Cardiomyocyte Na+ and Ca2+ mishandling drives vicious cycle involving CaMKII, ROS, and ryanodine receptors

Cardiomyocyte Na+ and Ca2+ mishandling, upregulated Ca2+/calmodulin-dependent kinase II (CaMKII), and increased reactive oxygen species (ROS) are characteristics of various heart diseases, including heart failure (HF), long QT (LQT) syndrome, and catecholaminergic polymorphic ventricular tachycardia (CPVT). These changes may form a vicious cycle of positive feedback to promote cardiac dysfunction and arrhythmias. In HF rabbit cardiomyocytes investigated in this study, the inhibition of CaMKII, late Na+ current (INaL), and leaky ryanodine receptors (RyRs) all attenuated the prolongation and increased short-term variability (STV) of action potential duration (APD), but in age-matched controls these inhibitors had no or minimal effects. In control cardiomyocytes, we enhanced RyR leak (by low [caffeine] plus isoproterenol mimicking CPVT) which markedly increased STV and delayed afterdepolarizations (DADs). These proarrhythmic changes were significantly attenuated by both CaMKII inhibition and mitochondrial ROS scavenging, with a slight synergy with INaL inhibition. Inducing LQT by elevating INaL (by Anemone toxin II, ATX-II) caused markedly prolonged APD, increased STV, and early afterdepolarizations (EADs). Those proarrhythmic ATX-II effects were largely attenuated by mitochondrial ROS scavenging, and partially reduced by inhibition of CaMKII and pathological leaky RyRs using dantrolene. In human induced pluripotent stem cell-derived cardiomyocytes (hiPSC-CMs) bearing LQT3 mutation SCN5A N406K, dantrolene significantly attenuated cell arrhythmias and APD prolongation. Targeting critical components of the Na+–Ca2+–CaMKII–ROS–INaL arrhythmogenic vicious cycle may exhibit important on-target and also trans-target effects (e.g., INaL and RyR inhibition can alter INaL-mediated LQT3 effects). Incorporating this vicious cycle into therapeutic strategies provides novel integrated insight for treating cardiac arrhythmias and diseases.


Introduction
Heart failure (HF) is characterized by cardiomyocyte Na + and Ca 2+ dysregulation including elevated intracellular [Na + ] ([Na + ] i ) and late Na + current (I NaL ), reduced sarcoplasmic reticulum (SR) Ca 2+ uptake, and increased diastolic SR Ca 2+ leak, Na + /Ca 2+ exchange (NCX), and reactive oxygen species (ROS) that contribute to systolic dysfunction and arrhythmias [1,2,8,27,54]. These alterations also frequently occur in many other heart diseases such as atrial fibrillation [47], ischemia/reperfusion injury [56], hypertrophic cardiomyopathy [6], long QT (LQT) syndromes [48], catecholaminergic polymorphic ventricular tachycardia (CPVT) [32,70], and diabetes [16,25]. Moreover, Ca 2+ / calmodulin-dependent protein kinase δ (CaMKIIδ) is also upregulated and chronically active in these diseases [1,20,61], and directly promotes I NaL [67] and diastolic SR Ca 2+ leak through the ryanodine receptor (RyR) [1]. Furthermore, reactive oxygen species (ROS) are increased by CaMKII [49], and elevated [Na + ] i and intracellular [Ca 2+ ] ([Ca 2+ ] i ) [7], which in turn further stimulate CaMKII [11] and RyR leak [50]. Thus, these pathological changes in HF are connected via a vicious cycle of positive feedback reinforcing systolic and diastolic dysfunction and arrhythmia mechanisms [18,45,68] (see Fig. 1). For example, a primary increase in SR Ca 2+ leak would promote CaMKII activation, which can promote I NaL , prolong action potential duration (APD), increase [Na + ] i , and ROS production that can further drive the cycle and amplify the functional impacts of initial insults at any given point. Based on the highly integrative nature of the vicious cycle and the significant impact it has on pathophysiological development, we hypothesized that targeting one of the key components (or combination of those) can prevent cellular proarrhythmia. Inhibition of one component in the vicious cycle with a selective drug is expected to induce the drug-specific on-target effect but also trans-target effects in the loop as it may reduce the feedback activation of the vicious cycle. The on-target drug effects have been the focus of research in recent decades and showed benefits in HF. (1) Selective Na + channel inhibitors (tetrodotoxin, GS-967) were shown to reverse the increased I NaL and the prolongation of the APD in HF [27,43].
