Distinct Occurrence of Proarrhythmic Afterdepolarizations in Atrial Versus Ventricular Cardiomyocytes: Implications for Translational Research on Atrial Arrhythmia

Background: Principal mechanisms of arrhythmia have been derived from ventricular but not atrial cardiomyocytes of animal models despite higher prevalence of atrial arrhythmia (e.g., atrial fibrillation). Due to significant ultrastructural and functional differences, a simple transfer of ventricular proneness toward arrhythmia to atrial arrhythmia is critical. The use of murine models in arrhythmia research is widespread, despite known translational limitations. We here directly compare atrial and ventricular mechanisms of arrhythmia to identify critical differences that should be considered in murine models for development of antiarrhythmic strategies for atrial arrhythmia. Methods and Results: Isolated murine atrial and ventricular myocytes were analyzed by wide field microscopy and subjected to a proarrhythmic protocol during patch-clamp experiments. As expected, the spindle shaped atrial myocytes showed decreased cell area and membrane capacitance compared to the rectangular shaped ventricular myocytes. Though delayed afterdepolarizations (DADs) could be evoked in a similar fraction of both cell types (80% of cells each), these led significantly more often to the occurrence of spontaneous action potentials (sAPs) in ventricular myocytes. Interestingly, numerous early afterdepolarizations (EADs) were observed in the majority of ventricular myocytes, but there was no EAD in any atrial myocyte (EADs per cell; atrial myocytes: 0 ± 0; n = 25/12 animals; ventricular myocytes: 1.5 [0–43]; n = 20/12 animals; p < 0.05). At the same time, the action potential duration to 90% decay (APD90) was unaltered and the APD50 even increased in atrial versus ventricular myocytes. However, the depolarizing L-type Ca2+ current (ICa) and Na+/Ca2+-exchanger inward current (INCX) were significantly smaller in atrial versus ventricular myocytes. Conclusion: In mice, atrial myocytes exhibit a substantially distinct occurrence of proarrhythmic afterdepolarizations compared to ventricular myocytes, since they are in a similar manner susceptible to DADs but interestingly seem to be protected against EADs and show less sAPs. Key factors in the generation of EADs like ICa and INCX were significantly reduced in atrial versus ventricular myocytes, which may offer a mechanistic explanation for the observed protection against EADs. These findings may be of relevance for current studies on atrial level in murine models to develop targeted strategies for the treatment of atrial arrhythmia.


