Animals were treated in accordance with the Technion Ethics Committee. The experimental protocols were approved by the Animal Care and Use Committee of the Technion (Ethics number: IL-001-01-19 for rabbits and 002-01-19 for mice).

SAN cells were extracted from New Zealand white rabbits weighing 2.3–2.7 kg. Rabbits were sedated with ketamine (0.1 ml/kg) and xylazine (0.1 ml/kg) and anesthetized with 200 mg/ml sodium pentobarbital diluted with heparin, administered through an intravenous cannula. The hearts were quickly removed and placed in warm (37°C) Tyrode’s solution containing: 125 mM NaCl, 5.6 mM KCl, 1.2 mM NaH₂PO₄, 24 mM NaHCO₃, 5.6 mM glucose, 21 mM MgCl₂, and 21.8 mM CaCl₂, bubbled with 95% O₂ and 5% CO₂. The SAN tissue was processed as previously described (Davoodi et al., 2017). Briefly, SAN cells were dispersed by gentle pipetting through a fresh KB solution containing 70 mM L-Glutamic acid, 30 mM KCl, 10 mM KH₂PO₄, 10 mM HEPES, 20 mM taurine, 10 mM glucose, 21 mM MgCl₂, and 0.3 mM EGTA (pH 7.38, with KOH), filtered through a 150 μm mesh.

For experiments involving cultured SAN cells, the cells were diluted in serum-free culture medium mixed with salts and blebbistatin containing 116 mM NaCl, 5.4 mM KCl, 0.8 mM MgCl₂, 0.9 mM NaH₂PO₄, 5.6 mM glucose, 20 mM HEPES, 1.8 mM CaCl₂, 26 mM NaHCO₃, 20% M199 (Sigma, supplemented with 5 mM creatine, 2 mM L-carnitine, 5 mM taurine, 0.1% ITS, and 1% penicillin and streptomycin and titrated to pH 7.4 with NaOH, at 37°C) and 25 μM blebbistatin. The suspended cells were seeded and incubated (37°C, 90% humidity, 5% CO₂, Galaxy 170R, Eppendorf) as previously described (Segal et al., 2019). After 1 h the medium was replaced with serum-enriched culture medium containing 116 mM NaCl, 5.4 mM KCl, 0.8 mM MgCl₂, 0.9 mM NaH₂PO₄, 5.6 mM glucose, 20 mM HEPES, 1.8 mM CaCl₂, 26 mM NaHCO₃), 20% M199 (Sigma, supplemented with 5 mM creatine, 2 mM L-carnitine, 5 mM taurine, 0.1% insulin-transferrin-selenium-X, 4% fetal bovine serum, 2% horse serum, and 1% penicillin and streptomycin and titrated to pH 7.4 with NaOH at 37°C) and 25 μM blebbistatin.

For experiments using whole SAN tissue, adult (12–14 weeks, 25–30 g) male C57BL mice were anesthetized with sodium pentobarbital (50 mg/kg, intraperitoneal (i.p.)) diluted with heparin. The hearts were quickly removed and placed in a 37°C Tyrode’s solution containing 140 mM NaCl, 1 mM MgCl₂, 5.4 mM KCl, 1.8 mM CaCl₂, 5 mM HEPES, and 5 mM glucose, pH 7.4 (titrated with NaOH). The SAN tissue was isolated from the intact heart as previously described (Neco et al., 2012; Wang et al., 2017). Briefly, the SAN and the surrounding atrial tissue were dissected and pinned down in custom-made silicone-covered optical chambers, bathed in the above SAN tissue Tyrode’s solution.

Human-induced pluripotent stem cells (hiPSCs; clone 24.5) were generated from human dermal fibroblasts using Sendai virus CytoTune-iPS 2.0 and Sendai Reprogramming Kit, #A16517 (Thermo Fisher, Waltham, MA, United States) for the transfection of Yamanaka’s four factors: 4 Oct, Klf4, c-Myc, Sox2, as previously described (Takahashi and Yamanaka, 2006; Novak et al., 2015). Before biopsy collection, the 42-year-old healthy female donor signed a consent form according to approval #3116 by the Helsinki Committee for experiments on human subjects at Rambam Health Care Campus, Haifa, Israel. hiPSCs were differentiated into cardiomyocytes (hiPSC-CMs) by modulating Wnt/β-catenin signaling, as previously described (Lian et al., 2013).

