Discovery of SARS-CoV-2 antiviral synergy between remdesivir and approved drugs in human lung cells

SARS coronavirus 2 (SARS-CoV-2) has caused an ongoing global pandemic with significant mortality and morbidity. At this time, the only FDA-approved therapeutic for COVID-19 is remdesivir, a broad-spectrum antiviral nucleoside analog. Efficacy is only moderate, and improved treatment strategies are urgently needed. To accomplish this goal, we devised a strategy to identify compounds that act synergistically with remdesivir in preventing SARS-CoV-2 replication. We conducted combinatorial high-throughput screening in the presence of submaximal remdesivir concentrations, using a human lung epithelial cell line infected with a clinical isolate of SARS-CoV-2. This identified 20 approved drugs that act synergistically with remdesivir, many with favorable pharmacokinetic and safety profiles. Strongest effects were observed with established antivirals, Hepatitis C virus nonstructural protein 5A (HCV NS5A) inhibitors velpatasvir and elbasvir. Combination with their partner drugs sofosbuvir and grazoprevir further increased efficacy, increasing remdesivir’s apparent potency > 25-fold. We report that HCV NS5A inhibitors act on the SARS-CoV-2 exonuclease proofreader, providing a possible explanation for the synergy observed with nucleoside analog remdesivir. FDA-approved Hepatitis C therapeutics Epclusa® (velpatasvir/sofosbuvir) and Zepatier® (elbasvir/grazoprevir) could be further optimized to achieve potency and pharmacokinetic properties that support clinical evaluation in combination with remdesivir.

. Primary screening results identifying compounds increasing antiviral effects of remdesivir (RDV). (a) Assay outline: Vero-E6 cells are added to 384 well plates, treated with DMSO (left panel) or drug (middle panel), infected with SARS-CoV-2 and incubated for 72 h to observe cytopathic effect (CPE; left panel). Effective drug treatment inhibits occurrence of CPE (middle panel). CPE is measured by quantifying ATP content in viable cells using a luminescent assay (Cell-Titer Glo). The right panel shows the cytotoxicity control, treating cells with drugs but without virus. (b) Screening assay performance. Average Luminescence is shown for the Vero E6 primary screen in presence of EC15 of remdesivir (n = 144, 24 wells each from 6 screening plates), error bars indicate standard deviation. "Uninfected": positive control (equivalent to 100% inhibition of CPE), "SARS-CoV-2": negative control, infected and treated with DMSO (equivalent to 0% inhibition of CPE). Z' = 0. 63  www.nature.com/scientificreports/ With this assay, we conducted a primary screen of a library of ~ 1200 FDA-approved drugs with remdesivir in Vero E6 cells, achieving an average Z' of 0.63 + / − 0.04 (Fig. 1b). A parallel screen in the absence of viral infection assessed compound toxicity. The primary screen identified 90 compounds with antiviral activity exclusively in the presence of EC15 remdesivir (Fig. 1d, red). None of the hit compounds showed significant toxicity in cells (Fig. 1e, red). More than 95% of hit compounds were confirmed in an additional Vero E6 assay (Fig. 1f). A secondary screen of these compounds was carried out in the Calu-3 assay; 28 of the initial hit compounds maintained strong antiviral activity in a background of EC15 of remdesivir across both cell lines (Fig. 1g, red).
As all tested compounds are approved drugs annotated with their molecular targets, we conducted a gene set enrichment analysis to identify pathways preferentially targeted by hit compounds in the combinatorial screen (Fig. 2). We observed a statistically significant enrichment of compounds affecting the corticosteroid pathway (Fig. 2, GSEA p = 0.0001), as well as for calcium channel, proton pump and HIV protease modulation (Fig. 2, GSEA p = 0.004, 0.003), all of which have been implicated with antiviral effects in the literature [29][30][31][32] . Without remdesivir, none of the mentioned targets were enriched (p > 0.1, FDR q value > 0. 36) 33 . This suggests that remdesivir makes SARS-CoV-2 uniquely vulnerable to inhibition of otherwise nonessential targets.

