Beta-adrenergic receptor stimulation limits the cellular proarrhythmic effects of chloroquine and azithromycin

Background: The antimalarial drug chloroquine and antimicrobial drug azithromycin have received significant attention during the current COVID-19 pandemic. Both drugs can alter cardiac electrophysiology and have been associated with drug-induced arrhythmias. Meanwhile, sympathetic activation is commonly observed during systemic inflammation and oxidative stress (e.g., in SARS-CoV-2 infection), and may influence the electrophysiological effects of chloroquine and azithromycin. Here, we investigated the effect of beta-adrenergic stimulation on proarrhythmic properties of chloroquine and azithromycin using a detailed in silico model of ventricular electrophysiology. Methods: Concentration-dependent chloroquine and azithromycin-induced alterations in ion-channel function were incorporated into the Heijman canine ventricular cardiomyocyte model. Single and combined drug effects on action-potential (AP) properties were analyzed using a population of 592 models accommodating inter-individual variability. Sympathetic stimulation was simulated by an increase in pacing rate and experimentally validated isoproterenol-induced changes in ion-channel function. Results: At 1 Hz pacing, therapeutic doses of chloroquine and azithromycin (5 and 20 µM, respectively) individually prolonged AP duration (APD) by 33% and 13%. Their combination produced synergistic APD prolongation (+161%) with incidence of proarrhythmic early afterdepolarizations in 53.5% of models. Increasing the pacing frequency to 2 Hz shortened APD and together with 1 µM isoproterenol corrected the drug-induced APD prolongation. No afterdepolarizations occurred following increased rate and simulated application of 0.1-1 µM isoproterenol. Conclusion: Sympathetic stimulation limits chloroquine- and azithromycin-induced proarrhythmia by reducing their APD-prolonging effect, suggesting the importance of heart rate and autonomic status monitoring in particular conditions (e.g., COVID-19).


Introduction
Six-months after its first identification in Wuhan, China in December 2019, Severe Acute Respiratory Syndrome-associated Coronavirus type-2 (SARS-CoV-2) infection (i.e., Coronavirus Disease 2019 / COVID-19) has contributed to more than 500,000 deaths worldwide and has been declared a pandemic with significant global socioeconomic impact (1). At the moment, the exact pathophysiology of the disease remains unclear and no definitive therapy is available. Several drugs are considered effective in preclinical studies and are currently being tested against SARS-CoV-2 in the clinic (e.g., the antivirals lopinavir, ritonavir, and remdesivir; the antimicrobial azithromycin; the antimalarial drugs chloroquine and hydroxychloroquine, and more recently antiparasitic ivermectin) (2)(3)(4). Of those, chloroquine (CQ) and azithromycin (AZM) have gained significant attention due to their high accessibility and low cost. Nonetheless, their effectivity against COVID-19 has not been confirmed by any large clinical trial and their use is controversial. Some studies reported the benefit of those drugs (5-7), while others reported no effect (8,9). This controversy is further complicated by the retraction of papers demonstrating the absence of benefit of these drugs in COVID-19 (8) and the termination of their emergency use by the United States Food and Drug Administration (FDA) due to their potential proarrhythmic effects (10).
CQ is a widely-used antimalarial drug that inhibits multiple cardiac ion-channels (11). It has been suggested to prevent the viral entry, transport and post-entry events in COVID- 19, although the exact mechanisms remain unknown (12). Meanwhile, AZM is a broadspectrum macrolide antibiotic that is believed to potentiate the effect of CQ, reducing the replication capabilities of SARS-CoV-2 (12). Similar to CQ, AZM also inhibits multiple cardiac ion-channels in a dose-dependent manner (11). Therefore, the administration of CQ and AZM, alone or in combination, can prolong the ventricular cardiomyocyte action potential duration (APD) and thereby the QT interval on the electrocardiogram. Excessive QT-interval prolongation has been implicated in drug-induced malignant arrhythmias, such as Torsade de Pointes (10,13,14), by promoting early afterdepolarizations (EADs) and a heterogeneous repolarization substrate.
Computational modeling has increasingly been used in cardiac safety pharmacology to predict the proarrhythmic effect of novel compounds (23)(24)(25). To the best of our knowledge, previous analyses of the potential proarrhythmic effects of CQ and AZM have not considered the role of beta-adrenergic receptor stimulation. Therefore, this study aimed to assess the potential cellular proarrhythmic effects of CQ and AZM in both the absence and presence of beta-adrenergic receptor stimulation using a population of detailed in silico models of ventricular electrophysiology.