Ca²⁺ fluorescence was imaged using a LSM880 confocal microscope, equipped with a 40x/1.2 water immersion lens, at 37 ± 0.5°C, as described before (Davoodi et al., 2017). Briefly, cells were excited with a 488 nm argon laser and fluorescence emission was collected with LP 505 nm. Images were acquired from spontaneously beating SAN cells in line scan mode (1.22 ms per scan; pixel size, 0.01 μm), from the long axes of pacemaker cells, from spontaneously beating hiPSC-CMs and from SAN cells within the primary spontaneously beating SAN tissue, where each cell was imaged before (control) and after drug administration.

Rabbit SAN cells were loaded with 5 μM Fluo-4 AM (Thermo Fisher, 30 μmol/L) for 20 min in the dark, at room temperature, and washed with 37°C HEPES buffer containing:140 mM NaCl, 5.4 mM KCl, 5 mM HEPES, 10 mM glucose, 22 mM MgCl₂, 21 mM CaCl₂ (pH 7.4, with NaOH). Experiments were performed on fresh cells less than 6 h after isolation or on cultured cells after washout of blebbistatin. hiPSC-CMs were loaded with 2.5 μM Flou-4 AM for 20 min in the dark, at room temperature, and then washed with Tyrode’s solution (at 37°C) containing 140 mM NaCl, 5.4 mM KCl, 10 mM HEPES, 2 mM Na-pyruvate, 10 mM glucose, 1 mM MgCl₂ and 2 mM CaCl₂ (pH 7.4, with NaOH). Intact SAN preparations were loaded with Fluo-4-AM (30 μmol/L) for 1 h at 37°C and place on a shaker at 60 RPM. The tissues were washed twice with 37°C Tyrode’s solution containing 140 mM NaCl, 1 mM MgCl₂, 5.4 mM KCl, 1.8 mM CaCl₂, 5 mM HEPES, and 5 mM glucose, pH 7.4 (titrated with NaOH).

Ca²⁺ analysis was performed using a modified version of Sparkalizer (Davoodi et al., 2017). The signal (F) was normalized by the minimal value between beats (F₀). Ca²⁺ transients were semi-automatically detected and Ca²⁺ sparks were manually marked. Beat interval was calculated as the mean beat interval of each cell, and the variability was calculated as its standard deviation (SD). Relaxation time (50% and 90%) and spark parameters were automatically calculated by the software as we have previously described (Davoodi et al., 2017).

To predict the mechanisms underlying the effect of HCQ on the spontaneous beat rate, simulations were performed using our previously published rabbit SAN cell model (Behar et al., 2016). Beat interval was calculated as the average distance between membrane potential peaks after model stabilization. Numerical integration was performed using the MATLAB (The MathWorks, Inc. Natick, MA, United States) ode15s stiff solver, and the model simulations were run for 100 s to ensure that steady state was attained. Computation was performed on an Intel(R) Core (TM) i7–4790 CPU @ 3.60 GHz machine with 8 GB RAM. The source code of the numerical model is available at: http://bioelectric-bioenergetic-lab.net.technion.ac.il.

HCQ and IBMX were purchased from Sigma-Aldrich.

Data are presented as mean ± standard error of the means (SEM), and compared using paired or unpaired student’s t-test or one-way analysis of variance (ANOVA). Differences were considered statistically significant at p < 0.05. To determine whether the shapes of two distributions were the same, the Kolmogorov-Smirnov two-sample test was applied to the standard scores (z-scores) from each sample. The standard scores were obtained by subtracting the sample-specific mean and dividing the result by the sample-specific standard deviation for each group.