Quantitation of synergistic effects with remdesivir in a dose response matrix
To identify the most promising drug combinations for use in the clinic, we conducted a dose-response interaction matrix analysis to quantitatively evaluate the synergy between screen hits and remdesivir. The matrix combined ten concentrations of remdesivir (up to 10 μM) with eleven concentrations of each screen hit (up to 40 μM), allowing us to test CPE in SARS-CoV-2-infected Calu-3 cells for 110 concentration combinations per remdesivir/compound pair. We then used computational zero interaction potency (ZIP) modeling to quantitatively determine if synergy was present 34 . The model combines both Loewe additivity and Bliss independence models,  www.nature.com/scientificreports/ systematically assessing drug interaction patterns that may arise in a drug combination matrix. In this model, a value of < 0 signifies antagonism, 0-10 additive effects, and values > 10 show synergy between compound pairs. Strikingly, 20 compounds showed pronounced synergy with remdesivir in counteracting SARS-CoV-2-induced CPE, with maximal ZIP-scores of 29-87 (Fig. 3, Figure ED2): velpatasvir, elbasvir, dabrafenib, cilostazol, nimodipine, conivaptan hydrochloride, clobetasol, budesonide, drosiprenone, ezetimibe, ivosidenib, selexipag, meprednisone, nifedipine, omeprazole sulfide, quinapril, rifaximin, telmisartan, valdecoxib and zafirlukast. Out of this list of candidates to combine with remdesivir, we prioritized velpatasvir, elbasvir, dabrafenib, cilostazol and nimodipine for detailed characterization based on the strength of the synergistic effect, mechanism of action, safety profile and the likelihood of clinical usefulness in context with best practices for COVID-19 treatment. Velpatasvir and elbasvir are hepatitis C antivirals targeting the HCV nonstructural protein 5 (NS5A), a replication co-factor with no clear homolog in SARS-CoV-2. Dabrafenib is a B-raf inhibitor used for melanoma chemotherapy, with an acceptable safety profile; B-raf inhibitors have been shown to have antiviral effects but have not been reported in the context of SARS-CoV-2 30,35 . Cilostazol is a widely prescribed, generically available PDE3 inhibitor, used to prevent stroke and treat intermittent claudication 36 . Nimodipine is a generically available calcium channel blocker used to treat hypertension with a favorable safety profile, acting on one of the druggable pathways enriched in the screen (Fig. 2) 37 .

Orthogonal validation of prioritized antiviral drug combinations
We next assessed viral infectivity in a tissue culture infectious dose 50 (TCID50) assay, which determines the titer of infectious viral particles that leads to cell death in 50% of assayed wells. In this experiment, Calu-3 cells infected with SARS-CoV-2 were treated with EC15 remdesivir by itself or in combination with velpatasvir, sofosbuvir, elbasvir, grazoprevir, velpatasvir/sofosbuvir (Epclusa®), elbasvir/grazoprevir (Zepatier®), dabrafenib, cilostazol or nimodipine. The TCID50 assays (Fig. 5b,f) confirmed results seen in the CPE screening assay (Fig. 5a,e): on its own, remdesivir at its EC15 had only modest effects and velpatasvir or sofosbuvir had no significant effect; yet in combination, remdesivir/velpatasvir/sofosbuvir reduced the titer of infectious viral particles by ~ 1500-fold (Fig. 5b). Similar results were observed for elbasvir: the viral titer was reduced ~ 1500-fold in combination of elbasvir and remdesivir, with little to no effect from single agent treatment (Fig. 5f). Consistent with earlier results, the co-formulated combination of grazoprevir and elbasvir was synergistic with remdesivir ( Fig. 5f;  Fig. 4d). Dabrafenib, cilostazol and nimodipine also reduced viral titer in presence of remdesivir ( Figure ED3h).