Methods
Concentration-dependent CQ and AZM-induced alterations in 7 ion-channels (rapid delayed-rectifier K + (IKr), fast Na + (INa), late Na + (INaL), transient-outward K + (Ito), inwardrectifier K + (IK1) and slow delayed-rectifier K + current (IKs)) (11) (Figure 2) were incorporated into the Heijman canine ventricular cardiomyocyte model (26) with betaadrenergic receptor signaling. A cellular concentration within the therapeutic range of CQ and AZM was selected (5 and 20 µM, respectively (27,28)) and cellular simulations were performed in Myokit (29). The effects of the drugs alone and in combination (assuming independent drug-binding sites) on action potential (AP) properties were assessed. Sympathetic stimulation was simulated by an increase in pacing rate and experimentally validated isoproterenol-induced changes in ion-channel function (26,30). All results are presented during steady-state pacing at the indicated pacing frequencies (after 1000 beats of prepacing). To evaluate the robustness of our findings and assess potential consequences of intra-and inter-subject variability on the electrophysiological effect of CQ and AZM, the maximum conductance of 9 major ionic currents (INa, INaL, ICa,L, IKr, IKs, IK1, Ito, INCX and INaK) were scaled based on a normal distribution with mean 1.0 and standard deviation 0.2, to create populations of models, as previously described (31). In brief, 1000 variants of the model were created and the variants displaying "nonphysiological" AP properties (defined as APD90 or RMP outside the range of 3 standard deviations of experimental APD90 and RMP from (32)) were excluded. In total, 592 out of 1000 models were included. The non-normally distributed data are presented as median and inter-quartile ranges (IQR). The model code is available at www.github.com/jordiheijman. . Furthermore, at pacing rates >1 Hz, AZM slightly hyperpolarized the RMP, which was opposed by the RMP-depolarizing effect of betaadrenergic activation, while CQ with or without ISO consistently showed a slight depolarization of RMP, likely due to its inhibition of IK1 (Figure 2). The RMP modulating effect is attenuated at low pacing rates ( Figure 3B, right panel).  (26) and preventing such phosphorylation resulted in repolarization failure (RF) in the CQ+AZM group in the presence of simulated beta-adrenergic stimulation ( Figure 5).

During
Next, a population-based study was conducted to accommodate intra-and interindividual variability. A population of 1000 models was created by varying 9 ionic currents  (Figure 6A-E).
Finally, the incidence of EADs and RF in the population of models was calculated ( Figure   6F). During 1 Hz pacing, no EAD/RF was documented in the non-treated group, while 6.4% of models in the CQ group, 2.4% of models in the AZM group, and 53.5% of models in the combined group exhibited EADs/RFs. Following the increase in pacing rate to 2 Hz, the incidence of EAD/RF was reduced to 0.5%, 0.7% and 11.5%, respectively. No EAD/RF was observed in any of the groups following the application of ISO in 2 Hz pacing.

Discussion
Here, we investigated the potential proarrhythmic effects of CQ and AZM in the ventricular cardiomyocyte in the absence or presence of beta-adrenergic stimulation using an in silico approach. First, our results indicate that CQ and AZM could significantly prolong the APD even within their therapeutic range. Moreover, their combination resulted in a synergistic APD prolongation, leading to the initiation of proarrhythmic EADs. Second, beta-adrenergic stimulation reduced APD prolongation and prevented EAD formation due to the upregulation of IKs and ICa,L. Finally, our population-based study confirmed the robustness of these findings and showed that beta-adrenergic stimulation completely cancelled the initiation of EADs and RFs in all groups, highlighting a potential important role for beta-adrenergic activity in preventing drug-induced proarrhythmia by CQ and AZM.

