Mitochondrial-Mediated Oxidative Ca2+/Calmodulin-Dependent Kinase II Activation Induces Early Afterdepolarizations in Guinea Pig Cardiomyocytes: An In Silico Study.

Background Oxidative stress-mediated Ca2+/calmodulin-dependent protein kinase II (Ca MKII) phosphorylation of cardiac ion channels has emerged as a critical contributor to arrhythmogenesis in cardiac pathology. However, the link between mitochondrial-derived reactive oxygen species (md ROS ) and increased Ca MKII activity in the context of cardiac arrhythmias has not been fully elucidated and is difficult to establish experimentally. Methods and Results We hypothesize that pathological md ROS can cause erratic action potentials through the oxidation-dependent Ca MKII activation pathway. We further propose that Ca MKII -dependent phosphorylation of sarcolemmal slow Na+ channels alone is sufficient to elicit early afterdepolarizations. To test the hypotheses, we expanded our well-established guinea pig cardiomyocyte excitation- contraction coupling, mitochondrial energetics, and ROS - induced- ROS - release model by incorporating oxidative Ca MKII activation and Ca MKII -dependent Na+ channel phosphorylation in silico. Simulations show that md ROS mediated-Ca MKII activation elicits early afterdepolarizations by augmenting the late Na+ currents, which can be suppressed by blocking L-type Ca2+ channels or Na+/Ca2+ exchangers. Interestingly, we found that oxidative Ca MKII activation-induced early afterdepolarizations are sustained even after md ROS has returned to its physiological levels. Moreover, mitochondrial-targeting antioxidant treatment can suppress the early afterdepolarizations, but only if given in an appropriate time window. Incorporating concurrent md ROS -induced ryanodine receptors activation further exacerbates the proarrhythmogenic effect of oxidative Ca MKII activation. Conclusions We conclude that oxidative Ca MKII activation-dependent Na channel phosphorylation is a critical pathway in mitochondria-mediated cardiac arrhythmogenesis.

Clinical Perspective

What Is New?

We developed a multiscale computational model linking cardiomyocyte mitochondrial energetics to Ca2+/calmodulin‐dependent protein kinase II activity and Ca2+ handling.

Mitochondrial‐mediated oxidative Ca2+/calmodulin‐dependent protein kinase II activation is sufficient to elicit early afterdepolarizations solely through enhanced late Na+ current.

Oxidative Ca2+/calmodulin‐dependent protein kinase II activation–elicited early afterdepolarizations are sustained even after mitochondrial‐derived reactive oxygen species has returned to its physiological levels.

What Are the Clinical Implications?

It is critically important to consider mitochondria when designing novel antiarrhythmic therapies.

It appears that there is a treatment window for antioxidants to suppress Ca2+/calmodulin‐dependent protein kinase II–mediated pathological effects.

Introduction

Cardiovascular disease is a major health problem in the United States and its incidence increases steadily as the general population ages.1 Despite advances in diagnosis and treatment, cardiovascular disease mortality remains high, accounting for nearly 500 000 American deaths each year.1 About one half of cardiovascular disease–related deaths occur suddenly because of sudden cardiac death resulting from ventricular arrhythmias.2, 3 Although the incidence rate is high, the precise molecular mechanisms underlying cardiac arrhythmogenesis are not fully understood, hindering the development of effective therapeutic strategies.

Recently, loss of mitochondrial function, which is often observed in many disease processes such as heart failure, ischemic cardiomyopathy, hypertrophic cardiomyopathy, and metabolic diseases, has emerged as a key contributor to the arrhythmogenic substrate. While the detailed mechanistic pathways remain incompletely understood, work from our laboratory and others4, 5, 6, 7 suggest that the proarrhythmic effect of mitochondria dysfunction is at least partially attributed to the organellar‐derived reactive oxygen species (ROS), which can influence multiple redox‐sensitive ion channels/transporters underlying Ca2+ handling such as ryanodine receptors (RyRs)6, 7, 8, 9 and sarcoplasmic reticulum (SR) Ca2+ ATPase (SERCA).10, 11 Beside Ca2+ handling proteins, excessive ROS can also affect sarcolemmal voltage–gated Na+ channels,12, 13 K+ channels, Na+/Ca2+ exchanger, and L‐type Ca2+ channels (LCCs).11, 14, 15, 16, 17, 18 In addition to direct modulation of redox‐sensitive ion channels, mitochondrial‐derived ROS (mdROS) may indirectly influence Ca2+ handling and action potentials (APs) via redox signaling pathways such as oxidation‐mediated Ca2+/calmodulin‐dependent kinase II (CaMKII) phosphorylation. CaMKII is a multifunctional protein kinase ubiquitously expressed in cardiomyocytes and is activated by binding to Ca2+/CaM and subsequent autophosphorylation.19, 20 A growing body of evidence has demonstrated that CaMKII can also be activated by ROS.21, 22, 23, 24 Once activated, CaMKII can phosphorylate a wide range of key Ca2+ and Na+ regulatory proteins such as LCCs,25, 26, 27, 28 RyRs,29, 30, 31, 32, 33, 34, 35 phosphalamban,29, 34, 36 and Na+ channels.37, 38 Importantly, Xie et al39 showed that H2O2 perfusion–induced oxidative CaMKII activation leads to afterdepolarizations in isolated rabbit cardiomyocytes, likely by phosphorylation of Na+ channels and LCCs. Given those advances, the detailed mechanistic pathways by which oxidation‐dependent CaMKII activation creates a proarrhythmia substrate in diseased hearts remain unclear, partially because of the multidirectional interaction loops between CaMKII activation and ion handling. As a powerful tool complementary to experimental measurement, computational modeling has been applied to elucidate how CaMKII activation may influence cardiac ion handling and electrophysiology. For instance, Onal et al40 explored the CaMKII‐dependent regulation of late Na+ current (INa,L), Ca2+ homeostasis, and cellular excitability in atrial myocytes using a computer model. In another computational study, Dai et al41 showed that CaMKII overexpression facilitates early afterdepolarization (EAD) by prolonging the deactivation of the INa,L, and combination with β‐adrenergic activation further increases EAD risk. Modeling studies also suggested that CaMKII activation–mediated SR Ca2+ overload and increased cytosolic Na+ elicit post‐acidosis arrhythmias in human myocytes.42 To examine the role of oxidation‐dependent CaMKII activation in regulating cardiac cell excitability following myocardial infarction, Christensen et al43 developed a mathematical model of CaMKII activity, which, for the first time, includes both oxidative and autophosphorylation activation pathways. More recently, an integrative cardiomyocyte model has been developed by Foteinou et al44 to study the mechanistic role of oxidized CaMKII in the genesis of H2O2‐induced EADs in the heart. In a similar study, Zhang et al45 developed a new Markov chain model of CaMKII δ‐isoform that involves both of the autophosphorylation and oxidation pathways to simulate CaMKII activation and its effect on APs under oxidative stress in cardiomyocytes.