Similar results were obtained when we quantified infected cells by immunofluorescence microscopy. We treated Calu-3 cells with remdesivir and compound, infected cells with SARS-CoV-2, and stained for nuclei (DAPI) and SARS-CoV-2 nucleocapsid protein (N-protein, NP; Fig. 5c,g,i). While EC15 concentrations of remdesivir had little effect on viral replication, as indicated by distinct N-protein staining, its combination with velpatasvir, elbasvir, velpatasvir/sofosbuvir, elbasvir/grazoprevir, dabrafenib, cilostazol or nimodipine strongly reduced the number of infected cells (Fig. 5c,g,i and figure ED3h). In fact, cells treated with the commercial HCV antiviral combinations were statistically not significantly different from uninfected cells (Fig. 5c,g,i; RDV/ velpatasvir/sofosbuvir p = 0.18, RDV/elbasvir/grazoprevir p = 0.07). Analyzing viral genome copy number from the supernatant of infected cells by RT-qPCR further confirmed the drastic effects of combining HCV antivirals with remdesivir in blocking SARS-CoV-2 replication (Fig. 5d,h and ED3j). We conclude that the approved HCV antiviral medications Epclusa® (velpatasvir/sofosbuvir) and Zepatier® (elbasvir/grazoprevir) are strongly synergistic with remdesivir in blocking SARS-CoV-2 replication, significantly reducing viral load of infected cells. www.nature.com/scientificreports/ www.nature.com/scientificreports/ To determine whether candidate compound antiviral synergy with remdesivir could be achieved in a more physiologically relevant cell type, we utilized primary normal human bronchial epithelial cells (NHBE) with transient overexpression of the human ACE2 receptor. As before, we found that treatment with remdesivir (EC15),  www.nature.com/scientificreports/ velpatasvir, sofosbuvir, elbasvir, and grazoprevir as single agents, and in various clinically used combinations, possessed little effect on viral replication relative to the vehicle control conditions (Fig. 6a,b). In contrast, when the EC15 concentration of remdesivir was combined with the above compounds in various concentrations, we observed a significant synergistic effect against SARS-CoV-2 replication (Fig. 6a,b), consistent with the data in previous assays (Figs. 3,4,5). Again, the clinically used coformulations Epclusa® (Velpatasvir/Sofosbuvir) and Zepatier® (Elbasvir / grazoprevir) exhibited the strongest antiviral effect when combined with remdesivir www.nature.com/scientificreports/ ( Fig. 6a,b). These data confirm the screening assay data in Calu-3 and Vero E6 cells and indicate antiviral synergy of HCV NS5A inhibitors with remdesivir can be observed in primary human lung cells infected with SARS-CoV-2.

HCV NS5A inhibitors are inhibiting the SARS-CoV-2 exonuclease proofreader
Next, we studied mechanism of action, focusing on the strongest synergistic compound, velpatasvir. Its described HCV target NS5A is a part of the viral replication machinery, but has no known homolog in SARS-CoV-2 or other coronaviruses [42][43][44][45] . As velpatasvir on its own was not strongly active, we concluded that the potential target had to be non-essential in the absence of remdesivir, but had to become essential as remdesivir terminated RNA strands were generated. We speculated that the observed synergy effect could be explained by inhibition of the SARS-CoV-2 exonuclease proofreader. We used a mass spectrometry-based in vitro assay to investigate the exonuclease function of the SARS-CoV-2 nsp10 and nsp14 complex, which had been experimentally validated using a library of more than 25 known inhibitors for the SARS-CoV-2 polymerase complex 24,[38][39][40][41] . RNA was incubated with pre-assembled SARS-CoV-2 exonuclease complex (nsp14/nsp10) in the absence or presence of varying amounts of velpatasvir. RNA and the cleavage products of the exonuclease reaction were analyzed by matrix-assisted laser desorption/ ionization mass spectrometry (MALDI-TOF MS). The peak at 8165 Da corresponds to intact RNA (8157 Da expected, Fig. 7a). In the absence of velpatasvir, exonuclease activity caused nucleotide cleavage from the 3'-end of the RNA, as shown by the 7 lower molecular weight fragments corresponding to cleavage of 1-7 nucleotides, with only ~ 24% intact RNA remaining (Fig. 7b). Velpatasvir at 25, 50 and 100 μM reduced exonuclease activity in a concentration-dependent manner as shown by the reduced intensities of the fragmentation peaks and www.nature.com/scientificreports/ more prominent intact RNA peak (Fig. 7c-e). This was also observed with HCV NS5A inhibitor elbasvir, as well as other representatives of that drug class, daclatasvir and ledipasvir (Fig. 7g-i). These data indicate that HCV NS5A inhibitors can inhibit the SARS-CoV-2 exonuclease proofreader (nsp14/nsp10) complex, with velpatasvir demonstrating such activity in a dose-dependent manner. www.nature.com/scientificreports/ Next, we determined the effect of velpatasvir on RNA terminated with the nucleotide analog remdesivir. Remdesivir (R)-terminated RNA was incubated with the pre-assembled nsp14/nsp10 complex in the absence or presence of 50 µM velpatasvir. The peak at 8209 Da represent intact R-terminated RNA (expected 8204, Fig. 7j). MALDI-TOF MS analysis demonstrated that in the absence of velpatasvir, exonuclease activity caused cleavage of 1-8 nucleotides from the 3'-end of the RNA as shown by the lower molecular weight fragments (Fig. 7k). When 50 µM velpatasvir was added, exonuclease activity was inhibited as shown by the reduced intensities of the fragmentation peaks and increased peak of the intact RNAs (Fig. 7l). Thus, the SARS-CoV-2 exonuclease activity is substantially inhibited by velpatasvir for both uracil-terminated and remdesivir-terminated RNA, offering a possible explanation for the synergistic effects of HCV NS5A inhibitors and remdesivir in Calu-3 and primary human lung cells infected with SARS-CoV-2.