We first tested the effect of HCQ on the spontaneous beating interval and rhythm of isolated single rabbit SAN cells. At 1 μM and 3 μM, HCQ increased beat interval; 24.7 ± 7.8% and 170.8 ± 22.7%, respectively, compared to control (Figure 1A). Further, both doses increased the beat interval variability, as quantified by the SD (Supplementary Table S1) and each increased the beat interval distribution compared to control (Figure 1B). Next, we investigated whether the increase in beat interval is associated with changes in the coupled clock function. As illustrated by the representative Ca²⁺ transients (Figures 1C–E) and Supplementary Table S1, we found that 1 µM and 3 μM HCQ increased the 50% and 90% Ca²⁺ transient relaxation times. Notably, the changes in sarcoplasmic reticulum (SR) local Ca²⁺ release (LCR) parameters (including its period) reflect alterations in both membrane currents and the Ca²⁺ clock machinery (Yaniv et al., 2014). Specifically, increased in HCQ concentration caused a progressive decrease in LCR length and Ca²⁺ signal magnitude of individual LCRs (Supplementary Table S1). Figure 1F depicts that the distribution of Ca²⁺ signal magnitude of individual LCRs does not change compared to control. Both HCQ concentrations increased LCR period (Supplementary Table S1) along with increased distribution compared to control (Figure 1G). In addition, the slope and correlation coefficient between beating interval and LCR period was altered by HCQ (Figure 1H); whereas at 1 μM, a correlation existed (changed from 0.87 in control to 0.86) with a different slope as compared to the control, at 3 μM, low correlation was found between the two (0.76).

To explore the mechanisms underlying HCQ-induced bradycardia, the effect of 1 μM HCQ on a single SAN cell was simulated using our previously published computational model (Behar et al., 2016). Based on the finding that HCQ (1–10 μM) inhibits I_f in guinea pig SAN cells (Capel et al., 2015), we firstly focused on this key pacemaker current. Decreasing I_f density by 35% (the maximal decrease in I_f observed at 10 μM HCQ in guinea pig SAN cells) led to only a 7% increase in beat interval (Figures 2A,B), which is smaller than the 25% increase observed experimentally (Figure 1A). Additionally, a 70% decrease in I_f density (the maximal decrease that still produces stable results by the model) increased the beat interval by 20%, which agrees with the effect found experimentally (Supplementary Figure S1). However, no changes in L-type Ca²⁺ current (I_Ca,L) were obtained, in contrast to experimental results (Capel et al., 2015). Because the model considers both direct and indirect (through changes in Ca²⁺ influx rate and cAMP/PKA levels) effects of I_f on pacemaker function, we concluded that a direct effect of HCQ solely on I_f cannot explain the effect of HCQ on the beat interval.

The guinea pig experiments further showed that 10 μM HCQ (3 μM were not tested) attenuated the I_Ca,L and potassium current (I_Kr) densities by 12% and 35%, respectively (Capel et al., 2015). Reduction of only the I_Ca,L density by 10% increased the beat interval by 16%, which is smaller than the increase observed here experimentally. Moreover, no change was observed in I_f, in contrast to the experimental results (Capel et al., 2015) (Supplementary Figure S2). Next, we tested the effect of reducing both I_Ca,L and I_f (as observed in experiments with 1 μM HCQ (Capel et al., 2015)) densities by 10% (Figure 2C). The resulting 21% increase in beat interval agreed with the effect found experimentally (Figure 1A). In contrast to the experimental observations, adding a direct effect on I_Kr shortened the beat interval. Under both simulated conditions (i.e., a direct effect on I_f with/without an effect on I_Ca,L) there was no effect on I_Kr. In contrast, due to an indirect simulated effect of HCQ on these current densities and kinetics, the transient outward potassium current (I_to) and the sodium-calcium exchanger current (I_NCX) were decreased by 30% (Figures 2D,E). Additionally, the model was used to explore Ca²⁺ kinetics in different intracellular compartments. 1 μM HCQ did not affect cytosolic Ca²⁺ (Figure 2F). However, the model also predicted reductions in junctional sarcoplasmic reticulum (JSR) Ca²⁺ (Figure 2G) and SR Ca²⁺ release (Figure 2H), resulting in decreased Ca²⁺ concentration in the subspace (Figure 2I), and smaller Ca²⁺ release from the SR.