(2) CaMKII inhibition using KN-93 or autocamide-2-related inhibitory peptide (AIP) was shown to reduce RyR leak (for matched SR Ca 2+ load) in rabbit HF [1], and reduced diastolic Ca 2+ spark rate in human HF [59]. (3) Mitochondrial-targeted antioxidant MitoTEMPOL normalized global cellular ROS and prevented arrhythmogenic remodelling in a guinea pig model of nonischaemic HF [9]. (4) The pathological leaky conformation of RyR, induced by CaMKII and ROS, can be selectively inhibited using dantrolene, which reduces SR Ca 2+ leak in CPVT and HF [64]. However, the trans-target effects and the strengths of interactions in this vicious cycle have not been systematically investigated.

Rabbit cardiomyocyte isolation
Enzymatic isolation of left ventricular cardiomyocytes from New Zealand White rabbits (male, 3-4-month-old) Fig. 1 Schematic of arrhythmogenic vicious cycle in heart disease. Heart failure (HF) is characterized by increases of ryanodine receptor (RyR) mediated Ca 2+ leak, Ca 2+ /calmodulin-dependent kinase II (CaMKII) activity, late Na current (I NaL ), intracellular [Na] i , action potential duration (APD) and reactive oxygen species (ROS) production, along with reduced repolarization reserve (K + currents, I K ). These factors form a vicious positive feedback cycle that perpetuates HF-associated dysfunction and arrhythmogenesis. For example, the RyR Ca 2+ leak increases local [Ca 2+ ], further activating cleft CaM-KII that further enhances RyR leak and I NaL (red arrows) and downregulates K + channel expression to reduce I K , which prolongs APD (as in genetic long QT (LQT) syndromes). Long APDs predispose myocytes to early afterdepolarizations (EADs) and increased intracellular [Na + ] and [Ca 2+ ] loading, which impairs mitochondrial Ca 2+ handling and may further promote ROS production. ROS can further promote I NaL and pathological leaky RyR (as in catecholaminergic polymorphic ventricular tachycardia, CPVT), and increase propensity for delayed afterdepolarizations (DADs). ROS also induces autonomous CaMKII activation closing the positive feedback loop was performed as previously described [21]. Briefly, animals were injected with heparin (400 U/kg body weight) and anesthetized with isoflurane (3-5%). Hearts were excised and retrograde perfused on constant flow Langendorff apparatus (5 min, 37 °C) with Ca 2+ -free normal Tyrode's solution, gassed with 100% O 2 . Then, ventricular myocytes were digested using collagenase type II (Worthington) and protease type XIV (Sigma-Aldrich). Ventricular myocytes were dispersed mechanically and filtered through a nylon mesh and allowed to sediment for ~ 10 min. The sedimentation was repeated three times using increasing [Ca 2+ ] from 0.125 to 0.25 then 0.5 mmol/L. Finally, ventricular myocytes were kept in Tyrode's solution at room temperature until use.

HF rabbit model
HF was induced in New Zealand White rabbits (male, 3-4-month-old) by aortic insufficiency and 4 weeks later by aortic constriction as previously described [22]. Data here reported were obtained from 10 HF and 10 age-matched control rabbits at 2-2.5 years of age. Echocardiography was performed periodically to monitor cardiac function. Cardiomyocytes were isolated from HF rabbits when left ventricular end-systolic dimension exceeded 1.45 cm. HF animals exhibited significant myocardial hypertrophy, enlarged left ventricular dimensions, pulmonary congestion, and abdominal ascites fluid accumulation, similar to our previous studies on this HF rabbit model [22,54].

Human iPSC-CMs
Patient specific hiPSC line carrying the SCN5A N406K mutation was generated as previously described [60]. Human iPSC-CMs were differentiated by methods developed in the laboratory of Mark Mercola [44]. At day 20, hiPSC-CMs were placed in a metabolic maturation media and cultured for 5 weeks to improve cardiomyocyte phenotype, including more negative diastolic membrane potentials and Na + current dependent action potentials [12]. Then, hiPSC-CM monolayers were dissociated and re-plated in low density onto Matrigel-coated coverslips 3-5 days before experiments.
