key: cord-0851741-7pbzampo
authors: Bertolin, Agustina P.; Weissmann, Florian; Zeng, Jingkun; Posse, Viktor; Milligan, Jennifer C.; Canal, Berta; Ulferts, Rachel; Wu, Mary; Drury, Lucy S.; Howell, Michael; Beale, Rupert; Diffley, John F.X.
title: Identifying SARS-CoV-2 Antiviral Compounds by Screening for Small Molecule Inhibitors of Nsp12/7/8 RNA-dependent RNA Polymerase
date: 2021-04-08
journal: bioRxiv
DOI: 10.1101/2021.04.07.438807
sha: f881d2547d2c92dd8deea30c4c44a565e52a5a86
doc_id: 851741
cord_uid: 7pbzampo

The coronavirus disease 2019 (COVID-19) global pandemic has turned into the largest public health and economic crisis in recent history impacting virtually all sectors of society. There is a need for effective therapeutics to battle the ongoing pandemic. Repurposing existing drugs with known pharmacological safety profiles is a fast and cost-effective approach to identify novel treatments. The COVID-19 etiologic agent is the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a single-stranded positive-sense RNA virus. Coronaviruses rely on the enzymatic activity of the replication-transcription complex (RTC) to multiply inside host cells. The RTC core catalytic component is the RNA-dependent RNA polymerase (RdRp) holoenzyme. The RdRp is one of the key druggable targets for CoVs due to its essential role in viral replication, high degree of sequence and structural conservation and the lack of homologs in human cells. Here, we have expressed, purified and biochemically characterised active SARS-CoV-2 RdRp complexes. We developed a novel fluorescence resonance energy transfer (FRET)-based strand displacement assay for monitoring SARS-CoV-2 RdRp activity suitable for a high-throughput format. As part of a larger research project to identify inhibitors for all the enzymatic activities encoded by SARS-CoV-2, we used this assay to screen a custom chemical library of over 5000 approved and investigational compounds for novel SARS-CoV-2 RdRp inhibitors. We identified 3 novel compounds (GSK-650394, C646 and BH3I-1) and confirmed suramin and suramin-like compounds as in vitro SARS-CoV-2 RdRp activity inhibitors. We also characterised the antiviral efficacy of these drugs in cell-based assays that we developed to monitor SARS-CoV-2 growth.

The COVID-19 outbreak caused by SARS-CoV-2 is considered to be the most severe global public health crisis since the influenza pandemic in 1918 (1, 2) . As of February 2021, there were more than 100 million confirmed cases and 2.5 million deaths globally (3). Remdesivir is currently the only antiviral drug approved for COVID-19 treatment. Remdesivir is a nucleoside analog that acts as a small molecule inhibitor of an essential coronavirus enzyme complex, the RNA-dependent RNA polymerase (4) . Nucleoside analogues are one of the largest and most effective classes of small molecule drugs against viruses like human immunodeficiency virus, hepatitis B virus and herpesviruses (5, 6) . The efficacy of remdesivir as a COVID-19 treatment remains controversial.

While one clinical trial found remdesivir to shorten recovery time in COVID-19 patients, other clinical trials found no effect of remdesivir on recovery time or mortality rate (7) (8) (9) (10) . Besides remdesivir, several other clinically approved nucleoside analog compounds are currently under study for treating COVID-19. Molnupiravir (11, 12) and are the most promising candidates so far, while ritonavir (14), ribavirin (15) , favipiravir (16) and sofosbuvir (6, 17) have not demonstrated significant antiviral effect against SARS-CoV-2 in laboratory or clinical settings.

Coronaviruses (CoVs) are enveloped single-stranded positive-sense RNA viruses that belong to the order Nidovirales (18) . CoV replication and transcription occur in the cytoplasm of infected cells and are mediated by the replication-transcription complex (RTC) (18, 19) . The core RTC is composed of i) nsp12-F/7H8). To be able to obtain better yields, we transferred the His6-tag to the catalytic subunit nsp12 and produced a complex containing a His6-3xFlag tag on nsp12 (Sf nsp12-HF) and a neutral 6 amino acid (Gly-Gly-Ser)2-linker between nsp7 and nsp8 (7L8). Purification of this complex by affinity to Ni-NTA agarose, ion exchange and size exclusion chromatography resulted in ~100x higher protein yields (Supplementary Table S1), albeit with less stoichiometric nsp7-nsp8 fusion protein ( Figure   1A , lane 4: nsp12-HF/7L8). We also tried co-expression of individual nsp7 and nsp8 subunits with Table S1 ). In a second approach, we expressed the three proteins individually in E. coli as N-terminal His-SUMO fusion proteins ( Figure 1B ). In this system, the affinity tag and SUMO fusion can be removed after affinity purification by a SUMO-specific protease (41), leaving behind the same N-terminus as would be generated by viral protease-mediated polyprotein cleavage in infected cells. We expressed nsp7, nsp8 and nsp12 using this system and purified the proteins by affinity to Ni-NTA agarose, fusion protein removal, ion exchange and size exclusion chromatography ( Figure 1B ).

