High Throughput Screening for SARS-CoV-2 Helicase Inhibitors

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for nearly 7 million deaths worldwide since its outbreak in late 2019. Even with the rapid development and production of vaccines and intensive research, there is still a huge need for specific anti-viral drugs that address the rapidly arising new variants. To address this concern, the National Institute of Allergy and Infectious Diseases (NIAID) established nine Antiviral Drug Discovery (AViDD) Centers, tasked with exploring approaches to target pathogens with pandemic potential, including SARS-CoV-2. In this study, we sought inhibitors of SARS-CoV2 non-structural protein 13 (nsP13) as potential antivirals, first developing a HTS-compatible assay to measure SARS-CoV2 nsP13 helicase activity. Here we present our effort in implementing the assay in a 1,536 well-plate format and in identifying nsP13 inhibitor hit compounds from a ~650,000 compound library. The primary screen was robust (average Z’ = 0.86 ± 0.05) and resulted in 7,009 primary hits. 1,763 of these compounds upon repeated retests were further confirmed, showing consistent inhibition. Following in-silico analysis, an additional orthogonal assay and titration assays, we identified 674 compounds with IC₅₀ <10 μM. We confirmed activity of independent compound batches from de novo powders while also incorporating multiple counterscreen assays. Our study highlights the potential of this assay for use on HTS platforms to discover novel compounds inhibiting SARS-CoV2 nsP13, which merit further development as an effective SARS-CoV2 antiviral.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS CoV-2) caused coronavirus disease 2019 (COVID-19) and its ensuing pandemic that resulted in nearly 7 million deaths worldwide [1] since December of that year. The impact of COVID-19 on our society is still enormous which has been driving momentum for developing prevention and treatment strategies. To date, there have been multiple successful vaccine developments [2, 3] and anti-viral drugs have also been identified [4–6], efforts helping to curb morbidity and mortality rates of COVID-19. All antiviral drugs are categorized into two types: those targeting viral proteins and those targeting host cellular components. Targeting a viral protein has potential for high selectivity but may lead to the emergence of mutation-driven drug resistance [7]. Targeting host cellular components, including cellular receptors and human enzymes, may be effective in thwarting viral entry and replication, but has a drawback in that other beneficial functions of those components may be impacted, conferring severe side effects and poor safety profiles [8, 9]. For this reason, the AViDD Centers are directed to focus on viral rather than host targets. This focus doesn’t eliminate the use of phenotypic approaches, which are especially important in identifying first-in-class drugs, but demands fast follow-on mechanistic studies to establish mode of action.

Currently for SARS CoV-2, there are drugs and monoclonal antibodies (mAbs) against three viral targets: RNA dependent RNA polymerase (RdRp), Main protease (M^pro) and spike proteins. The door is open for further SARS CoV-2 antiviral drug discovery, given that the virus expresses four other enzymes required for viral replication that can serve as target for an SARS CoV-2 drug: M^pro (3Cl^pro; nsP5), papain-like protease (PL^pro; nsP3), RdRP (nsP12) and helicase (nsP13) [10–18]. Unlike other viral enzyme proteins, viral helicases have been explored less as an antiviral target, due to the difficulty to identify specific inhibitors with low affinity for human helicase and the hardship of developing robust and cost-effective assays that could be implemented for an HTS campaign.

SARS CoV-2 nsP13 belongs to the 1B helicase superfamily. It has an ATP-dependent DNA and RNA unwinding activity with 5’ to 3’ polarity [19, 20]. SARS CoV-2 helicase (nsP13) is highly conserved as compared to the helicases in other coronaviruses, with only one amino acid residue difference with SARS CoV-1 nsP13 [21], and 99% similarity to Bat SARS-like coronavirus helicase [22]. This predicts that bona fide nsP13 inhibitors may have pan-coronavirus antiviral activity, a favorable attribute given that other coronaviruses have pandemic potential.