Given the advances, how endogenous ROS, especially those derived from mitochondria (mdROS), affect CaMKII activity and consequently ion homeostasis and AP remains largely unexplored. Dissecting direct mdROS effects and indirect effects caused by CaMKII phosphorylation is difficult to address experimentally, as is defining the contribution of individual targets to arrhythmogenesis. As the voltage‐gated Na+ currents (INa) are a significant contributor to the initiation and duration of the cardiac AP and a well‐recognized substrate of CaMKII phosphorylation, we hypothesize that mdROS‐mediated oxidative CaMKII activation could elicit abnormal APs by enhancing INa. To test the hypothesis, we expanded our well‐established cardiomyocyte excitation‐contraction coupling, mitochondrial energetics, and ROS‐induced‐ROS‐release (ECME‐RIRR)5, 46, 47 model by incorporating oxidative CaMKII activation and slow Na+ channel phosphorylation. Our simulations show that mdROS bursts–mediated oxidative CaMKII activation significantly augments INa,L, which alone is sufficient to cause EADs. Moreover, we show that under certain conditions the oxidative CaMKII activation–induced EADs persist even after mdROS have returned to physiological levels, an event that is likely attributed to CaMKII's property as a “memory molecule.” Finally, model simulations suggest that timing is critical for antioxidant treatments to effectively eliminate mdROS‐CaMKII activation–induced EADs.

Methods

No human or animal subject was involved in this theoretical study. Thus, there was no institutional review board approval required. The data, analytic methods, and codes will be made available to other researchers for purposes of reproducing the results or replicating the procedure. Model equations and parameters are available within the article or Tables S1 through S28.

Model Development

In this in silico study, we aimed to examine the effect of mitochondrial‐derived oxidative stress on CaMKII activation to induce arrhythmias. The model was based on our recently published guinea pig cardiomyocyte ECME‐RIRR model47 and incorporated several new model components including a CaMKII activity module, a Markov slow Na+ channel module, and an Na+ channel phosphorylation module. The scheme of the expanded ECME‐RIRR model is shown in Figure 1, and the newly added model components are described below.

Figure 1

Scheme of the expanded excitation‐contraction coupling, mitochondrial energetics, and ‐induced‐‐release (RIRR) model that incorporates oxidative Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation. The electrophysiological module describes the major ion channels underlying the action potential (eg, fast Na+ channel and Na+/Ca2+ exchanger) and processes involved in Ca2+ handling (eg, local Ca2+ control and transport of Ca2+ across the sarcoplasmic reticulum). The mitochondrial module accounts for major components of mitochondrial energetics such as tricarboxylic acid cycle and oxidative phosphorylation, as well as mitochondrial membrane ion channels (eg, Ca2+ uniport). The RIRR module describes reactive oxygen species (ROS) production (from the electron transport chain), transport (through inner membrane anion channel [IMAC]), and scavenging (eg, by the superoxide dismutase and glutathione peroxidase enzymes).

CaMKII activity module

The CaMKII module model was built based on the Markov chain models constructed by Foteinou et al44 and Zhang et al,45 which comprises CaMKII activation by ROS and experimental data from Erickson et al.22 For simplification, we assumed that Ca2+/CaM‐dependent activity, phosphorylation‐dependent activity, and oxidation‐dependent activity are homogeneous across the cell. With this assumption, the total activated CaMKII is defined as the sum of these activated CaMKII (ie, binding Ca2+/CaM, phosphorylated, and oxidized). The fraction of activated CaMKII can be calculated as:where [CaMKIIactive] is the concentration of total activated CaMKII and [CaMKIItotal] is the concentration of total CaMKII. The complete CaMKII activation model is described in Figure S1 and Tables S4 and S25.