INTRODUCTION
Triggered arrhythmia originates from early (EADs) and delayed (DADs) afterdepolarizations on the cellular level (Damiano and Rosen, 1984;Levine et al., 1985;January and Riddle, 1989). Underlying molecular mechanisms and principle concepts of proarrhythmic mechanisms have been well defined in ventricular cardiomyocytes in various experimental models (Pogwizd et al., 1999(Pogwizd et al., , 2001Sipido et al., 2000;Pott et al., 2012). In the past two decades, murine models have been increasingly investigated in arrhythmia research due to their accessibility for genetic manipulation despite significant differences between mouse and human physiology (for review, see Nerbonne, 2014).
Yet atrial -or in clinical terms supraventricular -arrhythmias are much more prevalent than ventricular arrhythmias and attract great attention in clinical and translational research, especially in the case of atrial fibrillation (Kirchhof et al., 2016). However, the electrophysiological properties and underlying molecular mechanisms in atrial cardiomyocytes are considerably less defined as compared to ventricular cardiomyocytes, especially in smaller rodents like mice and rats (Walden et al., 2009). Nevertheless, murine models are frequently used to investigate proarrhythmic substrates or to evaluate potential antiarrhythmic strategies to counter atrial arrhythmia even on multicellular or whole-heart level (Bao et al., 2016;Syeda et al., 2016;Wan et al., 2016;Chang et al., 2017;Wang et al., 2017). Yet, there is a lack of studies, which directly compare proarrhythmic substrates and their molecular basis in atrial versus ventricular cardiomyocytes of murine models, which appears unexpected in the face of the high clinical and translational relevance of atrial arrhythmia and the widespread use of murine models in the research of atrial arrhythmia.
One might suppose that ventricular proneness toward proarrhythmic afterdepolarizations can simply be applied on the atrial level. However, this requires a consistent functional and structural analogy between atrial and ventricular cardiomyocytes, which is not the case. Previous work demonstrated substantial differences regarding the morphology and microarchitecture of atrial and ventricular cardiomyocytes (Bossen et al., 1981;Blatter et al., 2003;Brette et al., 2005;Walden et al., 2009;Bootman et al., 2011;Frisk et al., 2014). Thus, atrial myocytes rather resemble smooth-muscle myocytes due to the diminished sarcomere structure and a fusiform shape. Also, t-tubules as a morphologic feature of crucial functional relevance for excitation-contraction coupling in ventricular myocytes are substantially reduced or even absent in a major fraction of atrial myocytes (Brette et al., 2005;Frisk et al., 2014), though this is species-dependent (Yue et al., 2017). This again leads to a considerably altered excitation pattern consisting of a slowly propagating centripetal wave in atrial myocytes as compared to the rapid simultaneous and homogenous excitation pattern in ventricular myocytes (Blatter et al., 2003). In addition, there are considerable differences in the expression of functional key proteins participating in excitationcontraction coupling Lüss et al., 1999Lüss et al., , 2000Walden et al., 2009;Nerbonne, 2014). Consequently, these findings suggest significant structural and functional differences and absence of a perfect analogy of both cell types. Moreover, these atrioventricular differences depend on the investigated species. For instance, the t-tubule system in atrial myocytes is reportedly poorly developed in rats, cats, rabbits, and guinea pigs, but well developed in mice or larger mammals like humans (Yue et al., 2017). Likewise, atrioventricular differences in the expression of critical Ca 2+ handling proteins are speciesdependent (Lüss et al., 1999).
The diverse ultrastructure and ion channel expression are reflected in the differing action potential kinetics between atrial and ventricular myocytes in mammals. In general, the resting membrane potential of atrial myocytes is more depolarized as compared to ventricular myocytes, which may be a result of a decreased I K1 function in atrial myocytes (Cordeiro et al., 2015). Both, the upstroke velocity and the amplitude of the atrial action potential are decreased in comparison to ventricular myocytes , which is promoted by the depolarized resting potential in atrial myocytes that leads to a slower and more incomplete recovery from inactivation of the cardiac sodium channels. Early repolarization in phase 1 is enhanced in atrial myocytes due to the synergistic activation of the transient K + currents I to(f) and atrial-specific I Kur in atrial myocytes (Nerbonne and Kass, 2005). The voltage-level of the action potential plateau in phase 2 is more hyperpolarized in atrial myocytes due to a higher cumulative driving force of repolarizing K v populations. Differences in the action potential duration (APD) seem to be species-dependent, but in higher mammals, the atrial action potential is abbreviated (Burashnikov et al., 2014).
When compared to human action potentials, typical murine ventricular action potentials are significantly abbreviated and exhibit a short plateau phase that starts at more negative voltage levels (Ahrens-Nicklas and Christini, 2009). A large I to in murine myocytes outweighs the depolarizing I Ca and therefore reduces the plateau voltage level, whereas the human I to is less dominant. Taken together, electrophysiological parameters differ between atrial and ventricular myocytes and between different species, as reflected in the characteristic referring action potential shapes.
To account for the significant atrioventricular differences, some recent studies applied computational approaches to mimic atrial electrophysiology to investigate mechanisms of atrial arrhythmia (Zhang et al., 2005;Maleckar et al., 2009;Grandi et al., 2011;Ni et al., 2017). However, despite great advancement during the past decade, computational models are based on assumptions and thus still differ from the genuine atrial electrophysiology, which might particularly apply to complex and stochastic parameters like proarrhythmic afterdepolarizations or atrial fibrillation, that result from a multitude of underlying processes and interactions.
Since murine models are frequently used in translational research on atrial arrhythmia, but data on the manifestation of proarrhythmic afterdepolarizations in atrial myocytes and their differences to ventricular myocytes is sparse, we investigated afterdepolarizations and their molecular bases in atrial and ventricular myocytes in a murine model. The obtained findings might carry several implications for contemporary and future studies investigating atrial arrhythmia in murine models.

Genetic Background of the Murine Model
Investigated wild-type mice were on a combined genetic background (CD-1 and C57BL/6) as previously described (Bögeholz et al., 2015(Bögeholz et al., , 2016 at the age of 7.5-12.5 weeks. The applied experimental methods confirmed to the instructions of Directive 2010/63/EU of the European Parliament on the protection of animals used for scientific purposes and were approved by the local authorities (Landesamt für Natur, Umwelt und Verbraucherschutz NRW; permission number: 84-02.05.50.17.002).