SARS CoV-2 nsP13 has a triangular pyramidal shape and consists of three domains: an N-terminal zinc-binding domain, a C-terminal RNA-binding ATPase with two RecA-like domains (1A and 2A domains), and finally stalk and 1B domains that bridge the N-terminal and C-terminal domains [23]. While a study using a fragment-based approach identified several potential binding sites for small molecules in addition to two well-defined conserved drug binding pockets in nsP13, an ATP/ADP−binding site and RNA-binding site, there are only small number of molecules, including Punicalagin, Cepharanthine, FPA-124 and 2-phenylquinoline derivatives, known to interact with nsP13 resulting in inhibition of its enzyme activity [5, 24, 25]. Hence, further development of nsP13-targeting antiviral drugs which are more potent than these early tool compounds is needed, provided that a margin of selectivity toward related host targets (human helicases) is maintained. Here we present our effort in developing, implementing, and executing a high throughput campaign that resulted in potential hits for anti-SARS-CoV2 therapeutic development.

Material and Methods

Purified SARS CoV-2 nsP13 helicase Production

Both His tagged (His-nsP13) and cleaved versions of SARS CoV-2 nsP13 (nsP13) were produced to yield the full-length nsP13. The nsP13 gene was initially cloned into a plasmid by the gene synthesis method by GenScript with nsP13 sequence from the genome sequence of SARS-CoV2 Wuhan-Hu-1 isolated (GenBank access no.: NC_045512). The full length nsP13 gene, covering aa. 1–601, was subcloned an in-house pET-based expression vector with expression and purification tags on the N-termini. The final expression form has an extra sequence of MPSDKPHHHHHHHHHHLLENLYFQG on the N-terminal part of nsP13. The underlined sequence indicates the TEV recognition site. The protein was expressed in E.coli BL21 (DE3) at 18 oC for overnight and purified with a standard His-tag purification method using Ni column (Qiagen 10 mL custom packed) followed by a cation exchange (GE 5mL SP column) and a size exclusion column (Superdex 200 26X60) in a buffer (100 mM NaCl, 10 mM HEPES 7.4, 1 mM DTT). For cleavage of the His tag, the protein was treated with TEV protease and further purified to remove the cleaved tag and TEV protease using a size exclusion column.

The final protein is in buffer contains 100 mM NaCl, 10mM HEPES (pH 7.4) and 1mM DTT.

dsDNA substrate and Trap DNA

dsDNA substrate was kindly provided by Dr. Chung (University of Louisville). The following single strand DNA was synthesized by IDT; T20D25BHQ: 5′-TTTTTTTTTTTTTTTTTTTTGAGCGGATTACTATACTACATTAGA(BHQ)-3′; FAM-T0D25: /56-FAM/TCTAATGTAGTATAGTAATCCGCTC; ATTO647-T0D25: /ATTO-647/TCTAATGTAGTATAGTAATCCGCTC 100 μM T20D25BHQ and 1mM FAM-T0D25 or ATTO647-IBRQ were annealed from 95 °C to room temperature (RT) over 2 hours with annealing buffer (10 mM Tris-HCl pH7.5–8.0, 50mM NaCl, 1mM EDTA). The sequence of trap DNA is TCTAATGTAGTATAGTAATCCGCTC and synthesized by IDT. The 100 μM stock was made by dissolving it in TE buffer.

SARS CoV-2 nsP13 384 Well Plate format protocol

First, 10 μL of mixture of His-nsP13 (final concentration is 15 nM) and trap DNA (final concentration is 500 nM) in assay buffer (100 mM NaCl, 2.5 mM MgCl₂, 20 mM HEPES (pH 7.4) and 2 mM ATP) were dispensed. Only trap DNA was dispensed to the wells for high control. Sequentially, 5 μL compound was added. For the wells designated as negative control, 5 μL of 5X stop solution (20 mM HEPES (pH 7.4), 0.2 M NaCl and 0.2M EDTA) was dispensed. Plates were incubated at RT for 10 minutes, followed by addition of 10 μL of dsDNA (final concentration is 100 nM). Plates were incubated at 30 °C for 60 minutes. After addition of 5 μL of 5X stop solution, the fluorescence intensity was measured using PHERAstar (BGM LABTECH) with filter set of Ex / Em: 485 / 520 nm.

