, as it is critical for normal excitation contraction coupling [11]. Electrical excitation occurs when a cardiac myocyte is depolarized via an action potential. The action potential of a cardiomyocyte is dependent on its location, but overall, there are several unifying characteristics. At the beginning of each depolarization cycle, cardiomyocytes are at a resting membrane potential. During phase 0 of the action potential, there is a rapid influx of sodium ions resulting in a fast depolarization, which is then followed by repolarization stages. Phase 1 of the action potential is characterized by potassium efflux (repolarizing current). Following this, voltage-gated calcium channels open, creating balance between potassium efflux (repolarizing current) and calcium and sodium entry into the cell (depolarizing current), resulting in a plateau phase (phase 2). Over the course of repolarization, sodium and calcium channels inactivate, and potassium current becomes dominant and the cell returns to its resting membrane potential (phase 3). At rest (phase 4), the membrane potential is maintained near the potassium equilibrium potential by I_K1 potassium channels. Membrane depolarization is coupled to cardiomyocyte contraction via the second messenger calcium in a process referred to as excitation–contraction coupling. Initial depolarization of the cell by sodium influx triggers the opening of voltage-gated L-type calcium channels, allowing extracellular calcium to flow down a steep electrochemical gradient into the cell. This initial calcium entry is sensed by ryanodine receptors (RyR2) and triggers release of a large burst of calcium from the sarcoplasmic reticulum (SR) into the cytosol via a calcium-induced calcium release mechanism. Free intracellular calcium binds to troponin C, thus promoting a conformational change that allows actin to interact with the myosin ATPase, allowing ATP hydrolysis and contraction to occur. During the relaxation phase, calcium must be removed from the cytosol through return into the SR via the SR Ca²⁺ ATPase (SERCA) or extruded from the cell through inward sodium calcium exchanger (NCX; 3Na⁺ entering for every Ca²⁺ being extruded). Decrease of intracellular calcium favors dissociation of calcium from troponin C, reversing contraction [11]. The flux of calcium into and back out of the cytosol is called a calcium transient, and the rising and falling phases follow the action potential depolarization and repolarization [10].

Intracellular calcium handling is tightly regulated and balanced in steady state such that the total calcium entering the cell during systole must leave the cell during diastole. The upstroke of the calcium transient corresponds to release of calcium from the SR and the downstroke corresponding to reuptake and extrusion. Both of these phases of the action potential and calcium transient can be regulated in multiple ways. For example, sympathetic regulation plays an important role in regulating function of proteins such as RyR2, L-type calcium channel, and phospholamban (a SERCA regulator), during the fight or flight response to increase contractility. This topic has been reviewed extensively [10, 11]. It is relevant to note that there are a number of structural and functional differences between atria and ventricle that impact calcium handling. For example, there is a rudimentary t-tubule system in the atria compared to the ventricle, and consequently calcium must diffuse from surface RyR2 clusters to non-junctional RyR2 clusters [17]. It is proposed that RyR2 is activated both by cytosolic Ca²⁺ as well as increased Ca²⁺ inside the SR, in contrast to the ventricle in which RyR2 is predominantly activated by cytosolic Ca²⁺ [84]. Furthermore, there are axial tubules as opposed to transverse tubules that have high-density clusters of RyR2 that allow for rapid calcium signaling [17]. Calcium handling proteins are also differentially expressed between atria and ventricle with SR calcium content and SERCA in the atria being significantly higher than in the ventricle and phospholamban being less abundant as SERCA is also regulated by sarcolipin. By contrast, RyR2 and NCX have lower expression in the atria [15, 16, 32]. Taken together all of these differences contribute to the shorter calcium transient durations and a distinct “U”-shaped calcium transient [17]. These are important considerations when trying to understand or model normal or inappropriate calcium handling in the atria, as models that are predominantly ventricular in characteristic may not fully capture aspects of atrial calcium handling. Such differences may also explain why perturbations that may have no significant effects to drive significant changes in EC coupling and calcium handling in ventricular myocardium can have functional implications in the atrium.

In our current understanding, disruption of normal calcium handling can play an important mechanistic role in arrhythmia development through inappropriately timed depolarizations. In theory, two components are required for an arrhythmia to occur: presence of an abnormal myocardial substrate and formation of an ectopic trigger. Abnormal substrate refers to altered electrical conduction between cardiomyocytes. Examples of arrhythmogenic substrate include reduced conduction speed via decreased gap junction conductance, intercalated disc abnormalities, fibrosis, reduced repolarization reserve, and any structural and functional obstacles that impede propagation in the appropriate path [153]. Loss of connexins and other intercalated disc proteins can result in decreased conduction velocity [134]. Additionally, functional obstacles could include changes that alter conduction or refractoriness of the tissue through inappropriate depolarization force during the action potential. For example, low-level depolarization of the membrane during phase 4 can inactivate sodium channels, leading to inactivated regions dispersed among excitable neighboring cells, creating a local conduction block [69].

The second component important to arrhythmia formation is the presence of an ectopic trigger, which refers to initiation of electrical activity at regions outside of the sinoatrial node. Reentry loops occur with increased frequency when an abnormal trigger initiates a wave of depolarization in the setting of abnormal substrate. The development of reentry circuits is prevented in healthy cardiac tissue by the brief refractory period immediately following the action potential, which prevents retrograde activation and promotes unidirectional myocardial activation. However, when obstacles are present that block or slow forward conduction, the depolarizing wave front may be delayed enough to allow depolarization to circle back to reactivate the initial area after the refractory period ends, creating a reentry circuit [3, 55, 129].