To assess RNA-dependent RNA polymerisation activity we generated a primed RNA substrate by annealing a 35 nucleotide (nt) RNA template that contained a Cy3 fluorophore on its 5' end with a 10 nt primer strand complementary to the 3'-end of the template. After incubation of RdRp with primed substrate and NTPs, reaction products were separated by native PAGE (Figure 1C) . We tested in this assay the activity of the RdRp preparation from insect cells containing C-terminally tagged nsp12 and a co-expressed nsp7-nsp8 fusion protein with a His6-linker (Sf nsp12-F/7H8). This RdRp preparation was able to extend the primed RNA substrate to generate duplex RNA in a time-dependent manner, confirming RNA-dependent RNA polymerisation activity ( Figure 1C ).

We developed a variation of this primer-extension assay that can measure activity directly in solution, without requiring running products in a gel (Supplementary Figure S1A,B) . As in Figure 1C , the Cy3labelled primed RNA substrate is extended by the RdRp complex and generates duplex RNA. Then, after 60 minutes, we added a 14 nt RNA strand partially complementary to the template strand with a quencher molecule at its 3' end. If RdRp has extended the primer, this quencher strand will not anneal to the template strand and will not be able to quench Cy3 fluorescence (Supplementary Figure   S1A ). We tested nsp12-F/7H8 in this assay and found that Cy3 fluorescence was greatly increased when RdRp was included in the reaction and the presence of Mn 2+ enhanced RdRp activity compared to Mg 2+ alone (Supplementary Figure S1B) , which is in line with a published SARS-CoV-1 nsp12 enzymatic characterization (42).

None of the primer-extension assays described above (Figure 1C and Supplementary Figure S1A and B) are amenable to accurate high throughput screening (HTS) as they involve multiple steps and rely only on end point values. Therefore, we designed a FRET-based assay suitable for HTS based on RNA synthesis-coupled strand displacement activity ( Figure 1D ). Strand displacement refers to the ability of certain DNA/RNA polymerases to displace downstream non-template strands from the template strand while polymerising nucleotides (43, 44) ( Figure 1D ). The RNA substrate was constructed by annealing the primed 35 nt RNA template with the 14 nt quencher strand ( Figure 1D ).

This structure places the Cy3 fluorophore in close proximity to the quencher localised on the opposite strand. As RdRp elongates the primer, it displaces the downstream quencher strand producing a fluorescent signal. As the final product is an RNA duplex, the quencher strand is prevented from reannealing ( Figure 1D ). When Sf nsp12-F/7H8 was incubated with the strand displacement substrate, fluorescence increased near-linearly with time and was dependent on enzyme concentration ( Figure 1E) . The presence of Mn 2+ was not required but again greatly enhanced RdRp activity compared to Mg 2+ alone (Supplementary Figure S1C-E) . The fluorescence increase was dependent on NTPs (Supplementary Figure S1F) suggesting that i) there were no contaminating nucleases in the reactions that could also have resulted in freeing the Cy3 fluorophore from the quencher and ii) RdRp polymerisation was driving strand displacement of the quencher strand.

We tested our different RdRp preparations from Figure 1A and 1B using the FRET-based strand displacement assay in a concentration range of 25 nM -400 nM (Figure 2 ). The C-terminally tagged nsp12 co-expressed with nsp7-nsp8 fusion protein with His linker (Sf nsp12-F/7H8, Figure 2A ) was slightly more active than Sf nsp12-F/7L8 (Figure 2B) , which we suspect is due to a lower stoichiometry of the 7L8 fusion protein (Figure 1A, lanes 3 and 4) . The RdRp complex in which nsp7 and nsp8 were co-expressed individually with nsp12 (Sf nsp12-HF/7/8) did not show measurable activity (Supplementary Figure S2A) , probably because nsp7 and nsp8 were sub-stoichiometric ( Figure 1A , lane 5). Pre-incubating individually expressed Sf nsp12-HF and Sf 7H8 fusion protein (1:3 ratio, Figure 2C ) resulted in slightly less activity than the co-purified Sf nsp12-F/7H8 (Figure 2A) .

Pre-incubating the three individually expressed proteins from E. coli (Ec) nsp12, nsp7, and nsp8 at a ratio of 1:3:6 produced the highest activity ( Figure 2D ). Mixing Ec nsp12 with insect cell expressed 7H8 at a ratio of 1:3 (Sf 7H8, Figure 2E ) and mixing Sf nsp12-HF with Ec nsp7 and Ec nsp8 at a ratio of 1:3:6 ( Figure 2F ) also produced moderate activity. All together, these results suggest that either co-purification of nsp12 with a nsp7-nsp8 fusion protein or mixing and preincubation of individually purified proteins, regardless of insect cell or bacterial expression, can generate active RdRp complexes.