The Markov slow Na+ channel module

Experimental studies showed that the effect of CaMKII activation on Na+ channels is mainly on the slow component, ie, augmenting the INa,L. As our ECME‐RIRR model consists only of the fast INa model, we adopted the Markov framework of the Na+ channel model developed by Grandi et al48 to incorporate the INa,L:where GNa,L is the conductance of the late Na+ channels (mS/μF) and PLO is the open possibility of the late Na+ channels. The complete Markov slow Na+ channel model and parameters are listed in Tables S1 and S26, respectively.

Slow Na+ channel dynamic phosphorylation module

For modeling purposes, we contended that Na+ channels are either phosphorylated by activated CaMKII or not. The transition between the unphosphorylated and phosphorylated Na+ channels can be described by a 2‐state Markov model (Figure S1). Specifically, the fraction of phosphorylated Na+ channels is governed bywhere is the fraction of phosphorylated Na+ channels, KPhos is the phosphorylation rate, which is proportional to the fraction of activated CaMKII, and KDephos is a constant, which can be determined by the experimental data as described previously.48 The total INa,L is calculated by the following equations:

In these equations, I′Na,L represents the late Na+ current caused by CaMKII activation, and P′LO is the open probability of phosphorylated late Na+ channels. The model parameters for the phosphorylated and unphosphorylated Na+ channels by activated CaMKII were refit with experimental data by Aiba et al49 and are listed in Table S27.

Simulation Protocol

The CaMKII activity and slow Na+ channel module models were integrated into the guinea pig ventricular myocytes ECME‐RIRR model47 after parameterization. To focus on the effect of mdROS‐mediated oxidative CaMKII activation on AP and dissect the underlying ionic mechanisms, we did not consider the direct effect of mdROS on redox‐sensitive ion handling proteins such as RyRs, SERCA, and Na+ channels in the present study unless otherwise specified. The formulas of other processes, such as ion channels and metabolic reactions, and model parameters were the same as those in the ECME‐RIRR model47 (Tables S1 through S28). The code of the new cell model was written in C++ (Visual Studio, Microsoft). The nonlinear ordinary differential equations were integrated numerically with CVODE as previously described.5, 50

The cardiomyocyte was stimulated at 0.25 Hz until the steady state was reached. The steady state values were then used as initial conditions for subsequent simulations. For model validation, we first simulated the effect of pacing cycle lengths (PCLs; 500, 1000, 2000, and 4000 ms) on AP duration (APD), then the ROS‐induced INa,L augmentation, and finally EAD incidence rate dependence on PCL under oxidative stress. Model simulations were compared with experimental data from the literature. After model validation, we simulated the effect of mdROS on CaMKII activation, INa,L, cytosolic Na+ and Ca2+ handling, and AP under various conditions. The production of mdROS was modeled as a fraction, or shunt, of electrons from the electron transport chain into the matrix, as previously described.5, 47, 51, 52 Studies have shown that under physiological conditions, up to 2% of the electron flowing the respiratory chain are partially reduced to form the superoxide,53 thus the physiological value of shunt was set as 2%. Pathological shunt was set as 10% or 14% to induce sustained mitochondrial oscillations, which is consistent with our previous computational studies.5, 47, 52 The simulation results were postprocessed and plotted using Origin software (OriginLab).

Results

Model Validation

To validate the built guinea pig cardiomyocyte ECME‐RIRR model that incorporates new model modules, we first simulated the effects of CaMKII phosphorylation and PCL on APD and compared the results with experimental data. Our simulations showed that increasing the PCL (from 500 to 4000 ms) caused stepwise APD elongation (Figure 2A, gray), which was further enhanced by CaMKII phosphorylation (data not shown). Those model predictions were comparable to experimental data reported by Aiba et al49 (Figure 2A, black).

Figure 2

Validation of cardiomyocyte excitation‐contraction coupling, mitochondrial energetics, and ‐induced‐‐release model that incorporates oxidative Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation and slow Na+ channels phosphorylation modules. A, The effect of pacing cycle length (PCL) on the duration of action potential (APD). For comparison, the APDs at different PCLs were normalized to the APD at PCL=1 second. Blank triangles are model simulations and black squares represent experimental data from Aiba et al.49 B, The effects of oxidative CaMKII activation on late Na+ current (INa,L). The integral of I a,L was calculated between 50 and 500 ms after the onset of depolarization. The CaMKII activation–induced INa,L change (ie, Δ integral INa,L) was obtained by subtracting the baseline INa,L integral from the CaMKII INa,L integral. Simulation was run with PCL=2 seconds. Experimental data are from Wagner et al.38 C, Effect of cytosolic H2O2 on action potential at different PCLs (1000 and 6000 seconds). Left panel shows model simulations and right panel shows experimental data from isolated rabbit ventricular myocytes (modified from Zhao et al55 with permission. Copyright© 2012, The American Physiological Society). WT indicates wild‐type.

We then simulated the effect of oxidative stress on the voltage‐gated INa (Figure 2B). In an experimental study, Wagner et al38 showed that H2O2 (200 μmol/L) perfusion caused a remarkable increase (≈177 A ms/F) in INa,L integral in wild‐type mouse ventricular myocytes and this enhancement was substantially reduced in CaMKII knockout cardiomyocytes (≈48.7 A ms/F). Our model simulations showed a similar trend: reducing CaMKII by 95% (estimated based on Western blot data in Backs et al54) notably reduced the oxidative stress–induced INa,L augmentation.

