key: cord-0852818-o8sougxa
authors: Milligan, Jennifer C; Zeisner, Theresa U; Papageorgiou, George; Joshi, Dhira; Soudy, Christelle; Ulferts, Rachel; Wu, Mary; Lim, Chew Theng; Tan, Kang Wei; Weissmann, Florian; Canal, Berta; Fujisawa, Ryo; Deegan, Tom; Nagara, Hema; Bineva-Todd, Ganka; Basier, Clovis; Curran, Joseph F; Howell, Michael; Beale, Rupert; Labib, Karim; O’Reilly, Nicola; Diffley, John F.X
title: Identifying SARS-CoV-2 Antiviral Compounds by Screening for Small Molecule Inhibitors of nsp5 Main Protease
date: 2021-04-08
journal: bioRxiv
DOI: 10.1101/2021.04.07.438806
sha: 69b41f6493ec38060039aaea3424de321cd3ee9f
doc_id: 852818
cord_uid: o8sougxa

The coronavirus 2019 (COVID-19) pandemic, caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), spread around the world with unprecedented health and socio-economic effects for the global population. While different vaccines are now being made available, very few antiviral drugs have been approved. The main viral protease (nsp5) of SARS-CoV-2 provides an excellent target for antivirals, due to its essential and conserved function in the viral replication cycle. We have expressed, purified and developed assays for nsp5 protease activity. We screened the nsp5 protease against a custom chemical library of over 5,000 characterised pharmaceuticals. We identified calpain inhibitor I and three different peptidyl fluoromethylketones (FMK) as inhibitors of nsp5 activity in vitro, with IC50 values in the low micromolar range. By altering the sequence of our peptidomimetic FMK inhibitors to better mimic the substrate sequence of nsp5, we generated an inhibitor with a subnanomolar IC50. Calpain inhibitor I inhibited viral infection in monkey-derived Vero E6 cells, with an EC50 in the low micromolar range. The most potent and commercially available peptidyl-FMK compound inhibited viral growth in Vero E6 cells to some extent, while our custom peptidyl FMK inhibitor offered a marked antiviral improvement.

The COVID-19 pandemic, caused by the SARS-CoV-2 virus, emerged late in 2019 and rapidly spread around the world, developing into the worst health crisis of the 21 st century [1, 2] . As of February 2021, a year after the WHO declared the outbreak a public health emergency of international concern, over 100 million cases have been confirmed, with more than 2.4 million deaths attributed to COVID-19 [3] . Over 60 different vaccines have now reached the clinical development stage, with several approved worldwide [4] . Vaccinating vulnerable people against SARS-CoV-2 is of paramount importance. However, new variants of the virus are emerging, and it is unclear how long vaccines will remain effective [5, 6] .

Thus, to combat this pandemic most effectively, antiviral drugs are needed as complements to vaccines. SARS-CoV-2 encodes at least nine enzymes that are important for viral proliferation and thus are attractive targets for antiviral drugs. In contrast to the rapidly evolving structural proteins of the virus that the vaccines are based on, these enzymes are highly conserved between different coronaviruses [7, 8] . This suggests that antivirals inhibiting these enzymes may be useful as pan-coronavirus treatments and as the virus becomes resistant to existing vaccines. COVID-19 is already the third zoonotic coronavirus after SARS-CoV-1 and MERS-CoV that emerged as global health threats during the last two decades [9] , so the availability of pan-coronavirus treatments as a first line of defence against novel coronaviruses may be crucial in the future.