Ectopic activity, at a cellular level, is often categorized as being due to early afterdepolarizations (EADs) or delayed afterdepolarizations (DADs). EADs are defined as slowing or reversing of repolarization during the plateau or repolarization phase (phases 2 and 3), with phase 2 EADs being more commonly associated with prolonged APs. Increased depolarization drive, decreased repolarization drive, or a combination of both during the plateau phase of the AP can promote EAD formation. High depolarization drive can be caused by increases in late sodium, L-type calcium, or NCX. Reduced repolarization drive can be caused by decreased potassium currents. Often, increased depolarizing force from one source causes a positive feedback loop that promotes continued depolarization. For example, depolarization via late sodium current can reopen L-type calcium channels, which results in further depolarization and higher calcium levels in the cytosol, which leads to increased activity of NCX [21, 44, 56, 126]. EAD-triggered APs can occur through reactivated L-type calcium current or inward NCX. EADs have been demonstrated to play an important role in AF pathogenesis [21, 25, 28, 31, 67, 68, 90, 97]. DADs occur during phase 4 of the AP and are driven by spontaneous release of SR calcium, either through activation of inward NCX or Ca²⁺-activated chloride current [100, 129]. DADs can also trigger APs if the depolarization reaches threshold for opening of fast sodium channels. DADs have also been demonstrated to initiate triggered activity resulting in AF [31, 36, 48, 59, 67].

The tight coupling of individual cardiomyocytes to its neighboring cells serves as a critical mechanism by which depolarizing current from an EAD or DAD from one cell is distributed across many cardiomyocytes. For example, “EAD islands” can be created due to coupling between cells. When EADs trigger APs within these EAD islands, these APs may propagate into neighboring regions to trigger EADs in those regions [142]. However, tight coupling between individual cardiomyocytes also serves to dilute the potential of an EAD or DAD to depolarize a large group of cells, a concept called source-sink mismatch [40, 142]. Thus, in order to generate a spontaneous depolarization in tissue, nearly 7000 contiguous EAD-generating cardiomyocytes would be necessary. However, in cases when there is an abnormal substrate favoring reentry, only 40 EAD generating cells would be required [98, 145]. EADs and prolonged action potentials may result in heterogeneity of the refractory period, also promoting local conduction block [69]. These examples demonstrate that trigger and substrate may not be distinct, but rather both can be influenced by calcium handling alterations.

In summary, a significant body of work has established the importance of both ectopic triggers and arrhythmogenic substrates in the development of AF [1, 28, 33, 61, 68, 70, 90, 92, 93, 96, 112, 118, 136, 137, 151]. However, recent studies demonstrating that AF has a strong genetic component are pushing the field to examine the mechanistic link between changes at gene level and ectopic trigger and arrhythmogenic substrate formation. Of particular interest has been the role of transcription factors in regulating calcium handling proteins to mediate changes in trigger and substrate to affect cardiac rhythm.

Transcriptional regulators in AF

Work beginning nearly eight decades ago has generated a body of literature demonstrating a role of genetics in atrial fibrillation [37, 78]. In fact, if a patient has AF, there is over a thirty percent chance that they will also have at least one close relative who also has arrhythmias [5, 35, 39]. The earliest studies were in families where AF, but not heart disease, was commonly found. Using linkage studies, a number of genetic loci have been identified, including in KCNQ1, KCNE1, KCNJ, GJA, CNAN, SCN5, SCN10, CACNA, and NPPA [29]. Most of these genes encode for ion channels or junctional channels, with the exception of atrial natriuretic peptide precursor. While these mutations had a strong effect in individuals, they were often rare mutants, which made determining their relevance to the general population difficult to ascertain. With the advent of GWAS technology, a new perspective on the role of genetic modulation of AF risk has started to take form. Using GWAS methods, the whole genome could be surveyed in a large population of people to identify loci that had lower effect size, but were more prevalent in the population. Subsequently, several key ideas were generated. First, a new class of genetic loci important in AF development was identified: transcription factors such as TBX5, PITX2, GATA4–6, NKX2.5, ZFHX3, PRRX1 [81]. Second, studies linking these transcription factors to a direct mechanism for AF development revealed regulatory roles of these transcription factors in the expression of ion channels and calcium handling proteins as well as complex interactions between multiple transcription factors [22, 28, 29, 68, 90, 125, 141]. Third, new work has highlighted the role of microRNAs (miRNAs) in the regulation of both AF relevant transcription factors as well as on genes encoding channels directly [80]. Furthermore, the combination of new knowledge of these genetic risk factors along with traditional clinically known risk factors such as diabetes, hypertension, and cardiomyopathy has provided support of a “multi-hit” hypothesis, where underlying genetic risk factors, which could have a lower effect size on their own, when combined with other genetic risk factors or with other clinical factors contribute to the development of AF [29, 81]. In this review, we will mainly focus on transcription factors with specific emphasis on those with known downstream functional calcium handling effects, and aim to provide a brief introduction of individual transcription factors and their association with AF. We will address how these transcription factors can tribute to AF by driving changes in cellular calcium handling or excitation contraction coupling. We will also address the role of microRNAs and long noncoding RNAs in driving changes which may contribute to a predisposition to the development of AF. Summary tables can be found at the end.

PITX2

Multiple GWAS studies have demonstrated a strong association between atrial fibrillation and a region on chromosome 4q25 in patient cohorts across numerous countries including Iceland, Sweden, the USA, and Hong Kong. At least four distinct susceptibility signals were discovered in this chromosome region, and a single copy of the 4q25 variant increases AF risk by 65% [77, 131]. The closest gene to this locus is PITX2, encoding for paired-like homeodomain 2 protein, located about 150,000 bp away, making it a strong candidate gene to investigate the genetic basis of AF. Indeed, a recent study found cardiomyocyte PITX2 concentration along with PITX2-dependent protein BMP10 levels was predictive of AF recurrence [107]. Prior to discovery of its role in AF, PITX2 was known to play a key role in cardiac development, particularly in atrial septation, and formation of the sinoatrial node, outflow tract, and myocardial sleeves of the pulmonary vein [87, 88]. There are four known isoforms of PITX2: PITX2a, PITX2b, PITX2c, and PITX2d (which lacks a DNA binding site). PITX2c is expressed predominantly in the left atria, and embryonic haploinsufficiency of this isoform was sufficient to increase AF susceptibility following burst pacing in mice. Atrial myocytes isolated from these Pitx2 haploinsufficient mice showed reduced action potential duration [65]. Atrial-specific deficiency of PITX2 caused electrophysiological defects including atrioventricular node block as well as missing p waves. However, interestingly, changes in action potential duration were not observed [24]. In order to separate out developmental from post-natal effects of PITX2 loss, a conditional deletion model in which atrial Pitx2 was deleted after birth. In this model, Pitx2 conditional knockout mice demonstrated sinus node dysfunction, a phenotype-associated with AF in humans [127]. Independent studies on adult-specific Pitx2 haploinsufficient mice showed increased susceptibility to AF following burst pacing. Like constitutive Pitx2 haploinsufficiency, adult-specific Pitx2 haploinsufficiency resulted in decreased action potential duration. Several studies have also demonstrated that PITX2 may play a role in regulating resting membrane potential, with constitutive Pitx2 loss resulting in a more depolarized resting membrane potential [24, 124].