Because the model predicts indirect changes in I_to and I_NCX, we used the computational model to test the direct effect of reduction in the conductance of these currents. Reduction of I_to by 80% decreased the beat interval by 3% (Supplementary Figure S3). Thus, its effect on spontaneous beating rate is negligible. Reduction of I_NCX by 40% increased the beat interval, in agreement with the effect found experimentally (Figure 1A). However, no change in I_f was documented, in contrast to the experimental results (Capel et al., 2015).

To demonstrate that the effect of HCQ on automaticity is not unique to rabbit SAN cells, experiments were repeated using spontaneously beating hiPSC-CMs. Whereas 1 μM (−5.2 ± 13.5% compared to control, n = 8) and 3 μM (5.4 ± 22.4% compared to control, n = 8) HCQ did not affect beat interval, 10 μM HCQ increased beat interval by 55 ± 11.8% (Figure 3A), as well as beat interval variability (Supplementary Table S2) and beat interval distribution (Figure 3B). As illustrated by the representative examples, 10 μM HCQ increased the 50% and 90% Ca²⁺ transient relaxation times (Figures 3C–E; Supplementary Table S2). Additionally, 50% and 90% Ca²⁺ transient relaxation times and increased their distribution by 10 μM HCQ (Figures 3E,F).

Mouse SAN tissue were treated with HCQ to determine its effect at the tissue level. As observed for the hiPSC-CMs, whereas 1 μM (5 ± 11.2% compared to control, n = 17) and 3 μM (4 ± 14.5% compared to control, n = 11) HCQ did not affect beating interval, 10 μM HCQ increased beat interval by 84.3 ± 10% (Figure 4A). Further, 10 µM HCQ increased beat interval variability (Supplementary Table S3) and its distribution (Figure 4B). Figures 4C,D show representative Ca²⁺ transients in control and in 10 μM HCQ-treated SANs; the 50% and 90% Ca²⁺ transient average relaxation times and the distribution were comparable to control (Figures 4E,F).

In an attempt to alleviate the bradycardic effect of HCQ, we searched for an FDA-approved drug that can prevent this effect, by restoring Ca²⁺ kinetics as well as I_f and I_Ca,L conductance (Capel et al., 2015). As a proof-of-concept, IBMX decreased SAN cell and tissue beat interval by increasing Ca²⁺ -dependent phosphorylation activity, and decreased beat interval in pacemaker cell and tissue in response to an increased beat interval associated with aging (Yaniv et al., 2016) or genetic mutation such as catecholaminergic polymorphic ventricular tachycardia (Arbel-Ganon et al., 2020). We choose an IBMX concentration that will reverse the beating interval shortening when concomitantly delivered with HCQ based on previous reports (Vinogradova et al., 2008).

Figure 5A shows that addition of 10 μM IBMX to 1 μM HCQ prevented the increase in beat interval induced by the latter (14.6 ± 10% compared to control) and prevented the increased beat interval variability (Supplementary Table S4) and increased beat interval distribution compared to pre-HCQ levels (Figure 5B). As visualized by the representative examples (Figures 5C–E), in the presence of IBMX, 50% and 90% Ca²⁺ transient relaxation times were comparable to pre-HCQ levels (Supplementary Table S4). Further, IBMX + HCQ restored average LCR length and average and distribution of Ca²⁺ signal of individual LCRs to pre-HCQ levels (Figure 5F; Supplementary Table S4). In the presence of HCQ and IBMX, the average (Supplementary Table S4) and the distribution of the LCR period (seen in Figure 5G) were similar to pre-HCQ levels. In addition, HCQ + IBMX partially restored the correlation (x = 0.76) between beat interval and LCR period, in contrast to the lack of correlation found in HCQ-treated cells (Figure 5H). We also tested the effect of application of 10 μM IBMX to 3 and 10 μM HCQ. Supplementary Figure S5 shows a shift in the dose response curve in response of HCQ in the presence of IBMX. Moreover, in the presence of 10 μM H-89 (PKA inhibitor) the effect of IBMX on restoration of HCQ associated bradycardia was lower.