SARS CoV-2 nsP13 Assay in 1536 Well Plate format for HTS

First, 2.5 μL of mixture of nsP13 (final concentration is 0.075 nM) and trap DNA in assay buffer (100 mM NaCl, 2.5 mM MgCl₂, 20 mM HEPES (pH 7.4), 2 mM ATP and 0.05% BSA) were dispensed. Only trap DNA was dispensed to the wells for high control. Sequentially, 30 nL compound was added. For the wells designated as negative control, 1 μL of 5X stop solution (20 mM HEPES (pH7.4), 0.2 M NaCl and 0.2M EDTA) was dispensed before addition of 2.5 μL of dsDNA (final concentration is 100 nM). Following by centrifugation of the plates at 1,200 rpm for 1 minute, plates are incubated at 30 °C for 30 minutes. The fluorescence intensity was measured using PHERAstar (BGM LABTECH) with filter set of Ex / Em: 485 / 520 nm. As an orthogonal readout, dsDNA substrates prepared with ATTO647-IBRQ were used. For this assay, we used exactly same conditions as we did with the FAM-T0D25 substrate but changed to 50 nM final concentration and of course utilized the proper filter set of Ex / Em: 490 / 675 nm on the PHERAstar (BGM LABTECH) plate reader.

Cell-based VSV-G entry Assay

This assay utilized Murine Leukemia virus (MLV) reporter luciferase virus pseudotyped with glycoprotein from Vesicular stomatitis virus (VSV-G), which was kindly provided by Dr. Choe (Boston Children’s Hospital, Harvard Medical School). This VSV-G pseudovirus was generated as previously described [12] [26]. In this assay, 2000 cells (hACE2–293T, kindly provided by Dr. Choe; Boston Children’s Hospital, Harvard Medical School) / 2.5 μL / well were incubated with 2.5 μL / well of VSV-G PV (Multiplicity of Infection (MOI) 1) for 48 hours at 37 °C and 5% CO₂. Viral infection was detected as luminescent signal using One-Glo reagents (Promega, 5 μL /well). The data was obtained using PHERAstar (BGM LABTECH).

dsDNA melting temperature assay

25 μM of test compound was mixed with 5 μM of dsDNA substrate described above in 1 X helicase buffer with 0.2 X ROX. The dsDNA Tm was measured with a real-time PCR machine (QuantaStudio 6-Pro, ThermoFisher) using the Tm analysis module with ROX as a passive dye.

Screening libraries

UF Scripps Drug Discovery Library (UF-SDDL)

The UF Scripps Drug Discovery Library (UF-SDDL) is a collection of 649,568 small drug-like compounds that belong to The High-Throughput Molecular Screening Center at Herbert Wertheim UF Scripps Institute for Biomedical Innovation & Technology. The compounds are unique and individually pure, representing a wide diversity of drug-like small organic molecules used in traditional and non-traditional drug-discovery biology. This library has been curated from over 20 commercial sources, supplemented with academic sources, including compounds and sub-libraries prepared internally. All UF-SDDL compounds were selected based on scaffold novelty, physical properties (i.e., “drug-likeness”) and spatial connectivity. In its current state, the UF-SDDL has several focused sub-libraries for screening popular drug-discovery target classes (e.g., kinases/transferases, GPCRs, ion channels, nuclear receptors, hydrolases, transporters), with diverse chemistries (e.g., click-chemistry, PAINS-free collections, Fsp3 enriched, covalent inhibitors and natural product collections) and with desirable physical properties (“rule-of-five”, “rule-of-three”, polar surface area, etc.) [27–32].

TargetMol Library

This library consists of 1,097 compounds that was selected based on ranking of the results of in silico molecular docking using the Swiss-Model Homology Modelling process to generate reliable protein models or 3D protein structures of Receptor Binding Domain (RBD) of spike protein, ACE2, viral papain like protease (Plpro) and main protease (3CLpro). This library was purchased from TargetMol Chemicals (Boston USA). https://www.targetmol.com/compound-library/anti_covid_19_compound_library

Helicase focused library

Life Chemicals (Ontario, Canada) created a collection of 2,531 compounds which consists of a reference set of molecules known to have helicase-related activity with analogs that display 84 % Tanimoto similarity to the reference set using MDL public keys. https://lifechemicals.com/news/helicase-screening-library-is-now-available-pre-plated