Chloroquine and azithromycin exhibit a synergistic APD-prolonging effect
CQ and AZM block multiple ion channels, including the rapid delayed-rectifier K + current (IKr) (11), which dose-dependently prolongs the APD and increases the propensity for EADs, creating a substrate for cardiac arrhythmias. In the clinic, they are known to prolong the QT interval, increasing the susceptibility for life-threatening arrhythmias, such as Torsade de Pointes. Several studies have also reported the potential proarrhythmic effects of CQ and AZM in COVID-19 patients (10,13,14). In this computational study, we confirmed the potentially harmful ventricular APD-prolonging effect of CQ and AZM.
However, within the therapeutic range, the incidence of EADs was relatively low (6.4% in CQ group and 2.4% in AZM group). However, the combination of both drugs, as proposed in the treatment of COVID-19, produced a synergistic APD-prolonging effect that further increased the likelihood of EADs, particularly at slow heart rates, suggesting the need for close monitoring of the QT interval during the administration of these drugs in the clinic.
In agreement, a previous prospective observational study also showed that the maximum corrected QT interval during treatment was significantly longer in the combination group compared to the mono therapy group, highlighting the synergy between CQ and AZM (14).

Beta-adrenergic activation reduces the APD and lowers the cellular proarrhythmic risk of chloroquine and azithromycin
Beta-adrenergic agonists have been used as an antidote against CQ intoxication for a long time (33,34). Their benefit in the management of CQ-induced arrhythmia has been experimentally demonstrated in anaesthetized rats, showing that the CQ-infused group treated with isoprenaline (a selective beta-adrenergic receptor agonist) displayed longer time to arrhythmias and death (35). Conversely, the administration of propranolol (a betaadrenergic receptor blocker) potentiated the electrocardiographic effects of CQ, indicating that beta-adrenergic receptor blockade might render the heart more vulnerable to the actions of CQ (36).
In this study, we demonstrated that beta-adrenergic stimulation could be a potential On the other hand, there has been clear evidence that long-term beta-adrenergic stimulation promotes cardiac remodeling, including hypertrophy, fibrosis and the downregulation of several ion channels via transcriptional and post-translational modifications, potentially creating a substrate for cardiac arrhythmias (22,37,38).
Therefore, transient activation of the beta-adrenergic response may be beneficial against drug-induced proarrhythmia and beta-blockers might not be appropriate under such circumstances. On the other hand, beta-blockers could be used to reduce the detrimental effect of long-term beta-adrenergic stimulation or to reduce the complications of COVID-

19-induced systemic inflammation in the absence of medications with proarrhythmic
behavior.

Limitations of the study
Here, we performed a computational study using an established canine ventricular cardiomyocyte model with beta-adrenergic signaling (26,30). In this study, we used the cellular concentrations of CQ and AZM. However, it can be challenging to correlate these cellular concentrations to the clinically relevant doses due to variability in the pharmacokinetics and -dynamics of the drugs, particularly in severely ill patients. Pharmacokinetics/-dynamics models exist and, in the future, could be implemented to obtain a more precise simulation of the electrophysiological consequences of the drugs.
The drug-induced ion-channel modifications incorporated in this study (Figure 2) were derived from previous publication using heterologous expression systems (in Chinese hamster ovary / human embryonic kidney cells), which could display different results from human cardiomyocytes (11). Since these data are the only available data to date, we assumed that the relative drug effects are retained across species and cell types.

Conclusions
CQ and AZM exhibit a proarrhythmic behavior due to their APD-prolonging effect, with their combination showing a pronounced potentiation of this APD prolongation, posing a bigger threat for cardiac arrhythmias. Acute activation of the sympathetic nervous system prevents CQ-and AZM-induced proarrhythmia by reducing their APD-prolonging effect, highlighting the importance of preserving the beta-adrenergic response in the presence of such proarrhythmic medications and the potential significance of heart-rate and autonomic-status monitoring in particular conditions such as COVID-19.

Author contributions
HS and JH conceived the study. HS performed the computational simulations. HS and JH performed the data analysis and drafted the manuscript. All authors critically revised the manuscript and approved the final version.