Given the genetic and functional studies indicating PITX2 plays in an important role in the regulation of atrial rhythm, determination of the mechanism through which Pitx2 changes may result in AF has become an area of great interest. Studies aimed at addressing such a mechanism have shown that PITX2 regulates a number of genes important to cellular calcium handling. Initial studies in embryonic Pitx2c haploinsufficiency mouse models demonstrated loss of PITX2C resulted in decreased Cacna1c (encoding for L-type calcium channel alpha 1c subunit) and decreased Kcnc4 (encoding voltage gated potassium channel 4) [65]. Mice with atrial-specific loss of Pitx2 showed decreased Kcnj2, Kcnj12, and Kcnj4 (encoding inward rectifier potassium channels) along with decreased Scn5a and Scn1b (encoding sodium channels) [24]. Interestingly, the opposite effect was observed in post-natal deletion of Pitx2. CHIP-seq revealed upregulation of calcium handling genes Cacna1d (Cav1.3 L-type calcium channel), Cacna2d2 (calcium channel auxiliary unit), Jph2 (junctophilin, linker between plasma membrane and sarcolemma), RyR2, and Atp2a2. Additionally, potassium channel genes Kcnq1, Kcnj11 were also upregulated [127]. These changes are supported by data in adult-specific Pitx2 haploinsufficiency, showing PITX2 and TBX5 co-regulate calcium handling genes Atp2a2 and Ryr2, with loss of PITX2 resulting in upregulation of both. Furthermore, modeling studies support that PITX2 loss-induced increase in SERCA activity can increase the probability of ectopic activity [7]. PITX2 has also been shown to upregulate Wnt8, which in turn regulates microRNAs involved in calcium handling. The complex interaction network between transcription factors suggests a mechanism by which AF can be caused by a number of different mechanisms. For example, both decreased and increased PITX2 are associated with AF, but with different transcriptional milieu resulting in different calcium handling and electrophysiological changes. Overexpression of PITX2c in HL-1 cells resulted in increased I_ks density but decreased I_Ca,L [105].

PITX2 can also regulate intercalated disc genes. In inducible Pitx2 knockout mice, clustering analysis showed that intercalated disc genes Gja1, Dsp, Plec, and Emd were downregulated and immunostaining found mislocalized β-catenin. These gene level changes were accompanied by disrupted intercalated discs caused by widened spaces between the junctions and dysfunctional swollen mitochondria. Furthermore, in human atrial fibroblasts with siRNA-induced PITX2c knockdown, profibrotic changes such as increased migration and matrix metalloproteinase-2 expression were observed [62]. Such changes may provide an arrhythmogenic substrate, increasing the likelihood of reentry circuits and sustained AF.

Thus, PITX2 is important to the regulation of a large number of ion channels and transporters providing multiple potential mechanisms by which electrical activity can be disrupted. Its net role in influencing myocardial electrical activity in AF likely depends critically on the summative effects of multiple pathways and downstream targets. Furthermore, the potential fibrillogenic role for PITX2 likely also relates to substrate level alterations. An important area for future study is to determine the predominant physiological mechanisms in humans and how Pitx2 mutations can drive arrhythmias in human disease.

TBX5

T-box transcription factor-5, TBX5, drives a gene regulatory network that includes PITX2 and GATA4 to tightly control calcium handling proteins and regulate AF susceptibility [90]. TBX5 was originally studied in a developmental context and is known to be critical for limb and cardiac development. Patients with mutations in TBX5 resulting in haploinsufficiency develop Holt-Oram syndrome, characterized by defects in the cardiac septation and formation of the cardiac conduction system [19, 20, 85]. Zebrafish Tbx5 mutants exhibit bradycardia and failure to develop properly with an elongated myocardium that fails to undergo looping morphogenesis [110]. TBX5 is expressed throughout most of the heart but is much more abundant in the atria compared to the ventricle [121]. Recently, Tbx5 variants have been implicated in increased AF risk by several GWAS [82, 131]. A mouse model of Tbx5 knockout was generated to directly examine its role in AF development. Adult-specific Tbx5 knockout in mice resulted in the development of spontaneous and sustained AF and isolated atrial cardiomyocytes from these mice show slowed calcium transient decay, prolonged AP, and increased incidence of EADs and DADs. RNA-sequencing of Tbx5 knockout atria revealed decreases in Ryr2 and Atp2a2, and presence of calcium chelating agent, BAPTA, inside the pipette normalized the AP, suggesting that calcium handling changes rather than potassium channel alterations likely contribute to AP alterations [90].

Recently, we found that Tbx5-deficient atrial cardiomyocytes exhibited concurrent decreases in SERCA-mediated SR calcium uptake, increases in NCX-mediated calcium efflux, and increases in calcium influx through L-type calcium channel. To demonstrate that these changes provide the mechanistic underpinnings for the AP abnormalities, we normalized SERCA function by removing SERCA inhibitor phospholamban and found that calcium transients were normalized. Furthermore, AP duration and triggered events could be rescued. Critically, Tbx5 knockout mice with normalized SERCA function no longer developed AF [28]. TBX5 regulation of SERCA has also been shown to be important in maintaining diastolic function. Namely, loss of one copy of Tbx5 results in decreased expression of SERCA2 and calcium uptake with prolonged calcium transients, leading to heart failure [152]. Computer modeling also showed that ectopic activity following TBX5 insufficiency could be accounted for by increased inward NCX current along with decreased outward IK1 current, which was not observed in animal studies [6].