Best protein yields were obtained in insect cells when nsp12 was co-expressed with 7L8 fusion protein ( Supplementary Table S1 ), and therefore, we optimised the HTS assay conditions using this RdRp preparation (Sf nsp12-HF/7L8). The 7L8 fusion protein was sub-stoichiometric ( Figure 1A , lane 4), so we tested if the activity of Sf nsp12-HF/7L8 could be increased by further addition of nsp12 cofactors: Ec nsp7 and Ec nsp8 or Sf 7H8. We preincubated the proteins at high concentration (10 µM for Sf nsp12-HF/7L8 either with Ec nsp7 and Ec nsp8 in 1:3:6 ratio, or with Sf 7H8 in 1:6 ratio) for 30 minutes and found they increased the activity of Sf nsp12-HF/7L8 (Figure 3A-C) , with Sf 7H8 fusion protein addition showing the highest activity ( Figure 3C) . We also mixed Sf nsp12-HF/7L8 and Sf 7H8 at a ratio of 1:3 or 1:5 at different concentrations and found both ratios conferred similar activity at 100 nM and 150 nM Sf nsp12-HF/7L8 (Figure 3D-E) . A final concentration of 150 nM RdRp complex in a 1:3 ratio was chosen and used for the high-throughput screen and all subsequent experiments. Finally, the presence of 5% DMSO, which is contributed by the chemical library in the high throughput screening assay, did not affect RdRp activity (Supplementary Figure S3A ).

The optimised strand displacement assay was used to screen a custom library of over 5000 compounds for inhibitors of SARS-CoV-2 RdRp activity in a 384-well format. The compounds were incubated with RdRp at room temperature for 10 minutes, and then reactions were started by substrate addition (Figure 4A) . Fluorescent signals were recorded in 90 second intervals for 10 cycles, which covered the linear phase of the reactions (Supplementary Figure 4A) , and fluorescence increase over time was used to calculate the reaction velocity for each well. The library was screened at two compound concentrations, 1.25 µM ( Figure 4B ) and 3.75 µM (Figure 4C ).

Reactions that showed >15% reduction in the velocity (Figure 4B ,C) or >10% reduction in the endpoint signal (Supplementary Figure S4C ,D) were manually inspected (example in Figure 4D ). 64 compounds were selected as primary hits. Compounds were then discounted if: i) the compound also scored as a strong hit in parallel HTSs of other SARS-CoV-2 enzymes (with the exception of sulphated naphthylamine derivatives, see below) (Biochem J, this issue); ii) the compound was reported or strongly predicted to be a promiscuous inhibitor due to colloidal aggregation (LogP value >3.6 and >85% similarity to a known aggregator) (45, 46) ; iii) the compound was reported or predicted to be a nucleic acid intercalator. This analysis resulted in the selection of 18 compounds for further validation experiments (Supplementary Table S2 Table S2 ) and their nsp13 inhibition was shown to be insensitive to detergent (Zeng et al, Biochem J, this issue), making nonspecific inhibition due to colloidal aggregation unlikely (60).

Suramin, as the main representative drug of this group, was selected to be included in further in vitro validation experiments as part of this study. Ultimately, 14 compounds were chosen for further validation experiments.

In order to confirm and quantify RdRp inhibition by the selected compounds in vitro, we performed compound titration experiments to determine the drug concentration at which half maximal inhibition is achieved (IC50). We carried these experiments out in the presence and in the absence of the nonionic detergent Triton X-100 to rule out unspecific inhibition due to compound aggregation (60). We Table S3) . Therefore, these compounds were excluded from further analysis. The remaining 6 compounds inhibited RdRp with IC50 values of 0.43 µM -56 µM and showed no or only minor sensitivity to detergent, indicating they might be bona fide RdRp inhibitors ( Figure 5A , Table 1 ).

We tested the 6 compounds in the gel-based polymerisation assay which monitors RNA synthesis more directly using the same primed RNA substrate as in Figure 1C . The RdRp preparation used here (Sf 12-HF/7L8 + Sf 7H8, ratio 1:3) also showed time-dependent RNA duplex formation ( Figure   5B ). We performed a 30-minute reaction with the compounds at 50 µM and found they inhibited RdRp consistent with their IC50 values from the kinetic assay. Suramin, C646 and BH3I-1 inhibited RNA duplex formation almost completely, and the compounds with higher IC50 values, GSK-650394, MDK83190 and Cefsulodin showed partial inhibition ( Figure 5C ).