The first two-thirds of the SARS-CoV-2 genome encodes sixteen non-structural proteins (nsps), which are required for viral proliferation [10, 11] . They are encoded in two large overlapping open reading frames (ORF 1a and ORF 1ab). Upon entry into the host cell, these are translated into two polyproteins (pp1a and pp1ab respectively), which are cleaved by two virus encoded cysteine proteases, generating sixteen functional nsps. The viral papain like protease (PLpro), which is encoded within nsp3, excises nsp1-3 [12] . The main viral protease (nsp5) is a chymotrypsin-related protease that cleaves the polyproteins at eleven sites, releasing nsp4-nsp16 [13] [14] [15] . The excised nsps are essential for the assembly of the viral replication transcription complex [11] . Inhibition of nsp5, therefore, blocks the viral replication cycle, making it an attractive antiviral drug target [16] . Nsp5 structure and function is conserved across all coronaviruses, with SARS-CoV-1 and 2 sharing roughly 96% sequence identity with the greatest degree of sequence conservation around the active site [14] . Nsp5 cleaves polyproteins after a glutamine residue, which is the case for the vast majority of coronavirus main proteases but is rare in human proteases [8, 17, 18] . This suggests that antivirals targeting the SARS-CoV-2 main protease may have broad spectrum anti-coronavirus inhibitory effects. As part of a larger project to identify inhibitors of all SARS-CoV-2 enzymes, here we describe a high-throughput drug screen to identify inhibitors of the main viral protease nsp5.

It has previously been shown that both C-and N-terminal epitope tags inhibit nsp5 dimerization and activity [17, 19] . Therefore, we optimised a purification method for a full length, untagged nsp5 from E. coli based on the SUMO/Ulp1 cleavable tag system ( Figure   1A ) [20] . Figure 1B shows fractions from the purification steps separated in an SDSpolyacrylamide gel; nsp5 migrates at around 36 kDa. This method resulted in a high yield of relatively pure nsp5.

To examine nsp5 protease activity, we designed a nsp9 gel-based cleavage assay. For this we synthesised a fusion protein substrate with a short linker sequence based on the nsp4-nsp5 junction (SAVLQ) in-between a FLAG-His epitope tag and the nsp9 protein ( Figure   1C ). We selected the nsp4-5 cleavage site, as it is a natural cleavage site of nsp5 predicted to have the highest affinity [17] . Cleavage of the substrate, which migrates at approximately 20 kDa, by nsp5 results in the release of untagged nsp9, which migrates at approximately 12 kDa. Cleavage of the Flag-His-SAVLQ-nsp9 substrate over time by nsp5 can be observed from as early as 30 minutes at both 0.5 µM and 1 µM of nsp5 ( Figure 1D ).

To carry out enzymatic characterisation and high throughput screening (HTS) for nsp5 inhibitors, we optimised a Förster (fluorescence) resonance energy transfer (FRET)-based assay ( Figure 2A ). We used a 10 amino acid-long peptide substrate based on the natural nsp4-nsp5 cleavage site. The peptide is covalently attached to the fluorophore 2aminobenzoyl (Abz) at its N-terminus and a quencher (N-Tyrosine) at its C-terminus ( Figure   2A ). Cleavage of the substrate releases the fluorophore from the proximity of the quencher, resulting in an increase of fluorescent signal. Using this substrate, we determined the KM of nsp5 for this substrate to be 33.7 ± 4.7 µM ( Figure 2B ). This is comparable to a previously reported value of 28.2 μM using a similar FRET substrate [21].

We found that, at an enzyme concentration of 10 nM, the reaction is roughly linear for the first 20 minutes at room temperature (19 -23 °C) ( Figure 2C-D) . We used a substrate concentration of 20 µM for the HTS, which is relatively close to the KM value to ensure that competitive as well as non-competitive inhibitors can be identified [22] . We also found that buffer conditions had a profound effect on activity, especially when using automated liquid handling ( Figure S1 ). In brief, it proved essential to include glycerol (between 5-10%) and detergent (e.g. 0.01-0.02% Tween 20) to stabilise the enzyme. Inclusion of detergent also prevents compound aggregation ([23] and Zeng et al. this issue).

We carried out a high throughput screen (HTS) of a custom library with over 5,000 compounds at two drug concentrations of 4 µM and 0.8 µM. For the HTS, we incubated the compounds with nsp5 before reaction initiation to allow for the detection of slow-binding inhibitors. We defined primary hits as compounds that reduced nsp5 activity by more than 30% at the higher compound concentration ( Figure 3A , see Methods). We identified 27 primary hits that met this criterion and ranked them by percentage inhibition of nsp5 activity at both drug concentrations (Table 1 ) and validated them according to the scheme presented in Figure 3B .

