A Repurposed Drug Interferes with Nucleic Acid to Inhibit the Dual Activities of Coronavirus Nsp13

The recent pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) highlighted a critical need to discover more effective antivirals. While therapeutics for SARS-CoV-2 exist, its nonstructural protein 13 (Nsp13) remains a clinically untapped target. Nsp13 is a helicase responsible for unwinding double-stranded RNA during viral replication and is essential for propagation. Like other helicases, Nsp13 has two active sites: a nucleotide binding site that hydrolyzes nucleoside triphosphates (NTPs) and a nucleic acid binding channel that unwinds double-stranded RNA or DNA. Targeting viral helicases with small molecules, as well as the identification of ligand binding pockets, have been ongoing challenges, partly due to the flexible nature of these proteins. Here, we use a virtual screen to identify ligands of Nsp13 from a collection of clinically used drugs. We find that a known ion channel inhibitor, IOWH-032, inhibits the dual ATPase and helicase activities of SARS-CoV-2 Nsp13 at low micromolar concentrations. Kinetic and binding assays, along with computational and mutational analyses, indicate that IOWH-032 interacts with the RNA binding interface, leading to displacement of nucleic acid substrate, but not bound ATP. Evaluation of IOWH-032 with microbial helicases from other superfamilies reveals that it is selective for coronavirus Nsp13. Furthermore, it remains active against mutants representative of observed SARS-CoV-2 variants. Overall, this work provides a new inhibitor for Nsp13 and provides a rationale for a recent observation that IOWH-032 lowers SARS-CoV-2 viral loads in human cells, setting the stage for the discovery of other potent viral helicase modulators.

Table S4.Structures of compounds selected for analysis from first round of screening.
Table S5.Structures of compounds selected for analysis from second round of screening.

Table S6. PDB ID and chain of the structural ensemble of SARS-CoV-2 Nsp13 cryo-EM structures following Principal Component Analysis (PCA) and Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN).
a These structures represent the Nsp13 dimer in the "backtracked" complex * Side chain contact frequency and residue interaction analysis from the WT Nsp13-IOWH-032 simulations with and without ATP bound.Average contribution of binding free energy per residue side chain (∆G )*+, -.*+/ -1*, (--)&4*,5 678*+ (46) ) with standard error and mean of the replicates reported.This analysis assumes that removal of the side chain does not affect the structure.
Purple colored text indicates results from the 5 out of the total 6 trajectories of IOWH-032 at the RNA site.Orange/brown colored text indicates 3 out of the total 6 of trajectories of IOWH-032 at the allosteric site.Removing these trajectories prioritized the determination of the residues that would interact with IOWH-032 during the analysis.22E).Percent prevalences were calculated using the Nextstrain bioinformatics software.The 3,972-genome library was filtered by selecting the variant of choice by clade, locating the desired mutation in Nsp13 (found between the coding nucleotide location 16240-18039 in the nucleotide/amino acid diversity map) and selecting the amino acid associated with the mutation.The number of sample genomes containing that mutation were then compared to the total for that variant. 5,6 e S1.Virtual screening of approximately 6000 compounds from drug repurposing library using thresholds of four different metrics. 7The left scatter plot shows the Lin_F9 score vs. ligand efficiency, while the right scatter plot shows the ligand β-score vs. ligand coverage.The compounds that had a Lin_F9 score <=-8.5 kcal/mol, ligand efficiency <= -0.28, ligand coverage >= 0.85, and ligand β-score <= -8.5 kcal/mol were selected from the screen.The selected compounds that meet the thresholds are highlighted in red.poses by predicted binding affinity were selected for each receptor.Each docking pose was represented by a ligand-receptor shortest distance vector.PCA and HDBSCAN are used to visualize and cluster the docking location to three sites, allosteric (brown), RNA #1 and #2 (purple), and ATP (orange).Table S7 contains the breakdown of the docking location, frequency, and binding affinity predictions per receptor cluster and ligand pose location cluster.(B) Top ranked pose of IOWH-032 (lime green stick) when blindly docked by ΔLin_F9XGB to an Nsp13nucleotide analog complex (PDB ID: 7NN0) indicates that IOWH-032 directly overlaps with ANP-PNP (Adenylyl imidodiphosphate, a non-hydrolyzable ATP-analog, orange stick).This pose suggests that IOWH-032 would be competitive with ATP, which was not observed experimentally, and so other poses were pursued.
The elution fractions were combined and immediately loaded onto 2 x 1 mL HiTrap SP HP columns (Cytiva) for cation exchange, using a 50 mL Super Loop (Cytiva), οn the ÄKTA FPLC.Both were pre-equilibrated with elution buffer.After loading, the column was then washed with 2 CV of elution buffer B (1 mL/min) and proteins were eluted with 3 CV of Hi-Salt buffer (1 mL/min).The collected sample was incubated overnight while rocking at 4 ºC with TEV protease (1:40 mass ratio).
The sample was concentrated using an Amicon centrifugal unit (MilliporeSigma, 30K MWCO, 3732 x g, 4ºC).The protein was further purified via gel filtration on a HiLoad 16/600 Superdex 200 prep grade column (Cytiva) that was equilibrated with gel filtration buffer (50 mM HEPES-NaOH pH 7.5, 500 mM NaCl, 5 mM MgCl2, 5% (v/v) glycerol, 0.5 mM TCEP) using the ÄKTA FPLC.500 μL of protein was injected on the column and eluted over an isocratic flow of gel filtration buffer at 1 mL/min, over 1 CV.Purified Nsp13 was concentrated as described above.Samples were then aliquoted, flash frozen with liquid N2, and stored at -80 ºC.Unless otherwise noted, this construct was used for most experiments due to improved stability and yields relative to the initial Nsp13 construct (6xHis PreScission SARS-CoV-2 Nsp13), above.
For purification of Nsp13 mutants, due to the reduced stability of constructs, some steps were slightly changed.Namely, post-cation exchange chromatography, the protein was diluted with gel filtration buffer (50 mM HEPES-NaOH pH 7.5, 500 mM NaCl, 5 mM MgCl2, 5% (v/v) glycerol, 0.5 mM TCEP) using a 1:5 dilution, and glycerol was added to increase the final glycerol percentage to 10% for the incubation with TEV Protease overnight.

