Three patients affected by LQTS were studied. One carries the KCNQ1-p.R594Q variant (Mura et al., 2018), associated with the most common form of LQTS (LQT1). One was affected by the Jervell and Lange-Nielsen syndrome (JLNS) carrying the KCNQ1-p.R594Q & KCNQ1-p.R190W variants in compound heterozygosity (Mura et al., 2018); JLNS is one of the most severe forms of LQTS, caused by the presence of homozygous or compound heterozygous variants on KCNQ1 or on KCNE1 associated to congenital deafness in addition to the typical cardiac features (Schwartz et al., 2006). Another patient was affected by the CALM1-p.F142L variant (Crotti et al., 2013) associated with CALM-LQTS, which manifests in infants, is extremely severe, and responds poorly to therapy. These hiPSC-CMs were previously characterized (Rocchetti et al., 2017; Crotti et al., 2019).

hiPSCs from healthy subjects: one was an unrelated bona fide healthy donor (WT (Lee et al., 2021) while hiPSCs from the second donor (WT2) were provided by the Coriell Institute for Medical Research (WTC-11 line, catalog No. GM25256). The electrophysiological maturation state of WT2, JLNS and CALM-LQTS hiPSC-CMs was confirmed with patch clamp in this study and it can be observed by the hyperpolarized resting membrane potential and the prominent upstroke velocity, indirectly indicating a robust presence of I_K1 and I_Na, respectively (Supplementary Figure 1).

hiPSCs were cultured on recombinant human vitronectin (rhVTN) in E8 Flex medium and differentiated on cell culture-grade Matrigel. hiPSCs were differentiated to hiPSC-CMs following a protocol based on the modulation of the Wnt-signaling pathway (Lian et al., 2012), purified through glucose starvation (> 90% CMs) and cryopreserved at days 17–20. Cryopreserved hiPSC-CMs were thawed before each experiment and maintained in culture for 7 days in RPMI medium supplemented with B27 Supplement (RPMI + B27). Data were obtained from 3 independent differentiations for each line. Details on reagents are provided in Supplementary Table 1.

Multiwell MEAs (MultiChannel Systems) were coated with 40 μg/mL bovine fibronectin and processed as previously described (Sala et al., 2017b). hiPSC-CMs were dissociated with TrypLE Select and replated on Multiwell MEAs as confluent monolayers. RPMI + B27 was refreshed every 48–72 h. The recordings started after 1 week and were made at Baseline and at 2, 24, and 48 h after the treatment with HCQ (Figure 1).

HCQ was dissolved in water to obtain a stock solution of 50 mM. After having recorded Field Potentials (FP) values at Baseline, RPMI + B27 in Multiwell MEAs was refreshed with working concentrations of HCQ prepared in RPMI + B27 (1, 10, 100 μM). Vehicle was balanced accordingly and used as a control in each experiment and each Multiwell MEA plate.

Given the significant non-zero slope of the FPD-RR linear fit for the majority of the lines, data were corrected with the clinically used Bazett’s QT correction formula, validated across all ages (Stramba-Badiale et al., 2018). As previously shown, corrected FPDs (cFPD) nullified in the majority of the cases (Sala et al., 2016) the positive correlation with RR when present (Figures 1B,C), normalizing the rate-dependency of FPD. Data are indicated in Supplementary Table 2.

Signs of spontaneous arrhythmias and electrical instability were detected at baseline in MEAs from the JLNS and CALM-LQTS (Supplementary Figure 4).

We first investigated the presence of potential detrimental effects of the exposure to HCQ on the spontaneous activity of hiPSC-CMs monolayers.

We observed a dose-dependent effect of HCQ quantified as an increasing number of wells that ceased to beat with higher concentrations (Figures 2, 3A). Here, a genotype-dependent effect was observed and it became particularly evident in two conditions: after the acute (2 h) exposure to 10 μM HCQ, hiPSC-CMs from the two controls showed no effect on the spontaneous beating while ∼20% of the monolayers from the LQT1, ∼35% of the JLNS and ∼50% of the monolayers from the very severe CALM-LQTS temporarily halted the beating, which only partially recovered at longer timepoints (Figure 2). Similarly, 100 μM HCQ induced a larger termination of the spontaneous beating in monolayers from JLNS and CALM-LQTS, while asymptomatic LQT1 and the two WT seems more resistant to higher concentrations of HCQ (Figures 2, 3A). A similar trend was also observed in isolated hiPSC-CMs paced at 1 Hz, with cells from JLNS and CALM-LQTS exhibiting a lower tolerance to 10 μM HCQ than those from WT2 (Supplementary Figure 1). Next, we developed a 6-point scoring system to incorporate information regarding the FP quality degradation due to HCQ into the classification of the drug effect. This is particularly relevant to provide differential weights to MEAs in which the main temporal FP parameters (i.e., FPD, RR, cFPD, etc.) could still be tracked and quantified but were spoiled by the drug treatment. We observed that HCQ had more severe effects on the FP quality of hiPSC-CMs from the two symptomatic subjects, with minor or modest effects in hiPSC-CMs from the asymptomatic LQT1 and from the healthy controls. These differences could be appreciated already at 1 μM and were particularly relevant at 10 μM. At 100 μM, only hiPSC-CMs from one of the two healthy donors maintained a higher proportion of good FPs despite their FP shapes being significantly modified. The deterioration of the FP signal quality has to be attributed to a significant decrease in action potential (AP) amplitude, AP peak, upstroke velocity and to a trend toward depolarization of the E_diast (Supplementary Figures 1, 2).