Action potentials (APs) were evoked by 2-ms-long suprathreshold depolarizing pulses delivered via the patch pipette. 50 consecutive APs were recorded to examine the average behaviour, and APD at 90% repolarization (APD 90 ) was determined. Series of 50 consecutive APs were analysed to estimate short-term variability of APD 90 (STV) according to the following formula: STV = Σ(│APD n+1 -APD n │)/ [(n beats −1)×√2], where APD n and APD n+1 indicate the durations of the nth and (n + 1)th APs, and n beats denotes the total number of consecutive beats analysed. APD alternans magnitude was calculated as the difference between the average APD 90 of odd and even numbered beats during 50 consecutive APs recorded. Diastolic arrhythmogenic activities were elicited by cessation of 1-min tachypacing, and membrane potential was recorded for additional 1 min. Delayed afterdepolarizations (DADs) were defined as > 1 mV depolarization within 0.5 s. Spontaneous APs (sAPs) were defined as depolarizations showing overshoot with a fast upstroke phase. Early afterdepolarizations (EADs) were assessed at 0.2 Hz pacing, and EADs were defined as > 3 mV depolarization during AP repolarization.
AP-clamp experiments were performed to measure I NaL as previously described [19]. A typical rabbit AP was used to AP-clamp cells at 2 Hz pacing frequency. I NaL was measured as GS-967 (1 μmol/L)-sensitive current in control and following enhancement with ATX-II (5 nmol/L).
Cell pretreatments with MitoTEMPOL and AIP (myristoylated) started 30 min before the experiments, and the drugs were also added to both the perfusion and pipette solutions.
Chemicals and reagents were purchased from Sigma-Aldrich
To measure [Ca 2+ ] i , cardiomyocytes were loaded with 10 μmol/L Rhod2-AM (ThermoFisher, Waltham, MA, USA) for 10 min at room temperature and subsequently left to de-esterify in fresh Tyrode's solution for a minimum of 30 min. Then, cardiomyocytes were placed in a RC-27NE2 recording chamber and stimulated at 0.5 Hz frequency in Tyrode's solution at room temperature. Rhod2 was excited at 561 nm wavelength using an Optoscan monochromator, and fluorescence was collected at 530 ± 20 nm. Fluorescence signals were recorded after steady state was reached in the cell during pacing.

Statistical analysis
Data are presented as Mean ± SEM. Statistical significance of differences for normally distributed data was tested by paired Student's t-test to compare two groups and ANOVA with Dunnett's or Tukey's post-hoc test to compare multiple groups. For non-normally distributed data, we used Wilcoxon matched-pairs signed rank test, Mann-Whitney test, and Kruskal-Wallis ANOVA with Dunn's post-hoc test. Differences were deemed significant if P < 0.05.

Enhanced RyR leak, CaMKII, and I NaL all contribute to arrhythmogenic AP changes in HF
Cardiomyocytes in our HF rabbit model exhibited significantly prolonged APD 90 and greater short-term APD variability (STV) vs. age-matched healthy controls at 1 Hz at 37 °C ( Fig. 2a-d). We tested the effects of specific inhibition of either CaMKII, I NaL or RyR leak on APs of rabbit ventricular myocytes isolated from failing and healthy hearts. Pretreatment with the selective CaMKII inhibitor peptide AIP (1 μmol/L) or the late Na + current inhibitor GS-967 (1 μmol/L) significantly shortened APD 90 and reduced STV in failing myocytes to the level of healthy age-matched myocytes ( Fig. 2a-d). Interestingly, the pathological RyR conformation inhibitor dantrolene (10 μmol/L) also shortened APD 90 and reduced STV in HF ( Fig. 2a-d). Importantly, in healthy control myocytes neither AIP nor dantrolene had significant effects on APD 90 , and GS-967 only slightly shortened APD 90 in healthy myocytes (Fig. 2c). In healthy myocytes GS-967 and AIP slightly reduced STV, but those differences were quantitatively small compared to those for HF myocytes (Fig. 2d). The effects of direct I NaL inhibition (GS-967) and CaMKII (AIP) on APD and STV could be expected because I NaL and CaMKII activity are known to be elevated in HF, and CaMKII has been shown to directly enhance I NaL [5,27,67]. However, the potent effect of dantrolene on APD and STV is evidence that the pathological RyR state in HF increases APD and STV, which may be mediated by the vicious cycle via SR Ca 2+ leak-promoted CaMKII and I NaL .