Purification of pETHisSUMO HCV NS3 Helicase Expression
A truncated form of the Hepatitis C NS3 helicase protein (GenBank No. AJ238799.1:3918 to 5321) obtained from Twist Biosciences was cloned into a pETHis-SUMO expression vector. 2 Expression and purification protocols followed previous literature. 10In brief, the plasmid was transformed into E. coli BL21 (DE3) cells (Novagen) and grown overnight at 37 °C on LB-agar plates containing carbenicillin (50 μg/mL).Isolated colonies were selected for small culture inoculation into 5 mL of LB media supplemented with carbenicillin (50 μg/mL) and grown with shaking overnight (37 °C, 200 RPM).These were then used to inoculate 500 mL of LB supplemented with carbenicillin (50 μg/mL) and grown to log-phase (OD600 = 0.6, optical density at 600 nm) with shaking (37 ºC, 200 RPM).Cultures were then induced with 0.5 mM IPTG for 17 hours with shaking (16 ºC, 200 RPM).The cells were pelleted through centrifugation (3732 x g, 30 min) and stored at -80 ºC.
A point mutation of NS3 was expressed and purified in the same manner.

Nsp13 ATPase Inhibition Assays
For Nsp13 ATPase assays, reactions were prepared with indicated amounts of protein in Nsp13 ATPase Buffer (25 mM HEPES-NaOH pH 7.5, 50 mM NaCl, 5 mM MgCl2, 1 mM DTT, 0.01% Tween-20, unless otherwise noted).Compound or DMSO vehicle was added, making the final DMSO concentration 10%.Reaction components were incubated at 25 ºC for 15-30 min and then initiated with a final concentration of 250 μM ATP, unless otherwise noted, bringing the final reaction volume to 15 μL.Reactions were then quenched with 15 μL of 0.05% formic acid at the designated time point.
For Kinase Glo (Promega) analysis, quenched reactions (15 μL) were added to a 96-well half area opaque plate (Corning) with 15 μL of Kinase Glo reagent (Promega) and incubated for 45 min with shaking (25 ºC, 200 RPM).Plates were then analyzed by measuring luminescence on a plate reader (Molecular Devices, Spectra Max iD5, 1 s integration; or Molecular Devices, FlexStation 3) using SoftMax Pro 7.1 software.Concentrations of ATP remaining were calculated using an ATP standard curve in Microsoft Excel, and analyzed/graphed in GraphPad Prism 9 software.Normalized percent inhibitions were calculated based on reactions lacking compound with only vehicle (DMSO) added (0%), and reactions without Nsp13 or with a catalytically inactive mutant containing vehicle only (100%).
For ADP-Glo (Promega) analysis, quenched reactions (10 μL) were added to a 96-well half area opaque plate (Corning) with 10 μL of ADP-Glo reagent (Promega) and incubated for 45 min with shaking (25 ºC, 200 RPM).Kinase Detection Reagent (20 μL) (Promega) was then added to each well and incubated for an additional 45 min with shaking (25 ºC, 200 RPM).Plates were then analyzed by measuring luminescence on a plate reader (Molecular Devices, Spectra Max iD5, 1 s integration) using SoftMax Pro 7.1 software.Concentrations of ADP were calculated using an ATP-to-ADP standard curve in Microsoft Excel, and analyzed/graphed in GraphPad Prism 9 software.Normalized percent inhibitions were calculated based on reactions lacking compound with only vehicle (DMSO) added (0%), and reactions without Nsp13 containing vehicle only (100%).
For ATPase IC50 measurements, reactions were set-up as stated above but with varying concentrations of compound.Compound was serially diluted from the designated maximum concentration, using 2-fold dilutions, and added to designated reactions making the final DMSO concentration 10%.Reactions were initiated with the noted ATP concentrations.KinaseGlo reagents (Promega) were used as previously described to determine the ATP remaining for each reaction.Normalized percent inhibitions were calculated based on reactions with only vehicle (DMSO) added (0%), and reactions without Nsp13 containing vehicle (100%).The data was fit in Prism 9 using the following equation: Y = 100/(1+10^((LogIC50-X)*HillSlope)).Hill slopes were allowed to vary to obtain the best fit and because Nsp13 is predicted to form a dimer, which suggests that compound:protein binding is likely not 1:1.
For ATPase kinetics assays, reactions were set-up as stated above but with varying [ATP].ATP was serial diluted from 1 mM, using a 2-fold serial dilution.The reactions were then initiated using these varying concentrations of ATP.ADP-Glo reagents (Promega) were used as previously described to determine the [ADP] produced for each reaction.Michaelis-Menten parameters Km and Vmax were calculated by fitting initial rates (μM/sec) vs. [ATP] (μM) in Prism 9 with the following equation: Y = Vmax*X/(Km+X).Kcat was calculated by dividing Vmax by total [enzyme].