Some compounds within the library interfered with the HTS emission wavelength (420 nm), increasing the fluorescence of the first time point and subsequent time points. Due to a maximum range of fluorescence detection these reactions became saturated over the course of the reaction resulting in an artificially reduced rate of reaction being determined for these compounds. We classed drug compounds that did not interfere with the HTS emission wavelength as "normal". Those that interfered slightly but a reduction in nsp5 activity was still evident we classed as "moderate". Finally, those that greatly interfered with the HTS fluorescence and so were likely artificial hits we classed as "high" (Table 1) . We excluded fourteen drugs from the primary hit list, which showed "high" autofluorescence. We were unable to source one of the compounds, so we tested twelve compounds further as nsp5 inhibitors.

We first re-tested these twelve compounds in the FRET-based assay over a range of drug concentrations from 0 to 500 µM and calculated IC50 values where possible. Four compounds did not show reproducible inhibition of nsp5 across any of the tested drug concentrations ( Figure S2A ) and we, therefore, discarded them. Tryphostin artificially increased fluorescence in the FRET-based assay at some concentrations tested, reducing the number of points available for IC50 calculation. Therefore, we also tested it in the nsp9 gel-based cleavage assay ( Figure S2A ). As it did not inhibit in either validation step, we discarded it. Fatostatin HBr and PDK1 /Akt /Flt Dual pathway showed some inhibition of nsp5 in the FRET-based assay. However, they had relatively high IC50 values for nsp5 inhibition and they did not show inhibition at any of the drug concentrations in the nsp9 gelbased assay, thus, we removed them from the final hit list ( Figure S2B ). [21, 24] . To test the specificity of the final five drugs we tested them against a different protease, thrombin. We used the same buffer as in the HTS, which contained a non-ionic detergent, which reduces the likelihood of unspecific inhibition via colloidal aggregation [23] . We used a substrate based on the natural thrombin substrate: Aminobenzoyl-LGARGHRPYD-N-Tyrosine [25] . Of the five drugs, only bromoenol lactone inhibited thrombin activity ( Figure 3D ) indicating that it is not a specific inhibitor of nsp5.

Lastly, we tested whether these compounds could inhibit the other viral coronavirus protease (nsp3 PLpro), which is also a cysteine protease [26] . The substrate and assay conditions used in this experiment are described elsewhere (see Lim et al, Biochemical J, this issue).

None of the remaining hits from the primary screen inhibited PLpro, indicating a specificity for nsp5 ( Figure 3E ). All together, we have identified four specific inhibitors of nsp5 in vitro ( Table 2) .

Ebselen has recently been described as an nsp5 inhibitor [14] , but was not identified as an hit in our screen. We wondered whether this discrepancy was due to the presence of reducing agent in our reaction buffers, which was not present in the previous work [14] . We Figure 4A ). In contrast, Calpain Inhibitor I inhibited nsp5 in the absence of reducing agent as well as the presence of either reducing agent ( Figure 4B ). The presence of reducing agents did not affect the activity of nsp5 in the absence of inhibitor ( Figure S3A ).

Taken together, these results indicate that Ebselen can only inhibit nsp5 under non-reducing conditions. A similar conclusion was also reached recently by Ma et al. [29] .

We next tested the ability of our best hit compounds to inhibit SARS-CoV-2 replication in monkey-derived Vero E6 cells as described by Zeng et al (see this issue) ( Figure 5A ).

Calpain inhibitor I, which has previously been tested for nsp5 inhibition in vitro but not in a cell-based assay [21] , was the most potent inhibitor of viral infection with an EC50 value of 0.28 ± 0.01 µM. Despite being a good inhibitor in vitro, the Z-VAD-FMK inhibitor displayed inhibition only at concentrations greater than 100 µM ( Figure 5B -C).

Combining two antiviral drugs with two different modes of action is a common strategy in treating viral infections as it can increase each drug's effectiveness and prevent the emergence of drug resistance [30] . We therefore tested the inhibitory capacity of our best hits in the same cell-based assay in combination with remdesivir. Remdesivir is a broad spectrum nucleoside analogue that is currently the only approved antiviral drug against SARS-CoV-2 and targets the RNA-dependent RNA polymerase [31] . None of the nsp5 inhibitors tested showed synergistic or additive effect with remdesivir ( Figure S4A-B) .