RyR leak increases APD-variability via mitoROS-CaMKII-I NaL feedback
To separate RyR leak from the complex HF phenotype, in terms of the arrhythmogenic feedback signalling network (i.e., the vicious cycle), we induced RyR leak by low [caffeine] (200 μmol/L) and isoproterenol (ISO; 100 nmol/L) in healthy rabbit ventricular myocytes. Caffeine (3 min) slightly prolonged APD 90 and increased STV (Fig. 3a, b). The additional application of GS-967 decreased APD 90 and STV back to control suggesting a role for Ca 2+ -dependent upregulation of I NaL (Fig. 3a, b). In contrast to caffeine effects, ISO (3 min) shortened APD 90 and reduced STV (Fig. 3c, d). Then, inhibition of the slow delayed rectifier K + current (I Ks ) using HMR-1556 (HMR, 1 μmol/L) in the presence of ISO markedly prolonged APD 90 (mimicking LQT1) and significantly increased STV, while HMR had no effect on APD 90 at basal conditions without ISO stimulation (Fig. 3c, d). These data suggest that the upregulation of I Ks counterbalances the increased I NaL during β-adrenergic stimulation. Nonetheless, during steady-state pacing at 1 Hz, only very few DADs occurred in a small fraction of cells treated with either caffeine or ISO alone (Fig. 3e, f). However, when caffeine and ISO were applied together, several DADs were observed in every cell measured (Fig. 3e, f). This suggests that the increased SR Ca 2+ leak (caffeine) must be combined with enhanced SR Ca 2+ loading (ISO) to induce DADs in healthy myocytes. Hence, in the following, we used a combination treatment of low [caffeine] and ISO to investigate the role of the vicious cycle in proarrhythmic AP changes.
Following 3-min caffeine + ISO treatment, the AP plateau was significantly elevated (Fig. 4a), but the APD 90 did not change (Fig. 4a, c); however, STV was markedly increased (Fig. 4b, d). Importantly, the I NaL inhibitor GS-967 (and not a direct SR Ca 2+ leak modulator) significantly shortened APD 90 following caffeine + ISO and attenuated the increase in STV (Fig. 4a-d). Cell pretreatment with the mitochondrial ROS (mitoROS) scavenger mitoTEMPOL (20 μmol/L) or AIP did not change significantly baseline APD 90 and STV (Fig. 4c, d). However, caffeine + ISO induced APD 90 shortening in both mitoTEMPOL and AIP pretreated cells ( Fig. 4a-d). The additional application of GS-967 no longer altered APD 90 in mitoTEMPOL and AIP pretreated cells (Fig. 4a-d). These data indicate that mitoROS-CaMKII signalling markedly upregulates I NaL following RyR leak enhancement and contributes to increased beat-to-beat APD 90 -variability. Thus, SR Ca 2+ leak enhancement recruits ROS, I NaL and CaMKII as part of its integrated response.
Next, we examined the stability of the SR Ca 2+ release system following cessation of pacing. Caffeine + ISO induced spontaneous SR Ca 2+ release events, which were attenuated in mitoTEMPOL-treated cells (Fig. 6a, b). Caffeine + ISO induced multiple DADs and, in a few instances, spontaneous APs (sAPs) following cessation of tachypacing (Fig. 6c). The frequency of DADs was markedly attenuated by cell pretreatment with mitoTEMPOL or AIP (Fig. 6d). Addition of GS-967 reduced DAD frequency only in MitoTEMPOL-pretreated cells but did not change DAD amplitude following cessation of pacing (Fig. 6e).
These data indicate that mitochondrial superoxide production and CaMKII activation markedly enhance DADs in cells with pronounced RyR leak, and a combined treatment of mitoTEMPOL or AIP and GS-967 is largely protective against DADs.