HCV NS3 ATPase Inhibition Assays
For NS3 ATPase assays, reactions were prepared with indicated amounts of protein in the ATPase Buffer B (25 mM 4-Morpholinepropanesulfonic acid (MOPS), pH 6.5, 1.25 mM MgCl2, 5% DMSO, 50 µg/mL BSA, 0.01% Tween-20).Compound or DMSO was added, making the final DMSO concentration 5%.Reaction components were incubated at 23 ºC for 10 min and then reactions were initiated with 250 μM ATP, bringing the final reaction volume to 15 μL.Reactions were then quenched with 15 μL of 0.05% formic acid at the designated time point.Kinase Glo (Promega) was used as previously described to quantify ATP remaining.Plates were then analyzed by measuring luminescence on a plate reader (Molecular Devices, Spectra Max iD5, 1 s integration) using SoftMax Pro 7.1 software and analyzed/graphed in GraphPad Prism 9 software.Normalized percent inhibitions were calculated based on reactions with only vehicle (DMSO) added (0%), and reactions without Nsp13 containing vehicle only (100%).
For ATPase IC50 measurements, reactions were set-up as stated above but with varying concentration of compound.Compound was serially diluted from the designated maximum concentration, using 2-fold dilutions, and added to designated reactions making the final DMSO concentration 5%.Compounds were initiated with ATP.Kinase Glo reagents (Promega) were used as previously described to determine the ATP remaining for each reaction.Normalized percent inhibitions were calculated based on reactions with only vehicle (DMSO) added (0%), and reactions without NS3 containing vehicle only (100%).The data was fit in Prism 9 using the following equation: Y = 100/(1+10^((LogIC50-X)*HillSlope)).Hill slopes were allowed to vary to obtain the best fit, as it was not known if the compound would bind with 1:1 stoichiometry.

Tte UvrD ATPase Inhibition Assays
Tte UvrD helicase protein was obtained from New England Biolabs (Cat.#M1202S) and the manufacturer's protocol was followed using their recommended assay conditions: 62.5 nM DNA (same DNA duplex as used in Nsp13 helicase inhibition assays), 1 mM ATP, 2 µL of Isothermal Amplification Buffer (NEB).Compound or DMSO was added, making the final DMSO concentration 5%.Reaction components were incubated at 65 ºC for 10 min and then initiated with 5 ng of Tte UvrD helicase, bringing the final reaction volume to 20 μL.Reactions were then quenched with 20 μL of 20 mM EDTA at the designated time point.Activity was quantified using the previously mentioned Kinase Glo reagents in a 384-well, flat bottom, opaque microplate (Grenier).Plates were then analyzed by measuring luminescence on a plate reader (Molecular Devices, Spectra Max iD5, 1 s integration) and analyzed/graphed in GraphPad Prism 9 software.

Nsp13 Helicase Inhibition Assays
FRET-based helicase assays were conducted with indicated amounts of protein in Nsp13 Helicase Buffer (20 mM HEPES-NaOH pH 7.5, 20 mM NaCl, 5 mM MgCl2, 1 mM DTT, 0.01% Tween-20 unless otherwise noted) with 50 nM labeled DNA-duplex, unless otherwise noted.The labeled DNA duplex was obtained from Integrated DNA Technologies by pre-annealing Oligomer 1 (5'-/Cy3/CCTCGTAGGTACGCATGGATCCTCT-3') and Oligomer 2 (5'-GCTTGCATGCCTGCAGGTGCACTCTAGAGGATCCATGCGTACCTACGAGG/BHQ_2/-3').These oligomers were annealed using a Bio-Rad Laboratories T100 TM Thermocycler (95 ºC for 5 minutes, -1ºC/cycle for 1 min/cycle (70 cycles), and then held at 4ºC).Compound or DMSO was added, making the final DMSO concentration 10%.Reaction components were mixed and incubated at 28 ºC for t = 15 min.Then 36 μL of reaction was added to a 96-well half area opaque plate (Corning) and the reaction was initiated with 40 mM ATP (4 μL), unless otherwise indicated, using auto-injectors in a plate reader (Molecular Devices, Spectra Max iD5), resulting in a final reaction volume of 40 μL and an initial ATP concentration of 4 mM, unless otherwise stated.Helicase unwinding was tracked at 28 ºC by measuring fluorescence (excitation: 515 nm and emission: 564 nm) every 0.5 s for 90 sec or over an indicated time course.The linear phase was analyzed and plotted using Prism 9 software.
For IC50 measurements, reactions were set-up as stated above but with varying concentration of compound.Compound was serially diluted from the designated maximum concentration, typically using 2-fold dilutions, and added to designated reactions lacking ATP making the final DMSO concentration 10%.Reactions were initiated with ATP using auto-injectors in the plate reader.Helicase unwinding was tracked as previously described.Normalized percent inhibition values were calculated using the RFU/sec of reactions with only vehicle (DMSO) added (0%), and the RFU/sec of reactions without Nsp13 containing vehicle only (100%).Each of the Nsp13 constructs were tested at concentrations that provides a measurable initial linear phase for helicase activity.The following concentrations were used: 25 nM: R21A, R21W, E142W; 12.5 nM: E341D, T380A, Y180A, E142A; 3.125 nM: P77L, R392C.The data was fit in Prism 9 using the following equation: Y = 100/(1+10^((LogIC50-X)*HillSlope)).Hill slopes were allowed to vary to obtain the best fit, as stated above.
For helicase kinetic analysis, the [labeled DNA-duplex] was varied as indicated.Michaelis-Menten parameters, KM and Vmax, were calculated by fitting initial rate (RFU/sec) vs.
The plates were rocked and measured using the same method and plate reader as described for the direct binding assays above.Data was analyzed in Prism with controls of duplex-FL alone and duplex-FL with saturating Nsp13 construct concentrations, and half-maximal effective concentration (EC50) values were calculated for each with the equations: logEC50=logEC50 -(1/HillSlope)*log(50/(100-50)) and Y=Bottom + (Top-Bottom)/(1+10^((LogEC50-X)*HillSlope)).KI values were then calculated from the EC50 and KD values using previously described methods. 14

