3.6 Computational simulations of HCQ and AZM on human ventricular action potentials
To translate the experimental data describing the multi-ion channel blocking effects of HCQ and/or AZM to human ventricular excitation, the ORd model was implemented to reconstruct the impacts of HCQ alone and combined with AZM, on human ventricular action potentials including APD90 and maximal upstroke velocities.
Figure 8A displays predicted action potentials under control conditions and in the presence of HCQ, and HCQ and AZM combined. It was shown that with the 10 µM HCQ alone, the APD was prolonged by 100 ms as compared to the control condition. With the combined action of HCQ and AZM, a synergistic action of the two drugs on APD prolongation was seen, which was manifested in a greater increase of APD90 of ~300 ms (Figure 8B). Such an APD prolongation can be attributable to the integrated action of reduced repolarisation potassium channel currents of IKr and IK1, and reduced depolarising currents of INa, ICaL and INaCa. A small reduction in action potential maximum upstroke velocities was also observed by both HCQ and HCQ+AZM due to the effects on INa.
DISCUSSION
The main findings of this study are: (1) Reported therapeutic levels of HCQ blocked multiple human cardiac ion channels, particular hERG (mediating IKr) and Kir2.1 (IK1,) and, to a lesser degree, hNav1.5 (INa). AZM potentiated the effects of HCQ on hERG and Kir2.1. (2) Computational molecular docking studies correspondingly predicted HCQ binding to hERG. Accordingly, (3) at the level of intact hearts, HCQ increased QT intervals as well as ventricular AP and Ca2+ transient durations and their heterogeneities. However, (4) HCQ further reduced heart rates, A-V and ventricular conduction as well as increasing conduction heterogeneities. All these effects were accentuated by combining HCQ challenge with AZM. Furthermore, (5) The HCQ+AZM combination, but not HCQ alone, then went on to increase susceptibilities to ventricular arrhythmic events consistent with significant clinical cardiac safety risks. (6) Humanin-silico modelling of ion channel effects of HCQ and AZM recapitulated the altered APDs but not the reduced conduction. This mechanistic electrophysiological data complements ongoing multiple randomized trials evaluating clinical efficacy and cardiac safety profiles of these drugs as a treatment for Covid-19. The findings directly align with recent FDA guidelines concerning their use.
These successive lines of investigation were combined to reach an integrated view of HCQ and AZM action. This began with the approach recommended by the CiPA guidelines (Strauss et al., 2018) in which we examined effects of HCQ and AZM on human cardiac ion channels responsible for the main inward (hNav1.5 (INa) Cav1.2 (ICa,L)) and outward (Kv4.3 (Ito), hERG (IKr), KCNQ1/E1 (IKs) and Kir2.1 (IK1)) currents in the human ventricular myocardium. HCQ and AZM were studied both alone and in combination through a wide concentration range encompassing and extending beyond reported therapeutic levels. The latter typically fell around 1.15-4.5 µM(Durcan, Clarke, Magder & Petri, 2015) and 0.5-0.7 µM(Miller, Fiechtner, Carpenter, Brown, Stroshane & Stecher, 1987) for HCQ in SLE and rheumatoid arthritis patients on a daily 400 mg dose. In contrast, its dose used in Covid-19, 1.2 g/day, would give substantially higher therapeutic concentrations of ~10 µM(Borba et al., 2020). Blood AZM levels following successive daily 500 mg and two 250 mg doses were reported at 0.22 μM(Lai, Ho, Jain & Walters, 2011). HCQ both alone and combined with AZM synergistically inhibited IKr and IK1 with IC50s at the high end of reported therapeutic levels, with some minor inhibition of INa in this range.
Consistent with these experimental findings, computational molecular docking simulations utilizing a recent hERG homotetramer structure (PDB ID: 5va2) demonstrate potential HCQ binding. This is predominantly with Y652 and F656 from one or more hERG subunits. Additional residues involved in interactions with HCQ frequently included F557, L622, T623, S624, and S660. In contrast, the model predicted approximately two-fold weaker AZM-hERG binding.
Others have also explored the electrophysiological effects of HCQ or AZM on some cardiac ion channels but did so at three times the clinically relevant concentration and above (Capel et al., 2015; Lai, Ho, Jain & Walters, 2011). Furthermore, the present study systematically evaluates HCQ and AZM in combination for the first time. Furthermore, the only report on electrophysiological effects of HCQ specifically studied sinoatrial cells, demonstrating significant reductions in hyperpolarization-activated currents (If), with reduced ICaL and IKr observed at 3 µM (Capel et al., 2015). Nevertheless, this HCQ action on If explains its bradycardic effects observed here, further illustrating its broad cardiac electrophysiological actions. Such a bradycardic medication in the presence of a hypotension accompanying sepsis or SARS could potentially exacerbate a septic shock.