Loss of Tbx5 can also play a role in regulating the formation of an arrhythmogenic substrate. Optical mapping studies demonstrated decreased conduction velocity in the atria of mice with adult-specific homozygous Tbx5 deletion. Loss of conduction velocity was likely due to the decreased expression of Tbx5 effector genes Gja1 and Dsp, which encode for intercalated disc proteins, as well as Scn5a, encoding for the sodium channel required for rapid depolarization [90]. In line with the evidence that the presence of a substrate and trigger increase the propensity for the formation of reentry loops, macro-reentrant arrhythmias were present in Tbx5-deficient mice [90].

TBX5 and PITX2 regulatory gene expression through an incoherent feed forward loop, such that Pitx2 haploinsufficiency can rescue Tbx5 haploinsufficiency-induced AF susceptibility. Pitx2 is regulated by Tbx5 via a cis-regulatory element (CRE) upstream of the Pitx2 promoter, and Tbx5 expression significantly increased transcriptional activation by this CRE. PITX2 and TBX5 have opposing effects on co-regulated target genes Ryr2, Atp2a2, as well as Scn5a, Gja1, and Dsp as PITX2 reduces TBX5-dependent activation of all of these genes [90]. In this way, Pitx2 haploinsufficiency relieves inhibition of Atp2a2 and Ryr2, allowing for restoration of the expression of these genes following Tbx5 loss, and thereby rescues calcium handling defects and AF susceptibility [90]. Data also support that in addition to TBX5 loss-off-function, TBX5 gain-of-function can potentially contribute to AF [119]. One possible mechanism is through upregulation of Cx40, transcripts of Tbx3 and Kcnj2 transcripts; however, functional studies in animal models have not yet been published [103].

In summary, TBX5 is an important regulator of cardiac rhythm through its regulation of calcium handling and arrhythmogenic substrate formation. Tbx5 loss-associated decrease in SERCA function and increase in NCX and L-type calcium current demonstrate a physiologically relevant mechanistic link between transcription factors and calcium handling to affect AF propensity. An important area for future study is the role of the interplay between TBX5 and other transcription factors such as PITX2, GATA4, and NKX2.5 in regulation of calcium handling and AF propensity through life.

GATA4

GATA4 was previously identified to be important for cardiac development and heart physiology [2, 43, 83, 91, 120]. Gata4 mutation has also been linked to the development of cardiac hypertrophy and apoptosis signaling [4, 13]. The role of GATA4 in these contexts has been reviewed elsewhere [123]. Evidence that GATA4 is an important regulator of cardiac rhythm is provided both by GWAS and studies of patients with lone AF. Mutations in the GATA4 gene found in AF families include M247T, A411V, G16C, H28D, S70T, S160T, y38d, and p103a. These sites are highly conserved evolutionarily and the G16C, H28D, S70T, S160T, y38d, and p103a mutations were found to decrease GATA4 transcriptional activity [58, 102, 140, 148]. Interestingly, inducing Gata4 haploinsufficiency alone in adult mice did not cause rhythm abnormalities but could rescue Tbx5 haploinsufficiency-induced AF in Tbx5/Gata4 compound haploinsufficient mice [68].

Similar to PITX2, GATA4 and TBX5 coregulate calcium handling proteins SERCA, RyR2, and NCX. Several studies have indicated GATA4 interacts with Atp2a2 promoters and enhancers. For example, in transverse aortic constriction hearts, Gata4 binding to Atp2a2 promoter region is decreased [75]. Gata4 is Rock dependent, and increased RhoA-Rock signaling induced Serca2 gene expression in neonatal ventricular cardiomyocytes [135]. While GATA4 was found to activate an Atp2a2 enhancer in HEK 293 T cells, Gata4 haploinsufficiency rescues SERCA function in Tbx5 haploinsufficient mice [68]. The antagonistic relationship between Gata4 and Tbx5 was also observed in the Gata4 repression of a Tbx5-dependent Ryr2 enhancer [68]. Additionally, many studies have demonstrated a role for GATA4 in the regulation of NCX, including mutation of GATA4 at −75, which fully abolishes Ncx1 promoter activity while mutation at −145 reduces Ncx1 promoter activity by 60% in rat neonatal cardiomyocytes [94]. GATA4 binds to the proximal portion of the endogenous Ncx1 promoter in adult cardiomyocytes at the −50 GATA element. While GATA4 binding region of Ncx1 promotor is important for expression of Ncx1, overexpression of GATA4 in adult cardiomyocytes did not increase Ncx1 expression likely because GATA4 association is not increased by GATA4 overexpression [146]. It is notable that NCX1 can also regulate Gata4. A mutation in a highly conserved leucine residue in Ncx1 led to reduced heart rate and a weak contraction in zebrafish. These mutants also downregulated Gata4 and further experiments with overexpressed Gata4 partially restored the cardiac defects seen in the mutants [27].

GATA4 can also regulate non-calcium handling genes that may play a role in AF development. For example, Hcn1 but not Hcn4 was decreased by GATA4 overexpression [141]. Furthermore, GATA4 in conjunction with SP1 also regulate the human CX40 promotor as an AF-associated polymorphism in GATA4 results in altered GATA4 interaction with the CX40 promoter, and thus decreased transcriptional activity [38]. Finally, luciferase assays showed significantly decreased activation activity of downstream ANP gene in GATA4 mutants found in families with lone AF [148].

In summary, GATA4 plays a role in AF formation as it has been implicated by GWAS and family studies, but the mechanisms underlying the regulation are still not fully elucidated. Recent studies illustrate an antagonistic relationship between Gata4 and Tbx5, in particular in the regulation of Ryr2 and the rescue of Tbx5 haploinsufficiency-induced AF by Gata4 insufficiency. Future mechanistic studies are required to understand the balance between Ryr2 regulation and other putative GATA4-dependent trans-sarcolemmal fluxes such as NCX1 and L-type calcium current.