Finally, we examined the effect of ROS on APs at different PCLs. To be consistent with experimental studies by Xie et al39 and Zhao et al,55 the concentration of cytosolic ROS was set as 200 μmol/L in this simulation. Our simulations showed that oxidative stress–induced EAD incidence rate is PCL dependent: EADs could be induced readily when PCL was long (eg, 6 seconds) but hardly when PCL was relatively short (eg, 1 second) (Figure 2C). Those simulations were in agreement with previous experimental39, 55 (Figure 2D) and computational studies.44

Effect of mdROS on CaMKII and Ion Handling During Mitochondrial Oscillations

After model validation, we simulated how mitochondrial‐derived oxidative stress could influence CaMKII activity and ion handling in a “beating” guinea pig cardiomyocyte. As previously reported,5, 46, 47, 52, 56 increasing shunt, the fraction of the electrons of the respiratory chain towards the generation of from 0.02 to 0.10 triggered sustained mitochondrial oscillations and cyclic ROS production (Figure S2A). The results indicate that addition of new model components (eg, mdROS‐mediated oxidative CaMKII activation and CaMKII‐dependent Na+ channels phosphorylation) did not alter the dynamics of the existing model subsystems. Simulations also show that ΔΨm depolarization (and the associated mdROS bursting) led to sustained EADs during each oscillatory cycle (Figure 3A, for better visualization only the first depolarization was shown). Cytosolic Na+ concentration ([Na+]i) climbed gradually during mitochondrial depolarization (Figure 3B). Cytosolic Ca2+ transient rose slightly and exhibited a large increase during the decay phase (Figure 3C). The dynamics of the fraction of activated CaMKII were similar to that of [Na+]i (Figure 3D).

Figure 3

Effect of mitochondrial depolarization and mitochondrial‐derived reactive oxygen species (ROS) bursts on action potential (A), cytosolic Na+ ([Na+]i) (B), cytosolic Ca2+ concentration ([Ca2+]i) (C), and fraction of activated Ca2+/calmodulin‐dependent protein kinase II (CaMKII) (D). shunt=0.1 and pacing cycle length=2 seconds.

To examine the ionic mechanisms underlying mdROS‐mediated EADs, we analyzed the dynamics of major Ca2+ and Na+ handling currents/fluxes before and after mitochondrial depolarization, and with or without mdROS‐induced oxidative CaMKII activation. In the absence of CaMKII activation, the mdROS bursts slightly elongated APD and Ca2+ transient, enhanced INa,L, and shifted the Na+‐Ca2+ exchanger current (INaCa) forward component to the right (Figure 4, red lines). The effects on L‐type calcium channel current (ICaL), SR Ca2+ release, and SERCA Ca2+ uptake were small. Addition of CaMKII activation had negligible effects on APD and cytosolic Ca2+ concentration ([Ca2+]i), as well as INaCa, ICaL, and SR Ca2+ handling under physiological mdROS conditions (ie, polarized ΔΨm) (Figure 4, blue lines). The INa,L integral was slightly increased, likely caused by [Ca2+]i‐induced CaMKII activation. During mitochondrial depolarization, mdROS‐mediated CaMKII activation did not change the peak INa but caused substantial INa,L augmentation (Figure 4C, olive line arrow #1), resulting in APD prolongation and AP reverse (Figure 4A, olive lines). The AP reverse reactivated ICaL (Figure 4E, olive line), which triggered Ca2+‐induced Ca2+ release (Figure 4F, olive lines), leading to a larger Ca2+ elevation (Figure 4B, olive line) and a second INa,L surge (Figure 4C, olive line arrow #2). The forward mode INaCa was initially inhibited and reversed (arrow #1) and then largely amplified (arrow #2) (Figure 4C, olive line), caused by AP reverse and altered Na+ and Ca2+ homeostasis. It is worth mentioning that oxidative CaMKII activation alone could not induce delayed afterdepolarizations (DADs; data not shown), which is consistent with the findings of Foteinou et al.44

Figure 4

Dynamics of action potential (A), cytosolic Ca2+ concentration ([Ca2+]i) (B), Na+ current (INa) (C), Na+‐Ca2+ exchanger current (INaCa) (D), L‐type calcium channel current (ICaL) (E), and sarcoplasmic reticulum Ca2+ release (Jrel) (F) before (Repo) and after (Depo) mitochondrial depolarization. Ca2+/calmodulin‐dependent protein kinase II (CaMKII) (+) represents the new excitation‐contraction coupling, mitochondrial energetics, and induced‐‐release (ECME‐RIRR) model consisting of the oxidative CaMKII activation module, whereas CaMKII (−) represents the previous ECME‐RIRR model that does not incorporate the oxidative CaMKII activation module. shunt=0.1 and pacing cycle length=2 seconds.

One unique characteristic of our ECME‐RIRR model is its capability to simulate sustained mitochondrial oscillations50, 56 (Figure S2A), allowing examination of the dynamics of AP upon mitochondrial repolarization.5, 52 As shown in Figure 5A, after ΔΨm repolarization, AP EADs surprisingly remained on the first several (eg, 7 in this simulation) beats, even though mdROS had reduced to basal levels (Figure S2A). The sustained EADs then turned to intermittent EADs and eventually became normal APs. [Ca2+]i (Figure 5B) and activation of INa,L (Figure 5C) followed the same pattern. [Na+]i (Figure 5D) and fraction of phosphorylated Na+ channels (Frac_NaP) (Figure 5E) gradually decreased during the repolarization phase. However, [Na+]i did not completely return to the predepolarization level, causing gradual [Na+]i accumulation along the progression of mitochondrial oscillations (Figure S2B). In the absence of mdROS‐induced oxidative CaMKII activation, [Na+]i remained relatively constant during mitochondrial oscillations (Figure S2C).