Discussion
Combination therapy cures Hepatitis C and enables long-term HIV suppression without significant development of resistance. Such therapy is highly desirable for the SARS-CoV-2 pandemic, but typically takes more than 10 years to develop. Here, we identify 20 FDA-approved compounds that have potential to improve efficacy of remdesivir and could be further optimized to make them clinically useful, significantly shortening timelines compared to de novo drug discovery.
Most strikingly, the drug class of HCV NS5A-inhibitors showed strong synergy with remdesivir in SARS-CoV-2 infected cells, both in immortalized and primary human lung cells, resulting in drastically reduced viral load. The molecular target of HCV NS5A inhibitors is a component of the HCV membrane-bound replication complex, with additional roles in virion assembly and modulation of host cell physiology, but no clear homolog in SARS-CoV-2 [42][43][44][45] . Our data indicate that HCV NS5A inhibitors inhibit the SARS-CoV-2 exonuclease proofreader. This could explain why NS5A inhibitors by themselves do not act as strong antivirals in cell culture 29 , where virus replication is optimal and the proofreader might not be essential; under conditions where virus replication is severely impaired-in the presence of nucleoside inhibitors such as remdesivir-the proofreader could become essential, demonstrating the principle of synthetic lethality. Importantly, this class of moleculesdespite being sought after-have remained elusive; they do not show activity when tested on their own in direct antiviral screening approaches. Only recently, molecular docking combined with biochemical, in vitro assays for Nsp14/10 activity have yielded exonuclease proofreader inhibitors 46 . Importantly, these less potent compounds also do show some synergy with remdesivir, providing independent evidence that inhibition of the proofreader can be synergistic with RdRp inhibition in cells. Indeed, the exonuclease proofreader of coronaviruses is highly divergent from human exonucleases while being strongly conserved within coronaviruses, making it an ideal antiviral target with potential for pan-coronavirus use 46 . Our unbiased combinatorial approach has identified new classes of potentially useful compounds that would have remained obscure in single-agent screening approaches.
Given the promising in vitro data in the current study, it is important to consider the potency and pharmacokinetics of these drugs to understand their potential to be used in the clinic. A critical question to consider is if the concentrations of remdesivir and HCV NS5A inhibitors required for antiviral synergy in our in vitro systems are achievable in patients. The intracellular concentration of the active remdesivir metabolite in the human lung is estimated to be between 4-10 μM, close to its 7 μM IC50 and well below the 18 μM IC90 that would be needed to fully inhibit the virus 19 . Due to systemic toxicity, remdesivir cannot be dosed higher 19 ; recent inhalation trials aim to increase lung concentration by changing route of administration. The 25-fold shift in potency reported in this study could move the IC90 from 18 μM to ~ 0.7 μM in the example above, well below the estimated intracellular concentration of 4-10 μM, putting more robust virus eradication within reach. However, the μM potency of velpatasvir and other NS5A inhibitors is likely too low to allow direct use in patients, in light of their nM exposure after oral administration 47 ; this is not unexpected for repurposing approaches, where the new target is not the one the drug was initially designed for. However, different avenues are possible: higher doses in preclinical animal models could be investigated to increase exposure, supported by an extremely low toxicity of velpatasvir in rodents (NOAEL > 1500 mg/kg/day in mice, no clinical signs observed) 47 . Different routes of administration such as inhalation and specialized formulations such as nanoparticles could be tried 48,49 . The identified compounds are excellent starting points with favorable toxicity profiles for lead optimization studies; many analogs have been generated during preclinical development of HCV NS5A inhibitors and should be tested in the reported assays; structure-activity-relationship knowledge already exists from numerous HCV optimization campaigns. While some optimization will be needed, the lead optimization process is likely significantly shortened relative to de novo drug discovery.