NKX2.5

NKX2.5 is a part of the NK-2 homeobox family of transcription activators. Mutations to the NKX2.5 have been associated with many cardiac diseases, particularly congenital heart disease and familial dilated cardiomyopathy as NKX2.5 was identified to be important in heart formation, particularly atrial septation and sinoatrial node development [12, 26, 88, 149]. Embryonic homozygous Nxk2.5 deletion is lethal. Sinoatrial node-specific Nkx2.5 loss caused inverted, diphasic, flattened, or lost P waves and perinatal loss of Nkx2.5 also results in prolonged PR and QRS interval, indicating potential regulation of the atrioventricular node [18]. During cardiac development, NKX2.5 acts synergistically with the GATA family of transcription factors as well as TBX5. As with TBX5 and GATA4, GWAS have implicated NKX2.5 in AF. Additionally, missense mutations (F145S, N19D, and F186S) identified in familial AF studies showed significantly reduced transcriptional activity [52, 144]. Additionally, overexpression of Nkx2.5 in neonatal rat myocytes caused increased beat frequency and decreased APD50 [141]. Given these data, there is significant interest in understanding the mechanism by which NKX2.5 can regulate cardiac rhythm.

While heterozygous Nkx2.5 deletion in adult mice did not cause AF or alteration of calcium handling, and no combinatorial effects of NKX2.5 loss with either GATA4 or TBX5 loss were observed, Atp2a2 and Ryr2 both have NKX2.5 binding sites [9, 43]. Heterozygous germline deletion of Nkx2.5 significantly decreased gene expression of Scn5a, Cacna1c, and Kcnh2 and Nkx2.5 was found to bind to the Kcnh2 promoter [9]. Expression of a mutant form of NKX2.5 in HL-1 cells also significantly reduced KCNH2 channel current [9]. NKX2.5 knockout via siRNA in HL-1 cells resulted in decreased SERCA and Cav1.2, but increased phosphorylated RYR protein expression while overexpression of NKX2.5 in neonatal rat hearts led to upregulation of T-type calcium channel Cav3.2, but downregulation of Cav1.3 (L-type) [22, 141].

In summary, NKX2.5 plays a role in cardiac development through regulation of ion channels and calcium handling proteins. Given recent data demonstrating adult heterozygous deletion of Nkx2.5 did not result in rhythm abnormalities, future studies aimed at linking developmental and early life NKX2.5-driven changes to adult onset atrial fibrillation will be valuable.

CREM

Cardiac-specific expression of the human isoform of CRE-binding protein modulator (CREM), CREM-IbΔC-X has also been shown to be important in the regulation of cardiac development and atrial rhythm. Transgenic mice expressing CREM-IbΔC-X showed atrial dilatation and AF development with severity and onset determined by gene dose [89, 122]. Although mice less than 8 weeks old with CREM-IbΔC-X overexpression did not yet show persistent AF, there was already atrial dilatation, downregulation of mRNA for connexin 40, and prolonged APD50 and APD90. Interestingly, no changes in connexin 43 levels were noted [113, 122]. Additionally, loss of contractile filaments and some metabolic pathways were noted in the early stage prior to AF onset [113]. After 12 weeks, CREM-IbΔC-X overexpression mice showed persistent AF along with calcium handling changes including increased calcium sparks, reduced SR calcium load, and increased diastolic calcium leak, all suggestive of increased RyR2 release [66]. Indeed, genetic inhibition of RyR2-S2814 phosphorylation in CREM-IbΔC-X overexpression mice normalized RyR2-mediated calcium leak and prevented spontaneous AF [71]. This presents an interesting contrast to the Tbx5 knockout model of AF in which APs were also prolonged and SR calcium load also decreased, but there was a reduction in calcium sparks. Additionally, CREM-IbΔC-X showed suppression of potassium channel genes and proteins leading to a decrease in repolarization reserve. Atrial Kcnk2, Kcnd2, Kcnd3, Kcnj3, and Kcnh2 mRNA and their corresponding protein products were significantly suppressed in CREM-IbΔC-X overexpression mice and inward rectifier potassium current was decreased [99, 111]. Taken together, these studies suggest the mechanisms of AF in CREM-IbΔC-X transgenic mice are alteration of calcium handling, potassium channel downregulation, and arrhythmogenic substrate formation.

SHOX2

Shox2 has been shown to play an important role in SA node function, particularly in conjunction with Nkx2.5 [14]. Specifically, Shox2 heterozygous mutation resulted in abnormally increased expression of Cx43 and Cx40 in the SAN of mouse embryos [14]. Concurrently, SHOX2 also plays a role in familial and early onset AF. SHOX2 expression levels in the right atrial appendage of AF patients were found to be significantly reduced compared to those with sinus rhythm and several variants within SHOX2 have been associated with AF. These mutations include G81E, H283Q, G77D, A293 =, R194X, and 3′ UTR variant (C.*28 T >C) with the R194X eliminating transcriptional activity and H283Q and G77D mutations reducing transcriptional activity [49, 72]. The H283Q and G77D mutations also result in bradycardia, illustrating the dual role of SHOX2 in AF and SA node function [50]. Notably, Shox2 is under the regulation of both Nkx2.5 and Tbx5 in the developing heart while Pitx2 directly binds Shox2 to suppress sinoatrial node programming in the left atrium [104, 139]. However, further study is still required to elucidate the potential interplay between SHOX2-dependent pathways and other fibrillogenic signaling pathways.

ZFHX3

SNPs rs2106261 and rs7193343 in the intronic region of transcription factor ZFHX3 (zinc finger homeobox 3 and also known as AT-motif binding factor 1, ATBF1) showed AF association by GWAS [8, 34, 46]. ZFHX3 was also associated with left atrium dilatation and AF recurrence after treatment by ablation [54]. In HL-1 cells, siRNA knockdown of ZFHX3 reduced APD50 and APD20 but did not affect APD90 [63].