Two-Dimensional Planimetric Cell Area
Quantification of the two-dimensional planimetric cell area was performed in suspensions of non-fixed atrial and ventricular myocytes using a Photometrics CoolSNAP HQ2 camera on a Nikon eclipse TI-E microscopy system. Quantification was performed with the NIS-Elements AR v. 4.13.05 (Nikon) imaging software.

Cellular Electrophysiology
For whole-cell patch-clamp experiments, cells were transferred to the stage of an inverted microscope (IX 50, Olympus) and investigated using an EPC 800 amplifier connected to an InstruTECH ITC-18 interface and a MS computer with the PatchMaster v2x53 software (all HEKA, Bellmore, NY, United States) (Bögeholz et al., 2015(Bögeholz et al., , 2016. The tip resistance of patch pipettes (GB150TF-8P, Science Products, Hofheim, Germany) was 4 ± 1 M .

Statistical Analysis
Numerical data are given as means ± standard error of the mean (SEM), or in case of non-normal distributed data, as the median value with the interquartile range (IQR) between the 25th and 75th percentile (Q3-Q1). In box plot figures, the boxes represent the data between 25th and 75th percentile, whiskers data between the 10th and 90th percentile, horizontal lines give the median value, whereas the mean value is shown by the small square. For direct comparison between atrial and ventricular myocytes, the unpaired two-tailed Student's t-test or the Mann-Whitney Rank Sum test was performed, where appropriate. Repeated measurements were analyzed using a two-way repeated measures ANOVA test followed by a Holm-Sidak post hoc test. Proportions were compared with a Fisher's exact test. P-values < 0.05 were considered statistical significant.

Reduced Translation of DADs Into Spontaneous Action Potentials in Atrial Myocytes
The susceptibility toward proarrhythmic afterdepolarizations was evaluated in atrial and ventricular myocytes using the same provocation protocol for both cell types, comprised of abrupt changes between various pacing frequencies and isoproterenol application as established previously with minor modifications (Bögeholz et al., 2015(Bögeholz et al., , 2016. In detail, the protocol consisted of 20 stimuli per cycle at 5 Hz; 2 Hz, 1 Hz; 0.5 Hz; 0.25 Hz, and 0.125 Hz followed by 3 min rest. Finally, 1 µM isoproterenol was added during steady-state 1 Hz pacing (1 min.) followed by 20 stimuli at 5 Hz, 2 Hz, 1 Hz, and 0.5 Hz resulting in a total protocol duration of 10 min and 28 s. Delayed afterdepolarizations (DADs) were defined as low-amplitude depolarizations exceeding the diastolic membrane potential for more than 2 mV ( Figure 3A). Spontaneous action potentials (sAPs) defined as fully developed action potentials arising from a DAD and exceeding −10 mV threshold, were counted ( Figure 3A).
The fraction of cells exhibiting DADs was similar in both cell types (fraction of cells with DADs in %; atrial myocytes: 80.0; n = 25/12 cells/animals; ventricular myocytes: 80.0; n = 20/12 cells/animals; p > 0.05) (Figure 3B). sAPs occurred in both cell types, but the fraction of cells with sAPs was significantly higher in ventricular myocytes (fraction of cells with spontaneous AP in %; atrial myocytes: 12.0; n = 25/12 cells/animals; ventricular myocytes: 70.0; n = 20/12 cells/animals; p < 0.05) (Figure 3C). The median number of sAPs per cell was significantly decreased in atrial versus ventricular myocytes ( Figure 3D). FIGURE 1 | Basic differences in morphometry between atrial and ventricular myocytes. (A) Exemplary microscopic images demonstrate morphological differences between murine atrial and ventricular myocytes. Atrial myocytes exhibit a smaller 2D cell area compared to ventricular myocytes (B) and consistently a reduced electrical cell capacitance (C). * p < 0.05 atrial vs. ventricular.

Absence of Early Afterdepolarizations in Atrial Myocytes
Throughout the above stated protocol, the occurrence of early afterdepolarizations (EADs) was quantified in atrial and ventricular myocytes. EADs were defined as re-upstrokes of the membrane potential during repolarization of a stimulated action potential, if the membrane potential of the action potential plateau shortly falls of and then exceeds −40 mV threshold ( Figure 4A). EADs showed various shapes as illustrated in Figure 4A.