The three most potent inhibitors in vitro were the FMK peptidomimetic inhibitors, originally designed to target caspases [32]. These FMK peptidomimetic inhibitors comprise an Nterminal group that increases cell permeability, a short peptidyl targeting sequence, and a Cterminal functional group which can covalently bind to and inactivate the active cysteine residue in proteases ( Figure 6C ). We initially tried to investigate the effect of different Cterminal functional groups on inhibitor potency. The functional group in all three of the peptidomimetic inhibitors picked up in our screen is a fluoromethylketone group. While these are non-toxic to cells in culture, experiments in mice showed that their metabolic conversion leads to the production of toxic fluoroacetate [32, 33] . Some safer alternatives of the FMK peptidomimetic inhibitors are available commercially. One such alternative has been developed based on the difluorophenoxymethylketone (OPh) functional group [32]. While they were not present in our drug library, three of them are available commercially; Q-VD-OPh, Q-DEVD-OPh and Q-IETD-OPh. We tested these three drugs in the nsp9 gel-based cleavage assay. Of these, Q-IETD-OPh showed the best nsp5 inhibition in the gel-based assay, Q-DEVD-OPh only displayed inhibition at the highest concentration of drug tested and Q-VD-OPh did not inhibit nsp5 activity ( Figure 6A ). Similarly, in the FRET-based assay, Q-IETD-OPh was the most effective nsp5 inhibitor, with an IC50 value of 17.4 ± 1 µM ( Figure   6B ). This comparison between peptides with the same amino acid sequence, but differing functional groups, shows that the FMK peptidomimetic inhibitors are more potent inhibitors of nsp5 in vitro than the OPh variants.

Next, we tested whether changing the sequence of the peptidyl moiety would improve the inhibitor potency of the FMK compounds. The three FMK inhibitors identified in the screen mimic the substrate sequence of different caspases [32] . All of them have a common length of three or four amino acids, two hydrophobic residues as well as one charged residue.

Work on SARS-CoV-1 nsp5 indicates that bulky hydrophobic residues are key for substrate recognition, especially in the P2 position ( Figure 6C ) [34] . We hypothesised that altering the amino acid targeting sequence to further mimic an nsp5 substrate might improve the inhibitory effect on nsp5. We synthesised three custom FMK inhibitors, with lengths varying from four to six amino acids ( Figure 6C ) [35] [36] [37] . Their peptidyl moiety mimicked the sequence around the natural nsp4/5 cut site [17] . To simplify chemical synthesis by basing it on previous approaches, we used an aspartic acid instead of a glutamine at the P1 site and an alanine instead of a threonine in the P6 site.

All three custom peptidomimetic compounds exhibited sub-micromolar IC50 values in the FRET-based assay ( Figure 6D ). Z-ASAVLD-FMK had the highest IC50 (0.26 ± 0.02 µM), followed by Z-SAVLD-FMK (0.02 ± 0.001 µM). Z-AVLD-FMK, the shortest of the peptides, exhibited by far the greatest inhibitory potency with an IC50 of less than 1 nM (IC50 = 0.8 ± 0.09 nM) which is a substantial improvement over the best commercially available peptidomimetic compound Z-VAD-FMK (IC50: 0.16 ± 0.01 µM).

To assess whether the customisation also improved the in vivo inhibitory capacity, we tested the best custom nsp5 inhibitor (Z-AVLD-FMK) in the viral infection assay ( Figure 7A -B). We determined the EC50 value for Z-AVLD-FMK to be 66.01 ± 7.28 µM. Compared to the commercial Z-VAD-FMK, this is a 2-fold increase in inhibitor potency ( Figure 7 and Figure   5 ). This showed that changing the peptidyl moiety to mimic the nsp5 substrate greatly improved its in vitro and cell-based inhibitory effect. Taken together, our work indicates that FMK inhibitors targeting nsp5 could be used as a starting point for the development of effective antiviral drugs against SARS-CoV-2.