Enhanced I NaL induces RyR leak that further prolongs APD
Next, we tested whether proarrhythmic AP changes induced by enhanced I NaL are attenuated by dantrolene in healthy rabbit ventricular myocytes. Anemone toxin II (ATX-II, 5 nmol/L) significantly enhanced I NaL during AP-clamp, and the net charge carried by I NaL increased by 3.9-fold (Fig. 7a, b). In current-clamp, ATX-II also prolonged APD 90 (mimicking LQT3) and markedly increased STV (Fig. 7c,  d). Importantly, dantrolene attenuated the ATX-II-induced APD prolongation (Fig. 7c), and this dantrolene effect was absent in cells preincubated with mitoTEMPOL and AIP (Fig. 7e). Moreover, mitoTEMPOL and AIP slightly reduced the increase in STV by ATX-II (Fig. 7f). These data indicate that the mitoROS-CaMKII-induced RyR leak contributes to APD prolongation when I NaL is enhanced by ATX-II. However, the dantrolene impact on APD prolongation is modest.

I NaL induced EADs are attenuated by dantrolene, mitoTEMPOL and AIP
ATX-II also increased systolic and diastolic intracellular Ca 2+ levels (measured as change in Rhod-2 fluorescence) and prolonged the [Ca 2+ ] i transient (CaT, Fig. 8a). ATX-II markedly prolonged APD at low pacing rates (Fig. 8b) and 10 nmol/L ATX-II induced early afterdepolarizations (EADs) (Fig. 8c). Dantrolene slightly attenuated both the frequency and amplitude of EADs in ATX-II (Fig. 8c-f). Interestingly, mitoTEMPOL preincubation markedly reduced the frequency (Fig. 8d) but not the amplitude of the EADs (Fig. 8e); however, mitoTEMPOL delayed the time to the first EAD significantly (Fig. 8g). In contrast to mitoTEM-POL, AIP was only slightly protective against EAD formation induced by ATX-II (Fig. 8d). These data indicate that ROS production, and also slightly CaMKII and RyR leak contribute to ATX-II induced EADs.

Dantrolene reduces arrhythmogenic activities in SCN5A N406K hiPSC-CMs
Next, we tested the effects of dantrolene in hiPSC-CMs carrying the SCN5A N406K LQT3 mutation, which has been associated with significant QT prolongation, increased risk of torsade de pointes-type ventricular tachycardia and sudden cardiac death [60]. Previous biophysical characterization [31] showed that the mutant channels exhibit an interesting mixed phenotype with increased I NaL as gain-of-function (long QT3) and a decreased peak I Na (due to reduced surface expression of Na + channels) as loss-of-function (Brugada syndrome). Importantly, these changes in Na + channel function are similar to the CaMKII-mediated effects [67] and remodelling in HF [65]. Moreover, hiPSC-CMs carrying the SCN5A N406K mutation also showed impaired intracellular Ca 2+ handling and Ca 2+ -dependent arrhythmias [60].
APs in SCN5A N406K and wild type (WT) hiPSC-CMs, cultured in a metabolic maturation media and paced at 1 Hz [12], exhibited sufficiently negative diastolic V m to enable robust Na + channel availability and AP rate of rise (Fig. 9a-c). Even so the N406K vs. WT cells exhibited lower maximal upstroke velocity (dV/dt max ), prolonged APD 90 , and significant AP triangulation, in line with data in literature and the expected consequences of decreased peak I Na and increased I NaL (Fig. 9a-c). Cells carrying the N406K mutation also had frequent spontaneous depolarizations (Fig. 9b). Moreover, significant APD alternans occurred in N406K mutants at higher pacing rates (starting at 3 Hz; Fig. 9b). Importantly, dantrolene treatment significantly reduced the spontaneous depolarizations and shortened APD 90 in N406K, while it had no effect on APD 90 in WT hiPSC-CMs (Fig. 9a-c). Dantrolene also increased the APD alternans threshold frequency (from 3 to 4 Hz) and reduced the amplitude of APD alternans (Fig. 9c). These data reinforce the suggested interplay between I NaL and RyR in the vicious cycle.