In addition to HCV antivirals, we found 18 more synergistic combinations between remdesivir and approved drugs with favorable safety profiles and a wide range of pharmacokinetic properties that could be studied further and evaluated for their therapeutic usefulness, including dabrafenib, nimodipine and cilostazol. We also identified the well-tolerated and widely used steroids budesonide and meprednisone as showing robust synergy with remdesivir, supporting the notion that steroids can have direct antiviral effects ( Figure ED2) 50,51 . These findings open up the possibility to find dual-action steroid-remdesivir combinations that have antiviral effects early in infection and exert immunomodulatory effects, as achieved with dexamethasone, later. Synergy with remdesivir was also observed for compounds modulating calcium channel and proton pump activity, consistent with wellestablished modulation and exploitation of host cell calcium signaling during infection 31,32,52 . We identified as synergistic with remdesivir the generic calcium-channel blockers nimodipine and nifedipine which are widely used as anti-hypertensives and have excellent safety profiles 37 . Omeprazole sulfide is a metabolite of omeprazole (Prilosec), an over-the-counter proton pump inhibitor to treat reflux, that has been previously identified as enhancing the effect of remdesivir on SARS-CoV-2 53 . Interestingly, several proton pump inhibitors have been strong hits in other SARS-CoV-2 repurposing screening campaigns 54  www.nature.com/scientificreports/ It is important to note that synergistic effects in cells could also be due to pharmacokinetic interactions. For example, NS5A inhibitors can inhibit the membrane transporters P-gp, BCRP, OATP1B1 and OATP1B3 55 . Those transporters reduce intracellular drug concentrations, and their inhibition could increase the apparent potency of remdesivir. However, analysis of known transporter inhibitors in our compound collection reveals only modest enrichment of OATP1B1 inhibitors, and no enrichment for P-gp, BCRP, and OATP1B3 inhibitors ( Figure ED4). This suggests that the observed synergy is not likely due to pharmacokinetic interactions only.
Taken together, our study leverages unbiased combinatorial screening to discover compounds synergistic with antiviral remdesivir. We identify 20 promising combinations between remdesivir and approved drugs with a favorable safety profile and a wide range of pharmacokinetic properties. Among these, combining remdesivir with the HCV NS5A inhibitor combinations Epclusa (velpatasvir/sofosbuvir) and Zepatier (elbasvir/grazoprevir) increased remdesivir potency 25-fold and practically eliminated SARS-CoV-2 from infected cells, including from primary human lung cells, to our knowledge the strongest synergy with remdesivir observed to date. Synergistic combinations are synthetically lethal and strong candidates for further studies in animal models of SARS-CoV-2 infection and optimization, including administration and formulation, for possible clinical evaluation in COVID-19 patients. We provide evidence that HCV NS5A compounds such as velpatasvir and elbasvir are SARS-CoV-2 exonuclease proofreader inhibitors, which provides an intriguing explanation for the observed synthetic lethality with RdRp inhibition through remdesivir, and identifies a class of molecules that has been sought after but largely remained elusive through direct antiviral screening approaches. Exonuclease proofreader inhibitors would be highly desirable in the clinic for potential combination treatment with nucleotide analogs and could improve efficacy, prevent the development of resistance and allow the addition of more therapeutic options as vaccine protection begins to wane. Indeed, encouraging clinical data has recently been reported for Epclusa® in mild to moderate COVID-19 56 . Hepatitis C NS5A therapeutics Epclusa® (velpatasvir/sofosbuvir) and Zepatier® (elbasvir/ grazoprevir) would be excellent starting points for optimization of potency and pharmacokinetic properties to enable clinical evaluation in COVID-19 patients in combination with remdesivir. To generate a master viral stock, Vero E6 were plated in T175 flasks (Nunc, Roskilde, Denmark) and allowed to grow to ~ 80% confluency before infection with the USA-WA1/2020 strain of SARS-CoV-2 (BEI Resources, Manassas, VA; NR-52281). At 72hpi, dramatic CPE was observed and the flasks were freeze-lysed at − 80°C. After thaw, lysate was collected and centrifuged at 3000 × rpm for 20 min to pellet cell debris (Beckman Coulter Allegra X-14R). This procedure was repeated for a second passage working stock with collection at 48hpi and titered by TCID50 assay.