After having assessed the detrimental effects of HCQ on FP quality, we investigated the effect of different concentrations of HCQ on MEA temporal parameters to verify whether HCQ could trigger a differential FPD prolongation based on the donor’s genotype. Higher susceptibility to media changes has been observed in the CALM-LQTS at all timepoints, even with vehicle, with spontaneous arrhythmic events often present (Supplementary Figure 4). Treatment with vehicle induced changes (shortening) in the cFPD of WT at 2 h, while it prolonged cFPD in CALM-LQTS at all the timepoints.

HCQ 1 μM induced prolongation in the cFPD of CALM-LQTS at all the timepoints and increased the number of MEAs without electrical activity.

HCQ 10 μM induced shortening in the cFPD of WT only at 2 h, while it significantly increased cFPD in WT2, LQT1, JLNS and CALM-LQTS at all the time points and was paired with a cessation of beating only affecting LQT1, JLNS, and CALM-LQTS.

HCQ 100 μM induced cFPD prolongation in WT, WT2, LQT1 while monolayers from JLNS and CALM-LQTS did not tolerate such high dosage and no electrical activity was observed at all timepoints.

Although the concentrations and the modality of administration (acute vs. cumulative) are different among in vitro experiments and many clinical trials, results for the WT subjects were similar to results from early timepoints published by other investigators with MEAs (TeBay et al., 2021) or in a more sophisticated organ-on-a-chip system (Charrez et al., 2021). Our study further extends these results including a susceptible population of patients affected by cLQTS that may be at higher risk of exhibiting diLQTS and developing TdP. The mean HCQ plasma concentration in subjects treated with HCQ has been measured as 50.3 ng/mL following a single 200 mg dose of HCQ (Plaquenil^®) and a mean peak blood concentration of 129.6 ng/mL (Plaquenil label indication); based on these data, it is reasonable to estimate an in vitro concentration of 0.15–0.387 μM.

Inhibition of SARS-CoV-2 infection in vitro was achieved at concentrations larger than 4 μM (EC₅₀ 4.06 μM or higher) regardless of the multiplicity of infection (MOI) used for SARS-CoV-2, Liu et al. (2020), while the cytotoxic concentration (CC₅₀) in ATCC-1586 cells was around 250 μM. Other studies identified a similar range of concentrations in Vero/VeroE6 cells, with an EC₅₀ of ∼6 μM after 24 h of treatment (Yao et al., 2020) or an IC₅₀ of 2–4 μM at 48 or 72, h respectively (Maisonnasse et al., 2020).

In the RECOVERY trial, patients were treated with 800 mg administered at zero and 6 h, followed by 400 mg starting at 12 h after the initial dose and subsequently every 12 h for the following 9 days or until discharge (RECOVERY Collaborative Group et al., 2020). The simulated whole-blood concentration-time plots derived the theoretical whole-blood concentration of HCQ to span from 1 to 6 μM, with an upper safety bound for whole-blood of ∼10 μM (3 μM at plasma concentration) (White et al., 2020).

The concentrations used in this study appropriately covered clinically relevant concentrations as well as the in vitro effective dosages for SARS-CoV-2 inhibition, but also take into consideration a potential higher tolerance to drugs by hiPSC-CMs and the unknown effect of serum-free cell culture media on HCQ bioavailability.

Concentration-dependent effects were recorded for HCQ in all the lines, with the main modifications being changes in the repolarization peak as well as in the FP amplitude and quality. We could detect indications of proarrhythmic events in many of the MEAs treated with high HCQ concentrations, with the main events being abnormal shape of the repolarization wave, an irregular beat rate, a variable amplitude and presence of multiple depolarization peaks among consecutive beats. Some of these features were already visible at baseline for hiPSC-CMs from the two most severe lines used in this study (i.e., JLNS, CALM-LQTS).