Screening data acquisition, normalization, representation, and analysis

All raw plate-based data files were uploaded into the UF Scripps institutional HTS database (Symyx Technologies, Santa Clara, CA) for plate QC and hit identification. Activity for each well was normalized on a per-plate basis using the following equation:

$$\%\text{activity}=100\times \left(\left(\text{Test well-Median Low Control}\right)/\left(\text{Median High Control-Median Low Control}\right)\right)$$

Where “High Control” indicates wells containing dsDNA + trap DNA + DMSO; while “Low Control” indicates wells containing nsP13 + dsDNA + trap DNA + DMSO and finally the “Data Wells” contain the same with test compounds. The Z’ and signal-to-background ratio (S:B) were calculated using the High Control (HC) and Low Control (LC) wells using the following equations:

$$Z\text{'}=\frac{(3\times \text{STDV HC}+3\times \text{STDV LC})}{\text{ABS}(\text{AVR HC}-\text{AVR LC})}$$

$$S:B=AVGHC÷AVGLC$$

In each case, a Z’ value greater than 0.5 was required for a plate to be considered acceptable [33].

Result

Optimization of nsP13 assay

The helicase activity of SARS CoV-2 nsP13 was measured by detecting the fluorescein amidites (FAM) signal increase which is only occurring when FAM is dissociated from the black hole quencher (BHQ) by unwinding activity of dsDNA substrate (Figure 1). This assay is a modification of a previously published method by Jang et al. [34] using a BHQ quencher instead of TAMRA to streamline the assay. The presence of trap DNA indicates existence of unannealed substrate, non-enzymatic unwinding activity and stabilizes the signal caused by enzymatic unwinding activity. This nsP13 assay was initially developed in 384 well-plate format (wpf). In this format, His tagged nsP13 (His-nsP13) was used. The Z’ was calculated using the average and standard deviation (STDV) of the high control which includes dsRNA and trap DNA and low control which includes His-nsP13, dsRNA and trap DNA. We obtained an excellent Z’ = 0.96, in this format. Next, we optimized the assay in 1536 wpf to ensure its amenable with our industrial scale screening platform and fits within an optimal time point. Operating at time points less than 30 minutes and clearly less than 15 minutes can be suboptimal on robotic screening platforms because it often leads to bottle necks and differential timing within the assay. The first optimization was done with His- nsP13 and we tested different concentrations of enzyme with a single concentration of dsDNA substrate (Supplementary Figure 1 A). All concentrations showed great Z’ but after 90 minutes, the reaction reached its maximum and the signal plateaued. In order to identify weak inhibitors during screening, it is necessary to screen while enzymatic reaction is linear and hence we decided to use 15 nM of nsP13 and 30 minutes reaction time for further optimization. Notably, when we performed the assay across a full plate in 1536 format, the variation across the plate was observed (Supplementary Figure 1 B). This caused an increase in STDV of low control (contains dsDNA + trap DNA + nsP13 + DMSO), hence a decrease in Z’. The addition of final concentration of 0.05% BSA in the rection buffer significantly decreased this variability and increased Z’ and improved the signal-to-background ratio (S:B; Supplementary Figure 1 B). Therefore, we decided to include BSA in the reaction buffer for further optimization. To avoid any potential interference from the His-tag at the N-terminus of the protein with the protein conformation or the interaction between the protein and small molecule, His cleaved nsP13 (nsP13) was tested and indeed we found higher activity from non-tagged nsP13. In an enzyme titration assay, we found that 0.038 nM of nsP13 represented a similar pattern of S:B progress to the previous outcome of 15 nM His-nsP13 and concentrations above 0.075 nM nsP13 showed exceptional Z’, greater than 0.8 (Supplementary Figure 1 C). Based on statistical data and assay robustness, an enzyme concentration of 0.075 nM, and 30 minutes incubation as the end point were chosen to perform large scale screen.

Figure 1:

SARS CoV-2 nsP13 assay principle. A schematic diagram of SARS CoV-2 nsP13 assay. The assay measures the fluorescence intensity of FMA that caused by unwinding of dsDNA substrate by SARS CoV-2 nsP13.