Figure 5

Dynamics of action potential (A), cytosolic Ca2+ concentration ([Ca2+]i) (B), Na+ current (INa) (C), cytosolic Na+ concentration ([Na+]i) (D), and fraction of phosphorylated Na+ channels (Frac_NaP) (E) during the first mitochondrial repolarization. shunt=0.1 and pacing cycle length=2 seconds.

Importantly, with the progression of mitochondrial oscillations, the time needed for AP to return to normal morphology during ΔΨm repolarization increased. When shunt was further increased to 0.14, EADs (constant and intermittent) were maintained throughout the whole repolarization phase after the third depolarization (Figure S3). This behavior seemed to be attributed to the elevated CaMKII activation and augmentation of INa,L. Thus, we analyzed the correlation between the number of sustained or intermittent EADs and the peak [Na+]i during the state transition (eg, from sustained EADs to intermittent EADs). Results show that the numbers of sustained EADs and intermittent EADs were both closely correlated with the peak [Na+]i, especially under more severe stressed conditions (ie, shunt=0.14) (Figure 6).

Figure 6

Correlation between the number of sustained early afterdepolarizations (EADs) (A and B) or intermittent EADs (C and D) and peak Na+ concentration during mitochondrial repolarization at different shunts (0.1 for A and C and 0.14 for B and D). Black solid lines represent regression and red dash lines represent 95% confidence bands; pacing cycle length=2 seconds.

Effect of Blocking Na+ or Ca2+ Handling Channel on mdROS‐CaMKII Activation–Induced EADs

Next, we examined the effect of completely blocking individual Na+ or Ca2+ handling channel on mdROS‐mediated oxidative CaMKII activation–induced AP abnormality during mitochondrial depolarization. Under control conditions (ie, physiological mdROS production or shunt=0.02), blocking INa,L had no effect on AP upstroke but slightly shortened APD. The effects on [Ca2+]i, [Na+]i, INaCa, ICaL, and the peak of INa were negligible (data not shown). Under pathological mdROS production conditions (eg, shunt=0.14), 100% elimination of INa,L abolished EADs, transient INaCa reverse, and ICaL reactivation, accompanied by reduced [Ca2+]i and [Na+]i overload (Figure 7A, blue lines).

Figure 7

Effect of complete blockage of late Na+ current (INa,L) (A), Na+‐Ca2+ exchanger current (INaCa) (B), or L‐type calcium channel current (ICaL) (C) on action potential (1), cytosolic Ca2+ concentration ([Ca2+]i) (2), cytosolic Na+ concentration ([Na+]i) (3), INaCa (4), INa (5), and ICaL (6) under normal and mitochondrial depolarization conditions. shunt=0.14 and pacing cycle length=2 seconds.

Similar to INa,L inhibition, completely blocking INaCa suppressed ICaL reactivation and the subsequent Ca2+‐induced Ca2+ release, which prevented Ca2+ elevation and abolished the EADs (Figure 7B, blue lines). INaCa blockage also reduced INa,L enhancement and [Na+]i (Figure 7B), which was consistent with published data.55 It is worth mentioning that although transient blockage seems beneficial, long‐term INaCa inhibition may cause significant alteration of [Ca2+]i and [Na+]i homeostasis and eventually lead to abnormal APs (such as EADs).

Our model simulations show that blocking ICaL also eliminated the oxidative CaMKII activation–induced EADs, which was consistent with previous experimental data showing that L‐type Ca2+ channel inhibitor suppressed oxidative stress–induced EADs.39 Lack of ICaL activation facilitated phase 2 AP repolarization, resulting in significant APD shortening that hindered the subsequent Ca2+‐induced Ca2+ release and Ca2+ overload. Consequently, Na+/Ca2+ exchanger inward current enhancement was suppressed (Figure 7C). The outcome of ICaL blockade in the absence of mdROS bursts were shortened APD, abolished AP plateau, and diminished Ca2+ transients (data not shown), which agreed with previous studies.57, 58

Effect of Antioxidant Treatment on mdROS‐CaMKII Activation–Induced EADs

We then examined whether antioxidant treatment, such as reducing mitochondrial ROS production or increasing ROS scavenging could eliminate mdROS‐CaMKII activation–induced EADs. In the simulations shown in Figure 8, shunt was initially set to the basal level (0.02) for 5 seconds, then increased to 0.1 or 0.14, and finally reduced to 0.02 at 700 seconds. Increasing shunt caused sustained mitochondrial oscillations and correlated fluctuations of [Na+]i and Frac_NaP. Interestingly, both [Na+]i and Frac_NaP increased gradually with the progression of mitochondrial oscillations, with the increases more evident at higher shunt (Figure 8A and 8B). Reducing shunt from 0.1 to the basal level rendered [Na+]i and Frac_NaP to their initial values and eliminated EADs (Figure 8, red lines). However, when the preceding shunt was higher (ie, 0.14), reducing ROS production at 700 seconds failed to normalize the elevated [Na+]i and Frac_NaP and suppress EADs (Figure 8, blue lines). Interestingly, we found that reducing mdROS production (shunt, from 0.14 to 0.02) earlier (eg, at 350 seconds) converted sustained EADs to intermittent EADs. When shunt reduction was induced earlier still (eg, 200 seconds), [Na+] overload and the oxidative stress–induced EADs were eliminated (Figure 9).