One effective strategy for combatting the COVID-19 pandemic is to re-purpose existing drugs to rapidly identify potential antivirals. The main viral protease is a promising antiviral drug target as it plays a critical role in viral replication by generating the functional viral replication proteins from the polyproteins [13, 14] . Three of the strongest hits identified in our HTS were cell permeable FMK peptides. Peptidyl

FMKs are widely used due to their ability to strongly and selectively inhibit serine and cysteine proteases, such as caspases, cathepsins and Sentrin/SUMO specific proteases [32, 36] . FMK peptides have been previously investigated as inhibitors of the SARS-CoV-1 nsp5 in vitro, and one FMK peptide (Z-DEVD-FMK) has recently been identified as an inhibitor of SARS-CoV-2 nsp5 [24, 40]. The peptidyl backbone of these FMK inhibitors allows them to be utilised for target-based inhibition of specific proteases by mimicking the substrate sequence that binds directly to the active site of the protease. Since the substrate sequence preference of nsp5 is known, we designed three custom inhibitors with the peptidyl moiety based on the nsp5 cleavage site with the predicted highest affinity (nsp4/5) [17] . The shortest of the custom inhibitors, Z-AVLD-FMK, showed extraordinary inhibitor potency in vitro with a sub-nanomolar IC50. It also improved the inhibitor potency in the cellbased viral infection assay implying that inhibition is a result of the FMK peptides action on nsp5. This shows that peptidomimetic inhibitors are an excellent tool for probing nsp5 protease activity in vitro and in cell culture due to their easily customisable nature.

Our custom inhibitors could be further optimised by incorporating a glutamine, found in all natural nsp5 substrate sequences, at the P1 site [8, 18] . Another way forward may be to exchange the FMK functional group with less toxic versions such as difluorophenoxymethylketone (OPh). The development of FMK peptides into clinical drugs was initially halted due to high in vivo host cell toxicity. This may be due to the metabolic conversion of the m-FMK group into toxic fluoroacetate, especially in the liver [32].

However, promising studies in mice treated with a peptidyl-OPh inhibitor suggested this alternative functional group it was well tolerated [41] . Furthermore, Q-VD-OPh, has been used in trials investigating AIDS disease progression in rhesus macaques with no toxic side effects [42] . While the commercially available OPh compounds we tested inhibited nsp5 to a lesser extent than the FMK counterparts, customisation of their substrate sequence will likely increase their inhibitory potency against nsp5. Thus, exchanging the functional group of the customised FMK inhibitors could allow them to be developed into clinical drugs.

The nsp5 coding sequence was subcloned from a bacterial expression vector of MBP-nsp5, Cell pellets were resuspended in nsp9 buffer (30 mM HEPES pH 7.6, 250 mM sodium chloride, 5 mM magnesium acetate, 10% glycerol, 0.02% NP-40 substitute, 1 mM DTT) supplemented with protease inhibitors (Roche Complete Ultra tablets, 1 mM AEBSF, 10 µg/ml pepstatin A, 10 µg/ml leupeptin) and lysed with a dounce homogenizer. The protein was purified from the cleared lysate by affinity to Anti-FLAG M2 Affinity gel (Sigma-Aldrich) and eluted with nsp9 buffer containing 0.1 mg/ml 3xFlag peptide.

Reactions were carried out in a buffer containing 50 mM HEPES-KOH pH 7.6, 1 mM EDTA, 2 mM DTT, 10% glycerol, and 0.02% Tween-20. A concentration of 1 µM nsp5 was used unless otherwise specified. This was incubated with drug compounds resuspended in DMSO or DMSO alone for 10 minutes at room temperature. Reactions were then initiated by the addition of 6.25 µM FLAG-His-SAVLQ-nsp9 substrate. Protease activity was allowed to continue for 1 hour at room temperature before reactions were quenched and denatured by the addition of SDS loading buffer. Products were then separated over a 12% Bis-Tris gel in MOPS buffer and stained with coomassie blue (Generon Cat. No NB-45-00078-1L).