Reduced L-Type Ca 2+ Current in Murine Atrial Myocytes
To further investigate the molecular mechanisms of EAD and DAD, the depolarizing L-type Ca 2+ current (I Ca ) was measured in atrial and ventricular myocytes ( Figure 5A). To account for differences in cell size, peak I Ca was normalized to the cell capacitance as a measure of cell surface area. I Ca peak amplitude was significantly reduced in atrial myocytes as compared to ventricular myocytes (peak amplitude of I Ca in response to a square wave pulse to +10 mV in pA/pF; atrial myocytes: 2.6 [2.0-3.1]; n = 13/3 cells/animals; ventricular myocytes: 8.2 [5.7-11.0]; n = 20/4 cells/animals; p < 0.05) (Figures 5B,C). Inactivation of I Ca was estimated by fitting a single exponential decay function to I Ca tracings in response to a voltage step to +10 mV. I Ca inactivation was unaltered between both cell types (τ in ms; atrial myocytes: 27.7 ± 2.2; n = 13/3 cells/animals; ventricular myocytes: 26.2 ± 1.4; n = 20/4 cells/animals; p > 0.05) ( Figure 5D).

Smaller NCX Inward Current but Increased SR Ca 2+ Load in Atrial Versus Ventricular Myocytes
The depolarizing NCX inward current (I NCX ; extrusion of 1 Ca 2+ from the cell in exchange for 3 Na + ) is thought to promote the occurrence of EAD and DAD. Thus, I NCX was investigated in atrial and ventricular myocytes. I NCX peak amplitude was significantly decreased in atrial compared to ventricular myocytes (I NCX amplitude in pA/pF; FIGURE 2 | Distinct action potential kinetics in atrial versus ventricular myocytes. (A) Representative action potentials in atrial (left) and ventricular cardiomyocytes without (mid) and with (right) an EAD. Atrial myocytes exhibited a depolarized resting membrane potential (RMP) (B), a decreased maximal action potential (AP) upstroke velocity (V max ) (C), a decreased AP amplitude (D), and prolonged action potential duration to 50% decay (APD 50 ) (E). (F) Though a fraction of ventricular cells exhibited incessant EADs leading to AP prolongation (raw values within the dashed box), the average APD 90 was unaltered between both cell types. * p < 0.05 atrial vs. ventricular; n.s., non-significant.  (Figures 6A,B). The area above the curve (AAC) of the inward current up to baseline ( Figure 6A) is considered as a measure for the sarcoplasmic Ca 2+ content. AAC was larger in atrial than ventricular myocytes (AAC of I NCX in pA * s; atrial myocytes: 293 [139-574]; n = 11/3 cells/animals; ventricular myocytes: 118.0 [98-232]; n = 14/3 cells/animals; p < 0.05) (Figure 6C).

DISCUSSION
The use of murine models is widespread in experimental research, due to facilitated accessibility to genetic manipulation, short reproduction cycles and comparable low housing and breeding costs. Hence, many experimental findings that describe molecular mechanisms of arrhythmia, originate from single cell experiments in murine models. However, the translational value of findings derived from murine models is often called into question (Kaese and Verheule, 2012;Bögeholz et al., 2014;Nerbonne, 2014), since major electrophysiological properties differ considerably between humans and mice. Furthermore, the majority of proarrhythmic mechanisms were derived from ventricular murine cardiomyocytes, despite the higher prevalence of atrial arrhythmia and the enormous clinical attention toward supraventricular arrhythmia, such as atrial fibrillation. This might be due to the robustness and ease of isolation of ventricular myocytes facilitating experimental approaches to investigate the cellular electrophysiology. However, due to ultrastructural and functional differences between ventricular and atrial cardiomyocytes the proneness toward proarrhythmic afterdepolarizations might differ between both cell types. Consequently, in face of these atrioventricular differences, several studies investigated murine models, particularly on the atrial level, using isolated cardiomyocytes (Lemoine et al., 2011;Faggioni et al., 2014), multicellular preparations (Bao et al., 2016;Syeda et al., 2016) and whole-heart approaches (Wan et al., 2016;Chang et al., 2017;Wang et al., 2017) to evaluate translational concepts in the generation of atrial fibrillation. We here present direct evidence that the manifestation of proarrhythmia in the form of the occurrence of EAD and DAD considerably differs between atrial and ventricular murine myocytes, which carries several potential implications for arrhythmia research in murine models, especially on the atrial level.