Discussion
Impairments in cardiomyocyte Na + and Ca 2+ handling are characteristic of HF and contribute to contractile dysfunction and arrhythmias [45,54,68]. In our HF rabbit model, [Na + ] i was found to be 3 mmol/L higher than in control [8]. In agreement with this, I NaL was increased by 82% in failing rabbit myocytes, and the I NaL upregulation was predominantly CaMKII-dependent [27]. CaMKIIδ C expression and autophosphorylation were increased by 112% and 260%, respectively, in HF rabbit hearts, and similar increases were found in human heart samples from patients with dilated and ischemic cardiomyopathies [5]. CaMKII-dependent phosphorylation of RyR2 at S2814 was increased by 105% in rabbit HF [1] and led to increased SR Ca 2+ leak at a given SR Ca 2+ load [58]. Moreover, NCX expression and NCX current were also increased by 93% and 120%, respectively, in HF rabbits [53]. Furthermore, the membrane resistance is increased in HF due to 25-50% reduction in inward rectifier K + current (I K1 ) [22,54], thus, a given depolarizing current can cause larger DADs. The magnitude of I K1 reduction quantitatively matches the downregulation of Kcnj2/ K ir 2.1 expression upon chronic CaMKII overexpression [21]. Taken together, less Δ[Ca 2+ ] i is required to trigger a spontaneous AP in failing cardiomyocytes [54]. Importantly, CaMKII inhibition was shown to prevent DADs in isolated failing cardiomyocytes [22] and reduced in vivo arrhythmia inducibility in HF [28]. Calcium-and CaMKII dependent arrhythmias were also demonstrated in long QT caused by either gain-of-function mutation in Na + channels [72] or loss-of-function mutation in K + channels [63]. Along the same lines, in RyR-mutant CPVT, inhibition of CaMKII markedly attenuated proarrhythmic activities [4,38]. These data indicate the activation of the vicious cycle and its pivotal role in arrhythmogenesis in HF, LQT, and CPVT.
ROS is a critical mediator of pathological cellular remodelling and contributes to impaired cardiomyocyte Na + and Ca 2+ homeostasis in heart diseases [15]. Our data support the concept of a strong, bidirectional feedback between SR Ca 2+ leak and increased ROS [17]. ROS can oxidize RyRs d, e DAD frequency and amplitude during a 1-min recording following cessation of tachypacing. Late Na + current was inhibited using GS-967 (GS, 1 μmol/L), and CaMKII was inhibited using AIP (1 μmol/L). Blue lines represent lognormal distribution curves. DAD and sCaR frequencies were compared using Kruskal-Wallis one-way ANOVA with Dunn's multiple comparisons test; DAD amplitudes were compared using Mann-Whitney test. (N = 6-7 animals in each treatment group, each individual myocyte (n) is shown as a data point.) [50] and induce autonomous CaMKII activation [11], both further increase SR Ca 2+ leak [62,69]. SR Ca 2+ leak then may increase Ca 2+ uptake into neighbouring mitochondria via the mitochondrial Ca 2+ uniporter (MCU) [3,36]. Oxidation of MCU can also increase its activity [10]. Furthermore, CaMKII can also increase ROS via NADPH oxidase 2 (NOX2) [41,49]. While some data in isolated mitochondria suggested elevated mitochondrial [Ca 2+ ] in HF (due to leaky RyRs and increased mitochondrial Ca 2+ uptake [57]), more direct HF measurements in intact guinea-pig ventricular myocytes indicated reduced mitochondrial [Ca 2+ ] (due to elevated [Na + ] i , lower CaTs and greater Ca 2+ extrusion via mitochondrial Na + /Ca 2+ exchange) [42]. Moreover, both increased and decreased mitochondrial [Ca 2+ ] may increase ROS production [7]. Interestingly, a recent paper showed that moderate overexpression of MCU that enhances mitochondrial Ca 2+ uptake also improves HF phenotype by reducing SR Ca 2+ leak [40]. This highlights the pathophysiological role of the vicious cycle and mitochondrial ROS therein.