The validation experiments indicated that 6 compounds are genuine SARS-CoV-2 RdRp inhibitors in biochemical assays ( Figure 5A and Table 1 ). We therefore evaluated their potential antiviral activity against SARS-CoV-2 in Vero E6 cells. Cells were treated with the compounds and then infected with SARS-CoV-2. After 22 hours, viral plaques were analysed by viral nucleocapsid (N) protein immunofluorescence ( Figure 6A) . We tested the 5 novel RdRp inhibitors for their potential antiviral activity using this assay ( Figure 6B ). Dose-response curves and the half maximal effective concentration (EC50) for each compound was calculated ( Figure 6C) . Two of them, GSK-650394 and C646, presented a dose-dependent reduction in the levels of SARS-CoV-2 N protein (green) in Vero E6 cells with low micromolar EC50 values (EC50=7.6 µM and EC50=19 µM respectively, Figure 6B ,C).

Suramin antiviral activity in Vero E6 cells following the same aforementioned protocol has been evaluated by us in the accompanying manuscript Zeng et al. and an EC50 of 9.9 µM against SARS-CoV-2 was observed (Zeng et al, Biochem J, this issue).

Combination therapy, the use of two or more antivirals with different modes of action, is a known strategy for limiting drug resistance as well as achieving better physiological outcomes (61) . The latter is usually caused by reduced toxicity due to the use of lower doses in combination protocols compared to monotherapy doses (61) . Consequently, a potential synergy between RdRp inhibitors with different mechanisms of action (non-nucleoside analogues identified here and a nucleoside analog, remdesivir) was assessed (38, 62, 63). Following the viral inhibition protocol described in Figure 6A , we performed dose-response experiments for each compound in the presence of a fixed concentration of 0.5 µM remdesivir, which as a single agent inhibited viral infection by less than 10%.

Under these assay conditions, none of the validated inhibitors showed significant synergy with remdesivir (Supplementary Figure S6 ).

The replication/transcription complex (RTC) of coronaviruses catalyses the synthesis of all genomic RNAs (viral replication) and sub-genomic RNAs (viral transcription). The core of SARS-CoV-2 RTC is the RNA-dependent RNA polymerase holoenzyme, composed of nsp12, the catalytic subunit, together with its cofactors nsp7 and nsp8 (19) (20) (21) 64) . SARS-CoV-2 RdRp is the target of remdesivir, the only antiviral approved for human use to treat COVID-19 by regulatory agencies (62) . Remdesivir is a broad-spectrum nucleoside analog first used for treatment of Ebola infections (4). Remdesivir's moderate success in COVID-19 treatment is still a matter of controversy as new clinical trials suggest that it is not as effective as first thought (7-10). Coronaviruses harbour a unique RNA-replication proofreading trait conferred by the 3'-to-5' exonuclease activity of nsp14 (29), offering a possible explanation as to why the development or repurposing of effective nucleoside analogues, with the exception of remdesivir, has been unsuccessful (31, 36). Therefore, the characterisation of nonnucleoside analog-type drugs that target SARS-CoV-2 RdRp might be needed more than ever.

Moreover, COVID-19 is the third known transfer, after SARS-CoV in 2003 and MERS-CoV in 2012, of an animal coronavirus to a human population in only 20 years and therefore it is highly probable that new zoonotic outbreaks will occur in the future (65) . Thus, it is vital to develop numerous broadspectrum antiviral strategies, that together with vaccination programs, could curb future cross-species transmissions.

Here, we have expressed, purified and characterised a diverse array of SARS-CoV-2 RdRp complexes. We used two expression hosts, a bacterial system (E. coli) and a baculovirus-insect cell system and showed that recombinant proteins from both systems individually expressed or expressed as a complex lead to active proteins. We characterised different constructs of nsp12, nsp7 and nsp8 and tested activities of RdRp complexes formed with different stoichiometries of the subunits. We showed that the SARS-CoV-2 RdRp can strand displace downstream RNA encountered during synthesis. Based on this characteristic of SARS-CoV-2 RdRp, we designed a novel and robust FRETbased assay that allows real-time monitoring of RdRp activity in vitro. We utilised this assay to screen a custom chemical library to identify compounds that inhibit SARS-CoV-2 RdRp enzymatic activity.

Our study identified GSK-650394, C646 and BH3I-1 as well as confirmed suramin and suramin-like compounds as in vitro inhibitors of SARS-CoV-2 RdRp activity. These compounds do not have typical chemical structures of nucleoside analogues and thus likely inhibit RdRp through a different mechanism of action from remdesivir. Further work is required to understand their mode of inhibition.

Cryo-electron microscopy structures of SARS-CoV-2 RdRp have been presented by several groups (20, 38, 40, 64, 66, 67) . Future compound-protein structures may help reveal the inhibitory mechanisms of the RdRp inhibitors identified in this study.