Next, computational modeling was conducted to identify putative mechanisms responsible for the ability of IBMX to prevent the effect of HCQ on the pacemaker. The model suggests that in response to IBMX and HCQ, the beat interval decreased by 15% compared to control, in contrast to the 21% increase (measured experimentally) caused by HCQ alone (Figure 6A). Addition of IBMX to HCQ prevented the decrease in I_f (Figure 6B, 8% reduction), I_Ca,L (Figure 6C, 7% reduction), I_to (Figure 6D, 6% increase) and I_NCX (Figure 6E, 4% increase). As found experimentally, adding IBMX restored Ca²⁺ kinetics (Figure 6F) and JSR Ca²⁺ (Figure 6G) and Ca²⁺ release from the SR (Figure 6H). Consequently, Ca²⁺ in the subspace returned to its pre-HCQ levels (Figure 6I).

In agreement with the experiments in the previous models, treatment of hiPSC-CMs with 50 μM IBMX +10 μM HCQ prevented the increase in beat interval (−1 ± 6% compared to control; Figure 7A), as well as the increase in beat interval variability (Supplementary Table S2) and beat interval scattering (Figure 7B). As visualized by the representative Ca²⁺ transients (Figures 7C–E) and quantified in Supplementary Table S2, 50% and 90% Ca²⁺ transient relaxation times and distributions (Figures 7F,G) were comparable to pre-HCQ levels. Note that a 10 μM IBMX and 10 μM HCQ did not reverse the increase in beat interval by 10 μM HCQ alone (5 ± 4%, n = 5).

Finally, in SAN tissue, 50 μM IBMX and 10 μM HCQ prevented the increase in beat interval induced by HCQ alone (−16 ± 6% compared to control). The combined treatment also prevented the increase in beat interval variability (Supplementary Table S3) and beat interval scattering (Figure 8B). As visualized by the representative Ca²⁺ transients (Figures 8C–E) and quantified in Supplementary Table S3, in the presence of IBMX and HCQ, 50% and 90% Ca²⁺ transient relaxation times were comparable to pre-HCQ levels including their distribution (Figures 8F,G). Note that a 10 μM IBMX plus 10 μM HCQ regimen did not reverse the increase in beat interval by 10 μM HCQ alone (10 ± 8%, n = 10).

The exact cardiac origin (i.e., ventricular, sinoatrial or atrial) of the hiPSC-CMs used in this study was not determined. Only cells that beat spontaneously at a rate similar to the human heart were selected for the experiments. The similarity between their response to HCQ or HCQ and IBMX compared to the responses of isolated SAN cells or SAN cells residing in the SAN tissue, suggest that they constitute a reliable and comparable model. Future use of SAN cells derived from hiPSC will clarify this point (Protze et al., 2016). Moreover, although Ca²⁺ and membrane clocks were discovered in hiPSCs (Zeevi-Levin et al., 2012), it remains to be determined whether they are coupled, and, if so, which signaling cascades connect them.

The concentrations of applied HCQ and IBMX were not consistent across the tested models. While a significant effect of 1 μM HCQ on beat interval was shown in single isolated SAN cells, a higher concentration was required to obtain a similar effect in hiPSC-CMs and SAN cells residing in SAN tissue. Similar trend was documented for IBMX. These results are not surprising when considering previous reports on the effects of other drugs on these experimental models (Janardhan et al., 2012; Ben-Ari et al., 2014). The lower sensitivity of drugs in SAN tissue compared to isolated cells may be related to diffusion limitations, specifically of drugs that affect internal signaling or which bind membrane channels from the cytosolic side. Lower sensitivity of hiPSC-CMs to drugs may be related to low densities of membrane channels such as I_f compared to rabbit SAN, and to different levels of cAMP/PKA activity. Future experiments are needed to further characterize these observations.

We predict here an indirect effect of HCQ on I_NCX. Future experiments are needed to verify that it is an indirect effect of HCQ. It is challenging to measure I_NCX because its major contribution is during the non-linear phase of the action potential and performing steady voltage steps results in unphysiological activity. Performing ramp voltage steps is feasible, but the HCQ affects the non-linear slope of the action potential and thus even a ramp step will not be of physiological relevance. Moreover, based on our model observations (Supplementary Figure S6), voltage clamping would induce no significant measurable changes in I_NCX in the presence versus absence of HCQ.