At the whole heart level, our optical mapping studies demonstrate that HCQ correspondingly prolongs ventricular APD as well as CaD, to extents further exacerbated by combination with AZM. This fulfils expectations from its inhibition of IKr and IK1 and translates to the observed electrocardiographic QT prolongations. It is interesting to note that even prior to the Covid-19 pandemic, series of case reports have led to both HCQ and AZM being listed as definite causes of prolonged QT interval and risk of torsade de pointes for patients with long QT syndrome on crediblemeds.org (Roden, Harrington, Poppas & Russo, 2020).
Prolonged APD predisposes to early afterdepolarisations (EADs) during the plateau phase of the ventricular action potential, classically attributed to ICaL reactivation (January & Riddle, 1989). EADs are also more common under conditions of intracellular Ca2+ overload, including increased sympathetic drive: ICaL can be enhanced by high intracellular Ca2+ levels and by L-type Ca2+channel phosphorylation by CaMKII(Herring, Kalla & Paterson, 2019). EADs may potentially trigger ectopic beats and initiate arrhythmic events given suitable substrate.
In addition to these recovery phenomena, we also demonstrate for the first time that HCQ reduces ventricular conduction on multi-electrode arrays and optical mapping recording, and compromise atrioventricular and intraventricular conduction, reflected in prolonged PR intervals and QRS durations, to extents accentuated by further addition of AZM. Such findings are consistent with Na channel block (Reiz & Nath, 1986) of the kind exemplified clinically by the CAST trials (Echt et al., 1991) and/or compromised gap junction function(King, Huang & Fraser, 2013). Together, the combination demonstrated here of prolonged and increased dispersions of repolarisation and calcium handling combined with slowed conduction would offer a suitable substrate for sustained re-entrant arrhythmia. Thus, the combined exposure to HCQ and AZM, but not HCQ alone provoked alternans, ventricular tachycardic episodes, and re-entrant excitation following progressively accelerated cardiac pacing.
Finally, a validated human ventricular myocyte mathematical model integrated the observed ion channel properties with the effects on action potential and intracellular calcium handling characteristics in intact hearts. It separated whole heart electrophysiological features that were explicable in terms of the single channel properties and those attributable to altered calcium homeostasis. The modelling reproduced effects of surface ion channel function and their interactions forming a ‘membrane clock’ underlying the measurable electrical activity (Fig. 8C)(Lei & Huang, 2019). It reproduced the action potential prolongation and consequent prolonged QT intervals with HCQ and the synergistic effects of combining HCQ with AZM, attributing these to their combined effects on IKr and IK1. These latter events have been described as feeding forward into a ‘Ca2+ clock’ driving excitation-contraction coupling but whose events exert feedback effects modifying action potential conduction, recovery, and post-recovery stability, exemplified by Ca2+ feedback effects on Na+ channel and gap junction function (Fig. 8C), as well as longer-term effects on channel expression(Huang, 2017; Lei, Wu, Terrar & Huang, 2018). The modelling predicted the increased calcium transient magnitudes and durations particularly with HCQ+AZM (Fig 8C), but only small changes in maximum upstroke velocity attributable to Na channel blockade. The observed conduction slowing nevertheless could arise from a Ca2+-dependent gap junction uncoupling.
These results together provide an integrated basis for the electrophysiological and pharmacological effects of HCQ, AZM and their combination. They thereby underpin recent FDA guidelines cautioning against combined HCQ/AZM administration for treatment of Covid-19 on grounds of cardiac safety. Admittedly, physiological though not pro-arrhythmic effects occurred with HCQ administered alone at the ~10 µM concentrations associated with Covid-19 therapy and not at the lower ~1.0 µM HCQ levels associated with therapy for rheumatoid arthritis and SLE. HCQ alone at therapeutic doses thus did not increase incidences of experimental ventricular arrhythmias in healthy hearts. Nevertheless, safety threshold doses could be reduced by accompanying cardiac injury arising from viral toxicity, sepsis, hypoxia-related myocyte injury, and immune-mediated cytokine storm in Covid-19. Furthermore, AZM potentiated the HCQ toxicity, now producing arrhythmic outcomes. The present findings thus strongly indicate a desirability for monitoring electrocardiographic, including QT, changes and cardiac injury during 4-aminoquinoline-macrolide therapy for Covid-19 patients.