Protection Against EADs in Atrial Myocytes
In line with previous studies on rats (Walden et al., 2009), murine atrial myocytes were significantly smaller and exhibited a fusiform shape with reduced cross-striation as compared to the larger and rectangular ventricular myocytes. The applied provocation protocol in our study was sufficient to elicit frequent DADs and several sAPs in both cell types. However, though the majority of ventricular myocytes exhibited frequent EADs, no single EAD occurred in any investigated atrial myocyte. Consistently, other studies found that genetic or pharmacological interventions that prolong the APD are required to elicit EADs on atrial level in murine models (Li et al., 2009;Qin et al., 2012). However, these findings were not compared to ventricular myocytes. The observed protection against EADs in this study is not a result of an abbreviated action potential, because APD 90 only tended to be increased in ventricular myocytes due to a fraction of cells that exhibited EADs (Figure 2F), but overall, the APD 90 was not statistically different between both cell types. APD 50 was prolonged in atrial myocytes, but this likely does not promote EADs, since EADs typically occur at a voltage below the level reached at APD 50 . The findings with regard to the APDs are in good accordance with previous findings from our group when comparing data from FVB/N wild-type controls, published recently in two separate studies, demonstrating APD 50 values of 8.75 ± 0.8 ms and 4.6 ± 0.5 ms and APD 90 values of 38.7 ± 2.7 ms and 32.5 ± 4.4 ms for atrial (Seidl et al., 2017) and ventricular  myocytes, respectively. Of note, these findings on the action potential kinetics of atrial and ventricular myocytes do not apply to higher mammals or even human action potential kinetics, since human atrial cardiomyocytes exhibit a substantially prolonged action potential plateau (APD 90 ∼ 300 ms) (Christ et al., 2016). Likewise, the human ventricular action potential is considerably prolonged as compared to mouse models (APD 90 ∼ 450-500 ms) (Johnson et al., 2018). These action potential kinetics may allow EADs in human atrial myocytes rather than in murine atrial myocytes. Nonetheless, vigorous provocation conditions like strong inhibition of repolarizing K + channels, resulting in a significant APD prolongation, is required to elicit EADs in human atrial myocytes (Olson et al., 2006). If the APD is prolonged by vigorous pharmacological interventions, EADs can even be elicited in murine atrial trabeculae (Zhou and Liu, 1995). Hence, the protection against EADs in murine atrial myocytes can be overridden by aggressive pharmacological interventions. However, murine atrial myocytes are still less susceptible to EADs as compared to ventricular myocytes, which is of relevance for studies on atrial arrhythmia using murine models.
What is the underlying mechanism of protection against EADs in murine atrial myocytes? Since the classical mechanism in the generation of an EAD consists of a reactivation of I Ca or enhanced NCX-mediated inward current (Volders et al., 1997), the prominent reduction of both, as observed in this study, suggests a plausible explanation for the protection against EADs in atrial myocytes. Beyond that, I Ca and I NCX may be functionally coupled. Our group has shown that the reduced NCX function in murine ventricular myocytes of homo-and heterozygous NCX knockout mouse models is accompanied by a reduction of I Ca (Pott et al., 2007;Bögeholz et al., 2015). In ventricular myocytes, the underlying mechanism seems to consist of an increased subsarcolemmal Ca 2+ concentration due to reduced NCX-mediated Ca 2+ extrusion capacitance, which enhances Ca 2+ -dependent inactivation of I Ca . Thus, the inhibitory interaction between I NCX and I Ca may synergistically protect against the occurrence of EADs. The same mechanism could apply to atrial myocytes, although this needs to be further investigated, since the functional coupling of I NCX and I Ca may be different in atrial myocytes due to a diverse ultrastructure. However, there may be also differences in other depolarizing and repolarizing ion currents that may reduce the proneness toward EADs in atrial myocytes. Recently, Edwards et al. (2014) proposed a role for non-equilibrium sodium current (I Na ) reactivation in the generation of EADs in mice, based on the much faster action potential repolarization velocity in comparison to higher mammals. A role for I Na reactivation is also likely in our study since the median take-off potential of EADs in ventricular myocytes was −44.8 mV, which is below the I Ca activation range. Though we did not directly measure I Na kinetics in our study, the reduced maximal action potential upstroke velocity and the reduced action potential amplitude in atrial vs. ventricular myocytes are both indicators for a reduced Na + channel availability. One possible cause is the relatively depolarized resting membrane potential in atrial myocytes, which will reduce the availability of Na + channels and driving force for I Na in atrial versus ventricular myocytes. Furthermore, Edwards et al. (2014) demonstrated that I Na reactivation enhances with increasing velocity of repolarization. Since atrial myocytes had a rather triangular action potential shape along with increased APD 50 , the reduced repolarization velocity in atrial versus ventricular myocytes may thus additionally contribute to the protection against EADs in murine atrial myocytes. Interestingly, in their study Edwards et al. (2014) also demonstrated in a computational model that ablation of I NCX was able to eliminate EADs and that I Ca was also reactivated to a small extent during EADs secondary to I Na . Thus, the reduction of both currents in atrial myocytes should contribute to the reduced propensity to EADs in atrial mouse cardiomyocytes as observed in our study.