While the mechanism for ZFHX3 regulation of cardiac rhythm remains poorly defined, studies in HL-1 suggest its loss results in increased SR calcium leak and potassium currents. HL-1 cells with reduced ZFHX3 exhibited increased SER-CA2A protein expression and activity along with increased RyR2 expression, along with increased SR calcium content, higher amplitude calcium transients, and increased SR leak compared to control cells. Furthermore, reduction of ZFHX3 also increased expression of potassium channel proteins Kv1.4, Kv1.5, and Kir3.4 and increased IKAch, IKur, and Ito. ZFHX3 may also play a role in substrate formation as tachypacing in HL-1 cells led to decreased ZFHX3 and increased STAT3 activation, which may mediate AF structural changes [60]. As with other transcription factors, ZFHX3 likely is a part of a gene regulatory network. ZFHX3 and PITX2c positively regulate each other and both modulate expression of TBX5 and NKX2.5 as knockdown of either by siRNA resulted in increased expression of both Tbx5 and Nkx2.5 [53].

ETV1

ETV1 is highly expressed in the fast conduction system and its knockout in mice resulted in prolonged atrial-His and His-ventricular conduction time [116]. E twenty-six variant 1 (ETV1) was found to upregulate in patients with permanent AF and a mouse model of Etv1 overexpression resulted in loss of p-waves accompanied by structural remodeling leading to atrial enlargement and dilatation. Etv1 may be critical in the modulation of the Angiotensin II-mediated signaling pathway important for atrial remodeling. Indeed, cardiomyocyte-specific deletion of Etv1 was protective against angiotensin II-induced atrial remodeling. In particular, loss of Etv1 prevented angiotensin II-induced decrease in calcium handling genes Ryr2 and Atp2a2 along with potassium channel genes Kcnh2 and Kcnk3 [108, 109]. As 178 potential ETV1 target genes have been identified and ETV1 shares binding sites with TBX5, NKX2.5, and GATA4 in atrial myocytes, future studies focused on elucidating the interplay between ETV1 and other AF transcription factors to regulate calcium handling are needed [108].

PRRX1

Meta-analysis identified an AF susceptibility locus within PRRX1, encoding for a homeodomain transcription factor [34, 51]. Furthermore, an AF-associated single nucleotide polymorphism, rs593479 (1q24), was validated in a Japanese population [74]. Prrx1 knockouts from stem cell–derived cardiomyocytes and zebrafish had shortened Aps [130]. Shortened APs may increase the propensity for development of AF by allowing for shorter refractory periods; however, the mechanism of PRRX1-mediated AP alterations is yet to be defined.

The transcription factors discussed above (summarized in Table 1) are not a complete list of genes contributing to AF. There are many other factors involved in atrial remodeling which on its own can contribute to a fibrillogenic substrate [81]. In addition, genes important in remodeling can also affect expression of ion channels and calcium handling genes. For example, NFAT can cause downregulation of the alpha subunit of Cav1.2, which can lead to shortening of the AP and shorten the refractory period [106, 150]. NF-κB can downregulate expression of NAV1.5, which can slow myocardial conduction [114].

Table 1.

Summary of transcription factors’ effects on their downstream targets. Subscripts a and v indicate location of recorded expression, in the atria or ventricle, respectively. Targets are bolded if they have been recorded with an enhancer assay

	PITX2			TBX5			GATA4
	mRNA	Protein	Function	mRNA	Protein	Function	mRNA	Protein	Function
Ryanodine receptor	↓ Ryr2a [90, 127]	↓ RYR2a [90]		↑ Ryr2a [28, 90]	↑ RYR2a [90]		↓ RYR2 [68]***
SR Ca2+ATPase	↓ Atp2a2a [90, 127]	↓ ATP2A2a [90]		↑ Atp2a2a [28, 68, 90]	↑ ATP2A2a [28, 90]	↓ SR Ca2+ uptake in knockout mice [28]	↓ ATP2A2 [68]***	↓ ATP2A2a [68]	↑ SR Ca2+ uptake in Gata4 haplosufficient atrial myocytes [68]
Ca2+ channels	↓ Cacna1da ↓ Cacna2d2a [127]					↑ ICa,L in knockout atrial myocytes [28]			↓ ICa,L in AdGata4 neonatal myocytes [42]
Na+ channel	↓ Scn5aa [90]	↓ SCN5Aa [90]		↑ Scn5aa [90]	↑ SCN5Aa [90]
Gap junctions	↓ Gja1a [90, 127]	↓ GJA1a [90]		↑ Gja1a ↑Gja5a [90]	↑ GJA1a [90]	Slowed myocardial conduction in knockout [90]
K+ (Inward rectifying)	↓ Kcnj11a [127]			↑ Kcnj2a ↑ Kcnj8a ↑ Kcnj11a [90] ↑ Kcnj3a ↑ Kcnj5a [68, 90] ↑ Kcnk3a (outward rectifying) [90]
K+ (Transient outward)				↑ Kcna7a ↑ Kcnd2a ↑ Kcnd3a [90]
K+ (Ca2+ activated)	↓ Kcnn3a [127]
K+ (Delayed rectifying)	↑ Kcnq1 ↑ Kcne1 [105]*		↓ Iks in knockout HL-1 cells[105]	↑ Kcng2a ↑ Kcnv2a [90] ↑ Kcnh2a [68]
Na+/Ca2+ exchanger					↓ SLC8A1a [28]	↑ NCX in knockout atrial myocytes [28]
PITX2				↑ Pitx2a [90]
TBX5 GATA4 NKX2.5	↓ Nkx2.5a [24]
ZFHX3	↓ Zfhx3a [127]
microRNA	↑ miR-17–92 ↑ miR-106b-25 ↓ miR-335 ↓ miR-423 [138]** ↓ miR-1a[24]			↓ miR-182–5p [47]
Substrate level changes	Knockout results in shortened AP durations[65] Knockout results in disrupted intercalated discs[127]			Knockout results in AP prolongation and slowed conduction velocity[90]
Trigger level changes				Knockout results in EADs, and DADs[68, 90]
	NKX2.5			ZFHX3			ETV1
	mRNA	Protein	Function	mRNA	Protein	Function	mRNA	Protein	Function
Ryanodine receptor		↓ RYR2a [22]			↓ RYR2a [63]		↓ Ryr2a [109]
SR Ca2+ATPase		↑ ATP2A2a [22]			↓ ATP2A2a [63]		↓ Atp2a2a [109]
Ca2+ channels	↑ Cacna1c [42]*	↓ CAV1.2v [141]	↓ ICa,L in AdCsx/Nkx2.5 neonatal myocytes 141]
Na+ channel	↑ Scn5a [42]*						↑ Scn5aa [115]
Gap junctions							↑ Gja5a [109]
K+ (Inward rectifying)					↓ KIR3.4a [63]
K+ (Transient outward)					↓ KV1.4a [63]		↓ Kcnk3a [109] (outward rectifying channel)
K+ (Delayed rectifying)	↑ Kcnh2 [42]**		↓ IKr in transcriptionally impaired NKX2.5 HL-1 cells[42]				↓ Kcnh2a [109]
Na+/Ca2+ exchanger
PITX2 TBX5 GATA4		↑ GATA4v [141]
NKX2.5							↑ Nkx2.5a [115]	↑ NKX2.5a [115]
microRNA
Substrate level changes	Overexpression results in decreased APD50[141] Knockout results in impaired atrioventricular conduction[57]			Knockdown results in shortened APD20 and APD50[63]			Overexpression results in enlarged and dilated atria[109]
Trigger level changes
	LKB1			CREM-IbΔC-X
	mRNA	Protein	Function	mRNA	Protein	Function
Ryanodine receptor
SR Ca2+ATPase
Ca2+ channels	↑ Cacna1ha [64]			↑ Cacna1ca [122]
Na+ channel	↑ Scn5aa [64]	↑ SCN5Aa [64]		↓ Scn5aa [122]
Gap junctions	↑ Gja1a ↑ Gja5a [64]			↓ Gja5a [122]
K+ (Inward rectifying)	↑ Kcnj8a ↑ Kcnj14a ↑ Kcnk3a [64]			↓ Kcnj3a ↓ Kcnh2a [99] ↓ Kcnk2a [111]	↓ KIR2.1a ↓ KIR3.a [99]	↓ IK1 [99]
K+ (Transient Outward)				↓ Kcnd2a ↓ Kcnd3a [99]
K+ (Delayed rectifying)	↑ Kcnq1a [64]			↑ Kcne1a [122]
Na+/Ca2+ exchanger PITX2 TBX5 GATA4 NKX2.5 microRNA
Substrate level changes				Overexpression results in atrial dilatation and prolonged action potential[122]
Trigger level changes				Overexpression results in increased calcium sparks [66]