Figure 8

Effect of reducing shunt on mitochondrial‐derived reactive oxygen species (mdROS)‐Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation–induced early afterdepolarizations (EADs). In these simulations, shunt was set as 0.02 (black line), 0.10 (red line), or 0.14 (blue line) during 0 to 700 seconds and 0.02 thereafter. A, Cytosolic Na+ concentration ([Na+]i), (B) fraction of phosphorylated Na+ channels (Frac_NaP), and (C) action potentials of the last 2 beats. Pacing cycle length=2 seconds.

Figure 9

Effect of timing of reducing shunt on mitochondrial‐derived reactive oxygen species‐Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation–induced early afterdepolarizations (EADs). In these simulations, shunt was initially set as 0.14 and then reduced to 0.02 at 200 seconds (black line), 300 seconds (red line), or 700 seconds (blue lines). A, Fraction of phosphorylated Na+ channels (Frac_NaP). B, Action potentials of the last 2 beats. Pacing cycle length=2 seconds.

Another strategy to reduce oxidative stress is to use antioxidant scavengers. In our model, this can be achieved by increasing et_SOD, a parameter that represents the total amount of superoxide dismutase. Similar to reducing shunt, increasing et_SOD, when introduced sufficiently soon after mitochondrial depolarization (eg, at 200 seconds), successfully reduced elevated [Na+]i and suppressed EADs. However, increasing ROS scavenging later (eg, 300 and 700 seconds) failed to eliminate EADs (Figure S4). Taken together, those simulations imply that: (1) altered ion (eg, [Na+]i and [Ca2+]i) homeostasis plays a critical role in mdROS‐CaMKII activation–induced EADs, and (2) timely antioxidant treatment is critical for its antiarrhythmic effect with the sooner the better.

RyRs Oxidation Effect for Inducing Arrhythmias

In a recent computational study, we showed that mdROS can induce abnormal Ca2+ cycling and elicit erratic APs by directly activating RyRs and inhibiting SERCA. Here, we examined whether concurrent oxidative RyRs activation would exacerbate the effect of oxidative CaMKII activation on Ca2+ mishandling and AP abnormality. As shown in Figure 10A, mdROS induced during phase 2 of the AP, when modeled to activate CaMKII only, caused an EAD. Adding the effect of mdROS on SR Ca2+ handling (eg, RyRs activation and SERCA inhibition) converted the single EAD to multiple EADs. Notably, when the mdROS burst was induced during phase 4 of the AP, concurrent oxidative RyRs and CaMKII activation elicited a DAD, which was otherwise barely seen with oxidative CaMKII activation alone (Figure 10B). Concurrent RyRs oxidation by mdROS also exacerbated Na+ and Ca2+ overload during mitochondrial depolarization (Figure 10C through 10F).

Figure 10

Concurrent mitochondrial‐derived reactive oxygen species (mdROS)–mediated oxidative ryanodine receptor (RyR) activation and oxidative Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation on action potential (A and B), cytosolic Ca2+ concentration ([Ca2+]i) (C and D), and cytosolic Na+ concentration ([Na+]i) (E and F). In simulations of (A, C, and E) mdROS bursting was induced at phase 2 of the action potential and in (B, D, and F) mdROS bursting was induced at phase 4 of the action potential. Pacing cycle length=2 seconds and shunt=0.1. CON: shunt = 0.02, or no mitochondrial‐derived reactive oxygen species bursting.

Discussion

In the present study, we expanded our recently published guinea pig cardiomyocyte ECME‐RIRR model47 by incorporating mdROS‐induced oxidative CaMKII activation and slow Na+ channel phosphorylation. Our new model was able to replicate previous model simulations (eg, sustained mitochondrial oscillations) and experimental data (eg, rate dependence of H2O2‐induced EADs and ROS‐induced increases in intracellular Na+ and Ca2+). We then simulated how the endogenous mitochondrial‐derived oxidative stress (ie, mdROS) may influence CaMKII activity and subsequently alter cardiomyocyte ion homeostasis and APs. Our main findings are: (1) mdROS‐mediated oxidative CaMKII activation–induced augmentation of INa,L alone is sufficient to elicit EADs; (2) mdROS‐CaMKII activation–induced EADs can be sustained even when mdROS reduces to a physiological level; and (3) mdROS burst‐induced EADs can be suppressed by antioxidant treatment only when it is given within a timely window.

It has been proposed that the proarrhythmic effect of oxidative CaMKII activation is attributed to its capability to phosphorylate multiple ion channels/transporters underlying Ca2+ handling and AP. However, the detailed mechanistic pathways remain incompletely understood, partially because of the difficulty in experimentally dissecting the contribution of individual ion currents. For instance, although CaMKII activation of INa,L has been implicated to be involved in oxidative stress–induced EADs,39, 44, 59 whether this INa,L augmentation alone can induce EADs under oxidative stress has never been examined. Therefore, in this computational study, we developed an ECME‐RIRR model that considered only the direct modulatory effect of CaMKII oxidation on INa. Model simulations showed that oxidative CaMKII activation–induced augmentation of INa,L, which is comparable to experimental data (Figure 2B), successfully elicits EADs in a cardiomyocyte exposure to increased mdROS. Further analysis suggests that the augmented INa,L causes EADs by altering both membrane potential and intracellular ion (eg, Na+ and Ca2+) homeostasis. In particular, our simulations revealed that the mdROS‐CaMKII activation–induced EADs involve the following: (1) increased INa,L leads to APD prolongation and AP reverse; (2) the AP reverse causes a shift in Na+/Ca2+ exchanger activity (reverse mode) and ICaL reactivation; (3) reactivation of ICaL triggers Ca2+‐induced Ca2+ release and results in a larger Ca2+ transient, which further augments ICaL via a dynamic positive feedback mechanism; and (4) the large Ca2+ increase activates the forward mode INaCa and CaMKII‐mediated INa,L, collectively resulting in EADs.