Cytotoxicity assay against infection cell lines
Vero E6 ACE2-T2A-TMPRSS2 (BEI Resources, NR-54970), Vero E6 (ATCC #CRL-1586), and A549-ACE2 cells 15 (kind gift of Meike Dittmann) were cultured in 1x Dulbecco's modified Eagle's medium (Corning #10-013-CV) supplemented with 10% heat-inactivated FBS, 1x penicillin-streptomycin, and 2 mM L-glutamine and grown at 37 °C with 5% CO2.Cells from each line were plated in tissue culture-treated black-walled and -bottomed 96 well plates with 1.5 x 10 4 cells in 100 µL of culture media per well.Plates were incubated overnight at 37 °C with 5% CO2.The following day cells were washed with 1 x PBS (+Ca 2+ +Mg 2+ ) and active compound, inactive compound, and DMSO vehicle control were added in triplicate, while remaining cells had no additives.All compounds were diluted in treatment media (1 x DMEM (Gibco #11-965-092) supplemented with 2% heat-inactivated FBS.Compounds were 2-fold serially diluted starting at 20 µM and ending at 156.25 nM (all in a final concentration of 0.1% DMSO when incubated with cells).Cells were incubated overnight (t = 22 hr) at 37 °C with 5% CO2.The next day cells were fixed with 4% paraformaldehyde for 1 hour at room temperature then permeabilized with 3% bovine serum albumin in 1 x PBS with 0.1% Triton X-100 for 30 min at room temperature.Following permeabilization, cells were stained with DAPI at 1.25 µg/mL diluted in 3% BSA in 1 x PBS for 1 hour at room temperature.Following staining, all plates were read on a CyTek CyTation 7 plate imager for cell counts.CC50 values were calculated in GraphPad Prism 9.5.1 using a nonlinear fit with variable slope.

Initial virtual screen
When the project began in 2020, the only structure available was a homolog to the SARS-CoV-2 Nsp13, the SARS-CoV-1 Nsp13 (PDB ID: 6JYT). 16This structure shares a 99.8% sequence identity to the SARS-CoV-2 Nsp13, with a single point mutation difference at I150V at a site located away from the ATP and RNA binding sites. 17Around 6,000 compounds from the Drug Repurposing Hub of the Broad Institute were first screened to the ATP binding site of this structure. 7DKit version 2020.03.1 18 was used to read in the compounds in the SMILES format, add hydrogens, and generate an initial low energy 3D conformer for each compound using a distance geometry algorithm with MMFF94 force field. 19OpenBabel 2.4.1 was then used to generate up to 10 conformers per compound using the genetic algorithm with the default options. 20AlphaSpace was then used to perform fragment-centric topographical mapping of the SARS-CoV Nsp13 surface. 21Two targetable binding pockets, the ATP and RNA binding sites, were identified as highly targetable binding sites.The OpenBabel-generated conformers were docked using flexible ligand docking with a fork of the Smina docking suite 22 that has Lin_F9 scoring function built-in 23 and retaining only the top 5 poses.The default docking parameters of Smina, such as exhaustiveness = 8, num_modes = 9, autobox_add = 4 were used for docking.The top 5 poses were retained, and 24 compounds were selected using thresholds of Lin_F9 score <= -8.5 kcal/mol, ligand efficiency <= -0.28, ligand coverage >= 0.85, ligand β-score <= -8.5 kcal/mol (Figure S1).Later, the ΔLin_F9XGB scoring function 24 was developed and used to rescore the top 5 docked poses.In the second screen, we replaced the Lin_F9 and ligand efficiency thresholds ∆Lin_F9XGB score <= -9.0 kcal/mol and ligand efficiency <= -0.30.From this screen, 104 compounds were selected (Figure S3).