Cells and virus. Vero E6 and Calu-3 cells (Calu
Compound preparation and drug screening. The FDA-approved drug library containing 1200 small molecule compounds (TargetMol, Wellesley Hills, MA, L4200) was stored at 10 mM in dimethyl sulfoxide (DMSO) in 384-well master plates. Resupply for velpatasvir, elbasvir, daclatasvir and ledipasvir was also obtained from TargetMol. Remdesivir was stored at 10 mM in DMSO (T7766, TargetMol). 2500 Vero E6 (12 μl/well) or 10,000 Calu-3 (12 μl/well) were seeded in 384-well white optical-bottom tissue culture plates (Nunc) with the Multidrop Combi liquid handling instrument (Thermo Fisher Scientific, Waltham, MA). Cells were allowed to adhere and expand, 24 h for Vero E6 and 48 h for Calu-3, at 37 °C and 5% CO 2 . For the primary screen, confirmation and synergy dose response interaction matrix analysis, compounds were prediluted to 8 × final concentration in high glucose DMEM. 3 μl compound was transferred from dilution plates using a Cybio Well vario liquid handler (Analytik Jena, Jena, Germany) to cells, leading to a final concentration of DMSO at 0.44% in the assay plate (v/v). Primary screen and confirmation was performed at 40 μM compound, dose responses were generated by 2 × dilutions starting at 40 μM or 10 μM. For synergy experiments, EC15(± 5) of remdesivir was empirically determined and used for each experiment in combination with other drugs as indicated above. Final DMSO was maintained at 0.44%-0.8% (v/v). Cells were incubated at 37 °C and 5% CO 2 for 1 h before infection. Viral inoculum was prepared such that the final MOI = 0.05 upon addition of 6 μl/well viral inoculum. After complete CPE was observed in DMSO-treated, infected wells 72hpi for Vero-E6 and 96hpi for Calu 3, opaque stickers (Nunc) were applied to plate optical bottoms, and plates were developed with the CellTiter-Glo 2.0 reagent (Promega, Madison, WI) according to the manufacturer's instructions. For Vero E6 reagent was diluted 1:1 (v/v) in PBS (Gibco, Waltham, MA, USA). Luminescence of developed plates was read on a Spectramax L (Molecular Devices, San Jose, CA). Each plate contained 24 wells uninfected/DMSO treated cells (100% CPE inhibition), and 24 wells infected/DMSO treated cells (0% CPE inhibition). Average values from those wells were used to normalize data and determine % CPE inhibition for each compound well. For duplicate plates, average values and standard deviations were determined. Z' was determined as described 57 . Stastical significance was assessed using a two-tailed, heteroscedastic student's t-test. Measurements were taken from distinct samples unless indicated otherwise. The data was plotted and analyzed with spotfire (Tibco) and GraphPad Prism. Synergy analysis was performed using synergyfinder, using a zero-interaction potency (ZIP) model 34 . GSEA analysis. Compounds were annotated with targets, pathways and mechanisms of actions using the Center for Emerging and Neglected Diseases' database and for pharmacokinetic data and transporter inhibition data, the DrugBank database 58 . Each annotation property was tested for enrichment among the screening hits www.nature.com/scientificreports/ using the gene set enrichment analysis (GSEA) software as described 29,59,60 . The compounds annotated for each property were treated as part of the "gene set". For each set of annotations, the background compound set was defined as the set of compounds annotated for any property. GSEA preranked analysis was performed using the compounds' % CPE inhibition from each screen. Compound sets included in the analysis were between 5 and 500 compounds. Enrichment results with p < 0.01 and false discovery rate (FDR) q value < 0.1 were considered statistically significant. P values were generated using a one-sided hypergeometric test 61 .