A genotype-specific drug response pattern could also be identified and it aligned with the severity of the underlying genotypes. Our data indicate that hiPSC-CMs from the healthy donors or the asymptomatic LQT1 mutation carrier tolerated higher concentrations of HCQ and showed lower susceptibility to HCQ regardless of baseline in vitro FPD values.

Importantly and in agreement with previous observations (Sala et al., 2016, 2017a), the genotype-specific effects were present and predominant over the different baseline FPD parameters, further confirming that absolute FPD (or AP duration, APD) values alone might provide misleading information when classifying drug effects in different cell lines, with the relative drug effects and a more comprehensive analysis of all FP parameters to take into account the underlying genotypes may provide more relevant results.

Healthy controls showed a higher tolerance for QTc prolongation and higher resistance to high HCQ concentrations. This might have impacted directly the patients: the asymptomatic LQT1 carrier (Female, Basal QTc = 458 ms), for example, was identified in our center only through a family screening of the sibling (JLNS, Female, Basal QTc = 578 ms, symptomatic) and might have been included in one of the COVID-19 clinical trials with HCQ because in most of them the exclusion criteria for QTc were a cutoff of 480 ms. According to our MEA data, this would have meant a significantly higher chance for this patient to experience QTc prolongation, a degradation of the electrical signal (Figure 3C) and a potentially pathological FPD prolongation (Figure 4).

Similarly to Chloroquine (Sánchez-Chapula et al., 2001), also HCQ blocks I_K1, and this could explain the increased beating frequency observed particularly in hiPSC-CMs from WT despite HCQ should inhibit also the funny current (I_f) (Capel et al., 2015); in our experimental settings, this effect did not clearly emerge in hiPSC-CMs from LQTS subjects; however, a drug-induced I_K1 blockade may generate an additive effect on the underlying LQTS, especially on spontaneously beating cells, with the effect on the beating frequency potentially being masked by an additive and abnormal FPD prolongation due to an already compromised repolarization reserve. Single-cell patch clamp data on hiPSC-CMs from WT2, LQTS and CALM-LQTS was performed and confirmed the trend toward depolarization (n.s. in all the lines), a large decrease in the upstroke velocity and a decrease in the plateau amplitude (p < 0.05 in all the lines) that overall could be linked to the decreased FP quality observed in Figure 3. These effects are consistent with patch clamp data in heterologous systems indicating HCQ as a blocker of multiple ion currents (Thomet et al., 2021). Some of the cells in all groups did not tolerate the acute exposure to 10 μM HCQ (Supplementary Figure 1), with a larger effects on cells from JLNS and CALM-LQTS groups; this confirms that isolated hiPSC-CMs have higher sensitivity to drugs than those included in monolayers, and that data from these two models may be interpreted and considered differently in terms of translational relevance (Sala et al., 2016). A trend toward APD shortening (p < 0.05 for WT2, n.s. for LQTS and CALM-LQTS) was observed and, for WT2, was similar to the results at 2 h in Figure 4 and Supplementary Figure 3. For JLNS and CALM-LQTS, the likely higher sensitivity to 10 μM HCQ exhibited by isolated hiPSC-CMs compared to monolayers may have contributed to exacerbate the effects on I_Na and dV/dt_max and thus the trend toward APD shortening.

Drug discovery and drug approval pipelines are already designed to prevent or minimize compounds that show a proarrhythmic potential due to a hERG blocking activity to be approved and commercialized (International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use, 2005). Even though important results have been achieved in the predictive potential of in vitro and in silico systems (Blinova et al., 2017; Li et al., 2017; Passini et al., 2017), comprehensive experimental models capable of recapitulating the complexity of human physiology are lacking and thus several compounds were withdrawn from the market after their approval [e.g., cisapride (Henney, 2000), astemizole (Gottlieb, 1999), terfenadine (Fung et al., 2001), etc.], with cardiovascular toxicity still being a major cause for molecules to be discarded in the preclinical phases of drug development (Ferri et al., 2013). Platforms to guide clinical decision making and orient research efforts are required, particularly in critical situations when there is a compelling need for repurposed compounds or when therapies are being recommended for home use. In recent years, the use of hiPSC-CMs for safety pharmacology has progressively gained momentum and multiple validation strategies have been attempted to generate experimental models with enhanced predictive potential, and recently the CiPA approach or similar strategies have been used to evaluate the proarrhythmic risk of HCQ on hiPSC-CMs from commercial hiPSC lines or healthy individuals (Delaunois et al., 2021). Platforms based on hiPSCs can offer rapid and relevant readouts for safety pharmacology (Sala et al., 2017a) but the possibility of incorporating patient-specific drug responses in addition to those from healthy individuals can increase the predictive capacity of these platforms in conditions where prompt actions are required.