GSK-650394, suramin and C646 showed also significant inhibition on viral growth in cell-based assays with an EC50 of 7.6 µM, 9.9 µM and 19 µM respectively ( Figure 6B ,C) (nsp13 paper, Biochem J, this issue). GSK-650394 is an inhibitor of SGK-1 kinase (68) that plays an important role in a diverse range of cellular processes such as stress response, ion transport, inflammation and cell proliferation (69) . GSK-650394 has been reported to impede growth of cellular models of cancers (68, 70) and has also been implicated in the inhibition of replication of influenza A virus, an enveloped negative-sense ssRNA virus, in human A549 cells (71) . In our study GSK-650394 showed a low micromolar EC50 of 7.6 µM in inhibiting SARS-CoV-2 in cell-based assays, with effects on cell death only at the highest concentration tested (300 µM). Suramin is an antiparasitic drug first introduced by Bayer in 1916 that is still used to treat sleeping sickness and river blindness (72) . Suramin has been repurposed for a number of diverse applications including the inhibition of purinergic signaling, treatment of viral infections and treatment for advanced malignancies reflecting its multitude of targets (47) . Suramin is a polyanionic compound and probably binds tightly to positively charged patches found in DNA or RNA binding proteins (47, 48) . This is consistent with our findings, as suramin and suramin analogues inhibit in vitro both SARS-CoV-2 RdRp and SARS-CoV-2 nsp13 helicase (Zeng et al, Biochem J, this issue). C646 is a competitive inhibitor of the histone acetyltransferase p300 that amongst other effects results in a reduction of pro-inflammatory gene expression in cells and animal models (73) (74) (75) (76) . C646 was also shown to suppress the replication of different strains of influenza A viruses in A549 cells with an EC50 of 16 μM by affecting several steps of the viral life cycle (77) .

Murine models of influenza infection treated with C646 also showed significant reduction in viral replication with no associated toxicity (77) . Further in vitro, cell-based and pre-clinical studies are needed to confirm and characterise these novel RdRp inhibitors. Moreover, they can be used as promising lead compounds to design improved drug candidates.

In summary, we have identified small-molecule non-nucleoside analog inhibitors of SARS-CoV-2

RdRp that also inhibited viral replication in cell-based assays. We provide evidence that GSK-650394, suramin and C646 could be considered as drug candidates that deserve further evaluation, as these compounds were found to exhibit antiviral activity against SARS-CoV-2 in a relevant cell culture model at non-cytotoxic concentrations. This screening strategy can be expanded to other small molecule compound libraries and together with the biochemical insights regarding SARS-CoV-2 RdRp expression and purification described here, these should accelerate the search for improved antivirals that target SARS-CoV-2 RdRp holoenzyme.

The following proteins were expressed in baculovirus-infected insect cells: nsp7-His6-nsp8 (Sf 7H8), nsp12-His6-3xFlag (Sf nsp12-HF), nsp12-3xFlag/nsp7-His6-nsp8 (Sf nsp12-F/7H8), nsp12-His6-3xFlag/nsp7-(GGS)2-nsp8 (Sf nsp12-HF/7L8), and nsp12-His6-3xFlag/nsp7/nsp8 (Sf nsp12-HF/7/8).

The coding sequences of SARS-CoV-2 nsp7, nsp8 and nsp12 (NCBI reference sequence NC_045512.2) were codon-optimised for S. frugiperda and synthesized (GeneArt, Thermo Fisher Scientific). Baculoviral expression constructs were generated using the biGBac vector system (78).

An empty polyhedrin (polh) expression cassette from pLIB was inserted into the vectors pBIG1a, pBIG1b, and pBIG1c to generate pBIG1a(polh), pBIG1b(polh), and pBIG1c(polh).

Nsp12 was subcloned into pBIG1a(polh) either to contain a C-terminal His6-3xFlag tag (protein sequence: M-nsp12-GGSHHHHHHGSDYKDHDGDYKDHDIDYKDDDDK, pBIG1a_nsp12-His6-3xFlag) or to contain a C-terminal 3xFlag tag (protein sequence: M-nsp12-GGSDYKDHDGDYKDHDIDYKDDDDK, pBIG1a_nsp12-3xFlag). Nsp7 and nsp8 were either subcloned individually into pBIG1b(polh) and pBIG1c(polh) to not contain a tag (pBIG1b_nsp7 and pBIG1c_nsp8), or into pBIG1b(polh) to be expressed as fusion protein with a His6-linker (protein sequence: M-nsp7-HHHHHH-nsp8, pBIG1b_nsp7-His6-nsp8) or with a (GGS)2-linker (protein sequence: M-nsp7-GGSGGS-nsp8, pBIG1b_nsp7-(GGS)2-nsp8). Co-expression constructs were generated by subcloning expression cassettes from pBIG1 vectors into pBIG2abc vectors:

pBIG1a_nsp12-3xFlag, pBIG1b_nsp7-His6-nsp8 and empty pBIG1c were used to generate pBIG2abc_nsp12-3xFlag/nsp7-His6-nsp8 (nsp12-F/7H8). The vectors pBIG1a_nsp12-His6-3xFlag, pBIG1b_nsp7-(GGS)2-nsp8, and empty pBIG1c were used to generate pBIG2abc_nsp12-His6-3xFlag/nsp7-(GGS)2-nsp8 (nsp12-HF/7L8). The vectors pBIG1a_nsp12-His6-3xFlag, pBIG1b_nsp7 and pBIG1c_nsp8 were used to generate pBIG2abc_nsp12-His6-3xFlag/nsp7/nsp8 (nsp12-HF/7/8).

Baculoviruses were generated and amplified in Sf9 insect cells (Thermo Fisher Scientific) using the EMBacY baculoviral genome (79). For protein expression, Sf9 cells were infected with baculovirus and collected 48 hours after infection, flash-frozen and stored at -70°C.

The following proteins were expressed in Escherichia coli (Ec): 14His-SUMO-nsp7 (Ec 7), 14His-SUMO-nsp8 (Ec 8) and 14His-SUMO-nsp12 (Ec 12). The coding sequences of SARS-CoV-2 nsp7, nsp8 and nsp12 were codon-optimised for E. coli and synthesized (Genewiz). Nsp7, nsp8 and nsp12 were subcloned into the K27SUMO expression vector (41) to code for Ulp1-cleavable 14His-SUMO fusion proteins. For expression the vectors were transformed into T7 Express lysY competent E. coli cells (New England Biolabs). Expression cultures were grown in LB containing 50 µg/ml kanamycin to an OD600 of 0.5 and expression was induced with 0.4 mM IPTG at 16°C. After overnight incubation, cells were collected, flash-frozen and stored at -70°C.

All proteins were purified at 4°C. The proteins Sf nsp12-F/7H8 and Sf nsp12-HF/7/8 were purified by FLAG affinity, Heparin affinity and size exclusion chromatography. Insect cell pellets containing these proteins were resuspended in Flag buffer (50 mM HEPES pH 7.4, 150 mM KCl, 5 mM MgCl2, 10% glycerol, 1 mM DTT) supplemented with protease inhibitors (Roche Complete Ultra EDTA-free tablets, 10 µg/ml leupeptin, 10 µg/ml pepstatin A, 1 mM AEBSF) and lysed by Dounce homogenization. After centrifugation at 39000g, 60 min, 4°C, the protein was purified from the cleared lysate by affinity to Anti-FLAG M2 Affinity gel (Sigma-Aldrich) and eluted in Flag buffer containing 0.5 mg/ml 3xFLAG peptide. The Flag eluate was diluted 1:6 with dilution buffer ( 

The Cy3 template strand and the primer were annealed at 1:2 ratio using the protocol described above. A reaction was typically started at room temperature by adding 10 µL of a 2x substrate solution containing the RNA substrate and NTPs in reaction buffer to 10 µL of a 2x enzyme solution containing the RdRp complex in reaction buffer. The quencher strand was added at 2-fold concentration (relative to the Cy3 strand) after a 60-min reaction. The quencher strand was allowed to anneal with any unpolymerized template RNA for 35 min at room temperature before the fluorescent signal was read.

The Cy3 template strand and the primer were annealed at 1:2 ratio using the protocol described above. A reaction was typically started at room temperature by adding 5 µL of a 2x substrate solution containing the RNA substrate and NTPs in reaction buffer to 5 µL of a 2x enzyme solution containing the RdRp complex in reaction buffer. The reaction was stopped by adding 5 µL stopping buffer (20 mM EDTA, 0.5% SDS, 2 mg/ml proteinase K, 10% glycerol) and incubating at 37°C for 5 min.

Reaction products were then analysed by running on a 20% non-denaturing PAGE gel in 0.5X TBE buffer (Life Technologies) for 20 min at room temperature and 200 V. An Amersham Imager 600 was used to image Cy3 fluorescence. To test compound inhibition, a compound volume of 0.5 µL at a concentration of 1 mM was incubated with 5 µL 2x enzyme solution for 10 min before starting the reaction.

The screening plates containing the custom library of compounds were prepared and stored as described in (Zeng et al, Biochem J, this issue). For the HTS assay, first, 10 µM of Sf nsp12-HF/7L8 complex was incubated with 30 µM Sf 7H8 for 30 min on ice to form the RdRp complex. Then the RdRp complex was diluted in reaction buffer, before being dispensed at 10 µL volume to the plates to incubate with compounds for 10 min at room temperature. To start the reaction, 10 µL 2x substrate mix (600 µM NTPs, 200 nM RNA substrate) in reaction buffer was dispensed and the plates were spun briefly before being transferred to a Tecan Spark microplate reader. The fluorescent signal was first read at 1 min of reaction and then read every 1.5 min for 10 cycles. The screening was performed twice, one with a final compound concentration of 1.25 µM and one with 3.75 µM.

