Experiments were performed using the human induced pluripotent stem cell (hiPSC) line UMGi014-C clone 14 (isWT1.14) provided by LC (Stem Cell Unit, University Medical Center Göttingen). The iPSC generation and characterization for pluripotency and genomic stability was described previously (18). The line was reprogrammed from somatic cells of a healthy 35-year-old Caucasian male individual. Human iPSC cultures were maintained in Stem MACS iPS Brew XF Medium (Miltenyi Biotec; #130-104-368) in 17 μg/cm² Growth Factor Reduced Matrigel-coated (Corning; #354230) 6-well dishes in a humidified normoxic incubator (37°C, 5% CO₂). Cultures were routinely passaged in colonies at a ratio of 1:16 every 4–5 days after dissociation using Versene Solution [0.48 mM ethylenediamine tetraacetic acid (EDTA)] (Thermo Fisher; #15040033). Cells were plated in Stem MACS iPS Brew XF Medium and 1 μl/ml of 2 mM Thiazovivin (Millipore; #420220) in DMSO (final concentration 2 μM) for the first 24 h. The culture medium was changed daily with 2 ml per well Stem MACS iPS Brew XF Medium.

HiPSCs were differentiated along the ventricular lineage via the modulation of the WNT signaling pathway as previously described (19). Briefly, single cells were harvested using Versene Solution and plated on 6-well dishes at 120,000–160,000 per well into a final volume of 2 ml per well. Differentiation was started when the hiPSC cultures reached a confluency of 80–95% using 3 ml per well Cardio Differentiation Medium [RPMI 1640 (with GlutaMAX and HEPES) (Thermo Fisher; #72400021) supplemented with 0.2 mg/ml L-ascorbic acid 2-phosphate (Sigma; #A8960) and 0.5 mg/ml human recombinant albumin (Sigma; A9731)] and freshly added 4 μM CHIR99021 (Millipore; #361559; 0.4 μl/ml of 10 mM stock solution in DMSO). After 24 h, medium was exchanged to Cardio Differentiation Medium. On day 2, medium was changed to 3 ml Cardio Differentiation Medium supplemented with freshly added 5 μM IWP2 (Millipore; #681671; 1 μl/ml of 5 mM stock solution in DMSO) for 2 days and afterward medium was changed to 3 ml Cardio Differentiation Medium for another 2 days and then again medium was changed to 3 ml Cardio Differentiation Medium for another 2 days. From day 8 onward, medium was changed to 2 or 3 ml of Cardio Culture Medium RPMI 1640 (with GlutaMAX and HEPES) supplemented with the final concentration of 1 × B-27 Supplement (from 50×; Thermo Fisher; #17504044) per well every 2 or 3 days, respectively. On day 15, cells were detached using 0.25% trypsin-EDTA solution (Thermo Fisher; #25200056) and re-plated in 2 ml Cardio Culture Medium supplemented with 20% fetal bovine serum (Thermo Fisher; #10270106, Lot no: 2243865, South American FBS) and 2 μM Thiazovivin at 1 × 10⁶ cells per well in 6-well dishes and medium was changed to Cardio Culture Medium the following day and again on day 18. On day 20, cardiac myocyte selection (20) was performed by changing medium to 2 ml Cardio Selection Medium RPMI 1640 (without glucose and HEPES) with 0.2 mg/ml L-ascorbic acid 2-phosphate and 0.5 mg/ml human recombinant albumin, as well as a final concentration of 4 mM lactate/HEPES (1:250 from 1 M stock) for 2 days. On day 23, medium was replaced with 2 ml Cardio Selection Medium. Starting on day 25, cultures were maintained in Cardio Culture Medium, with regular media changes every 2–3 days. Cells were re-plated after day 30 at 750,000 cells per well in 6-well dishes or at 160,000 cells per well in 24-well dishes for MTT. Experiments were performed on 60-day differentiated cells (hiPSC-CMs).