^*Assay done in CHO cell line.

^**Expression level was recorded in whole heart.

^***Enhancer assay done in HEK293T cells

MicroRNAs

MiRNAs are a class of small noncoding RNAs that perform post-transcriptional repression of target genes of more than 60% of all protein-coding genes [41]. There is increasing evidence that miRNAs are regulators of cardiac rhythm and that alterations in miRNA expression can result in AF. Altered miRNA expression patterns in the heart and circulating in the blood have been observed in both patients with AF and animal models [80]. Furthermore, studies altering individual miRNAs have revealed several mechanisms by which miRNAs may regulate cardiac rhythm. MiRNAs play a role in the regulation of potassium channels. For example, inhibition of miR-26 increased inward rectifier current (I_K1) and increased incidence of AF in mice [79]. Furthermore, miRNAs can regulate calcium handling, as shown by canine and murine miR-328 overexpression models which show decreased L-type calcium current and increased AF susceptibility [76]. Additionally, alteration in miRNA expression can also induce structural remodeling, leading to atrial fibrosis, and thus providing an arrhythmogenic substrate [80, 132]. Currently, there are still many miRNAs identified by microarray analysis (comparing AF patients to those with sinus rhythm) for which we do not yet understand the mechanism of action.

MiRNAs can serve as an intermediate player through which transcription factors may regulate ion channels and calcium handling genes as well as other transcription factors. For example, PITX2 transactivates miR-17–92 and miR-106b-25 and downregulates miR-335 and miR-423. MiR-17–92 then represses Shox2 and Tbx3 [138]. miR-17–92- and miR-106b-25-deficient mice also exhibit arrhythmias following rapid pacing [138]. Pitx2 also downregulates miR-1, which can regulate I_k1 amplitude [24]. PITX2 is not alone in its regulation of miRNAs as TBX5 also regulates a number of miRNAs including downregulation of miR-182–5p, which was found to be involved in myocardial differentiation [47]. On the other hand, overexpression of miR-182 led to an increase in CACNB2.1, but a decrease in CACNB2.2, CAC-NB4b, and CACNA2d1 subunits [47]. Conversely, miRNAs may also regulate transcription factors. For example, miR-200c directly downregulates TBX5 and GATA4 and miR-21 negatively regulates PITX2c expression [101, 128].

There has been a significant number of microRNAs found associated with calcium handling channels and 11 miRNAs have been consistently been associated with AF [86]. Increased miR-208b has been linked with downregulation of SERCA2a [86]. miR-93 negatively regulates miRRYR at the 3′-UTR site while miR-106b-25 deficiency causes RYR2 upregulation in atria resulting in increased spontaneous calcium waves [23]. miR-328 targets CACNA1c and CACNB1 and overexpressed miR-200 decreases CACNA1c protein levels [76, 101].

Thus, there are many possible mechanisms whereby microRNAs could potentially contribute to altered electrical activity in atrial cardiomyocytes (Summarized in Table 2). This provides an additional level of complexity beyond channel level and transcription factor level alterations to drive atrial fibrillation. In the future, it is possible that small molecule inhibitors of specific microRNAs could provide personalized therapeutic approaches to prevent AF in some patients.

Table 2.