Ion channel remodelling in HF leads to APD prolongation and increased STV [22], creating a vulnerable arrhythmia substrate. APD prolongation then may promote further cellular Na + and Ca 2+ loading and CaMKII activation (Fig. 1). Inhibition of the upregulated I NaL , CaMKII and leaky RyRs all reduced APD prolongation and STV in HF (Fig. 2). In contrast, acute pharmacological induction of RyR leak by caffeine + isoproterenol did not change APD (Fig. 4). The more pronounced APD change by RyR leak in HF cardiomyocytes might reflect the effect of reduced repolarization reserve (downregulated K + channels [46]) and altered balance between inward and outward ionic currents [23,24]. In line with this, inhibition of I Ks led to APD prolongation following β-adrenergic stimulation in rabbit (Fig. 3) and human [33] ventricular myocytes. Hamilton et al. [18] also showed that caffeine + isoproterenol but not isoproterenol alone increased mitoROS production. Moreover, we have shown a two-hit arrhythmia model in which hyperglycaemia-induced CaMKII activation and RyR leak alone did not change APD, but when repolarization reserve was reduced, a marked APD prolongation occurred [26]. Like with arrhythmia induction, arrhythmia termination may require two simultaneous targets. Such synergy was observed when either MitoTEMPOL or AIP was combined with GS-967 leading to a marked reduction in DADs (Figs. 5, 6). Then, in an inverse experimental setting, enhanced I NaL prolonged APD (Fig. 7), increased [Ca 2+ ] i , and induced EADs (Fig. 8).
Multiple mechanisms can contribute to EADs, including spontaneous SR Ca 2+ release and inward NCX, reopening of L-type Ca 2+ channels (LTCC), and augmentation of I NaL , and all of these are modulated by [Ca 2+ ] i and CaMKII [30,55]. CaMKII inhibition attenuated EADs (Fig. 8) and buffering [Ca 2+ ] i has been previously shown to abolish EADs induced by ATX-II [29]. Experimental [71] and computational modelling [13] studies mechanistically investigated the EAD mechanisms upon H 2 O 2 treatment and showed that EADs emerge at slow pacing rates upon simultaneous activation of both LTCC and Na + channels via ROS-dependent CaMKII activation (and alone, neither RyR nor I NaL nor LTCC effects were sufficient to produce EADs). Intracellular Na + loading induced by either ouabain [39] or ATX-II Fig. 8 Enhanced late Na + current induces RyR leak, mito-ROS and CaMKII to promote EADs. a Intracellular Ca 2+ transient (CaT) measured as Rhod-2 fluorescence in control and ATX-II (10 nmol/L) in rabbit ventricular myocytes paced at 0.5 Hz. The error bar on the control diastolic value indicates the degree of variability in the baseline raw F (F/non-cellular background), and F 0 is the control baseline F in each cell. Wilcoxon matched-pairs signed rank test. b Reverse-rate dependent APD 90 prolongation by ATX-II (5 and 10 nmol/L). c Early afterdepolarizations (EADs, red arrowheads) at 0.2 Hz steadystate pacing in a representative rabbit ventricular cell in control and ATX-II (10 nmol/L), and after application of dantrolene (DAN, 10 μmol/L). d, e EAD frequency and amplitude during pacing in cells without pretreatment and following pretreatment with MitoTEMPOL (20 μmol/L) and CaMKII inhibitor AIP (1 μmol/L). Blue lines represent lognormal distribution curves. f Cumulative EAD frequencies as a function of EAD amplitudes. g Time to first EAD after application of ATX-II. Cells were paced at 0.2 Hz steady-state. EAD frequencies were compared using Friedman repeated measure ANOVA with Dunn's multiple comparisons test. EAD amplitudes were compared using Mann-Whitney test. (N = 3-7 animals in each treatment group, each individual myocyte (n) is shown as a data point.) [34,66] treatment has been shown to increase mitoROS and diastolic Ca 2+ -triggered arrhythmias. Here we showed that mitoROS also plays an important role in mediating EADs induced by the ATX-II-enhanced I NaL (Fig. 8), which may reflect spatial and functional coupling between Na V 1.5 channels and mitochondria [52]. Moreover, multiscale modelling of the mitochondria-SR microdomain showed that elevated ROS production increases [Ca 2+ ] i and arrhythmia propensity by stimulating RyRs and inhibiting SERCA [37]. Consistent with this, MitoTEMPOL pretreatment significantly prolonged EAD latency (Fig. 8) suggesting that the increase in ROS is an early response to [Na + ] i loading. Interestingly, inhibition of SR Ca 2+ leak by dantrolene attenuated APD prolongation following ATX-II treatment (Fig. 7) and in HF (Fig. 2). The effects of dantrolene on APD (and EAD formation) might be explained by the attenuation of SR Ca 2+ leak-induced CaMKII activity, changes in myocyte Na + and Ca 2+ loading, enhanced inward NCX, and late Ca 2+ sparks, which can activate the vicious cycle and influence AP configuration [14]. Dantrolene also markedly attenuated APD prolongation, alternans, and spontaneous diastolic activities (i.e., DADs, sAPs) in hiPSC-CMs carrying SCN5A N406K mutation, highlighting the critical role of SR Ca 2+ leak (and the activated vicious cycle) in these arrhythmias (Fig. 9).