DADs and Spontaneous Action Potentials in Atrial Versus Ventricular Myocytes
Both atrial and ventricular myocytes exhibited frequent DADs on average in 4 out of 5 (80%) cells during the applied provocation protocol, which contains abrupt changes in pacing cycle lengths and additional catecholaminergic stimulation. Hence, the conditions were sufficient to elicit proarrhythmia in both cell types. Since DADs result from spontaneous Ca 2+ release from the sarcoplasmic reticulum (SR), the observed increase of SR Ca 2+ load as assessed by the integrated NCX inward current in atrial compared to ventricular myocytes may promote the occurrence of DADs in atrial myocytes. An increased SR Ca 2+ load of atrial compared to ventricular myocytes in smaller rodents has been demonstrated previously (Walden et al., 2009). However, the occurrence of sAPs as the final proarrhythmic consequence of DADs was significantly reduced in atrial compared to ventricular myocytes. This finding might be surprising at first sight, since the resting membrane potential in atrial myocytes is more depolarized as compared to ventricular myocytes, which should facilitate the translation of DADs in sAPs. However, I NCX peak current as the key translator of spontaneous Ca 2+ release events into sAPs is significantly reduced in atrial myocytes, which could potentially explain for the reduced occurrence of sAPs in atrial myocytes. Next to the reduced I NCX peak current, two potential alternative alterations may protect against the occurrence of afterdepolarizations in atrial myocytes. These could be a reduced availability of I Na and/or increased repolarization reserve. Although these components were not directly measured, the upstroke velocity of the action potential is slowed in atrial versus ventricular myocytes, which can be considered as a proxy measure for reduced I Na as the final translator of I NCX mediated membrane depolarization into a sAP. This may also be a result of the depolarized resting membrane potential in atrial myocytes. In principle, atrial myocytes could also be protected against afterdepolarizations by an increased repolarization reserve potentially mediated via increased I K1 . However, since I K1 is the major hyperpolarizing current in diastole and the resting membrane potential is depolarized in atrial myocytes, the latter possibility seems less likely. Not at least, the substantially diverse ultrastructure of the SR and its altered coupling to the plasma membrane may also contribute to the reduced translation of DADs into sAPs (Frisk et al., 2014).