Summary of microRNA effects on their downstream targets. Subscripts a and v indicate location of recorded expression, in the atria or ventricle, respectively. Targets are bolded if they have been recorded with an enhancer assay

	miR-26			miR-1			miR-182
	mRNA	Protein	Function	mRNA	Protein	Function	mRNA	Protein	Function
Ca2+ channels							↑ Cacnb2.1 ↓ Cacnb2.2 ↓ Cacnb4b ↓ Cacna2d1 [47]
Gap junctions				↓ Gja1a [24]
K+ (Inward rectifying)	↓ Kcnj2v [79]	↓ KIR2.1v [79]	↓ Ik1 in miR-26 transfected rat ventricular cells[79]	↓ Kcnj2a ↓ Kcnj4a [24]
TBX5 GATA4
	miR-200			miR-328			miR-106b-25
	mRNA	Protein	Function	mRNA	Protein	Function	mRNA	Protein	Function
Ca2+ channels	↓ CACNA1C [101]*	↓ CACNA1C [101]*		↓ Cacna1ca [76] ↓ Cacna1ba [76]	↓ CAV1.2a [76] ↓ CAVβ1a [76]
Ryanodine receptor								↓ RYR2a [23]	↑ SR Ca2+ increased leak in knockout atrial myocytes [23]
K+ (Inward rectifying)	↓ KCNJ2 [101]*
TBX5	↓ TBX5 [101]*
GATA4	↓ GATA4 [101]*

^*Recorded in human pluripotent stem cell–derived cardiomyocytes

Long noncoding RNAs

A number of long noncoding RNAs have been found to be important in AF and may regulate expression of calcium handling proteins. KCNQ1OT1 was found upregulated in an angiotensin II-induced AF mouse model. Further studies of the lncRNA showed miR-384b involvement where miR-384b simultaneously bound both KCNQ1OT1 and CACNA1C. Thus, KCNQ1OT1 functioned to upregulate CACNA1C by competitively binding to miR-384b. Furthermore, transcription factor YY1 overexpression led to the upregulation of KCNQ1OT1 and CACNA1C, implying a relationship between the three [117]. LncRNA TCONS_00075467 was downregulated in a rabbit AF model and selected for further study. This downregulation also led to upregulation of miR-328, which then inhibited CACNA1C protein expression. RNAi against TCONS_0075467 was found to inhibit CACNA1C and could be relieved with an miR-328 inhibitor [73]. A third AF relevant lnRNA is PANCR, which is close to the 4q25 locus. PANCR was hypothesized to possibly stabilize PITX2c mRNA [45]. Additionally, RACER, a TBX5-dependent enhancer-associated lncRNA, was found required for RYR expression [147].

Conclusions and future studies

In this review, we have outlined some of the changes which are known to contribute to a genetic predisposition of AF. What we have learned so far has helped us understand why certain individuals and families are at risk of developing AF, and this has tremendous clinical implications. It will help with early detection and improve our ability to counsel high-risk individuals. Nevertheless, although many of the genetic factors are now defined, our understanding of the downstream implications remains poorly understood. Specifically, while we understand many of the changes at the level of RNA and protein, in most cases, we do not yet fully understand how these changes affect cellular electrophysiology and calcium handling or how these functional alterations at the level of cells translate to development of atrial fibrillation at the level of the myocardium. Furthermore, we do not understand the complex interplay between multiple genetically driven changes that vary from patient-to-patient. Research to uncover these biophysical mechanisms and complex genetic regulatory networks will be critical to developing novel personalized AF therapies in the near future.

A “threefold model” that links transcription factor genetic variants to AF predisposition was recently proposed [133]. In this model, genetic variants of AF transcription factors can result in altered expression of effector genes directly, change in other AF relevant transcription factors, or interaction as cofactors with other transcription factors. Here, we add further detail to the threefold model and propose that transcriptional regulation of ion channels and calcium handling genes can be multi-fold, and ultimately affects the probability of ectopic triggers and altered substrate in atrial myocytes. These probabilities are determined by the functional balance between calcium efflux and influx and in parallel depolarizing and repolarizing membrane forces. Over the years, many functional studies of AF relevant transcription factors (e.g., PITX2, TBX5, GATA4, NKX2.5, SHOX2, ZFHX3, ETV1, PRRX1) have shown that the regulation of calcium handling often occurs through the alteration of multiple components (Fig. 1). For example, Tbx5 loss results in concurrent decrease of SERCA activity and increased NCX and I_Ca flux. This demonstrates the need to perform functional studies in conjunction with others to determine the unique functional consequences of AF relevant transcription factor variants.

Fig. 1.

Genetic regulation of ion channels and transporters can lead to altered conduction, membrane potential, and cellular calcium handling which contribute to the formation of a fibrillogenic substrate and triggered activity

Through many studies, we have learned that transcription factor function and anatomical localization are dynamic through development, mature heart, and disease states. For example, Rommel et al. highlight with ETV1, transcription factors can orchestrate chamber, time, and disease-specific responses via its 178 potential target genes [108]. We and others show that in the case of AF, there is a complex transcriptional network which contains regulatory loops between multiple transcription factors, unidirectional regulation by transcription factors on effector ion channels and calcium handling genes as well as bidirectional regulation with calcium handling genes such as sodium-calcium exchanger regulating transcription factor expression. These regulatory relationships illustrate the complexity of AF as a disease and the ways in which alteration of one transcription factor may have a multitude of effects beyond its direct effectors. Future studies of the electrophysiological alterations resulting from changes in transcriptome and epigenome through the progression of AF could help elucidate disease progression.

We further propose that AF relevant transcription factor variant alteration of effector genes may occur through a multimodal approach including changes in enhancer site binding, microRNA-mediated effects, and long non-coding RNA-mediated changes. For example, microRNA miR-17–92 is transactivated by Pitx2 but can in turn repress Shox2 and Tbx3. Such studies suggest that several different genetic strategies can be employed solo or in conjunction with each other to address AF predisposition. Future functional studies of microRNA and long non-coding RNA-mediated changes in calcium handling and ion channels are needed to continue to elucidate regulatory loops within the AF relevant transcriptional network.