Primary HTS

The first step of the HTS campaign was screening the nsP13 assay against the 649,568 UF-SDDL. In the primary screen, compounds were tested at a single concentration in singlicate at a final nominal concentration of 7 μM. Raw assay data was imported into UF Scripps’ corporate database and subsequently analyzed using Symyx software. Activity of each compound was calculated on a per-plate basis using the following equation:

$$\%\text{activity}=100\times \left(\left(\text{Test well-Median Low Control}\right)/\left(\text{Median High Control-Median Low Control}\right)\right)$$

Where the “High Control” represents wells containing ds DNA + trap DNA + DMSO, “Low Control” represents wells containing nsP13 + dsDNA + trap DNA + DMSO and “Data Wells” contain the same including test compound. The Z’ and S:B for this assay is calculated using the High Control and Low Control wells.

The nsP13 assay performance was consistent with an average Z’ of 0.86 ± 0.05 and an average S:B of 5.14 ± 0.55 (n = 522 plates). Figure 2 shows the z-score value throughout the primary screen, showing a very reasonable distribution demonstrating the reliability and robustness of the assay. A statistical summary of the results of the primary screening assay are shown in Supplementary Figure 2. A mathematical algorithm was used to determine active compounds on a per plate basis. Three values were calculated: (1) the average activity value for all sample wells (2) 3 times the STDV value for the same set of wells (3) The sum of these two values was used as a cutoff parameter, i.e. any compound that exhibited greater percent inhibition than the cutoff parameter was declared active; again this was applied to each individual plate. Using the “Interval-based AVG + 3*SD Cut-off” criteria yielded 7,009 active compounds (“hits”). SSYA10–001, a 1,2,4-triazole, was used as a reference compound. This compound is a known coronavirus helicase inhibitor which blocks dsDNA and dsRNA unwinding activities of nsp13 [35, 36] and enabled us to monitor activity of nsP13 throughout the screening. The concentration response curve (CRC) of SSYA10–001 showed nsP13 activity was consistent throughout the primary screen, with an average IC₅₀ of 18.5 μM.

Figure 2:

Z-score analysis of SARS CoV-2 nsP13 primary HTS. Graphed is Z-score distribution across all 649,568 compounds tested. The red dots represent hit compound, black dots are non-hit compound, blue dots are low controls and orange dots are high controls.

Secondary Assays

Confirmation Screen

After completion of cherry-picking, a confirmation screen was run. The confirmation screen used the same reagents and detection system as the primary screening assays but tested each of the available compounds at a single concentration (nominally 7 μM) in triplicate. 15 compounds were unavailable, therefore 6,994 compounds were screened in secondary assay. Figure 3 A shows the scatter plots for this assay. The nsP13 confirmation assay performance yielded an average Z’ of 0.86 ± 0.1 and a S:B of 4.5 ± 0.1. Using a hit cut-off of 22.29 % inhibition (primary screen hit cut-off), 1,763 compounds showed their averages of all three replicates above this cut-off (hit rate: 25.21%). The CRC of SSYA10–001 showed the IC₅₀ was within expected range = 9 μM. In addition, total of 3,628 compounds from Helicase focused library (Life Chemicals; 2,531 compounds) and TargetMol (1,097 compounds) were also tested in triplicate 7 μM (Figure 3 B). These libraries were chosen retrospective to the large UF SDDL and obtained during the primary HTS campaign, so they were included as a separate work stream and run in triplicate, albeit as part of the same assay and time frame for the confirmation screen described just above. This assay performance yielded an average Z’ of 0.78 ± 0.1 and a S:B of 3.81 ± 0.2. Using a hit cut-off of 22.29 % inhibition (primary screen hit cut-off), 34 compounds showed inhibition activity above 22.29% in triplicate (hit rate: 0.94%).

Figure 3:

SARS CoV-2 nsP13 confirmation HTS result. Graphed is single point scatterplots of all compounds tested for each assay. Each dot graphed represents the activity result of a well containing test compound (black dots) or controls (red, pink and green dots). Overall screening statistics are described in the box below the graph. A. Summary of the result of SARS CoV-2 nsP13 confirmation screen including CRC data of the reference compound, SSYA10–001. B. Summary of the result of additional 3.6K compounds screening.