Cultures were treated with or without 5 μM chloroquine (Sigma; #C6628) (21), 5 μM hydroxychloroquine (Sigma; # H0915), 5 μM remdesivir (Biosynth Carbosynth; AG170167), 25 μM lopinavir/ritonavir (Sigma; #SML1222-10MG, #SML0491-10MG), and lopinavir/ritonavir/8 U interferon-β (Sigma; #IF014) for 24 h. After treatment, cells were analyzed as described below. Drug concentrations chosen are based on literature data and are below the 50% cytotoxic concentrations (22, 23). Incubation of cardiac myocytes with drugs for 24 h under normal culture conditions did not result in cytotoxicity.

Cell viability was determined using the MTT assay, which measures the reduction of thiazolyl blue tetrazolium bromide (MTT) into an insoluble formazan product by the mitochondria of viable cells. In brief, hiPSC-CMs were seeded into 24-well plates at a density of 160,000 cells per well. The cells were treated with various concentrations of drugs or DMSO as a vehicle control. After 24 h incubation, 50 μl 0.5 mg/ml MTT solution (Sigma-Aldrich; Merck KGaA) was added to each well, followed by further incubation for 4 h. The medium was then removed, and the formazan crystals were dissolved in 500 μl solubilization buffer (10% SDS in 0.01 M HCl). The absorbance was measured at 570 nm on a plate reader (Perkin Elmer, EnSpire Multimode Plate Reader). The relative cell viability was expressed as a percentage of the control group.

Total RNA was isolated from cultured cells by using QIAzol Lysis Reagent (Qiagen; #79306) according to manufacturer’s instructions. RNA sequencing (RNA-seq) was carried out via a commercially available long non-coding RNA service (BGI, Shenzhen, China). Briefly, total RNA was fragmented into short fragments and ribosomal RNA was removed. The cDNA synthesis was performed using random priming. Double-stranded cDNA was purified and enriched by PCR amplification, after which the library products were sequenced using BGISEQ-500.

Total RNA was isolated from cultured cells by using QIAzol Lysis Reagent (Qiagen; #79306) according to manufacturer’s instructions, and reverse-transcribed into complementary DNA (cDNA) by using iScript cDNA Reverse Transcription Kit (Biorad; #1708891). Quantitative real-time PCR was performed using iTAQ SYBR Green PCR Kit (Biorad; #1725124) according to the manufacturer’s instructions.

Paired-end rRNA-depleted sequencing data were analyzed in detail. After initial quality assessment, low quality regions and adapter sequences were removed with Flexbar (24) (version 3.5). Residual rRNA reads were removed using Bowtie2 with an rRNA sequence-based index (25). Principal read mapping against the ENSEMBL human reference genome build 100 (hg38) was performed with the STAR RNA-seq aligner (26) (version 2.7.5a). Mapped reads were assigned to genes using the Rsubread package (27) (version 2.2.6). Quality of sequencing data and mapping results was assessed with MultiQC (28) (version 1.9). Differential gene expression was analyzed with edgeR (29) (version 3.30.3). Gene Ontology (GO) analyses were performed using topGO version 2.40.0 with all genes having an RPKM ≥ 1 throughout all samples acting as background list. Pathway analyses were performed using the PathVisio software (30) (version 3.3.0) and individual pathways provided by WikiPathways (31). Heatmaps were generated using the ComplexHeatmap package (32) (version 2.5.4). All downstream analyses were performed with R version 3.6.3.

Cell culture experiments were performed in at least three independent experiments with at least three biological replicates per experiment. Statistical analysis was performed using GraphPad Prism 7.0 (Graphpad Software Inc.¹) or R. Data values are mean ± standard error of the mean (SEM). For statistical analyses, when only two conditions where compared, unpaired two tailed t-test was used.