Blind ensemble docking
The hit compound IOWH-032 was blindly docked to the ensemble of SARS-CoV-2 receptor structures.
The ensemble of structures was parsed and analyzed using the beta version of the python package dist_analy (https://github.com/echen1214/dist_analy).The corresponding data and tutorials are available on (https://github.com/echen1214/nsp13_sbdd_md_data).The available 12 and 60 cyro-electron microscopy (cryo-EM) and X-ray crystallography structures, respectively, of SARS-CoV-2 Nsp13 as of 08/01/22 were retrieved from the Protein Data Bank using the pyPDB API [UniProt: P0DTD1]. 25,26 urther parsing and splitting of the structures resulted in 24 and 61 monomer chains from cryo-EM and X-ray crystallography, respectively.Cryo-EM structures of Nsp13 in the replication-transcription complex (RTC) is often observed with two protomers in complex with the RNA-dependent RNA-Polymerase (RdRp).We followed the nomenclature proposed by Hillen, 27 which defines Nsp13T (binding on the 'Thumb' side of the RdRp) and Nsp13F (binding on the 'Fingers' side of the RdRp).Next, a machine learning protocol using Scikit-learn was applied to analyze the conformational ensemble of the structure. 28A pairwise residueresidue shortest distance matrix was calculated for each monomer chain to represent the protein structure.Then, two unsupervised machine learning algorithms were used to analyze the structural ensemble.Principal Component Analysis (PCA) was used to visualize the variance of the cryo-EM structural features set along the principal components. 29The data points were projected on the first two principal components which comprise 92.6% of the explained variance ratio.Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) was used to cluster the projected data points into three clusters. 30The X-ray crystallography structures are then projected onto the PCA plot (Figure S8, Table S6).The resulting plot and clustering underscore the conformational diversity of the cryo-EM structures in contrast to the conformational homogeneity of the resolved X-ray crystal structures.Cluster 1 (green) was characterized by Nsp13T monomer in the open state, whereby it is proposed the RNA was not engaged with the helicase and allowing the RdRp to perform its translocation and RNA synthesis activity.Cluster 2 (red) was characterized by Nsp13T monomer in the engaged and swivel conformations, whereby it was proposed RdRp was in the backtracking complex.Cluster 2 (cyan) was represented by the Nsp13F or apo Nsp13T conformations.When projected on the PCA plot, all the X-ray crystal structures were located near cluster 2. Each cryo-EM and X-ray crystal receptor conformation is then prepared by using the pdb2pqr30 v3.5.2 to add hydrogens and rebuild any missing side chains, and MGLTools 1.5.6 `prepare_receptor4.py` to add Gasteiger charges and AutoDock Vina types and to remove the non-polar hydrogens to the receptor. 31,32 arting from the SMILES string of IOWH-032, Open Babel 3.1.0was used to add hydrogens and generate 20 conformers using the using the genetic algorithm with the default options. 20Each ligand conformer was prepared by using Meeko 0.3.3 to add Gasteiger charges and AutoDock Vina types, and remove the non-polar hydrogens.Each conformer was blindly docked to the ensemble of receptors with a fork of the Smina docking suite 8 that built in the Lin_F9 scoring function. 23ach docking score was re-scored using ΔLin_F9XGB to provide more accurate binding affinity predictions. 24ΔLin_F9XGB used an ensemble of eXtreme Gradient Boosting (XGBoost) trees to predict a correction to the initial Lin_F9 score.ΔLin_F9XGB required a multitude of features to input into XGBoost version 1.2.0.The 48 Vina features were generated using AutoDock Vina 1.1.2The SASA features were calculated with MSMS 2.6.1.The β-cluster features were calculated using AlphaSpace 2.0 and the Vina scoring function.The ligand features were generated with RDKit 2020.09.3.The sum of the Lin_F9 score and the XGBoost corrections results in the final binding affinity prediction (pKd).The top 5 scoring poses for each receptor were used for further analysis.
To analyze the binding location of IOWH-032 following ensemble docking, the shortest ligand-receptor distance vector for each pose was calculated.This vector characterized the location of the docking pose relative to the receptor.These vectors were then analyzed with PCA and HDBSCAN.Here the docking poses and corresponding binding scores were clustered into three distinct binding sites, the ATP, RNA and one allosteric binding site (Figure S9A, Table S7).The RNA binding pose could be distinguished into two sub-clusters RNA #1 and #2.The RNA #1 pose bound in the center of the RNA binding pocket at the core of the protein.The RNA #2 pose bound at the proximal region of the RNA binding site (with respect to the view seen in Figure 1B).IOWH-032 was only found docked to the allosteric binding site when docked to 7RDX chain E from cluster 1.The cluster 1 structures reflected the 1B-domain swiveled to be near the Zincfinger domain and created the allosteric binding pocket that was unique to this cluster of structures.