Titration Assay

The list of 1,797 compounds (1,763 from UF-SDDL and 34 from the Helicase focused and TargetMol libraries) that confirmed as hits were filtered based on following criteria: molecular weight between 142 and 700 (Da), did not have a high hit rate in our screening history at UF-Scripps, and were not Pan-assay interference compounds (PAINS) [37]. This resulted in 890 compounds which were selected for the titration assay. The titration assay employed the same reagents, protocols, and detection systems as the confirmation screen but tested each of the selected compounds as 10-point dose-response titrations (3-fold dilutions) in triplicate. The nsP13 titration assay was divided into two screens. The first screen tested 881 compounds (from UF SDDL) and the assay performance was consistent with an average Z’ of 0.79 ± 0.1 and a S:B of 4.19 ± 0.2 (Supplementary Figure 2). The second screen tested 9 compounds (from TargetMol and helicase collection hits) and the assay performance was consistent with an average Z’ of 0.90 ± 0.1 and a S:B of 3.72 ± 0.1 (Supplementary Figure 2). A four-parameter equation describing a sigmoidal dose-response curve was then fitted with adjustable baseline using Assay Explorer software (Symyx Technologies Inc.). The reported IC₅₀ values were generated from fitted curves by solving for the X-intercept value at the 50% inhibition level of the Y-intercept value. The following rule was used to declare a compound as “active” or “inactive”: Compounds with an IC₅₀ greater than 10 μM were considered inactive. Compounds with an IC₅₀ less than 10 μM were considered active. Of those, 678 compounds demonstrated nominal potency (IC₅₀ < 10 μM) in the nsP13 assay.

As an orthogonal readout assay, the same reagents and protocol were utilized but ATTO647-IBRQ substrate (final concentration of substrate, 50 nM, was used in the assay) and Ex / Em: 490 / 675 nm filters were used. The same 890 compounds (881 from UF SDDL and 9 from TargetMol and helicase collection) were tested as 10-point dose-response titrations (3-fold dilutions) in triplicate. The orthogonal read titration assay performance yielded an average Z’ of 0.83 ± 0.1 and a S:B of 6.16 ± 0.2 (Supplementary Figure 2). 666 compounds demonstrated IC₅₀ < 10 μM in this assay. The Venn analysis of active compounds (IC₅₀ < 10 μM) from these two titration assays indicated 625compounds overlapped, representing high specificity of the inhibition activity of compounds to SARS CoV-2 nsP13 (Supplementary Figure 3). Of those, 175 compounds were selected based on their activities (IC₅₀ < 5 μM), structures and structural properties for Liquid chromatography–mass spectrometry (LC–MS) analysis to confirm purity and identify mass. Of the 175 samples submitted for LC-MS analysis, 149 compounds showed the correct mass, i.e., the molecular weight of the structure in UF Scripps’ database matched that identified by LC-MS analysis of the screening sample. Of these 149 compounds, as determined by nominal methods (UV-vis spectroscopy, MS and ELSD), 53 samples demonstrated purity of < 80%. Hence, 96 compounds demonstrated the correct mass and were found pure (> 80%). While not available at the time of compound selection for medchem, we performed the LC-MS analysis as we believed this information may provide value to us retrospectively as it could be used to correlate to compound activity following the outcomes of medchem round 1.