Incubation of 60-day differentiated hiPSC-CMs with a range of concentrations of the drugs chosen based on literature data for 24 h under normal culture conditions did not result in cytotoxicity as assessed by MTT assay (Figure 1A). Next, hiPSC-CMs were treated with the drugs at selected, relevant experimental doses, which are most commonly used. Total RNA was isolated using QIAzol Lysis Reagent and RNA samples (RNA integrity number [RIN] ≥ 9.6) were subjected to deep sequencing library preparation and RNA sequencing. Principal component analysis of RNA-seq data showed a clear clustering of samples by treatments (Figure 1B). Specifically, samples treated with LR, LRI, and Rem cluster far apart from each other as well as the control samples. Investigation of the differentially expressed genes in each candidate treatment versus vehicle control yielded noticeable differences in the number of differentially expressed genes (Figure 1C). The numbers of differentially expressed genes in CQ and HCQ treatments were comparable with 350–400 genes (FDR ≤ 0.05, |log₂ fold change| ≥ 0.5) differentially expressed with CQ or HCQ compared to vehicle control. The number of differentially expressed genes was higher by one order of magnitude for LR, LRI, and Rem treatment compared to control; moreover, the number of upregulated and downregulated genes was balanced. We continued to specifically investigate differentially expressed genes shared between LR, LRI, and Rem. On the one hand, overall, nearly 900 genes were shared between the three treatments (Figure 1D). On the other hand, LR and LRI share more than 3,700 genes, with 2,054 differentially regulated genes that were specific to the LRI treatment. The number of genes specific to LR was one order of magnitude less, while around 1,000 differentially expressed genes were specific to Rem. Further inspection of the top differentially expressed genes for LR and LRI showed a strong downregulation of Caveolin-3 (CAV3) that has been implicated in the biogenesis of t-tubules (33) and moreover been shown to be associated with cardiac hypertrophy and heart failure when expression is decreased (34) (Figures 1E,F). On the other side of the spectrum, LR treatment showed a strong upregulation of the transcript of calcitonin gene-related peptide (CALCA), a protein that is secreted from the heart during ischemia or simulated ischemia (35, 36). LRI treatment yielded in upregulation of two genes of the 2′,5′-oligo(A) synthetase family, OAS1 and OAS2 (Figure 1F), which are known to be interferon-inducible (37) and thus is in line with the treatment. Interestingly, one of the top upregulated transcripts after Rem treatment is MIR4430 (Figure 1G), a hardly characterized microRNA that however was recently associated with other repurposed drugs to combat COVID-19 in an in silico study (38). A comprehensive list of significantly differentially expressed genes for each comparison is provided in Supplementary Table 1. Moreover, we set up an interactive web portal for visualization of differential gene expression results, which can be accessed via https://covid19drugs.jakobilab.org.

To gain insights into the biological basis for the differences in gene expression, we performed GO-Term enrichment analyses for each of the differential gene expression experiments and identified key pathways. While the number of differentially expressed genes is lower in CQ and HCQ when compared to LR, LRI, and Rem, pathway analyses showed significant enrichment and upregulation of GO categories of cholesterol biosynthesis, cholesterol homeostasis, and tricarboxylic acid cycle. In contrast, genes in GO categories for sarcomere organization, muscle filament sliding, and cardiac conduction were downregulated with HCQ and in part with CQ treatment (Figures 2A,B). Decrease of gene products in these categories may underlie reduced contractile function. Strikingly, for LR we identified a strong upregulation of the endoplasmic reticulum (ER) unfolded protein response (Figure 2C). During LRI treatment we observed an overall downregulation throughout all enriched GO molecular process categories. Specifically, cardiac muscle cell development and homophilic cell adhesion were strongly downregulated (Figure 2D). However, for both, LR and LRI the enriched GO categories showed both, up and downregulation to comparable extents, while for Rem we observed a more global trend to downregulation in the enriched GO categories, with the notable exception of the PERK-mediated branch of the unfolded protein response, which was strongly upregulated (Figure 2E).

Next, we performed hierarchical clustering based on the RNA-seq data from LR, LRI, and Rem treatments, mapped genes involved in these key pathways onto the heatmap, and observed a clear grouping based on treatments (Figure 3). Compellingly, we identified a compact cluster of around 400 genes upregulated in LRI (Figure 3), many of which are associated with type I interferon signaling and thus clearly differentiating between the LR and LRI treatments. Furthermore, we recognized an enrichment for upregulated genes involved in the ER stress response for LR and LRI treatments compared to control. In contrast to the ER stress response genes upregulated with LR and LRI treatments, several other key pathways showed large-scale downregulation of genes such as cardiac muscle contraction, regulation of heart contraction, sarcomere organization, and heart development (Figure 3). While downregulation of genes involved in these pathways was most pronounced with LRI treatment, LR treatment showed a similar trend, although with lower log₂ fold changes. In comparison to LR and LRI, only few genes of these pathways show downregulation with Rem treatment. Moreover, the log₂ fold changes of those genes are lower compared to LR and LRI, thus indicating a less significant contribution of Rem treatment to potentially unfavorable changes in these gene programs.