Here, we aimed to preserve physiological regulation in our cellular experiments to uncover interactions within the feedback loops. However, this approach has limitations on quantifying the exact role that each component plays in the vicious cycle. Full inhibition of one key component may break the whole loop and have a marked arrhythmia reducing effect, like that seen in HF (Fig. 2). However, this approach may overestimate the individual contribution of one arm in the feedback loop. On the contrary, the APD shortening effect of dantrolene in HF (Fig. 2), ATX-IIinduced long QT3 (Fig. 7), and SCN5A N406K (Fig. 9), and the antiarrhythmic effects of MitoTEMPOL and AIP in pharmacologically enhanced I NaL (Fig. 5) and RyR leak (Fig. 8) clearly demonstrate the importance and strength of the vicious cycle. The interaction between [Na + ] i loading and ROS in promoting arrhythmias was found to be particularly strong, which then can lead to further RyR leak and CaMKII activation. In line with this, such synergy between c Summary data on maximal upstroke velocity (dV/dt max ), AP duration at 90% repolarization (APD 90 ), AP triangulation (APD 90 -APD 50 ) at 1 Hz pacing, and the magnitude of APD 90 alternans in subsequent beats at 4 Hz pacing. Student's paired t-test and ANOVA with Tukey's multiple comparisons test. (Each individual hiPSC-CM (n) is shown as a data point.) multiple antiarrhythmic targets (e.g., inhibition of both Na + channels and leaky RyRs) may contribute to the clinical benefit of flecainide [35] and ranolazine [51]. Future, mechanistic experiments (e.g., using permeabilized myocytes) could determine the quantitative relationship between [Na + ] i and mitochondrial ROS production at a given [Ca 2+ ] i . Incorporating these data may help to constrain and improve computational models in the future, which then would allow more controlled analysis of different branches of the vicious cycle.
As discussed above, many components of the [Na + ] i -[Ca 2+ ] i -ROS-CaMKII-RyR leak vicious cycle signalling have already been shown; however, the strength of feedback interactions have not been previously investigated. Our conceptual novelty here is the identification of important trans-target effects beyond the on-target effects of the otherwise selective inhibitors. It may have important clinical implications suggesting that potentially a combination therapy targeting the major components of the arrhythmogenic vicious cycle described here can be synergistic and may provide substantial benefits in heart diseases by reducing cellular proarrhythmia. The use of combination therapy may also be advantageous by reducing the effective dose of each drug, thus reducing their adverse effects. Moreover, our data show that the most favourable drug target(s) may vary among heart diseases, and thus, personalized medicine approaches are required to identify the optimal drug combinations.
Funding This work was supported by grants from the National Institutes of Health (NIH) P01-141084 to DMB and MM, and R01-142282 to DMB and JB, the Finnish Cultural Foundation 00200088 to RPP, and the Osk. Huttunen Foundation to RPP.
Availability of data and materials All data and materials are available on reasonable request to the corresponding author.

Conflict of interest
The authors declare that they have no conflict of interest.
Ethical approval All animal handling and laboratory procedures were in accordance with the approved protocols (#21572 and #21137) of the Institutional Animal Care and Use Committee at University of California, Davis conforming to the NIH Guide for the Care and Use of Laboratory Animals (8 th edition, 2011).
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.