Computational Methods
Molecular dynamics (MD) simulations were performed for IOWH-032 bound to Nsp13 helicase with and without ATP bound.The corresponding parameter files are available on (https://github.com/echen1214/nsp13_sbdd_md_data).The starting poses of the IOWH-032 and Nsp13 complex were determined following the blind docking of IOWH-032 to the structural ensemble of Nsp13 (Figure S8).The ensemble docking results suggested three sites that IOWH-032 could bind: ATP, RNA, and allosteric sites.We did not run MD simulations of the ATP site because initial experiments indicated that IOWH-032 did not bind at that site (Figure 4A).The allosteric site-bound state was represented by the top 1 ranking ΔLin_F9XGB pose docked to the Nsp13T protomer taken from the class IV (1B-open) cryo-EM structure of the replicationtranscription complex [PDB ID: 7RDX; chain E]. 33 This pose was found to be clustered to the allosteric cluster.The resolved ADP molecule was modified to ATP by providing the ATP parameter file and using tleap to rebuild the missing phosphate. 34,35 he RNA site-bound state was represented by the top 1 ranking ΔLin_F9XGB score pose docked to the Nsp13 monomer taken from chain A of the apo Nsp13 structure [PDB ID: 6ZSL; chain A]. 9 This pose was found to be clustered to the RNA #1 cluster.The ATP and Mg 2+ was modelled into this structure by aligning the AMP-PNP-and Mg 2+ -bound Nsp13 (PDB ID: 7NN0) 9 and modifying the amide between the β-and γ-phosphates in AMP-PNP to oxygen.The missing loop regions were remodeled using Modeller. 36he system preparation, minimization and equilibration followed the methods reported by Weber and McCullagh 37 and are summarized below.The protein was modelled using ff14SB parameters 38 , and ATP 34 and Mg 2+ 39 were modelled using parameters obtained from the AMBER parameter database.Each individual IOWH-032 molecule pose was parameterized.First, Open Babel 3.1.0was used to protonate the ligand molecule. 20Then, the restrained electrostatic potential (RESP) charges of the pose was determined in three steps using Gaussian 16, Revision A.03 using B3LYP/6-31G* functional/basis set and antechamber. 40,41 he pose was first geometry optimized and then the electronic potential was calculated with Merz-Singh-Kollman scheme. 42,43 nally, antechamber was used to calculate the RESP charges and any missing bonded parameters were parameterized with GAFF2.The protein structures were protonated using the H++ 4.0 webserver at pH 7. 44,45 All crystallographic waters and zinc ions were retained.The Nsp13 zinc binding domain contains three zinc binding pockets.The three zincs and neighboring residues within 2.8 Å of the zinc were parameterized in CYM-CYM-CYM-CYM, CYM-CYM-CYM-HID, and CYM-CYM-HID-HIE environments using MCPB.pyversion 7.0 tool. 46TIP3P water was added to each system with at least a 12 Å buffer.Na + and Cl -ions were added to neutralize the charge and provide a 0.1 M salt concentration using the SLTCAP webserver. 47ll-atom, explicit solvent MD simulations for the two initial starting poses of the IOWH-032, ATP, Nsp13 complex were performed using the GPU-enabled AMBER22.03 software. 35ydrogen atoms were constrained using the SHAKE algorithm. 48Direct nonbonding interactions were cut off at 12, and long-range electrostatic interactions were modeled using the particle-mesh Ewald treatment. 49Each simulation in the NPT ensemble was run with the Monte Carlo barostat set to 1 atm, Langevin thermostat set to 300 K and integration time step of 2 fs.
Each system followed the same minimization, equilibration, and production protocols.The system was minimized in 10 stages of 2000 steps of steepest descent minimization with varying harmonic restraints.The first stage used 500 kcal mol -1 Å -2 restraints on all protein, ATP, and ligand atoms.In the next 4 stages, the restraint on the protein side chains was progressively lowered to 10.0, 1.0, 0.1, and 0.0 kcal mol -1 Å -2 .In the final 5 stages, the restraint on the protein backbone, ATP, and IOWH-032 ligand was progressively lowered to 50.0, 5.0, 1.0, 0.1, and 0.0 kcal mol -1 Å -2 .This minimized system spawned a triplicate of simulation that were equilibrated and then run separately.The system was heated from 0 to 300 K with a harmonic restraint of 40 kcal mol -1 Å -2 on all protein, ATP, and ligand atoms.The pressure was then equilibrated in 7 stages.First, a 1 ns of NVT was run with the same harmonic restraints.In the following 6 pressure equilibration stages of 200 ps each, the restraint was reduced to 20.0, 10.0, 5.0, 1.0, 0.1, and 0 kcal mol -1 Å -2 .Finally, a conventional simulation was run for 100 ns with a snapshot saved at every 0.1 ns.This output was then used for analysis of the potential contacts between the IOWH-032 and Nsp13 at each site.

Analysis of MD simulations to predict residues that interact with IOWH-032
Analyses of the conventional MD simulations to select the residues for alanine/tryptophan mutations were performed using the CPPTRAJ V6.18.1 tool and the MMPBSA.pyversion 14.0 tools. 46,50 rajectories and frames that did not retain the IOWH-032 molecule in the binding pocket were discarded from further analysis.This was determined by aligning each frame to the starting frame and calculating the ligand heavy atom RMSD with the `rmsd` command.The trajectories with many frames with RMSD > 6 Å were discarded (Figure S11).Removing these trajectories prioritized the determination of the residues that would be interacting with IOWH-032 during the analysis.This analysis assumes that removal of the side chain does not affect the structure.
To select residues for mutation, the side chain contact frequency (< 4.5 Å) with the ligand and the side chain energetic contributions to binding were measured per residue.The side chain contact frequency was measured using the `nativecontacts` command.The side chain energetic contributions to binding were calculated by performing binding free energy calculations using the MMPBSA.pytool. 513][54] The calculations were carried out using "single-trajectory" protocol on all frames. 51This protocol stripped the trajectory to create bound/unbound and solvated/un-solvated ensembles.The polar solvation free energy contribution was determined by solving the PB equation using the classical nonpolar solvent model (inp = 1) and atomic radii according to the parameter file 54 or by applying the "OBC" GB method using mbondi2 radii (igb = 5). 55Both were computed at 100 mM ionic strength and with dielectric constant of 1 and 80 for the interior and exterior of the solute, respectively.The nonpolar solvation free energy contribution was calculated with the solvent accessible surface area (SASA) method. 56The change in conformational entropy was neglected.The per residue contribution of the side chain to the free energy was calculated for both the GB and PB models. 57The scheme that added the 1-4 terms to the internal potential terms (idecomp = 1) was used.This analysis is referenced in the main text as "per-residue side chain binding energy decomposition".
The residues that retained side chain contacts frequency > 0.9 and had an absolute side chain energetic contribution in at least one trajectory > 0.75 kcal/mol were selected.Of this subset of residues, five residues from each binding site that interacted with the ligand across the binding interface and range in polarity were selected for experimental mutagenesis.