To obtain further insights into the details of the altered pathways, we combined differential gene expression results and pathway structures. We performed the pathway analysis using KEGG/wikipathway due to the ability to directly integrate not only the gene names into pathways, but also directly map expression values into the graphical pathway. This way, we were able provide more information within the representative pathway as a Reactome pathway analysis could have provided. We selected a panel of 11 key ER stress response genes and profiled their expression after LR, LRI, and Rem treatment (Figure 4A, FDR ≤ 0.05, log₂ fold change ≥ 0.5). We observed a strong upregulation of the ER stress response genes for LR and LRI that was less pronounced in Rem-treated samples. Specifically, mesencephalic astrocyte-derived neurotrophic factor (MANF) and heat shock protein family A (Hsp70) member 5 (HSPA5) show significant downregulation after Rem treatment but are upregulated after LR and LRI treatment. To validate our findings, we performed qRT-PCR on a set of eight ER stress response genes and found, in line with the RNA-seq data, upregulation of nodal ER stress response regulators, such as C/EBP homologous protein (CHOP), activating transcription factor 4 (ATF4), X-box binding protein 1 (XBP1), and endoplasmic reticulum to nucleus signaling 1 (ERN1) in LR and LRI compared to control samples (Figure 4B). With Rem treatment, only CHOP, ATF4, and XBP1 showed upregulation, while HSPA5 and MANF were downregulated, which is in line with the RNA-seq data. Furthermore, we overlaid gene expression data with a pathway representation of the ER stress pathway and observed an upregulation of ER stress response genes at several key positions in the pathway, such as XBP1, CHOP, HSPA5, ERN1, ATF6, and PERK (Figure 4C). These results demonstrate dysregulation of the ER stress response with LR and LRI treatments.

Our initial global pathway analysis (Figure 3) showed downregulated genes in LR and LRI treated samples that were associated with cardiac muscle contraction, regulation of heart contraction, and sarcomere organization. Based on these GO term associations, we curated a list of 100 genes, profiled their expression in detail (Figure 5A, FDR ≤ 0.05, |log₂ fold change| ≥ 0.5), and observed that 80% of these genes show decreased expression after treatment with LR and LRI, while only 20% show increased expression. In contrast, only 14 of the 100 selected genes show overall significant differential expression with Rem treatment, including the upregulated troponin T1, slow skeletal type (TNNT1), ATPase sarcoplasmic/endoplasmic reticulum Ca2+ transporting 1 (ATP2A1), and Myozenin-1 (MYOZ1). We validated the RNA-seq data by performing qRT-PCR for two sets of genes, comprising heavy and light myosin chains (Figure 5B). We observed a strong decrease in myosin heavy and light chain transcripts for LR and LRI, whereas the heavy chain transcripts were more strikingly downregulated than the light chain transcripts in LR (Figure 5B). In LRI-treated cells, we observed a strong decrease in expression of both types of myosin chains. In contrast, Rem treatment significantly decreased MYL3 and MYL4, while MYL2 and MYH7 were only slightly, but significantly decreased, and MYH6 levels were unchanged compared to control (Figure 5B).

We subsequently examined a transcript panel of troponins, tropomyosins, and cardiac muscle alpha actin and found a general downregulation after LR and LRI treatment with a striking effect on cardiac actin (ACTC1, Figure 5C). In contrast, Rem treatment resulted in upregulation of tropomyosin 1 and 2 (TPM1, TPM2), while TNNC1 (troponin C1, slow skeletal and cardiac type) and TNNT2 (troponin T2, cardiac type) were slightly but significantly downregulated and ACTC1 and troponin I3, cardiac type (TNNI3) were significantly downregulated compared to vehicle control. These findings suggest significant negative effects of LR and LRI on contractile components of cardiac myocytes.