MATLAB was used to process data. The slope of fluorescent signal increase over time was calculated using a linear regression model for each compound as the reaction velocity (V). Then V for each compound is normalised against the DMSO controls in each plate as following:

Normalised_V=V/mean(V_DMSO), Where mean(V_DMSO) is the average of velocity values of DMSO controls in a particular plate.

The endpoint signal of each compound was also normalised against DMSO controls. Both the normalised reaction velocity and endpoint signal were used for hit selection. Compounds that gave >15% reduction in the reaction velocity or >10% reduction in the endpoint signal were further inspected. A kinetic curve (example in Figure 4D ), in which the y-axis shows fluorescent signals, and the x-axis shows timepoints, was plotted for each of the compounds that passed the threshold. 10 wells that were before and after the compound well were plotted together in the same graph. We manually inspected the curves and confirmed that 64 compounds gave obvious inhibition.

The Z-score was also calculated for each compound (Supplementary Figure S4B To assess the quality of the screen, the Z-factor was calculated for each plate as following:

Where SD(V_DMSO) is the standard deviation of reaction velocity values of all DMSO controls in a particular plate, SD(background) is the standard deviation of reaction velocity values of background wells containing no enzyme, and mean(background) is the average of reaction velocity values of background wells. The average Z-factor for the screen is 0.71, indicating the screen is good, as described by (80).

The method used for predicting drug aggregation propensity is described in (46) and can be accessed in http://advisor.bkslab.org/.

Compounds selected for in vitro experimental validation and cell-based studies were purchased and resuspended following manufacturer's instructions. See Supplementary Table S4 for more details.

The production of the SARS-CoV-2 virus isolate used in these assays and the in-house generation of the SARS-CoV-2 nucleocapsid (N) recombinant antibody is described elsewhere (Biochem J, nsp13 paper). The conditions for the viral inhibition assay were the following: 15.000 Vero E6 cells (NIBC, UK) resuspended in DMEM containing 10% FBS were seeded into each well of 96-well imaging plates (Greiner 655090) and cultured overnight at 37°C and 5% CO2. The next day, 5x solutions of compounds were generated by dispensing 10 mM stocks of compounds into a V-bottom 96-well plate (Thermo 249946) and back filling with DMSO to equalise the DMSO concentration in all wells using an Echo 550 (Labcyte) before resuspending in DMEM containing 10% FBS. The assay plates with seeded Vero E6 cells had the media replaced with 60 µl of fresh growth media, then 20 µl of the 5x compounds were stamped into the wells of the assay plates using a Biomek Fx automated liquid handler. Finally, the cells were infected by adding 20 µl of SARS-CoV-2 with a final MOI of 0.5 PFU/cell. 22 h post infection, cells were fixed, permeabilised, and stained for SARS-CoV-2 N protein using Alexa488-labelled-CR3009 antibody and cellular DNA using DRAQ7 (Abcam). Whole-well imaging at 5x was carried out using an Opera Phenix (Perkin Elmer) and fluorescent areas and intensity calculated using the Phenix-associated software Harmony (Perkin Elmer).

All data associated with this paper will be deposited in FigShare 

Editing, Visualization. Viktor Posse: Conceptualization, Methodology, Validation, Investigation, Writing -Review & Editing. Jennifer Milligan: Investigation, Resources. Berta Canal: Conceptualization, Investigation, Writing -Review & Editing. Rachel Ulferts: Investigation. Mary Wu: Investigation. Lucy S. Drury: Resources. Michael Howell: Resources, Supervision. Rupert Beale: Resources, Supervision. John FX Diffley: Conceptualization, Methodology, Formal analysis, Writing -Review & Editing, Supervision

A pneumonia outbreak associated with a new coronavirus of probable bat origin

A new coronavirus associated with human respiratory disease in China

Coronavirus Susceptibility to the Antiviral Remdesivir (GS-5734) Is Mediated by the Viral Polymerase and the Proofreading Exoribonuclease. mBio

Nucleosides for the treatment of respiratory RNA virus infections

Repurposing Nucleoside Analogues for Human Coronaviruses

Remdesivir in adults with severe COVID-19: a randomised, double-blind, placebo-controlled, multicentre trial

Effect of Remdesivir vs Standard Care on Clinical Status at 11 Days in Patients With Moderate COVID-19: A Randomized Clinical Trial

Remdesivir for 5 or 10 Days in Patients with Severe Covid-19

Repurposed Antiviral Drugs for Covid-19 -Interim WHO Solidarity Trial Results

Small-Molecule Antiviral beta-d-N (4)-Hydroxycytidine Inhibits a Proofreading-Intact Coronavirus with a High Genetic Barrier to Resistance