Immunofluorescence microscopy analysis (IFA) and RT-qPCR. 50 62 . Custom code written in MATLAB (available at https:// gitlab. com/ tjian-darza cq-lab/ second-deriv ative-cq-analy sis) was used to take the numerical second derivative of fluorescence intensity with respect to cycle number, using a sliding window of ± 3 cycles. The peak of the second derivative was fit to a parabola, whose center was taken to be the Cq value 62 . 96 well CellTiter-Glo 2.0 and TCID50 assay. 40,000 Calu-3 cells (50 µl/well) were seeded in 96-well white optical-bottom tissue culture plates (Nunc). 48 h post-seeding, drug combinations were added to the cells in 25 µl DMEM and incubated at 37 °C and 5% CO 2 for 1 h before infection. 25 µl viral inoculum was added for MOI = 0.05. At 24hpi, 25 µl supernatant was saved for TCID50 assay. After complete CPE was observed in DMSO-treated, infected wells 96hpi, opaque stickers (Nunc) were applied to plate optical bottoms, and plates were developed with the CellTiter-Glo 2.0 reagent (Promega, Madison, WI), according to the manufacturer's instructions. Luminescence of developed plates was read on a Spectramax L (Molecular Devices, San Jose, CA).
To quantify infectious particles secreted by cells in a TCID50 assay, 25 μl of supernatant from infected, combination-treated cells was collected at 24hpi/drug treatment and tenfold serially diluted in DMEM. Each dilution was applied directly to eight wells in 96-well plates (Corning) pre-prepared with Vero E6 cells, then incubated for three days at 37 °C and 5% CO 2 . TCID50/mL for each sample was calculated by determining the dilution factor required to produce CPE, including syncytia formation, cell clearing and cell rounding, in half, or 4/8, of the wells. Limit of detection was determined as the concentration of virus resulting in CPE in 50% of the wells treated with the lowest dilution of sample.
Reagents and purification of the SARS-CoV-2 exonuclease nsp14/nsp10 complex. For exonuclease experiments, remdesivir triphosphate (RDV-TP) was purchased from MedChemExpress (Monmouth Junction, NJ), sofosbuvir triphosphate (SOF-TP) was purchased from Sierra Bioresearch (Tucson, AZ), and UTP was purchased from Fisher Scientific. The RNA oligonucleotide (template-loop-primer) was purchased from Dharmacon (Horizon Discovery, Lafayette, CO). The 3′-exonuclease, referred to as nsp14, and its protein cofactor, nsp10, were cloned and expressed based on the SARS-CoV-2 genome sequence. The pRSFDuet-1 plasmids (Novagen) coding SARS-Cov-2 Nsp14 or nsp10 engineered with an N-terminal His-SUMO tag were prepared as follows: SARS-CoV-2 RNA isolated from the supernatant of SARS-CoV-2-infected Vero E6 cells was provided by Benjamin R. tenOever 63 . The sequence encoding nsp10 and nsp14 was reverse transcribed into cDNA using gene-specific primers and SuperScript III Reverse Transcriptase (ThermoFisher). nsp10 and nsp14 coding sequences were PCR amplified using forward and reverse gene specific primers flanked by BamHI and XhoI recombination sites, respectively. PCR products were digested with BamHI and XhoI and subsequently ligated into BamHI and XhoI-digested pRSFDuet-His6-sumo vector and sequence verified. pRSFDuet-His6-sumo is a  GGA TCC GCT GAA AAT GTA ACA GGA CTC TTT AAA.  NSP14 pRSF XhoI rev CCC GCT CGA GCG G TCA CTG AAG TCT TGT AAA AGT GTT CCA GAGG.  RT primer NSP14 TTC TTG GCT ATG TCA GTC ATA GAA CAAAC.  NSP10 pRSF BamHI fw CGC GGA TCC GCT GGT AAT GCA ACA GAA GTG CCT GCC.  NSP10 pRSF XhoI_rev CCC GCT CGA GCG G TCA CTG AAG CAT GGG TTC GCG GAG TTG ATC.  