It is worth mentioning that while our model suggests that direct CaMKII activation of ICaL is not required in mdROS‐CaMKII activation–induced EADs, ICaL reactivation, caused by AP reverse, plays a critical role in the EAD generation, as blocking ICaL eliminates the mdROS‐CaMKII activation–induced EADs. Interestingly, blocking ICaL also caused APD shortening. This finding is different from that reported in our recent computational studies47 focusing on mdROS‐induced abnormal SR Ca2+ handling, in which blocking ICaL suppressed EADs but did not reduce AP prolongation (as compared with normal mdROS). This suggests that the ionic mechanisms underlying oxidative CaMKII activation and oxidative RyRs activation–mediated arrhythmogenesis are different. We also found that oxidative CaMKII activation alone cannot generate DADs but concurrent oxidative RyRs and CaMKII activations can. As previous experimental studies38, 39 have recorded both EADs and DADs in H2O2‐perfused isolated cardiomyocytes, it is likely that both pathways are presented in cardiomyocytes undergoing oxidative stress. In addition, our channel blocking simulations indicated that INaCa activation is also involved in the generation of oxidative CaMKII activation–induced EADs. Thus, although we showed that oxidative CaMKII activation–induced INa,L augmentation is capable of eliciting EAD, the actual arrhythmogenic effects of mdROS are clearly multifactorial and new antiarrhythmic treatments targeting both ion channels/transporters/proteins and mitochondria are essential.

Another intriguing finding from the present modeling study is that the mdROS‐CaMKII activation–induced EADs may not terminate immediately upon mitochondrial repolarization, even though mdROS has been reduced to basal physiological levels. While the phenomenon of sustained EADs postmitochondrial repolarization in cardiomyocytes needs further experimental verification, it may indeed occur in cells undergoing oxidative stress such as ischemia reperfusion, as a result of the specific property of CaMKII as “a memory molecule.”60 The memory refers to the autophosphorylation‐mediated sustained CaMKII activation even after the dissociation of Ca2+/CaM or the fall of Ca2+ concentration to baseline levels, which is essential for memory storage in the brain. Recently, Song et al61 showed that short‐term (5 minutes) ROS exposure caused persistent (more than 60 minutes) activation of ICaL in isolated rat cardiomyocytes, likely via the oxidative stress–induced sustained CaMKII activation, indicating that CaMKII may act as a redox‐sensitive “memory molecule” in cardiomyocytes. Notably, our model simulations showed that the duration of persistent EADs, or the proarrhythmic “memory” of CaMKII activation, is linked to the severity of mitochondrial dysfunction: the higher the mdROS bursting, the stronger the memory. Permanent memory might form if mitochondrial malfunction lasts long enough; in this case persistent arrhythmias will occur (Figure 8). Thus, as suggested by our antioxidant treatment simulations (Figure 9 and Figure S4) the ideal antiarrhythmic treatment would require the intervention be given timely and before the permanent memory of CaMKII is formed. Our computational analysis further showed that there is a close correlation between peak [Na+]i and the robustness of CaMKII's memory, or the durations of the sustained and intermittent EADs during mitochondrial repolarization, suggesting that cytosolic [Na+]i may be used as a risk factor of oxidative CaMKII activation–mediated arrhythmogenesis. The gradual [Na+]i accumulation and sustained EADs facilitated by the oxidative stress–induced CaMKII‐dependent INa,L augmentation has also been reported by Wagner et al.38

In addition to antioxidant treatment, we also examined several other possible antiarrhythmic strategies such as blocking Na+ or Ca2+ channels. Although our simulations showed that blocking INaCa, ICaL, or INa,L all eliminated the mdROS‐mediated oxidative CaMKII activation–induced EADs, their antiarrhythmic roles should be further assessed experimentally, as long‐term ion channel inhibition may break ion homeostasis and induce new arrhythmogenic substrates. For instance, it has been shown that inhibition of Na+/Ca2+ exchanger–mediated Ca2+ extrusion increases Ca2+ spark frequency in resting cardiac myocytes. Long‐term ICaL inhibition, especially by nondihydropyridine Ca2+ channel blockers, can cause a shortening of APD and reduce cardiac contractility and conduction. In addition, the role of Ca2+ channel blockers in ventricular arrhythmias is limited and less well defined.62, 63, 64 Compared with INaCa and ICaL inhibition, blocking the INa,L reduces depolarizing current during the plateau phase of the AP and thus may be more potent and safer. In line with this, it has been reported that a selective INa,L inhibitor, GS‐458967, prevents APD prolongation without affecting AP upstroke velocity in guinea pig ventricular myocytes.65, 66 Ranolazine, another Na+ channel blocker, has been shown to reduce EAD and DAD occurrence in various settings where INaL is enhanced.59