Mutated Nsp13 simulation experiments
The mutations Y180A, E142A and E142W were introduced in the simulations to confirm the observed effect of mutation on binding.The ChimeraX `swapaa` command was used to mutate these residues to the desired side chain through sampling conformations from the Dunbrack rotamer library. 58,59 o determine the initial pose of E142W-IOWH-032 complex, IOWH-032 was re-docked to the E142W structure and the top ΔLin_F9XGB pose with IOWH-032 in the RNA site was selected.Otherwise, the force field parameters, system setup, minimization, equilibration, and production used for the wild-type complex were used in these simulations.No trajectories were removed during this analysis and ATP-apo state simulations were not run.
The binding free energies of IOWH-032 to the Nsp13-ATP complex (ΔG WT ) were determined with MM-PBSA and MM-GBSA calculations.These calculations were performed on each replicate using the "single-trajectory" protocol and following the same parameters as described above in the "Analysis of MD simulations to predict residues that interact with IOWH-032" section.The difference in binding energies between wildtype and mutants are calculated by ∆∆G = ∆G <=: − ∆G 9:

Figure S2 .
Figure S2.First round of inhibitors tested for inhibition of Nsp13 ATPase activity led to identification of initial hits.Compounds were incubated with Nsp13 (pET28a 6xHis PreScission SARS-CoV-2 Nsp13) at indicated concentrations and ATP consumption was calculated within the initial linear phase (t = 5 min).Normalized percent inhibitions were calculated based on reactions lacking compound with only vehicle (DMSO) added (0%), and reactions with inactive mutant E375A (pSRC323) containing vehicle only (100%), as shown.The following compounds were considered "active" based on this assay: AST, ABT, BAY, ST, TEL and DBZ, with the latter two showing the greatest inhibition.Structures are shown in Table S4.No detergent was used in these assays.Error bars represent standard deviation (SD) for n = 3 experiments.

Figure S3 .
Figure S3.Virtual screening of approximately 6000 compounds from the drug repurposing library using thresholds of four different metrics and an updated scoring function.The left scatter plot shows the ΔLin_F9XGB score vs. ligand efficiency, while the right scatter plot shows the ligand β-score vs. ligand coverage.The compounds that had ∆Lin_F9XGB score <= -9.0 kcal/mol, ligand efficiency <= -0.30, ligand coverage >= 0.85, and ligand β-score <= -8.5 kcal/mol were selected.The selected compounds that meet the thresholds are shown as red dots.Previously screened active compounds (telatinib (TEL) and dibenzazepine (DBZ)) are shown as cyan dots.

Figure S5 .
Figure S5.Helicase activity and inhibition can be measured using a FRET-based assay.(A) The rate and extent of labeled double-stranded DNA unwinding varies based on the concentration of Nsp13 used (pET28a 6xHis PreScission SARS-CoV-2 Nsp13).(B) Analysis of helicase unwinding by Nsp13 (35 nM) in the presence and absence of ATP indicates that double stranded DNA unwinding is ATP-dependent.(C-D) Titrations of Telatinib and IOWH-032, respectively, into Nsp13 helicase reactions lacking detergent.Normalization was performed by comparison to reactions lacking compound (0% inhibition, only DMSO present) and reactions lacking enzyme (100% inhibition).Experiments were performed at a final volume of 40 uL with 4 mM ATP and Nsp13 (pET28a 6xHis PreScission SARS-CoV-2 Nsp13, 3.12 nM).Error bars represent SD for n = 3 (parts B-D).

Figure S6 .
Figure S6.Kinetic analyses indicate that IOWH-032 does not compete with ATP.(A) Michaelis-Menten analysis of Nsp13 (25 nM) ATPase activity with increasing [ATP] in the presence and absence of IOWH-032 indicates an uncompetitive mode of inhibition, as seen in Figure 4A with a higher concentration of IOWH-032 tested.(B) Michaelis-Menten analysis of Nsp13 (25 nM) helicase activity with increasing [oligonucleotide duplex] in the presence and absence of increasing [IOWH-032] indicates mixed inhibition, as seen in Figure 4B.Error bars indicate SD for n = 2 in part A and standard error of the mean (SEM) for n = 2 for part B.

Figure S7 .
Figure S7.Fluorescence polarization (FP)-based analysis indicates that IOWH-032 competes with oligonucleotide for binding to Nsp13.(A) Competition experiment using unlabeled duplex oligonucleotide containing a scrambled sequence (based on that of the duplex DNA probe) titrated into Nsp13 (200 nM) leads to displacement of duplex-FL (100 nM) in 10% DMSO (EC50 = 421 nM).(B) Titration of IOWH-032 into solution containing duplex DNA-FL (100 nM) and 10% DMSO shows that IOWH-032 does not interact with probe alone.For A-B, dashed line indicates the polarization of probe only, and incubations were performed for t = 60 min each.(C-G) Saturation binding experiments for Nsp13 wild-type (WT) and indicated mutants using 20 nM of duplex-FL in 10% DMSO.Note that the calculated KD for WT is different than that in Figure 4D, as a different concentration of probe was used, and incubations extended t = 30-180 min for these experiments.Error bars represent SD (n = 3 except n = 2 for part B).