Translational Perspective on Murine Models in Research on Atrial Arrhythmia
Early and delayed afterdepolarizations are thought to promote the occurrence of human atrial arrhythmia, such as atrial fibrillation (Patterson et al., 2006;Beavers et al., 2013). Mice are commonly used as models for the investigation of basic proarrhythmic mechanisms and therapeutic strategies for treatment of atrial fibrillation. Thus, it is of relevance to characterize cellular proarrhythmic substrates like EADs and DADs and existing specific differences between atrial and ventricular myocytes in the mouse. Our findings show that the occurrence of EADs, DADs, and sAPs significantly differs between atrial and ventricular myocytes. Thus, proarrhythmia in both cell types cannot be considered equal.
Thus, a direct translation of the ventricular proarrhythmia to the atrial level is questionable. Although basic causes for afterdepolarizations (i.e., I NCX and reactivation of I Ca and I Na ) might be similar, the relative quantitative contribution and their functional interaction seem to differ in an environment of a diverse expression of other ion channel populations and the referring cellular microarchitecture. Together, these differences finally result in the distinct probability for the occurrence of afterdepolarizations in atrial and ventricular myocytes. This finding is of relevance for the evaluation of antiarrhythmic strategies targeting atrial arrhythmia at least when using murine models on the single cell and multicellular level. However, it is debatable, how good the current data situation actually is regarding a direct comparison of atrial and ventricular proarrhythmia in other species -particularly in humans, which is complicated by the availability of material to examine. In view of this limited availability -especially of human myocardial tissue -computational modeling of atrial electrophysiology is a promising alternative approach, but its reliability strongly depends on the completeness of the integrated data derived from experimental research (Grandi et al., 2010;Bers and Grandi, 2011;Pandit et al., 2011). There are computational models of human atrial myocytes that simulate physiological and pathophysiological conditions in terms of chronic atrial fibrillation based on ventricular models . Atrial electrophysiology is simulated by adopting several ion current functions in part based on the findings of studies that reported the protein expression but not function (Wang et al., 1996). Furthermore, some parameters regarding functional ion channel properties are derived from single references, since experimental data from human atrial versus ventricular myocytes is sparse.
Occurrence of whole-heart arrhythmias is even more complex than basic cellular mechanisms of arrhythmia, since its initiation depends on the stochastic coincidence of cellular arrhythmic events in relation to the regular heart beat. The occurrence and proarrhythmic significance of afterdepolarizations in isolated cardiomyocytes differs from those of cardiomyocytes that are embedded within the myocardial syncytium. Neighboring cardiomyocytes, coupled via gap junctions, might serve as a current-sink and thus counterbalance proarrhythmic alterations of the membrane potential of a single myocyte, which reduces the probability for a sAP to occur. Thus, depolarization of a critical number of adjacent cardiomyocytes is required to initiate an action potential on the multicellular level (Rubart and Zipes, 2005). Furthermore, the likelihood for initiation of sustained whole-heart arrhythmia does not only depend on the mere number of premature ventricular complexes (PVCs), but also on the timing of the PVC in relation to the prior heartbeat as reflected in the typical short-long-short initiation pattern of torsade-de-pointes tachycardias (Kay et al., 1983) or its short-coupled variant (Leenhardt et al., 1994). Beyond the initiation of wholeheart arrhythmia, its perpetuation is even more complex and underlies a multitude of factors including the anatomical structure, pathophysiological alterations like myocardial scars and the influence of antiarrhythmic drug or catheter-based therapy.

Limitations
Experiments were conducted at room temperature. Thus, functional parameters may differ, when these experiments are performed at physiological temperatures. The current study was performed on cardiomyocytes isolated from a non-diseased mouse model. It is possible that remodeled atrial cardiomyocytes from certain disease-models are not protected against the occurrence of EADs and DADs. However, especially murine models of atrial fibrillation are very diverse in showing shortened, unchanged or even prolonged APDs as a reflection of a heterogeneous underlying electrophysiology (Riley et al., 2012) and thus may not always reflect the pathophysiology of human atrial fibrillation where action potentials are usually prolonged. Consequently, every model should be evaluated individually regarding its validity and limitations.

CONCLUSION
Taken together, we have demonstrated that the manifestation of proarrhythmic afterdepolarizations like EADs and the development of sAPs from DADs differs between atrial and ventricular myocytes in mice. This has to be considered when using murine models for translational research particularly with regard to atrial arrhythmia like atrial fibrillation. When using other animal, in vitro or computational models, we should critically scrutinize the underlying electrophysiological features and other fundamental aspects of the cell types before transferring findings derived from the ventricular level to atrial level, especially with regard to arrhythmia.

AUTHOR CONTRIBUTIONS
NB and JS designed the study. NB and PP acquired the data. NB, PP, DD, and JS performed the statistical analysis. NB, PP, and JS wrote the manuscript. DD, GF, JG, CP, FM, and LE critically revised the manuscript. All authors contributed substantially to the interpretation of the data and approved the final version of the manuscript.

FUNDING
This work was supported by the German Heart Foundation (Deutsche Stiftung für Herzforschung; F/44/12) to CP, FM, and LE, F/11/17 to NB and LE, and the Deutsche Forschungsgemeinschaft, Sonderforschungsbereich SFB 656 (project C11) to LE, and Mu1376/11-3 to FM. JG was supported by the National Institutes of Health (NIHR01HL048509).

ACKNOWLEDGMENTS
This work contains data from the doctoral thesis of PP.