An orally bioavailable broad-spectrum antiviral inhibits SARS-CoV-2 in human airway epithelial cell cultures and multiple coronaviruses in mice

AT-527 is a potent in vitro replication inhibitor of SARS-CoV-2, the virus responsible for the COVID-19 pandemic

Remdesivir, lopinavir, emetine, and homoharringtonine inhibit SARS-CoV-2 replication in vitro

Efficacy and safety of favipiravir, an oral RNA-dependent RNA polymerase inhibitor, in mild-to-moderate COVID-19: A randomized, comparative, open-label, multicenter, phase 3 clinical trial

Coronavirus biology and replication: implications for SARS-CoV-2

The Nonstructural Proteins Directing Coronavirus RNA Synthesis and Processing

Structure of the SARS-CoV nsp12 polymerase bound to nsp7 and nsp8 co-factors

One severe acute respiratory syndrome coronavirus protein complex integrates processive RNA polymerase and exonuclease activities

Human coronavirus 229E nonstructural protein 13: characterization of duplex-unwinding, nucleoside triphosphatase, and RNA 5'-triphosphatase activities

Multiple enzymatic activities associated with severe acute respiratory syndrome coronavirus helicase

The human coronavirus 229E superfamily 1 helicase has RNA and DNA duplex-unwinding activities with 5'-to-3' polarity

Functional screen reveals SARS coronavirus nonstructural protein nsp14 as a novel cap N7 methyltransferase

Okazaki fragment processing: modulation of the strand displacement activity of DNA polymerase delta by the concerted action of replication protein A, proliferating cell nuclear antigen, and flap endonuclease-1

Apparent activity in high-throughput screening: origins of compound-dependent assay interference

An Aggregation Advisor for Ligand Discovery

100 Years of Suramin

Suramin inhibits cullin-RING E3 ubiquitin ligases

Identification of HMGA2 inhibitors by AlphaScreen-based ultra-high-throughput screening assays

Suramin potently inhibits cGAMP synthase, cGAS, in THP1 cells to modulate IFN-beta levels

Protease Inhibitors by a Quantitative High-Throughput Screening

Suramin interacts with the positively charged region surrounding the 5-fold axis of the EV-A71 capsid and inhibits multiple enterovirus

Phosphorothioate oligonucleotides, suramin and heparin inhibit DNA-dependent protein kinase activity

Suramin inhibits helicase activity of NS3 protein of dengue virus in a fluorescence-based high throughput assay format

Naphthalene-sulfonate inhibitors of human norovirus RNA-dependent RNA-polymerase

Structural bases of norovirus RNA dependent RNA polymerase inhibition by novel suramin-related compounds

Structure-based inhibition of Norovirus RNA-dependent RNA polymerases

In silico screening for human norovirus antivirals reveals a novel non-nucleoside inhibitor of the viral polymerase

Feng BY, Shoichet BK. A detergent-based assay for the detection of promiscuous inhibitors

Combinatorial drug therapy for cancer in the post-genomic era

Remdesivir for the Treatment of Covid-19 -Final Report

Remdesivir and chloroquine effectively inhibit the recently emerged novel coronavirus (2019-nCoV) in vitro

Structural Basis for RNA Replication by the SARS-CoV-2 Polymerase

Viruses in bats and potential spillover to animals and humans

Structural basis for inhibition of the RNA-dependent RNA polymerase from SARS-CoV-2 by remdesivir

Structural Basis for Helicase-Polymerase Coupling in the SARS-CoV-2 Replication-Transcription Complex

Development of a small-molecule serum-and glucocorticoid-regulated kinase-1 antagonist and its evaluation as a prostate cancer therapeutic

The Dark Side of PI3K Signaling

Development of a new analog of SGK1 inhibitor and its evaluation as a therapeutic molecule of colorectal cancer

Serumand glucocorticoid-regulated kinase 1 is required for nuclear export of the ribonucleoprotein of influenza A virus

The Drugs of Sleeping Sickness: Their Mechanisms of Action and Resistance, and a Brief History

Virtual ligand screening of the p300/CBP histone acetyltransferase: identification of a selective small molecule inhibitor

p300 exerts an epigenetic role in chronic neuropathic pain through its acetyltransferase activity in rats following chronic constriction injury (CCI)

Inhibition of the acetyltransferases p300 and CBP reveals a targetable function for p300 in the survival and invasion pathways of prostate cancer cell lines

C646, a Novel p300/CREB-Binding Protein-Specific Inhibitor of Histone Acetyltransferase, Attenuates Influenza A Virus Infection

We thank Anne Early for assistance. This work was supported by the Francis Crick Institute, which receives its core funding from Cancer Research UK (FC001066), the UK Medical Research Council