Figure S8 .
Figure S8.Structural ensemble analysis of the SARS-CoV-2 Nsp13 structures.Each structure is represented by a pairwise residue-residue shortest distance matrix.PCA and HDBSCAN are used to visualize and cluster the cryo-EM data representations.Table S6 contains the list of the corresponding structures and chains.The X-ray crystal structures are projected onto the corresponding PCA plot.Highlighted in pink are the two structures used for the initial pose of the MD simulations.All structures are used to perform the ensemble docking of IOWH-032.

Figure S9 .
Figure S9.Ensemble docking pose analysis of the docking location of IOWH-032 to every structure of SARS-CoV-2 Nsp13 and analysis of possible ATP binding pose.(A) The top 5 poses by predicted binding affinity were selected for each receptor.Each docking pose was represented by a ligand-receptor shortest distance vector.PCA and HDBSCAN are used to visualize and cluster the docking location to three sites, allosteric (brown), RNA #1 and #2 (purple), and ATP (orange).TableS7contains the breakdown of the docking location, frequency, and binding affinity predictions per receptor cluster and ligand pose location cluster.(B) Top ranked pose of IOWH-032 (lime green stick) when blindly docked by ΔLin_F9XGB to an Nsp13nucleotide analog complex (PDB ID: 7NN0) indicates that IOWH-032 directly overlaps with ANP-PNP (Adenylyl imidodiphosphate, a non-hydrolyzable ATP-analog, orange stick).This pose suggests that IOWH-032 would be competitive with ATP, which was not observed experimentally, and so other poses were pursued.

Figure S10 .
Figure S10.Helicase mutants demonstrate helicase activity at nanomolar concentrations, some with lower rates compared to wild-type protein.(A-K) The initial rates of each tested mutant of Nsp13 at 25, 12.5, 6.25, and 3.125 nM concentrations.Each graph of was normalized based on the maximum RFU/sec value of each data set.(L) Standard curve generated by measuring the fluorescence of increasing concentrations of labeled double stranded DNA (DNA-FL: Oligomer 1-(5'-/Cy3/CCTCGTAGGTACGCATGGATCCTCT-3') and Oligomer 2-(5'-GCTTGCATGCCTGCAGGTGCACTCTAGAGGATCCATGCGTACCTACGAGG-3').This probe mimics unquenched DNA as it unwinds via helicase activity.(M) Initial rates of indicated Nsp13

Figure S11 .
Figure S11.Heavy atom RMSD of IOWH-032 to the first frame after aligning each frame to the initial structure by using the protein structure heavy atoms as the reference.The 6ZSL chain A simulation is run with IOWH-032 bound in the RNA binding site and the 7RDX chain E simulation is run with IOWH-032 bound in the allosteric site.Trajectories with RMSD(IOWH-032) > 6 Å were discarded prior to the initial residue interaction analysis.Trajectory 1 of the 6ZSL chain A simulation, trajectory 1 and the first 30 ps of trajectory 2 of the 7RDX chain E simulations, and trajectories 2 and 3 of the 7RDX chain E with ATP bound simulations were discarded.The high RMSD observed for half of the allosterically bound trajectories suggest the instability of the compound at that site.In contrast, the lower RMSD observed for most of the RNA-site bound trajectories support binding of IOWH-032 to the RNA interaction site.

Figure S12 .
Figure S12.IOWH-032 does not inhibit other microbial helicases at micromolar concentrations.(A) Titration of 250 nM Hepatitis C Virus (HCV) RNA helicase NS3 with IOWH-032 and 250 μM ATP showed that ATPase activity decreases only when concentrations of IOWH-032 exceed 100 µM.NS3 (250 nM) was incubated with 250 µM ATP, and indicated concentrations of IOWH-032.(B) Analysis of indicated concentrations of WT NS3 versus a point mutant in a catalytic residue shows that ATP (250 µM) consumption relies on active NS3.(C) The consumption of ATP (1 mM) at 65 ºC by the bacterial DNA helicase Thermoanaerobacter tengcongensis (Tte) UvrD (3.02 nM) is not affected by increasing concentrations of IOWH-032.Error bars represent SD for n = 3.

Figure S13 .
Figure S13.IOWH-032 is toxic against commonly used mammalian infection cell lines.(A-B) IOWH-032 and analog 2 were titrated into indicated cell lines in 0.1 % DMSO.(C) Constant concentration of vehicle (0.1% DMSO) was added to indicated cell lines across a multiwell plate row.Cells were incubated at 37 °C for t = 22 hr prior to measuring viability.Percent viability was calculated by comparison to cells that lacked any added inhibitor or vehicle.Vero cell lines are African Green Monkey kidney cells; the ACE2-TMPRSS2 subline overexpresses the human receptor and protease.A549 ACE2 cells are human lung adenocarcinoma cell lines that overexpress human ACE2.Error bars indicate SD for n = 3.

Table S7 . Ensemble docking pose binding affinity analysis per receptor cluster and ligand pose location cluster.
b Indicates chain E c Structures bound by ADP are indicated in normal font d Structures not bound by ADP are indicated in italics e Structures which make direct contacts with RNA are bolded

Table S9 . Change in average binding free energy upon mutation of select residues
*Average binding free energy of IOWH-032 to the Nsp13-ATP complex (ΔG WT ) with standard error of the mean of the three replicates reported.Change in average binding free energy (∆∆G )*+, 9:→<=: ) with standard errors of the difference of means.The binding free energy is calculated with Molecular Mechanics-Poisson-Boltzmann (or Generalized Born) Surface Area MM-PB/GBSA.Positive ∆∆G )*+, 9:→<=: suggest weaker binding and negative ∆∆G )*+, 9:→<=: suggest improved binding following the mutation.