The assay uses a FRET peptide substrate 

To determine the KM for our FRET substrate, the FRET-based nsp5 activity assay was done 

To identify inhibitors of nsp5 the FRET-based nsp5 activity assay, as described above, was used to screen a custom library of over 5,000 compounds assembled from different sources 

Time-course data was analysed using MATLAB (R2020_a) to determine the rate of reaction then normalised to the rate of reaction of positive (no drug compound) control reaction wells.

We assessed each individual HTS screen plate and the HTS as a whole by calculating a Z' factor score based on normalised rate of reaction compared to no enzyme controls. The screen had a Z' factor greater than 0.6 indicating that we could reliably identify hits.

Percentage reduction in rate of reaction or "activity" were then calculated for each drug compound. Those with a reduction in activity that was greater than 30% in the highest drug concentration (4 µM) were selected as primary hits. Hits were then ranked based on percentage reduction in activity and whether they were hits in both HTS concentrations.

Compounds selected for in vitro experimental validation and cell-based experiments were purchased and resuspended into DMSO following manufacturer's instruction.

Drug compounds for IC50 determination were titrated over a wide range from 1500 mM to either 1 nM over 20 steps or from 500 mM to 10 pM over 30 steps depending on the drug stock concentrations available. FRET-based nsp5 activity assays were carried out, as 

All data associated with this paper will be deposited in FigShare (https://figshare.com/). The five hits from the screen were tested for specificity against PLpro protease. Tables   Table 1: Full primary hit list   Table showing ranked initial hits with compound name, fluorescence reading type and percentage inhibition at both primary screen concentrations (4 and 0.8 µM). Those with "HIGH" fluorescence readings were either reading at over the detectable range in the screen wavelength or completely saturated. Such a reading would result in a decreased rate in reaction due to how reaction rates were determined via MATLAB. Those reading at "moderate' had slightly increased fluorescence at both screen concentrations, but a rate of reaction could still be determined. Finally, those with "normal" fluorescence readings were drugs that did not interfere with the fluorescence signal at the primary screen emission wavelength at all. Step 1:

Step 2:

Step 3:

Gel-based Nsp9 cleavage assay 5 drugs

Step 4:

FRET-based thrombin assay 4 drugs

Step 5: Normalised activity (%)

Calpain Inhibitor I 10 -3 10 -2 10 -1 10 0 10 1 10 2 10 3 10 -3 10 -2 10 -1 10 0 10 1 10 2 10 3 

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

A new coronavirus associated with human respiratory disease in China

Coronavirus disease (Covid-19) pandemic

Draft landscape and tracker of COVID-19 candidate vaccines

SARS-CoV-2 escape <em>in vitro</em> from a highly neutralizing COVID-19 convalescent plasma

Comprehensive mapping of mutations in the SARS-CoV-2 receptor-binding domain that affect recognition by polyclonal human plasma antibodies

From SARS to MERS: crystallographic studies on coronaviral proteases enable antiviral drug design

Conservation of substrate specificities among coronavirus main proteases

Origin and evolution of pathogenic coronaviruses

The Nonstructural Proteins Directing Coronavirus RNA Synthesis and Processing

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

Identification of Severe Acute Respiratory Syndrome Coronavirus Replicase Products and Characterization of Papain-Like Protease Activity

Mechanisms and enzymes involved in SARS coronavirus genome expression

Structure of Mpro from SARS-CoV-2 and discovery of its inhibitors

Processing of the SARS-CoV pp1a/ab nsp7-10 region

Virus-encoded proteinases and proteolytic processing in the Nidovirales

Evaluating the 3C-like protease activity of SARS-Coronavirus: recommendations for standardized assays for drug discovery

Mechanism of the maturation process of SARS-CoV 3CL protease

Ligand-induced Dimerization of Middle East Respiratory Syndrome (MERS) Coronavirus nsp5 Protease (3CLpro): IMPLICATIONS FOR nsp5 REGULATION AND THE DEVELOPMENT OF ANTIVIRALS

We thank all members of the involved labs for their help and comments on the manuscript. 

 2-(acetylamino)-N-[(1S)-1-({[(1S)-1-formylpentyl]