RT_primer NSP10 GAT GTT GAT ATG ACA TGG TCG TAA CAGC. The nsp14 and nsp10 proteins were expressed in Escherichia coli BL21-CodonPlus(DE3)-RIL (Stratagene). The bacteria were grown in Luria-Bertani medium supplemented with 50 mg/mL kanamycin at 37 °C to an OD600 of 0.6, induced with 0.4 mM isopropyl β-D-1-thiogalactopyranoside and 50 µM ZnCl 2 overnight at 18 °C. Cells were collected via centrifugation at 5000 × g and equal volumes of the nsp14 and nsp10 bacterial cells were then mixed for nsp14-nsp10 protein complex purification. They were lysed via sonication in Lysis Buffer (500 mM NaCl, 20 mM imidazole, 20 mM Tris·HCl, pH 8.0, 1 mM phenylmethylsulfonyl fluoride). After centrifugation at 40,000 × g, the supernatant was loaded onto 5 mL Nickel Sepharose 6 fast flow resins (GE Healthcare) in a gravity flow column. The target protein was eluted using Lysis Buffer supplemented with 500 mM imidazole. The eluted protein was incubated with ULP1 (lab stock) during dialysis at 4 °C overnight against a buffer containing 20 mM Tris·HCl, pH 7.5, 20 mM imidazole, 150 mM NaCl, 100 µM ZnCl 2 , and 5 mM β-mercaptoethanol. Then the sample was loaded onto the HisTrap FF column (GE Healthcare) to remove His-SUMO tag, and the flowthrough was collected. The target proteins were further purified through a Superdex200 10/300 gel filtration column (GE Healthcare) in a buffer containing 20 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , and 1 mM dithiothreitol. The fractions corresponding to the nsp14 and nsp10 complex were detected by SDS-PAGE and collected. The protein sample was flash-frozen in liquid nitrogen and stored at − 80 °C.
Extension reactions with SARS-CoV-2 RNA-dependent RNA polymerase to produce Remdesivir (RDV) terminated RNAs. 10 µL of 10 µM RNA template-loop-primers (5'-UUU UCA UCG CGU AGU UUU CUA CGC G-3' for RDV-TP extension) in 1 × RdRp reaction buffer was annealed by heating to 75 °C for 3 min and cooling to room temperature. 5 µL of 8 µM RdRp complex (nsp12/nsp7/nsp8) 38 in 1 × reaction buffer was added to the annealed RNA template-loop-primer solution and incubated for an additional 10 min at room temperature. Finally, 5 µL of a solution containing 0.2 mM RDV-TP in 1 × reaction buffer was added and incubation was carried out for 2 h at 30 °C. The final concentrations of reagents in the 20 µL extension reactions were 2 µM nsp12/nsp7/nsp8, 5 µM RNA template-loop-primer, and 50 µM RDV-TP. The 1 × reaction buffer contains the following reagents: 10 mM Tris-HCl pH 8, 10 mM KCl, 2 mM MgCl 2 and 1 mM β-mercaptoethanol. Desalting of the reaction mixture was performed with an Oligo Clean & Concentrator kit (Zymo Research) resulting in ~ 10 µL purified aqueous RNA solutions. 1 µL of each solution was subjected to MALDI-TOF MS (Bruker ultrafleXtreme) analysis. The remaining ~ 9 µL extended template-loop-primer solutions were used to test exonuclease activity as described below.
SARS-CoV-2 exonuclease reactions in the presence and absence of velpatasvir. The U-terminated RNA (Fig. 7), and the RDV extended RNA product from above (sequences shown in Fig. 7), were annealed by heating to 75 °C for 3 min and cooling to room temperature in 1 × exonuclease reaction buffer. To a 14 µL solution of 71.4 nM exonuclease complex (nsp14/nsp10) in 1 × exonuclease reaction buffer, 1 µL of DMSO with or without various concentrations of velpatasvir was added and incubated for 15 min at room temperature. Then 5 µL of the annealed RNA (2 µM) in 1 × exonuclease reaction buffer was added to the exonuclease/velpatasvir mixture and incubated at 37 °C for 15 min. The final concentrations of reagents in the 20 µL reactions were 50 nM nsp14/nsp10, 500 nM RNA, 0-100 µM velpatasvir and 5% DMSO. The 1 × exonuclease reaction buffer contains the following reagents: 40 mM Tris-HCl pH 8, 1.5 mM MgCl 2 and 5 mM DDT. Following desalting using an Oligo Clean & Concentrator (Zymo Research), the samples were subjected to MALDI-TOF MS (Bruker ultrafleXtreme) analysis.

Data availability
The authors declare that the data supporting the findings of this study are available within the paper and its supplementary information files. www.nature.com/scientificreports/