Model Limitations

As our major goal was to examine whether, and if so, mdROS‐CaMKII activation–induced INaL augmentation can induce EADs, the present model only incorporates CaMKII‐dependent phosphorylation of Na+ channels. However, it is well appreciated that many other ion channels/transporters such as LCCs, RyRs, and K+ channels can be phosphorylated by CaMKII.67, 68 CaMKII can increase ICaL and SR Ca2+ release, thereby exacerbating Ca2+ overload and increasing the risk of arrhythmogenesis. Moreover, the activity of LCCs, RyRs, Na+ channels, and K+ channels can be directly influenced by mdROS.69 The effects of oxidative CaMKII activation and direct oxidation on ion channels and homeostasis as well as AP may or may not overlap. Furthermore, previous studies from our laboratory and others have shown that deregulated cytosolic ion handling perturbs mitochondrial energetics and leads to oxidative stress, which can positively feedback on CaMKII activity. Finally, recent studies suggested that CaMKII may directly phosphorylate mitochondrial ion channels such as the Ca2+ uniporter,23 thus altering mitochondrial ion homeostasis and bioenergetics.23 Those components can be added to the ECME‐RIRR model in the future. That being said, lack of these mechanisms should have little impact on the present study, as our main goal was to develop a computational model to examine whether the mdROS‐mediated CaMKII activation can induce EADs in cardiomyocytes and to understand the underlying ionic mechanisms.

Conclusions

The present study provides a novel computational tool to quantitatively investigate the proarrhythmic effects of mdROS‐mediated oxidative CaMKII activation in cardiomyocytes. The results indicate that CaMKII activation is sufficient to initiate downstream molecular events that promote aberrant Ca2+ handling and abnormal APs by sensing elevated mitochondrial‐derived oxidative stress. Our simulations also underscore the importance of timely treatments in the context of oxidative CaMKII activation–induced arrhythmias.

Sources of Funding

The authors acknowledge funding support from China Scholarships Council (grant No. 201506250131) and Key Technologies Research and Development Program of China (grant No. 2017YFC0110400) (to Yang) and National Institutes of Health/National Heart, Lung, and Blood Institute R01HL121206A1 (to Zhou).

Disclosures

None.

Table S1. Sarcolemmal Membrane Ionic Currents

Table S2. Ca2+ Handling System

Table S3. SR Ca2+ Dynamics

Table S4. CaMKII Activation Module

Table S5. Ionic Concentrations Balance Equations

Table S6. Force Generation Model

Table S7. Mitochondrial Membrane Potential (ΔΨm)

Table S8. Energy Metabolism System

Table S9. General Parameters

Table S10. Sarcolemmal Membrane Current Parameters

Table S11. Na+/K+ Pump Parameters

Table S12. Nonspecific Channel Current Parameters

Table S13. Background Ca2+ Current Parameters

Table S14. Sarcolemmal Ca2+ Current Parameters

Table S15. Sarcoplasmic Reticulum Ca2+ ATPase Parameters

Table S16. L‐type Ca2+ Current Parameters

Table S17. Ca2+ Release Channel Parameters

Table S18. Force Generation Parameters

Table S19. Cytoplasmic Energy Handling Parameters

Table S20. Tricarboxylic Acid Cycle Parameters

Table S21. Oxidative Phosphorylation Parameters

Table S22. Mitochondrial Ca2+ Handling Parameters

Table S23. ROS‐Induced‐ROS‐Release Parameters

Table S24. ROS Diffusion Model Parameters

Table S25. CaMKII Activation Parameters

Table S26. Markov Late Na+ Channels Parameters

Table S27. Na+ Channels Phosphorylation Parameters

Table S28. States Variables Initial Values

Figure S1. Top: The scheme of Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation model modified from Foteinou et al.1 In this model, there is only 1 inactive state (state I). CaMKII can be activated via binding Ca2+/calmodulin (CaM), autophosphorylating, and oxidizing. Bottom: The scheme of Markov model for CaMKII activation–dependent Na+ channel phosphorylation. We hypothesized that the phosphorylated rate was proportional to CaMKII autophosphorylation rate and the fraction of activated CaMKII. The dephosphorylation rate was a constant, which was fit using the experiment data from Wagner et al.2

Figure S2. A, Increasing shunt from 0.02 to 0.1 caused sustained mitochondrial oscillations and cyclic reactive oxygen species (ROS) bursts in a cardiomyocyte. B, Dynamics of cytosolic Na+ concentration ([Na+]i) during mitochondrial oscillations in the presence of oxidative Ca2+/calmodulin‐dependent protein kinase II (CaMKII) oxidation. C, Dynamics of [Na+]i during mitochondrial oscillations in the absence of oxidative CaMKII oxidation.

Figure S3. After the third depolarization, the early afterdepolarizations (EADs) were sustained throughout the whole repolarization phase. Shunt=0.14, pacing cycle length=2 seconds.

Figure S4. Effect of increasing reactive oxygen species (ROS) scavenging (ie, et_SOD) on mitochondrial‐derived ROS (mdROS)–Ca2+/calmodulin‐dependent protein kinase II (CaMKII) activation–induced early afterdepolarizations (EADs). et_SOD was increased from 1.43×10−3 mmol/L (baseline value) to 1.8×10−3 mmol/L at 200 seconds (black line), 300 seconds (red line), or 700 seconds (blue line). A, Fraction of phosphorylated Na+ channels. B, Action potentials of the last 2 beats. shunt=0.14 and pacing cycle length=2 seconds.

Click here for additional data file.