The anti-malarial activity of CQ is mainly attributed to the diffusion of CQ into lysosomes, which neutralizes the pH and becomes trapped in lysosomes by protonation, thus resulting in the inhibition of normal autophagy activity (43). Moreover, CQ is also used as adjuvant in several cancer therapy trials which employ the autophagy inhibiting properties of CQ. In this context, a recent study revealed a role for cholesterol biosynthesis in maintaining lysosomal integrity under stress; since inhibiting autophagy interferes with processing extracellularly derived cholesterol esters, thus making cells dependent on the cholesterol biosynthesis pathway (44). Moreover, lysosomal membrane cholesterol decreases permeability to water and ions which suppresses swelling and destabilization under osmotic stress. Intriguingly, in our study, we observed that with CQ and HCQ treatment, most of the enriched pathways are downregulated, except for the regulation of the cholesterol biosynthetic process, where CQ- and HCQ-treated samples show strong upregulation for key genes. We found key enzymes for cholesterol synthesis, 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMGCR), hydroxymethylglutaryl-CoA synthase (HMGC1), farnesyl pyrophosphate synthase (FDPS), and farnesyl-diphosphate farnesyltransferase 1 (FDFT1) robustly upregulated with CQ and HCQ treatment. Thus, in line with a previous report (44), our findings may indicate that CQ-treated cells upregulate the cholesterol biosynthesis in order to compensate for reduced processing of cholesterol esters and to counteract CQ-toxicity by adapting the cholesterol biosynthesis pathway. Nevertheless, CQ or HCQ may lead to lower levels of cholesterol, as HCQ was recently found to lower total cholesterol in a large study of patients with systemic lupus erythematosus and rheumatoid arthritis, where CQ and HCQ are common treatment options (45). In contrast, low cholesterol concentrations in COVID-19 patients have recently been linked to more severe outcomes (46, 47). However, due to the short duration of treatment of acute COVID-19 patients compared to long term patients with rheumatoid arthritis treated with CQ or HCQ, the cholesterol-lowering effects might not play a role in progression of COVID-19 in these patients.

The autophagy-inhibiting properties of CQ were found to affect several cellular processes such as bioenergetics (48). Experiments in primary rat cortical neurons showed that the inhibition of autophagy by CQ increased mitochondrial DNA damage and at the same time decreased bioenergetics (48). CQ was found to reduce glycolysis activity, as well as decrease intermediate products of the tricarboxylic acid (TCA) cycle and key components of glutaminolysis (48). In our study, we observed an upregulation of genes enriched for TCA activity with CQ treatment. Indeed, this upregulation may be consistent with an upregulation of glutaminolysis itself as it was previously found that succinate, fumarate, and malate were not affected by CQ-treatment, thus hinting at adaptive changes of the TCA cycle (48). Therefore, our study may contribute to uncovering the transcriptional changes of the TCA cycle components upon treatment with autophagy inhibitors such as CQ or HCQ.

Treatment of hiPSC-CMs with LR or LRI showed significant changes in key cardiac gene programs such as Ca²⁺ handling and sarcomere organization. The downregulation of several key genes encoding subunits of cardiac ion channels we observed such as KCNQ1, KCNH2, and CACNA1C, is in line with findings that show QT prolongation and torsade de pointes in patients treated with protease inhibitors (PI) (49), such as lopinavir and ritonavir. Moreover, the strict downregulation of other key categories of cardiac function such as cardiac muscle cell development with LR and LRI is coordinate with a recent study describing risks of bradycardia with LR treatment for COVID-19 patients (50). Similarly, we observed perturbations in the regulation of PP1, a critical regulator of cardiac function that mediates restoration of contractility to basal levels after beta-adrenergic stimulation. Dysregulation of PP1, in turn, has been suggested to contribute to impaired function of the heart (51), thus highlighting unfavorable effects of LR and LRI treatment on hiPSC-CMs.

Recent guidelines recommend the use of low- to moderate-dose statins for patients with one or more CVD risk factor as preemptive measure (52). However, treatment with statins together with LR or LRI, should be implemented with selected statins such as pitavastatin or pravastatin due to known drug-drug interactions between statins and antiviral PIs (53), thus requiring special attention for COVID-19 patients on statins.