Medchem Round 1 (Medchem −1): Selection of HTS compounds for powder confirmation

Following the titration assays, 218 of the 625 compounds were chosen for further follow-up study of a second, independently acquired or synthesized sample (termed a “powder confirmation”). Figure 4 shows representative examples of structural classes obtained for powder confirmation, derived based on perceived tractability. Less tractable compounds have issues such as potential chemical instability, high reactivity, toxicity/PAINS-like sub-structures, and high assay hit rates. Inhibition curves (obtained in duplicate runs) are shown for a representative compound in each class. All of these 218 compounds were confirmed for their correct mass and purity by LCMS. The follow-up assay employed the same reagents, protocols, and detection systems as the titration assay described above. The nsP13 MedChem −1 assay performance was consistent with an average Z’ of 0.73 ± 0.082 and a S:B of 5.34 ± 0.44 (Supplementary Figure 2). 110 compounds demonstrated IC₅₀ < 10 μM in this assay. To see if the compounds had non-specific inhibitory activity of viral infection, a cell-based VSV-G pseudovirus entry assay was applied to screen the same 218 compounds. In this assay, 2000 cells (hACE2–293T) / 2.5 μL /well were incubated with 2.5 μL /well of VSV-G PV (MOI 1) for 48 hours at 37 °C and 5% CO2. Viral infection was detected as luminescent signal using One-Glo reagents (Promega, 5uL/well). The low control for this assay was wells that contain cells, pseudovirus and DMSO. The data wells contain same but with compounds. The high control contains cell and DMSO. This counterscreen performance yielded an average Z’ of 0.61 ± 0.027 and a S:B of 124.13 ± 6.37 (Supplementary Figure 2). In this assay, 22 compounds showed > 30 % average max inhibition activity and within these 22 compounds, 7 compounds demonstrated IC₅₀ < 10 μM. These 22 compounds were thus non-specific inhibitors and were dropped from further hit triage efforts. The range of IC₅₀s of 218 compounds that were tested in this cell-based assay was from 145 nM to > 15 μM.

Figure 4:

Selected hit series studied in medchem round 1. Examples of purchased compounds subjected to the biochemical medchem assays utilizing either the FAM substrate (Red curve) or ATTO 647 substrate (Blue curve) and corresponding IC₅₀ are shown. The graphs display concentration response curves generated by 10pt 3-fold serial dilutions for each compound tested in triplicate. The X-axis is in log molar concentration and the Y-axis is normalized percent activity. Error bars are shown in SD and each point is an N=3 replicates.

As a counter assay to rule out false-positive compounds, 217 compounds (one compound was not available) were tested in a dsDNA melting temperature assay to see if there was any potential interference on dsDNA melting. In the assay, we found only one compound affecting the Tm of dsDNA and confirmed that all of the other 216 compounds do not have any significant effect on the Tm of the substrate with an dTm of 0.029 ± 0.20 (AVG ± STDV).

Medchem-2: Selected commercial analogs

Following Medchem-1 / powder confirmation and analysis of CRCs, preference for additional follow-up was given to commercially available analogs of confirmed active compounds with relatively few structural stability / tractability / selectivity concerns. This set of 101 commercial compounds is termed Medchem-2. The assays used were unchanged from earlier rounds. The nsP13 MedChem-2 assay performance was consistent with earlier runs, with an average Z’ of 0.76 ± 0.025 and a S:B of 4.78 ± 0.17 (Supplementary Figure 2). The orthogonal read assay was also performed using ATTO647-IBRQ substrate and this assay performance yielded an average Z’ of 0.76 ± 0.092 and a S:B of 8.68 ± 0.27 (Supplementary Figure 2). In both assays the same six top performing compounds were identified, with IC₅₀ < 7.5 μM as shown in Figure 5, making it clear that the activity assessments were consistent between the two assay formats. Compounds from Figures 4 & 5, and analogs, will progress to assessment of cell-based antiviral activity, with a focus on novel and non-commercial compounds, prepared by in-house synthesis with design guided by assay performance and by structural biology insights (e.g. X-ray crystallography, cryo electron microscopy).

Figure 5:

Selected most active hits identified in medchem round 2. Chemical structures, IC50, vendor and part number, and potential target class when known are shown for each molecule. The IC₅₀s shown are associated to the SCV2-NSP13 biochemical helicase assay.