When comparing LR and LRI, we observed specific effects of INF-β as sharply defined enrichment of genes specifically associated in the type I interferon response for LRI. The interferon response employs double-stranded RNA-activated protein kinase (PKR) to reduce viral replication via phosphorylation of eIF2α and subsequent reduction of protein synthesis (54). Interestingly, genes induced by INF-β were shown to be associated with ER stress (55) and while in general we observed that the addition of INF-β increased the effects of LR treatment on gene expression, this was especially pronounced for key genes of the ER stress response, such as CHOP, which has been shown to play a role at the interface between ER stress and CVD (56). While the ER stress response exerts an initial protective effect, the strength and duration of activation of the ER stress response will determine if the response will eventually switch to the proapoptotic phase and yield to cell death (56). Alterations in protein folding demand, such as those that occur during cardiac ischemia, hypertrophy, and remodeling, result in homeostatic imbalance in ER, causing activation of the ER stress response and thus induction of translation inhibition and gene expression tailored to stress conditions in the ER (57). We and others have shown that in cardiovascular disease, stresses such as oxidative stress or hypoxia can perturb ER homeostasis and activate the ER stress response, which induces expression of proteins that can function to protect the myocardium (58). Both, the interferon response and ER stress response are known to be activated during viral infection and are linked by the phosphorylation of the α subunit of translation initiation factor eIF-2 (eIF2α) (59, 60), which results in inhibition of synthesis of viral proteins. The upregulation of the ER stress response genes in LR-treated cells, in addition to the ER stress response the viral infection itself, may negatively affect the replication of the virus. This effect might further be enhanced by IFN-β treatment as part of the LRI regime, adding the burden of the activation of the interferon response to the already activated ER stress response. Taken together, our results show elevated expression of key ER stress response genes in LR and even more pronounced in LRI that might result in cell death depending on the duration of LR/LRI treatment.

The activation of the ER stress response caused by PIs was studied in the context of the HIV/AIDS, where PIs are routinely employed as treatment option. While long-term studies for HIV treatment and prevention with LR report general long-term safety (61), cardio-metabolic side effects in the heart have been reported (62). Specifically, lopinavir, which has been shown to induce the highest levels of ER stress amongst PIs (41), may cause adverse effects that have been observed in macrophages (63) and hepatocytes (64). Moreover, clinical studies linked ER stress-associated diseases like metabolic syndrome to patients administered PIs over longer terms (41). Thus, the combination of potential adverse effects and unfavorable dysregulation of key cardiac gene programs after LR and LRI treatment likely outweigh negative effects of LR and LRI on viral infection. Besides dysregulation of critical cardiac gene programs, we observed significant dysregulation of several G protein-coupled receptor-associated (GPCR) genes which in turn might lead to different cardiovascular pathologies, such as hypertrophic and fibrotic remodeling of left and right cardiac ventricles and systemic and pulmonary hypertension (65).

In hiPSC-CMs treated with Rem, we detected a slight, but significant upregulation of some of the ER stress response genes, specifically CHOP, ATF4, and XBP1. Interestingly, we found MANF is significantly downregulated in Rem compared to LR, LRI, and even vehicle-treated control, even though ATF6, a known inducer of MANF (66) is significantly upregulated after Rem treatment. Similarly, HSPA5, a key chaperon in the ER stress response is also significantly downregulated in Rem. Interestingly, HSPA5 which encodes the GRP78 protein, has been identified as a potential target for the treatment of Ebola virus (67) and is currently discussed as a potential treatment for COVID-19 (68, 69). Thus, the downregulation of HSPA5 might be one cause for the more promising results of treating COVID-19 with Rem such as the Adaptive COVID-19 Treatment Trial (ACTT-1) found that patients receiving Rem treatment tend to profit from a significantly shorter time to recovery compared to placebo (70). Available data on the effect of Rem on the heart is limited to a single study of 681 patients infected with the Ebola virus, where one of the patients suffered hypotension and subsequent death by cardiac arrest (71), which might be in line with our observation that several genes associates with hypotension show dysregulation, however, the effect of LR and LRI is more pronounced.