Discussion

In this study we successfully developed and optimized the fluorescence-based SARS CoV-2 nsP13 helicase assay in 1536 wpf for HTS to identify potential helicase inhibitors. We have screened about 650K compounds at single concentration, in singlicate as the first step. The assay throughout the screening showed excellent statistics and proved to be very robust. To increase the chance to find selective nsP13 inhibitors we also included other libraries. They are the TargetMol and helicase focused libraries which consists of small molecules that target SARS CoV-2 infection and helicase. The hit rate was just under 1 % and only 34 compounds showed % inhibition above 22.29 %. While not surprising since nsP13 had roughly the same hit rate using the UF-SDDL, this could be due to the narrow targetability of the enzyme and reinforces the specificity of the screening outcome. We have selected total of 881 compounds based on these results to test dose-response manner (10 points seral dilution). Here 77% compounds showed IC₅₀ > 10 μM. We have also applied an orthogonal read of this assay to see if any compounds have fluorophore related activity. For this we used substrate with ATTO 647 instead of FAM, resulting in identifying 625 compounds also showed IC₅₀ > 10 μM. Many screening hits and hit series were confirmed following acquisition and testing of pure second samples (powder confirmation, medchem- 1) and acquisition and testing of selected analogs (medchem-2). Compounds shown in Figure 4 & 5, and analogs, are nsP13 inhibitors in the context of the assays used herein. Follow-up cell-based antiviral assays and ligand optimization to identify tractable leads are planned.

Helicases are essential for host cells; thus, a counter-screen that evaluates a potential cross activity to human homologs would be important to select antiviral candidate. Superfamily 2 (SF2) helicase to which SARS CoV-2 belongs include DEAD-box, RecQ-like, and RIG-I-like helicases [38]. Not all SF2 helicases exhibit unwinding activity, and they have various preference in the substrates. One of candidate enzymes as a counter screen for SARS CoV-2 nsP13 assay would be human RecQ4 or WRN (Werner’s syndrome gene), which is a superfamily II DNA helicase involving various biochemical process of cellular metabolism. Traditionally, monitoring of the unwinding activity of helicases has been based on a gel-based assay, which limits utility as a high-throughput format [39]. However, recently there have been development of non-gel-based helicase assays, which are based on the dsDNA/dsRNA activated ATPase activity [40–42]. In addition, recent publications have shown an HTS-compatible assays for WRN helicase. Therefore, it would be worthwhile to test our selected compounds in these assays to rule out any potential interference with cellular helicases near future [43, 44].

Although SARS2 CoV-2 nsP13 is an RNA helicase, it exhibits a robust unwinding activity both for dsDNA and for dsRNA. Further in biochemical assays, nsP13 showed a higher processability to dsDNA substrates over dsRNA substrate [45]. Recent publication demonstrated that this higher dsDNA unwinding activity might be due to the lack of the Fe-S cluster in bacterially produced nsP13 [46]. We took advantage of the robust DNA unwinding activity as the cost of dsRNA substrates is generally higher than that of dsDNA. Additionally, the stability of dsDNA is better than dsRNA, which can help the robustness of HTS performance. It has not been confirmed that dsDNA unwinding inhibitors of nsP13 are effective in inhibiting dsRNA unwinding due to a potential difference in mechanistic or structural difference. As there are no well-established potent nor selective inhibitors that effectively block the unwinding activity of SARS CoV-2 nsP13, this question remains to be addressed. More importantly, anti-viral activity of the selected compounds from this HTS campaign would be critical to nominate chemical entities for developing antiviral compounds for nsP13. The helicase functions as a component of viral replicase complex, and intracellular environments are expected to be dramatically different from the enzymatic reaction condition. The hit compounds discovered in this HTS campaign might provide us a way to address the question.

Our HTS assay utilizes the unwinding activity of the nsP13; however, other important functions could be investigated as a future screen. For example, the back tracking of replicase complex from a mismatched newly synthesized RNA is proposed as a critical function of SARS CoV-2 helicase in viral RNA synthesis based on the structural study [47]. In this notion, nsP13 plays a key role in translocating the replicase complex as a motor protein. Furthermore, because of the high similarity in amino acid sequence of nsP13 amongst all coronaviruses, it would be interesting if such compounds would have pan-coronavirus antiviral activity that may benefit the next coronavirus pandemic. Development of an HTS assay that involves the other nsPs such as nsP12 (viral polymerase) would provide novel antiviral hits or targets that are formed in the complex structure, not in the single protein. Such compounds should complement or synergize with antivirals having a different target / MOA and a drug cocktail approach is likely to lessen the risk of treatment resistance and enhance treatment effectiveness. Currently, this type of complex assay remains to be developed due to its technical difficulty in defining readouts for such assays although others appear to have successfully accomplished this task with alternative helicases [48].